diff --git a/app/.gitignore b/app/.gitignore index 796b96d1c..080495e55 100644 --- a/app/.gitignore +++ b/app/.gitignore @@ -1 +1,3 @@ /build +/src/main/obj +/src/main/libs diff --git a/app/app.iml b/app/app.iml index 9648c37ef..290a8c0b8 100644 --- a/app/app.iml +++ b/app/app.iml @@ -53,7 +53,6 @@ - diff --git a/app/build.gradle b/app/build.gradle index cd85420e7..e3f1b1b18 100644 --- a/app/build.gradle +++ b/app/build.gradle @@ -1,8 +1,10 @@ +import org.apache.tools.ant.taskdefs.condition.Os + apply plugin: 'com.android.application' android { compileSdkVersion 22 - buildToolsVersion "22" + buildToolsVersion "22.0.1" defaultConfig { applicationId "com.hippo.ehviewer" @@ -21,6 +23,37 @@ android { proguardFiles getDefaultProguardFile('proguard-android.txt'), 'proguard-rules.pro' } } + + // Disable default NdkCompile task + // Set new jni lib dir + sourceSets.main { + jni.srcDirs = [] + jniLibs.srcDir 'src/main/libs' + } + + // noinspection GroovyAssignabilityCheck + task ndkBuild(type: Exec, description: 'Compile JNI source via NDK') { + def ndkDir = project.plugins.findPlugin('com.android.application').sdkHandler.getNdkFolder() + def ndkBuilder + if (Os.isFamily(Os.FAMILY_WINDOWS)) { + ndkBuilder = 'ndk-build.cmd' + } else { + ndkBuilder = 'ndk-build' + } + commandLine "$ndkDir\\$ndkBuilder", + '-C', file('src/main/jni').absolutePath, + 'NDK_DEBUG=0', + '-j4' + } + + tasks.withType(JavaCompile) { + compileTask -> compileTask.dependsOn ndkBuild + } +} + +clean { + delete file('src/main/obj').absolutePath + delete file('src/main/libs').absolutePath } dependencies { diff --git a/app/src/main/jni/Android.mk b/app/src/main/jni/Android.mk new file mode 100644 index 000000000..cfa8c2bcb --- /dev/null +++ b/app/src/main/jni/Android.mk @@ -0,0 +1,232 @@ +LOCAL_PATH := $(call my-dir) + +################################################## +### libjpeg-turbo ### +################################################## +include $(CLEAR_VARS) + +LOCAL_MODULE := libjpeg-turbo +LOCAL_SRC_FILES:= \ +libjpeg-turbo/jaricom.c \ +libjpeg-turbo/jcomapi.c \ +libjpeg-turbo/jdapimin.c \ +libjpeg-turbo/jdapistd.c \ +libjpeg-turbo/jdarith.c \ +libjpeg-turbo/jdatadst.c \ +libjpeg-turbo/jdatadst-tj.c \ +libjpeg-turbo/jdatasrc.c \ +libjpeg-turbo/jdatasrc-tj.c \ +libjpeg-turbo/jdcoefct.c \ +libjpeg-turbo/jdcolor.c \ +libjpeg-turbo/jddctmgr.c \ +libjpeg-turbo/jdhuff.c \ +libjpeg-turbo/jdinput.c \ +libjpeg-turbo/jdmainct.c \ +libjpeg-turbo/jdmarker.c \ +libjpeg-turbo/jdmaster.c \ +libjpeg-turbo/jdmerge.c \ +libjpeg-turbo/jdphuff.c \ +libjpeg-turbo/jdpostct.c \ +libjpeg-turbo/jdsample.c \ +libjpeg-turbo/jdtrans.c \ +libjpeg-turbo/jerror.c \ +libjpeg-turbo/jfdctflt.c \ +libjpeg-turbo/jfdctfst.c \ +libjpeg-turbo/jfdctint.c \ +libjpeg-turbo/jidctflt.c \ +libjpeg-turbo/jidctfst.c \ +libjpeg-turbo/jidctint.c \ +libjpeg-turbo/jidctred.c \ +libjpeg-turbo/jmemmgr.c \ +libjpeg-turbo/jmemnobs.c \ +libjpeg-turbo/jquant1.c \ +libjpeg-turbo/jquant2.c \ +libjpeg-turbo/jutils.c +LOCAL_CFLAGS := -O3 -DANDROID + +ifeq ($(TARGET_ARCH_ABI),$(filter $(TARGET_ARCH_ABI), armeabi armeabi-v7a)) + # Don't add -mfpu=neon + # Some arm v7 cpu do not support neon, + # libjpeg-turbo can check neon in init_simd() + # through parse_proc_cpuinfo() + #LOCAL_CFLAGS += -mfpu=neon + LOCAL_SRC_FILES += \ + libjpeg-turbo/simd/jsimd_arm.c \ + libjpeg-turbo/simd/jsimd_arm_neon.S +endif + +ifeq ($(TARGET_ARCH_ABI),arm64-v8a) + LOCAL_SRC_FILES += \ + libjpeg-turbo/simd/jsimd_arm64.c \ + libjpeg-turbo/simd/jsimd_arm64_neon.S +endif + +ifeq ($(TARGET_ARCH_ABI),x86) + LOCAL_SRC_FILES += \ + libjpeg-turbo/simd/jsimd_i386.c \ + libjpeg-turbo/simd/jsimdcpu.asm \ + libjpeg-turbo/simd/jfdctflt-3dn.asm \ + libjpeg-turbo/simd/jidctflt-3dn.asm \ + libjpeg-turbo/simd/jquant-3dn.asm \ + libjpeg-turbo/simd/jccolor-mmx.asm \ + libjpeg-turbo/simd/jcgray-mmx.asm \ + libjpeg-turbo/simd/jcsample-mmx.asm \ + libjpeg-turbo/simd/jdcolor-mmx.asm \ + libjpeg-turbo/simd/jdmerge-mmx.asm \ + libjpeg-turbo/simd/jdsample-mmx.asm \ + libjpeg-turbo/simd/jfdctfst-mmx.asm \ + libjpeg-turbo/simd/jfdctint-mmx.asm \ + libjpeg-turbo/simd/jidctfst-mmx.asm \ + libjpeg-turbo/simd/jidctint-mmx.asm \ + libjpeg-turbo/simd/jidctred-mmx.asm \ + libjpeg-turbo/simd/jquant-mmx.asm \ + libjpeg-turbo/simd/jfdctflt-sse.asm \ + libjpeg-turbo/simd/jidctflt-sse.asm \ + libjpeg-turbo/simd/jquant-sse.asm \ + libjpeg-turbo/simd/jccolor-sse2.asm \ + libjpeg-turbo/simd/jcgray-sse2.asm \ + libjpeg-turbo/simd/jcsample-sse2.asm \ + libjpeg-turbo/simd/jdcolor-sse2.asm \ + libjpeg-turbo/simd/jdmerge-sse2.asm \ + libjpeg-turbo/simd/jdsample-sse2.asm \ + libjpeg-turbo/simd/jfdctfst-sse2.asm \ + libjpeg-turbo/simd/jfdctint-sse2.asm \ + libjpeg-turbo/simd/jidctflt-sse2.asm \ + libjpeg-turbo/simd/jidctfst-sse2.asm \ + libjpeg-turbo/simd/jidctint-sse2.asm \ + libjpeg-turbo/simd/jidctred-sse2.asm \ + libjpeg-turbo/simd/jquantf-sse2.asm \ + libjpeg-turbo/simd/jquanti-sse2.asm +endif + +ifeq ($(TARGET_ARCH_ABI),x86_64) + LOCAL_SRC_FILES += \ + libjpeg-turbo/simd/jsimd_x86_64.c \ + libjpeg-turbo/simd/jfdctflt-sse-64.asm \ + libjpeg-turbo/simd/jccolor-sse2-64.asm \ + libjpeg-turbo/simd/jcgray-sse2-64.asm \ + libjpeg-turbo/simd/jcsample-sse2-64.asm \ + libjpeg-turbo/simd/jdcolor-sse2-64.asm \ + libjpeg-turbo/simd/jdmerge-sse2-64.asm \ + libjpeg-turbo/simd/jdsample-sse2-64.asm \ + libjpeg-turbo/simd/jfdctfst-sse2-64.asm \ + libjpeg-turbo/simd/jfdctint-sse2-64.asm \ + libjpeg-turbo/simd/jidctflt-sse2-64.asm \ + libjpeg-turbo/simd/jidctfst-sse2-64.asm \ + libjpeg-turbo/simd/jidctint-sse2-64.asm \ + libjpeg-turbo/simd/jidctred-sse2-64.asm \ + libjpeg-turbo/simd/jquantf-sse2-64.asm \ + libjpeg-turbo/simd/jquanti-sse2-64.asm +endif + +ifeq ($(TARGET_ARCH_ABI),mips) + LOCAL_SRC_FILES += \ + libjpeg-turbo/simd/jsimd_mips.c \ + libjpeg-turbo/simd/jsimd_mips_dspr2.S +endif + +ifeq ($(TARGET_ARCH_ABI),mips64) + LOCAL_SRC_FILES += \ + libjpeg-turbo/jsimd_none.c +endif + +include $(BUILD_STATIC_LIBRARY) + + +################################################## +### libz ### +################################################## +include $(CLEAR_VARS) + +LOCAL_MODULE := libz +LOCAL_SRC_FILES := \ +zlib/adler32.c \ +zlib/crc32.c \ +zlib/inffast.c \ +zlib/inflate.c \ +zlib/inftrees.c \ +zlib/trees.c \ +zlib/zutil.c +LOCAL_CFLAGS := -O3 -DANDROID + +include $(BUILD_STATIC_LIBRARY) + + +################################################## +### libpng ### +################################################## +include $(CLEAR_VARS) + +LOCAL_MODULE := libpng +LOCAL_SRC_FILES := \ +libpng/png.c \ +libpng/pngerror.c \ +libpng/pngget.c \ +libpng/pngmem.c \ +libpng/pngpread.c \ +libpng/pngread.c \ +libpng/pngrio.c \ +libpng/pngrtran.c \ +libpng/pngrutil.c \ +libpng/pngset.c \ +libpng/pngtrans.c +LOCAL_CFLAGS := -O3 -DANDROID + +ifeq ($(TARGET_ARCH_ABI),armeabi-v7a) + # Add -mfpu=neon + # libpng only check neon when add -mfpu=neon + LOCAL_CFLAGS += -mfpu=neon +endif + +ifeq ($(TARGET_ARCH_ABI),$(filter $(TARGET_ARCH_ABI), armeabi armeabi-v7a arm64-v8a)) + LOCAL_SRC_FILES += \ + libpng/arm/arm_init.c \ + libpng/arm/filter_neon_intrinsics.c \ + libpng/arm/filter_neon.S +endif + +LOCAL_STATIC_LIBRARIES := libz + +include $(BUILD_STATIC_LIBRARY) + + +################################################## +### libgif ### +################################################## +include $(CLEAR_VARS) + +LOCAL_MODULE := libgif +LOCAL_SRC_FILES := \ +giflib/dgif_lib.c \ +giflib/gif_hash.c \ +giflib/gifalloc.c +LOCAL_CFLAGS := -O3 -DANDROID + +include $(BUILD_STATIC_LIBRARY) + + +################################################## +### libimage ### +################################################## +include $(CLEAR_VARS) + +RELEASE_FLAG := true + +LOCAL_MODULE := libimage +LOCAL_SRC_FILES := \ +image.c \ +jpeg.c \ +png.c \ +gif.c \ +utils.c +LOCAL_CFLAGS := -O3 -DANDROID +LOCAL_STATIC_LIBRARIES := libjpeg-turbo libpng libgif +LOCAL_LDLIBS := -lGLESv2 + +ifdef RELEASE_FLAG + LOCAL_CFLAGS += -DRELEASE +else + LOCAL_LDLIBS += -llog +endif + +include $(BUILD_SHARED_LIBRARY) diff --git a/app/src/main/jni/Application.mk b/app/src/main/jni/Application.mk new file mode 100644 index 000000000..7e938c7c9 --- /dev/null +++ b/app/src/main/jni/Application.mk @@ -0,0 +1,3 @@ +APP_ABI := armeabi armeabi-v7a x86 +APP_PLATFORM := android-7 +APP_OPTIM := release diff --git a/app/src/main/jni/gif.c b/app/src/main/jni/gif.c new file mode 100644 index 000000000..2e6232b9e --- /dev/null +++ b/app/src/main/jni/gif.c @@ -0,0 +1,574 @@ +#include "gif.h" +#include "utils.h" + +//#define STRICT_FORMAT_89A + +#define USE_PIXELS 0 +#define USE_BAK 1 +#define SET_BG 2 + +static int errorCode; + +// TODO is GIF87a work fine ? + +/* + * Return NULL if error + */ +static JNIEnv* getEnv(GifFileType* gif) { + if (gif == NULL) + return NULL; + + JNIEnv* env = NULL; + StreamContainer* sc = (StreamContainer*) (gif->UserData); + if (sc != NULL) { + JavaVM* jvm = sc->jvm; + (*jvm)->AttachCurrentThread(jvm, &env, NULL); + } + return env; +} + +static int streamReadFun(GifFileType* gif, GifByteType* bytes, int size) { + StreamContainer* sc = (StreamContainer*) gif->UserData; + JNIEnv* env = getEnv(gif); + + if (env == NULL) + return 0; + + (*env)->MonitorEnter(env, sc->stream); + + if (sc->buffer == NULL) { + jbyteArray buffer = (*env)->NewByteArray(env, size < 256 ? 256 : size); + sc->buffer = (jbyteArray) (*env)->NewGlobalRef(env, buffer); + } else { + jsize bufLen = (*env)->GetArrayLength(env, sc->buffer); + if (bufLen < size) { + (*env)->DeleteGlobalRef(env, sc->buffer); + sc->buffer = NULL; + + jbyteArray buffer = (*env)->NewByteArray(env, size); + sc->buffer = (*env)->NewGlobalRef(env, buffer); + } + } + + int len = (*env)->CallIntMethod(env, sc->stream, sc->readMID, sc->buffer, 0, + size); + if ((*env)->ExceptionOccurred(env)) { + (*env)->ExceptionClear(env); + len = 0; + } else if (len > 0) { + (*env)->GetByteArrayRegion(env, sc->buffer, 0, len, bytes); + } + + (*env)->MonitorExit(env, sc->stream); + + return len >= 0 ? len : 0; +} + +static int fileReadFun(GifFileType* gif, GifByteType* bytes, int size) { + FILE* fp = (FILE*) gif->UserData; + return fread(bytes, 1, size, fp); +} + +/* + * stack store the index to draw + * stack capacity should be imageCount * sizeof(int). + * When read it, use from tail to head + */ +static void getDrawStack(GIF* gif, int targetIndex, int* which, int* stack, + int* stackSize) { + int i; + int *disposals = gif->disposals; + int pxlIndex = gif->pxlIndex; + int bakIndex = gif->bakIndex; + + *which = SET_BG; + *stackSize = 0; + + for (i = targetIndex; i >= 0;) { + if (i == pxlIndex) { + *which = USE_PIXELS; + break; + } else if (i == bakIndex) { + *which = USE_BAK; + break; + } + + // Add index to stack + stack[*stackSize] = i; + (*stackSize)++; + + // If it is first index, break; + if (i == 0) + break; + + switch (disposals[--i]) { + case DISPOSAL_UNSPECIFIED: + case DISPOSE_DO_NOT: + break; + case DISPOSE_BACKGROUND: + // Just out of while(), + i = -1; + break; + case DISPOSE_PREVIOUS: + i--; + break; + } + } +} + +static inline bool getColorFromTableRGB(const ColorMapObject* cmap, int index, + rgb* color) { + if (cmap == NULL || index < 0 || index >= cmap->ColorCount) + return false; + GifColorType gct = cmap->Colors[index]; + color->red = gct.Red; + color->green = gct.Green; + color->blue = gct.Blue; + return true; +} + +static inline bool getColorFromTableRGBA(const ColorMapObject* cmap, int index, + rgba* color) { + if (cmap == NULL || index < 0 || index >= cmap->ColorCount) + return false; + GifColorType gct = cmap->Colors[index]; + color->red = gct.Red; + color->green = gct.Green; + color->blue = gct.Blue; + color->alpha = 0xff; + return true; +} + +static inline bool getColorFromTableLUM(const ColorMapObject* cmap, int index, + lum* color) { + if (cmap == NULL || index < 0 || index >= cmap->ColorCount) + return false; + GifColorType gct = cmap->Colors[index]; + color->l = getVFrowRGB(gct.Red, gct.Green, gct.Blue); + return true; +} + +static inline bool getColorFromTableLUMA(const ColorMapObject* cmap, int index, + luma* color) { + if (cmap == NULL || index < 0 || index >= cmap->ColorCount) + return false; + GifColorType gct = cmap->Colors[index]; + color->l = getVFrowRGB(gct.Red, gct.Green, gct.Blue); + color->a = 0xff; + return true; +} + +static void setBgRGB(GIF* gif, rgb* pixels) { + GifFileType* gifFile = gif->gifFile; + int num = gifFile->SWidth * gifFile->SHeight; + int size = num * sizeof(rgb); + rgb color; + if (getColorFromTableRGB(gifFile->SColorMap, gifFile->SBackGroundColor, &color)) + eraseRGB(pixels, num, color); + else + memset(pixels, BG_LUM, size); +} + +static void setBgRGBA(GIF* gif, rgba* pixels) { + GifFileType* gifFile = gif->gifFile; + int num = gifFile->SWidth * gifFile->SHeight; + int size = num * sizeof(rgba); + rgba color; + if (getColorFromTableRGBA(gifFile->SColorMap, gifFile->SBackGroundColor, &color)) { + eraseRGBA(pixels, num, color); + color.red = BG_LUM; + color.green = BG_LUM; + color.blue = BG_LUM; + color.red = 0xff; + } + eraseRGBA(pixels, num, color); +} + +static void setBgLUM(GIF* gif, lum* pixels) { + GifFileType* gifFile = gif->gifFile; + int num = gifFile->SWidth * gifFile->SHeight; + int size = num * sizeof(lum); + lum color; + if (getColorFromTableLUM(gifFile->SColorMap, gifFile->SBackGroundColor, &color)) + eraseLUM(pixels, num, color); + else + memset(pixels, BG_LUM, size); +} + +static void setBgLUMA(GIF* gif, luma* pixels) { + GifFileType* gifFile = gif->gifFile; + int num = gifFile->SWidth * gifFile->SHeight; + int size = num * sizeof(lum); + luma color; + if (!getColorFromTableLUMA(gifFile->SColorMap, gifFile->SBackGroundColor, &color)) { + color.l = BG_LUM; + color.a = 0xff; + } + eraseLUMA(pixels, num, color); +} + +static void setBg(GIF* gif, void* pixels) { + switch (gif->format) { + case GL_RGB: + setBgRGB(gif, pixels); + break; + case GL_RGBA: + setBgRGBA(gif, pixels); + break; + case GL_LUMINANCE: + setBgLUM(gif, pixels); + break; + case GL_LUMINANCE_ALPHA: + setBgLUMA(gif, pixels); + break; + } +} + +static inline void copyLineRGB(GifByteType* src, rgb* dst, + const ColorMapObject* cmap, int tran, int width) { + for (; width > 0; width--, src++, dst++) { + int index = *src; + if (tran == -1 || *src != tran) + getColorFromTableRGB(cmap, *src, dst); + } +} + +static inline void copyLineRGBA(GifByteType* src, rgba* dst, + const ColorMapObject* cmap, int tran, int width) { + for (; width > 0; width--, src++, dst++) { + int index = *src; + if (tran == -1 || *src != tran) + getColorFromTableRGBA(cmap, *src, dst); + } +} + +static inline void copyLineLUM(GifByteType* src, lum* dst, + const ColorMapObject* cmap, int tran, int width) { + for (; width > 0; width--, src++, dst++) { + int index = *src; + if (tran == -1 || *src != tran) + getColorFromTableLUM(cmap, *src, dst); + } +} + +static inline void copyLineLUMA(GifByteType* src, luma* dst, + const ColorMapObject* cmap, int tran, int width) { + for (; width > 0; width--, src++, dst++) { + int index = *src; + if (tran == -1 || *src != tran) + getColorFromTableLUMA(cmap, *src, dst); + } +} + +static void drawFrame(GIF* gif, void* pixels, int index) { + GifFileType* gifFile = gif->gifFile; + int ScreenWidth = gifFile->SWidth; + int ScreenHeight = gifFile->SHeight; + SavedImage cur = gifFile->SavedImages[index]; + GifImageDesc gid = cur.ImageDesc; + ColorMapObject *cmap = gid.ColorMap; + int tran = gif->trans[index]; + GifByteType* src; + rgb* dst1; + rgba* dst2; + lum* dst3; + luma* dst4; + + if (cmap == NULL) + cmap = gifFile->SColorMap; + if (cmap != NULL) { + src = cur.RasterBits; + int copyWidth = gid.Width; + int copyHeight = gid.Height; + if (copyWidth + gid.Left > ScreenWidth) + copyWidth = ScreenWidth - gid.Left; + if (copyHeight + gid.Top > ScreenHeight) + copyHeight = ScreenHeight - gid.Top; + + if (copyWidth > 0 && copyHeight > 0) { + switch (gif->format) { + case GL_RGB: + dst1 = (rgb*)pixels + gid.Top * ScreenWidth + gid.Left; + for (; copyHeight > 0; copyHeight--, src += gid.Width, dst1 += ScreenWidth) + copyLineRGB(src, dst1, cmap, tran, copyWidth); + break; + case GL_RGBA: + dst2 = (rgba*)pixels + gid.Top * ScreenWidth + gid.Left; + for (; copyHeight > 0; copyHeight--, src += gid.Width, dst2 += ScreenWidth) + copyLineRGBA(src, dst2, cmap, tran, copyWidth); + break; + case GL_LUMINANCE: + dst3 = (lum*)pixels + gid.Top * ScreenWidth + gid.Left; + for (; copyHeight > 0; copyHeight--, src += gid.Width, dst3 += ScreenWidth) + copyLineLUM(src, dst3, cmap, tran, copyWidth); + break; + case GL_LUMINANCE_ALPHA: + dst4 = (luma*)pixels + gid.Top * ScreenWidth + gid.Left; + for (; copyHeight > 0; copyHeight--, src += gid.Width, dst4 += ScreenWidth) + copyLineLUMA(src, dst4, cmap, tran, copyWidth); + break; + } + } + } +} + +static void render(GIF* gif, int index) { + + int i; + int width = gif->gifFile->SWidth; + int height = gif->gifFile->SHeight; + int imageCount = gif->gifFile->ImageCount; + int which, stackSize; + int stack[imageCount]; + void* tempPoint; + int tempInt; + + getDrawStack(gif, index, &which, stack, &stackSize); + + // prepare pixels + switch (which) { + case USE_PIXELS: + break; + case USE_BAK: + // switch pixel and bak + tempPoint = gif->bak; + gif->bak = gif->pixels; + gif->pixels = tempPoint; + tempInt = gif->bakIndex; + gif->bakIndex = gif->pxlIndex; + gif->pxlIndex = tempInt; + break; + case SET_BG: + // Set bg + setBg(gif, gif->pixels); + break; + } + + // Draw frame + for (stackSize--; stackSize >= 0; stackSize--) + drawFrame(gif, gif->pixels, stack[stackSize]); + + // render in screen + glTexSubImage2D(DEFAULT_TARGET, 0, 0, 0, width, + height, gif->format, DEFAULT_TYPE, gif->pixels); + + // update pxlIndex + gif->pxlIndex = index; + + // Set back if necessary + if (index + 1 < imageCount && gif->disposals[index + 1] == DISPOSE_PREVIOUS && + gif->bakIndex != gif->pxlIndex) { + int size; + switch (gif->format) { + case GL_RGB: + size = width * height * sizeof(rgb); + break; + case GL_RGBA: + size = width * height * sizeof(rgba); + break; + case GL_LUMINANCE: + size = width * height * sizeof(lum); + break; + case GL_LUMINANCE_ALPHA: + size = width * height * sizeof(luma); + break; + } + if (gif->bak == NULL) + gif->bak = malloc(size); + if (gif->bak != NULL) { + memcpy(gif->bak, gif->pixels, size); + gif->bakIndex = index; + } + } +} + +static GIF* analysisGifFileType(GifFileType* gifFile, int format, int* delays) { + + int i; + int* trans; + int* disposals; + GIF* gif; + GraphicsControlBlock gcb; + + int imageCount = gifFile->ImageCount; + + gif = (GIF*) malloc(sizeof(GIF)); + if (gif == NULL) + return NULL; + + gif->gifFile = gifFile; + gif->trans = (int*) malloc(imageCount * sizeof(int)); + gif->disposals = (int*) malloc(imageCount * sizeof(int)); + gif->format = format; + switch (format) { + case GL_RGB: + gif->pixels = malloc(gifFile->SWidth * gifFile->SHeight * sizeof(rgb)); + break; + case GL_RGBA: + gif->pixels = malloc(gifFile->SWidth * gifFile->SHeight * sizeof(rgba)); + break; + case GL_LUMINANCE: + gif->pixels = malloc(gifFile->SWidth * gifFile->SHeight * sizeof(lum)); + break; + case GL_LUMINANCE_ALPHA: + gif->pixels = malloc(gifFile->SWidth * gifFile->SHeight * sizeof(luma)); + break; + } + gif->pxlIndex = -1; + gif->bak = NULL; + gif->bakIndex = -1; + + if (gif->trans == NULL || gif->disposals == NULL || gif->pixels == NULL) { + free(gif->pixels); + free(gif->disposals); + free(gif->trans); + free(gif); + return NULL; + } + + trans = gif->trans; + disposals = gif->disposals; + for (i = 0; i < imageCount; i++) { + if (DGifSavedExtensionToGCB(gifFile, i, &gcb) != GIF_OK) { + delays[i] = DEFAULT_DELAY; + trans[i] = -1; + disposals[i] = 0; + } else { + delays[i] = gcb.DelayTime * 10; + trans[i] = gcb.TransparentColor; + disposals[i] = gcb.DisposalMode; + if (delays[i] <= 0) + delays[i] = DEFAULT_DELAY; + } + } + + return gif; +} + +jobject getObjFromGifFileType(JNIEnv* env, GifFileType* gifFile, int format) { + + int realWidth, slurp, imageCount; + int* delays; + GIF* gif; + jintArray delayArray; + jclass gifClazz; + jmethodID constructor; + + // Fix width, for not RGBA, but I will tell others the real width + realWidth = gifFile->SWidth; + if (format != GL_RGBA) + gifFile->SWidth = nextMulOf4(realWidth); + + // Slurp + slurp = DGifSlurp(gifFile); +#ifdef STRICT_FORMAT_89A + if (slurp != GIF_OK) + return NULL; +#endif + + // Check image count + imageCount = gifFile->ImageCount; + if (imageCount < 1) + return NULL; + + // delays malloc + delays = (int*) malloc(imageCount * sizeof(int)); + if (delays == NULL) + return NULL; + + // analysis GifFileType + gif = analysisGifFileType(gifFile, format, delays); + if (gif == NULL) { + free(delays); + return NULL; + } + + // Copy value + delayArray = (*env)->NewIntArray(env, imageCount); + (*env)->SetIntArrayRegion(env, delayArray, 0, imageCount, delays); + free(delays); + + gifClazz = (*env)->FindClass(env, + "com/hippo/ehviewer/gallery/image/GifImage"); + constructor = (*env)->GetMethodID(env, gifClazz, "", + "(JIIIII[I)V"); + if (constructor == 0) { + GIF_Free((JNIEnv*)NULL, gif); + return NULL; + } else { + return (*env)->NewObject(env, gifClazz, constructor, (jlong) (intptr_t) gif, + FILE_FORMAT_GIF, realWidth, gifFile->SHeight, format, + DEFAULT_TYPE, delayArray); + } +} + +jobject GIF_DecodeStream(JNIEnv* env, jobject is, jint format) { + + StreamContainer* sc; + GifFileType* gifFile; + jobject gifImage; + + if (format == FORMAT_AUTO) + format = FORMAT_RGB; + + sc = getStreamContainer(env, is); + if (sc == NULL) + return NULL; + + // Open gif file + gifFile = DGifOpen(sc, &streamReadFun, &errorCode); + if (gifFile == NULL) { + closeStreamContainer(env, sc); + clearException(env); + freeStreamContainer(env, sc); + return NULL; + } + + gifImage = getObjFromGifFileType(env, gifFile, format); + if (gifImage == NULL) + DGifCloseFile(gifFile, &errorCode); + + closeStreamContainer(env, sc); + clearException(env); + freeStreamContainer(env, sc); + + return gifImage; +} + +jobject GIF_DecodeFileHandler(JNIEnv* env, FILE* fp, jint format) { + + GifFileType* gifFile; + jobject gifImage; + + if (format == FORMAT_AUTO) + format = FORMAT_RGB; + + gifFile = DGifOpen(fp, &fileReadFun, &errorCode); + if (gifFile == NULL) + return NULL; + + gifImage = getObjFromGifFileType(env, gifFile, format); + if (gifImage == NULL) + DGifCloseFile(gifFile, &errorCode); + + return gifImage; +} + +void GIF_Render(JNIEnv* env, GIF* gif, int format, int index) { + + if (format != gif->format) + return; + + render(gif, index); +} + +void GIF_Free(JNIEnv* env, GIF* gif) { + + DGifCloseFile(gif->gifFile, &errorCode); + free(gif->trans); + free(gif->disposals); + free(gif->pixels); + free(gif->bak); + free(gif); +} + diff --git a/app/src/main/jni/gif.h b/app/src/main/jni/gif.h new file mode 100644 index 000000000..22faa91d1 --- /dev/null +++ b/app/src/main/jni/gif.h @@ -0,0 +1,27 @@ +#ifndef GIF_H_ +#define GIF_H_ + +#include + +#include "image.h" +#include "giflib/gif_lib.h" + +#define DEFAULT_DELAY 100 + +typedef struct { + GifFileType* gifFile; + int* trans; + int* disposals; + int format; + void* pixels; + int pxlIndex; + void* bak; + int bakIndex; +} GIF; + +jobject GIF_DecodeStream(JNIEnv* env, jobject is, jint format); +jobject GIF_DecodeFileHandler(JNIEnv* env, FILE* fp, jint format); +void GIF_Render(JNIEnv* env, GIF* gif, int format, int index); +void GIF_Free(JNIEnv* env, GIF* gif); + +#endif /* GIF_H_ */ diff --git a/app/src/main/jni/giflib/config.h b/app/src/main/jni/giflib/config.h new file mode 100644 index 000000000..235cdb7e3 --- /dev/null +++ b/app/src/main/jni/giflib/config.h @@ -0,0 +1,72 @@ +/* config.h. Generated from config.h.in by configure. */ +/* config.h.in. Generated from configure.ac by autoheader. */ + +/* Define to 1 if you have the header file. */ +#define HAVE_DLFCN_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_FCNTL_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_INTTYPES_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_LIMITS_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_MEMORY_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STDARG_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STDINT_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STDLIB_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STRINGS_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STRING_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_SYS_STAT_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_SYS_TYPES_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_UNISTD_H 1 + +/* Define to the sub-directory in which libtool stores uninstalled libraries. + */ +#define LT_OBJDIR ".libs/" + +/* Name of package */ +#define PACKAGE "giflib" + +/* Define to the address where bug reports for this package should be sent. */ +#define PACKAGE_BUGREPORT "esr@thyrsus.com" + +/* Define to the full name of this package. */ +#define PACKAGE_NAME "giflib" + +/* Define to the full name and version of this package. */ +#define PACKAGE_STRING "giflib 5.1.1" + +/* Define to the one symbol short name of this package. */ +#define PACKAGE_TARNAME "giflib" + +/* Define to the home page for this package. */ +#define PACKAGE_URL "" + +/* Define to the version of this package. */ +#define PACKAGE_VERSION "5.1.1" + +/* Define to 1 if you have the ANSI C header files. */ +#define STDC_HEADERS 1 + +/* Version number of package */ +#define VERSION "5.1.1" diff --git a/app/src/main/jni/giflib/dgif_lib.c b/app/src/main/jni/giflib/dgif_lib.c new file mode 100644 index 000000000..320fdb9e8 --- /dev/null +++ b/app/src/main/jni/giflib/dgif_lib.c @@ -0,0 +1,1176 @@ +/****************************************************************************** + +dgif_lib.c - GIF decoding + +The functions here and in egif_lib.c are partitioned carefully so that +if you only require one of read and write capability, only one of these +two modules will be linked. Preserve this property! + +*****************************************************************************/ + +#include +#include +#include +#include +#include +#include +#include + +#ifdef _WIN32 +#include +#endif /* _WIN32 */ + +#include "gif_lib.h" +#include "gif_lib_private.h" + +/* compose unsigned little endian value */ +#define UNSIGNED_LITTLE_ENDIAN(lo, hi) ((lo) | ((hi) << 8)) + +/* avoid extra function call in case we use fread (TVT) */ +#define READ(_gif,_buf,_len) \ + (((GifFilePrivateType*)_gif->Private)->Read ? \ + ((GifFilePrivateType*)_gif->Private)->Read(_gif,_buf,_len) : \ + fread(_buf,1,_len,((GifFilePrivateType*)_gif->Private)->File)) + +static int DGifGetWord(GifFileType *GifFile, GifWord *Word); +static int DGifSetupDecompress(GifFileType *GifFile); +static int DGifDecompressLine(GifFileType *GifFile, GifPixelType *Line, + int LineLen); +static int DGifGetPrefixChar(GifPrefixType *Prefix, int Code, int ClearCode); +static int DGifDecompressInput(GifFileType *GifFile, int *Code); +static int DGifBufferedInput(GifFileType *GifFile, GifByteType *Buf, + GifByteType *NextByte); + +/****************************************************************************** + Open a new GIF file for read, given by its name. + Returns dynamically allocated GifFileType pointer which serves as the GIF + info record. +******************************************************************************/ +GifFileType * +DGifOpenFileName(const char *FileName, int *Error) +{ + int FileHandle; + GifFileType *GifFile; + + if ((FileHandle = open(FileName, O_RDONLY)) == -1) { + if (Error != NULL) + *Error = D_GIF_ERR_OPEN_FAILED; + return NULL; + } + + GifFile = DGifOpenFileHandle(FileHandle, Error); + return GifFile; +} + +/****************************************************************************** + Update a new GIF file, given its file handle. + Returns dynamically allocated GifFileType pointer which serves as the GIF + info record. +******************************************************************************/ +GifFileType * +DGifOpenFileHandle(int FileHandle, int *Error) +{ + char Buf[GIF_STAMP_LEN + 1]; + GifFileType *GifFile; + GifFilePrivateType *Private; + FILE *f; + + GifFile = (GifFileType *)malloc(sizeof(GifFileType)); + if (GifFile == NULL) { + if (Error != NULL) + *Error = D_GIF_ERR_NOT_ENOUGH_MEM; + (void)close(FileHandle); + return NULL; + } + + /*@i1@*/memset(GifFile, '\0', sizeof(GifFileType)); + + /* Belt and suspenders, in case the null pointer isn't zero */ + GifFile->SavedImages = NULL; + GifFile->SColorMap = NULL; + + Private = (GifFilePrivateType *)malloc(sizeof(GifFilePrivateType)); + if (Private == NULL) { + if (Error != NULL) + *Error = D_GIF_ERR_NOT_ENOUGH_MEM; + (void)close(FileHandle); + free((char *)GifFile); + return NULL; + } +#ifdef _WIN32 + _setmode(FileHandle, O_BINARY); /* Make sure it is in binary mode. */ +#endif /* _WIN32 */ + + f = fdopen(FileHandle, "rb"); /* Make it into a stream: */ + + /*@-mustfreeonly@*/ + GifFile->Private = (void *)Private; + Private->FileHandle = FileHandle; + Private->File = f; + Private->FileState = FILE_STATE_READ; + Private->Read = NULL; /* don't use alternate input method (TVT) */ + GifFile->UserData = NULL; /* TVT */ + /*@=mustfreeonly@*/ + + /* Let's see if this is a GIF file: */ + /* coverity[check_return] */ + if (READ(GifFile, (unsigned char *)Buf, GIF_STAMP_LEN) != GIF_STAMP_LEN) { + if (Error != NULL) + *Error = D_GIF_ERR_READ_FAILED; + (void)fclose(f); + free((char *)Private); + free((char *)GifFile); + return NULL; + } + + /* Check for GIF prefix at start of file */ + Buf[GIF_STAMP_LEN] = 0; + if (strncmp(GIF_STAMP, Buf, GIF_VERSION_POS) != 0) { + if (Error != NULL) + *Error = D_GIF_ERR_NOT_GIF_FILE; + (void)fclose(f); + free((char *)Private); + free((char *)GifFile); + return NULL; + } + + if (DGifGetScreenDesc(GifFile) == GIF_ERROR) { + (void)fclose(f); + free((char *)Private); + free((char *)GifFile); + return NULL; + } + + GifFile->Error = 0; + + /* What version of GIF? */ + Private->gif89 = (Buf[GIF_VERSION_POS] == '9'); + + return GifFile; +} + +/****************************************************************************** + GifFileType constructor with user supplied input function (TVT) +******************************************************************************/ +GifFileType * +DGifOpen(void *userData, InputFunc readFunc, int *Error) +{ + char Buf[GIF_STAMP_LEN + 1]; + GifFileType *GifFile; + GifFilePrivateType *Private; + + GifFile = (GifFileType *)malloc(sizeof(GifFileType)); + if (GifFile == NULL) { + if (Error != NULL) + *Error = D_GIF_ERR_NOT_ENOUGH_MEM; + return NULL; + } + + memset(GifFile, '\0', sizeof(GifFileType)); + + /* Belt and suspenders, in case the null pointer isn't zero */ + GifFile->SavedImages = NULL; + GifFile->SColorMap = NULL; + + Private = (GifFilePrivateType *)malloc(sizeof(GifFilePrivateType)); + if (!Private) { + if (Error != NULL) + *Error = D_GIF_ERR_NOT_ENOUGH_MEM; + free((char *)GifFile); + return NULL; + } + + GifFile->Private = (void *)Private; + Private->FileHandle = 0; + Private->File = NULL; + Private->FileState = FILE_STATE_READ; + + Private->Read = readFunc; /* TVT */ + GifFile->UserData = userData; /* TVT */ + + /* Lets see if this is a GIF file: */ + /* coverity[check_return] */ + if (READ(GifFile, (unsigned char *)Buf, GIF_STAMP_LEN) != GIF_STAMP_LEN) { + if (Error != NULL) + *Error = D_GIF_ERR_READ_FAILED; + free((char *)Private); + free((char *)GifFile); + return NULL; + } + + /* Check for GIF prefix at start of file */ + Buf[GIF_STAMP_LEN] = '\0'; + if (strncmp(GIF_STAMP, Buf, GIF_VERSION_POS) != 0) { + if (Error != NULL) + *Error = D_GIF_ERR_NOT_GIF_FILE; + free((char *)Private); + free((char *)GifFile); + return NULL; + } + + if (DGifGetScreenDesc(GifFile) == GIF_ERROR) { + free((char *)Private); + free((char *)GifFile); + if (Error != NULL) + *Error = D_GIF_ERR_NO_SCRN_DSCR; + return NULL; + } + + GifFile->Error = 0; + + /* What version of GIF? */ + Private->gif89 = (Buf[GIF_VERSION_POS] == '9'); + + return GifFile; +} + +/****************************************************************************** + This routine should be called before any other DGif calls. Note that + this routine is called automatically from DGif file open routines. +******************************************************************************/ +int +DGifGetScreenDesc(GifFileType *GifFile) +{ + int BitsPerPixel; + bool SortFlag; + GifByteType Buf[3]; + GifFilePrivateType *Private = (GifFilePrivateType *)GifFile->Private; + + if (!IS_READABLE(Private)) { + /* This file was NOT open for reading: */ + GifFile->Error = D_GIF_ERR_NOT_READABLE; + return GIF_ERROR; + } + + /* Put the screen descriptor into the file: */ + if (DGifGetWord(GifFile, &GifFile->SWidth) == GIF_ERROR || + DGifGetWord(GifFile, &GifFile->SHeight) == GIF_ERROR) + return GIF_ERROR; + + if (READ(GifFile, Buf, 3) != 3) { + GifFile->Error = D_GIF_ERR_READ_FAILED; + GifFreeMapObject(GifFile->SColorMap); + GifFile->SColorMap = NULL; + return GIF_ERROR; + } + GifFile->SColorResolution = (((Buf[0] & 0x70) + 1) >> 4) + 1; + SortFlag = (Buf[0] & 0x08) != 0; + BitsPerPixel = (Buf[0] & 0x07) + 1; + GifFile->SBackGroundColor = Buf[1]; + GifFile->AspectByte = Buf[2]; + if (Buf[0] & 0x80) { /* Do we have global color map? */ + int i; + + GifFile->SColorMap = GifMakeMapObject(1 << BitsPerPixel, NULL); + if (GifFile->SColorMap == NULL) { + GifFile->Error = D_GIF_ERR_NOT_ENOUGH_MEM; + return GIF_ERROR; + } + + /* Get the global color map: */ + GifFile->SColorMap->SortFlag = SortFlag; + for (i = 0; i < GifFile->SColorMap->ColorCount; i++) { + /* coverity[check_return] */ + if (READ(GifFile, Buf, 3) != 3) { + GifFreeMapObject(GifFile->SColorMap); + GifFile->SColorMap = NULL; + GifFile->Error = D_GIF_ERR_READ_FAILED; + return GIF_ERROR; + } + GifFile->SColorMap->Colors[i].Red = Buf[0]; + GifFile->SColorMap->Colors[i].Green = Buf[1]; + GifFile->SColorMap->Colors[i].Blue = Buf[2]; + } + } else { + GifFile->SColorMap = NULL; + } + + return GIF_OK; +} + +/****************************************************************************** + This routine should be called before any attempt to read an image. +******************************************************************************/ +int +DGifGetRecordType(GifFileType *GifFile, GifRecordType* Type) +{ + GifByteType Buf; + GifFilePrivateType *Private = (GifFilePrivateType *)GifFile->Private; + + if (!IS_READABLE(Private)) { + /* This file was NOT open for reading: */ + GifFile->Error = D_GIF_ERR_NOT_READABLE; + return GIF_ERROR; + } + + /* coverity[check_return] */ + if (READ(GifFile, &Buf, 1) != 1) { + GifFile->Error = D_GIF_ERR_READ_FAILED; + return GIF_ERROR; + } + + switch (Buf) { + case DESCRIPTOR_INTRODUCER: + *Type = IMAGE_DESC_RECORD_TYPE; + break; + case EXTENSION_INTRODUCER: + *Type = EXTENSION_RECORD_TYPE; + break; + case TERMINATOR_INTRODUCER: + *Type = TERMINATE_RECORD_TYPE; + break; + default: + *Type = UNDEFINED_RECORD_TYPE; + GifFile->Error = D_GIF_ERR_WRONG_RECORD; + return GIF_ERROR; + } + + return GIF_OK; +} + +/****************************************************************************** + This routine should be called before any attempt to read an image. + Note it is assumed the Image desc. header has been read. +******************************************************************************/ +int +DGifGetImageDesc(GifFileType *GifFile) +{ + unsigned int BitsPerPixel; + GifByteType Buf[3]; + GifFilePrivateType *Private = (GifFilePrivateType *)GifFile->Private; + SavedImage *sp; + + if (!IS_READABLE(Private)) { + /* This file was NOT open for reading: */ + GifFile->Error = D_GIF_ERR_NOT_READABLE; + return GIF_ERROR; + } + + if (DGifGetWord(GifFile, &GifFile->Image.Left) == GIF_ERROR || + DGifGetWord(GifFile, &GifFile->Image.Top) == GIF_ERROR || + DGifGetWord(GifFile, &GifFile->Image.Width) == GIF_ERROR || + DGifGetWord(GifFile, &GifFile->Image.Height) == GIF_ERROR) + return GIF_ERROR; + if (READ(GifFile, Buf, 1) != 1) { + GifFile->Error = D_GIF_ERR_READ_FAILED; + GifFreeMapObject(GifFile->Image.ColorMap); + GifFile->Image.ColorMap = NULL; + return GIF_ERROR; + } + BitsPerPixel = (Buf[0] & 0x07) + 1; + GifFile->Image.Interlace = (Buf[0] & 0x40) ? true : false; + + /* Setup the colormap */ + if (GifFile->Image.ColorMap) { + GifFreeMapObject(GifFile->Image.ColorMap); + GifFile->Image.ColorMap = NULL; + } + /* Does this image have local color map? */ + if (Buf[0] & 0x80) { + unsigned int i; + + GifFile->Image.ColorMap = GifMakeMapObject(1 << BitsPerPixel, NULL); + if (GifFile->Image.ColorMap == NULL) { + GifFile->Error = D_GIF_ERR_NOT_ENOUGH_MEM; + return GIF_ERROR; + } + + /* Get the image local color map: */ + for (i = 0; i < GifFile->Image.ColorMap->ColorCount; i++) { + /* coverity[check_return] */ + if (READ(GifFile, Buf, 3) != 3) { + GifFreeMapObject(GifFile->Image.ColorMap); + GifFile->Error = D_GIF_ERR_READ_FAILED; + GifFile->Image.ColorMap = NULL; + return GIF_ERROR; + } + GifFile->Image.ColorMap->Colors[i].Red = Buf[0]; + GifFile->Image.ColorMap->Colors[i].Green = Buf[1]; + GifFile->Image.ColorMap->Colors[i].Blue = Buf[2]; + } + } + + if (GifFile->SavedImages) { + SavedImage* new_saved_images = + (SavedImage *)realloc(GifFile->SavedImages, + sizeof(SavedImage) * (GifFile->ImageCount + 1)); + if (new_saved_images == NULL) { + GifFile->Error = D_GIF_ERR_NOT_ENOUGH_MEM; + return GIF_ERROR; + } + GifFile->SavedImages = new_saved_images; + } else { + if ((GifFile->SavedImages = + (SavedImage *) malloc(sizeof(SavedImage))) == NULL) { + GifFile->Error = D_GIF_ERR_NOT_ENOUGH_MEM; + return GIF_ERROR; + } + } + + sp = &GifFile->SavedImages[GifFile->ImageCount]; + memcpy(&sp->ImageDesc, &GifFile->Image, sizeof(GifImageDesc)); + if (GifFile->Image.ColorMap != NULL) { + sp->ImageDesc.ColorMap = GifMakeMapObject( + GifFile->Image.ColorMap->ColorCount, + GifFile->Image.ColorMap->Colors); + if (sp->ImageDesc.ColorMap == NULL) { + GifFile->Error = D_GIF_ERR_NOT_ENOUGH_MEM; + return GIF_ERROR; + } + } + sp->RasterBits = (unsigned char *)NULL; + sp->ExtensionBlockCount = 0; + sp->ExtensionBlocks = (ExtensionBlock *) NULL; + + GifFile->ImageCount++; + + Private->PixelCount = (long)GifFile->Image.Width * + (long)GifFile->Image.Height; + + /* Reset decompress algorithm parameters. */ + return DGifSetupDecompress(GifFile); +} + +/****************************************************************************** + Get one full scanned line (Line) of length LineLen from GIF file. +******************************************************************************/ +int +DGifGetLine(GifFileType *GifFile, GifPixelType *Line, int LineLen) +{ + GifByteType *Dummy; + GifFilePrivateType *Private = (GifFilePrivateType *) GifFile->Private; + + if (!IS_READABLE(Private)) { + /* This file was NOT open for reading: */ + GifFile->Error = D_GIF_ERR_NOT_READABLE; + return GIF_ERROR; + } + + if (!LineLen) + LineLen = GifFile->Image.Width; + + if ((Private->PixelCount -= LineLen) > 0xffff0000UL) { + GifFile->Error = D_GIF_ERR_DATA_TOO_BIG; + return GIF_ERROR; + } + + if (DGifDecompressLine(GifFile, Line, LineLen) == GIF_OK) { + if (Private->PixelCount == 0) { + /* We probably won't be called any more, so let's clean up + * everything before we return: need to flush out all the + * rest of image until an empty block (size 0) + * detected. We use GetCodeNext. + */ + do + if (DGifGetCodeNext(GifFile, &Dummy) == GIF_ERROR) + return GIF_ERROR; + while (Dummy != NULL) ; + } + return GIF_OK; + } else + return GIF_ERROR; +} + +/****************************************************************************** + Put one pixel (Pixel) into GIF file. +******************************************************************************/ +int +DGifGetPixel(GifFileType *GifFile, GifPixelType Pixel) +{ + GifByteType *Dummy; + GifFilePrivateType *Private = (GifFilePrivateType *) GifFile->Private; + + if (!IS_READABLE(Private)) { + /* This file was NOT open for reading: */ + GifFile->Error = D_GIF_ERR_NOT_READABLE; + return GIF_ERROR; + } + if (--Private->PixelCount > 0xffff0000UL) + { + GifFile->Error = D_GIF_ERR_DATA_TOO_BIG; + return GIF_ERROR; + } + + if (DGifDecompressLine(GifFile, &Pixel, 1) == GIF_OK) { + if (Private->PixelCount == 0) { + /* We probably won't be called any more, so let's clean up + * everything before we return: need to flush out all the + * rest of image until an empty block (size 0) + * detected. We use GetCodeNext. + */ + do + if (DGifGetCodeNext(GifFile, &Dummy) == GIF_ERROR) + return GIF_ERROR; + while (Dummy != NULL) ; + } + return GIF_OK; + } else + return GIF_ERROR; +} + +/****************************************************************************** + Get an extension block (see GIF manual) from GIF file. This routine only + returns the first data block, and DGifGetExtensionNext should be called + after this one until NULL extension is returned. + The Extension should NOT be freed by the user (not dynamically allocated). + Note it is assumed the Extension description header has been read. +******************************************************************************/ +int +DGifGetExtension(GifFileType *GifFile, int *ExtCode, GifByteType **Extension) +{ + GifByteType Buf; + GifFilePrivateType *Private = (GifFilePrivateType *)GifFile->Private; + + if (!IS_READABLE(Private)) { + /* This file was NOT open for reading: */ + GifFile->Error = D_GIF_ERR_NOT_READABLE; + return GIF_ERROR; + } + + /* coverity[check_return] */ + if (READ(GifFile, &Buf, 1) != 1) { + GifFile->Error = D_GIF_ERR_READ_FAILED; + return GIF_ERROR; + } + *ExtCode = Buf; + + return DGifGetExtensionNext(GifFile, Extension); +} + +/****************************************************************************** + Get a following extension block (see GIF manual) from GIF file. This + routine should be called until NULL Extension is returned. + The Extension should NOT be freed by the user (not dynamically allocated). +******************************************************************************/ +int +DGifGetExtensionNext(GifFileType *GifFile, GifByteType ** Extension) +{ + GifByteType Buf; + GifFilePrivateType *Private = (GifFilePrivateType *)GifFile->Private; + + if (READ(GifFile, &Buf, 1) != 1) { + GifFile->Error = D_GIF_ERR_READ_FAILED; + return GIF_ERROR; + } + if (Buf > 0) { + *Extension = Private->Buf; /* Use private unused buffer. */ + (*Extension)[0] = Buf; /* Pascal strings notation (pos. 0 is len.). */ + /* coverity[tainted_data,check_return] */ + if (READ(GifFile, &((*Extension)[1]), Buf) != Buf) { + GifFile->Error = D_GIF_ERR_READ_FAILED; + return GIF_ERROR; + } + } else + *Extension = NULL; + + return GIF_OK; +} + +/****************************************************************************** + Extract a Graphics Control Block from raw extension data +******************************************************************************/ + +int DGifExtensionToGCB(const size_t GifExtensionLength, + const GifByteType *GifExtension, + GraphicsControlBlock *GCB) +{ + if (GifExtensionLength != 4) { + return GIF_ERROR; + } + + GCB->DisposalMode = (GifExtension[0] >> 2) & 0x07; + GCB->UserInputFlag = (GifExtension[0] & 0x02) != 0; + GCB->DelayTime = UNSIGNED_LITTLE_ENDIAN(GifExtension[1], GifExtension[2]); + if (GifExtension[0] & 0x01) + GCB->TransparentColor = (int)GifExtension[3]; + else + GCB->TransparentColor = NO_TRANSPARENT_COLOR; + + return GIF_OK; +} + +/****************************************************************************** + Extract the Graphics Control Block for a saved image, if it exists. +******************************************************************************/ + +int DGifSavedExtensionToGCB(GifFileType *GifFile, + int ImageIndex, GraphicsControlBlock *GCB) +{ + int i; + + if (ImageIndex < 0 || ImageIndex > GifFile->ImageCount - 1) + return GIF_ERROR; + + GCB->DisposalMode = DISPOSAL_UNSPECIFIED; + GCB->UserInputFlag = false; + GCB->DelayTime = 0; + GCB->TransparentColor = NO_TRANSPARENT_COLOR; + + for (i = 0; i < GifFile->SavedImages[ImageIndex].ExtensionBlockCount; i++) { + ExtensionBlock *ep = &GifFile->SavedImages[ImageIndex].ExtensionBlocks[i]; + if (ep->Function == GRAPHICS_EXT_FUNC_CODE) + return DGifExtensionToGCB(ep->ByteCount, ep->Bytes, GCB); + } + + return GIF_ERROR; +} + +/****************************************************************************** + This routine should be called last, to close the GIF file. +******************************************************************************/ +int +DGifCloseFile(GifFileType *GifFile, int *ErrorCode) +{ + GifFilePrivateType *Private; + + if (GifFile == NULL || GifFile->Private == NULL) + return GIF_ERROR; + + if (GifFile->Image.ColorMap) { + GifFreeMapObject(GifFile->Image.ColorMap); + GifFile->Image.ColorMap = NULL; + } + + if (GifFile->SColorMap) { + GifFreeMapObject(GifFile->SColorMap); + GifFile->SColorMap = NULL; + } + + if (GifFile->SavedImages) { + GifFreeSavedImages(GifFile); + GifFile->SavedImages = NULL; + } + + GifFreeExtensions(&GifFile->ExtensionBlockCount, &GifFile->ExtensionBlocks); + + Private = (GifFilePrivateType *) GifFile->Private; + + if (!IS_READABLE(Private)) { + /* This file was NOT open for reading: */ + if (ErrorCode != NULL) + *ErrorCode = D_GIF_ERR_NOT_READABLE; + free((char *)GifFile->Private); + free(GifFile); + return GIF_ERROR; + } + + if (Private->File && (fclose(Private->File) != 0)) { + if (ErrorCode != NULL) + *ErrorCode = D_GIF_ERR_CLOSE_FAILED; + free((char *)GifFile->Private); + free(GifFile); + return GIF_ERROR; + } + + free((char *)GifFile->Private); + free(GifFile); + if (ErrorCode != NULL) + *ErrorCode = D_GIF_SUCCEEDED; + return GIF_OK; +} + +/****************************************************************************** + Get 2 bytes (word) from the given file: +******************************************************************************/ +static int +DGifGetWord(GifFileType *GifFile, GifWord *Word) +{ + unsigned char c[2]; + + /* coverity[check_return] */ + if (READ(GifFile, c, 2) != 2) { + GifFile->Error = D_GIF_ERR_READ_FAILED; + return GIF_ERROR; + } + + *Word = (GifWord)UNSIGNED_LITTLE_ENDIAN(c[0], c[1]); + return GIF_OK; +} + +/****************************************************************************** + Get the image code in compressed form. This routine can be called if the + information needed to be piped out as is. Obviously this is much faster + than decoding and encoding again. This routine should be followed by calls + to DGifGetCodeNext, until NULL block is returned. + The block should NOT be freed by the user (not dynamically allocated). +******************************************************************************/ +int +DGifGetCode(GifFileType *GifFile, int *CodeSize, GifByteType **CodeBlock) +{ + GifFilePrivateType *Private = (GifFilePrivateType *)GifFile->Private; + + if (!IS_READABLE(Private)) { + /* This file was NOT open for reading: */ + GifFile->Error = D_GIF_ERR_NOT_READABLE; + return GIF_ERROR; + } + + *CodeSize = Private->BitsPerPixel; + + return DGifGetCodeNext(GifFile, CodeBlock); +} + +/****************************************************************************** + Continue to get the image code in compressed form. This routine should be + called until NULL block is returned. + The block should NOT be freed by the user (not dynamically allocated). +******************************************************************************/ +int +DGifGetCodeNext(GifFileType *GifFile, GifByteType **CodeBlock) +{ + GifByteType Buf; + GifFilePrivateType *Private = (GifFilePrivateType *)GifFile->Private; + + /* coverity[tainted_data_argument] */ + /* coverity[check_return] */ + if (READ(GifFile, &Buf, 1) != 1) { + GifFile->Error = D_GIF_ERR_READ_FAILED; + return GIF_ERROR; + } + + /* coverity[lower_bounds] */ + if (Buf > 0) { + *CodeBlock = Private->Buf; /* Use private unused buffer. */ + (*CodeBlock)[0] = Buf; /* Pascal strings notation (pos. 0 is len.). */ + /* coverity[tainted_data] */ + if (READ(GifFile, &((*CodeBlock)[1]), Buf) != Buf) { + GifFile->Error = D_GIF_ERR_READ_FAILED; + return GIF_ERROR; + } + } else { + *CodeBlock = NULL; + Private->Buf[0] = 0; /* Make sure the buffer is empty! */ + Private->PixelCount = 0; /* And local info. indicate image read. */ + } + + return GIF_OK; +} + +/****************************************************************************** + Setup the LZ decompression for this image: +******************************************************************************/ +static int +DGifSetupDecompress(GifFileType *GifFile) +{ + int i, BitsPerPixel; + GifByteType CodeSize; + GifPrefixType *Prefix; + GifFilePrivateType *Private = (GifFilePrivateType *)GifFile->Private; + + /* coverity[check_return] */ + if (READ(GifFile, &CodeSize, 1) < 1) { /* Read Code size from file. */ + return GIF_ERROR; /* Failed to read Code size. */ + } + BitsPerPixel = CodeSize; + + Private->Buf[0] = 0; /* Input Buffer empty. */ + Private->BitsPerPixel = BitsPerPixel; + Private->ClearCode = (1 << BitsPerPixel); + Private->EOFCode = Private->ClearCode + 1; + Private->RunningCode = Private->EOFCode + 1; + Private->RunningBits = BitsPerPixel + 1; /* Number of bits per code. */ + Private->MaxCode1 = 1 << Private->RunningBits; /* Max. code + 1. */ + Private->StackPtr = 0; /* No pixels on the pixel stack. */ + Private->LastCode = NO_SUCH_CODE; + Private->CrntShiftState = 0; /* No information in CrntShiftDWord. */ + Private->CrntShiftDWord = 0; + + Prefix = Private->Prefix; + for (i = 0; i <= LZ_MAX_CODE; i++) + Prefix[i] = NO_SUCH_CODE; + + return GIF_OK; +} + +/****************************************************************************** + The LZ decompression routine: + This version decompress the given GIF file into Line of length LineLen. + This routine can be called few times (one per scan line, for example), in + order the complete the whole image. +******************************************************************************/ +static int +DGifDecompressLine(GifFileType *GifFile, GifPixelType *Line, int LineLen) +{ + int i = 0; + int j, CrntCode, EOFCode, ClearCode, CrntPrefix, LastCode, StackPtr; + GifByteType *Stack, *Suffix; + GifPrefixType *Prefix; + GifFilePrivateType *Private = (GifFilePrivateType *) GifFile->Private; + + StackPtr = Private->StackPtr; + Prefix = Private->Prefix; + Suffix = Private->Suffix; + Stack = Private->Stack; + EOFCode = Private->EOFCode; + ClearCode = Private->ClearCode; + LastCode = Private->LastCode; + + if (StackPtr > LZ_MAX_CODE) { + return GIF_ERROR; + } + + if (StackPtr != 0) { + /* Let pop the stack off before continueing to read the GIF file: */ + while (StackPtr != 0 && i < LineLen) + Line[i++] = Stack[--StackPtr]; + } + + while (i < LineLen) { /* Decode LineLen items. */ + if (DGifDecompressInput(GifFile, &CrntCode) == GIF_ERROR) + return GIF_ERROR; + + if (CrntCode == EOFCode) { + /* Note however that usually we will not be here as we will stop + * decoding as soon as we got all the pixel, or EOF code will + * not be read at all, and DGifGetLine/Pixel clean everything. */ + GifFile->Error = D_GIF_ERR_EOF_TOO_SOON; + return GIF_ERROR; + } else if (CrntCode == ClearCode) { + /* We need to start over again: */ + for (j = 0; j <= LZ_MAX_CODE; j++) + Prefix[j] = NO_SUCH_CODE; + Private->RunningCode = Private->EOFCode + 1; + Private->RunningBits = Private->BitsPerPixel + 1; + Private->MaxCode1 = 1 << Private->RunningBits; + LastCode = Private->LastCode = NO_SUCH_CODE; + } else { + /* Its regular code - if in pixel range simply add it to output + * stream, otherwise trace to codes linked list until the prefix + * is in pixel range: */ + if (CrntCode < ClearCode) { + /* This is simple - its pixel scalar, so add it to output: */ + Line[i++] = CrntCode; + } else { + /* Its a code to needed to be traced: trace the linked list + * until the prefix is a pixel, while pushing the suffix + * pixels on our stack. If we done, pop the stack in reverse + * (thats what stack is good for!) order to output. */ + if (Prefix[CrntCode] == NO_SUCH_CODE) { + CrntPrefix = LastCode; + + /* Only allowed if CrntCode is exactly the running code: + * In that case CrntCode = XXXCode, CrntCode or the + * prefix code is last code and the suffix char is + * exactly the prefix of last code! */ + if (CrntCode == Private->RunningCode - 2) { + Suffix[Private->RunningCode - 2] = + Stack[StackPtr++] = DGifGetPrefixChar(Prefix, + LastCode, + ClearCode); + } else { + Suffix[Private->RunningCode - 2] = + Stack[StackPtr++] = DGifGetPrefixChar(Prefix, + CrntCode, + ClearCode); + } + } else + CrntPrefix = CrntCode; + + /* Now (if image is O.K.) we should not get a NO_SUCH_CODE + * during the trace. As we might loop forever, in case of + * defective image, we use StackPtr as loop counter and stop + * before overflowing Stack[]. */ + while (StackPtr < LZ_MAX_CODE && + CrntPrefix > ClearCode && CrntPrefix <= LZ_MAX_CODE) { + Stack[StackPtr++] = Suffix[CrntPrefix]; + CrntPrefix = Prefix[CrntPrefix]; + } + if (StackPtr >= LZ_MAX_CODE || CrntPrefix > LZ_MAX_CODE) { + GifFile->Error = D_GIF_ERR_IMAGE_DEFECT; + return GIF_ERROR; + } + /* Push the last character on stack: */ + Stack[StackPtr++] = CrntPrefix; + + /* Now lets pop all the stack into output: */ + while (StackPtr != 0 && i < LineLen) + Line[i++] = Stack[--StackPtr]; + } + if (LastCode != NO_SUCH_CODE && Prefix[Private->RunningCode - 2] == NO_SUCH_CODE) { + Prefix[Private->RunningCode - 2] = LastCode; + + if (CrntCode == Private->RunningCode - 2) { + /* Only allowed if CrntCode is exactly the running code: + * In that case CrntCode = XXXCode, CrntCode or the + * prefix code is last code and the suffix char is + * exactly the prefix of last code! */ + Suffix[Private->RunningCode - 2] = + DGifGetPrefixChar(Prefix, LastCode, ClearCode); + } else { + Suffix[Private->RunningCode - 2] = + DGifGetPrefixChar(Prefix, CrntCode, ClearCode); + } + } + LastCode = CrntCode; + } + } + + Private->LastCode = LastCode; + Private->StackPtr = StackPtr; + + return GIF_OK; +} + +/****************************************************************************** + Routine to trace the Prefixes linked list until we get a prefix which is + not code, but a pixel value (less than ClearCode). Returns that pixel value. + If image is defective, we might loop here forever, so we limit the loops to + the maximum possible if image O.k. - LZ_MAX_CODE times. +******************************************************************************/ +static int +DGifGetPrefixChar(GifPrefixType *Prefix, int Code, int ClearCode) +{ + int i = 0; + + while (Code > ClearCode && i++ <= LZ_MAX_CODE) { + if (Code > LZ_MAX_CODE) { + return NO_SUCH_CODE; + } + Code = Prefix[Code]; + } + return Code; +} + +/****************************************************************************** + Interface for accessing the LZ codes directly. Set Code to the real code + (12bits), or to -1 if EOF code is returned. +******************************************************************************/ +int +DGifGetLZCodes(GifFileType *GifFile, int *Code) +{ + GifByteType *CodeBlock; + GifFilePrivateType *Private = (GifFilePrivateType *)GifFile->Private; + + if (!IS_READABLE(Private)) { + /* This file was NOT open for reading: */ + GifFile->Error = D_GIF_ERR_NOT_READABLE; + return GIF_ERROR; + } + + if (DGifDecompressInput(GifFile, Code) == GIF_ERROR) + return GIF_ERROR; + + if (*Code == Private->EOFCode) { + /* Skip rest of codes (hopefully only NULL terminating block): */ + do { + if (DGifGetCodeNext(GifFile, &CodeBlock) == GIF_ERROR) + return GIF_ERROR; + } while (CodeBlock != NULL) ; + + *Code = -1; + } else if (*Code == Private->ClearCode) { + /* We need to start over again: */ + Private->RunningCode = Private->EOFCode + 1; + Private->RunningBits = Private->BitsPerPixel + 1; + Private->MaxCode1 = 1 << Private->RunningBits; + } + + return GIF_OK; +} + +/****************************************************************************** + The LZ decompression input routine: + This routine is responsable for the decompression of the bit stream from + 8 bits (bytes) packets, into the real codes. + Returns GIF_OK if read successfully. +******************************************************************************/ +static int +DGifDecompressInput(GifFileType *GifFile, int *Code) +{ + static const unsigned short CodeMasks[] = { + 0x0000, 0x0001, 0x0003, 0x0007, + 0x000f, 0x001f, 0x003f, 0x007f, + 0x00ff, 0x01ff, 0x03ff, 0x07ff, + 0x0fff + }; + + GifFilePrivateType *Private = (GifFilePrivateType *)GifFile->Private; + + GifByteType NextByte; + + /* The image can't contain more than LZ_BITS per code. */ + if (Private->RunningBits > LZ_BITS) { + GifFile->Error = D_GIF_ERR_IMAGE_DEFECT; + return GIF_ERROR; + } + + while (Private->CrntShiftState < Private->RunningBits) { + /* Needs to get more bytes from input stream for next code: */ + if (DGifBufferedInput(GifFile, Private->Buf, &NextByte) == GIF_ERROR) { + return GIF_ERROR; + } + Private->CrntShiftDWord |= + ((unsigned long)NextByte) << Private->CrntShiftState; + Private->CrntShiftState += 8; + } + *Code = Private->CrntShiftDWord & CodeMasks[Private->RunningBits]; + + Private->CrntShiftDWord >>= Private->RunningBits; + Private->CrntShiftState -= Private->RunningBits; + + /* If code cannot fit into RunningBits bits, must raise its size. Note + * however that codes above 4095 are used for special signaling. + * If we're using LZ_BITS bits already and we're at the max code, just + * keep using the table as it is, don't increment Private->RunningCode. + */ + if (Private->RunningCode < LZ_MAX_CODE + 2 && + ++Private->RunningCode > Private->MaxCode1 && + Private->RunningBits < LZ_BITS) { + Private->MaxCode1 <<= 1; + Private->RunningBits++; + } + return GIF_OK; +} + +/****************************************************************************** + This routines read one GIF data block at a time and buffers it internally + so that the decompression routine could access it. + The routine returns the next byte from its internal buffer (or read next + block in if buffer empty) and returns GIF_OK if succesful. +******************************************************************************/ +static int +DGifBufferedInput(GifFileType *GifFile, GifByteType *Buf, GifByteType *NextByte) +{ + if (Buf[0] == 0) { + /* Needs to read the next buffer - this one is empty: */ + /* coverity[check_return] */ + if (READ(GifFile, Buf, 1) != 1) { + GifFile->Error = D_GIF_ERR_READ_FAILED; + return GIF_ERROR; + } + /* There shouldn't be any empty data blocks here as the LZW spec + * says the LZW termination code should come first. Therefore we + * shouldn't be inside this routine at that point. + */ + if (Buf[0] == 0) { + GifFile->Error = D_GIF_ERR_IMAGE_DEFECT; + return GIF_ERROR; + } + if (READ(GifFile, &Buf[1], Buf[0]) != Buf[0]) { + GifFile->Error = D_GIF_ERR_READ_FAILED; + return GIF_ERROR; + } + *NextByte = Buf[1]; + Buf[1] = 2; /* We use now the second place as last char read! */ + Buf[0]--; + } else { + *NextByte = Buf[Buf[1]++]; + Buf[0]--; + } + + return GIF_OK; +} + +/****************************************************************************** + This routine reads an entire GIF into core, hanging all its state info off + the GifFileType pointer. Call DGifOpenFileName() or DGifOpenFileHandle() + first to initialize I/O. Its inverse is EGifSpew(). +*******************************************************************************/ +int +DGifSlurp(GifFileType *GifFile) +{ + size_t ImageSize; + GifRecordType RecordType; + SavedImage *sp; + GifByteType *ExtData; + int ExtFunction; + + GifFile->ExtensionBlocks = NULL; + GifFile->ExtensionBlockCount = 0; + + do { + if (DGifGetRecordType(GifFile, &RecordType) == GIF_ERROR) + return (GIF_ERROR); + + switch (RecordType) { + case IMAGE_DESC_RECORD_TYPE: + if (DGifGetImageDesc(GifFile) == GIF_ERROR) + return (GIF_ERROR); + + sp = &GifFile->SavedImages[GifFile->ImageCount - 1]; + /* Allocate memory for the image */ + if (sp->ImageDesc.Width < 0 && sp->ImageDesc.Height < 0 && + sp->ImageDesc.Width > (INT_MAX / sp->ImageDesc.Height)) { + return GIF_ERROR; + } + ImageSize = sp->ImageDesc.Width * sp->ImageDesc.Height; + + if (ImageSize > (SIZE_MAX / sizeof(GifPixelType))) { + return GIF_ERROR; + } + sp->RasterBits = (unsigned char *)malloc(ImageSize * + sizeof(GifPixelType)); + + if (sp->RasterBits == NULL) { + return GIF_ERROR; + } + + if (sp->ImageDesc.Interlace) { + int i, j; + /* + * The way an interlaced image should be read - + * offsets and jumps... + */ + int InterlacedOffset[] = { 0, 4, 2, 1 }; + int InterlacedJumps[] = { 8, 8, 4, 2 }; + /* Need to perform 4 passes on the image */ + for (i = 0; i < 4; i++) + for (j = InterlacedOffset[i]; + j < sp->ImageDesc.Height; + j += InterlacedJumps[i]) { + if (DGifGetLine(GifFile, + sp->RasterBits+j*sp->ImageDesc.Width, + sp->ImageDesc.Width) == GIF_ERROR) + return GIF_ERROR; + } + } + else { + if (DGifGetLine(GifFile,sp->RasterBits,ImageSize)==GIF_ERROR) + return (GIF_ERROR); + } + + if (GifFile->ExtensionBlocks) { + sp->ExtensionBlocks = GifFile->ExtensionBlocks; + sp->ExtensionBlockCount = GifFile->ExtensionBlockCount; + + GifFile->ExtensionBlocks = NULL; + GifFile->ExtensionBlockCount = 0; + } + break; + + case EXTENSION_RECORD_TYPE: + if (DGifGetExtension(GifFile,&ExtFunction,&ExtData) == GIF_ERROR) + return (GIF_ERROR); + /* Create an extension block with our data */ + if (ExtData != NULL) { + if (GifAddExtensionBlock(&GifFile->ExtensionBlockCount, + &GifFile->ExtensionBlocks, + ExtFunction, ExtData[0], &ExtData[1]) + == GIF_ERROR) + return (GIF_ERROR); + } + while (ExtData != NULL) { + if (DGifGetExtensionNext(GifFile, &ExtData) == GIF_ERROR) + return (GIF_ERROR); + /* Continue the extension block */ + if (ExtData != NULL) + if (GifAddExtensionBlock(&GifFile->ExtensionBlockCount, + &GifFile->ExtensionBlocks, + CONTINUE_EXT_FUNC_CODE, + ExtData[0], &ExtData[1]) == GIF_ERROR) + return (GIF_ERROR); + } + break; + + case TERMINATE_RECORD_TYPE: + break; + + default: /* Should be trapped by DGifGetRecordType */ + break; + } + } while (RecordType != TERMINATE_RECORD_TYPE); + + return (GIF_OK); +} + +/* end */ diff --git a/app/src/main/jni/giflib/gif_hash.c b/app/src/main/jni/giflib/gif_hash.c new file mode 100644 index 000000000..61a4d139c --- /dev/null +++ b/app/src/main/jni/giflib/gif_hash.c @@ -0,0 +1,132 @@ +/***************************************************************************** + +gif_hash.c -- module to support the following operations: + +1. InitHashTable - initialize hash table. +2. ClearHashTable - clear the hash table to an empty state. +2. InsertHashTable - insert one item into data structure. +3. ExistsHashTable - test if item exists in data structure. + +This module is used to hash the GIF codes during encoding. + +*****************************************************************************/ + +#include +#include +#include +#include +#include +#include + +#include "gif_lib.h" +#include "gif_hash.h" +#include "gif_lib_private.h" + +/* #define DEBUG_HIT_RATE Debug number of misses per hash Insert/Exists. */ + +#ifdef DEBUG_HIT_RATE +static long NumberOfTests = 0, + NumberOfMisses = 0; +#endif /* DEBUG_HIT_RATE */ + +static int KeyItem(uint32_t Item); + +/****************************************************************************** + Initialize HashTable - allocate the memory needed and clear it. * +******************************************************************************/ +GifHashTableType *_InitHashTable(void) +{ + GifHashTableType *HashTable; + + if ((HashTable = (GifHashTableType *) malloc(sizeof(GifHashTableType))) + == NULL) + return NULL; + + _ClearHashTable(HashTable); + + return HashTable; +} + +/****************************************************************************** + Routine to clear the HashTable to an empty state. * + This part is a little machine depended. Use the commented part otherwise. * +******************************************************************************/ +void _ClearHashTable(GifHashTableType *HashTable) +{ + memset(HashTable -> HTable, 0xFF, HT_SIZE * sizeof(uint32_t)); +} + +/****************************************************************************** + Routine to insert a new Item into the HashTable. The data is assumed to be * + new one. * +******************************************************************************/ +void _InsertHashTable(GifHashTableType *HashTable, uint32_t Key, int Code) +{ + int HKey = KeyItem(Key); + uint32_t *HTable = HashTable -> HTable; + +#ifdef DEBUG_HIT_RATE + NumberOfTests++; + NumberOfMisses++; +#endif /* DEBUG_HIT_RATE */ + + while (HT_GET_KEY(HTable[HKey]) != 0xFFFFFL) { +#ifdef DEBUG_HIT_RATE + NumberOfMisses++; +#endif /* DEBUG_HIT_RATE */ + HKey = (HKey + 1) & HT_KEY_MASK; + } + HTable[HKey] = HT_PUT_KEY(Key) | HT_PUT_CODE(Code); +} + +/****************************************************************************** + Routine to test if given Key exists in HashTable and if so returns its code * + Returns the Code if key was found, -1 if not. * +******************************************************************************/ +int _ExistsHashTable(GifHashTableType *HashTable, uint32_t Key) +{ + int HKey = KeyItem(Key); + uint32_t *HTable = HashTable -> HTable, HTKey; + +#ifdef DEBUG_HIT_RATE + NumberOfTests++; + NumberOfMisses++; +#endif /* DEBUG_HIT_RATE */ + + while ((HTKey = HT_GET_KEY(HTable[HKey])) != 0xFFFFFL) { +#ifdef DEBUG_HIT_RATE + NumberOfMisses++; +#endif /* DEBUG_HIT_RATE */ + if (Key == HTKey) return HT_GET_CODE(HTable[HKey]); + HKey = (HKey + 1) & HT_KEY_MASK; + } + + return -1; +} + +/****************************************************************************** + Routine to generate an HKey for the hashtable out of the given unique key. * + The given Key is assumed to be 20 bits as follows: lower 8 bits are the * + new postfix character, while the upper 12 bits are the prefix code. * + Because the average hit ratio is only 2 (2 hash references per entry), * + evaluating more complex keys (such as twin prime keys) does not worth it! * +******************************************************************************/ +static int KeyItem(uint32_t Item) +{ + return ((Item >> 12) ^ Item) & HT_KEY_MASK; +} + +#ifdef DEBUG_HIT_RATE +/****************************************************************************** + Debugging routine to print the hit ratio - number of times the hash table * + was tested per operation. This routine was used to test the KeyItem routine * +******************************************************************************/ +void HashTablePrintHitRatio(void) +{ + printf("Hash Table Hit Ratio is %ld/%ld = %ld%%.\n", + NumberOfMisses, NumberOfTests, + NumberOfMisses * 100 / NumberOfTests); +} +#endif /* DEBUG_HIT_RATE */ + +/* end */ diff --git a/app/src/main/jni/giflib/gif_hash.h b/app/src/main/jni/giflib/gif_hash.h new file mode 100644 index 000000000..ac20a43cb --- /dev/null +++ b/app/src/main/jni/giflib/gif_hash.h @@ -0,0 +1,39 @@ +/****************************************************************************** + +gif_hash.h - magfic constants and declarations for GIF LZW + +******************************************************************************/ + +#ifndef _GIF_HASH_H_ +#define _GIF_HASH_H_ + +#include +#include + +#define HT_SIZE 8192 /* 12bits = 4096 or twice as big! */ +#define HT_KEY_MASK 0x1FFF /* 13bits keys */ +#define HT_KEY_NUM_BITS 13 /* 13bits keys */ +#define HT_MAX_KEY 8191 /* 13bits - 1, maximal code possible */ +#define HT_MAX_CODE 4095 /* Biggest code possible in 12 bits. */ + +/* The 32 bits of the long are divided into two parts for the key & code: */ +/* 1. The code is 12 bits as our compression algorithm is limited to 12bits */ +/* 2. The key is 12 bits Prefix code + 8 bit new char or 20 bits. */ +/* The key is the upper 20 bits. The code is the lower 12. */ +#define HT_GET_KEY(l) (l >> 12) +#define HT_GET_CODE(l) (l & 0x0FFF) +#define HT_PUT_KEY(l) (l << 12) +#define HT_PUT_CODE(l) (l & 0x0FFF) + +typedef struct GifHashTableType { + uint32_t HTable[HT_SIZE]; +} GifHashTableType; + +GifHashTableType *_InitHashTable(void); +void _ClearHashTable(GifHashTableType *HashTable); +void _InsertHashTable(GifHashTableType *HashTable, uint32_t Key, int Code); +int _ExistsHashTable(GifHashTableType *HashTable, uint32_t Key); + +#endif /* _GIF_HASH_H_ */ + +/* end */ diff --git a/app/src/main/jni/giflib/gif_lib.h b/app/src/main/jni/giflib/gif_lib.h new file mode 100644 index 000000000..ac0307d75 --- /dev/null +++ b/app/src/main/jni/giflib/gif_lib.h @@ -0,0 +1,309 @@ +/****************************************************************************** + +gif_lib.h - service library for decoding and encoding GIF images + +*****************************************************************************/ + +#ifndef _GIF_LIB_H_ +#define _GIF_LIB_H_ 1 + +#ifdef __cplusplus +extern "C" { +#endif /* __cplusplus */ + +#define GIFLIB_MAJOR 5 +#define GIFLIB_MINOR 1 +#define GIFLIB_RELEASE 1 + +#define GIF_ERROR 0 +#define GIF_OK 1 + +#include +#include + +#define GIF_STAMP "GIFVER" /* First chars in file - GIF stamp. */ +#define GIF_STAMP_LEN sizeof(GIF_STAMP) - 1 +#define GIF_VERSION_POS 3 /* Version first character in stamp. */ +#define GIF87_STAMP "GIF87a" /* First chars in file - GIF stamp. */ +#define GIF89_STAMP "GIF89a" /* First chars in file - GIF stamp. */ + +typedef unsigned char GifPixelType; +typedef unsigned char *GifRowType; +typedef unsigned char GifByteType; +typedef unsigned int GifPrefixType; +typedef int GifWord; + +typedef struct GifColorType { + GifByteType Red, Green, Blue; +} GifColorType; + +typedef struct ColorMapObject { + int ColorCount; + int BitsPerPixel; + bool SortFlag; + GifColorType *Colors; /* on malloc(3) heap */ +} ColorMapObject; + +typedef struct GifImageDesc { + GifWord Left, Top, Width, Height; /* Current image dimensions. */ + bool Interlace; /* Sequential/Interlaced lines. */ + ColorMapObject *ColorMap; /* The local color map */ +} GifImageDesc; + +typedef struct ExtensionBlock { + int ByteCount; + GifByteType *Bytes; /* on malloc(3) heap */ + int Function; /* The block function code */ +#define CONTINUE_EXT_FUNC_CODE 0x00 /* continuation subblock */ +#define COMMENT_EXT_FUNC_CODE 0xfe /* comment */ +#define GRAPHICS_EXT_FUNC_CODE 0xf9 /* graphics control (GIF89) */ +#define PLAINTEXT_EXT_FUNC_CODE 0x01 /* plaintext */ +#define APPLICATION_EXT_FUNC_CODE 0xff /* application block */ +} ExtensionBlock; + +typedef struct SavedImage { + GifImageDesc ImageDesc; + GifByteType *RasterBits; /* on malloc(3) heap */ + int ExtensionBlockCount; /* Count of extensions before image */ + ExtensionBlock *ExtensionBlocks; /* Extensions before image */ +} SavedImage; + +typedef struct GifFileType { + GifWord SWidth, SHeight; /* Size of virtual canvas */ + GifWord SColorResolution; /* How many colors can we generate? */ + GifWord SBackGroundColor; /* Background color for virtual canvas */ + GifByteType AspectByte; /* Used to compute pixel aspect ratio */ + ColorMapObject *SColorMap; /* Global colormap, NULL if nonexistent. */ + int ImageCount; /* Number of current image (both APIs) */ + GifImageDesc Image; /* Current image (low-level API) */ + SavedImage *SavedImages; /* Image sequence (high-level API) */ + int ExtensionBlockCount; /* Count extensions past last image */ + ExtensionBlock *ExtensionBlocks; /* Extensions past last image */ + int Error; /* Last error condition reported */ + void *UserData; /* hook to attach user data (TVT) */ + void *Private; /* Don't mess with this! */ +} GifFileType; + +#define GIF_ASPECT_RATIO(n) ((n)+15.0/64.0) + +typedef enum { + UNDEFINED_RECORD_TYPE, + SCREEN_DESC_RECORD_TYPE, + IMAGE_DESC_RECORD_TYPE, /* Begin with ',' */ + EXTENSION_RECORD_TYPE, /* Begin with '!' */ + TERMINATE_RECORD_TYPE /* Begin with ';' */ +} GifRecordType; + +/* func type to read gif data from arbitrary sources (TVT) */ +typedef int (*InputFunc) (GifFileType *, GifByteType *, int); + +/* func type to write gif data to arbitrary targets. + * Returns count of bytes written. (MRB) + */ +typedef int (*OutputFunc) (GifFileType *, const GifByteType *, int); + +/****************************************************************************** + GIF89 structures +******************************************************************************/ + +typedef struct GraphicsControlBlock { + int DisposalMode; +#define DISPOSAL_UNSPECIFIED 0 /* No disposal specified. */ +#define DISPOSE_DO_NOT 1 /* Leave image in place */ +#define DISPOSE_BACKGROUND 2 /* Set area too background color */ +#define DISPOSE_PREVIOUS 3 /* Restore to previous content */ + bool UserInputFlag; /* User confirmation required before disposal */ + int DelayTime; /* pre-display delay in 0.01sec units */ + int TransparentColor; /* Palette index for transparency, -1 if none */ +#define NO_TRANSPARENT_COLOR -1 +} GraphicsControlBlock; + +/****************************************************************************** + GIF encoding routines +******************************************************************************/ + +/* Main entry points */ +GifFileType *EGifOpenFileName(const char *GifFileName, + const bool GifTestExistence, int *Error); +GifFileType *EGifOpenFileHandle(const int GifFileHandle, int *Error); +GifFileType *EGifOpen(void *userPtr, OutputFunc writeFunc, int *Error); +int EGifSpew(GifFileType * GifFile); +const char *EGifGetGifVersion(GifFileType *GifFile); /* new in 5.x */ +int EGifCloseFile(GifFileType *GifFile, int *ErrorCode); + +#define E_GIF_SUCCEEDED 0 +#define E_GIF_ERR_OPEN_FAILED 1 /* And EGif possible errors. */ +#define E_GIF_ERR_WRITE_FAILED 2 +#define E_GIF_ERR_HAS_SCRN_DSCR 3 +#define E_GIF_ERR_HAS_IMAG_DSCR 4 +#define E_GIF_ERR_NO_COLOR_MAP 5 +#define E_GIF_ERR_DATA_TOO_BIG 6 +#define E_GIF_ERR_NOT_ENOUGH_MEM 7 +#define E_GIF_ERR_DISK_IS_FULL 8 +#define E_GIF_ERR_CLOSE_FAILED 9 +#define E_GIF_ERR_NOT_WRITEABLE 10 + +/* These are legacy. You probably do not want to call them directly */ +int EGifPutScreenDesc(GifFileType *GifFile, + const int GifWidth, const int GifHeight, + const int GifColorRes, + const int GifBackGround, + const ColorMapObject *GifColorMap); +int EGifPutImageDesc(GifFileType *GifFile, + const int GifLeft, const int GifTop, + const int GifWidth, const int GifHeight, + const bool GifInterlace, + const ColorMapObject *GifColorMap); +void EGifSetGifVersion(GifFileType *GifFile, const bool gif89); +int EGifPutLine(GifFileType *GifFile, GifPixelType *GifLine, + int GifLineLen); +int EGifPutPixel(GifFileType *GifFile, const GifPixelType GifPixel); +int EGifPutComment(GifFileType *GifFile, const char *GifComment); +int EGifPutExtensionLeader(GifFileType *GifFile, const int GifExtCode); +int EGifPutExtensionBlock(GifFileType *GifFile, + const int GifExtLen, const void *GifExtension); +int EGifPutExtensionTrailer(GifFileType *GifFile); +int EGifPutExtension(GifFileType *GifFile, const int GifExtCode, + const int GifExtLen, + const void *GifExtension); +int EGifPutCode(GifFileType *GifFile, int GifCodeSize, + const GifByteType *GifCodeBlock); +int EGifPutCodeNext(GifFileType *GifFile, + const GifByteType *GifCodeBlock); + +/****************************************************************************** + GIF decoding routines +******************************************************************************/ + +/* Main entry points */ +GifFileType *DGifOpenFileName(const char *GifFileName, int *Error); +GifFileType *DGifOpenFileHandle(int GifFileHandle, int *Error); +int DGifSlurp(GifFileType * GifFile); +GifFileType *DGifOpen(void *userPtr, InputFunc readFunc, int *Error); /* new one (TVT) */ + int DGifCloseFile(GifFileType * GifFile, int *ErrorCode); + +#define D_GIF_SUCCEEDED 0 +#define D_GIF_ERR_OPEN_FAILED 101 /* And DGif possible errors. */ +#define D_GIF_ERR_READ_FAILED 102 +#define D_GIF_ERR_NOT_GIF_FILE 103 +#define D_GIF_ERR_NO_SCRN_DSCR 104 +#define D_GIF_ERR_NO_IMAG_DSCR 105 +#define D_GIF_ERR_NO_COLOR_MAP 106 +#define D_GIF_ERR_WRONG_RECORD 107 +#define D_GIF_ERR_DATA_TOO_BIG 108 +#define D_GIF_ERR_NOT_ENOUGH_MEM 109 +#define D_GIF_ERR_CLOSE_FAILED 110 +#define D_GIF_ERR_NOT_READABLE 111 +#define D_GIF_ERR_IMAGE_DEFECT 112 +#define D_GIF_ERR_EOF_TOO_SOON 113 + +/* These are legacy. You probably do not want to call them directly */ +int DGifGetScreenDesc(GifFileType *GifFile); +int DGifGetRecordType(GifFileType *GifFile, GifRecordType *GifType); +int DGifGetImageDesc(GifFileType *GifFile); +int DGifGetLine(GifFileType *GifFile, GifPixelType *GifLine, int GifLineLen); +int DGifGetPixel(GifFileType *GifFile, GifPixelType GifPixel); +int DGifGetComment(GifFileType *GifFile, char *GifComment); +int DGifGetExtension(GifFileType *GifFile, int *GifExtCode, + GifByteType **GifExtension); +int DGifGetExtensionNext(GifFileType *GifFile, GifByteType **GifExtension); +int DGifGetCode(GifFileType *GifFile, int *GifCodeSize, + GifByteType **GifCodeBlock); +int DGifGetCodeNext(GifFileType *GifFile, GifByteType **GifCodeBlock); +int DGifGetLZCodes(GifFileType *GifFile, int *GifCode); + + +/****************************************************************************** + Color table quantization (deprecated) +******************************************************************************/ +int GifQuantizeBuffer(unsigned int Width, unsigned int Height, + int *ColorMapSize, GifByteType * RedInput, + GifByteType * GreenInput, GifByteType * BlueInput, + GifByteType * OutputBuffer, + GifColorType * OutputColorMap); + +/****************************************************************************** + Error handling and reporting. +******************************************************************************/ +extern const char *GifErrorString(int ErrorCode); /* new in 2012 - ESR */ + +/***************************************************************************** + Everything below this point is new after version 1.2, supporting `slurp + mode' for doing I/O in two big belts with all the image-bashing in core. +******************************************************************************/ + +/****************************************************************************** + Color map handling from gif_alloc.c +******************************************************************************/ + +extern ColorMapObject *GifMakeMapObject(int ColorCount, + const GifColorType *ColorMap); +extern void GifFreeMapObject(ColorMapObject *Object); +extern ColorMapObject *GifUnionColorMap(const ColorMapObject *ColorIn1, + const ColorMapObject *ColorIn2, + GifPixelType ColorTransIn2[]); +extern int GifBitSize(int n); + +/****************************************************************************** + Support for the in-core structures allocation (slurp mode). +******************************************************************************/ + +extern void GifApplyTranslation(SavedImage *Image, GifPixelType Translation[]); +extern int GifAddExtensionBlock(int *ExtensionBlock_Count, + ExtensionBlock **ExtensionBlocks, + int Function, + unsigned int Len, unsigned char ExtData[]); +extern void GifFreeExtensions(int *ExtensionBlock_Count, + ExtensionBlock **ExtensionBlocks); +extern SavedImage *GifMakeSavedImage(GifFileType *GifFile, + const SavedImage *CopyFrom); +extern void GifFreeSavedImages(GifFileType *GifFile); + +/****************************************************************************** + 5.x functions for GIF89 graphics control blocks +******************************************************************************/ + +int DGifExtensionToGCB(const size_t GifExtensionLength, + const GifByteType *GifExtension, + GraphicsControlBlock *GCB); +size_t EGifGCBToExtension(const GraphicsControlBlock *GCB, + GifByteType *GifExtension); + +int DGifSavedExtensionToGCB(GifFileType *GifFile, + int ImageIndex, + GraphicsControlBlock *GCB); +int EGifGCBToSavedExtension(const GraphicsControlBlock *GCB, + GifFileType *GifFile, + int ImageIndex); + +/****************************************************************************** + The library's internal utility font +******************************************************************************/ + +#define GIF_FONT_WIDTH 8 +#define GIF_FONT_HEIGHT 8 +extern const unsigned char GifAsciiTable8x8[][GIF_FONT_WIDTH]; + +extern void GifDrawText8x8(SavedImage *Image, + const int x, const int y, + const char *legend, const int color); + +extern void GifDrawBox(SavedImage *Image, + const int x, const int y, + const int w, const int d, const int color); + +extern void GifDrawRectangle(SavedImage *Image, + const int x, const int y, + const int w, const int d, const int color); + +extern void GifDrawBoxedText8x8(SavedImage *Image, + const int x, const int y, + const char *legend, + const int border, const int bg, const int fg); + +#ifdef __cplusplus +} +#endif /* __cplusplus */ +#endif /* _GIF_LIB_H */ + +/* end */ diff --git a/app/src/main/jni/giflib/gif_lib_private.h b/app/src/main/jni/giflib/gif_lib_private.h new file mode 100644 index 000000000..adaf5571e --- /dev/null +++ b/app/src/main/jni/giflib/gif_lib_private.h @@ -0,0 +1,59 @@ +/**************************************************************************** + +gif_lib_private.h - internal giflib routines and structures + +****************************************************************************/ + +#ifndef _GIF_LIB_PRIVATE_H +#define _GIF_LIB_PRIVATE_H + +#include "gif_lib.h" +#include "gif_hash.h" + +#define EXTENSION_INTRODUCER 0x21 +#define DESCRIPTOR_INTRODUCER 0x2c +#define TERMINATOR_INTRODUCER 0x3b + +#define LZ_MAX_CODE 4095 /* Biggest code possible in 12 bits. */ +#define LZ_BITS 12 + +#define FLUSH_OUTPUT 4096 /* Impossible code, to signal flush. */ +#define FIRST_CODE 4097 /* Impossible code, to signal first. */ +#define NO_SUCH_CODE 4098 /* Impossible code, to signal empty. */ + +#define FILE_STATE_WRITE 0x01 +#define FILE_STATE_SCREEN 0x02 +#define FILE_STATE_IMAGE 0x04 +#define FILE_STATE_READ 0x08 + +#define IS_READABLE(Private) (Private->FileState & FILE_STATE_READ) +#define IS_WRITEABLE(Private) (Private->FileState & FILE_STATE_WRITE) + +typedef struct GifFilePrivateType { + GifWord FileState, FileHandle, /* Where all this data goes to! */ + BitsPerPixel, /* Bits per pixel (Codes uses at least this + 1). */ + ClearCode, /* The CLEAR LZ code. */ + EOFCode, /* The EOF LZ code. */ + RunningCode, /* The next code algorithm can generate. */ + RunningBits, /* The number of bits required to represent RunningCode. */ + MaxCode1, /* 1 bigger than max. possible code, in RunningBits bits. */ + LastCode, /* The code before the current code. */ + CrntCode, /* Current algorithm code. */ + StackPtr, /* For character stack (see below). */ + CrntShiftState; /* Number of bits in CrntShiftDWord. */ + unsigned long CrntShiftDWord; /* For bytes decomposition into codes. */ + unsigned long PixelCount; /* Number of pixels in image. */ + FILE *File; /* File as stream. */ + InputFunc Read; /* function to read gif input (TVT) */ + OutputFunc Write; /* function to write gif output (MRB) */ + GifByteType Buf[256]; /* Compressed input is buffered here. */ + GifByteType Stack[LZ_MAX_CODE]; /* Decoded pixels are stacked here. */ + GifByteType Suffix[LZ_MAX_CODE + 1]; /* So we can trace the codes. */ + GifPrefixType Prefix[LZ_MAX_CODE + 1]; + GifHashTableType *HashTable; + bool gif89; +} GifFilePrivateType; + +#endif /* _GIF_LIB_PRIVATE_H */ + +/* end */ diff --git a/app/src/main/jni/giflib/gifalloc.c b/app/src/main/jni/giflib/gifalloc.c new file mode 100644 index 000000000..f4118d6c7 --- /dev/null +++ b/app/src/main/jni/giflib/gifalloc.c @@ -0,0 +1,411 @@ +/***************************************************************************** + + GIF construction tools + +****************************************************************************/ + +#include +#include +#include + +#include "gif_lib.h" + +#define MAX(x, y) (((x) > (y)) ? (x) : (y)) + +/****************************************************************************** + Miscellaneous utility functions +******************************************************************************/ + +/* return smallest bitfield size n will fit in */ +int +GifBitSize(int n) +{ + register int i; + + for (i = 1; i <= 8; i++) + if ((1 << i) >= n) + break; + return (i); +} + +/****************************************************************************** + Color map object functions +******************************************************************************/ + +/* + * Allocate a color map of given size; initialize with contents of + * ColorMap if that pointer is non-NULL. + */ +ColorMapObject * +GifMakeMapObject(int ColorCount, const GifColorType *ColorMap) +{ + ColorMapObject *Object; + + /*** FIXME: Our ColorCount has to be a power of two. Is it necessary to + * make the user know that or should we automatically round up instead? */ + if (ColorCount != (1 << GifBitSize(ColorCount))) { + return ((ColorMapObject *) NULL); + } + + Object = (ColorMapObject *)malloc(sizeof(ColorMapObject)); + if (Object == (ColorMapObject *) NULL) { + return ((ColorMapObject *) NULL); + } + + Object->Colors = (GifColorType *)calloc(ColorCount, sizeof(GifColorType)); + if (Object->Colors == (GifColorType *) NULL) { + free(Object); + return ((ColorMapObject *) NULL); + } + + Object->ColorCount = ColorCount; + Object->BitsPerPixel = GifBitSize(ColorCount); + Object->SortFlag = false; + + if (ColorMap != NULL) { + memcpy((char *)Object->Colors, + (char *)ColorMap, ColorCount * sizeof(GifColorType)); + } + + return (Object); +} + +/******************************************************************************* +Free a color map object +*******************************************************************************/ +void +GifFreeMapObject(ColorMapObject *Object) +{ + if (Object != NULL) { + (void)free(Object->Colors); + (void)free(Object); + } +} + +#ifdef DEBUG +void +DumpColorMap(ColorMapObject *Object, + FILE * fp) +{ + if (Object != NULL) { + int i, j, Len = Object->ColorCount; + + for (i = 0; i < Len; i += 4) { + for (j = 0; j < 4 && j < Len; j++) { + (void)fprintf(fp, "%3d: %02x %02x %02x ", i + j, + Object->Colors[i + j].Red, + Object->Colors[i + j].Green, + Object->Colors[i + j].Blue); + } + (void)fprintf(fp, "\n"); + } + } +} +#endif /* DEBUG */ + +/******************************************************************************* + Compute the union of two given color maps and return it. If result can't + fit into 256 colors, NULL is returned, the allocated union otherwise. + ColorIn1 is copied as is to ColorUnion, while colors from ColorIn2 are + copied iff they didn't exist before. ColorTransIn2 maps the old + ColorIn2 into the ColorUnion color map table./ +*******************************************************************************/ +ColorMapObject * +GifUnionColorMap(const ColorMapObject *ColorIn1, + const ColorMapObject *ColorIn2, + GifPixelType ColorTransIn2[]) +{ + int i, j, CrntSlot, RoundUpTo, NewGifBitSize; + ColorMapObject *ColorUnion; + + /* + * We don't worry about duplicates within either color map; if + * the caller wants to resolve those, he can perform unions + * with an empty color map. + */ + + /* Allocate table which will hold the result for sure. */ + ColorUnion = GifMakeMapObject(MAX(ColorIn1->ColorCount, + ColorIn2->ColorCount) * 2, NULL); + + if (ColorUnion == NULL) + return (NULL); + + /* + * Copy ColorIn1 to ColorUnion. + */ + for (i = 0; i < ColorIn1->ColorCount; i++) + ColorUnion->Colors[i] = ColorIn1->Colors[i]; + CrntSlot = ColorIn1->ColorCount; + + /* + * Potentially obnoxious hack: + * + * Back CrntSlot down past all contiguous {0, 0, 0} slots at the end + * of table 1. This is very useful if your display is limited to + * 16 colors. + */ + while (ColorIn1->Colors[CrntSlot - 1].Red == 0 + && ColorIn1->Colors[CrntSlot - 1].Green == 0 + && ColorIn1->Colors[CrntSlot - 1].Blue == 0) + CrntSlot--; + + /* Copy ColorIn2 to ColorUnion (use old colors if they exist): */ + for (i = 0; i < ColorIn2->ColorCount && CrntSlot <= 256; i++) { + /* Let's see if this color already exists: */ + for (j = 0; j < ColorIn1->ColorCount; j++) + if (memcmp (&ColorIn1->Colors[j], &ColorIn2->Colors[i], + sizeof(GifColorType)) == 0) + break; + + if (j < ColorIn1->ColorCount) + ColorTransIn2[i] = j; /* color exists in Color1 */ + else { + /* Color is new - copy it to a new slot: */ + ColorUnion->Colors[CrntSlot] = ColorIn2->Colors[i]; + ColorTransIn2[i] = CrntSlot++; + } + } + + if (CrntSlot > 256) { + GifFreeMapObject(ColorUnion); + return ((ColorMapObject *) NULL); + } + + NewGifBitSize = GifBitSize(CrntSlot); + RoundUpTo = (1 << NewGifBitSize); + + if (RoundUpTo != ColorUnion->ColorCount) { + register GifColorType *Map = ColorUnion->Colors; + + /* + * Zero out slots up to next power of 2. + * We know these slots exist because of the way ColorUnion's + * start dimension was computed. + */ + for (j = CrntSlot; j < RoundUpTo; j++) + Map[j].Red = Map[j].Green = Map[j].Blue = 0; + + /* perhaps we can shrink the map? */ + if (RoundUpTo < ColorUnion->ColorCount) { + GifColorType *new_map = (GifColorType *)realloc(Map, + sizeof(GifColorType) * RoundUpTo); + if( new_map == NULL ) { + GifFreeMapObject(ColorUnion); + return ((ColorMapObject *) NULL); + } + ColorUnion->Colors = new_map; + } + } + + ColorUnion->ColorCount = RoundUpTo; + ColorUnion->BitsPerPixel = NewGifBitSize; + + return (ColorUnion); +} + +/******************************************************************************* + Apply a given color translation to the raster bits of an image +*******************************************************************************/ +void +GifApplyTranslation(SavedImage *Image, GifPixelType Translation[]) +{ + register int i; + register int RasterSize = Image->ImageDesc.Height * Image->ImageDesc.Width; + + for (i = 0; i < RasterSize; i++) + Image->RasterBits[i] = Translation[Image->RasterBits[i]]; +} + +/****************************************************************************** + Extension record functions +******************************************************************************/ +int +GifAddExtensionBlock(int *ExtensionBlockCount, + ExtensionBlock **ExtensionBlocks, + int Function, + unsigned int Len, + unsigned char ExtData[]) +{ + ExtensionBlock *ep; + + if (*ExtensionBlocks == NULL) + *ExtensionBlocks=(ExtensionBlock *)malloc(sizeof(ExtensionBlock)); + else { + ExtensionBlock* ep_new = (ExtensionBlock *)realloc(*ExtensionBlocks, + sizeof(ExtensionBlock) * + (*ExtensionBlockCount + 1)); + if( ep_new == NULL ) + return (GIF_ERROR); + *ExtensionBlocks = ep_new; + } + + if (*ExtensionBlocks == NULL) + return (GIF_ERROR); + + ep = &(*ExtensionBlocks)[(*ExtensionBlockCount)++]; + + ep->Function = Function; + ep->ByteCount=Len; + ep->Bytes = (GifByteType *)malloc(ep->ByteCount); + if (ep->Bytes == NULL) + return (GIF_ERROR); + + if (ExtData != NULL) { + memcpy(ep->Bytes, ExtData, Len); + } + + return (GIF_OK); +} + +void +GifFreeExtensions(int *ExtensionBlockCount, + ExtensionBlock **ExtensionBlocks) +{ + ExtensionBlock *ep; + + if (*ExtensionBlocks == NULL) + return; + + for (ep = *ExtensionBlocks; + ep < (*ExtensionBlocks + *ExtensionBlockCount); + ep++) + (void)free((char *)ep->Bytes); + (void)free((char *)*ExtensionBlocks); + *ExtensionBlocks = NULL; + *ExtensionBlockCount = 0; +} + +/****************************************************************************** + Image block allocation functions +******************************************************************************/ + +/* Private Function: + * Frees the last image in the GifFile->SavedImages array + */ +void +FreeLastSavedImage(GifFileType *GifFile) +{ + SavedImage *sp; + + if ((GifFile == NULL) || (GifFile->SavedImages == NULL)) + return; + + /* Remove one SavedImage from the GifFile */ + GifFile->ImageCount--; + sp = &GifFile->SavedImages[GifFile->ImageCount]; + + /* Deallocate its Colormap */ + if (sp->ImageDesc.ColorMap != NULL) { + GifFreeMapObject(sp->ImageDesc.ColorMap); + sp->ImageDesc.ColorMap = NULL; + } + + /* Deallocate the image data */ + if (sp->RasterBits != NULL) + free((char *)sp->RasterBits); + + /* Deallocate any extensions */ + GifFreeExtensions(&sp->ExtensionBlockCount, &sp->ExtensionBlocks); + + /*** FIXME: We could realloc the GifFile->SavedImages structure but is + * there a point to it? Saves some memory but we'd have to do it every + * time. If this is used in GifFreeSavedImages then it would be inefficient + * (The whole array is going to be deallocated.) If we just use it when + * we want to free the last Image it's convenient to do it here. + */ +} + +/* + * Append an image block to the SavedImages array + */ +SavedImage * +GifMakeSavedImage(GifFileType *GifFile, const SavedImage *CopyFrom) +{ + if (GifFile->SavedImages == NULL) + GifFile->SavedImages = (SavedImage *)malloc(sizeof(SavedImage)); + else + GifFile->SavedImages = (SavedImage *)realloc(GifFile->SavedImages, + sizeof(SavedImage) * (GifFile->ImageCount + 1)); + + if (GifFile->SavedImages == NULL) + return ((SavedImage *)NULL); + else { + SavedImage *sp = &GifFile->SavedImages[GifFile->ImageCount++]; + memset((char *)sp, '\0', sizeof(SavedImage)); + + if (CopyFrom != NULL) { + memcpy((char *)sp, CopyFrom, sizeof(SavedImage)); + + /* + * Make our own allocated copies of the heap fields in the + * copied record. This guards against potential aliasing + * problems. + */ + + /* first, the local color map */ + if (sp->ImageDesc.ColorMap != NULL) { + sp->ImageDesc.ColorMap = GifMakeMapObject( + CopyFrom->ImageDesc.ColorMap->ColorCount, + CopyFrom->ImageDesc.ColorMap->Colors); + if (sp->ImageDesc.ColorMap == NULL) { + FreeLastSavedImage(GifFile); + return (SavedImage *)(NULL); + } + } + + /* next, the raster */ + sp->RasterBits = (unsigned char *)malloc(sizeof(GifPixelType) * + CopyFrom->ImageDesc.Height * + CopyFrom->ImageDesc.Width); + if (sp->RasterBits == NULL) { + FreeLastSavedImage(GifFile); + return (SavedImage *)(NULL); + } + memcpy(sp->RasterBits, CopyFrom->RasterBits, + sizeof(GifPixelType) * CopyFrom->ImageDesc.Height * + CopyFrom->ImageDesc.Width); + + /* finally, the extension blocks */ + if (sp->ExtensionBlocks != NULL) { + sp->ExtensionBlocks = (ExtensionBlock *)malloc( + sizeof(ExtensionBlock) * + CopyFrom->ExtensionBlockCount); + if (sp->ExtensionBlocks == NULL) { + FreeLastSavedImage(GifFile); + return (SavedImage *)(NULL); + } + memcpy(sp->ExtensionBlocks, CopyFrom->ExtensionBlocks, + sizeof(ExtensionBlock) * CopyFrom->ExtensionBlockCount); + } + } + + return (sp); + } +} + +void +GifFreeSavedImages(GifFileType *GifFile) +{ + SavedImage *sp; + + if ((GifFile == NULL) || (GifFile->SavedImages == NULL)) { + return; + } + for (sp = GifFile->SavedImages; + sp < GifFile->SavedImages + GifFile->ImageCount; sp++) { + if (sp->ImageDesc.ColorMap != NULL) { + GifFreeMapObject(sp->ImageDesc.ColorMap); + sp->ImageDesc.ColorMap = NULL; + } + + if (sp->RasterBits != NULL) + free((char *)sp->RasterBits); + + GifFreeExtensions(&sp->ExtensionBlockCount, &sp->ExtensionBlocks); + } + free((char *)GifFile->SavedImages); + GifFile->SavedImages = NULL; +} + +/* end */ diff --git a/app/src/main/jni/image.c b/app/src/main/jni/image.c new file mode 100644 index 000000000..2e3a0dc96 --- /dev/null +++ b/app/src/main/jni/image.c @@ -0,0 +1,176 @@ +#include "image.h" +#include "jpeg.h" +#include "gif.h" +#include "png.h" +#include "giflib/gif_lib.h" + +static int checkFormat(int format) { + if (format != FORMAT_AUTO && format != FORMAT_GRAY && format != FORMAT_GRAY_ALPHA + && format != FORMAT_RGB && format != FORMAT_RGBA) + return FORMAT_AUTO; + else + return format; +} + +// TODO +static int detectFileFormat() { + return FILE_FORMAT_GIF; +} + +/** + * + */ +static FILE* detectFileFormatPathName(const char* namePath, int* format) { + + char jpegSig[2] = {0xFF, 0xD8}; + char pngSig[2] = {137, 80}; + char gifSig[2] = {71, 73}; + char buf[2]; + + FILE* fp = fopen(namePath, "r"); + if (fp == NULL) { + *format = FILE_FORMAT_UNKNOWN; + return NULL; + } + + if (fread(buf, 1, 2, fp) != 2) { + fclose(fp); + *format = FILE_FORMAT_UNKNOWN; + return NULL; + } + rewind(fp); + + if (!memcmp(buf, jpegSig, 2)) + *format = FILE_FORMAT_JPEG; + else if (!memcmp(buf, pngSig, 2)) + *format = FILE_FORMAT_PNG; + else if (!memcmp(buf, gifSig, 2)) + *format = FILE_FORMAT_GIF; + else + *format = FILE_FORMAT_UNKNOWN; + + return fp; +} + +JNIEXPORT jobject JNICALL +Java_com_hippo_ehviewer_gallery_image_Image_nativeDecodeStream(JNIEnv* env, + jclass clazz, jobject is, jint format) { + + format = checkFormat(format); + + switch (detectFileFormat()) { + case FILE_FORMAT_JPEG: + return JPEG_DecodeStream(env, is, format); + + case FILE_FORMAT_PNG: + return PNG_DecodeStream(env, is, format); + + case FILE_FORMAT_BMP: + return NULL; + + case FILE_FORMAT_GIF: + return GIF_DecodeStream(env, is, format); + + default: + return NULL; + } +} + +JNIEXPORT jobject JNICALL +Java_com_hippo_ehviewer_gallery_image_Image_nativeDecodeFile(JNIEnv* env, + jclass clazz, jstring namePath, jint format) { + + int fileFormat; + const char *str; + FILE* fp; + jobject image; + + format = checkFormat(format); + str = (*env)->GetStringUTFChars(env, namePath, 0); + + fp = detectFileFormatPathName(str, &fileFormat); + if (fp == NULL) { + (*env)->ReleaseStringUTFChars(env, namePath, str); + return NULL; + } + + switch (fileFormat) { + case FILE_FORMAT_JPEG: + image = JPEG_DecodeFileHandler(env, fp, format); + break; + case FILE_FORMAT_PNG: + image = PNG_DecodeFileHandler(env, fp, format); + break; + case FILE_FORMAT_BMP: + image = NULL; + break; + case FILE_FORMAT_GIF: + image = GIF_DecodeFileHandler(env, fp, format); + break; + default: + image = NULL; + break; + } + + (*env)->ReleaseStringUTFChars(env, namePath, str); + fclose(fp); + + return image; +} + +JNIEXPORT void JNICALL +Java_com_hippo_ehviewer_gallery_image_Image_nativeFree(JNIEnv* env, + jclass clazz, jlong nativeImage, jint fileFormat) { + + switch (fileFormat) { + case FILE_FORMAT_JPEG: + JPEG_Free(env, (JPEG*) (intptr_t) nativeImage); + break; + case FILE_FORMAT_PNG: + PNG_Free(env, (PNG*) (intptr_t) nativeImage); + break; + case FILE_FORMAT_BMP: + break; + case FILE_FORMAT_GIF: + GIF_Free(env, (GIF*) (intptr_t) nativeImage); + break; + } +} + +JNIEXPORT void JNICALL +Java_com_hippo_ehviewer_gallery_image_Image_nativeRender(JNIEnv* env, + jclass clazz, jint format, jint type, jlong nativeImage, jint fileFormat) { + + switch (fileFormat) { + case FILE_FORMAT_JPEG: + JPEG_Render(env, (JPEG*) (intptr_t) nativeImage, format); + break; + case FILE_FORMAT_PNG: + PNG_Render(env, (PNG*) (intptr_t) nativeImage, format); + break; + case FILE_FORMAT_BMP: + break; + } +} + +JNIEXPORT void JNICALL +Java_com_hippo_ehviewer_gallery_image_GifImage_nativeRender(JNIEnv* env, + jclass clazz, jint format, jint type, jlong nativeImage, jint fileFormat, + jint index) { + + if (type != DEFAULT_TYPE || fileFormat != FILE_FORMAT_GIF) + return; + + GIF_Render(env, (GIF*) (intptr_t) nativeImage, format, index); +} + +jint JNI_OnLoad(JavaVM *vm, void *reserved) { + JNIEnv* env; + if ((*vm)->GetEnv(vm, (void**) (&env), JNI_VERSION_1_6) != JNI_OK) + return -1; + g_jvm = vm; + return JNI_VERSION_1_6; +} + +void JNI_OnUnload(JavaVM *vm, void *reserved) { +} diff --git a/app/src/main/jni/image.h b/app/src/main/jni/image.h new file mode 100644 index 000000000..0c77b39d5 --- /dev/null +++ b/app/src/main/jni/image.h @@ -0,0 +1,75 @@ +#include +#include +#include +#include +#include + +#ifndef IMAGE_H_ +#define IMAGE_H_ + +#ifdef RELEASE + +#define LOGV(...) +#define LOGD(...) +#define LOGI(...) +#define LOGW(...) +#define LOGE(...) +#define LOGF(...) + +#else + +#define TAG "libimage" +#define LOGV(...) __android_log_print(ANDROID_LOG_VERBOSE, TAG ,__VA_ARGS__) +#define LOGD(...) __android_log_print(ANDROID_LOG_DEBUG, TAG ,__VA_ARGS__) +#define LOGI(...) __android_log_print(ANDROID_LOG_INFO, TAG ,__VA_ARGS__) +#define LOGW(...) __android_log_print(ANDROID_LOG_WARN, TAG ,__VA_ARGS__) +#define LOGE(...) __android_log_print(ANDROID_LOG_ERROR, TAG ,__VA_ARGS__) +#define LOGF(...) __android_log_print(ANDROID_LOG_FATAL, TAG ,__VA_ARGS__) + +#endif /* RELEASE */ + +typedef unsigned char byte; + +typedef struct { + byte l; +} lum; + +typedef struct { + byte l; + byte a; +} luma; + +typedef struct { + byte red; + byte green; + byte blue; +} rgb; + +typedef struct { + byte red; + byte green; + byte blue; + byte alpha; +} rgba; + +#define DEFAULT_TARGET GL_TEXTURE_2D +#define DEFAULT_TYPE GL_UNSIGNED_BYTE + +#define FILE_FORMAT_UNKNOWN -1 +#define FILE_FORMAT_JPEG 0x0 +#define FILE_FORMAT_PNG 0x1 +#define FILE_FORMAT_BMP 0x2 +#define FILE_FORMAT_GIF 0x3 + +#define FORMAT_AUTO 0x0 +#define FORMAT_GRAY GL_LUMINANCE +#define FORMAT_GRAY_ALPHA GL_LUMINANCE_ALPHA +#define FORMAT_RGB GL_RGB +#define FORMAT_RGBA GL_RGBA + +#define BG_LUM 0x21 +static rgb defaultBgColorRGB ={0x21, 0x21, 0x21}; + +JavaVM *g_jvm; + +#endif /* IMAGE_H_ */ diff --git a/app/src/main/jni/jpeg.c b/app/src/main/jni/jpeg.c new file mode 100644 index 000000000..999382200 --- /dev/null +++ b/app/src/main/jni/jpeg.c @@ -0,0 +1,147 @@ +#include + +#include "image.h" +#include "jpeg.h" +#include "libjpeg-turbo/jpeglib.h" + +struct my_error_mgr { + struct jpeg_error_mgr pub; + jmp_buf setjmp_buffer; +}; + +typedef struct my_error_mgr * my_error_ptr; + +char emsg[JMSG_LENGTH_MAX]; + +static void my_error_exit(j_common_ptr cinfo) { + my_error_ptr myerr = (my_error_ptr) cinfo->err; + (*cinfo->err->format_message)(cinfo, emsg); + longjmp(myerr->setjmp_buffer, 1); +} + +jobject JPEG_DecodeStream(JNIEnv* env, jobject is, jint format) { + // TODO + return NULL; +} + +jobject JPEG_DecodeFileHandler(JNIEnv* env, FILE* fp, jint format) { + + JPEG* jpeg; + struct jpeg_decompress_struct cinfo; + struct my_error_mgr jerr; + JSAMPARRAY buffer; + void* data; + int realStride; + int fakeWidth; + int fakeStride; + int readLines; + jclass imageClazz; + jmethodID constructor; + + jpeg = (JPEG*) malloc(sizeof(JPEG)); + if (jpeg == NULL) + return NULL; + + cinfo.err = jpeg_std_error(&jerr.pub); + jerr.pub.error_exit = my_error_exit; + if (setjmp(jerr.setjmp_buffer)) { + LOGE("%s", emsg); + free(jpeg); + jpeg_destroy_decompress(&cinfo); + return NULL; + } + jpeg_create_decompress(&cinfo); + jpeg_stdio_src(&cinfo, fp); + jpeg_read_header(&cinfo, TRUE); + + // Set png format + switch (format) { + case FORMAT_AUTO: + if (cinfo.jpeg_color_space == JCS_GRAYSCALE) { + format = FORMAT_GRAY; + cinfo.out_color_space == JCS_GRAYSCALE; + } else { + format = FORMAT_RGB; + cinfo.out_color_space = JCS_RGB; + } + break; + case FORMAT_RGB: + cinfo.out_color_space = JCS_RGB; + break; + case FORMAT_RGBA: + cinfo.out_color_space = JCS_EXT_RGBA; + break; + case FORMAT_GRAY: + cinfo.out_color_space = JCS_GRAYSCALE; + break; + default: + format = FORMAT_RGB; + cinfo.out_color_space = JCS_RGB; + break; + } + + jpeg_start_decompress(&cinfo); + + // Set + realStride = cinfo.output_width * cinfo.output_components; + fakeWidth = format != FORMAT_RGBA ? nextMulOf4(cinfo.output_width) : cinfo.output_width; + fakeStride = fakeWidth * cinfo.output_components; + data = malloc(fakeStride * cinfo.output_height); + if (data == NULL) { + free(jpeg); + jpeg_finish_decompress(&cinfo); + jpeg_destroy_decompress(&cinfo); + return NULL; + } + buffer = (*cinfo.mem->alloc_sarray)((j_common_ptr) &cinfo, JPOOL_IMAGE, + realStride, 3); + jpeg->data = data; + jpeg->format = format; + jpeg->width = fakeWidth; + jpeg->height = cinfo.output_height; + + while (cinfo.output_scanline < cinfo.output_height) { + readLines = jpeg_read_scanlines(&cinfo, buffer, 3); + switch(readLines) { + case 3: + memcpy(data, buffer[0], realStride); + data += fakeStride; + case 2: + memcpy(data, buffer[readLines - 2], realStride); + data += fakeStride; + case 1: + memcpy(data, buffer[readLines - 1], realStride); + data += fakeStride; + } + } + + jpeg_finish_decompress(&cinfo); + jpeg_destroy_decompress(&cinfo); + + imageClazz = (*env)->FindClass(env, + "com/hippo/ehviewer/gallery/image/Image"); + constructor = (*env)->GetMethodID(env, imageClazz, "", + "(JIIIII)V"); + if (constructor == 0) { + JPEG_Free((JNIEnv*)NULL, jpeg); + return NULL; + } else { + return (*env)->NewObject(env, imageClazz, constructor, (jlong) (intptr_t) jpeg, + FILE_FORMAT_JPEG, cinfo.output_width, cinfo.output_height, format, + DEFAULT_TYPE); + } +} + +void JPEG_Render(JNIEnv* env, JPEG* jpeg, int format) { + if (format != jpeg->format) + return; + + glTexSubImage2D(DEFAULT_TARGET, 0, 0, 0, jpeg->width, + jpeg->height, format, DEFAULT_TYPE, jpeg->data); +} + +void JPEG_Free(JNIEnv* env, JPEG* jpeg) { + + free(jpeg->data); + free(jpeg); +} diff --git a/app/src/main/jni/jpeg.h b/app/src/main/jni/jpeg.h new file mode 100644 index 000000000..bc3465d7e --- /dev/null +++ b/app/src/main/jni/jpeg.h @@ -0,0 +1,20 @@ +#ifndef JPEG_H_ +#define JPEG_H_ + +#include + +#include "image.h" + +typedef struct { + void* data; + int width; + int height; + int format; +} JPEG; + +jobject JPEG_DecodeStream(JNIEnv* env, jobject is, jint format); +jobject JPEG_DecodeFileHandler(JNIEnv* env, FILE* fp, jint format); +void JPEG_Render(JNIEnv* env, JPEG* jpeg, int format); +void JPEG_Free(JNIEnv* env, JPEG* jpeg); + +#endif /* JPEG_H_ */ diff --git a/app/src/main/jni/libjpeg-turbo/README.md b/app/src/main/jni/libjpeg-turbo/README.md new file mode 100644 index 000000000..fcdb395d5 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/README.md @@ -0,0 +1,10 @@ +# simd/jsimdext.inc + +%define EXTN(name) _ %+ name
+to
+%define EXTN(name) name
+or ld error for x86 and x86_64 + +# Add cmyk to rgb, cmyk to gray + +In jdcolor.c, jdcolext.c diff --git a/app/src/main/jni/libjpeg-turbo/config.h b/app/src/main/jni/libjpeg-turbo/config.h new file mode 100644 index 000000000..35dd2050d --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/config.h @@ -0,0 +1,145 @@ +/* config.h. Generated from config.h.in by configure. */ +/* config.h.in. Generated from configure.ac by autoheader. */ + +/* use 8 or 12 */ +/* #undef BITS_IN_JSAMPLE */ + +/* libjpeg-turbo build number */ +#define BUILD "20150109" + +/* Support arithmetic encoding */ +#define C_ARITH_CODING_SUPPORTED 1 + +/* Support arithmetic decoding */ +#define D_ARITH_CODING_SUPPORTED 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_DLFCN_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_INTTYPES_H 1 + +/* Define to 1 if you have the header file. */ +/* #undef HAVE_JNI_H */ + +/* Define to 1 if you have the header file. */ +#define HAVE_LOCALE_H 1 + +/* Define to 1 if you have the `memcpy' function. */ +#define HAVE_MEMCPY 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_MEMORY_H 1 + +/* Define to 1 if you have the `memset' function. */ +#define HAVE_MEMSET 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STDDEF_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STDINT_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STDLIB_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STRINGS_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STRING_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_SYS_STAT_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_SYS_TYPES_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_UNISTD_H 1 + +/* Define to 1 if the system has the type `unsigned char'. */ +#define HAVE_UNSIGNED_CHAR 1 + +/* Define to 1 if the system has the type `unsigned short'. */ +#define HAVE_UNSIGNED_SHORT 1 + +/* Compiler does not support pointers to undefined structures. */ +/* #undef INCOMPLETE_TYPES_BROKEN */ + +/* How to obtain function inlining. */ +#define INLINE inline __attribute__((always_inline)) + +/* libjpeg API version */ +#define JPEG_LIB_VERSION 62 + +/* libjpeg-turbo version */ +#define LIBJPEG_TURBO_VERSION 1.4.0 + +/* Define to the sub-directory in which libtool stores uninstalled libraries. + */ +#define LT_OBJDIR ".libs/" + +/* Support in-memory source/destination managers */ +#define MEM_SRCDST_SUPPORTED 1 + +/* Define if you have BSD-like bzero and bcopy in rather than + memset/memcpy in . */ +/* #undef NEED_BSD_STRINGS */ + +/* Define if you need to include to get size_t. */ +#define NEED_SYS_TYPES_H 1 + +/* Name of package */ +#define PACKAGE "libjpeg-turbo" + +/* Define to the address where bug reports for this package should be sent. */ +#define PACKAGE_BUGREPORT "" + +/* Define to the full name of this package. */ +#define PACKAGE_NAME "libjpeg-turbo" + +/* Define to the full name and version of this package. */ +#define PACKAGE_STRING "libjpeg-turbo 1.4.0" + +/* Define to the one symbol short name of this package. */ +#define PACKAGE_TARNAME "libjpeg-turbo" + +/* Define to the home page for this package. */ +#define PACKAGE_URL "" + +/* Define to the version of this package. */ +#define PACKAGE_VERSION "1.4.0" + +/* Define if your (broken) compiler shifts signed values as if they were + unsigned. */ +/* #undef RIGHT_SHIFT_IS_UNSIGNED */ + +/* The size of `long', as computed by sizeof. */ +#define SIZEOF_LONG 8 + +/* Define to 1 if you have the ANSI C header files. */ +#define STDC_HEADERS 1 + +/* Version number of package */ +#define VERSION "1.4.0" + +/* Use accelerated SIMD routines. */ +#define WITH_SIMD 1 + +/* Define to 1 if type `char' is unsigned and you are not using gcc. */ +#ifndef __CHAR_UNSIGNED__ +/* # undef __CHAR_UNSIGNED__ */ +#endif + +/* Define to empty if `const' does not conform to ANSI C. */ +/* #undef const */ + +/* Define to `__inline__' or `__inline' if that's what the C compiler + calls it, or to nothing if 'inline' is not supported under any name. */ +#ifndef __cplusplus +/* #undef inline */ +#endif + +/* Define to `unsigned int' if does not define. */ +/* #undef size_t */ diff --git a/app/src/main/jni/libjpeg-turbo/jaricom.c b/app/src/main/jni/libjpeg-turbo/jaricom.c new file mode 100644 index 000000000..f43e2ea7f --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jaricom.c @@ -0,0 +1,153 @@ +/* + * jaricom.c + * + * Developed 1997-2009 by Guido Vollbeding. + * This file is part of the Independent JPEG Group's software. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains probability estimation tables for common use in + * arithmetic entropy encoding and decoding routines. + * + * This data represents Table D.2 in the JPEG spec (ISO/IEC IS 10918-1 + * and CCITT Recommendation ITU-T T.81) and Table 24 in the JBIG spec + * (ISO/IEC IS 11544 and CCITT Recommendation ITU-T T.82). + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" + +/* The following #define specifies the packing of the four components + * into the compact INT32 representation. + * Note that this formula must match the actual arithmetic encoder + * and decoder implementation. The implementation has to be changed + * if this formula is changed. + * The current organization is leaned on Markus Kuhn's JBIG + * implementation (jbig_tab.c). + */ + +#define V(i,a,b,c,d) (((INT32)a << 16) | ((INT32)c << 8) | ((INT32)d << 7) | b) + +const INT32 jpeg_aritab[113+1] = { +/* + * Index, Qe_Value, Next_Index_LPS, Next_Index_MPS, Switch_MPS + */ + V( 0, 0x5a1d, 1, 1, 1 ), + V( 1, 0x2586, 14, 2, 0 ), + V( 2, 0x1114, 16, 3, 0 ), + V( 3, 0x080b, 18, 4, 0 ), + V( 4, 0x03d8, 20, 5, 0 ), + V( 5, 0x01da, 23, 6, 0 ), + V( 6, 0x00e5, 25, 7, 0 ), + V( 7, 0x006f, 28, 8, 0 ), + V( 8, 0x0036, 30, 9, 0 ), + V( 9, 0x001a, 33, 10, 0 ), + V( 10, 0x000d, 35, 11, 0 ), + V( 11, 0x0006, 9, 12, 0 ), + V( 12, 0x0003, 10, 13, 0 ), + V( 13, 0x0001, 12, 13, 0 ), + V( 14, 0x5a7f, 15, 15, 1 ), + V( 15, 0x3f25, 36, 16, 0 ), + V( 16, 0x2cf2, 38, 17, 0 ), + V( 17, 0x207c, 39, 18, 0 ), + V( 18, 0x17b9, 40, 19, 0 ), + V( 19, 0x1182, 42, 20, 0 ), + V( 20, 0x0cef, 43, 21, 0 ), + V( 21, 0x09a1, 45, 22, 0 ), + V( 22, 0x072f, 46, 23, 0 ), + V( 23, 0x055c, 48, 24, 0 ), + V( 24, 0x0406, 49, 25, 0 ), + V( 25, 0x0303, 51, 26, 0 ), + V( 26, 0x0240, 52, 27, 0 ), + V( 27, 0x01b1, 54, 28, 0 ), + V( 28, 0x0144, 56, 29, 0 ), + V( 29, 0x00f5, 57, 30, 0 ), + V( 30, 0x00b7, 59, 31, 0 ), + V( 31, 0x008a, 60, 32, 0 ), + V( 32, 0x0068, 62, 33, 0 ), + V( 33, 0x004e, 63, 34, 0 ), + V( 34, 0x003b, 32, 35, 0 ), + V( 35, 0x002c, 33, 9, 0 ), + V( 36, 0x5ae1, 37, 37, 1 ), + V( 37, 0x484c, 64, 38, 0 ), + V( 38, 0x3a0d, 65, 39, 0 ), + V( 39, 0x2ef1, 67, 40, 0 ), + V( 40, 0x261f, 68, 41, 0 ), + V( 41, 0x1f33, 69, 42, 0 ), + V( 42, 0x19a8, 70, 43, 0 ), + V( 43, 0x1518, 72, 44, 0 ), + V( 44, 0x1177, 73, 45, 0 ), + V( 45, 0x0e74, 74, 46, 0 ), + V( 46, 0x0bfb, 75, 47, 0 ), + V( 47, 0x09f8, 77, 48, 0 ), + V( 48, 0x0861, 78, 49, 0 ), + V( 49, 0x0706, 79, 50, 0 ), + V( 50, 0x05cd, 48, 51, 0 ), + V( 51, 0x04de, 50, 52, 0 ), + V( 52, 0x040f, 50, 53, 0 ), + V( 53, 0x0363, 51, 54, 0 ), + V( 54, 0x02d4, 52, 55, 0 ), + V( 55, 0x025c, 53, 56, 0 ), + V( 56, 0x01f8, 54, 57, 0 ), + V( 57, 0x01a4, 55, 58, 0 ), + V( 58, 0x0160, 56, 59, 0 ), + V( 59, 0x0125, 57, 60, 0 ), + V( 60, 0x00f6, 58, 61, 0 ), + V( 61, 0x00cb, 59, 62, 0 ), + V( 62, 0x00ab, 61, 63, 0 ), + V( 63, 0x008f, 61, 32, 0 ), + V( 64, 0x5b12, 65, 65, 1 ), + V( 65, 0x4d04, 80, 66, 0 ), + V( 66, 0x412c, 81, 67, 0 ), + V( 67, 0x37d8, 82, 68, 0 ), + V( 68, 0x2fe8, 83, 69, 0 ), + V( 69, 0x293c, 84, 70, 0 ), + V( 70, 0x2379, 86, 71, 0 ), + V( 71, 0x1edf, 87, 72, 0 ), + V( 72, 0x1aa9, 87, 73, 0 ), + V( 73, 0x174e, 72, 74, 0 ), + V( 74, 0x1424, 72, 75, 0 ), + V( 75, 0x119c, 74, 76, 0 ), + V( 76, 0x0f6b, 74, 77, 0 ), + V( 77, 0x0d51, 75, 78, 0 ), + V( 78, 0x0bb6, 77, 79, 0 ), + V( 79, 0x0a40, 77, 48, 0 ), + V( 80, 0x5832, 80, 81, 1 ), + V( 81, 0x4d1c, 88, 82, 0 ), + V( 82, 0x438e, 89, 83, 0 ), + V( 83, 0x3bdd, 90, 84, 0 ), + V( 84, 0x34ee, 91, 85, 0 ), + V( 85, 0x2eae, 92, 86, 0 ), + V( 86, 0x299a, 93, 87, 0 ), + V( 87, 0x2516, 86, 71, 0 ), + V( 88, 0x5570, 88, 89, 1 ), + V( 89, 0x4ca9, 95, 90, 0 ), + V( 90, 0x44d9, 96, 91, 0 ), + V( 91, 0x3e22, 97, 92, 0 ), + V( 92, 0x3824, 99, 93, 0 ), + V( 93, 0x32b4, 99, 94, 0 ), + V( 94, 0x2e17, 93, 86, 0 ), + V( 95, 0x56a8, 95, 96, 1 ), + V( 96, 0x4f46, 101, 97, 0 ), + V( 97, 0x47e5, 102, 98, 0 ), + V( 98, 0x41cf, 103, 99, 0 ), + V( 99, 0x3c3d, 104, 100, 0 ), + V( 100, 0x375e, 99, 93, 0 ), + V( 101, 0x5231, 105, 102, 0 ), + V( 102, 0x4c0f, 106, 103, 0 ), + V( 103, 0x4639, 107, 104, 0 ), + V( 104, 0x415e, 103, 99, 0 ), + V( 105, 0x5627, 105, 106, 1 ), + V( 106, 0x50e7, 108, 107, 0 ), + V( 107, 0x4b85, 109, 103, 0 ), + V( 108, 0x5597, 110, 109, 0 ), + V( 109, 0x504f, 111, 107, 0 ), + V( 110, 0x5a10, 110, 111, 1 ), + V( 111, 0x5522, 112, 109, 0 ), + V( 112, 0x59eb, 112, 111, 1 ), +/* + * This last entry is used for fixed probability estimate of 0.5 + * as recommended in Section 10.3 Table 5 of ITU-T Rec. T.851. + */ + V( 113, 0x5a1d, 113, 113, 0 ) +}; diff --git a/app/src/main/jni/libjpeg-turbo/jcomapi.c b/app/src/main/jni/libjpeg-turbo/jcomapi.c new file mode 100644 index 000000000..d8f396d16 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jcomapi.c @@ -0,0 +1,108 @@ +/* + * jcomapi.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1997, Thomas G. Lane.0 + * It was modified by The libjpeg-turbo Project to include only code relevant + * to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains application interface routines that are used for both + * compression and decompression. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" + + +/* + * Abort processing of a JPEG compression or decompression operation, + * but don't destroy the object itself. + * + * For this, we merely clean up all the nonpermanent memory pools. + * Note that temp files (virtual arrays) are not allowed to belong to + * the permanent pool, so we will be able to close all temp files here. + * Closing a data source or destination, if necessary, is the application's + * responsibility. + */ + +GLOBAL(void) +jpeg_abort (j_common_ptr cinfo) +{ + int pool; + + /* Do nothing if called on a not-initialized or destroyed JPEG object. */ + if (cinfo->mem == NULL) + return; + + /* Releasing pools in reverse order might help avoid fragmentation + * with some (brain-damaged) malloc libraries. + */ + for (pool = JPOOL_NUMPOOLS-1; pool > JPOOL_PERMANENT; pool--) { + (*cinfo->mem->free_pool) (cinfo, pool); + } + + /* Reset overall state for possible reuse of object */ + if (cinfo->is_decompressor) { + cinfo->global_state = DSTATE_START; + /* Try to keep application from accessing now-deleted marker list. + * A bit kludgy to do it here, but this is the most central place. + */ + ((j_decompress_ptr) cinfo)->marker_list = NULL; + } else { + cinfo->global_state = CSTATE_START; + } +} + + +/* + * Destruction of a JPEG object. + * + * Everything gets deallocated except the master jpeg_compress_struct itself + * and the error manager struct. Both of these are supplied by the application + * and must be freed, if necessary, by the application. (Often they are on + * the stack and so don't need to be freed anyway.) + * Closing a data source or destination, if necessary, is the application's + * responsibility. + */ + +GLOBAL(void) +jpeg_destroy (j_common_ptr cinfo) +{ + /* We need only tell the memory manager to release everything. */ + /* NB: mem pointer is NULL if memory mgr failed to initialize. */ + if (cinfo->mem != NULL) + (*cinfo->mem->self_destruct) (cinfo); + cinfo->mem = NULL; /* be safe if jpeg_destroy is called twice */ + cinfo->global_state = 0; /* mark it destroyed */ +} + + +/* + * Convenience routines for allocating quantization and Huffman tables. + * (Would jutils.c be a more reasonable place to put these?) + */ + +GLOBAL(JQUANT_TBL *) +jpeg_alloc_quant_table (j_common_ptr cinfo) +{ + JQUANT_TBL *tbl; + + tbl = (JQUANT_TBL *) + (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, sizeof(JQUANT_TBL)); + tbl->sent_table = FALSE; /* make sure this is false in any new table */ + return tbl; +} + + +GLOBAL(JHUFF_TBL *) +jpeg_alloc_huff_table (j_common_ptr cinfo) +{ + JHUFF_TBL *tbl; + + tbl = (JHUFF_TBL *) + (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, sizeof(JHUFF_TBL)); + tbl->sent_table = FALSE; /* make sure this is false in any new table */ + return tbl; +} diff --git a/app/src/main/jni/libjpeg-turbo/jconfig.h b/app/src/main/jni/libjpeg-turbo/jconfig.h new file mode 100644 index 000000000..1a479d2ac --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jconfig.h @@ -0,0 +1,71 @@ +/* jconfig.h. Generated from jconfig.h.in by configure. */ +/* Version ID for the JPEG library. + * Might be useful for tests like "#if JPEG_LIB_VERSION >= 60". + */ +#define JPEG_LIB_VERSION 62 + +/* libjpeg-turbo version */ +#define LIBJPEG_TURBO_VERSION 1.4.0 + +/* Support arithmetic encoding */ +#define C_ARITH_CODING_SUPPORTED 1 + +/* Support arithmetic decoding */ +#define D_ARITH_CODING_SUPPORTED 1 + +/* + * Define BITS_IN_JSAMPLE as either + * 8 for 8-bit sample values (the usual setting) + * 12 for 12-bit sample values + * Only 8 and 12 are legal data precisions for lossy JPEG according to the + * JPEG standard, and the IJG code does not support anything else! + * We do not support run-time selection of data precision, sorry. + */ + +#define BITS_IN_JSAMPLE 8 /* use 8 or 12 */ + +/* Define to 1 if you have the header file. */ +#define HAVE_LOCALE_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STDDEF_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STDLIB_H 1 + +/* Define to 1 if the system has the type `unsigned char'. */ +#define HAVE_UNSIGNED_CHAR 1 + +/* Define to 1 if the system has the type `unsigned short'. */ +#define HAVE_UNSIGNED_SHORT 1 + +/* Compiler does not support pointers to undefined structures. */ +/* #undef INCOMPLETE_TYPES_BROKEN */ + +/* Support in-memory source/destination managers */ +#define MEM_SRCDST_SUPPORTED 1 + +/* Define if you have BSD-like bzero and bcopy in rather than + memset/memcpy in . */ +/* #undef NEED_BSD_STRINGS */ + +/* Define if you need to include to get size_t. */ +#define NEED_SYS_TYPES_H 1 + +/* Define if your (broken) compiler shifts signed values as if they were + unsigned. */ +/* #undef RIGHT_SHIFT_IS_UNSIGNED */ + +/* Use accelerated SIMD routines. */ +#define WITH_SIMD 1 + +/* Define to 1 if type `char' is unsigned and you are not using gcc. */ +#ifndef __CHAR_UNSIGNED__ +/* # undef __CHAR_UNSIGNED__ */ +#endif + +/* Define to empty if `const' does not conform to ANSI C. */ +/* #undef const */ + +/* Define to `unsigned int' if does not define. */ +/* #undef size_t */ diff --git a/app/src/main/jni/libjpeg-turbo/jconfigint.h b/app/src/main/jni/libjpeg-turbo/jconfigint.h new file mode 100644 index 000000000..361754241 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jconfigint.h @@ -0,0 +1,12 @@ +/* jconfigint.h. Generated from jconfigint.h.in by configure. */ +/* libjpeg-turbo build number */ +#define BUILD "20150109" + +/* How to obtain function inlining. */ +#define INLINE inline __attribute__((always_inline)) + +/* Define to the full name of this package. */ +#define PACKAGE_NAME "libjpeg-turbo" + +/* Version number of package */ +#define VERSION "1.4.0" diff --git a/app/src/main/jni/libjpeg-turbo/jdapimin.c b/app/src/main/jni/libjpeg-turbo/jdapimin.c new file mode 100644 index 000000000..fc8898f8b --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdapimin.c @@ -0,0 +1,397 @@ +/* + * jdapimin.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1998, Thomas G. Lane. + * It was modified by The libjpeg-turbo Project to include only code relevant + * to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains application interface code for the decompression half + * of the JPEG library. These are the "minimum" API routines that may be + * needed in either the normal full-decompression case or the + * transcoding-only case. + * + * Most of the routines intended to be called directly by an application + * are in this file or in jdapistd.c. But also see jcomapi.c for routines + * shared by compression and decompression, and jdtrans.c for the transcoding + * case. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" + + +/* + * Initialization of a JPEG decompression object. + * The error manager must already be set up (in case memory manager fails). + */ + +GLOBAL(void) +jpeg_CreateDecompress (j_decompress_ptr cinfo, int version, size_t structsize) +{ + int i; + + /* Guard against version mismatches between library and caller. */ + cinfo->mem = NULL; /* so jpeg_destroy knows mem mgr not called */ + if (version != JPEG_LIB_VERSION) + ERREXIT2(cinfo, JERR_BAD_LIB_VERSION, JPEG_LIB_VERSION, version); + if (structsize != sizeof(struct jpeg_decompress_struct)) + ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE, + (int) sizeof(struct jpeg_decompress_struct), (int) structsize); + + /* For debugging purposes, we zero the whole master structure. + * But the application has already set the err pointer, and may have set + * client_data, so we have to save and restore those fields. + * Note: if application hasn't set client_data, tools like Purify may + * complain here. + */ + { + struct jpeg_error_mgr * err = cinfo->err; + void * client_data = cinfo->client_data; /* ignore Purify complaint here */ + MEMZERO(cinfo, sizeof(struct jpeg_decompress_struct)); + cinfo->err = err; + cinfo->client_data = client_data; + } + cinfo->is_decompressor = TRUE; + + /* Initialize a memory manager instance for this object */ + jinit_memory_mgr((j_common_ptr) cinfo); + + /* Zero out pointers to permanent structures. */ + cinfo->progress = NULL; + cinfo->src = NULL; + + for (i = 0; i < NUM_QUANT_TBLS; i++) + cinfo->quant_tbl_ptrs[i] = NULL; + + for (i = 0; i < NUM_HUFF_TBLS; i++) { + cinfo->dc_huff_tbl_ptrs[i] = NULL; + cinfo->ac_huff_tbl_ptrs[i] = NULL; + } + + /* Initialize marker processor so application can override methods + * for COM, APPn markers before calling jpeg_read_header. + */ + cinfo->marker_list = NULL; + jinit_marker_reader(cinfo); + + /* And initialize the overall input controller. */ + jinit_input_controller(cinfo); + + /* OK, I'm ready */ + cinfo->global_state = DSTATE_START; +} + + +/* + * Destruction of a JPEG decompression object + */ + +GLOBAL(void) +jpeg_destroy_decompress (j_decompress_ptr cinfo) +{ + jpeg_destroy((j_common_ptr) cinfo); /* use common routine */ +} + + +/* + * Abort processing of a JPEG decompression operation, + * but don't destroy the object itself. + */ + +GLOBAL(void) +jpeg_abort_decompress (j_decompress_ptr cinfo) +{ + jpeg_abort((j_common_ptr) cinfo); /* use common routine */ +} + + +/* + * Set default decompression parameters. + */ + +LOCAL(void) +default_decompress_parms (j_decompress_ptr cinfo) +{ + /* Guess the input colorspace, and set output colorspace accordingly. */ + /* (Wish JPEG committee had provided a real way to specify this...) */ + /* Note application may override our guesses. */ + switch (cinfo->num_components) { + case 1: + cinfo->jpeg_color_space = JCS_GRAYSCALE; + cinfo->out_color_space = JCS_GRAYSCALE; + break; + + case 3: + if (cinfo->saw_JFIF_marker) { + cinfo->jpeg_color_space = JCS_YCbCr; /* JFIF implies YCbCr */ + } else if (cinfo->saw_Adobe_marker) { + switch (cinfo->Adobe_transform) { + case 0: + cinfo->jpeg_color_space = JCS_RGB; + break; + case 1: + cinfo->jpeg_color_space = JCS_YCbCr; + break; + default: + WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform); + cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */ + break; + } + } else { + /* Saw no special markers, try to guess from the component IDs */ + int cid0 = cinfo->comp_info[0].component_id; + int cid1 = cinfo->comp_info[1].component_id; + int cid2 = cinfo->comp_info[2].component_id; + + if (cid0 == 1 && cid1 == 2 && cid2 == 3) + cinfo->jpeg_color_space = JCS_YCbCr; /* assume JFIF w/out marker */ + else if (cid0 == 82 && cid1 == 71 && cid2 == 66) + cinfo->jpeg_color_space = JCS_RGB; /* ASCII 'R', 'G', 'B' */ + else { + TRACEMS3(cinfo, 1, JTRC_UNKNOWN_IDS, cid0, cid1, cid2); + cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */ + } + } + /* Always guess RGB is proper output colorspace. */ + cinfo->out_color_space = JCS_RGB; + break; + + case 4: + if (cinfo->saw_Adobe_marker) { + switch (cinfo->Adobe_transform) { + case 0: + cinfo->jpeg_color_space = JCS_CMYK; + break; + case 2: + cinfo->jpeg_color_space = JCS_YCCK; + break; + default: + WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform); + cinfo->jpeg_color_space = JCS_YCCK; /* assume it's YCCK */ + break; + } + } else { + /* No special markers, assume straight CMYK. */ + cinfo->jpeg_color_space = JCS_CMYK; + } + cinfo->out_color_space = JCS_CMYK; + break; + + default: + cinfo->jpeg_color_space = JCS_UNKNOWN; + cinfo->out_color_space = JCS_UNKNOWN; + break; + } + + /* Set defaults for other decompression parameters. */ + cinfo->scale_num = 1; /* 1:1 scaling */ + cinfo->scale_denom = 1; + cinfo->output_gamma = 1.0; + cinfo->buffered_image = FALSE; + cinfo->raw_data_out = FALSE; + cinfo->dct_method = JDCT_DEFAULT; + cinfo->do_fancy_upsampling = TRUE; + cinfo->do_block_smoothing = TRUE; + cinfo->quantize_colors = FALSE; + /* We set these in case application only sets quantize_colors. */ + cinfo->dither_mode = JDITHER_FS; +#ifdef QUANT_2PASS_SUPPORTED + cinfo->two_pass_quantize = TRUE; +#else + cinfo->two_pass_quantize = FALSE; +#endif + cinfo->desired_number_of_colors = 256; + cinfo->colormap = NULL; + /* Initialize for no mode change in buffered-image mode. */ + cinfo->enable_1pass_quant = FALSE; + cinfo->enable_external_quant = FALSE; + cinfo->enable_2pass_quant = FALSE; +} + + +/* + * Decompression startup: read start of JPEG datastream to see what's there. + * Need only initialize JPEG object and supply a data source before calling. + * + * This routine will read as far as the first SOS marker (ie, actual start of + * compressed data), and will save all tables and parameters in the JPEG + * object. It will also initialize the decompression parameters to default + * values, and finally return JPEG_HEADER_OK. On return, the application may + * adjust the decompression parameters and then call jpeg_start_decompress. + * (Or, if the application only wanted to determine the image parameters, + * the data need not be decompressed. In that case, call jpeg_abort or + * jpeg_destroy to release any temporary space.) + * If an abbreviated (tables only) datastream is presented, the routine will + * return JPEG_HEADER_TABLES_ONLY upon reaching EOI. The application may then + * re-use the JPEG object to read the abbreviated image datastream(s). + * It is unnecessary (but OK) to call jpeg_abort in this case. + * The JPEG_SUSPENDED return code only occurs if the data source module + * requests suspension of the decompressor. In this case the application + * should load more source data and then re-call jpeg_read_header to resume + * processing. + * If a non-suspending data source is used and require_image is TRUE, then the + * return code need not be inspected since only JPEG_HEADER_OK is possible. + * + * This routine is now just a front end to jpeg_consume_input, with some + * extra error checking. + */ + +GLOBAL(int) +jpeg_read_header (j_decompress_ptr cinfo, boolean require_image) +{ + int retcode; + + if (cinfo->global_state != DSTATE_START && + cinfo->global_state != DSTATE_INHEADER) + ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state); + + retcode = jpeg_consume_input(cinfo); + + switch (retcode) { + case JPEG_REACHED_SOS: + retcode = JPEG_HEADER_OK; + break; + case JPEG_REACHED_EOI: + if (require_image) /* Complain if application wanted an image */ + ERREXIT(cinfo, JERR_NO_IMAGE); + /* Reset to start state; it would be safer to require the application to + * call jpeg_abort, but we can't change it now for compatibility reasons. + * A side effect is to free any temporary memory (there shouldn't be any). + */ + jpeg_abort((j_common_ptr) cinfo); /* sets state = DSTATE_START */ + retcode = JPEG_HEADER_TABLES_ONLY; + break; + case JPEG_SUSPENDED: + /* no work */ + break; + } + + return retcode; +} + + +/* + * Consume data in advance of what the decompressor requires. + * This can be called at any time once the decompressor object has + * been created and a data source has been set up. + * + * This routine is essentially a state machine that handles a couple + * of critical state-transition actions, namely initial setup and + * transition from header scanning to ready-for-start_decompress. + * All the actual input is done via the input controller's consume_input + * method. + */ + +GLOBAL(int) +jpeg_consume_input (j_decompress_ptr cinfo) +{ + int retcode = JPEG_SUSPENDED; + + /* NB: every possible DSTATE value should be listed in this switch */ + switch (cinfo->global_state) { + case DSTATE_START: + /* Start-of-datastream actions: reset appropriate modules */ + (*cinfo->inputctl->reset_input_controller) (cinfo); + /* Initialize application's data source module */ + (*cinfo->src->init_source) (cinfo); + cinfo->global_state = DSTATE_INHEADER; + /*FALLTHROUGH*/ + case DSTATE_INHEADER: + retcode = (*cinfo->inputctl->consume_input) (cinfo); + if (retcode == JPEG_REACHED_SOS) { /* Found SOS, prepare to decompress */ + /* Set up default parameters based on header data */ + default_decompress_parms(cinfo); + /* Set global state: ready for start_decompress */ + cinfo->global_state = DSTATE_READY; + } + break; + case DSTATE_READY: + /* Can't advance past first SOS until start_decompress is called */ + retcode = JPEG_REACHED_SOS; + break; + case DSTATE_PRELOAD: + case DSTATE_PRESCAN: + case DSTATE_SCANNING: + case DSTATE_RAW_OK: + case DSTATE_BUFIMAGE: + case DSTATE_BUFPOST: + case DSTATE_STOPPING: + retcode = (*cinfo->inputctl->consume_input) (cinfo); + break; + default: + ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state); + } + return retcode; +} + + +/* + * Have we finished reading the input file? + */ + +GLOBAL(boolean) +jpeg_input_complete (j_decompress_ptr cinfo) +{ + /* Check for valid jpeg object */ + if (cinfo->global_state < DSTATE_START || + cinfo->global_state > DSTATE_STOPPING) + ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state); + return cinfo->inputctl->eoi_reached; +} + + +/* + * Is there more than one scan? + */ + +GLOBAL(boolean) +jpeg_has_multiple_scans (j_decompress_ptr cinfo) +{ + /* Only valid after jpeg_read_header completes */ + if (cinfo->global_state < DSTATE_READY || + cinfo->global_state > DSTATE_STOPPING) + ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state); + return cinfo->inputctl->has_multiple_scans; +} + + +/* + * Finish JPEG decompression. + * + * This will normally just verify the file trailer and release temp storage. + * + * Returns FALSE if suspended. The return value need be inspected only if + * a suspending data source is used. + */ + +GLOBAL(boolean) +jpeg_finish_decompress (j_decompress_ptr cinfo) +{ + if ((cinfo->global_state == DSTATE_SCANNING || + cinfo->global_state == DSTATE_RAW_OK) && ! cinfo->buffered_image) { + /* Terminate final pass of non-buffered mode */ + if (cinfo->output_scanline < cinfo->output_height) + ERREXIT(cinfo, JERR_TOO_LITTLE_DATA); + (*cinfo->master->finish_output_pass) (cinfo); + cinfo->global_state = DSTATE_STOPPING; + } else if (cinfo->global_state == DSTATE_BUFIMAGE) { + /* Finishing after a buffered-image operation */ + cinfo->global_state = DSTATE_STOPPING; + } else if (cinfo->global_state != DSTATE_STOPPING) { + /* STOPPING = repeat call after a suspension, anything else is error */ + ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state); + } + /* Read until EOI */ + while (! cinfo->inputctl->eoi_reached) { + if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED) + return FALSE; /* Suspend, come back later */ + } + /* Do final cleanup */ + (*cinfo->src->term_source) (cinfo); + /* We can use jpeg_abort to release memory and reset global_state */ + jpeg_abort((j_common_ptr) cinfo); + return TRUE; +} diff --git a/app/src/main/jni/libjpeg-turbo/jdapistd.c b/app/src/main/jni/libjpeg-turbo/jdapistd.c new file mode 100644 index 000000000..3be527cb2 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdapistd.c @@ -0,0 +1,278 @@ +/* + * jdapistd.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1996, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright (C) 2010, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains application interface code for the decompression half + * of the JPEG library. These are the "standard" API routines that are + * used in the normal full-decompression case. They are not used by a + * transcoding-only application. Note that if an application links in + * jpeg_start_decompress, it will end up linking in the entire decompressor. + * We thus must separate this file from jdapimin.c to avoid linking the + * whole decompression library into a transcoder. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jpegcomp.h" + + +/* Forward declarations */ +LOCAL(boolean) output_pass_setup (j_decompress_ptr cinfo); + + +/* + * Decompression initialization. + * jpeg_read_header must be completed before calling this. + * + * If a multipass operating mode was selected, this will do all but the + * last pass, and thus may take a great deal of time. + * + * Returns FALSE if suspended. The return value need be inspected only if + * a suspending data source is used. + */ + +GLOBAL(boolean) +jpeg_start_decompress (j_decompress_ptr cinfo) +{ + if (cinfo->global_state == DSTATE_READY) { + /* First call: initialize master control, select active modules */ + jinit_master_decompress(cinfo); + if (cinfo->buffered_image) { + /* No more work here; expecting jpeg_start_output next */ + cinfo->global_state = DSTATE_BUFIMAGE; + return TRUE; + } + cinfo->global_state = DSTATE_PRELOAD; + } + if (cinfo->global_state == DSTATE_PRELOAD) { + /* If file has multiple scans, absorb them all into the coef buffer */ + if (cinfo->inputctl->has_multiple_scans) { +#ifdef D_MULTISCAN_FILES_SUPPORTED + for (;;) { + int retcode; + /* Call progress monitor hook if present */ + if (cinfo->progress != NULL) + (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo); + /* Absorb some more input */ + retcode = (*cinfo->inputctl->consume_input) (cinfo); + if (retcode == JPEG_SUSPENDED) + return FALSE; + if (retcode == JPEG_REACHED_EOI) + break; + /* Advance progress counter if appropriate */ + if (cinfo->progress != NULL && + (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) { + if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) { + /* jdmaster underestimated number of scans; ratchet up one scan */ + cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows; + } + } + } +#else + ERREXIT(cinfo, JERR_NOT_COMPILED); +#endif /* D_MULTISCAN_FILES_SUPPORTED */ + } + cinfo->output_scan_number = cinfo->input_scan_number; + } else if (cinfo->global_state != DSTATE_PRESCAN) + ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state); + /* Perform any dummy output passes, and set up for the final pass */ + return output_pass_setup(cinfo); +} + + +/* + * Set up for an output pass, and perform any dummy pass(es) needed. + * Common subroutine for jpeg_start_decompress and jpeg_start_output. + * Entry: global_state = DSTATE_PRESCAN only if previously suspended. + * Exit: If done, returns TRUE and sets global_state for proper output mode. + * If suspended, returns FALSE and sets global_state = DSTATE_PRESCAN. + */ + +LOCAL(boolean) +output_pass_setup (j_decompress_ptr cinfo) +{ + if (cinfo->global_state != DSTATE_PRESCAN) { + /* First call: do pass setup */ + (*cinfo->master->prepare_for_output_pass) (cinfo); + cinfo->output_scanline = 0; + cinfo->global_state = DSTATE_PRESCAN; + } + /* Loop over any required dummy passes */ + while (cinfo->master->is_dummy_pass) { +#ifdef QUANT_2PASS_SUPPORTED + /* Crank through the dummy pass */ + while (cinfo->output_scanline < cinfo->output_height) { + JDIMENSION last_scanline; + /* Call progress monitor hook if present */ + if (cinfo->progress != NULL) { + cinfo->progress->pass_counter = (long) cinfo->output_scanline; + cinfo->progress->pass_limit = (long) cinfo->output_height; + (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo); + } + /* Process some data */ + last_scanline = cinfo->output_scanline; + (*cinfo->main->process_data) (cinfo, (JSAMPARRAY) NULL, + &cinfo->output_scanline, (JDIMENSION) 0); + if (cinfo->output_scanline == last_scanline) + return FALSE; /* No progress made, must suspend */ + } + /* Finish up dummy pass, and set up for another one */ + (*cinfo->master->finish_output_pass) (cinfo); + (*cinfo->master->prepare_for_output_pass) (cinfo); + cinfo->output_scanline = 0; +#else + ERREXIT(cinfo, JERR_NOT_COMPILED); +#endif /* QUANT_2PASS_SUPPORTED */ + } + /* Ready for application to drive output pass through + * jpeg_read_scanlines or jpeg_read_raw_data. + */ + cinfo->global_state = cinfo->raw_data_out ? DSTATE_RAW_OK : DSTATE_SCANNING; + return TRUE; +} + + +/* + * Read some scanlines of data from the JPEG decompressor. + * + * The return value will be the number of lines actually read. + * This may be less than the number requested in several cases, + * including bottom of image, data source suspension, and operating + * modes that emit multiple scanlines at a time. + * + * Note: we warn about excess calls to jpeg_read_scanlines() since + * this likely signals an application programmer error. However, + * an oversize buffer (max_lines > scanlines remaining) is not an error. + */ + +GLOBAL(JDIMENSION) +jpeg_read_scanlines (j_decompress_ptr cinfo, JSAMPARRAY scanlines, + JDIMENSION max_lines) +{ + JDIMENSION row_ctr; + + if (cinfo->global_state != DSTATE_SCANNING) + ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state); + if (cinfo->output_scanline >= cinfo->output_height) { + WARNMS(cinfo, JWRN_TOO_MUCH_DATA); + return 0; + } + + /* Call progress monitor hook if present */ + if (cinfo->progress != NULL) { + cinfo->progress->pass_counter = (long) cinfo->output_scanline; + cinfo->progress->pass_limit = (long) cinfo->output_height; + (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo); + } + + /* Process some data */ + row_ctr = 0; + (*cinfo->main->process_data) (cinfo, scanlines, &row_ctr, max_lines); + cinfo->output_scanline += row_ctr; + return row_ctr; +} + + +/* + * Alternate entry point to read raw data. + * Processes exactly one iMCU row per call, unless suspended. + */ + +GLOBAL(JDIMENSION) +jpeg_read_raw_data (j_decompress_ptr cinfo, JSAMPIMAGE data, + JDIMENSION max_lines) +{ + JDIMENSION lines_per_iMCU_row; + + if (cinfo->global_state != DSTATE_RAW_OK) + ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state); + if (cinfo->output_scanline >= cinfo->output_height) { + WARNMS(cinfo, JWRN_TOO_MUCH_DATA); + return 0; + } + + /* Call progress monitor hook if present */ + if (cinfo->progress != NULL) { + cinfo->progress->pass_counter = (long) cinfo->output_scanline; + cinfo->progress->pass_limit = (long) cinfo->output_height; + (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo); + } + + /* Verify that at least one iMCU row can be returned. */ + lines_per_iMCU_row = cinfo->max_v_samp_factor * cinfo->_min_DCT_scaled_size; + if (max_lines < lines_per_iMCU_row) + ERREXIT(cinfo, JERR_BUFFER_SIZE); + + /* Decompress directly into user's buffer. */ + if (! (*cinfo->coef->decompress_data) (cinfo, data)) + return 0; /* suspension forced, can do nothing more */ + + /* OK, we processed one iMCU row. */ + cinfo->output_scanline += lines_per_iMCU_row; + return lines_per_iMCU_row; +} + + +/* Additional entry points for buffered-image mode. */ + +#ifdef D_MULTISCAN_FILES_SUPPORTED + +/* + * Initialize for an output pass in buffered-image mode. + */ + +GLOBAL(boolean) +jpeg_start_output (j_decompress_ptr cinfo, int scan_number) +{ + if (cinfo->global_state != DSTATE_BUFIMAGE && + cinfo->global_state != DSTATE_PRESCAN) + ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state); + /* Limit scan number to valid range */ + if (scan_number <= 0) + scan_number = 1; + if (cinfo->inputctl->eoi_reached && + scan_number > cinfo->input_scan_number) + scan_number = cinfo->input_scan_number; + cinfo->output_scan_number = scan_number; + /* Perform any dummy output passes, and set up for the real pass */ + return output_pass_setup(cinfo); +} + + +/* + * Finish up after an output pass in buffered-image mode. + * + * Returns FALSE if suspended. The return value need be inspected only if + * a suspending data source is used. + */ + +GLOBAL(boolean) +jpeg_finish_output (j_decompress_ptr cinfo) +{ + if ((cinfo->global_state == DSTATE_SCANNING || + cinfo->global_state == DSTATE_RAW_OK) && cinfo->buffered_image) { + /* Terminate this pass. */ + /* We do not require the whole pass to have been completed. */ + (*cinfo->master->finish_output_pass) (cinfo); + cinfo->global_state = DSTATE_BUFPOST; + } else if (cinfo->global_state != DSTATE_BUFPOST) { + /* BUFPOST = repeat call after a suspension, anything else is error */ + ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state); + } + /* Read markers looking for SOS or EOI */ + while (cinfo->input_scan_number <= cinfo->output_scan_number && + ! cinfo->inputctl->eoi_reached) { + if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED) + return FALSE; /* Suspend, come back later */ + } + cinfo->global_state = DSTATE_BUFIMAGE; + return TRUE; +} + +#endif /* D_MULTISCAN_FILES_SUPPORTED */ diff --git a/app/src/main/jni/libjpeg-turbo/jdarith.c b/app/src/main/jni/libjpeg-turbo/jdarith.c new file mode 100644 index 000000000..c6a1a99ac --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdarith.c @@ -0,0 +1,763 @@ +/* + * jdarith.c + * + * This file was part of the Independent JPEG Group's software: + * Developed 1997-2009 by Guido Vollbeding. + * It was modified by The libjpeg-turbo Project to include only code relevant + * to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains portable arithmetic entropy decoding routines for JPEG + * (implementing the ISO/IEC IS 10918-1 and CCITT Recommendation ITU-T T.81). + * + * Both sequential and progressive modes are supported in this single module. + * + * Suspension is not currently supported in this module. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" + + +/* Expanded entropy decoder object for arithmetic decoding. */ + +typedef struct { + struct jpeg_entropy_decoder pub; /* public fields */ + + INT32 c; /* C register, base of coding interval + input bit buffer */ + INT32 a; /* A register, normalized size of coding interval */ + int ct; /* bit shift counter, # of bits left in bit buffer part of C */ + /* init: ct = -16 */ + /* run: ct = 0..7 */ + /* error: ct = -1 */ + int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */ + int dc_context[MAX_COMPS_IN_SCAN]; /* context index for DC conditioning */ + + unsigned int restarts_to_go; /* MCUs left in this restart interval */ + + /* Pointers to statistics areas (these workspaces have image lifespan) */ + unsigned char * dc_stats[NUM_ARITH_TBLS]; + unsigned char * ac_stats[NUM_ARITH_TBLS]; + + /* Statistics bin for coding with fixed probability 0.5 */ + unsigned char fixed_bin[4]; +} arith_entropy_decoder; + +typedef arith_entropy_decoder * arith_entropy_ptr; + +/* The following two definitions specify the allocation chunk size + * for the statistics area. + * According to sections F.1.4.4.1.3 and F.1.4.4.2, we need at least + * 49 statistics bins for DC, and 245 statistics bins for AC coding. + * + * We use a compact representation with 1 byte per statistics bin, + * thus the numbers directly represent byte sizes. + * This 1 byte per statistics bin contains the meaning of the MPS + * (more probable symbol) in the highest bit (mask 0x80), and the + * index into the probability estimation state machine table + * in the lower bits (mask 0x7F). + */ + +#define DC_STAT_BINS 64 +#define AC_STAT_BINS 256 + + +LOCAL(int) +get_byte (j_decompress_ptr cinfo) +/* Read next input byte; we do not support suspension in this module. */ +{ + struct jpeg_source_mgr * src = cinfo->src; + + if (src->bytes_in_buffer == 0) + if (! (*src->fill_input_buffer) (cinfo)) + ERREXIT(cinfo, JERR_CANT_SUSPEND); + src->bytes_in_buffer--; + return GETJOCTET(*src->next_input_byte++); +} + + +/* + * The core arithmetic decoding routine (common in JPEG and JBIG). + * This needs to go as fast as possible. + * Machine-dependent optimization facilities + * are not utilized in this portable implementation. + * However, this code should be fairly efficient and + * may be a good base for further optimizations anyway. + * + * Return value is 0 or 1 (binary decision). + * + * Note: I've changed the handling of the code base & bit + * buffer register C compared to other implementations + * based on the standards layout & procedures. + * While it also contains both the actual base of the + * coding interval (16 bits) and the next-bits buffer, + * the cut-point between these two parts is floating + * (instead of fixed) with the bit shift counter CT. + * Thus, we also need only one (variable instead of + * fixed size) shift for the LPS/MPS decision, and + * we can get away with any renormalization update + * of C (except for new data insertion, of course). + * + * I've also introduced a new scheme for accessing + * the probability estimation state machine table, + * derived from Markus Kuhn's JBIG implementation. + */ + +LOCAL(int) +arith_decode (j_decompress_ptr cinfo, unsigned char *st) +{ + register arith_entropy_ptr e = (arith_entropy_ptr) cinfo->entropy; + register unsigned char nl, nm; + register INT32 qe, temp; + register int sv, data; + + /* Renormalization & data input per section D.2.6 */ + while (e->a < 0x8000L) { + if (--e->ct < 0) { + /* Need to fetch next data byte */ + if (cinfo->unread_marker) + data = 0; /* stuff zero data */ + else { + data = get_byte(cinfo); /* read next input byte */ + if (data == 0xFF) { /* zero stuff or marker code */ + do data = get_byte(cinfo); + while (data == 0xFF); /* swallow extra 0xFF bytes */ + if (data == 0) + data = 0xFF; /* discard stuffed zero byte */ + else { + /* Note: Different from the Huffman decoder, hitting + * a marker while processing the compressed data + * segment is legal in arithmetic coding. + * The convention is to supply zero data + * then until decoding is complete. + */ + cinfo->unread_marker = data; + data = 0; + } + } + } + e->c = (e->c << 8) | data; /* insert data into C register */ + if ((e->ct += 8) < 0) /* update bit shift counter */ + /* Need more initial bytes */ + if (++e->ct == 0) + /* Got 2 initial bytes -> re-init A and exit loop */ + e->a = 0x8000L; /* => e->a = 0x10000L after loop exit */ + } + e->a <<= 1; + } + + /* Fetch values from our compact representation of Table D.2: + * Qe values and probability estimation state machine + */ + sv = *st; + qe = jpeg_aritab[sv & 0x7F]; /* => Qe_Value */ + nl = qe & 0xFF; qe >>= 8; /* Next_Index_LPS + Switch_MPS */ + nm = qe & 0xFF; qe >>= 8; /* Next_Index_MPS */ + + /* Decode & estimation procedures per sections D.2.4 & D.2.5 */ + temp = e->a - qe; + e->a = temp; + temp <<= e->ct; + if (e->c >= temp) { + e->c -= temp; + /* Conditional LPS (less probable symbol) exchange */ + if (e->a < qe) { + e->a = qe; + *st = (sv & 0x80) ^ nm; /* Estimate_after_MPS */ + } else { + e->a = qe; + *st = (sv & 0x80) ^ nl; /* Estimate_after_LPS */ + sv ^= 0x80; /* Exchange LPS/MPS */ + } + } else if (e->a < 0x8000L) { + /* Conditional MPS (more probable symbol) exchange */ + if (e->a < qe) { + *st = (sv & 0x80) ^ nl; /* Estimate_after_LPS */ + sv ^= 0x80; /* Exchange LPS/MPS */ + } else { + *st = (sv & 0x80) ^ nm; /* Estimate_after_MPS */ + } + } + + return sv >> 7; +} + + +/* + * Check for a restart marker & resynchronize decoder. + */ + +LOCAL(void) +process_restart (j_decompress_ptr cinfo) +{ + arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy; + int ci; + jpeg_component_info * compptr; + + /* Advance past the RSTn marker */ + if (! (*cinfo->marker->read_restart_marker) (cinfo)) + ERREXIT(cinfo, JERR_CANT_SUSPEND); + + /* Re-initialize statistics areas */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + if (! cinfo->progressive_mode || (cinfo->Ss == 0 && cinfo->Ah == 0)) { + MEMZERO(entropy->dc_stats[compptr->dc_tbl_no], DC_STAT_BINS); + /* Reset DC predictions to 0 */ + entropy->last_dc_val[ci] = 0; + entropy->dc_context[ci] = 0; + } + if (! cinfo->progressive_mode || cinfo->Ss) { + MEMZERO(entropy->ac_stats[compptr->ac_tbl_no], AC_STAT_BINS); + } + } + + /* Reset arithmetic decoding variables */ + entropy->c = 0; + entropy->a = 0; + entropy->ct = -16; /* force reading 2 initial bytes to fill C */ + + /* Reset restart counter */ + entropy->restarts_to_go = cinfo->restart_interval; +} + + +/* + * Arithmetic MCU decoding. + * Each of these routines decodes and returns one MCU's worth of + * arithmetic-compressed coefficients. + * The coefficients are reordered from zigzag order into natural array order, + * but are not dequantized. + * + * The i'th block of the MCU is stored into the block pointed to by + * MCU_data[i]. WE ASSUME THIS AREA IS INITIALLY ZEROED BY THE CALLER. + */ + +/* + * MCU decoding for DC initial scan (either spectral selection, + * or first pass of successive approximation). + */ + +METHODDEF(boolean) +decode_mcu_DC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data) +{ + arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy; + JBLOCKROW block; + unsigned char *st; + int blkn, ci, tbl, sign; + int v, m; + + /* Process restart marker if needed */ + if (cinfo->restart_interval) { + if (entropy->restarts_to_go == 0) + process_restart(cinfo); + entropy->restarts_to_go--; + } + + if (entropy->ct == -1) return TRUE; /* if error do nothing */ + + /* Outer loop handles each block in the MCU */ + + for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) { + block = MCU_data[blkn]; + ci = cinfo->MCU_membership[blkn]; + tbl = cinfo->cur_comp_info[ci]->dc_tbl_no; + + /* Sections F.2.4.1 & F.1.4.4.1: Decoding of DC coefficients */ + + /* Table F.4: Point to statistics bin S0 for DC coefficient coding */ + st = entropy->dc_stats[tbl] + entropy->dc_context[ci]; + + /* Figure F.19: Decode_DC_DIFF */ + if (arith_decode(cinfo, st) == 0) + entropy->dc_context[ci] = 0; + else { + /* Figure F.21: Decoding nonzero value v */ + /* Figure F.22: Decoding the sign of v */ + sign = arith_decode(cinfo, st + 1); + st += 2; st += sign; + /* Figure F.23: Decoding the magnitude category of v */ + if ((m = arith_decode(cinfo, st)) != 0) { + st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */ + while (arith_decode(cinfo, st)) { + if ((m <<= 1) == 0x8000) { + WARNMS(cinfo, JWRN_ARITH_BAD_CODE); + entropy->ct = -1; /* magnitude overflow */ + return TRUE; + } + st += 1; + } + } + /* Section F.1.4.4.1.2: Establish dc_context conditioning category */ + if (m < (int) ((1L << cinfo->arith_dc_L[tbl]) >> 1)) + entropy->dc_context[ci] = 0; /* zero diff category */ + else if (m > (int) ((1L << cinfo->arith_dc_U[tbl]) >> 1)) + entropy->dc_context[ci] = 12 + (sign * 4); /* large diff category */ + else + entropy->dc_context[ci] = 4 + (sign * 4); /* small diff category */ + v = m; + /* Figure F.24: Decoding the magnitude bit pattern of v */ + st += 14; + while (m >>= 1) + if (arith_decode(cinfo, st)) v |= m; + v += 1; if (sign) v = -v; + entropy->last_dc_val[ci] += v; + } + + /* Scale and output the DC coefficient (assumes jpeg_natural_order[0]=0) */ + (*block)[0] = (JCOEF) (entropy->last_dc_val[ci] << cinfo->Al); + } + + return TRUE; +} + + +/* + * MCU decoding for AC initial scan (either spectral selection, + * or first pass of successive approximation). + */ + +METHODDEF(boolean) +decode_mcu_AC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data) +{ + arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy; + JBLOCKROW block; + unsigned char *st; + int tbl, sign, k; + int v, m; + + /* Process restart marker if needed */ + if (cinfo->restart_interval) { + if (entropy->restarts_to_go == 0) + process_restart(cinfo); + entropy->restarts_to_go--; + } + + if (entropy->ct == -1) return TRUE; /* if error do nothing */ + + /* There is always only one block per MCU */ + block = MCU_data[0]; + tbl = cinfo->cur_comp_info[0]->ac_tbl_no; + + /* Sections F.2.4.2 & F.1.4.4.2: Decoding of AC coefficients */ + + /* Figure F.20: Decode_AC_coefficients */ + for (k = cinfo->Ss; k <= cinfo->Se; k++) { + st = entropy->ac_stats[tbl] + 3 * (k - 1); + if (arith_decode(cinfo, st)) break; /* EOB flag */ + while (arith_decode(cinfo, st + 1) == 0) { + st += 3; k++; + if (k > cinfo->Se) { + WARNMS(cinfo, JWRN_ARITH_BAD_CODE); + entropy->ct = -1; /* spectral overflow */ + return TRUE; + } + } + /* Figure F.21: Decoding nonzero value v */ + /* Figure F.22: Decoding the sign of v */ + sign = arith_decode(cinfo, entropy->fixed_bin); + st += 2; + /* Figure F.23: Decoding the magnitude category of v */ + if ((m = arith_decode(cinfo, st)) != 0) { + if (arith_decode(cinfo, st)) { + m <<= 1; + st = entropy->ac_stats[tbl] + + (k <= cinfo->arith_ac_K[tbl] ? 189 : 217); + while (arith_decode(cinfo, st)) { + if ((m <<= 1) == 0x8000) { + WARNMS(cinfo, JWRN_ARITH_BAD_CODE); + entropy->ct = -1; /* magnitude overflow */ + return TRUE; + } + st += 1; + } + } + } + v = m; + /* Figure F.24: Decoding the magnitude bit pattern of v */ + st += 14; + while (m >>= 1) + if (arith_decode(cinfo, st)) v |= m; + v += 1; if (sign) v = -v; + /* Scale and output coefficient in natural (dezigzagged) order */ + (*block)[jpeg_natural_order[k]] = (JCOEF) (v << cinfo->Al); + } + + return TRUE; +} + + +/* + * MCU decoding for DC successive approximation refinement scan. + */ + +METHODDEF(boolean) +decode_mcu_DC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data) +{ + arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy; + unsigned char *st; + int p1, blkn; + + /* Process restart marker if needed */ + if (cinfo->restart_interval) { + if (entropy->restarts_to_go == 0) + process_restart(cinfo); + entropy->restarts_to_go--; + } + + st = entropy->fixed_bin; /* use fixed probability estimation */ + p1 = 1 << cinfo->Al; /* 1 in the bit position being coded */ + + /* Outer loop handles each block in the MCU */ + + for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) { + /* Encoded data is simply the next bit of the two's-complement DC value */ + if (arith_decode(cinfo, st)) + MCU_data[blkn][0][0] |= p1; + } + + return TRUE; +} + + +/* + * MCU decoding for AC successive approximation refinement scan. + */ + +METHODDEF(boolean) +decode_mcu_AC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data) +{ + arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy; + JBLOCKROW block; + JCOEFPTR thiscoef; + unsigned char *st; + int tbl, k, kex; + int p1, m1; + + /* Process restart marker if needed */ + if (cinfo->restart_interval) { + if (entropy->restarts_to_go == 0) + process_restart(cinfo); + entropy->restarts_to_go--; + } + + if (entropy->ct == -1) return TRUE; /* if error do nothing */ + + /* There is always only one block per MCU */ + block = MCU_data[0]; + tbl = cinfo->cur_comp_info[0]->ac_tbl_no; + + p1 = 1 << cinfo->Al; /* 1 in the bit position being coded */ + m1 = (-1) << cinfo->Al; /* -1 in the bit position being coded */ + + /* Establish EOBx (previous stage end-of-block) index */ + for (kex = cinfo->Se; kex > 0; kex--) + if ((*block)[jpeg_natural_order[kex]]) break; + + for (k = cinfo->Ss; k <= cinfo->Se; k++) { + st = entropy->ac_stats[tbl] + 3 * (k - 1); + if (k > kex) + if (arith_decode(cinfo, st)) break; /* EOB flag */ + for (;;) { + thiscoef = *block + jpeg_natural_order[k]; + if (*thiscoef) { /* previously nonzero coef */ + if (arith_decode(cinfo, st + 2)) { + if (*thiscoef < 0) + *thiscoef += m1; + else + *thiscoef += p1; + } + break; + } + if (arith_decode(cinfo, st + 1)) { /* newly nonzero coef */ + if (arith_decode(cinfo, entropy->fixed_bin)) + *thiscoef = m1; + else + *thiscoef = p1; + break; + } + st += 3; k++; + if (k > cinfo->Se) { + WARNMS(cinfo, JWRN_ARITH_BAD_CODE); + entropy->ct = -1; /* spectral overflow */ + return TRUE; + } + } + } + + return TRUE; +} + + +/* + * Decode one MCU's worth of arithmetic-compressed coefficients. + */ + +METHODDEF(boolean) +decode_mcu (j_decompress_ptr cinfo, JBLOCKROW *MCU_data) +{ + arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy; + jpeg_component_info * compptr; + JBLOCKROW block; + unsigned char *st; + int blkn, ci, tbl, sign, k; + int v, m; + + /* Process restart marker if needed */ + if (cinfo->restart_interval) { + if (entropy->restarts_to_go == 0) + process_restart(cinfo); + entropy->restarts_to_go--; + } + + if (entropy->ct == -1) return TRUE; /* if error do nothing */ + + /* Outer loop handles each block in the MCU */ + + for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) { + block = MCU_data[blkn]; + ci = cinfo->MCU_membership[blkn]; + compptr = cinfo->cur_comp_info[ci]; + + /* Sections F.2.4.1 & F.1.4.4.1: Decoding of DC coefficients */ + + tbl = compptr->dc_tbl_no; + + /* Table F.4: Point to statistics bin S0 for DC coefficient coding */ + st = entropy->dc_stats[tbl] + entropy->dc_context[ci]; + + /* Figure F.19: Decode_DC_DIFF */ + if (arith_decode(cinfo, st) == 0) + entropy->dc_context[ci] = 0; + else { + /* Figure F.21: Decoding nonzero value v */ + /* Figure F.22: Decoding the sign of v */ + sign = arith_decode(cinfo, st + 1); + st += 2; st += sign; + /* Figure F.23: Decoding the magnitude category of v */ + if ((m = arith_decode(cinfo, st)) != 0) { + st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */ + while (arith_decode(cinfo, st)) { + if ((m <<= 1) == 0x8000) { + WARNMS(cinfo, JWRN_ARITH_BAD_CODE); + entropy->ct = -1; /* magnitude overflow */ + return TRUE; + } + st += 1; + } + } + /* Section F.1.4.4.1.2: Establish dc_context conditioning category */ + if (m < (int) ((1L << cinfo->arith_dc_L[tbl]) >> 1)) + entropy->dc_context[ci] = 0; /* zero diff category */ + else if (m > (int) ((1L << cinfo->arith_dc_U[tbl]) >> 1)) + entropy->dc_context[ci] = 12 + (sign * 4); /* large diff category */ + else + entropy->dc_context[ci] = 4 + (sign * 4); /* small diff category */ + v = m; + /* Figure F.24: Decoding the magnitude bit pattern of v */ + st += 14; + while (m >>= 1) + if (arith_decode(cinfo, st)) v |= m; + v += 1; if (sign) v = -v; + entropy->last_dc_val[ci] += v; + } + + (*block)[0] = (JCOEF) entropy->last_dc_val[ci]; + + /* Sections F.2.4.2 & F.1.4.4.2: Decoding of AC coefficients */ + + tbl = compptr->ac_tbl_no; + + /* Figure F.20: Decode_AC_coefficients */ + for (k = 1; k <= DCTSIZE2 - 1; k++) { + st = entropy->ac_stats[tbl] + 3 * (k - 1); + if (arith_decode(cinfo, st)) break; /* EOB flag */ + while (arith_decode(cinfo, st + 1) == 0) { + st += 3; k++; + if (k > DCTSIZE2 - 1) { + WARNMS(cinfo, JWRN_ARITH_BAD_CODE); + entropy->ct = -1; /* spectral overflow */ + return TRUE; + } + } + /* Figure F.21: Decoding nonzero value v */ + /* Figure F.22: Decoding the sign of v */ + sign = arith_decode(cinfo, entropy->fixed_bin); + st += 2; + /* Figure F.23: Decoding the magnitude category of v */ + if ((m = arith_decode(cinfo, st)) != 0) { + if (arith_decode(cinfo, st)) { + m <<= 1; + st = entropy->ac_stats[tbl] + + (k <= cinfo->arith_ac_K[tbl] ? 189 : 217); + while (arith_decode(cinfo, st)) { + if ((m <<= 1) == 0x8000) { + WARNMS(cinfo, JWRN_ARITH_BAD_CODE); + entropy->ct = -1; /* magnitude overflow */ + return TRUE; + } + st += 1; + } + } + } + v = m; + /* Figure F.24: Decoding the magnitude bit pattern of v */ + st += 14; + while (m >>= 1) + if (arith_decode(cinfo, st)) v |= m; + v += 1; if (sign) v = -v; + (*block)[jpeg_natural_order[k]] = (JCOEF) v; + } + } + + return TRUE; +} + + +/* + * Initialize for an arithmetic-compressed scan. + */ + +METHODDEF(void) +start_pass (j_decompress_ptr cinfo) +{ + arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy; + int ci, tbl; + jpeg_component_info * compptr; + + if (cinfo->progressive_mode) { + /* Validate progressive scan parameters */ + if (cinfo->Ss == 0) { + if (cinfo->Se != 0) + goto bad; + } else { + /* need not check Ss/Se < 0 since they came from unsigned bytes */ + if (cinfo->Se < cinfo->Ss || cinfo->Se > DCTSIZE2 - 1) + goto bad; + /* AC scans may have only one component */ + if (cinfo->comps_in_scan != 1) + goto bad; + } + if (cinfo->Ah != 0) { + /* Successive approximation refinement scan: must have Al = Ah-1. */ + if (cinfo->Ah-1 != cinfo->Al) + goto bad; + } + if (cinfo->Al > 13) { /* need not check for < 0 */ + bad: + ERREXIT4(cinfo, JERR_BAD_PROGRESSION, + cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al); + } + /* Update progression status, and verify that scan order is legal. + * Note that inter-scan inconsistencies are treated as warnings + * not fatal errors ... not clear if this is right way to behave. + */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + int coefi, cindex = cinfo->cur_comp_info[ci]->component_index; + int *coef_bit_ptr = & cinfo->coef_bits[cindex][0]; + if (cinfo->Ss && coef_bit_ptr[0] < 0) /* AC without prior DC scan */ + WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0); + for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) { + int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi]; + if (cinfo->Ah != expected) + WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi); + coef_bit_ptr[coefi] = cinfo->Al; + } + } + /* Select MCU decoding routine */ + if (cinfo->Ah == 0) { + if (cinfo->Ss == 0) + entropy->pub.decode_mcu = decode_mcu_DC_first; + else + entropy->pub.decode_mcu = decode_mcu_AC_first; + } else { + if (cinfo->Ss == 0) + entropy->pub.decode_mcu = decode_mcu_DC_refine; + else + entropy->pub.decode_mcu = decode_mcu_AC_refine; + } + } else { + /* Check that the scan parameters Ss, Se, Ah/Al are OK for sequential JPEG. + * This ought to be an error condition, but we make it a warning. + */ + if (cinfo->Ss != 0 || cinfo->Ah != 0 || cinfo->Al != 0 || + (cinfo->Se < DCTSIZE2 && cinfo->Se != DCTSIZE2 - 1)) + WARNMS(cinfo, JWRN_NOT_SEQUENTIAL); + /* Select MCU decoding routine */ + entropy->pub.decode_mcu = decode_mcu; + } + + /* Allocate & initialize requested statistics areas */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + if (! cinfo->progressive_mode || (cinfo->Ss == 0 && cinfo->Ah == 0)) { + tbl = compptr->dc_tbl_no; + if (tbl < 0 || tbl >= NUM_ARITH_TBLS) + ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl); + if (entropy->dc_stats[tbl] == NULL) + entropy->dc_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small) + ((j_common_ptr) cinfo, JPOOL_IMAGE, DC_STAT_BINS); + MEMZERO(entropy->dc_stats[tbl], DC_STAT_BINS); + /* Initialize DC predictions to 0 */ + entropy->last_dc_val[ci] = 0; + entropy->dc_context[ci] = 0; + } + if (! cinfo->progressive_mode || cinfo->Ss) { + tbl = compptr->ac_tbl_no; + if (tbl < 0 || tbl >= NUM_ARITH_TBLS) + ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl); + if (entropy->ac_stats[tbl] == NULL) + entropy->ac_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small) + ((j_common_ptr) cinfo, JPOOL_IMAGE, AC_STAT_BINS); + MEMZERO(entropy->ac_stats[tbl], AC_STAT_BINS); + } + } + + /* Initialize arithmetic decoding variables */ + entropy->c = 0; + entropy->a = 0; + entropy->ct = -16; /* force reading 2 initial bytes to fill C */ + + /* Initialize restart counter */ + entropy->restarts_to_go = cinfo->restart_interval; +} + + +/* + * Module initialization routine for arithmetic entropy decoding. + */ + +GLOBAL(void) +jinit_arith_decoder (j_decompress_ptr cinfo) +{ + arith_entropy_ptr entropy; + int i; + + entropy = (arith_entropy_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(arith_entropy_decoder)); + cinfo->entropy = (struct jpeg_entropy_decoder *) entropy; + entropy->pub.start_pass = start_pass; + + /* Mark tables unallocated */ + for (i = 0; i < NUM_ARITH_TBLS; i++) { + entropy->dc_stats[i] = NULL; + entropy->ac_stats[i] = NULL; + } + + /* Initialize index for fixed probability estimation */ + entropy->fixed_bin[0] = 113; + + if (cinfo->progressive_mode) { + /* Create progression status table */ + int *coef_bit_ptr, ci; + cinfo->coef_bits = (int (*)[DCTSIZE2]) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + cinfo->num_components*DCTSIZE2*sizeof(int)); + coef_bit_ptr = & cinfo->coef_bits[0][0]; + for (ci = 0; ci < cinfo->num_components; ci++) + for (i = 0; i < DCTSIZE2; i++) + *coef_bit_ptr++ = -1; + } +} diff --git a/app/src/main/jni/libjpeg-turbo/jdatadst-tj.c b/app/src/main/jni/libjpeg-turbo/jdatadst-tj.c new file mode 100644 index 000000000..1f6f3a5e7 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdatadst-tj.c @@ -0,0 +1,196 @@ +/* + * jdatadst-tj.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1996, Thomas G. Lane. + * Modified 2009-2012 by Guido Vollbeding. + * libjpeg-turbo Modifications: + * Copyright (C) 2011, 2014 D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains compression data destination routines for the case of + * emitting JPEG data to memory or to a file (or any stdio stream). + * While these routines are sufficient for most applications, + * some will want to use a different destination manager. + * IMPORTANT: we assume that fwrite() will correctly transcribe an array of + * JOCTETs into 8-bit-wide elements on external storage. If char is wider + * than 8 bits on your machine, you may need to do some tweaking. + */ + +/* this is not a core library module, so it doesn't define JPEG_INTERNALS */ +#include "jinclude.h" +#include "jpeglib.h" +#include "jerror.h" + +#ifndef HAVE_STDLIB_H /* should declare malloc(),free() */ +extern void * malloc (size_t size); +extern void free (void *ptr); +#endif + + +#define OUTPUT_BUF_SIZE 4096 /* choose an efficiently fwrite'able size */ + + +/* Expanded data destination object for memory output */ + +typedef struct { + struct jpeg_destination_mgr pub; /* public fields */ + + unsigned char ** outbuffer; /* target buffer */ + unsigned long * outsize; + unsigned char * newbuffer; /* newly allocated buffer */ + JOCTET * buffer; /* start of buffer */ + size_t bufsize; + boolean alloc; +} my_mem_destination_mgr; + +typedef my_mem_destination_mgr * my_mem_dest_ptr; + + +/* + * Initialize destination --- called by jpeg_start_compress + * before any data is actually written. + */ + +METHODDEF(void) +init_mem_destination (j_compress_ptr cinfo) +{ + /* no work necessary here */ +} + + +/* + * Empty the output buffer --- called whenever buffer fills up. + * + * In typical applications, this should write the entire output buffer + * (ignoring the current state of next_output_byte & free_in_buffer), + * reset the pointer & count to the start of the buffer, and return TRUE + * indicating that the buffer has been dumped. + * + * In applications that need to be able to suspend compression due to output + * overrun, a FALSE return indicates that the buffer cannot be emptied now. + * In this situation, the compressor will return to its caller (possibly with + * an indication that it has not accepted all the supplied scanlines). The + * application should resume compression after it has made more room in the + * output buffer. Note that there are substantial restrictions on the use of + * suspension --- see the documentation. + * + * When suspending, the compressor will back up to a convenient restart point + * (typically the start of the current MCU). next_output_byte & free_in_buffer + * indicate where the restart point will be if the current call returns FALSE. + * Data beyond this point will be regenerated after resumption, so do not + * write it out when emptying the buffer externally. + */ + +METHODDEF(boolean) +empty_mem_output_buffer (j_compress_ptr cinfo) +{ + size_t nextsize; + JOCTET * nextbuffer; + my_mem_dest_ptr dest = (my_mem_dest_ptr) cinfo->dest; + + if (!dest->alloc) ERREXIT(cinfo, JERR_BUFFER_SIZE); + + /* Try to allocate new buffer with double size */ + nextsize = dest->bufsize * 2; + nextbuffer = (JOCTET *) malloc(nextsize); + + if (nextbuffer == NULL) + ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10); + + MEMCOPY(nextbuffer, dest->buffer, dest->bufsize); + + if (dest->newbuffer != NULL) + free(dest->newbuffer); + + dest->newbuffer = nextbuffer; + + dest->pub.next_output_byte = nextbuffer + dest->bufsize; + dest->pub.free_in_buffer = dest->bufsize; + + dest->buffer = nextbuffer; + dest->bufsize = nextsize; + + return TRUE; +} + + +/* + * Terminate destination --- called by jpeg_finish_compress + * after all data has been written. Usually needs to flush buffer. + * + * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding + * application must deal with any cleanup that should happen even + * for error exit. + */ + +METHODDEF(void) +term_mem_destination (j_compress_ptr cinfo) +{ + my_mem_dest_ptr dest = (my_mem_dest_ptr) cinfo->dest; + + if(dest->alloc) *dest->outbuffer = dest->buffer; + *dest->outsize = (unsigned long)(dest->bufsize - dest->pub.free_in_buffer); +} + + +/* + * Prepare for output to a memory buffer. + * The caller may supply an own initial buffer with appropriate size. + * Otherwise, or when the actual data output exceeds the given size, + * the library adapts the buffer size as necessary. + * The standard library functions malloc/free are used for allocating + * larger memory, so the buffer is available to the application after + * finishing compression, and then the application is responsible for + * freeing the requested memory. + */ + +GLOBAL(void) +jpeg_mem_dest_tj (j_compress_ptr cinfo, + unsigned char ** outbuffer, unsigned long * outsize, + boolean alloc) +{ + boolean reused = FALSE; + my_mem_dest_ptr dest; + + if (outbuffer == NULL || outsize == NULL) /* sanity check */ + ERREXIT(cinfo, JERR_BUFFER_SIZE); + + /* The destination object is made permanent so that multiple JPEG images + * can be written to the same buffer without re-executing jpeg_mem_dest. + */ + if (cinfo->dest == NULL) { /* first time for this JPEG object? */ + cinfo->dest = (struct jpeg_destination_mgr *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT, + sizeof(my_mem_destination_mgr)); + dest = (my_mem_dest_ptr) cinfo->dest; + dest->newbuffer = NULL; + dest->buffer = NULL; + } + + dest = (my_mem_dest_ptr) cinfo->dest; + dest->pub.init_destination = init_mem_destination; + dest->pub.empty_output_buffer = empty_mem_output_buffer; + dest->pub.term_destination = term_mem_destination; + if (dest->buffer == *outbuffer && *outbuffer != NULL && alloc) + reused = TRUE; + dest->outbuffer = outbuffer; + dest->outsize = outsize; + dest->alloc = alloc; + + if (*outbuffer == NULL || *outsize == 0) { + if (alloc) { + /* Allocate initial buffer */ + dest->newbuffer = *outbuffer = (unsigned char *) malloc(OUTPUT_BUF_SIZE); + if (dest->newbuffer == NULL) + ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10); + *outsize = OUTPUT_BUF_SIZE; + } + else ERREXIT(cinfo, JERR_BUFFER_SIZE); + } + + dest->pub.next_output_byte = dest->buffer = *outbuffer; + if (!reused) + dest->bufsize = *outsize; + dest->pub.free_in_buffer = dest->bufsize; +} diff --git a/app/src/main/jni/libjpeg-turbo/jdatadst.c b/app/src/main/jni/libjpeg-turbo/jdatadst.c new file mode 100644 index 000000000..7a40e4f37 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdatadst.c @@ -0,0 +1,279 @@ +/* + * jdatadst.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1996, Thomas G. Lane. + * Modified 2009-2012 by Guido Vollbeding. + * libjpeg-turbo Modifications: + * Copyright (C) 2013, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains compression data destination routines for the case of + * emitting JPEG data to memory or to a file (or any stdio stream). + * While these routines are sufficient for most applications, + * some will want to use a different destination manager. + * IMPORTANT: we assume that fwrite() will correctly transcribe an array of + * JOCTETs into 8-bit-wide elements on external storage. If char is wider + * than 8 bits on your machine, you may need to do some tweaking. + */ + +/* this is not a core library module, so it doesn't define JPEG_INTERNALS */ +#include "jinclude.h" +#include "jpeglib.h" +#include "jerror.h" + +#ifndef HAVE_STDLIB_H /* should declare malloc(),free() */ +extern void * malloc (size_t size); +extern void free (void *ptr); +#endif + + +/* Expanded data destination object for stdio output */ + +typedef struct { + struct jpeg_destination_mgr pub; /* public fields */ + + FILE * outfile; /* target stream */ + JOCTET * buffer; /* start of buffer */ +} my_destination_mgr; + +typedef my_destination_mgr * my_dest_ptr; + +#define OUTPUT_BUF_SIZE 4096 /* choose an efficiently fwrite'able size */ + + +#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED) +/* Expanded data destination object for memory output */ + +typedef struct { + struct jpeg_destination_mgr pub; /* public fields */ + + unsigned char ** outbuffer; /* target buffer */ + unsigned long * outsize; + unsigned char * newbuffer; /* newly allocated buffer */ + JOCTET * buffer; /* start of buffer */ + size_t bufsize; +} my_mem_destination_mgr; + +typedef my_mem_destination_mgr * my_mem_dest_ptr; +#endif + + +/* + * Initialize destination --- called by jpeg_start_compress + * before any data is actually written. + */ + +METHODDEF(void) +init_destination (j_compress_ptr cinfo) +{ + my_dest_ptr dest = (my_dest_ptr) cinfo->dest; + + /* Allocate the output buffer --- it will be released when done with image */ + dest->buffer = (JOCTET *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + OUTPUT_BUF_SIZE * sizeof(JOCTET)); + + dest->pub.next_output_byte = dest->buffer; + dest->pub.free_in_buffer = OUTPUT_BUF_SIZE; +} + +#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED) +METHODDEF(void) +init_mem_destination (j_compress_ptr cinfo) +{ + /* no work necessary here */ +} +#endif + + +/* + * Empty the output buffer --- called whenever buffer fills up. + * + * In typical applications, this should write the entire output buffer + * (ignoring the current state of next_output_byte & free_in_buffer), + * reset the pointer & count to the start of the buffer, and return TRUE + * indicating that the buffer has been dumped. + * + * In applications that need to be able to suspend compression due to output + * overrun, a FALSE return indicates that the buffer cannot be emptied now. + * In this situation, the compressor will return to its caller (possibly with + * an indication that it has not accepted all the supplied scanlines). The + * application should resume compression after it has made more room in the + * output buffer. Note that there are substantial restrictions on the use of + * suspension --- see the documentation. + * + * When suspending, the compressor will back up to a convenient restart point + * (typically the start of the current MCU). next_output_byte & free_in_buffer + * indicate where the restart point will be if the current call returns FALSE. + * Data beyond this point will be regenerated after resumption, so do not + * write it out when emptying the buffer externally. + */ + +METHODDEF(boolean) +empty_output_buffer (j_compress_ptr cinfo) +{ + my_dest_ptr dest = (my_dest_ptr) cinfo->dest; + + if (JFWRITE(dest->outfile, dest->buffer, OUTPUT_BUF_SIZE) != + (size_t) OUTPUT_BUF_SIZE) + ERREXIT(cinfo, JERR_FILE_WRITE); + + dest->pub.next_output_byte = dest->buffer; + dest->pub.free_in_buffer = OUTPUT_BUF_SIZE; + + return TRUE; +} + +#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED) +METHODDEF(boolean) +empty_mem_output_buffer (j_compress_ptr cinfo) +{ + size_t nextsize; + JOCTET * nextbuffer; + my_mem_dest_ptr dest = (my_mem_dest_ptr) cinfo->dest; + + /* Try to allocate new buffer with double size */ + nextsize = dest->bufsize * 2; + nextbuffer = (JOCTET *) malloc(nextsize); + + if (nextbuffer == NULL) + ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10); + + MEMCOPY(nextbuffer, dest->buffer, dest->bufsize); + + if (dest->newbuffer != NULL) + free(dest->newbuffer); + + dest->newbuffer = nextbuffer; + + dest->pub.next_output_byte = nextbuffer + dest->bufsize; + dest->pub.free_in_buffer = dest->bufsize; + + dest->buffer = nextbuffer; + dest->bufsize = nextsize; + + return TRUE; +} +#endif + + +/* + * Terminate destination --- called by jpeg_finish_compress + * after all data has been written. Usually needs to flush buffer. + * + * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding + * application must deal with any cleanup that should happen even + * for error exit. + */ + +METHODDEF(void) +term_destination (j_compress_ptr cinfo) +{ + my_dest_ptr dest = (my_dest_ptr) cinfo->dest; + size_t datacount = OUTPUT_BUF_SIZE - dest->pub.free_in_buffer; + + /* Write any data remaining in the buffer */ + if (datacount > 0) { + if (JFWRITE(dest->outfile, dest->buffer, datacount) != datacount) + ERREXIT(cinfo, JERR_FILE_WRITE); + } + fflush(dest->outfile); + /* Make sure we wrote the output file OK */ + if (ferror(dest->outfile)) + ERREXIT(cinfo, JERR_FILE_WRITE); +} + +#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED) +METHODDEF(void) +term_mem_destination (j_compress_ptr cinfo) +{ + my_mem_dest_ptr dest = (my_mem_dest_ptr) cinfo->dest; + + *dest->outbuffer = dest->buffer; + *dest->outsize = (unsigned long)(dest->bufsize - dest->pub.free_in_buffer); +} +#endif + + +/* + * Prepare for output to a stdio stream. + * The caller must have already opened the stream, and is responsible + * for closing it after finishing compression. + */ + +GLOBAL(void) +jpeg_stdio_dest (j_compress_ptr cinfo, FILE * outfile) +{ + my_dest_ptr dest; + + /* The destination object is made permanent so that multiple JPEG images + * can be written to the same file without re-executing jpeg_stdio_dest. + * This makes it dangerous to use this manager and a different destination + * manager serially with the same JPEG object, because their private object + * sizes may be different. Caveat programmer. + */ + if (cinfo->dest == NULL) { /* first time for this JPEG object? */ + cinfo->dest = (struct jpeg_destination_mgr *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT, + sizeof(my_destination_mgr)); + } + + dest = (my_dest_ptr) cinfo->dest; + dest->pub.init_destination = init_destination; + dest->pub.empty_output_buffer = empty_output_buffer; + dest->pub.term_destination = term_destination; + dest->outfile = outfile; +} + + +#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED) +/* + * Prepare for output to a memory buffer. + * The caller may supply an own initial buffer with appropriate size. + * Otherwise, or when the actual data output exceeds the given size, + * the library adapts the buffer size as necessary. + * The standard library functions malloc/free are used for allocating + * larger memory, so the buffer is available to the application after + * finishing compression, and then the application is responsible for + * freeing the requested memory. + */ + +GLOBAL(void) +jpeg_mem_dest (j_compress_ptr cinfo, + unsigned char ** outbuffer, unsigned long * outsize) +{ + my_mem_dest_ptr dest; + + if (outbuffer == NULL || outsize == NULL) /* sanity check */ + ERREXIT(cinfo, JERR_BUFFER_SIZE); + + /* The destination object is made permanent so that multiple JPEG images + * can be written to the same buffer without re-executing jpeg_mem_dest. + */ + if (cinfo->dest == NULL) { /* first time for this JPEG object? */ + cinfo->dest = (struct jpeg_destination_mgr *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT, + sizeof(my_mem_destination_mgr)); + } + + dest = (my_mem_dest_ptr) cinfo->dest; + dest->pub.init_destination = init_mem_destination; + dest->pub.empty_output_buffer = empty_mem_output_buffer; + dest->pub.term_destination = term_mem_destination; + dest->outbuffer = outbuffer; + dest->outsize = outsize; + dest->newbuffer = NULL; + + if (*outbuffer == NULL || *outsize == 0) { + /* Allocate initial buffer */ + dest->newbuffer = *outbuffer = (unsigned char *) malloc(OUTPUT_BUF_SIZE); + if (dest->newbuffer == NULL) + ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10); + *outsize = OUTPUT_BUF_SIZE; + } + + dest->pub.next_output_byte = dest->buffer = *outbuffer; + dest->pub.free_in_buffer = dest->bufsize = *outsize; +} +#endif diff --git a/app/src/main/jni/libjpeg-turbo/jdatasrc-tj.c b/app/src/main/jni/libjpeg-turbo/jdatasrc-tj.c new file mode 100644 index 000000000..09f959602 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdatasrc-tj.c @@ -0,0 +1,185 @@ +/* + * jdatasrc-tj.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1996, Thomas G. Lane. + * Modified 2009-2011 by Guido Vollbeding. + * libjpeg-turbo Modifications: + * Copyright (C) 2011, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains decompression data source routines for the case of + * reading JPEG data from memory or from a file (or any stdio stream). + * While these routines are sufficient for most applications, + * some will want to use a different source manager. + * IMPORTANT: we assume that fread() will correctly transcribe an array of + * JOCTETs from 8-bit-wide elements on external storage. If char is wider + * than 8 bits on your machine, you may need to do some tweaking. + */ + +/* this is not a core library module, so it doesn't define JPEG_INTERNALS */ +#include "jinclude.h" +#include "jpeglib.h" +#include "jerror.h" + + +/* + * Initialize source --- called by jpeg_read_header + * before any data is actually read. + */ + +METHODDEF(void) +init_mem_source (j_decompress_ptr cinfo) +{ + /* no work necessary here */ +} + + +/* + * Fill the input buffer --- called whenever buffer is emptied. + * + * In typical applications, this should read fresh data into the buffer + * (ignoring the current state of next_input_byte & bytes_in_buffer), + * reset the pointer & count to the start of the buffer, and return TRUE + * indicating that the buffer has been reloaded. It is not necessary to + * fill the buffer entirely, only to obtain at least one more byte. + * + * There is no such thing as an EOF return. If the end of the file has been + * reached, the routine has a choice of ERREXIT() or inserting fake data into + * the buffer. In most cases, generating a warning message and inserting a + * fake EOI marker is the best course of action --- this will allow the + * decompressor to output however much of the image is there. However, + * the resulting error message is misleading if the real problem is an empty + * input file, so we handle that case specially. + * + * In applications that need to be able to suspend compression due to input + * not being available yet, a FALSE return indicates that no more data can be + * obtained right now, but more may be forthcoming later. In this situation, + * the decompressor will return to its caller (with an indication of the + * number of scanlines it has read, if any). The application should resume + * decompression after it has loaded more data into the input buffer. Note + * that there are substantial restrictions on the use of suspension --- see + * the documentation. + * + * When suspending, the decompressor will back up to a convenient restart point + * (typically the start of the current MCU). next_input_byte & bytes_in_buffer + * indicate where the restart point will be if the current call returns FALSE. + * Data beyond this point must be rescanned after resumption, so move it to + * the front of the buffer rather than discarding it. + */ + +METHODDEF(boolean) +fill_mem_input_buffer (j_decompress_ptr cinfo) +{ + static const JOCTET mybuffer[4] = { + (JOCTET) 0xFF, (JOCTET) JPEG_EOI, 0, 0 + }; + + /* The whole JPEG data is expected to reside in the supplied memory + * buffer, so any request for more data beyond the given buffer size + * is treated as an error. + */ + WARNMS(cinfo, JWRN_JPEG_EOF); + + /* Insert a fake EOI marker */ + + cinfo->src->next_input_byte = mybuffer; + cinfo->src->bytes_in_buffer = 2; + + return TRUE; +} + + +/* + * Skip data --- used to skip over a potentially large amount of + * uninteresting data (such as an APPn marker). + * + * Writers of suspendable-input applications must note that skip_input_data + * is not granted the right to give a suspension return. If the skip extends + * beyond the data currently in the buffer, the buffer can be marked empty so + * that the next read will cause a fill_input_buffer call that can suspend. + * Arranging for additional bytes to be discarded before reloading the input + * buffer is the application writer's problem. + */ + +METHODDEF(void) +skip_input_data (j_decompress_ptr cinfo, long num_bytes) +{ + struct jpeg_source_mgr * src = cinfo->src; + + /* Just a dumb implementation for now. Could use fseek() except + * it doesn't work on pipes. Not clear that being smart is worth + * any trouble anyway --- large skips are infrequent. + */ + if (num_bytes > 0) { + while (num_bytes > (long) src->bytes_in_buffer) { + num_bytes -= (long) src->bytes_in_buffer; + (void) (*src->fill_input_buffer) (cinfo); + /* note we assume that fill_input_buffer will never return FALSE, + * so suspension need not be handled. + */ + } + src->next_input_byte += (size_t) num_bytes; + src->bytes_in_buffer -= (size_t) num_bytes; + } +} + + +/* + * An additional method that can be provided by data source modules is the + * resync_to_restart method for error recovery in the presence of RST markers. + * For the moment, this source module just uses the default resync method + * provided by the JPEG library. That method assumes that no backtracking + * is possible. + */ + + +/* + * Terminate source --- called by jpeg_finish_decompress + * after all data has been read. Often a no-op. + * + * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding + * application must deal with any cleanup that should happen even + * for error exit. + */ + +METHODDEF(void) +term_source (j_decompress_ptr cinfo) +{ + /* no work necessary here */ +} + + +/* + * Prepare for input from a supplied memory buffer. + * The buffer must contain the whole JPEG data. + */ + +GLOBAL(void) +jpeg_mem_src_tj (j_decompress_ptr cinfo, + unsigned char * inbuffer, unsigned long insize) +{ + struct jpeg_source_mgr * src; + + if (inbuffer == NULL || insize == 0) /* Treat empty input as fatal error */ + ERREXIT(cinfo, JERR_INPUT_EMPTY); + + /* The source object is made permanent so that a series of JPEG images + * can be read from the same buffer by calling jpeg_mem_src only before + * the first one. + */ + if (cinfo->src == NULL) { /* first time for this JPEG object? */ + cinfo->src = (struct jpeg_source_mgr *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT, + sizeof(struct jpeg_source_mgr)); + } + + src = cinfo->src; + src->init_source = init_mem_source; + src->fill_input_buffer = fill_mem_input_buffer; + src->skip_input_data = skip_input_data; + src->resync_to_restart = jpeg_resync_to_restart; /* use default method */ + src->term_source = term_source; + src->bytes_in_buffer = (size_t) insize; + src->next_input_byte = (JOCTET *) inbuffer; +} diff --git a/app/src/main/jni/libjpeg-turbo/jdatasrc.c b/app/src/main/jni/libjpeg-turbo/jdatasrc.c new file mode 100644 index 000000000..bf70422ba --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdatasrc.c @@ -0,0 +1,283 @@ +/* + * jdatasrc.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1996, Thomas G. Lane. + * Modified 2009-2011 by Guido Vollbeding. + * libjpeg-turbo Modifications: + * Copyright (C) 2013, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains decompression data source routines for the case of + * reading JPEG data from memory or from a file (or any stdio stream). + * While these routines are sufficient for most applications, + * some will want to use a different source manager. + * IMPORTANT: we assume that fread() will correctly transcribe an array of + * JOCTETs from 8-bit-wide elements on external storage. If char is wider + * than 8 bits on your machine, you may need to do some tweaking. + */ + +/* this is not a core library module, so it doesn't define JPEG_INTERNALS */ +#include "jinclude.h" +#include "jpeglib.h" +#include "jerror.h" + + +/* Expanded data source object for stdio input */ + +typedef struct { + struct jpeg_source_mgr pub; /* public fields */ + + FILE * infile; /* source stream */ + JOCTET * buffer; /* start of buffer */ + boolean start_of_file; /* have we gotten any data yet? */ +} my_source_mgr; + +typedef my_source_mgr * my_src_ptr; + +#define INPUT_BUF_SIZE 4096 /* choose an efficiently fread'able size */ + + +/* + * Initialize source --- called by jpeg_read_header + * before any data is actually read. + */ + +METHODDEF(void) +init_source (j_decompress_ptr cinfo) +{ + my_src_ptr src = (my_src_ptr) cinfo->src; + + /* We reset the empty-input-file flag for each image, + * but we don't clear the input buffer. + * This is correct behavior for reading a series of images from one source. + */ + src->start_of_file = TRUE; +} + +#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED) +METHODDEF(void) +init_mem_source (j_decompress_ptr cinfo) +{ + /* no work necessary here */ +} +#endif + + +/* + * Fill the input buffer --- called whenever buffer is emptied. + * + * In typical applications, this should read fresh data into the buffer + * (ignoring the current state of next_input_byte & bytes_in_buffer), + * reset the pointer & count to the start of the buffer, and return TRUE + * indicating that the buffer has been reloaded. It is not necessary to + * fill the buffer entirely, only to obtain at least one more byte. + * + * There is no such thing as an EOF return. If the end of the file has been + * reached, the routine has a choice of ERREXIT() or inserting fake data into + * the buffer. In most cases, generating a warning message and inserting a + * fake EOI marker is the best course of action --- this will allow the + * decompressor to output however much of the image is there. However, + * the resulting error message is misleading if the real problem is an empty + * input file, so we handle that case specially. + * + * In applications that need to be able to suspend compression due to input + * not being available yet, a FALSE return indicates that no more data can be + * obtained right now, but more may be forthcoming later. In this situation, + * the decompressor will return to its caller (with an indication of the + * number of scanlines it has read, if any). The application should resume + * decompression after it has loaded more data into the input buffer. Note + * that there are substantial restrictions on the use of suspension --- see + * the documentation. + * + * When suspending, the decompressor will back up to a convenient restart point + * (typically the start of the current MCU). next_input_byte & bytes_in_buffer + * indicate where the restart point will be if the current call returns FALSE. + * Data beyond this point must be rescanned after resumption, so move it to + * the front of the buffer rather than discarding it. + */ + +METHODDEF(boolean) +fill_input_buffer (j_decompress_ptr cinfo) +{ + my_src_ptr src = (my_src_ptr) cinfo->src; + size_t nbytes; + + nbytes = JFREAD(src->infile, src->buffer, INPUT_BUF_SIZE); + + if (nbytes <= 0) { + if (src->start_of_file) /* Treat empty input file as fatal error */ + ERREXIT(cinfo, JERR_INPUT_EMPTY); + WARNMS(cinfo, JWRN_JPEG_EOF); + /* Insert a fake EOI marker */ + src->buffer[0] = (JOCTET) 0xFF; + src->buffer[1] = (JOCTET) JPEG_EOI; + nbytes = 2; + } + + src->pub.next_input_byte = src->buffer; + src->pub.bytes_in_buffer = nbytes; + src->start_of_file = FALSE; + + return TRUE; +} + +#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED) +METHODDEF(boolean) +fill_mem_input_buffer (j_decompress_ptr cinfo) +{ + static const JOCTET mybuffer[4] = { + (JOCTET) 0xFF, (JOCTET) JPEG_EOI, 0, 0 + }; + + /* The whole JPEG data is expected to reside in the supplied memory + * buffer, so any request for more data beyond the given buffer size + * is treated as an error. + */ + WARNMS(cinfo, JWRN_JPEG_EOF); + + /* Insert a fake EOI marker */ + + cinfo->src->next_input_byte = mybuffer; + cinfo->src->bytes_in_buffer = 2; + + return TRUE; +} +#endif + + +/* + * Skip data --- used to skip over a potentially large amount of + * uninteresting data (such as an APPn marker). + * + * Writers of suspendable-input applications must note that skip_input_data + * is not granted the right to give a suspension return. If the skip extends + * beyond the data currently in the buffer, the buffer can be marked empty so + * that the next read will cause a fill_input_buffer call that can suspend. + * Arranging for additional bytes to be discarded before reloading the input + * buffer is the application writer's problem. + */ + +METHODDEF(void) +skip_input_data (j_decompress_ptr cinfo, long num_bytes) +{ + struct jpeg_source_mgr * src = cinfo->src; + + /* Just a dumb implementation for now. Could use fseek() except + * it doesn't work on pipes. Not clear that being smart is worth + * any trouble anyway --- large skips are infrequent. + */ + if (num_bytes > 0) { + while (num_bytes > (long) src->bytes_in_buffer) { + num_bytes -= (long) src->bytes_in_buffer; + (void) (*src->fill_input_buffer) (cinfo); + /* note we assume that fill_input_buffer will never return FALSE, + * so suspension need not be handled. + */ + } + src->next_input_byte += (size_t) num_bytes; + src->bytes_in_buffer -= (size_t) num_bytes; + } +} + + +/* + * An additional method that can be provided by data source modules is the + * resync_to_restart method for error recovery in the presence of RST markers. + * For the moment, this source module just uses the default resync method + * provided by the JPEG library. That method assumes that no backtracking + * is possible. + */ + + +/* + * Terminate source --- called by jpeg_finish_decompress + * after all data has been read. Often a no-op. + * + * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding + * application must deal with any cleanup that should happen even + * for error exit. + */ + +METHODDEF(void) +term_source (j_decompress_ptr cinfo) +{ + /* no work necessary here */ +} + + +/* + * Prepare for input from a stdio stream. + * The caller must have already opened the stream, and is responsible + * for closing it after finishing decompression. + */ + +GLOBAL(void) +jpeg_stdio_src (j_decompress_ptr cinfo, FILE * infile) +{ + my_src_ptr src; + + /* The source object and input buffer are made permanent so that a series + * of JPEG images can be read from the same file by calling jpeg_stdio_src + * only before the first one. (If we discarded the buffer at the end of + * one image, we'd likely lose the start of the next one.) + * This makes it unsafe to use this manager and a different source + * manager serially with the same JPEG object. Caveat programmer. + */ + if (cinfo->src == NULL) { /* first time for this JPEG object? */ + cinfo->src = (struct jpeg_source_mgr *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT, + sizeof(my_source_mgr)); + src = (my_src_ptr) cinfo->src; + src->buffer = (JOCTET *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT, + INPUT_BUF_SIZE * sizeof(JOCTET)); + } + + src = (my_src_ptr) cinfo->src; + src->pub.init_source = init_source; + src->pub.fill_input_buffer = fill_input_buffer; + src->pub.skip_input_data = skip_input_data; + src->pub.resync_to_restart = jpeg_resync_to_restart; /* use default method */ + src->pub.term_source = term_source; + src->infile = infile; + src->pub.bytes_in_buffer = 0; /* forces fill_input_buffer on first read */ + src->pub.next_input_byte = NULL; /* until buffer loaded */ +} + + +#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED) +/* + * Prepare for input from a supplied memory buffer. + * The buffer must contain the whole JPEG data. + */ + +GLOBAL(void) +jpeg_mem_src (j_decompress_ptr cinfo, + unsigned char * inbuffer, unsigned long insize) +{ + struct jpeg_source_mgr * src; + + if (inbuffer == NULL || insize == 0) /* Treat empty input as fatal error */ + ERREXIT(cinfo, JERR_INPUT_EMPTY); + + /* The source object is made permanent so that a series of JPEG images + * can be read from the same buffer by calling jpeg_mem_src only before + * the first one. + */ + if (cinfo->src == NULL) { /* first time for this JPEG object? */ + cinfo->src = (struct jpeg_source_mgr *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT, + sizeof(struct jpeg_source_mgr)); + } + + src = cinfo->src; + src->init_source = init_mem_source; + src->fill_input_buffer = fill_mem_input_buffer; + src->skip_input_data = skip_input_data; + src->resync_to_restart = jpeg_resync_to_restart; /* use default method */ + src->term_source = term_source; + src->bytes_in_buffer = (size_t) insize; + src->next_input_byte = (JOCTET *) inbuffer; +} +#endif diff --git a/app/src/main/jni/libjpeg-turbo/jdcoefct.c b/app/src/main/jni/libjpeg-turbo/jdcoefct.c new file mode 100644 index 000000000..199a628b8 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdcoefct.c @@ -0,0 +1,747 @@ +/* + * jdcoefct.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1997, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright (C) 2010, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains the coefficient buffer controller for decompression. + * This controller is the top level of the JPEG decompressor proper. + * The coefficient buffer lies between entropy decoding and inverse-DCT steps. + * + * In buffered-image mode, this controller is the interface between + * input-oriented processing and output-oriented processing. + * Also, the input side (only) is used when reading a file for transcoding. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jpegcomp.h" + +/* Block smoothing is only applicable for progressive JPEG, so: */ +#ifndef D_PROGRESSIVE_SUPPORTED +#undef BLOCK_SMOOTHING_SUPPORTED +#endif + +/* Private buffer controller object */ + +typedef struct { + struct jpeg_d_coef_controller pub; /* public fields */ + + /* These variables keep track of the current location of the input side. */ + /* cinfo->input_iMCU_row is also used for this. */ + JDIMENSION MCU_ctr; /* counts MCUs processed in current row */ + int MCU_vert_offset; /* counts MCU rows within iMCU row */ + int MCU_rows_per_iMCU_row; /* number of such rows needed */ + + /* The output side's location is represented by cinfo->output_iMCU_row. */ + + /* In single-pass modes, it's sufficient to buffer just one MCU. + * We allocate a workspace of D_MAX_BLOCKS_IN_MCU coefficient blocks, + * and let the entropy decoder write into that workspace each time. + * In multi-pass modes, this array points to the current MCU's blocks + * within the virtual arrays; it is used only by the input side. + */ + JBLOCKROW MCU_buffer[D_MAX_BLOCKS_IN_MCU]; + + /* Temporary workspace for one MCU */ + JCOEF * workspace; + +#ifdef D_MULTISCAN_FILES_SUPPORTED + /* In multi-pass modes, we need a virtual block array for each component. */ + jvirt_barray_ptr whole_image[MAX_COMPONENTS]; +#endif + +#ifdef BLOCK_SMOOTHING_SUPPORTED + /* When doing block smoothing, we latch coefficient Al values here */ + int * coef_bits_latch; +#define SAVED_COEFS 6 /* we save coef_bits[0..5] */ +#endif +} my_coef_controller; + +typedef my_coef_controller * my_coef_ptr; + +/* Forward declarations */ +METHODDEF(int) decompress_onepass + (j_decompress_ptr cinfo, JSAMPIMAGE output_buf); +#ifdef D_MULTISCAN_FILES_SUPPORTED +METHODDEF(int) decompress_data + (j_decompress_ptr cinfo, JSAMPIMAGE output_buf); +#endif +#ifdef BLOCK_SMOOTHING_SUPPORTED +LOCAL(boolean) smoothing_ok (j_decompress_ptr cinfo); +METHODDEF(int) decompress_smooth_data + (j_decompress_ptr cinfo, JSAMPIMAGE output_buf); +#endif + + +LOCAL(void) +start_iMCU_row (j_decompress_ptr cinfo) +/* Reset within-iMCU-row counters for a new row (input side) */ +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + + /* In an interleaved scan, an MCU row is the same as an iMCU row. + * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows. + * But at the bottom of the image, process only what's left. + */ + if (cinfo->comps_in_scan > 1) { + coef->MCU_rows_per_iMCU_row = 1; + } else { + if (cinfo->input_iMCU_row < (cinfo->total_iMCU_rows-1)) + coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->v_samp_factor; + else + coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->last_row_height; + } + + coef->MCU_ctr = 0; + coef->MCU_vert_offset = 0; +} + + +/* + * Initialize for an input processing pass. + */ + +METHODDEF(void) +start_input_pass (j_decompress_ptr cinfo) +{ + cinfo->input_iMCU_row = 0; + start_iMCU_row(cinfo); +} + + +/* + * Initialize for an output processing pass. + */ + +METHODDEF(void) +start_output_pass (j_decompress_ptr cinfo) +{ +#ifdef BLOCK_SMOOTHING_SUPPORTED + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + + /* If multipass, check to see whether to use block smoothing on this pass */ + if (coef->pub.coef_arrays != NULL) { + if (cinfo->do_block_smoothing && smoothing_ok(cinfo)) + coef->pub.decompress_data = decompress_smooth_data; + else + coef->pub.decompress_data = decompress_data; + } +#endif + cinfo->output_iMCU_row = 0; +} + + +/* + * Decompress and return some data in the single-pass case. + * Always attempts to emit one fully interleaved MCU row ("iMCU" row). + * Input and output must run in lockstep since we have only a one-MCU buffer. + * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED. + * + * NB: output_buf contains a plane for each component in image, + * which we index according to the component's SOF position. + */ + +METHODDEF(int) +decompress_onepass (j_decompress_ptr cinfo, JSAMPIMAGE output_buf) +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + JDIMENSION MCU_col_num; /* index of current MCU within row */ + JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1; + JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1; + int blkn, ci, xindex, yindex, yoffset, useful_width; + JSAMPARRAY output_ptr; + JDIMENSION start_col, output_col; + jpeg_component_info *compptr; + inverse_DCT_method_ptr inverse_DCT; + + /* Loop to process as much as one whole iMCU row */ + for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row; + yoffset++) { + for (MCU_col_num = coef->MCU_ctr; MCU_col_num <= last_MCU_col; + MCU_col_num++) { + /* Try to fetch an MCU. Entropy decoder expects buffer to be zeroed. */ + jzero_far((void *) coef->MCU_buffer[0], + (size_t) (cinfo->blocks_in_MCU * sizeof(JBLOCK))); + if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) { + /* Suspension forced; update state counters and exit */ + coef->MCU_vert_offset = yoffset; + coef->MCU_ctr = MCU_col_num; + return JPEG_SUSPENDED; + } + /* Determine where data should go in output_buf and do the IDCT thing. + * We skip dummy blocks at the right and bottom edges (but blkn gets + * incremented past them!). Note the inner loop relies on having + * allocated the MCU_buffer[] blocks sequentially. + */ + blkn = 0; /* index of current DCT block within MCU */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + /* Don't bother to IDCT an uninteresting component. */ + if (! compptr->component_needed) { + blkn += compptr->MCU_blocks; + continue; + } + inverse_DCT = cinfo->idct->inverse_DCT[compptr->component_index]; + useful_width = (MCU_col_num < last_MCU_col) ? compptr->MCU_width + : compptr->last_col_width; + output_ptr = output_buf[compptr->component_index] + + yoffset * compptr->_DCT_scaled_size; + start_col = MCU_col_num * compptr->MCU_sample_width; + for (yindex = 0; yindex < compptr->MCU_height; yindex++) { + if (cinfo->input_iMCU_row < last_iMCU_row || + yoffset+yindex < compptr->last_row_height) { + output_col = start_col; + for (xindex = 0; xindex < useful_width; xindex++) { + (*inverse_DCT) (cinfo, compptr, + (JCOEFPTR) coef->MCU_buffer[blkn+xindex], + output_ptr, output_col); + output_col += compptr->_DCT_scaled_size; + } + } + blkn += compptr->MCU_width; + output_ptr += compptr->_DCT_scaled_size; + } + } + } + /* Completed an MCU row, but perhaps not an iMCU row */ + coef->MCU_ctr = 0; + } + /* Completed the iMCU row, advance counters for next one */ + cinfo->output_iMCU_row++; + if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) { + start_iMCU_row(cinfo); + return JPEG_ROW_COMPLETED; + } + /* Completed the scan */ + (*cinfo->inputctl->finish_input_pass) (cinfo); + return JPEG_SCAN_COMPLETED; +} + + +/* + * Dummy consume-input routine for single-pass operation. + */ + +METHODDEF(int) +dummy_consume_data (j_decompress_ptr cinfo) +{ + return JPEG_SUSPENDED; /* Always indicate nothing was done */ +} + + +#ifdef D_MULTISCAN_FILES_SUPPORTED + +/* + * Consume input data and store it in the full-image coefficient buffer. + * We read as much as one fully interleaved MCU row ("iMCU" row) per call, + * ie, v_samp_factor block rows for each component in the scan. + * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED. + */ + +METHODDEF(int) +consume_data (j_decompress_ptr cinfo) +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + JDIMENSION MCU_col_num; /* index of current MCU within row */ + int blkn, ci, xindex, yindex, yoffset; + JDIMENSION start_col; + JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN]; + JBLOCKROW buffer_ptr; + jpeg_component_info *compptr; + + /* Align the virtual buffers for the components used in this scan. */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + buffer[ci] = (*cinfo->mem->access_virt_barray) + ((j_common_ptr) cinfo, coef->whole_image[compptr->component_index], + cinfo->input_iMCU_row * compptr->v_samp_factor, + (JDIMENSION) compptr->v_samp_factor, TRUE); + /* Note: entropy decoder expects buffer to be zeroed, + * but this is handled automatically by the memory manager + * because we requested a pre-zeroed array. + */ + } + + /* Loop to process one whole iMCU row */ + for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row; + yoffset++) { + for (MCU_col_num = coef->MCU_ctr; MCU_col_num < cinfo->MCUs_per_row; + MCU_col_num++) { + /* Construct list of pointers to DCT blocks belonging to this MCU */ + blkn = 0; /* index of current DCT block within MCU */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + start_col = MCU_col_num * compptr->MCU_width; + for (yindex = 0; yindex < compptr->MCU_height; yindex++) { + buffer_ptr = buffer[ci][yindex+yoffset] + start_col; + for (xindex = 0; xindex < compptr->MCU_width; xindex++) { + coef->MCU_buffer[blkn++] = buffer_ptr++; + } + } + } + /* Try to fetch the MCU. */ + if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) { + /* Suspension forced; update state counters and exit */ + coef->MCU_vert_offset = yoffset; + coef->MCU_ctr = MCU_col_num; + return JPEG_SUSPENDED; + } + } + /* Completed an MCU row, but perhaps not an iMCU row */ + coef->MCU_ctr = 0; + } + /* Completed the iMCU row, advance counters for next one */ + if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) { + start_iMCU_row(cinfo); + return JPEG_ROW_COMPLETED; + } + /* Completed the scan */ + (*cinfo->inputctl->finish_input_pass) (cinfo); + return JPEG_SCAN_COMPLETED; +} + + +/* + * Decompress and return some data in the multi-pass case. + * Always attempts to emit one fully interleaved MCU row ("iMCU" row). + * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED. + * + * NB: output_buf contains a plane for each component in image. + */ + +METHODDEF(int) +decompress_data (j_decompress_ptr cinfo, JSAMPIMAGE output_buf) +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1; + JDIMENSION block_num; + int ci, block_row, block_rows; + JBLOCKARRAY buffer; + JBLOCKROW buffer_ptr; + JSAMPARRAY output_ptr; + JDIMENSION output_col; + jpeg_component_info *compptr; + inverse_DCT_method_ptr inverse_DCT; + + /* Force some input to be done if we are getting ahead of the input. */ + while (cinfo->input_scan_number < cinfo->output_scan_number || + (cinfo->input_scan_number == cinfo->output_scan_number && + cinfo->input_iMCU_row <= cinfo->output_iMCU_row)) { + if ((*cinfo->inputctl->consume_input)(cinfo) == JPEG_SUSPENDED) + return JPEG_SUSPENDED; + } + + /* OK, output from the virtual arrays. */ + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + /* Don't bother to IDCT an uninteresting component. */ + if (! compptr->component_needed) + continue; + /* Align the virtual buffer for this component. */ + buffer = (*cinfo->mem->access_virt_barray) + ((j_common_ptr) cinfo, coef->whole_image[ci], + cinfo->output_iMCU_row * compptr->v_samp_factor, + (JDIMENSION) compptr->v_samp_factor, FALSE); + /* Count non-dummy DCT block rows in this iMCU row. */ + if (cinfo->output_iMCU_row < last_iMCU_row) + block_rows = compptr->v_samp_factor; + else { + /* NB: can't use last_row_height here; it is input-side-dependent! */ + block_rows = (int) (compptr->height_in_blocks % compptr->v_samp_factor); + if (block_rows == 0) block_rows = compptr->v_samp_factor; + } + inverse_DCT = cinfo->idct->inverse_DCT[ci]; + output_ptr = output_buf[ci]; + /* Loop over all DCT blocks to be processed. */ + for (block_row = 0; block_row < block_rows; block_row++) { + buffer_ptr = buffer[block_row]; + output_col = 0; + for (block_num = 0; block_num < compptr->width_in_blocks; block_num++) { + (*inverse_DCT) (cinfo, compptr, (JCOEFPTR) buffer_ptr, + output_ptr, output_col); + buffer_ptr++; + output_col += compptr->_DCT_scaled_size; + } + output_ptr += compptr->_DCT_scaled_size; + } + } + + if (++(cinfo->output_iMCU_row) < cinfo->total_iMCU_rows) + return JPEG_ROW_COMPLETED; + return JPEG_SCAN_COMPLETED; +} + +#endif /* D_MULTISCAN_FILES_SUPPORTED */ + + +#ifdef BLOCK_SMOOTHING_SUPPORTED + +/* + * This code applies interblock smoothing as described by section K.8 + * of the JPEG standard: the first 5 AC coefficients are estimated from + * the DC values of a DCT block and its 8 neighboring blocks. + * We apply smoothing only for progressive JPEG decoding, and only if + * the coefficients it can estimate are not yet known to full precision. + */ + +/* Natural-order array positions of the first 5 zigzag-order coefficients */ +#define Q01_POS 1 +#define Q10_POS 8 +#define Q20_POS 16 +#define Q11_POS 9 +#define Q02_POS 2 + +/* + * Determine whether block smoothing is applicable and safe. + * We also latch the current states of the coef_bits[] entries for the + * AC coefficients; otherwise, if the input side of the decompressor + * advances into a new scan, we might think the coefficients are known + * more accurately than they really are. + */ + +LOCAL(boolean) +smoothing_ok (j_decompress_ptr cinfo) +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + boolean smoothing_useful = FALSE; + int ci, coefi; + jpeg_component_info *compptr; + JQUANT_TBL * qtable; + int * coef_bits; + int * coef_bits_latch; + + if (! cinfo->progressive_mode || cinfo->coef_bits == NULL) + return FALSE; + + /* Allocate latch area if not already done */ + if (coef->coef_bits_latch == NULL) + coef->coef_bits_latch = (int *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + cinfo->num_components * + (SAVED_COEFS * sizeof(int))); + coef_bits_latch = coef->coef_bits_latch; + + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + /* All components' quantization values must already be latched. */ + if ((qtable = compptr->quant_table) == NULL) + return FALSE; + /* Verify DC & first 5 AC quantizers are nonzero to avoid zero-divide. */ + if (qtable->quantval[0] == 0 || + qtable->quantval[Q01_POS] == 0 || + qtable->quantval[Q10_POS] == 0 || + qtable->quantval[Q20_POS] == 0 || + qtable->quantval[Q11_POS] == 0 || + qtable->quantval[Q02_POS] == 0) + return FALSE; + /* DC values must be at least partly known for all components. */ + coef_bits = cinfo->coef_bits[ci]; + if (coef_bits[0] < 0) + return FALSE; + /* Block smoothing is helpful if some AC coefficients remain inaccurate. */ + for (coefi = 1; coefi <= 5; coefi++) { + coef_bits_latch[coefi] = coef_bits[coefi]; + if (coef_bits[coefi] != 0) + smoothing_useful = TRUE; + } + coef_bits_latch += SAVED_COEFS; + } + + return smoothing_useful; +} + + +/* + * Variant of decompress_data for use when doing block smoothing. + */ + +METHODDEF(int) +decompress_smooth_data (j_decompress_ptr cinfo, JSAMPIMAGE output_buf) +{ + my_coef_ptr coef = (my_coef_ptr) cinfo->coef; + JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1; + JDIMENSION block_num, last_block_column; + int ci, block_row, block_rows, access_rows; + JBLOCKARRAY buffer; + JBLOCKROW buffer_ptr, prev_block_row, next_block_row; + JSAMPARRAY output_ptr; + JDIMENSION output_col; + jpeg_component_info *compptr; + inverse_DCT_method_ptr inverse_DCT; + boolean first_row, last_row; + JCOEF * workspace; + int *coef_bits; + JQUANT_TBL *quanttbl; + INT32 Q00,Q01,Q02,Q10,Q11,Q20, num; + int DC1,DC2,DC3,DC4,DC5,DC6,DC7,DC8,DC9; + int Al, pred; + + /* Keep a local variable to avoid looking it up more than once */ + workspace = coef->workspace; + + /* Force some input to be done if we are getting ahead of the input. */ + while (cinfo->input_scan_number <= cinfo->output_scan_number && + ! cinfo->inputctl->eoi_reached) { + if (cinfo->input_scan_number == cinfo->output_scan_number) { + /* If input is working on current scan, we ordinarily want it to + * have completed the current row. But if input scan is DC, + * we want it to keep one row ahead so that next block row's DC + * values are up to date. + */ + JDIMENSION delta = (cinfo->Ss == 0) ? 1 : 0; + if (cinfo->input_iMCU_row > cinfo->output_iMCU_row+delta) + break; + } + if ((*cinfo->inputctl->consume_input)(cinfo) == JPEG_SUSPENDED) + return JPEG_SUSPENDED; + } + + /* OK, output from the virtual arrays. */ + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + /* Don't bother to IDCT an uninteresting component. */ + if (! compptr->component_needed) + continue; + /* Count non-dummy DCT block rows in this iMCU row. */ + if (cinfo->output_iMCU_row < last_iMCU_row) { + block_rows = compptr->v_samp_factor; + access_rows = block_rows * 2; /* this and next iMCU row */ + last_row = FALSE; + } else { + /* NB: can't use last_row_height here; it is input-side-dependent! */ + block_rows = (int) (compptr->height_in_blocks % compptr->v_samp_factor); + if (block_rows == 0) block_rows = compptr->v_samp_factor; + access_rows = block_rows; /* this iMCU row only */ + last_row = TRUE; + } + /* Align the virtual buffer for this component. */ + if (cinfo->output_iMCU_row > 0) { + access_rows += compptr->v_samp_factor; /* prior iMCU row too */ + buffer = (*cinfo->mem->access_virt_barray) + ((j_common_ptr) cinfo, coef->whole_image[ci], + (cinfo->output_iMCU_row - 1) * compptr->v_samp_factor, + (JDIMENSION) access_rows, FALSE); + buffer += compptr->v_samp_factor; /* point to current iMCU row */ + first_row = FALSE; + } else { + buffer = (*cinfo->mem->access_virt_barray) + ((j_common_ptr) cinfo, coef->whole_image[ci], + (JDIMENSION) 0, (JDIMENSION) access_rows, FALSE); + first_row = TRUE; + } + /* Fetch component-dependent info */ + coef_bits = coef->coef_bits_latch + (ci * SAVED_COEFS); + quanttbl = compptr->quant_table; + Q00 = quanttbl->quantval[0]; + Q01 = quanttbl->quantval[Q01_POS]; + Q10 = quanttbl->quantval[Q10_POS]; + Q20 = quanttbl->quantval[Q20_POS]; + Q11 = quanttbl->quantval[Q11_POS]; + Q02 = quanttbl->quantval[Q02_POS]; + inverse_DCT = cinfo->idct->inverse_DCT[ci]; + output_ptr = output_buf[ci]; + /* Loop over all DCT blocks to be processed. */ + for (block_row = 0; block_row < block_rows; block_row++) { + buffer_ptr = buffer[block_row]; + if (first_row && block_row == 0) + prev_block_row = buffer_ptr; + else + prev_block_row = buffer[block_row-1]; + if (last_row && block_row == block_rows-1) + next_block_row = buffer_ptr; + else + next_block_row = buffer[block_row+1]; + /* We fetch the surrounding DC values using a sliding-register approach. + * Initialize all nine here so as to do the right thing on narrow pics. + */ + DC1 = DC2 = DC3 = (int) prev_block_row[0][0]; + DC4 = DC5 = DC6 = (int) buffer_ptr[0][0]; + DC7 = DC8 = DC9 = (int) next_block_row[0][0]; + output_col = 0; + last_block_column = compptr->width_in_blocks - 1; + for (block_num = 0; block_num <= last_block_column; block_num++) { + /* Fetch current DCT block into workspace so we can modify it. */ + jcopy_block_row(buffer_ptr, (JBLOCKROW) workspace, (JDIMENSION) 1); + /* Update DC values */ + if (block_num < last_block_column) { + DC3 = (int) prev_block_row[1][0]; + DC6 = (int) buffer_ptr[1][0]; + DC9 = (int) next_block_row[1][0]; + } + /* Compute coefficient estimates per K.8. + * An estimate is applied only if coefficient is still zero, + * and is not known to be fully accurate. + */ + /* AC01 */ + if ((Al=coef_bits[1]) != 0 && workspace[1] == 0) { + num = 36 * Q00 * (DC4 - DC6); + if (num >= 0) { + pred = (int) (((Q01<<7) + num) / (Q01<<8)); + if (Al > 0 && pred >= (1< 0 && pred >= (1<= 0) { + pred = (int) (((Q10<<7) + num) / (Q10<<8)); + if (Al > 0 && pred >= (1< 0 && pred >= (1<= 0) { + pred = (int) (((Q20<<7) + num) / (Q20<<8)); + if (Al > 0 && pred >= (1< 0 && pred >= (1<= 0) { + pred = (int) (((Q11<<7) + num) / (Q11<<8)); + if (Al > 0 && pred >= (1< 0 && pred >= (1<= 0) { + pred = (int) (((Q02<<7) + num) / (Q02<<8)); + if (Al > 0 && pred >= (1< 0 && pred >= (1<_DCT_scaled_size; + } + output_ptr += compptr->_DCT_scaled_size; + } + } + + if (++(cinfo->output_iMCU_row) < cinfo->total_iMCU_rows) + return JPEG_ROW_COMPLETED; + return JPEG_SCAN_COMPLETED; +} + +#endif /* BLOCK_SMOOTHING_SUPPORTED */ + + +/* + * Initialize coefficient buffer controller. + */ + +GLOBAL(void) +jinit_d_coef_controller (j_decompress_ptr cinfo, boolean need_full_buffer) +{ + my_coef_ptr coef; + + coef = (my_coef_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(my_coef_controller)); + cinfo->coef = (struct jpeg_d_coef_controller *) coef; + coef->pub.start_input_pass = start_input_pass; + coef->pub.start_output_pass = start_output_pass; +#ifdef BLOCK_SMOOTHING_SUPPORTED + coef->coef_bits_latch = NULL; +#endif + + /* Create the coefficient buffer. */ + if (need_full_buffer) { +#ifdef D_MULTISCAN_FILES_SUPPORTED + /* Allocate a full-image virtual array for each component, */ + /* padded to a multiple of samp_factor DCT blocks in each direction. */ + /* Note we ask for a pre-zeroed array. */ + int ci, access_rows; + jpeg_component_info *compptr; + + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + access_rows = compptr->v_samp_factor; +#ifdef BLOCK_SMOOTHING_SUPPORTED + /* If block smoothing could be used, need a bigger window */ + if (cinfo->progressive_mode) + access_rows *= 3; +#endif + coef->whole_image[ci] = (*cinfo->mem->request_virt_barray) + ((j_common_ptr) cinfo, JPOOL_IMAGE, TRUE, + (JDIMENSION) jround_up((long) compptr->width_in_blocks, + (long) compptr->h_samp_factor), + (JDIMENSION) jround_up((long) compptr->height_in_blocks, + (long) compptr->v_samp_factor), + (JDIMENSION) access_rows); + } + coef->pub.consume_data = consume_data; + coef->pub.decompress_data = decompress_data; + coef->pub.coef_arrays = coef->whole_image; /* link to virtual arrays */ +#else + ERREXIT(cinfo, JERR_NOT_COMPILED); +#endif + } else { + /* We only need a single-MCU buffer. */ + JBLOCKROW buffer; + int i; + + buffer = (JBLOCKROW) + (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE, + D_MAX_BLOCKS_IN_MCU * sizeof(JBLOCK)); + for (i = 0; i < D_MAX_BLOCKS_IN_MCU; i++) { + coef->MCU_buffer[i] = buffer + i; + } + coef->pub.consume_data = dummy_consume_data; + coef->pub.decompress_data = decompress_onepass; + coef->pub.coef_arrays = NULL; /* flag for no virtual arrays */ + } + + /* Allocate the workspace buffer */ + coef->workspace = (JCOEF *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(JCOEF) * DCTSIZE2); +} diff --git a/app/src/main/jni/libjpeg-turbo/jdcol565.c b/app/src/main/jni/libjpeg-turbo/jdcol565.c new file mode 100644 index 000000000..695f2621e --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdcol565.c @@ -0,0 +1,383 @@ +/* + * jdcol565.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1997, Thomas G. Lane. + * Modifications: + * Copyright (C) 2013, Linaro Limited. + * Copyright (C) 2014, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains output colorspace conversion routines. + */ + +/* This file is included by jdcolor.c */ + + +INLINE +LOCAL(void) +ycc_rgb565_convert_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert; + register int y, cb, cr; + register JSAMPROW outptr; + register JSAMPROW inptr0, inptr1, inptr2; + register JDIMENSION col; + JDIMENSION num_cols = cinfo->output_width; + /* copy these pointers into registers if possible */ + register JSAMPLE * range_limit = cinfo->sample_range_limit; + register int * Crrtab = cconvert->Cr_r_tab; + register int * Cbbtab = cconvert->Cb_b_tab; + register INT32 * Crgtab = cconvert->Cr_g_tab; + register INT32 * Cbgtab = cconvert->Cb_g_tab; + SHIFT_TEMPS + + while (--num_rows >= 0) { + INT32 rgb; + unsigned int r, g, b; + inptr0 = input_buf[0][input_row]; + inptr1 = input_buf[1][input_row]; + inptr2 = input_buf[2][input_row]; + input_row++; + outptr = *output_buf++; + + if (PACK_NEED_ALIGNMENT(outptr)) { + y = GETJSAMPLE(*inptr0++); + cb = GETJSAMPLE(*inptr1++); + cr = GETJSAMPLE(*inptr2++); + r = range_limit[y + Crrtab[cr]]; + g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], + SCALEBITS))]; + b = range_limit[y + Cbbtab[cb]]; + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr = rgb; + outptr += 2; + num_cols--; + } + for (col = 0; col < (num_cols >> 1); col++) { + y = GETJSAMPLE(*inptr0++); + cb = GETJSAMPLE(*inptr1++); + cr = GETJSAMPLE(*inptr2++); + r = range_limit[y + Crrtab[cr]]; + g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], + SCALEBITS))]; + b = range_limit[y + Cbbtab[cb]]; + rgb = PACK_SHORT_565(r, g, b); + + y = GETJSAMPLE(*inptr0++); + cb = GETJSAMPLE(*inptr1++); + cr = GETJSAMPLE(*inptr2++); + r = range_limit[y + Crrtab[cr]]; + g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], + SCALEBITS))]; + b = range_limit[y + Cbbtab[cb]]; + rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b)); + + WRITE_TWO_ALIGNED_PIXELS(outptr, rgb); + outptr += 4; + } + if (num_cols & 1) { + y = GETJSAMPLE(*inptr0); + cb = GETJSAMPLE(*inptr1); + cr = GETJSAMPLE(*inptr2); + r = range_limit[y + Crrtab[cr]]; + g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], + SCALEBITS))]; + b = range_limit[y + Cbbtab[cb]]; + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr = rgb; + } + } +} + + +INLINE +LOCAL(void) +ycc_rgb565D_convert_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert; + register int y, cb, cr; + register JSAMPROW outptr; + register JSAMPROW inptr0, inptr1, inptr2; + register JDIMENSION col; + JDIMENSION num_cols = cinfo->output_width; + /* copy these pointers into registers if possible */ + register JSAMPLE * range_limit = cinfo->sample_range_limit; + register int * Crrtab = cconvert->Cr_r_tab; + register int * Cbbtab = cconvert->Cb_b_tab; + register INT32 * Crgtab = cconvert->Cr_g_tab; + register INT32 * Cbgtab = cconvert->Cb_g_tab; + INT32 d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK]; + SHIFT_TEMPS + + while (--num_rows >= 0) { + INT32 rgb; + unsigned int r, g, b; + + inptr0 = input_buf[0][input_row]; + inptr1 = input_buf[1][input_row]; + inptr2 = input_buf[2][input_row]; + input_row++; + outptr = *output_buf++; + if (PACK_NEED_ALIGNMENT(outptr)) { + y = GETJSAMPLE(*inptr0++); + cb = GETJSAMPLE(*inptr1++); + cr = GETJSAMPLE(*inptr2++); + r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)]; + g = range_limit[DITHER_565_G(y + + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], + SCALEBITS)), d0)]; + b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)]; + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr = rgb; + outptr += 2; + num_cols--; + } + for (col = 0; col < (num_cols >> 1); col++) { + y = GETJSAMPLE(*inptr0++); + cb = GETJSAMPLE(*inptr1++); + cr = GETJSAMPLE(*inptr2++); + r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)]; + g = range_limit[DITHER_565_G(y + + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], + SCALEBITS)), d0)]; + b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)]; + d0 = DITHER_ROTATE(d0); + rgb = PACK_SHORT_565(r, g, b); + + y = GETJSAMPLE(*inptr0++); + cb = GETJSAMPLE(*inptr1++); + cr = GETJSAMPLE(*inptr2++); + r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)]; + g = range_limit[DITHER_565_G(y + + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], + SCALEBITS)), d0)]; + b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)]; + d0 = DITHER_ROTATE(d0); + rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b)); + + WRITE_TWO_ALIGNED_PIXELS(outptr, rgb); + outptr += 4; + } + if (num_cols & 1) { + y = GETJSAMPLE(*inptr0); + cb = GETJSAMPLE(*inptr1); + cr = GETJSAMPLE(*inptr2); + r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)]; + g = range_limit[DITHER_565_G(y + + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], + SCALEBITS)), d0)]; + b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)]; + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr = rgb; + } + } +} + + +INLINE +LOCAL(void) +rgb_rgb565_convert_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + register JSAMPROW outptr; + register JSAMPROW inptr0, inptr1, inptr2; + register JDIMENSION col; + JDIMENSION num_cols = cinfo->output_width; + SHIFT_TEMPS + + while (--num_rows >= 0) { + INT32 rgb; + unsigned int r, g, b; + + inptr0 = input_buf[0][input_row]; + inptr1 = input_buf[1][input_row]; + inptr2 = input_buf[2][input_row]; + input_row++; + outptr = *output_buf++; + if (PACK_NEED_ALIGNMENT(outptr)) { + r = GETJSAMPLE(*inptr0++); + g = GETJSAMPLE(*inptr1++); + b = GETJSAMPLE(*inptr2++); + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr = rgb; + outptr += 2; + num_cols--; + } + for (col = 0; col < (num_cols >> 1); col++) { + r = GETJSAMPLE(*inptr0++); + g = GETJSAMPLE(*inptr1++); + b = GETJSAMPLE(*inptr2++); + rgb = PACK_SHORT_565(r, g, b); + + r = GETJSAMPLE(*inptr0++); + g = GETJSAMPLE(*inptr1++); + b = GETJSAMPLE(*inptr2++); + rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b)); + + WRITE_TWO_ALIGNED_PIXELS(outptr, rgb); + outptr += 4; + } + if (num_cols & 1) { + r = GETJSAMPLE(*inptr0); + g = GETJSAMPLE(*inptr1); + b = GETJSAMPLE(*inptr2); + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr = rgb; + } + } +} + + +INLINE +LOCAL(void) +rgb_rgb565D_convert_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + register JSAMPROW outptr; + register JSAMPROW inptr0, inptr1, inptr2; + register JDIMENSION col; + register JSAMPLE * range_limit = cinfo->sample_range_limit; + JDIMENSION num_cols = cinfo->output_width; + INT32 d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK]; + SHIFT_TEMPS + + while (--num_rows >= 0) { + INT32 rgb; + unsigned int r, g, b; + + inptr0 = input_buf[0][input_row]; + inptr1 = input_buf[1][input_row]; + inptr2 = input_buf[2][input_row]; + input_row++; + outptr = *output_buf++; + if (PACK_NEED_ALIGNMENT(outptr)) { + r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)]; + g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)]; + b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)]; + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr = rgb; + outptr += 2; + num_cols--; + } + for (col = 0; col < (num_cols >> 1); col++) { + r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)]; + g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)]; + b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)]; + d0 = DITHER_ROTATE(d0); + rgb = PACK_SHORT_565(r, g, b); + + r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)]; + g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)]; + b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)]; + d0 = DITHER_ROTATE(d0); + rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b)); + + WRITE_TWO_ALIGNED_PIXELS(outptr, rgb); + outptr += 4; + } + if (num_cols & 1) { + r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0), d0)]; + g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1), d0)]; + b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2), d0)]; + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr = rgb; + } + } +} + + +INLINE +LOCAL(void) +gray_rgb565_convert_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + register JSAMPROW inptr, outptr; + register JDIMENSION col; + JDIMENSION num_cols = cinfo->output_width; + + while (--num_rows >= 0) { + INT32 rgb; + unsigned int g; + + inptr = input_buf[0][input_row++]; + outptr = *output_buf++; + if (PACK_NEED_ALIGNMENT(outptr)) { + g = *inptr++; + rgb = PACK_SHORT_565(g, g, g); + *(INT16*)outptr = rgb; + outptr += 2; + num_cols--; + } + for (col = 0; col < (num_cols >> 1); col++) { + g = *inptr++; + rgb = PACK_SHORT_565(g, g, g); + g = *inptr++; + rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(g, g, g)); + WRITE_TWO_ALIGNED_PIXELS(outptr, rgb); + outptr += 4; + } + if (num_cols & 1) { + g = *inptr; + rgb = PACK_SHORT_565(g, g, g); + *(INT16*)outptr = rgb; + } + } +} + + +INLINE +LOCAL(void) +gray_rgb565D_convert_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + register JSAMPROW inptr, outptr; + register JDIMENSION col; + register JSAMPLE * range_limit = cinfo->sample_range_limit; + JDIMENSION num_cols = cinfo->output_width; + INT32 d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK]; + + while (--num_rows >= 0) { + INT32 rgb; + unsigned int g; + + inptr = input_buf[0][input_row++]; + outptr = *output_buf++; + if (PACK_NEED_ALIGNMENT(outptr)) { + g = *inptr++; + g = range_limit[DITHER_565_R(g, d0)]; + rgb = PACK_SHORT_565(g, g, g); + *(INT16*)outptr = rgb; + outptr += 2; + num_cols--; + } + for (col = 0; col < (num_cols >> 1); col++) { + g = *inptr++; + g = range_limit[DITHER_565_R(g, d0)]; + rgb = PACK_SHORT_565(g, g, g); + d0 = DITHER_ROTATE(d0); + + g = *inptr++; + g = range_limit[DITHER_565_R(g, d0)]; + rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(g, g, g)); + d0 = DITHER_ROTATE(d0); + + WRITE_TWO_ALIGNED_PIXELS(outptr, rgb); + outptr += 4; + } + if (num_cols & 1) { + g = *inptr; + g = range_limit[DITHER_565_R(g, d0)]; + rgb = PACK_SHORT_565(g, g, g); + *(INT16*)outptr = rgb; + } + } +} diff --git a/app/src/main/jni/libjpeg-turbo/jdcolext.c b/app/src/main/jni/libjpeg-turbo/jdcolext.c new file mode 100644 index 000000000..d1ac25d88 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdcolext.c @@ -0,0 +1,182 @@ +/* + * jdcolext.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1997, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright (C) 2009, 2011, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains output colorspace conversion routines. + */ + + +/* This file is included by jdcolor.c */ + + +/* + * Convert some rows of samples to the output colorspace. + * + * Note that we change from noninterleaved, one-plane-per-component format + * to interleaved-pixel format. The output buffer is therefore three times + * as wide as the input buffer. + * A starting row offset is provided only for the input buffer. The caller + * can easily adjust the passed output_buf value to accommodate any row + * offset required on that side. + */ + +INLINE +LOCAL(void) +ycc_rgb_convert_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert; + register int y, cb, cr; + register JSAMPROW outptr; + register JSAMPROW inptr0, inptr1, inptr2; + register JDIMENSION col; + JDIMENSION num_cols = cinfo->output_width; + /* copy these pointers into registers if possible */ + register JSAMPLE * range_limit = cinfo->sample_range_limit; + register int * Crrtab = cconvert->Cr_r_tab; + register int * Cbbtab = cconvert->Cb_b_tab; + register INT32 * Crgtab = cconvert->Cr_g_tab; + register INT32 * Cbgtab = cconvert->Cb_g_tab; + SHIFT_TEMPS + + while (--num_rows >= 0) { + inptr0 = input_buf[0][input_row]; + inptr1 = input_buf[1][input_row]; + inptr2 = input_buf[2][input_row]; + input_row++; + outptr = *output_buf++; + for (col = 0; col < num_cols; col++) { + y = GETJSAMPLE(inptr0[col]); + cb = GETJSAMPLE(inptr1[col]); + cr = GETJSAMPLE(inptr2[col]); + /* Range-limiting is essential due to noise introduced by DCT losses. */ + outptr[RGB_RED] = range_limit[y + Crrtab[cr]]; + outptr[RGB_GREEN] = range_limit[y + + ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], + SCALEBITS))]; + outptr[RGB_BLUE] = range_limit[y + Cbbtab[cb]]; + /* Set unused byte to 0xFF so it can be interpreted as an opaque */ + /* alpha channel value */ +#ifdef RGB_ALPHA + outptr[RGB_ALPHA] = 0xFF; +#endif + outptr += RGB_PIXELSIZE; + } + } +} + + +/* + * Convert grayscale to RGB: just duplicate the graylevel three times. + * This is provided to support applications that don't want to cope + * with grayscale as a separate case. + */ + +INLINE +LOCAL(void) +gray_rgb_convert_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + register JSAMPROW inptr, outptr; + register JDIMENSION col; + JDIMENSION num_cols = cinfo->output_width; + + while (--num_rows >= 0) { + inptr = input_buf[0][input_row++]; + outptr = *output_buf++; + for (col = 0; col < num_cols; col++) { + /* We can dispense with GETJSAMPLE() here */ + outptr[RGB_RED] = outptr[RGB_GREEN] = outptr[RGB_BLUE] = inptr[col]; + /* Set unused byte to 0xFF so it can be interpreted as an opaque */ + /* alpha channel value */ +#ifdef RGB_ALPHA + outptr[RGB_ALPHA] = 0xFF; +#endif + outptr += RGB_PIXELSIZE; + } + } +} + + +/* + * Convert RGB to extended RGB: just swap the order of source pixels + */ + +INLINE +LOCAL(void) +rgb_rgb_convert_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + register JSAMPROW inptr0, inptr1, inptr2; + register JSAMPROW outptr; + register JDIMENSION col; + JDIMENSION num_cols = cinfo->output_width; + + while (--num_rows >= 0) { + inptr0 = input_buf[0][input_row]; + inptr1 = input_buf[1][input_row]; + inptr2 = input_buf[2][input_row]; + input_row++; + outptr = *output_buf++; + for (col = 0; col < num_cols; col++) { + /* We can dispense with GETJSAMPLE() here */ + outptr[RGB_RED] = inptr0[col]; + outptr[RGB_GREEN] = inptr1[col]; + outptr[RGB_BLUE] = inptr2[col]; + /* Set unused byte to 0xFF so it can be interpreted as an opaque */ + /* alpha channel value */ +#ifdef RGB_ALPHA + outptr[RGB_ALPHA] = 0xFF; +#endif + outptr += RGB_PIXELSIZE; + } + } +} + + +/* + * Convert cmyk to extended RGB + */ + +INLINE +LOCAL(void) +cmyk_rgb_convert_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) { + + register float lum; + register JSAMPROW inptr0, inptr1, inptr2, inptr3; + register JSAMPROW outptr; + register JDIMENSION col; + JDIMENSION num_cols = cinfo->output_width; + + while (--num_rows >= 0) { + inptr0 = input_buf[0][input_row]; + inptr1 = input_buf[1][input_row]; + inptr2 = input_buf[2][input_row]; + inptr3 = input_buf[3][input_row]; + input_row++; + outptr = *output_buf++; + + for (col = 0; col < num_cols; col++) { + lum = inptr3[col] / 255.0f; + outptr[RGB_RED] = inptr0[col] * lum; + outptr[RGB_GREEN] = inptr1[col] * lum; + outptr[RGB_BLUE] = inptr2[col] * lum; + /* Set unused byte to 0xFF so it can be interpreted as an opaque */ + /* alpha channel value */ +#ifdef RGB_ALPHA + outptr[RGB_ALPHA] = 0xFF; +#endif + outptr += RGB_PIXELSIZE; + } + } +} diff --git a/app/src/main/jni/libjpeg-turbo/jdcolor.c b/app/src/main/jni/libjpeg-turbo/jdcolor.c new file mode 100644 index 000000000..cf30690d1 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdcolor.c @@ -0,0 +1,967 @@ +/* + * jdcolor.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1997, Thomas G. Lane. + * Modified 2011 by Guido Vollbeding. + * libjpeg-turbo Modifications: + * Copyright 2009 Pierre Ossman for Cendio AB + * Copyright (C) 2009, 2011-2012, 2014, D. R. Commander. + * Copyright (C) 2013, Linaro Limited. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains output colorspace conversion routines. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jsimd.h" +#include "jconfigint.h" + + +/* Private subobject */ + +typedef struct { + struct jpeg_color_deconverter pub; /* public fields */ + + /* Private state for YCC->RGB conversion */ + int * Cr_r_tab; /* => table for Cr to R conversion */ + int * Cb_b_tab; /* => table for Cb to B conversion */ + INT32 * Cr_g_tab; /* => table for Cr to G conversion */ + INT32 * Cb_g_tab; /* => table for Cb to G conversion */ + + /* Private state for RGB->Y conversion */ + INT32 * rgb_y_tab; /* => table for RGB to Y conversion */ +} my_color_deconverter; + +typedef my_color_deconverter * my_cconvert_ptr; + + +/**************** YCbCr -> RGB conversion: most common case **************/ +/**************** RGB -> Y conversion: less common case **************/ + +/* + * YCbCr is defined per CCIR 601-1, except that Cb and Cr are + * normalized to the range 0..MAXJSAMPLE rather than -0.5 .. 0.5. + * The conversion equations to be implemented are therefore + * + * R = Y + 1.40200 * Cr + * G = Y - 0.34414 * Cb - 0.71414 * Cr + * B = Y + 1.77200 * Cb + * + * Y = 0.29900 * R + 0.58700 * G + 0.11400 * B + * + * where Cb and Cr represent the incoming values less CENTERJSAMPLE. + * (These numbers are derived from TIFF 6.0 section 21, dated 3-June-92.) + * + * To avoid floating-point arithmetic, we represent the fractional constants + * as integers scaled up by 2^16 (about 4 digits precision); we have to divide + * the products by 2^16, with appropriate rounding, to get the correct answer. + * Notice that Y, being an integral input, does not contribute any fraction + * so it need not participate in the rounding. + * + * For even more speed, we avoid doing any multiplications in the inner loop + * by precalculating the constants times Cb and Cr for all possible values. + * For 8-bit JSAMPLEs this is very reasonable (only 256 entries per table); + * for 12-bit samples it is still acceptable. It's not very reasonable for + * 16-bit samples, but if you want lossless storage you shouldn't be changing + * colorspace anyway. + * The Cr=>R and Cb=>B values can be rounded to integers in advance; the + * values for the G calculation are left scaled up, since we must add them + * together before rounding. + */ + +#define SCALEBITS 16 /* speediest right-shift on some machines */ +#define ONE_HALF ((INT32) 1 << (SCALEBITS-1)) +#define FIX(x) ((INT32) ((x) * (1L<Y conversion and divide it up into + * three parts, instead of doing three alloc_small requests. This lets us + * use a single table base address, which can be held in a register in the + * inner loops on many machines (more than can hold all three addresses, + * anyway). + */ + +#define R_Y_OFF 0 /* offset to R => Y section */ +#define G_Y_OFF (1*(MAXJSAMPLE+1)) /* offset to G => Y section */ +#define B_Y_OFF (2*(MAXJSAMPLE+1)) /* etc. */ +#define TABLE_SIZE (3*(MAXJSAMPLE+1)) + + +/* Include inline routines for colorspace extensions */ + +#include "jdcolext.c" +#undef RGB_RED +#undef RGB_GREEN +#undef RGB_BLUE +#undef RGB_PIXELSIZE + +#define RGB_RED EXT_RGB_RED +#define RGB_GREEN EXT_RGB_GREEN +#define RGB_BLUE EXT_RGB_BLUE +#define RGB_PIXELSIZE EXT_RGB_PIXELSIZE +#define ycc_rgb_convert_internal ycc_extrgb_convert_internal +#define gray_rgb_convert_internal gray_extrgb_convert_internal +#define rgb_rgb_convert_internal rgb_extrgb_convert_internal +#define cmyk_rgb_convert_internal cmyk_extrgb_convert_internal +#include "jdcolext.c" +#undef RGB_RED +#undef RGB_GREEN +#undef RGB_BLUE +#undef RGB_PIXELSIZE +#undef ycc_rgb_convert_internal +#undef gray_rgb_convert_internal +#undef rgb_rgb_convert_internal +#undef cmyk_rgb_convert_internal + +#define RGB_RED EXT_RGBX_RED +#define RGB_GREEN EXT_RGBX_GREEN +#define RGB_BLUE EXT_RGBX_BLUE +#define RGB_ALPHA 3 +#define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE +#define ycc_rgb_convert_internal ycc_extrgbx_convert_internal +#define gray_rgb_convert_internal gray_extrgbx_convert_internal +#define rgb_rgb_convert_internal rgb_extrgbx_convert_internal +#define cmyk_rgb_convert_internal cmyk_extrgbx_convert_internal +#include "jdcolext.c" +#undef RGB_RED +#undef RGB_GREEN +#undef RGB_BLUE +#undef RGB_ALPHA +#undef RGB_PIXELSIZE +#undef ycc_rgb_convert_internal +#undef gray_rgb_convert_internal +#undef rgb_rgb_convert_internal +#undef cmyk_rgb_convert_internal + +#define RGB_RED EXT_BGR_RED +#define RGB_GREEN EXT_BGR_GREEN +#define RGB_BLUE EXT_BGR_BLUE +#define RGB_PIXELSIZE EXT_BGR_PIXELSIZE +#define ycc_rgb_convert_internal ycc_extbgr_convert_internal +#define gray_rgb_convert_internal gray_extbgr_convert_internal +#define rgb_rgb_convert_internal rgb_extbgr_convert_internal +#define cmyk_rgb_convert_internal cmyk_extbgr_convert_internal +#include "jdcolext.c" +#undef RGB_RED +#undef RGB_GREEN +#undef RGB_BLUE +#undef RGB_PIXELSIZE +#undef ycc_rgb_convert_internal +#undef gray_rgb_convert_internal +#undef rgb_rgb_convert_internal +#undef cmyk_rgb_convert_internal + +#define RGB_RED EXT_BGRX_RED +#define RGB_GREEN EXT_BGRX_GREEN +#define RGB_BLUE EXT_BGRX_BLUE +#define RGB_ALPHA 3 +#define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE +#define ycc_rgb_convert_internal ycc_extbgrx_convert_internal +#define gray_rgb_convert_internal gray_extbgrx_convert_internal +#define rgb_rgb_convert_internal rgb_extbgrx_convert_internal +#define cmyk_rgb_convert_internal cmyk_extbgrx_convert_internal +#include "jdcolext.c" +#undef RGB_RED +#undef RGB_GREEN +#undef RGB_BLUE +#undef RGB_ALPHA +#undef RGB_PIXELSIZE +#undef ycc_rgb_convert_internal +#undef gray_rgb_convert_internal +#undef rgb_rgb_convert_internal +#undef cmyk_rgb_convert_internal + +#define RGB_RED EXT_XBGR_RED +#define RGB_GREEN EXT_XBGR_GREEN +#define RGB_BLUE EXT_XBGR_BLUE +#define RGB_ALPHA 0 +#define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE +#define ycc_rgb_convert_internal ycc_extxbgr_convert_internal +#define gray_rgb_convert_internal gray_extxbgr_convert_internal +#define rgb_rgb_convert_internal rgb_extxbgr_convert_internal +#define cmyk_rgb_convert_internal cmyk_extxbgr_convert_internal +#include "jdcolext.c" +#undef RGB_RED +#undef RGB_GREEN +#undef RGB_BLUE +#undef RGB_ALPHA +#undef RGB_PIXELSIZE +#undef ycc_rgb_convert_internal +#undef gray_rgb_convert_internal +#undef rgb_rgb_convert_internal +#undef cmyk_rgb_convert_internal + +#define RGB_RED EXT_XRGB_RED +#define RGB_GREEN EXT_XRGB_GREEN +#define RGB_BLUE EXT_XRGB_BLUE +#define RGB_ALPHA 0 +#define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE +#define ycc_rgb_convert_internal ycc_extxrgb_convert_internal +#define gray_rgb_convert_internal gray_extxrgb_convert_internal +#define rgb_rgb_convert_internal rgb_extxrgb_convert_internal +#define cmyk_rgb_convert_internal cmyk_extxrgb_convert_internal +#include "jdcolext.c" +#undef RGB_RED +#undef RGB_GREEN +#undef RGB_BLUE +#undef RGB_ALPHA +#undef RGB_PIXELSIZE +#undef ycc_rgb_convert_internal +#undef gray_rgb_convert_internal +#undef rgb_rgb_convert_internal +#undef cmyk_rgb_convert_internal + + +/* + * Initialize tables for YCC->RGB colorspace conversion. + */ + +LOCAL(void) +build_ycc_rgb_table (j_decompress_ptr cinfo) +{ + my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert; + int i; + INT32 x; + SHIFT_TEMPS + + cconvert->Cr_r_tab = (int *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + (MAXJSAMPLE+1) * sizeof(int)); + cconvert->Cb_b_tab = (int *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + (MAXJSAMPLE+1) * sizeof(int)); + cconvert->Cr_g_tab = (INT32 *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + (MAXJSAMPLE+1) * sizeof(INT32)); + cconvert->Cb_g_tab = (INT32 *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + (MAXJSAMPLE+1) * sizeof(INT32)); + + for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) { + /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */ + /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */ + /* Cr=>R value is nearest int to 1.40200 * x */ + cconvert->Cr_r_tab[i] = (int) + RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS); + /* Cb=>B value is nearest int to 1.77200 * x */ + cconvert->Cb_b_tab[i] = (int) + RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS); + /* Cr=>G value is scaled-up -0.71414 * x */ + cconvert->Cr_g_tab[i] = (- FIX(0.71414)) * x; + /* Cb=>G value is scaled-up -0.34414 * x */ + /* We also add in ONE_HALF so that need not do it in inner loop */ + cconvert->Cb_g_tab[i] = (- FIX(0.34414)) * x + ONE_HALF; + } +} + + +/* + * Convert some rows of samples to the output colorspace. + */ + +METHODDEF(void) +ycc_rgb_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + switch (cinfo->out_color_space) { + case JCS_EXT_RGB: + ycc_extrgb_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + ycc_extrgbx_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_BGR: + ycc_extbgr_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + ycc_extbgrx_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + ycc_extxbgr_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + ycc_extxrgb_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + default: + ycc_rgb_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + } +} + + +/**************** Cases other than YCbCr -> RGB **************/ + + +/* + * Initialize for RGB->grayscale colorspace conversion. + */ + +LOCAL(void) +build_rgb_y_table (j_decompress_ptr cinfo) +{ + my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert; + INT32 * rgb_y_tab; + INT32 i; + + /* Allocate and fill in the conversion tables. */ + cconvert->rgb_y_tab = rgb_y_tab = (INT32 *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + (TABLE_SIZE * sizeof(INT32))); + + for (i = 0; i <= MAXJSAMPLE; i++) { + rgb_y_tab[i+R_Y_OFF] = FIX(0.29900) * i; + rgb_y_tab[i+G_Y_OFF] = FIX(0.58700) * i; + rgb_y_tab[i+B_Y_OFF] = FIX(0.11400) * i + ONE_HALF; + } +} + + +/* + * Convert RGB to grayscale. + */ + +METHODDEF(void) +rgb_gray_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert; + register int r, g, b; + register INT32 * ctab = cconvert->rgb_y_tab; + register JSAMPROW outptr; + register JSAMPROW inptr0, inptr1, inptr2; + register JDIMENSION col; + JDIMENSION num_cols = cinfo->output_width; + + while (--num_rows >= 0) { + inptr0 = input_buf[0][input_row]; + inptr1 = input_buf[1][input_row]; + inptr2 = input_buf[2][input_row]; + input_row++; + outptr = *output_buf++; + for (col = 0; col < num_cols; col++) { + r = GETJSAMPLE(inptr0[col]); + g = GETJSAMPLE(inptr1[col]); + b = GETJSAMPLE(inptr2[col]); + /* Y */ + outptr[col] = (JSAMPLE) + ((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF]) + >> SCALEBITS); + } + } +} + +/* + * Convert CMYK to grayscale. + */ + +METHODDEF(void) +cmyk_gray_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) { + + my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert; + register float lum; + register int r, g, b; + register INT32 * ctab = cconvert->rgb_y_tab; + register JSAMPROW outptr; + register JSAMPROW inptr0, inptr1, inptr2, inptr3; + register JDIMENSION col; + JDIMENSION num_cols = cinfo->output_width; + + while (--num_rows >= 0) { + inptr0 = input_buf[0][input_row]; + inptr1 = input_buf[1][input_row]; + inptr2 = input_buf[2][input_row]; + inptr3 = input_buf[3][input_row]; + input_row++; + outptr = *output_buf++; + + for (col = 0; col < num_cols; col++) { + lum = (inptr3[col]) / 255.0f; + r = inptr0[col] * lum; + g = inptr1[col] * lum; + b = inptr2[col] * lum; + /* Y */ + outptr[col] = (JSAMPLE) + ((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF]) + >> SCALEBITS); + } + } +} + + +/* + * Color conversion for no colorspace change: just copy the data, + * converting from separate-planes to interleaved representation. + */ + +METHODDEF(void) +null_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + register JSAMPROW inptr, outptr; + register JDIMENSION count; + register int num_components = cinfo->num_components; + JDIMENSION num_cols = cinfo->output_width; + int ci; + + while (--num_rows >= 0) { + for (ci = 0; ci < num_components; ci++) { + inptr = input_buf[ci][input_row]; + outptr = output_buf[0] + ci; + for (count = num_cols; count > 0; count--) { + *outptr = *inptr++; /* needn't bother with GETJSAMPLE() here */ + outptr += num_components; + } + } + input_row++; + output_buf++; + } +} + + +/* + * Color conversion for grayscale: just copy the data. + * This also works for YCbCr -> grayscale conversion, in which + * we just copy the Y (luminance) component and ignore chrominance. + */ + +METHODDEF(void) +grayscale_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + jcopy_sample_rows(input_buf[0], (int) input_row, output_buf, 0, + num_rows, cinfo->output_width); +} + + +/* + * Convert grayscale to RGB + */ + +METHODDEF(void) +gray_rgb_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + switch (cinfo->out_color_space) { + case JCS_EXT_RGB: + gray_extrgb_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + gray_extrgbx_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_BGR: + gray_extbgr_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + gray_extbgrx_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + gray_extxbgr_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + gray_extxrgb_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + default: + gray_rgb_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + } +} + + +/* + * Convert plain RGB to extended RGB + */ + +METHODDEF(void) +rgb_rgb_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + switch (cinfo->out_color_space) { + case JCS_EXT_RGB: + rgb_extrgb_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + rgb_extrgbx_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_BGR: + rgb_extbgr_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + rgb_extbgrx_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + rgb_extxbgr_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + rgb_extxrgb_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + default: + rgb_rgb_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + } +} + +/* + * Convert CMYK to RGB + */ + +METHODDEF(void) +cmyk_rgb_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + switch (cinfo->out_color_space) { + case JCS_EXT_RGB: + cmyk_extrgb_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + cmyk_extrgbx_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_BGR: + cmyk_extbgr_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + cmyk_extbgrx_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + cmyk_extxbgr_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + cmyk_extxrgb_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + default: + cmyk_rgb_convert_internal(cinfo, input_buf, input_row, output_buf, + num_rows); + break; + } +} + + +/* + * Adobe-style YCCK->CMYK conversion. + * We convert YCbCr to R=1-C, G=1-M, and B=1-Y using the same + * conversion as above, while passing K (black) unchanged. + * We assume build_ycc_rgb_table has been called. + */ + +METHODDEF(void) +ycck_cmyk_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert; + register int y, cb, cr; + register JSAMPROW outptr; + register JSAMPROW inptr0, inptr1, inptr2, inptr3; + register JDIMENSION col; + JDIMENSION num_cols = cinfo->output_width; + /* copy these pointers into registers if possible */ + register JSAMPLE * range_limit = cinfo->sample_range_limit; + register int * Crrtab = cconvert->Cr_r_tab; + register int * Cbbtab = cconvert->Cb_b_tab; + register INT32 * Crgtab = cconvert->Cr_g_tab; + register INT32 * Cbgtab = cconvert->Cb_g_tab; + SHIFT_TEMPS + + while (--num_rows >= 0) { + inptr0 = input_buf[0][input_row]; + inptr1 = input_buf[1][input_row]; + inptr2 = input_buf[2][input_row]; + inptr3 = input_buf[3][input_row]; + input_row++; + outptr = *output_buf++; + for (col = 0; col < num_cols; col++) { + y = GETJSAMPLE(inptr0[col]); + cb = GETJSAMPLE(inptr1[col]); + cr = GETJSAMPLE(inptr2[col]); + /* Range-limiting is essential due to noise introduced by DCT losses. */ + outptr[0] = range_limit[MAXJSAMPLE - (y + Crrtab[cr])]; /* red */ + outptr[1] = range_limit[MAXJSAMPLE - (y + /* green */ + ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], + SCALEBITS)))]; + outptr[2] = range_limit[MAXJSAMPLE - (y + Cbbtab[cb])]; /* blue */ + /* K passes through unchanged */ + outptr[3] = inptr3[col]; /* don't need GETJSAMPLE here */ + outptr += 4; + } + } +} + + +/* + * RGB565 conversion + */ + +#define PACK_SHORT_565_LE(r, g, b) ((((r) << 8) & 0xF800) | \ + (((g) << 3) & 0x7E0) | ((b) >> 3)) +#define PACK_SHORT_565_BE(r, g, b) (((r) & 0xF8) | ((g) >> 5) | \ + (((g) << 11) & 0xE000) | \ + (((b) << 5) & 0x1F00)) + +#define PACK_TWO_PIXELS_LE(l, r) ((r << 16) | l) +#define PACK_TWO_PIXELS_BE(l, r) ((l << 16) | r) + +#define PACK_NEED_ALIGNMENT(ptr) (((size_t)(ptr)) & 3) + +#define WRITE_TWO_ALIGNED_PIXELS(addr, pixels) ((*(int *)(addr)) = pixels) + +#define DITHER_565_R(r, dither) ((r) + ((dither) & 0xFF)) +#define DITHER_565_G(g, dither) ((g) + (((dither) & 0xFF) >> 1)) +#define DITHER_565_B(b, dither) ((b) + ((dither) & 0xFF)) + + +/* Declarations for ordered dithering + * + * We use a 4x4 ordered dither array packed into 32 bits. This array is + * sufficent for dithering RGB888 to RGB565. + */ + +#define DITHER_MASK 0x3 +#define DITHER_ROTATE(x) (((x) << 24) | (((x) >> 8) & 0x00FFFFFF)) +static const INT32 dither_matrix[4] = { + 0x0008020A, + 0x0C040E06, + 0x030B0109, + 0x0F070D05 +}; + + +static INLINE boolean is_big_endian(void) +{ + int test_value = 1; + if(*(char *)&test_value != 1) + return TRUE; + return FALSE; +} + + +/* Include inline routines for RGB565 conversion */ + +#define PACK_SHORT_565 PACK_SHORT_565_LE +#define PACK_TWO_PIXELS PACK_TWO_PIXELS_LE +#define ycc_rgb565_convert_internal ycc_rgb565_convert_le +#define ycc_rgb565D_convert_internal ycc_rgb565D_convert_le +#define rgb_rgb565_convert_internal rgb_rgb565_convert_le +#define rgb_rgb565D_convert_internal rgb_rgb565D_convert_le +#define gray_rgb565_convert_internal gray_rgb565_convert_le +#define gray_rgb565D_convert_internal gray_rgb565D_convert_le +#include "jdcol565.c" +#undef PACK_SHORT_565 +#undef PACK_TWO_PIXELS +#undef ycc_rgb565_convert_internal +#undef ycc_rgb565D_convert_internal +#undef rgb_rgb565_convert_internal +#undef rgb_rgb565D_convert_internal +#undef gray_rgb565_convert_internal +#undef gray_rgb565D_convert_internal + +#define PACK_SHORT_565 PACK_SHORT_565_BE +#define PACK_TWO_PIXELS PACK_TWO_PIXELS_BE +#define ycc_rgb565_convert_internal ycc_rgb565_convert_be +#define ycc_rgb565D_convert_internal ycc_rgb565D_convert_be +#define rgb_rgb565_convert_internal rgb_rgb565_convert_be +#define rgb_rgb565D_convert_internal rgb_rgb565D_convert_be +#define gray_rgb565_convert_internal gray_rgb565_convert_be +#define gray_rgb565D_convert_internal gray_rgb565D_convert_be +#include "jdcol565.c" +#undef PACK_SHORT_565 +#undef PACK_TWO_PIXELS +#undef ycc_rgb565_convert_internal +#undef ycc_rgb565D_convert_internal +#undef rgb_rgb565_convert_internal +#undef rgb_rgb565D_convert_internal +#undef gray_rgb565_convert_internal +#undef gray_rgb565D_convert_internal + + +METHODDEF(void) +ycc_rgb565_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + if (is_big_endian()) + ycc_rgb565_convert_be(cinfo, input_buf, input_row, output_buf, num_rows); + else + ycc_rgb565_convert_le(cinfo, input_buf, input_row, output_buf, num_rows); +} + + +METHODDEF(void) +ycc_rgb565D_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + if (is_big_endian()) + ycc_rgb565D_convert_be(cinfo, input_buf, input_row, output_buf, num_rows); + else + ycc_rgb565D_convert_le(cinfo, input_buf, input_row, output_buf, num_rows); +} + + +METHODDEF(void) +rgb_rgb565_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + if (is_big_endian()) + rgb_rgb565_convert_be(cinfo, input_buf, input_row, output_buf, num_rows); + else + rgb_rgb565_convert_le(cinfo, input_buf, input_row, output_buf, num_rows); +} + + +METHODDEF(void) +rgb_rgb565D_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + if (is_big_endian()) + rgb_rgb565D_convert_be(cinfo, input_buf, input_row, output_buf, num_rows); + else + rgb_rgb565D_convert_le(cinfo, input_buf, input_row, output_buf, num_rows); +} + + +METHODDEF(void) +gray_rgb565_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + if (is_big_endian()) + gray_rgb565_convert_be(cinfo, input_buf, input_row, output_buf, num_rows); + else + gray_rgb565_convert_le(cinfo, input_buf, input_row, output_buf, num_rows); +} + + +METHODDEF(void) +gray_rgb565D_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + if (is_big_endian()) + gray_rgb565D_convert_be(cinfo, input_buf, input_row, output_buf, num_rows); + else + gray_rgb565D_convert_le(cinfo, input_buf, input_row, output_buf, num_rows); +} + + +/* + * Empty method for start_pass. + */ + +METHODDEF(void) +start_pass_dcolor (j_decompress_ptr cinfo) +{ + /* no work needed */ +} + + +/* + * Module initialization routine for output colorspace conversion. + */ + +GLOBAL(void) +jinit_color_deconverter (j_decompress_ptr cinfo) +{ + my_cconvert_ptr cconvert; + int ci; + + cconvert = (my_cconvert_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(my_color_deconverter)); + cinfo->cconvert = (struct jpeg_color_deconverter *) cconvert; + cconvert->pub.start_pass = start_pass_dcolor; + + /* Make sure num_components agrees with jpeg_color_space */ + switch (cinfo->jpeg_color_space) { + case JCS_GRAYSCALE: + if (cinfo->num_components != 1) + ERREXIT(cinfo, JERR_BAD_J_COLORSPACE); + break; + + case JCS_RGB: + case JCS_YCbCr: + if (cinfo->num_components != 3) + ERREXIT(cinfo, JERR_BAD_J_COLORSPACE); + break; + + case JCS_CMYK: + case JCS_YCCK: + if (cinfo->num_components != 4) + ERREXIT(cinfo, JERR_BAD_J_COLORSPACE); + break; + + default: /* JCS_UNKNOWN can be anything */ + if (cinfo->num_components < 1) + ERREXIT(cinfo, JERR_BAD_J_COLORSPACE); + break; + } + + /* Set out_color_components and conversion method based on requested space. + * Also clear the component_needed flags for any unused components, + * so that earlier pipeline stages can avoid useless computation. + */ + + switch (cinfo->out_color_space) { + case JCS_GRAYSCALE: + cinfo->out_color_components = 1; + if (cinfo->jpeg_color_space == JCS_GRAYSCALE || + cinfo->jpeg_color_space == JCS_YCbCr) { + cconvert->pub.color_convert = grayscale_convert; + /* For color->grayscale conversion, only the Y (0) component is needed */ + for (ci = 1; ci < cinfo->num_components; ci++) + cinfo->comp_info[ci].component_needed = FALSE; + } else if (cinfo->jpeg_color_space == JCS_RGB) { + cconvert->pub.color_convert = rgb_gray_convert; + build_rgb_y_table(cinfo); + } else if (cinfo->jpeg_color_space == JCS_CMYK) { + cconvert->pub.color_convert = cmyk_gray_convert; + build_rgb_y_table(cinfo); + } else + ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL); + break; + + case JCS_RGB: + case JCS_EXT_RGB: + case JCS_EXT_RGBX: + case JCS_EXT_BGR: + case JCS_EXT_BGRX: + case JCS_EXT_XBGR: + case JCS_EXT_XRGB: + case JCS_EXT_RGBA: + case JCS_EXT_BGRA: + case JCS_EXT_ABGR: + case JCS_EXT_ARGB: + cinfo->out_color_components = rgb_pixelsize[cinfo->out_color_space]; + if (cinfo->jpeg_color_space == JCS_YCbCr) { + if (jsimd_can_ycc_rgb()) + cconvert->pub.color_convert = jsimd_ycc_rgb_convert; + else { + cconvert->pub.color_convert = ycc_rgb_convert; + build_ycc_rgb_table(cinfo); + } + } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) { + cconvert->pub.color_convert = gray_rgb_convert; + } else if (cinfo->jpeg_color_space == JCS_RGB) { + if (rgb_red[cinfo->out_color_space] == 0 && + rgb_green[cinfo->out_color_space] == 1 && + rgb_blue[cinfo->out_color_space] == 2 && + rgb_pixelsize[cinfo->out_color_space] == 3) + cconvert->pub.color_convert = null_convert; + else + cconvert->pub.color_convert = rgb_rgb_convert; + } else if (cinfo->jpeg_color_space == JCS_CMYK) { + cconvert->pub.color_convert = cmyk_rgb_convert; + } else + ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL); + break; + + case JCS_RGB565: + cinfo->out_color_components = 3; + if (cinfo->dither_mode == JDITHER_NONE) { + if (cinfo->jpeg_color_space == JCS_YCbCr) { + if (jsimd_can_ycc_rgb565()) + cconvert->pub.color_convert = jsimd_ycc_rgb565_convert; + else { + cconvert->pub.color_convert = ycc_rgb565_convert; + build_ycc_rgb_table(cinfo); + } + } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) { + cconvert->pub.color_convert = gray_rgb565_convert; + } else if (cinfo->jpeg_color_space == JCS_RGB) { + cconvert->pub.color_convert = rgb_rgb565_convert; + } else + ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL); + } else { + /* only ordered dithering is supported */ + if (cinfo->jpeg_color_space == JCS_YCbCr) { + cconvert->pub.color_convert = ycc_rgb565D_convert; + build_ycc_rgb_table(cinfo); + } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) { + cconvert->pub.color_convert = gray_rgb565D_convert; + } else if (cinfo->jpeg_color_space == JCS_RGB) { + cconvert->pub.color_convert = rgb_rgb565D_convert; + } else + ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL); + } + break; + + case JCS_CMYK: + cinfo->out_color_components = 4; + if (cinfo->jpeg_color_space == JCS_YCCK) { + cconvert->pub.color_convert = ycck_cmyk_convert; + build_ycc_rgb_table(cinfo); + } else if (cinfo->jpeg_color_space == JCS_CMYK) { + cconvert->pub.color_convert = null_convert; + } else + ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL); + break; + + default: + /* Permit null conversion to same output space */ + if (cinfo->out_color_space == cinfo->jpeg_color_space) { + cinfo->out_color_components = cinfo->num_components; + cconvert->pub.color_convert = null_convert; + } else /* unsupported non-null conversion */ + ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL); + break; + } + + if (cinfo->quantize_colors) + cinfo->output_components = 1; /* single colormapped output component */ + else + cinfo->output_components = cinfo->out_color_components; +} diff --git a/app/src/main/jni/libjpeg-turbo/jdct.h b/app/src/main/jni/libjpeg-turbo/jdct.h new file mode 100644 index 000000000..6f8b1591a --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdct.h @@ -0,0 +1,207 @@ +/* + * jdct.h + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1996, Thomas G. Lane. + * It was modified by The libjpeg-turbo Project to include only code relevant + * to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This include file contains common declarations for the forward and + * inverse DCT modules. These declarations are private to the DCT managers + * (jcdctmgr.c, jddctmgr.c) and the individual DCT algorithms. + * The individual DCT algorithms are kept in separate files to ease + * machine-dependent tuning (e.g., assembly coding). + */ + + +/* + * A forward DCT routine is given a pointer to a work area of type DCTELEM[]; + * the DCT is to be performed in-place in that buffer. Type DCTELEM is int + * for 8-bit samples, INT32 for 12-bit samples. (NOTE: Floating-point DCT + * implementations use an array of type FAST_FLOAT, instead.) + * The DCT inputs are expected to be signed (range +-CENTERJSAMPLE). + * The DCT outputs are returned scaled up by a factor of 8; they therefore + * have a range of +-8K for 8-bit data, +-128K for 12-bit data. This + * convention improves accuracy in integer implementations and saves some + * work in floating-point ones. + * Quantization of the output coefficients is done by jcdctmgr.c. This + * step requires an unsigned type and also one with twice the bits. + */ + +#if BITS_IN_JSAMPLE == 8 +#ifndef WITH_SIMD +typedef int DCTELEM; /* 16 or 32 bits is fine */ +typedef unsigned int UDCTELEM; +typedef unsigned long long UDCTELEM2; +#else +typedef short DCTELEM; /* prefer 16 bit with SIMD for parellelism */ +typedef unsigned short UDCTELEM; +typedef unsigned int UDCTELEM2; +#endif +#else +typedef INT32 DCTELEM; /* must have 32 bits */ +typedef unsigned long long UDCTELEM2; +#endif + + +/* + * An inverse DCT routine is given a pointer to the input JBLOCK and a pointer + * to an output sample array. The routine must dequantize the input data as + * well as perform the IDCT; for dequantization, it uses the multiplier table + * pointed to by compptr->dct_table. The output data is to be placed into the + * sample array starting at a specified column. (Any row offset needed will + * be applied to the array pointer before it is passed to the IDCT code.) + * Note that the number of samples emitted by the IDCT routine is + * DCT_scaled_size * DCT_scaled_size. + */ + +/* typedef inverse_DCT_method_ptr is declared in jpegint.h */ + +/* + * Each IDCT routine has its own ideas about the best dct_table element type. + */ + +typedef MULTIPLIER ISLOW_MULT_TYPE; /* short or int, whichever is faster */ +#if BITS_IN_JSAMPLE == 8 +typedef MULTIPLIER IFAST_MULT_TYPE; /* 16 bits is OK, use short if faster */ +#define IFAST_SCALE_BITS 2 /* fractional bits in scale factors */ +#else +typedef INT32 IFAST_MULT_TYPE; /* need 32 bits for scaled quantizers */ +#define IFAST_SCALE_BITS 13 /* fractional bits in scale factors */ +#endif +typedef FAST_FLOAT FLOAT_MULT_TYPE; /* preferred floating type */ + + +/* + * Each IDCT routine is responsible for range-limiting its results and + * converting them to unsigned form (0..MAXJSAMPLE). The raw outputs could + * be quite far out of range if the input data is corrupt, so a bulletproof + * range-limiting step is required. We use a mask-and-table-lookup method + * to do the combined operations quickly. See the comments with + * prepare_range_limit_table (in jdmaster.c) for more info. + */ + +#define IDCT_range_limit(cinfo) ((cinfo)->sample_range_limit + CENTERJSAMPLE) + +#define RANGE_MASK (MAXJSAMPLE * 4 + 3) /* 2 bits wider than legal samples */ + + +/* Extern declarations for the forward and inverse DCT routines. */ + +EXTERN(void) jpeg_fdct_islow (DCTELEM * data); +EXTERN(void) jpeg_fdct_ifast (DCTELEM * data); +EXTERN(void) jpeg_fdct_float (FAST_FLOAT * data); + +EXTERN(void) jpeg_idct_islow + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_ifast + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_float + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_7x7 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_6x6 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_5x5 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_4x4 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_3x3 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_2x2 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_1x1 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_9x9 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_10x10 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_11x11 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_12x12 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_13x13 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_14x14 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_15x15 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); +EXTERN(void) jpeg_idct_16x16 + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col); + + +/* + * Macros for handling fixed-point arithmetic; these are used by many + * but not all of the DCT/IDCT modules. + * + * All values are expected to be of type INT32. + * Fractional constants are scaled left by CONST_BITS bits. + * CONST_BITS is defined within each module using these macros, + * and may differ from one module to the next. + */ + +#define ONE ((INT32) 1) +#define CONST_SCALE (ONE << CONST_BITS) + +/* Convert a positive real constant to an integer scaled by CONST_SCALE. + * Caution: some C compilers fail to reduce "FIX(constant)" at compile time, + * thus causing a lot of useless floating-point operations at run time. + */ + +#define FIX(x) ((INT32) ((x) * CONST_SCALE + 0.5)) + +/* Descale and correctly round an INT32 value that's scaled by N bits. + * We assume RIGHT_SHIFT rounds towards minus infinity, so adding + * the fudge factor is correct for either sign of X. + */ + +#define DESCALE(x,n) RIGHT_SHIFT((x) + (ONE << ((n)-1)), n) + +/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result. + * This macro is used only when the two inputs will actually be no more than + * 16 bits wide, so that a 16x16->32 bit multiply can be used instead of a + * full 32x32 multiply. This provides a useful speedup on many machines. + * Unfortunately there is no way to specify a 16x16->32 multiply portably + * in C, but some C compilers will do the right thing if you provide the + * correct combination of casts. + */ + +#ifdef SHORTxSHORT_32 /* may work if 'int' is 32 bits */ +#define MULTIPLY16C16(var,const) (((INT16) (var)) * ((INT16) (const))) +#endif +#ifdef SHORTxLCONST_32 /* known to work with Microsoft C 6.0 */ +#define MULTIPLY16C16(var,const) (((INT16) (var)) * ((INT32) (const))) +#endif + +#ifndef MULTIPLY16C16 /* default definition */ +#define MULTIPLY16C16(var,const) ((var) * (const)) +#endif + +/* Same except both inputs are variables. */ + +#ifdef SHORTxSHORT_32 /* may work if 'int' is 32 bits */ +#define MULTIPLY16V16(var1,var2) (((INT16) (var1)) * ((INT16) (var2))) +#endif + +#ifndef MULTIPLY16V16 /* default definition */ +#define MULTIPLY16V16(var1,var2) ((var1) * (var2)) +#endif diff --git a/app/src/main/jni/libjpeg-turbo/jddctmgr.c b/app/src/main/jni/libjpeg-turbo/jddctmgr.c new file mode 100644 index 000000000..40e68f1f2 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jddctmgr.c @@ -0,0 +1,349 @@ +/* + * jddctmgr.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1996, Thomas G. Lane. + * Modified 2002-2010 by Guido Vollbeding. + * libjpeg-turbo Modifications: + * Copyright 2009 Pierre Ossman for Cendio AB + * Copyright (C) 2010, D. R. Commander. + * Copyright (C) 2013, MIPS Technologies, Inc., California + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains the inverse-DCT management logic. + * This code selects a particular IDCT implementation to be used, + * and it performs related housekeeping chores. No code in this file + * is executed per IDCT step, only during output pass setup. + * + * Note that the IDCT routines are responsible for performing coefficient + * dequantization as well as the IDCT proper. This module sets up the + * dequantization multiplier table needed by the IDCT routine. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jdct.h" /* Private declarations for DCT subsystem */ +#include "jsimddct.h" +#include "jpegcomp.h" + + +/* + * The decompressor input side (jdinput.c) saves away the appropriate + * quantization table for each component at the start of the first scan + * involving that component. (This is necessary in order to correctly + * decode files that reuse Q-table slots.) + * When we are ready to make an output pass, the saved Q-table is converted + * to a multiplier table that will actually be used by the IDCT routine. + * The multiplier table contents are IDCT-method-dependent. To support + * application changes in IDCT method between scans, we can remake the + * multiplier tables if necessary. + * In buffered-image mode, the first output pass may occur before any data + * has been seen for some components, and thus before their Q-tables have + * been saved away. To handle this case, multiplier tables are preset + * to zeroes; the result of the IDCT will be a neutral gray level. + */ + + +/* Private subobject for this module */ + +typedef struct { + struct jpeg_inverse_dct pub; /* public fields */ + + /* This array contains the IDCT method code that each multiplier table + * is currently set up for, or -1 if it's not yet set up. + * The actual multiplier tables are pointed to by dct_table in the + * per-component comp_info structures. + */ + int cur_method[MAX_COMPONENTS]; +} my_idct_controller; + +typedef my_idct_controller * my_idct_ptr; + + +/* Allocated multiplier tables: big enough for any supported variant */ + +typedef union { + ISLOW_MULT_TYPE islow_array[DCTSIZE2]; +#ifdef DCT_IFAST_SUPPORTED + IFAST_MULT_TYPE ifast_array[DCTSIZE2]; +#endif +#ifdef DCT_FLOAT_SUPPORTED + FLOAT_MULT_TYPE float_array[DCTSIZE2]; +#endif +} multiplier_table; + + +/* The current scaled-IDCT routines require ISLOW-style multiplier tables, + * so be sure to compile that code if either ISLOW or SCALING is requested. + */ +#ifdef DCT_ISLOW_SUPPORTED +#define PROVIDE_ISLOW_TABLES +#else +#ifdef IDCT_SCALING_SUPPORTED +#define PROVIDE_ISLOW_TABLES +#endif +#endif + + +/* + * Prepare for an output pass. + * Here we select the proper IDCT routine for each component and build + * a matching multiplier table. + */ + +METHODDEF(void) +start_pass (j_decompress_ptr cinfo) +{ + my_idct_ptr idct = (my_idct_ptr) cinfo->idct; + int ci, i; + jpeg_component_info *compptr; + int method = 0; + inverse_DCT_method_ptr method_ptr = NULL; + JQUANT_TBL * qtbl; + + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + /* Select the proper IDCT routine for this component's scaling */ + switch (compptr->_DCT_scaled_size) { +#ifdef IDCT_SCALING_SUPPORTED + case 1: + method_ptr = jpeg_idct_1x1; + method = JDCT_ISLOW; /* jidctred uses islow-style table */ + break; + case 2: + if (jsimd_can_idct_2x2()) + method_ptr = jsimd_idct_2x2; + else + method_ptr = jpeg_idct_2x2; + method = JDCT_ISLOW; /* jidctred uses islow-style table */ + break; + case 3: + method_ptr = jpeg_idct_3x3; + method = JDCT_ISLOW; /* jidctint uses islow-style table */ + break; + case 4: + if (jsimd_can_idct_4x4()) + method_ptr = jsimd_idct_4x4; + else + method_ptr = jpeg_idct_4x4; + method = JDCT_ISLOW; /* jidctred uses islow-style table */ + break; + case 5: + method_ptr = jpeg_idct_5x5; + method = JDCT_ISLOW; /* jidctint uses islow-style table */ + break; + case 6: +#if defined(__mips__) + if (jsimd_can_idct_6x6()) + method_ptr = jsimd_idct_6x6; + else +#endif + method_ptr = jpeg_idct_6x6; + method = JDCT_ISLOW; /* jidctint uses islow-style table */ + break; + case 7: + method_ptr = jpeg_idct_7x7; + method = JDCT_ISLOW; /* jidctint uses islow-style table */ + break; +#endif + case DCTSIZE: + switch (cinfo->dct_method) { +#ifdef DCT_ISLOW_SUPPORTED + case JDCT_ISLOW: + if (jsimd_can_idct_islow()) + method_ptr = jsimd_idct_islow; + else + method_ptr = jpeg_idct_islow; + method = JDCT_ISLOW; + break; +#endif +#ifdef DCT_IFAST_SUPPORTED + case JDCT_IFAST: + if (jsimd_can_idct_ifast()) + method_ptr = jsimd_idct_ifast; + else + method_ptr = jpeg_idct_ifast; + method = JDCT_IFAST; + break; +#endif +#ifdef DCT_FLOAT_SUPPORTED + case JDCT_FLOAT: + if (jsimd_can_idct_float()) + method_ptr = jsimd_idct_float; + else + method_ptr = jpeg_idct_float; + method = JDCT_FLOAT; + break; +#endif + default: + ERREXIT(cinfo, JERR_NOT_COMPILED); + break; + } + break; + case 9: + method_ptr = jpeg_idct_9x9; + method = JDCT_ISLOW; /* jidctint uses islow-style table */ + break; + case 10: + method_ptr = jpeg_idct_10x10; + method = JDCT_ISLOW; /* jidctint uses islow-style table */ + break; + case 11: + method_ptr = jpeg_idct_11x11; + method = JDCT_ISLOW; /* jidctint uses islow-style table */ + break; + case 12: +#if defined(__mips__) + if (jsimd_can_idct_12x12()) + method_ptr = jsimd_idct_12x12; + else +#endif + method_ptr = jpeg_idct_12x12; + method = JDCT_ISLOW; /* jidctint uses islow-style table */ + break; + case 13: + method_ptr = jpeg_idct_13x13; + method = JDCT_ISLOW; /* jidctint uses islow-style table */ + break; + case 14: + method_ptr = jpeg_idct_14x14; + method = JDCT_ISLOW; /* jidctint uses islow-style table */ + break; + case 15: + method_ptr = jpeg_idct_15x15; + method = JDCT_ISLOW; /* jidctint uses islow-style table */ + break; + case 16: + method_ptr = jpeg_idct_16x16; + method = JDCT_ISLOW; /* jidctint uses islow-style table */ + break; + default: + ERREXIT1(cinfo, JERR_BAD_DCTSIZE, compptr->_DCT_scaled_size); + break; + } + idct->pub.inverse_DCT[ci] = method_ptr; + /* Create multiplier table from quant table. + * However, we can skip this if the component is uninteresting + * or if we already built the table. Also, if no quant table + * has yet been saved for the component, we leave the + * multiplier table all-zero; we'll be reading zeroes from the + * coefficient controller's buffer anyway. + */ + if (! compptr->component_needed || idct->cur_method[ci] == method) + continue; + qtbl = compptr->quant_table; + if (qtbl == NULL) /* happens if no data yet for component */ + continue; + idct->cur_method[ci] = method; + switch (method) { +#ifdef PROVIDE_ISLOW_TABLES + case JDCT_ISLOW: + { + /* For LL&M IDCT method, multipliers are equal to raw quantization + * coefficients, but are stored as ints to ensure access efficiency. + */ + ISLOW_MULT_TYPE * ismtbl = (ISLOW_MULT_TYPE *) compptr->dct_table; + for (i = 0; i < DCTSIZE2; i++) { + ismtbl[i] = (ISLOW_MULT_TYPE) qtbl->quantval[i]; + } + } + break; +#endif +#ifdef DCT_IFAST_SUPPORTED + case JDCT_IFAST: + { + /* For AA&N IDCT method, multipliers are equal to quantization + * coefficients scaled by scalefactor[row]*scalefactor[col], where + * scalefactor[0] = 1 + * scalefactor[k] = cos(k*PI/16) * sqrt(2) for k=1..7 + * For integer operation, the multiplier table is to be scaled by + * IFAST_SCALE_BITS. + */ + IFAST_MULT_TYPE * ifmtbl = (IFAST_MULT_TYPE *) compptr->dct_table; +#define CONST_BITS 14 + static const INT16 aanscales[DCTSIZE2] = { + /* precomputed values scaled up by 14 bits */ + 16384, 22725, 21407, 19266, 16384, 12873, 8867, 4520, + 22725, 31521, 29692, 26722, 22725, 17855, 12299, 6270, + 21407, 29692, 27969, 25172, 21407, 16819, 11585, 5906, + 19266, 26722, 25172, 22654, 19266, 15137, 10426, 5315, + 16384, 22725, 21407, 19266, 16384, 12873, 8867, 4520, + 12873, 17855, 16819, 15137, 12873, 10114, 6967, 3552, + 8867, 12299, 11585, 10426, 8867, 6967, 4799, 2446, + 4520, 6270, 5906, 5315, 4520, 3552, 2446, 1247 + }; + SHIFT_TEMPS + + for (i = 0; i < DCTSIZE2; i++) { + ifmtbl[i] = (IFAST_MULT_TYPE) + DESCALE(MULTIPLY16V16((INT32) qtbl->quantval[i], + (INT32) aanscales[i]), + CONST_BITS-IFAST_SCALE_BITS); + } + } + break; +#endif +#ifdef DCT_FLOAT_SUPPORTED + case JDCT_FLOAT: + { + /* For float AA&N IDCT method, multipliers are equal to quantization + * coefficients scaled by scalefactor[row]*scalefactor[col], where + * scalefactor[0] = 1 + * scalefactor[k] = cos(k*PI/16) * sqrt(2) for k=1..7 + */ + FLOAT_MULT_TYPE * fmtbl = (FLOAT_MULT_TYPE *) compptr->dct_table; + int row, col; + static const double aanscalefactor[DCTSIZE] = { + 1.0, 1.387039845, 1.306562965, 1.175875602, + 1.0, 0.785694958, 0.541196100, 0.275899379 + }; + + i = 0; + for (row = 0; row < DCTSIZE; row++) { + for (col = 0; col < DCTSIZE; col++) { + fmtbl[i] = (FLOAT_MULT_TYPE) + ((double) qtbl->quantval[i] * + aanscalefactor[row] * aanscalefactor[col]); + i++; + } + } + } + break; +#endif + default: + ERREXIT(cinfo, JERR_NOT_COMPILED); + break; + } + } +} + + +/* + * Initialize IDCT manager. + */ + +GLOBAL(void) +jinit_inverse_dct (j_decompress_ptr cinfo) +{ + my_idct_ptr idct; + int ci; + jpeg_component_info *compptr; + + idct = (my_idct_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(my_idct_controller)); + cinfo->idct = (struct jpeg_inverse_dct *) idct; + idct->pub.start_pass = start_pass; + + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + /* Allocate and pre-zero a multiplier table for each component */ + compptr->dct_table = + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(multiplier_table)); + MEMZERO(compptr->dct_table, sizeof(multiplier_table)); + /* Mark multiplier table not yet set up for any method */ + idct->cur_method[ci] = -1; + } +} diff --git a/app/src/main/jni/libjpeg-turbo/jdhuff.c b/app/src/main/jni/libjpeg-turbo/jdhuff.c new file mode 100644 index 000000000..7dc132896 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdhuff.c @@ -0,0 +1,816 @@ +/* + * jdhuff.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1997, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright (C) 2009-2011, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains Huffman entropy decoding routines. + * + * Much of the complexity here has to do with supporting input suspension. + * If the data source module demands suspension, we want to be able to back + * up to the start of the current MCU. To do this, we copy state variables + * into local working storage, and update them back to the permanent + * storage only upon successful completion of an MCU. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jdhuff.h" /* Declarations shared with jdphuff.c */ +#include "jpegcomp.h" +#include "jstdhuff.c" + + +/* + * Expanded entropy decoder object for Huffman decoding. + * + * The savable_state subrecord contains fields that change within an MCU, + * but must not be updated permanently until we complete the MCU. + */ + +typedef struct { + int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */ +} savable_state; + +/* This macro is to work around compilers with missing or broken + * structure assignment. You'll need to fix this code if you have + * such a compiler and you change MAX_COMPS_IN_SCAN. + */ + +#ifndef NO_STRUCT_ASSIGN +#define ASSIGN_STATE(dest,src) ((dest) = (src)) +#else +#if MAX_COMPS_IN_SCAN == 4 +#define ASSIGN_STATE(dest,src) \ + ((dest).last_dc_val[0] = (src).last_dc_val[0], \ + (dest).last_dc_val[1] = (src).last_dc_val[1], \ + (dest).last_dc_val[2] = (src).last_dc_val[2], \ + (dest).last_dc_val[3] = (src).last_dc_val[3]) +#endif +#endif + + +typedef struct { + struct jpeg_entropy_decoder pub; /* public fields */ + + /* These fields are loaded into local variables at start of each MCU. + * In case of suspension, we exit WITHOUT updating them. + */ + bitread_perm_state bitstate; /* Bit buffer at start of MCU */ + savable_state saved; /* Other state at start of MCU */ + + /* These fields are NOT loaded into local working state. */ + unsigned int restarts_to_go; /* MCUs left in this restart interval */ + + /* Pointers to derived tables (these workspaces have image lifespan) */ + d_derived_tbl * dc_derived_tbls[NUM_HUFF_TBLS]; + d_derived_tbl * ac_derived_tbls[NUM_HUFF_TBLS]; + + /* Precalculated info set up by start_pass for use in decode_mcu: */ + + /* Pointers to derived tables to be used for each block within an MCU */ + d_derived_tbl * dc_cur_tbls[D_MAX_BLOCKS_IN_MCU]; + d_derived_tbl * ac_cur_tbls[D_MAX_BLOCKS_IN_MCU]; + /* Whether we care about the DC and AC coefficient values for each block */ + boolean dc_needed[D_MAX_BLOCKS_IN_MCU]; + boolean ac_needed[D_MAX_BLOCKS_IN_MCU]; +} huff_entropy_decoder; + +typedef huff_entropy_decoder * huff_entropy_ptr; + + +/* + * Initialize for a Huffman-compressed scan. + */ + +METHODDEF(void) +start_pass_huff_decoder (j_decompress_ptr cinfo) +{ + huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy; + int ci, blkn, dctbl, actbl; + jpeg_component_info * compptr; + + /* Check that the scan parameters Ss, Se, Ah/Al are OK for sequential JPEG. + * This ought to be an error condition, but we make it a warning because + * there are some baseline files out there with all zeroes in these bytes. + */ + if (cinfo->Ss != 0 || cinfo->Se != DCTSIZE2-1 || + cinfo->Ah != 0 || cinfo->Al != 0) + WARNMS(cinfo, JWRN_NOT_SEQUENTIAL); + + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + dctbl = compptr->dc_tbl_no; + actbl = compptr->ac_tbl_no; + /* Compute derived values for Huffman tables */ + /* We may do this more than once for a table, but it's not expensive */ + jpeg_make_d_derived_tbl(cinfo, TRUE, dctbl, + & entropy->dc_derived_tbls[dctbl]); + jpeg_make_d_derived_tbl(cinfo, FALSE, actbl, + & entropy->ac_derived_tbls[actbl]); + /* Initialize DC predictions to 0 */ + entropy->saved.last_dc_val[ci] = 0; + } + + /* Precalculate decoding info for each block in an MCU of this scan */ + for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) { + ci = cinfo->MCU_membership[blkn]; + compptr = cinfo->cur_comp_info[ci]; + /* Precalculate which table to use for each block */ + entropy->dc_cur_tbls[blkn] = entropy->dc_derived_tbls[compptr->dc_tbl_no]; + entropy->ac_cur_tbls[blkn] = entropy->ac_derived_tbls[compptr->ac_tbl_no]; + /* Decide whether we really care about the coefficient values */ + if (compptr->component_needed) { + entropy->dc_needed[blkn] = TRUE; + /* we don't need the ACs if producing a 1/8th-size image */ + entropy->ac_needed[blkn] = (compptr->_DCT_scaled_size > 1); + } else { + entropy->dc_needed[blkn] = entropy->ac_needed[blkn] = FALSE; + } + } + + /* Initialize bitread state variables */ + entropy->bitstate.bits_left = 0; + entropy->bitstate.get_buffer = 0; /* unnecessary, but keeps Purify quiet */ + entropy->pub.insufficient_data = FALSE; + + /* Initialize restart counter */ + entropy->restarts_to_go = cinfo->restart_interval; +} + + +/* + * Compute the derived values for a Huffman table. + * This routine also performs some validation checks on the table. + * + * Note this is also used by jdphuff.c. + */ + +GLOBAL(void) +jpeg_make_d_derived_tbl (j_decompress_ptr cinfo, boolean isDC, int tblno, + d_derived_tbl ** pdtbl) +{ + JHUFF_TBL *htbl; + d_derived_tbl *dtbl; + int p, i, l, si, numsymbols; + int lookbits, ctr; + char huffsize[257]; + unsigned int huffcode[257]; + unsigned int code; + + /* Note that huffsize[] and huffcode[] are filled in code-length order, + * paralleling the order of the symbols themselves in htbl->huffval[]. + */ + + /* Find the input Huffman table */ + if (tblno < 0 || tblno >= NUM_HUFF_TBLS) + ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno); + htbl = + isDC ? cinfo->dc_huff_tbl_ptrs[tblno] : cinfo->ac_huff_tbl_ptrs[tblno]; + if (htbl == NULL) + ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno); + + /* Allocate a workspace if we haven't already done so. */ + if (*pdtbl == NULL) + *pdtbl = (d_derived_tbl *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(d_derived_tbl)); + dtbl = *pdtbl; + dtbl->pub = htbl; /* fill in back link */ + + /* Figure C.1: make table of Huffman code length for each symbol */ + + p = 0; + for (l = 1; l <= 16; l++) { + i = (int) htbl->bits[l]; + if (i < 0 || p + i > 256) /* protect against table overrun */ + ERREXIT(cinfo, JERR_BAD_HUFF_TABLE); + while (i--) + huffsize[p++] = (char) l; + } + huffsize[p] = 0; + numsymbols = p; + + /* Figure C.2: generate the codes themselves */ + /* We also validate that the counts represent a legal Huffman code tree. */ + + code = 0; + si = huffsize[0]; + p = 0; + while (huffsize[p]) { + while (((int) huffsize[p]) == si) { + huffcode[p++] = code; + code++; + } + /* code is now 1 more than the last code used for codelength si; but + * it must still fit in si bits, since no code is allowed to be all ones. + */ + if (((INT32) code) >= (((INT32) 1) << si)) + ERREXIT(cinfo, JERR_BAD_HUFF_TABLE); + code <<= 1; + si++; + } + + /* Figure F.15: generate decoding tables for bit-sequential decoding */ + + p = 0; + for (l = 1; l <= 16; l++) { + if (htbl->bits[l]) { + /* valoffset[l] = huffval[] index of 1st symbol of code length l, + * minus the minimum code of length l + */ + dtbl->valoffset[l] = (INT32) p - (INT32) huffcode[p]; + p += htbl->bits[l]; + dtbl->maxcode[l] = huffcode[p-1]; /* maximum code of length l */ + } else { + dtbl->maxcode[l] = -1; /* -1 if no codes of this length */ + } + } + dtbl->valoffset[17] = 0; + dtbl->maxcode[17] = 0xFFFFFL; /* ensures jpeg_huff_decode terminates */ + + /* Compute lookahead tables to speed up decoding. + * First we set all the table entries to 0, indicating "too long"; + * then we iterate through the Huffman codes that are short enough and + * fill in all the entries that correspond to bit sequences starting + * with that code. + */ + + for (i = 0; i < (1 << HUFF_LOOKAHEAD); i++) + dtbl->lookup[i] = (HUFF_LOOKAHEAD + 1) << HUFF_LOOKAHEAD; + + p = 0; + for (l = 1; l <= HUFF_LOOKAHEAD; l++) { + for (i = 1; i <= (int) htbl->bits[l]; i++, p++) { + /* l = current code's length, p = its index in huffcode[] & huffval[]. */ + /* Generate left-justified code followed by all possible bit sequences */ + lookbits = huffcode[p] << (HUFF_LOOKAHEAD-l); + for (ctr = 1 << (HUFF_LOOKAHEAD-l); ctr > 0; ctr--) { + dtbl->lookup[lookbits] = (l << HUFF_LOOKAHEAD) | htbl->huffval[p]; + lookbits++; + } + } + } + + /* Validate symbols as being reasonable. + * For AC tables, we make no check, but accept all byte values 0..255. + * For DC tables, we require the symbols to be in range 0..15. + * (Tighter bounds could be applied depending on the data depth and mode, + * but this is sufficient to ensure safe decoding.) + */ + if (isDC) { + for (i = 0; i < numsymbols; i++) { + int sym = htbl->huffval[i]; + if (sym < 0 || sym > 15) + ERREXIT(cinfo, JERR_BAD_HUFF_TABLE); + } + } +} + + +/* + * Out-of-line code for bit fetching (shared with jdphuff.c). + * See jdhuff.h for info about usage. + * Note: current values of get_buffer and bits_left are passed as parameters, + * but are returned in the corresponding fields of the state struct. + * + * On most machines MIN_GET_BITS should be 25 to allow the full 32-bit width + * of get_buffer to be used. (On machines with wider words, an even larger + * buffer could be used.) However, on some machines 32-bit shifts are + * quite slow and take time proportional to the number of places shifted. + * (This is true with most PC compilers, for instance.) In this case it may + * be a win to set MIN_GET_BITS to the minimum value of 15. This reduces the + * average shift distance at the cost of more calls to jpeg_fill_bit_buffer. + */ + +#ifdef SLOW_SHIFT_32 +#define MIN_GET_BITS 15 /* minimum allowable value */ +#else +#define MIN_GET_BITS (BIT_BUF_SIZE-7) +#endif + + +GLOBAL(boolean) +jpeg_fill_bit_buffer (bitread_working_state * state, + register bit_buf_type get_buffer, register int bits_left, + int nbits) +/* Load up the bit buffer to a depth of at least nbits */ +{ + /* Copy heavily used state fields into locals (hopefully registers) */ + register const JOCTET * next_input_byte = state->next_input_byte; + register size_t bytes_in_buffer = state->bytes_in_buffer; + j_decompress_ptr cinfo = state->cinfo; + + /* Attempt to load at least MIN_GET_BITS bits into get_buffer. */ + /* (It is assumed that no request will be for more than that many bits.) */ + /* We fail to do so only if we hit a marker or are forced to suspend. */ + + if (cinfo->unread_marker == 0) { /* cannot advance past a marker */ + while (bits_left < MIN_GET_BITS) { + register int c; + + /* Attempt to read a byte */ + if (bytes_in_buffer == 0) { + if (! (*cinfo->src->fill_input_buffer) (cinfo)) + return FALSE; + next_input_byte = cinfo->src->next_input_byte; + bytes_in_buffer = cinfo->src->bytes_in_buffer; + } + bytes_in_buffer--; + c = GETJOCTET(*next_input_byte++); + + /* If it's 0xFF, check and discard stuffed zero byte */ + if (c == 0xFF) { + /* Loop here to discard any padding FF's on terminating marker, + * so that we can save a valid unread_marker value. NOTE: we will + * accept multiple FF's followed by a 0 as meaning a single FF data + * byte. This data pattern is not valid according to the standard. + */ + do { + if (bytes_in_buffer == 0) { + if (! (*cinfo->src->fill_input_buffer) (cinfo)) + return FALSE; + next_input_byte = cinfo->src->next_input_byte; + bytes_in_buffer = cinfo->src->bytes_in_buffer; + } + bytes_in_buffer--; + c = GETJOCTET(*next_input_byte++); + } while (c == 0xFF); + + if (c == 0) { + /* Found FF/00, which represents an FF data byte */ + c = 0xFF; + } else { + /* Oops, it's actually a marker indicating end of compressed data. + * Save the marker code for later use. + * Fine point: it might appear that we should save the marker into + * bitread working state, not straight into permanent state. But + * once we have hit a marker, we cannot need to suspend within the + * current MCU, because we will read no more bytes from the data + * source. So it is OK to update permanent state right away. + */ + cinfo->unread_marker = c; + /* See if we need to insert some fake zero bits. */ + goto no_more_bytes; + } + } + + /* OK, load c into get_buffer */ + get_buffer = (get_buffer << 8) | c; + bits_left += 8; + } /* end while */ + } else { + no_more_bytes: + /* We get here if we've read the marker that terminates the compressed + * data segment. There should be enough bits in the buffer register + * to satisfy the request; if so, no problem. + */ + if (nbits > bits_left) { + /* Uh-oh. Report corrupted data to user and stuff zeroes into + * the data stream, so that we can produce some kind of image. + * We use a nonvolatile flag to ensure that only one warning message + * appears per data segment. + */ + if (! cinfo->entropy->insufficient_data) { + WARNMS(cinfo, JWRN_HIT_MARKER); + cinfo->entropy->insufficient_data = TRUE; + } + /* Fill the buffer with zero bits */ + get_buffer <<= MIN_GET_BITS - bits_left; + bits_left = MIN_GET_BITS; + } + } + + /* Unload the local registers */ + state->next_input_byte = next_input_byte; + state->bytes_in_buffer = bytes_in_buffer; + state->get_buffer = get_buffer; + state->bits_left = bits_left; + + return TRUE; +} + + +/* Macro version of the above, which performs much better but does not + handle markers. We have to hand off any blocks with markers to the + slower routines. */ + +#define GET_BYTE \ +{ \ + register int c0, c1; \ + c0 = GETJOCTET(*buffer++); \ + c1 = GETJOCTET(*buffer); \ + /* Pre-execute most common case */ \ + get_buffer = (get_buffer << 8) | c0; \ + bits_left += 8; \ + if (c0 == 0xFF) { \ + /* Pre-execute case of FF/00, which represents an FF data byte */ \ + buffer++; \ + if (c1 != 0) { \ + /* Oops, it's actually a marker indicating end of compressed data. */ \ + cinfo->unread_marker = c1; \ + /* Back out pre-execution and fill the buffer with zero bits */ \ + buffer -= 2; \ + get_buffer &= ~0xFF; \ + } \ + } \ +} + +#if __WORDSIZE == 64 || defined(_WIN64) + +/* Pre-fetch 48 bytes, because the holding register is 64-bit */ +#define FILL_BIT_BUFFER_FAST \ + if (bits_left < 16) { \ + GET_BYTE GET_BYTE GET_BYTE GET_BYTE GET_BYTE GET_BYTE \ + } + +#else + +/* Pre-fetch 16 bytes, because the holding register is 32-bit */ +#define FILL_BIT_BUFFER_FAST \ + if (bits_left < 16) { \ + GET_BYTE GET_BYTE \ + } + +#endif + + +/* + * Out-of-line code for Huffman code decoding. + * See jdhuff.h for info about usage. + */ + +GLOBAL(int) +jpeg_huff_decode (bitread_working_state * state, + register bit_buf_type get_buffer, register int bits_left, + d_derived_tbl * htbl, int min_bits) +{ + register int l = min_bits; + register INT32 code; + + /* HUFF_DECODE has determined that the code is at least min_bits */ + /* bits long, so fetch that many bits in one swoop. */ + + CHECK_BIT_BUFFER(*state, l, return -1); + code = GET_BITS(l); + + /* Collect the rest of the Huffman code one bit at a time. */ + /* This is per Figure F.16 in the JPEG spec. */ + + while (code > htbl->maxcode[l]) { + code <<= 1; + CHECK_BIT_BUFFER(*state, 1, return -1); + code |= GET_BITS(1); + l++; + } + + /* Unload the local registers */ + state->get_buffer = get_buffer; + state->bits_left = bits_left; + + /* With garbage input we may reach the sentinel value l = 17. */ + + if (l > 16) { + WARNMS(state->cinfo, JWRN_HUFF_BAD_CODE); + return 0; /* fake a zero as the safest result */ + } + + return htbl->pub->huffval[ (int) (code + htbl->valoffset[l]) ]; +} + + +/* + * Figure F.12: extend sign bit. + * On some machines, a shift and add will be faster than a table lookup. + */ + +#define AVOID_TABLES +#ifdef AVOID_TABLES + +#define HUFF_EXTEND(x,s) ((x) + ((((x) - (1<<((s)-1))) >> 31) & (((-1)<<(s)) + 1))) + +#else + +#define HUFF_EXTEND(x,s) ((x) < extend_test[s] ? (x) + extend_offset[s] : (x)) + +static const int extend_test[16] = /* entry n is 2**(n-1) */ + { 0, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080, + 0x0100, 0x0200, 0x0400, 0x0800, 0x1000, 0x2000, 0x4000 }; + +static const int extend_offset[16] = /* entry n is (-1 << n) + 1 */ + { 0, ((-1)<<1) + 1, ((-1)<<2) + 1, ((-1)<<3) + 1, ((-1)<<4) + 1, + ((-1)<<5) + 1, ((-1)<<6) + 1, ((-1)<<7) + 1, ((-1)<<8) + 1, + ((-1)<<9) + 1, ((-1)<<10) + 1, ((-1)<<11) + 1, ((-1)<<12) + 1, + ((-1)<<13) + 1, ((-1)<<14) + 1, ((-1)<<15) + 1 }; + +#endif /* AVOID_TABLES */ + + +/* + * Check for a restart marker & resynchronize decoder. + * Returns FALSE if must suspend. + */ + +LOCAL(boolean) +process_restart (j_decompress_ptr cinfo) +{ + huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy; + int ci; + + /* Throw away any unused bits remaining in bit buffer; */ + /* include any full bytes in next_marker's count of discarded bytes */ + cinfo->marker->discarded_bytes += entropy->bitstate.bits_left / 8; + entropy->bitstate.bits_left = 0; + + /* Advance past the RSTn marker */ + if (! (*cinfo->marker->read_restart_marker) (cinfo)) + return FALSE; + + /* Re-initialize DC predictions to 0 */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) + entropy->saved.last_dc_val[ci] = 0; + + /* Reset restart counter */ + entropy->restarts_to_go = cinfo->restart_interval; + + /* Reset out-of-data flag, unless read_restart_marker left us smack up + * against a marker. In that case we will end up treating the next data + * segment as empty, and we can avoid producing bogus output pixels by + * leaving the flag set. + */ + if (cinfo->unread_marker == 0) + entropy->pub.insufficient_data = FALSE; + + return TRUE; +} + + +LOCAL(boolean) +decode_mcu_slow (j_decompress_ptr cinfo, JBLOCKROW *MCU_data) +{ + huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy; + BITREAD_STATE_VARS; + int blkn; + savable_state state; + /* Outer loop handles each block in the MCU */ + + /* Load up working state */ + BITREAD_LOAD_STATE(cinfo,entropy->bitstate); + ASSIGN_STATE(state, entropy->saved); + + for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) { + JBLOCKROW block = MCU_data[blkn]; + d_derived_tbl * dctbl = entropy->dc_cur_tbls[blkn]; + d_derived_tbl * actbl = entropy->ac_cur_tbls[blkn]; + register int s, k, r; + + /* Decode a single block's worth of coefficients */ + + /* Section F.2.2.1: decode the DC coefficient difference */ + HUFF_DECODE(s, br_state, dctbl, return FALSE, label1); + if (s) { + CHECK_BIT_BUFFER(br_state, s, return FALSE); + r = GET_BITS(s); + s = HUFF_EXTEND(r, s); + } + + if (entropy->dc_needed[blkn]) { + /* Convert DC difference to actual value, update last_dc_val */ + int ci = cinfo->MCU_membership[blkn]; + s += state.last_dc_val[ci]; + state.last_dc_val[ci] = s; + /* Output the DC coefficient (assumes jpeg_natural_order[0] = 0) */ + (*block)[0] = (JCOEF) s; + } + + if (entropy->ac_needed[blkn]) { + + /* Section F.2.2.2: decode the AC coefficients */ + /* Since zeroes are skipped, output area must be cleared beforehand */ + for (k = 1; k < DCTSIZE2; k++) { + HUFF_DECODE(s, br_state, actbl, return FALSE, label2); + + r = s >> 4; + s &= 15; + + if (s) { + k += r; + CHECK_BIT_BUFFER(br_state, s, return FALSE); + r = GET_BITS(s); + s = HUFF_EXTEND(r, s); + /* Output coefficient in natural (dezigzagged) order. + * Note: the extra entries in jpeg_natural_order[] will save us + * if k >= DCTSIZE2, which could happen if the data is corrupted. + */ + (*block)[jpeg_natural_order[k]] = (JCOEF) s; + } else { + if (r != 15) + break; + k += 15; + } + } + + } else { + + /* Section F.2.2.2: decode the AC coefficients */ + /* In this path we just discard the values */ + for (k = 1; k < DCTSIZE2; k++) { + HUFF_DECODE(s, br_state, actbl, return FALSE, label3); + + r = s >> 4; + s &= 15; + + if (s) { + k += r; + CHECK_BIT_BUFFER(br_state, s, return FALSE); + DROP_BITS(s); + } else { + if (r != 15) + break; + k += 15; + } + } + } + } + + /* Completed MCU, so update state */ + BITREAD_SAVE_STATE(cinfo,entropy->bitstate); + ASSIGN_STATE(entropy->saved, state); + return TRUE; +} + + +LOCAL(boolean) +decode_mcu_fast (j_decompress_ptr cinfo, JBLOCKROW *MCU_data) +{ + huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy; + BITREAD_STATE_VARS; + JOCTET *buffer; + int blkn; + savable_state state; + /* Outer loop handles each block in the MCU */ + + /* Load up working state */ + BITREAD_LOAD_STATE(cinfo,entropy->bitstate); + buffer = (JOCTET *) br_state.next_input_byte; + ASSIGN_STATE(state, entropy->saved); + + for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) { + JBLOCKROW block = MCU_data[blkn]; + d_derived_tbl * dctbl = entropy->dc_cur_tbls[blkn]; + d_derived_tbl * actbl = entropy->ac_cur_tbls[blkn]; + register int s, k, r, l; + + HUFF_DECODE_FAST(s, l, dctbl); + if (s) { + FILL_BIT_BUFFER_FAST + r = GET_BITS(s); + s = HUFF_EXTEND(r, s); + } + + if (entropy->dc_needed[blkn]) { + int ci = cinfo->MCU_membership[blkn]; + s += state.last_dc_val[ci]; + state.last_dc_val[ci] = s; + (*block)[0] = (JCOEF) s; + } + + if (entropy->ac_needed[blkn]) { + + for (k = 1; k < DCTSIZE2; k++) { + HUFF_DECODE_FAST(s, l, actbl); + r = s >> 4; + s &= 15; + + if (s) { + k += r; + FILL_BIT_BUFFER_FAST + r = GET_BITS(s); + s = HUFF_EXTEND(r, s); + (*block)[jpeg_natural_order[k]] = (JCOEF) s; + } else { + if (r != 15) break; + k += 15; + } + } + + } else { + + for (k = 1; k < DCTSIZE2; k++) { + HUFF_DECODE_FAST(s, l, actbl); + r = s >> 4; + s &= 15; + + if (s) { + k += r; + FILL_BIT_BUFFER_FAST + DROP_BITS(s); + } else { + if (r != 15) break; + k += 15; + } + } + } + } + + if (cinfo->unread_marker != 0) { + cinfo->unread_marker = 0; + return FALSE; + } + + br_state.bytes_in_buffer -= (buffer - br_state.next_input_byte); + br_state.next_input_byte = buffer; + BITREAD_SAVE_STATE(cinfo,entropy->bitstate); + ASSIGN_STATE(entropy->saved, state); + return TRUE; +} + + +/* + * Decode and return one MCU's worth of Huffman-compressed coefficients. + * The coefficients are reordered from zigzag order into natural array order, + * but are not dequantized. + * + * The i'th block of the MCU is stored into the block pointed to by + * MCU_data[i]. WE ASSUME THIS AREA HAS BEEN ZEROED BY THE CALLER. + * (Wholesale zeroing is usually a little faster than retail...) + * + * Returns FALSE if data source requested suspension. In that case no + * changes have been made to permanent state. (Exception: some output + * coefficients may already have been assigned. This is harmless for + * this module, since we'll just re-assign them on the next call.) + */ + +#define BUFSIZE (DCTSIZE2 * 2) + +METHODDEF(boolean) +decode_mcu (j_decompress_ptr cinfo, JBLOCKROW *MCU_data) +{ + huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy; + int usefast = 1; + + /* Process restart marker if needed; may have to suspend */ + if (cinfo->restart_interval) { + if (entropy->restarts_to_go == 0) + if (! process_restart(cinfo)) + return FALSE; + usefast = 0; + } + + if (cinfo->src->bytes_in_buffer < BUFSIZE * (size_t)cinfo->blocks_in_MCU + || cinfo->unread_marker != 0) + usefast = 0; + + /* If we've run out of data, just leave the MCU set to zeroes. + * This way, we return uniform gray for the remainder of the segment. + */ + if (! entropy->pub.insufficient_data) { + + if (usefast) { + if (!decode_mcu_fast(cinfo, MCU_data)) goto use_slow; + } + else { + use_slow: + if (!decode_mcu_slow(cinfo, MCU_data)) return FALSE; + } + + } + + /* Account for restart interval (no-op if not using restarts) */ + entropy->restarts_to_go--; + + return TRUE; +} + + +/* + * Module initialization routine for Huffman entropy decoding. + */ + +GLOBAL(void) +jinit_huff_decoder (j_decompress_ptr cinfo) +{ + huff_entropy_ptr entropy; + int i; + + /* Motion JPEG frames typically do not include the Huffman tables if they + are the default tables. Thus, if the tables are not set by the time + the Huffman decoder is initialized (usually within the body of + jpeg_start_decompress()), we set them to default values. */ + std_huff_tables((j_common_ptr) cinfo); + + entropy = (huff_entropy_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(huff_entropy_decoder)); + cinfo->entropy = (struct jpeg_entropy_decoder *) entropy; + entropy->pub.start_pass = start_pass_huff_decoder; + entropy->pub.decode_mcu = decode_mcu; + + /* Mark tables unallocated */ + for (i = 0; i < NUM_HUFF_TBLS; i++) { + entropy->dc_derived_tbls[i] = entropy->ac_derived_tbls[i] = NULL; + } +} diff --git a/app/src/main/jni/libjpeg-turbo/jdhuff.h b/app/src/main/jni/libjpeg-turbo/jdhuff.h new file mode 100644 index 000000000..f2805e46c --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdhuff.h @@ -0,0 +1,227 @@ +/* + * jdhuff.h + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1997, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright (C) 2010-2011, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains declarations for Huffman entropy decoding routines + * that are shared between the sequential decoder (jdhuff.c) and the + * progressive decoder (jdphuff.c). No other modules need to see these. + */ + + +/* Derived data constructed for each Huffman table */ + +#define HUFF_LOOKAHEAD 8 /* # of bits of lookahead */ + +typedef struct { + /* Basic tables: (element [0] of each array is unused) */ + INT32 maxcode[18]; /* largest code of length k (-1 if none) */ + /* (maxcode[17] is a sentinel to ensure jpeg_huff_decode terminates) */ + INT32 valoffset[18]; /* huffval[] offset for codes of length k */ + /* valoffset[k] = huffval[] index of 1st symbol of code length k, less + * the smallest code of length k; so given a code of length k, the + * corresponding symbol is huffval[code + valoffset[k]] + */ + + /* Link to public Huffman table (needed only in jpeg_huff_decode) */ + JHUFF_TBL *pub; + + /* Lookahead table: indexed by the next HUFF_LOOKAHEAD bits of + * the input data stream. If the next Huffman code is no more + * than HUFF_LOOKAHEAD bits long, we can obtain its length and + * the corresponding symbol directly from this tables. + * + * The lower 8 bits of each table entry contain the number of + * bits in the corresponding Huffman code, or HUFF_LOOKAHEAD + 1 + * if too long. The next 8 bits of each entry contain the + * symbol. + */ + int lookup[1< 32 bits on your machine, and shifting/masking longs is + * reasonably fast, making bit_buf_type be long and setting BIT_BUF_SIZE + * appropriately should be a win. Unfortunately we can't define the size + * with something like #define BIT_BUF_SIZE (sizeof(bit_buf_type)*8) + * because not all machines measure sizeof in 8-bit bytes. + */ + +typedef struct { /* Bitreading state saved across MCUs */ + bit_buf_type get_buffer; /* current bit-extraction buffer */ + int bits_left; /* # of unused bits in it */ +} bitread_perm_state; + +typedef struct { /* Bitreading working state within an MCU */ + /* Current data source location */ + /* We need a copy, rather than munging the original, in case of suspension */ + const JOCTET * next_input_byte; /* => next byte to read from source */ + size_t bytes_in_buffer; /* # of bytes remaining in source buffer */ + /* Bit input buffer --- note these values are kept in register variables, + * not in this struct, inside the inner loops. + */ + bit_buf_type get_buffer; /* current bit-extraction buffer */ + int bits_left; /* # of unused bits in it */ + /* Pointer needed by jpeg_fill_bit_buffer. */ + j_decompress_ptr cinfo; /* back link to decompress master record */ +} bitread_working_state; + +/* Macros to declare and load/save bitread local variables. */ +#define BITREAD_STATE_VARS \ + register bit_buf_type get_buffer; \ + register int bits_left; \ + bitread_working_state br_state + +#define BITREAD_LOAD_STATE(cinfop,permstate) \ + br_state.cinfo = cinfop; \ + br_state.next_input_byte = cinfop->src->next_input_byte; \ + br_state.bytes_in_buffer = cinfop->src->bytes_in_buffer; \ + get_buffer = permstate.get_buffer; \ + bits_left = permstate.bits_left; + +#define BITREAD_SAVE_STATE(cinfop,permstate) \ + cinfop->src->next_input_byte = br_state.next_input_byte; \ + cinfop->src->bytes_in_buffer = br_state.bytes_in_buffer; \ + permstate.get_buffer = get_buffer; \ + permstate.bits_left = bits_left + +/* + * These macros provide the in-line portion of bit fetching. + * Use CHECK_BIT_BUFFER to ensure there are N bits in get_buffer + * before using GET_BITS, PEEK_BITS, or DROP_BITS. + * The variables get_buffer and bits_left are assumed to be locals, + * but the state struct might not be (jpeg_huff_decode needs this). + * CHECK_BIT_BUFFER(state,n,action); + * Ensure there are N bits in get_buffer; if suspend, take action. + * val = GET_BITS(n); + * Fetch next N bits. + * val = PEEK_BITS(n); + * Fetch next N bits without removing them from the buffer. + * DROP_BITS(n); + * Discard next N bits. + * The value N should be a simple variable, not an expression, because it + * is evaluated multiple times. + */ + +#define CHECK_BIT_BUFFER(state,nbits,action) \ + { if (bits_left < (nbits)) { \ + if (! jpeg_fill_bit_buffer(&(state),get_buffer,bits_left,nbits)) \ + { action; } \ + get_buffer = (state).get_buffer; bits_left = (state).bits_left; } } + +#define GET_BITS(nbits) \ + (((int) (get_buffer >> (bits_left -= (nbits)))) & ((1<<(nbits))-1)) + +#define PEEK_BITS(nbits) \ + (((int) (get_buffer >> (bits_left - (nbits)))) & ((1<<(nbits))-1)) + +#define DROP_BITS(nbits) \ + (bits_left -= (nbits)) + +/* Load up the bit buffer to a depth of at least nbits */ +EXTERN(boolean) jpeg_fill_bit_buffer + (bitread_working_state * state, register bit_buf_type get_buffer, + register int bits_left, int nbits); + + +/* + * Code for extracting next Huffman-coded symbol from input bit stream. + * Again, this is time-critical and we make the main paths be macros. + * + * We use a lookahead table to process codes of up to HUFF_LOOKAHEAD bits + * without looping. Usually, more than 95% of the Huffman codes will be 8 + * or fewer bits long. The few overlength codes are handled with a loop, + * which need not be inline code. + * + * Notes about the HUFF_DECODE macro: + * 1. Near the end of the data segment, we may fail to get enough bits + * for a lookahead. In that case, we do it the hard way. + * 2. If the lookahead table contains no entry, the next code must be + * more than HUFF_LOOKAHEAD bits long. + * 3. jpeg_huff_decode returns -1 if forced to suspend. + */ + +#define HUFF_DECODE(result,state,htbl,failaction,slowlabel) \ +{ register int nb, look; \ + if (bits_left < HUFF_LOOKAHEAD) { \ + if (! jpeg_fill_bit_buffer(&state,get_buffer,bits_left, 0)) {failaction;} \ + get_buffer = state.get_buffer; bits_left = state.bits_left; \ + if (bits_left < HUFF_LOOKAHEAD) { \ + nb = 1; goto slowlabel; \ + } \ + } \ + look = PEEK_BITS(HUFF_LOOKAHEAD); \ + if ((nb = (htbl->lookup[look] >> HUFF_LOOKAHEAD)) <= HUFF_LOOKAHEAD) { \ + DROP_BITS(nb); \ + result = htbl->lookup[look] & ((1 << HUFF_LOOKAHEAD) - 1); \ + } else { \ +slowlabel: \ + if ((result=jpeg_huff_decode(&state,get_buffer,bits_left,htbl,nb)) < 0) \ + { failaction; } \ + get_buffer = state.get_buffer; bits_left = state.bits_left; \ + } \ +} + +#define HUFF_DECODE_FAST(s,nb,htbl) \ + FILL_BIT_BUFFER_FAST; \ + s = PEEK_BITS(HUFF_LOOKAHEAD); \ + s = htbl->lookup[s]; \ + nb = s >> HUFF_LOOKAHEAD; \ + /* Pre-execute the common case of nb <= HUFF_LOOKAHEAD */ \ + DROP_BITS(nb); \ + s = s & ((1 << HUFF_LOOKAHEAD) - 1); \ + if (nb > HUFF_LOOKAHEAD) { \ + /* Equivalent of jpeg_huff_decode() */ \ + /* Don't use GET_BITS() here because we don't want to modify bits_left */ \ + s = (get_buffer >> bits_left) & ((1 << (nb)) - 1); \ + while (s > htbl->maxcode[nb]) { \ + s <<= 1; \ + s |= GET_BITS(1); \ + nb++; \ + } \ + s = htbl->pub->huffval[ (int) (s + htbl->valoffset[nb]) & 0xFF ]; \ + } + +/* Out-of-line case for Huffman code fetching */ +EXTERN(int) jpeg_huff_decode + (bitread_working_state * state, register bit_buf_type get_buffer, + register int bits_left, d_derived_tbl * htbl, int min_bits); diff --git a/app/src/main/jni/libjpeg-turbo/jdinput.c b/app/src/main/jni/libjpeg-turbo/jdinput.c new file mode 100644 index 000000000..6f4ea7ba3 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdinput.c @@ -0,0 +1,398 @@ +/* + * jdinput.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1997, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright (C) 2010, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains input control logic for the JPEG decompressor. + * These routines are concerned with controlling the decompressor's input + * processing (marker reading and coefficient decoding). The actual input + * reading is done in jdmarker.c, jdhuff.c, and jdphuff.c. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jpegcomp.h" + + +/* Private state */ + +typedef struct { + struct jpeg_input_controller pub; /* public fields */ + + boolean inheaders; /* TRUE until first SOS is reached */ +} my_input_controller; + +typedef my_input_controller * my_inputctl_ptr; + + +/* Forward declarations */ +METHODDEF(int) consume_markers (j_decompress_ptr cinfo); + + +/* + * Routines to calculate various quantities related to the size of the image. + */ + +LOCAL(void) +initial_setup (j_decompress_ptr cinfo) +/* Called once, when first SOS marker is reached */ +{ + int ci; + jpeg_component_info *compptr; + + /* Make sure image isn't bigger than I can handle */ + if ((long) cinfo->image_height > (long) JPEG_MAX_DIMENSION || + (long) cinfo->image_width > (long) JPEG_MAX_DIMENSION) + ERREXIT1(cinfo, JERR_IMAGE_TOO_BIG, (unsigned int) JPEG_MAX_DIMENSION); + + /* For now, precision must match compiled-in value... */ + if (cinfo->data_precision != BITS_IN_JSAMPLE) + ERREXIT1(cinfo, JERR_BAD_PRECISION, cinfo->data_precision); + + /* Check that number of components won't exceed internal array sizes */ + if (cinfo->num_components > MAX_COMPONENTS) + ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components, + MAX_COMPONENTS); + + /* Compute maximum sampling factors; check factor validity */ + cinfo->max_h_samp_factor = 1; + cinfo->max_v_samp_factor = 1; + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + if (compptr->h_samp_factor<=0 || compptr->h_samp_factor>MAX_SAMP_FACTOR || + compptr->v_samp_factor<=0 || compptr->v_samp_factor>MAX_SAMP_FACTOR) + ERREXIT(cinfo, JERR_BAD_SAMPLING); + cinfo->max_h_samp_factor = MAX(cinfo->max_h_samp_factor, + compptr->h_samp_factor); + cinfo->max_v_samp_factor = MAX(cinfo->max_v_samp_factor, + compptr->v_samp_factor); + } + +#if JPEG_LIB_VERSION >=80 + cinfo->block_size = DCTSIZE; + cinfo->natural_order = jpeg_natural_order; + cinfo->lim_Se = DCTSIZE2-1; +#endif + + /* We initialize DCT_scaled_size and min_DCT_scaled_size to DCTSIZE. + * In the full decompressor, this will be overridden by jdmaster.c; + * but in the transcoder, jdmaster.c is not used, so we must do it here. + */ +#if JPEG_LIB_VERSION >= 70 + cinfo->min_DCT_h_scaled_size = cinfo->min_DCT_v_scaled_size = DCTSIZE; +#else + cinfo->min_DCT_scaled_size = DCTSIZE; +#endif + + /* Compute dimensions of components */ + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { +#if JPEG_LIB_VERSION >= 70 + compptr->DCT_h_scaled_size = compptr->DCT_v_scaled_size = DCTSIZE; +#else + compptr->DCT_scaled_size = DCTSIZE; +#endif + /* Size in DCT blocks */ + compptr->width_in_blocks = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor, + (long) (cinfo->max_h_samp_factor * DCTSIZE)); + compptr->height_in_blocks = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor, + (long) (cinfo->max_v_samp_factor * DCTSIZE)); + /* downsampled_width and downsampled_height will also be overridden by + * jdmaster.c if we are doing full decompression. The transcoder library + * doesn't use these values, but the calling application might. + */ + /* Size in samples */ + compptr->downsampled_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor, + (long) cinfo->max_h_samp_factor); + compptr->downsampled_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor, + (long) cinfo->max_v_samp_factor); + /* Mark component needed, until color conversion says otherwise */ + compptr->component_needed = TRUE; + /* Mark no quantization table yet saved for component */ + compptr->quant_table = NULL; + } + + /* Compute number of fully interleaved MCU rows. */ + cinfo->total_iMCU_rows = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height, + (long) (cinfo->max_v_samp_factor*DCTSIZE)); + + /* Decide whether file contains multiple scans */ + if (cinfo->comps_in_scan < cinfo->num_components || cinfo->progressive_mode) + cinfo->inputctl->has_multiple_scans = TRUE; + else + cinfo->inputctl->has_multiple_scans = FALSE; +} + + +LOCAL(void) +per_scan_setup (j_decompress_ptr cinfo) +/* Do computations that are needed before processing a JPEG scan */ +/* cinfo->comps_in_scan and cinfo->cur_comp_info[] were set from SOS marker */ +{ + int ci, mcublks, tmp; + jpeg_component_info *compptr; + + if (cinfo->comps_in_scan == 1) { + + /* Noninterleaved (single-component) scan */ + compptr = cinfo->cur_comp_info[0]; + + /* Overall image size in MCUs */ + cinfo->MCUs_per_row = compptr->width_in_blocks; + cinfo->MCU_rows_in_scan = compptr->height_in_blocks; + + /* For noninterleaved scan, always one block per MCU */ + compptr->MCU_width = 1; + compptr->MCU_height = 1; + compptr->MCU_blocks = 1; + compptr->MCU_sample_width = compptr->_DCT_scaled_size; + compptr->last_col_width = 1; + /* For noninterleaved scans, it is convenient to define last_row_height + * as the number of block rows present in the last iMCU row. + */ + tmp = (int) (compptr->height_in_blocks % compptr->v_samp_factor); + if (tmp == 0) tmp = compptr->v_samp_factor; + compptr->last_row_height = tmp; + + /* Prepare array describing MCU composition */ + cinfo->blocks_in_MCU = 1; + cinfo->MCU_membership[0] = 0; + + } else { + + /* Interleaved (multi-component) scan */ + if (cinfo->comps_in_scan <= 0 || cinfo->comps_in_scan > MAX_COMPS_IN_SCAN) + ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->comps_in_scan, + MAX_COMPS_IN_SCAN); + + /* Overall image size in MCUs */ + cinfo->MCUs_per_row = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width, + (long) (cinfo->max_h_samp_factor*DCTSIZE)); + cinfo->MCU_rows_in_scan = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height, + (long) (cinfo->max_v_samp_factor*DCTSIZE)); + + cinfo->blocks_in_MCU = 0; + + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + /* Sampling factors give # of blocks of component in each MCU */ + compptr->MCU_width = compptr->h_samp_factor; + compptr->MCU_height = compptr->v_samp_factor; + compptr->MCU_blocks = compptr->MCU_width * compptr->MCU_height; + compptr->MCU_sample_width = compptr->MCU_width * compptr->_DCT_scaled_size; + /* Figure number of non-dummy blocks in last MCU column & row */ + tmp = (int) (compptr->width_in_blocks % compptr->MCU_width); + if (tmp == 0) tmp = compptr->MCU_width; + compptr->last_col_width = tmp; + tmp = (int) (compptr->height_in_blocks % compptr->MCU_height); + if (tmp == 0) tmp = compptr->MCU_height; + compptr->last_row_height = tmp; + /* Prepare array describing MCU composition */ + mcublks = compptr->MCU_blocks; + if (cinfo->blocks_in_MCU + mcublks > D_MAX_BLOCKS_IN_MCU) + ERREXIT(cinfo, JERR_BAD_MCU_SIZE); + while (mcublks-- > 0) { + cinfo->MCU_membership[cinfo->blocks_in_MCU++] = ci; + } + } + + } +} + + +/* + * Save away a copy of the Q-table referenced by each component present + * in the current scan, unless already saved during a prior scan. + * + * In a multiple-scan JPEG file, the encoder could assign different components + * the same Q-table slot number, but change table definitions between scans + * so that each component uses a different Q-table. (The IJG encoder is not + * currently capable of doing this, but other encoders might.) Since we want + * to be able to dequantize all the components at the end of the file, this + * means that we have to save away the table actually used for each component. + * We do this by copying the table at the start of the first scan containing + * the component. + * The JPEG spec prohibits the encoder from changing the contents of a Q-table + * slot between scans of a component using that slot. If the encoder does so + * anyway, this decoder will simply use the Q-table values that were current + * at the start of the first scan for the component. + * + * The decompressor output side looks only at the saved quant tables, + * not at the current Q-table slots. + */ + +LOCAL(void) +latch_quant_tables (j_decompress_ptr cinfo) +{ + int ci, qtblno; + jpeg_component_info *compptr; + JQUANT_TBL * qtbl; + + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + /* No work if we already saved Q-table for this component */ + if (compptr->quant_table != NULL) + continue; + /* Make sure specified quantization table is present */ + qtblno = compptr->quant_tbl_no; + if (qtblno < 0 || qtblno >= NUM_QUANT_TBLS || + cinfo->quant_tbl_ptrs[qtblno] == NULL) + ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, qtblno); + /* OK, save away the quantization table */ + qtbl = (JQUANT_TBL *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(JQUANT_TBL)); + MEMCOPY(qtbl, cinfo->quant_tbl_ptrs[qtblno], sizeof(JQUANT_TBL)); + compptr->quant_table = qtbl; + } +} + + +/* + * Initialize the input modules to read a scan of compressed data. + * The first call to this is done by jdmaster.c after initializing + * the entire decompressor (during jpeg_start_decompress). + * Subsequent calls come from consume_markers, below. + */ + +METHODDEF(void) +start_input_pass (j_decompress_ptr cinfo) +{ + per_scan_setup(cinfo); + latch_quant_tables(cinfo); + (*cinfo->entropy->start_pass) (cinfo); + (*cinfo->coef->start_input_pass) (cinfo); + cinfo->inputctl->consume_input = cinfo->coef->consume_data; +} + + +/* + * Finish up after inputting a compressed-data scan. + * This is called by the coefficient controller after it's read all + * the expected data of the scan. + */ + +METHODDEF(void) +finish_input_pass (j_decompress_ptr cinfo) +{ + cinfo->inputctl->consume_input = consume_markers; +} + + +/* + * Read JPEG markers before, between, or after compressed-data scans. + * Change state as necessary when a new scan is reached. + * Return value is JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI. + * + * The consume_input method pointer points either here or to the + * coefficient controller's consume_data routine, depending on whether + * we are reading a compressed data segment or inter-segment markers. + */ + +METHODDEF(int) +consume_markers (j_decompress_ptr cinfo) +{ + my_inputctl_ptr inputctl = (my_inputctl_ptr) cinfo->inputctl; + int val; + + if (inputctl->pub.eoi_reached) /* After hitting EOI, read no further */ + return JPEG_REACHED_EOI; + + val = (*cinfo->marker->read_markers) (cinfo); + + switch (val) { + case JPEG_REACHED_SOS: /* Found SOS */ + if (inputctl->inheaders) { /* 1st SOS */ + initial_setup(cinfo); + inputctl->inheaders = FALSE; + /* Note: start_input_pass must be called by jdmaster.c + * before any more input can be consumed. jdapimin.c is + * responsible for enforcing this sequencing. + */ + } else { /* 2nd or later SOS marker */ + if (! inputctl->pub.has_multiple_scans) + ERREXIT(cinfo, JERR_EOI_EXPECTED); /* Oops, I wasn't expecting this! */ + start_input_pass(cinfo); + } + break; + case JPEG_REACHED_EOI: /* Found EOI */ + inputctl->pub.eoi_reached = TRUE; + if (inputctl->inheaders) { /* Tables-only datastream, apparently */ + if (cinfo->marker->saw_SOF) + ERREXIT(cinfo, JERR_SOF_NO_SOS); + } else { + /* Prevent infinite loop in coef ctlr's decompress_data routine + * if user set output_scan_number larger than number of scans. + */ + if (cinfo->output_scan_number > cinfo->input_scan_number) + cinfo->output_scan_number = cinfo->input_scan_number; + } + break; + case JPEG_SUSPENDED: + break; + } + + return val; +} + + +/* + * Reset state to begin a fresh datastream. + */ + +METHODDEF(void) +reset_input_controller (j_decompress_ptr cinfo) +{ + my_inputctl_ptr inputctl = (my_inputctl_ptr) cinfo->inputctl; + + inputctl->pub.consume_input = consume_markers; + inputctl->pub.has_multiple_scans = FALSE; /* "unknown" would be better */ + inputctl->pub.eoi_reached = FALSE; + inputctl->inheaders = TRUE; + /* Reset other modules */ + (*cinfo->err->reset_error_mgr) ((j_common_ptr) cinfo); + (*cinfo->marker->reset_marker_reader) (cinfo); + /* Reset progression state -- would be cleaner if entropy decoder did this */ + cinfo->coef_bits = NULL; +} + + +/* + * Initialize the input controller module. + * This is called only once, when the decompression object is created. + */ + +GLOBAL(void) +jinit_input_controller (j_decompress_ptr cinfo) +{ + my_inputctl_ptr inputctl; + + /* Create subobject in permanent pool */ + inputctl = (my_inputctl_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT, + sizeof(my_input_controller)); + cinfo->inputctl = (struct jpeg_input_controller *) inputctl; + /* Initialize method pointers */ + inputctl->pub.consume_input = consume_markers; + inputctl->pub.reset_input_controller = reset_input_controller; + inputctl->pub.start_input_pass = start_input_pass; + inputctl->pub.finish_input_pass = finish_input_pass; + /* Initialize state: can't use reset_input_controller since we don't + * want to try to reset other modules yet. + */ + inputctl->pub.has_multiple_scans = FALSE; /* "unknown" would be better */ + inputctl->pub.eoi_reached = FALSE; + inputctl->inheaders = TRUE; +} diff --git a/app/src/main/jni/libjpeg-turbo/jdmainct.c b/app/src/main/jni/libjpeg-turbo/jdmainct.c new file mode 100644 index 000000000..7f7bd33e8 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdmainct.c @@ -0,0 +1,515 @@ +/* + * jdmainct.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1996, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright (C) 2010, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains the main buffer controller for decompression. + * The main buffer lies between the JPEG decompressor proper and the + * post-processor; it holds downsampled data in the JPEG colorspace. + * + * Note that this code is bypassed in raw-data mode, since the application + * supplies the equivalent of the main buffer in that case. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jpegcomp.h" + + +/* + * In the current system design, the main buffer need never be a full-image + * buffer; any full-height buffers will be found inside the coefficient or + * postprocessing controllers. Nonetheless, the main controller is not + * trivial. Its responsibility is to provide context rows for upsampling/ + * rescaling, and doing this in an efficient fashion is a bit tricky. + * + * Postprocessor input data is counted in "row groups". A row group + * is defined to be (v_samp_factor * DCT_scaled_size / min_DCT_scaled_size) + * sample rows of each component. (We require DCT_scaled_size values to be + * chosen such that these numbers are integers. In practice DCT_scaled_size + * values will likely be powers of two, so we actually have the stronger + * condition that DCT_scaled_size / min_DCT_scaled_size is an integer.) + * Upsampling will typically produce max_v_samp_factor pixel rows from each + * row group (times any additional scale factor that the upsampler is + * applying). + * + * The coefficient controller will deliver data to us one iMCU row at a time; + * each iMCU row contains v_samp_factor * DCT_scaled_size sample rows, or + * exactly min_DCT_scaled_size row groups. (This amount of data corresponds + * to one row of MCUs when the image is fully interleaved.) Note that the + * number of sample rows varies across components, but the number of row + * groups does not. Some garbage sample rows may be included in the last iMCU + * row at the bottom of the image. + * + * Depending on the vertical scaling algorithm used, the upsampler may need + * access to the sample row(s) above and below its current input row group. + * The upsampler is required to set need_context_rows TRUE at global selection + * time if so. When need_context_rows is FALSE, this controller can simply + * obtain one iMCU row at a time from the coefficient controller and dole it + * out as row groups to the postprocessor. + * + * When need_context_rows is TRUE, this controller guarantees that the buffer + * passed to postprocessing contains at least one row group's worth of samples + * above and below the row group(s) being processed. Note that the context + * rows "above" the first passed row group appear at negative row offsets in + * the passed buffer. At the top and bottom of the image, the required + * context rows are manufactured by duplicating the first or last real sample + * row; this avoids having special cases in the upsampling inner loops. + * + * The amount of context is fixed at one row group just because that's a + * convenient number for this controller to work with. The existing + * upsamplers really only need one sample row of context. An upsampler + * supporting arbitrary output rescaling might wish for more than one row + * group of context when shrinking the image; tough, we don't handle that. + * (This is justified by the assumption that downsizing will be handled mostly + * by adjusting the DCT_scaled_size values, so that the actual scale factor at + * the upsample step needn't be much less than one.) + * + * To provide the desired context, we have to retain the last two row groups + * of one iMCU row while reading in the next iMCU row. (The last row group + * can't be processed until we have another row group for its below-context, + * and so we have to save the next-to-last group too for its above-context.) + * We could do this most simply by copying data around in our buffer, but + * that'd be very slow. We can avoid copying any data by creating a rather + * strange pointer structure. Here's how it works. We allocate a workspace + * consisting of M+2 row groups (where M = min_DCT_scaled_size is the number + * of row groups per iMCU row). We create two sets of redundant pointers to + * the workspace. Labeling the physical row groups 0 to M+1, the synthesized + * pointer lists look like this: + * M+1 M-1 + * master pointer --> 0 master pointer --> 0 + * 1 1 + * ... ... + * M-3 M-3 + * M-2 M + * M-1 M+1 + * M M-2 + * M+1 M-1 + * 0 0 + * We read alternate iMCU rows using each master pointer; thus the last two + * row groups of the previous iMCU row remain un-overwritten in the workspace. + * The pointer lists are set up so that the required context rows appear to + * be adjacent to the proper places when we pass the pointer lists to the + * upsampler. + * + * The above pictures describe the normal state of the pointer lists. + * At top and bottom of the image, we diddle the pointer lists to duplicate + * the first or last sample row as necessary (this is cheaper than copying + * sample rows around). + * + * This scheme breaks down if M < 2, ie, min_DCT_scaled_size is 1. In that + * situation each iMCU row provides only one row group so the buffering logic + * must be different (eg, we must read two iMCU rows before we can emit the + * first row group). For now, we simply do not support providing context + * rows when min_DCT_scaled_size is 1. That combination seems unlikely to + * be worth providing --- if someone wants a 1/8th-size preview, they probably + * want it quick and dirty, so a context-free upsampler is sufficient. + */ + + +/* Private buffer controller object */ + +typedef struct { + struct jpeg_d_main_controller pub; /* public fields */ + + /* Pointer to allocated workspace (M or M+2 row groups). */ + JSAMPARRAY buffer[MAX_COMPONENTS]; + + boolean buffer_full; /* Have we gotten an iMCU row from decoder? */ + JDIMENSION rowgroup_ctr; /* counts row groups output to postprocessor */ + + /* Remaining fields are only used in the context case. */ + + /* These are the master pointers to the funny-order pointer lists. */ + JSAMPIMAGE xbuffer[2]; /* pointers to weird pointer lists */ + + int whichptr; /* indicates which pointer set is now in use */ + int context_state; /* process_data state machine status */ + JDIMENSION rowgroups_avail; /* row groups available to postprocessor */ + JDIMENSION iMCU_row_ctr; /* counts iMCU rows to detect image top/bot */ +} my_main_controller; + +typedef my_main_controller * my_main_ptr; + +/* context_state values: */ +#define CTX_PREPARE_FOR_IMCU 0 /* need to prepare for MCU row */ +#define CTX_PROCESS_IMCU 1 /* feeding iMCU to postprocessor */ +#define CTX_POSTPONED_ROW 2 /* feeding postponed row group */ + + +/* Forward declarations */ +METHODDEF(void) process_data_simple_main + (j_decompress_ptr cinfo, JSAMPARRAY output_buf, + JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail); +METHODDEF(void) process_data_context_main + (j_decompress_ptr cinfo, JSAMPARRAY output_buf, + JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail); +#ifdef QUANT_2PASS_SUPPORTED +METHODDEF(void) process_data_crank_post + (j_decompress_ptr cinfo, JSAMPARRAY output_buf, + JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail); +#endif + + +LOCAL(void) +alloc_funny_pointers (j_decompress_ptr cinfo) +/* Allocate space for the funny pointer lists. + * This is done only once, not once per pass. + */ +{ + my_main_ptr main_ptr = (my_main_ptr) cinfo->main; + int ci, rgroup; + int M = cinfo->_min_DCT_scaled_size; + jpeg_component_info *compptr; + JSAMPARRAY xbuf; + + /* Get top-level space for component array pointers. + * We alloc both arrays with one call to save a few cycles. + */ + main_ptr->xbuffer[0] = (JSAMPIMAGE) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + cinfo->num_components * 2 * sizeof(JSAMPARRAY)); + main_ptr->xbuffer[1] = main_ptr->xbuffer[0] + cinfo->num_components; + + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) / + cinfo->_min_DCT_scaled_size; /* height of a row group of component */ + /* Get space for pointer lists --- M+4 row groups in each list. + * We alloc both pointer lists with one call to save a few cycles. + */ + xbuf = (JSAMPARRAY) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + 2 * (rgroup * (M + 4)) * sizeof(JSAMPROW)); + xbuf += rgroup; /* want one row group at negative offsets */ + main_ptr->xbuffer[0][ci] = xbuf; + xbuf += rgroup * (M + 4); + main_ptr->xbuffer[1][ci] = xbuf; + } +} + + +LOCAL(void) +make_funny_pointers (j_decompress_ptr cinfo) +/* Create the funny pointer lists discussed in the comments above. + * The actual workspace is already allocated (in main_ptr->buffer), + * and the space for the pointer lists is allocated too. + * This routine just fills in the curiously ordered lists. + * This will be repeated at the beginning of each pass. + */ +{ + my_main_ptr main_ptr = (my_main_ptr) cinfo->main; + int ci, i, rgroup; + int M = cinfo->_min_DCT_scaled_size; + jpeg_component_info *compptr; + JSAMPARRAY buf, xbuf0, xbuf1; + + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) / + cinfo->_min_DCT_scaled_size; /* height of a row group of component */ + xbuf0 = main_ptr->xbuffer[0][ci]; + xbuf1 = main_ptr->xbuffer[1][ci]; + /* First copy the workspace pointers as-is */ + buf = main_ptr->buffer[ci]; + for (i = 0; i < rgroup * (M + 2); i++) { + xbuf0[i] = xbuf1[i] = buf[i]; + } + /* In the second list, put the last four row groups in swapped order */ + for (i = 0; i < rgroup * 2; i++) { + xbuf1[rgroup*(M-2) + i] = buf[rgroup*M + i]; + xbuf1[rgroup*M + i] = buf[rgroup*(M-2) + i]; + } + /* The wraparound pointers at top and bottom will be filled later + * (see set_wraparound_pointers, below). Initially we want the "above" + * pointers to duplicate the first actual data line. This only needs + * to happen in xbuffer[0]. + */ + for (i = 0; i < rgroup; i++) { + xbuf0[i - rgroup] = xbuf0[0]; + } + } +} + + +LOCAL(void) +set_wraparound_pointers (j_decompress_ptr cinfo) +/* Set up the "wraparound" pointers at top and bottom of the pointer lists. + * This changes the pointer list state from top-of-image to the normal state. + */ +{ + my_main_ptr main_ptr = (my_main_ptr) cinfo->main; + int ci, i, rgroup; + int M = cinfo->_min_DCT_scaled_size; + jpeg_component_info *compptr; + JSAMPARRAY xbuf0, xbuf1; + + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) / + cinfo->_min_DCT_scaled_size; /* height of a row group of component */ + xbuf0 = main_ptr->xbuffer[0][ci]; + xbuf1 = main_ptr->xbuffer[1][ci]; + for (i = 0; i < rgroup; i++) { + xbuf0[i - rgroup] = xbuf0[rgroup*(M+1) + i]; + xbuf1[i - rgroup] = xbuf1[rgroup*(M+1) + i]; + xbuf0[rgroup*(M+2) + i] = xbuf0[i]; + xbuf1[rgroup*(M+2) + i] = xbuf1[i]; + } + } +} + + +LOCAL(void) +set_bottom_pointers (j_decompress_ptr cinfo) +/* Change the pointer lists to duplicate the last sample row at the bottom + * of the image. whichptr indicates which xbuffer holds the final iMCU row. + * Also sets rowgroups_avail to indicate number of nondummy row groups in row. + */ +{ + my_main_ptr main_ptr = (my_main_ptr) cinfo->main; + int ci, i, rgroup, iMCUheight, rows_left; + jpeg_component_info *compptr; + JSAMPARRAY xbuf; + + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + /* Count sample rows in one iMCU row and in one row group */ + iMCUheight = compptr->v_samp_factor * compptr->_DCT_scaled_size; + rgroup = iMCUheight / cinfo->_min_DCT_scaled_size; + /* Count nondummy sample rows remaining for this component */ + rows_left = (int) (compptr->downsampled_height % (JDIMENSION) iMCUheight); + if (rows_left == 0) rows_left = iMCUheight; + /* Count nondummy row groups. Should get same answer for each component, + * so we need only do it once. + */ + if (ci == 0) { + main_ptr->rowgroups_avail = (JDIMENSION) ((rows_left-1) / rgroup + 1); + } + /* Duplicate the last real sample row rgroup*2 times; this pads out the + * last partial rowgroup and ensures at least one full rowgroup of context. + */ + xbuf = main_ptr->xbuffer[main_ptr->whichptr][ci]; + for (i = 0; i < rgroup * 2; i++) { + xbuf[rows_left + i] = xbuf[rows_left-1]; + } + } +} + + +/* + * Initialize for a processing pass. + */ + +METHODDEF(void) +start_pass_main (j_decompress_ptr cinfo, J_BUF_MODE pass_mode) +{ + my_main_ptr main_ptr = (my_main_ptr) cinfo->main; + + switch (pass_mode) { + case JBUF_PASS_THRU: + if (cinfo->upsample->need_context_rows) { + main_ptr->pub.process_data = process_data_context_main; + make_funny_pointers(cinfo); /* Create the xbuffer[] lists */ + main_ptr->whichptr = 0; /* Read first iMCU row into xbuffer[0] */ + main_ptr->context_state = CTX_PREPARE_FOR_IMCU; + main_ptr->iMCU_row_ctr = 0; + } else { + /* Simple case with no context needed */ + main_ptr->pub.process_data = process_data_simple_main; + } + main_ptr->buffer_full = FALSE; /* Mark buffer empty */ + main_ptr->rowgroup_ctr = 0; + break; +#ifdef QUANT_2PASS_SUPPORTED + case JBUF_CRANK_DEST: + /* For last pass of 2-pass quantization, just crank the postprocessor */ + main_ptr->pub.process_data = process_data_crank_post; + break; +#endif + default: + ERREXIT(cinfo, JERR_BAD_BUFFER_MODE); + break; + } +} + + +/* + * Process some data. + * This handles the simple case where no context is required. + */ + +METHODDEF(void) +process_data_simple_main (j_decompress_ptr cinfo, + JSAMPARRAY output_buf, JDIMENSION *out_row_ctr, + JDIMENSION out_rows_avail) +{ + my_main_ptr main_ptr = (my_main_ptr) cinfo->main; + JDIMENSION rowgroups_avail; + + /* Read input data if we haven't filled the main buffer yet */ + if (! main_ptr->buffer_full) { + if (! (*cinfo->coef->decompress_data) (cinfo, main_ptr->buffer)) + return; /* suspension forced, can do nothing more */ + main_ptr->buffer_full = TRUE; /* OK, we have an iMCU row to work with */ + } + + /* There are always min_DCT_scaled_size row groups in an iMCU row. */ + rowgroups_avail = (JDIMENSION) cinfo->_min_DCT_scaled_size; + /* Note: at the bottom of the image, we may pass extra garbage row groups + * to the postprocessor. The postprocessor has to check for bottom + * of image anyway (at row resolution), so no point in us doing it too. + */ + + /* Feed the postprocessor */ + (*cinfo->post->post_process_data) (cinfo, main_ptr->buffer, + &main_ptr->rowgroup_ctr, rowgroups_avail, + output_buf, out_row_ctr, out_rows_avail); + + /* Has postprocessor consumed all the data yet? If so, mark buffer empty */ + if (main_ptr->rowgroup_ctr >= rowgroups_avail) { + main_ptr->buffer_full = FALSE; + main_ptr->rowgroup_ctr = 0; + } +} + + +/* + * Process some data. + * This handles the case where context rows must be provided. + */ + +METHODDEF(void) +process_data_context_main (j_decompress_ptr cinfo, + JSAMPARRAY output_buf, JDIMENSION *out_row_ctr, + JDIMENSION out_rows_avail) +{ + my_main_ptr main_ptr = (my_main_ptr) cinfo->main; + + /* Read input data if we haven't filled the main buffer yet */ + if (! main_ptr->buffer_full) { + if (! (*cinfo->coef->decompress_data) (cinfo, + main_ptr->xbuffer[main_ptr->whichptr])) + return; /* suspension forced, can do nothing more */ + main_ptr->buffer_full = TRUE; /* OK, we have an iMCU row to work with */ + main_ptr->iMCU_row_ctr++; /* count rows received */ + } + + /* Postprocessor typically will not swallow all the input data it is handed + * in one call (due to filling the output buffer first). Must be prepared + * to exit and restart. This switch lets us keep track of how far we got. + * Note that each case falls through to the next on successful completion. + */ + switch (main_ptr->context_state) { + case CTX_POSTPONED_ROW: + /* Call postprocessor using previously set pointers for postponed row */ + (*cinfo->post->post_process_data) (cinfo, main_ptr->xbuffer[main_ptr->whichptr], + &main_ptr->rowgroup_ctr, main_ptr->rowgroups_avail, + output_buf, out_row_ctr, out_rows_avail); + if (main_ptr->rowgroup_ctr < main_ptr->rowgroups_avail) + return; /* Need to suspend */ + main_ptr->context_state = CTX_PREPARE_FOR_IMCU; + if (*out_row_ctr >= out_rows_avail) + return; /* Postprocessor exactly filled output buf */ + /*FALLTHROUGH*/ + case CTX_PREPARE_FOR_IMCU: + /* Prepare to process first M-1 row groups of this iMCU row */ + main_ptr->rowgroup_ctr = 0; + main_ptr->rowgroups_avail = (JDIMENSION) (cinfo->_min_DCT_scaled_size - 1); + /* Check for bottom of image: if so, tweak pointers to "duplicate" + * the last sample row, and adjust rowgroups_avail to ignore padding rows. + */ + if (main_ptr->iMCU_row_ctr == cinfo->total_iMCU_rows) + set_bottom_pointers(cinfo); + main_ptr->context_state = CTX_PROCESS_IMCU; + /*FALLTHROUGH*/ + case CTX_PROCESS_IMCU: + /* Call postprocessor using previously set pointers */ + (*cinfo->post->post_process_data) (cinfo, main_ptr->xbuffer[main_ptr->whichptr], + &main_ptr->rowgroup_ctr, main_ptr->rowgroups_avail, + output_buf, out_row_ctr, out_rows_avail); + if (main_ptr->rowgroup_ctr < main_ptr->rowgroups_avail) + return; /* Need to suspend */ + /* After the first iMCU, change wraparound pointers to normal state */ + if (main_ptr->iMCU_row_ctr == 1) + set_wraparound_pointers(cinfo); + /* Prepare to load new iMCU row using other xbuffer list */ + main_ptr->whichptr ^= 1; /* 0=>1 or 1=>0 */ + main_ptr->buffer_full = FALSE; + /* Still need to process last row group of this iMCU row, */ + /* which is saved at index M+1 of the other xbuffer */ + main_ptr->rowgroup_ctr = (JDIMENSION) (cinfo->_min_DCT_scaled_size + 1); + main_ptr->rowgroups_avail = (JDIMENSION) (cinfo->_min_DCT_scaled_size + 2); + main_ptr->context_state = CTX_POSTPONED_ROW; + } +} + + +/* + * Process some data. + * Final pass of two-pass quantization: just call the postprocessor. + * Source data will be the postprocessor controller's internal buffer. + */ + +#ifdef QUANT_2PASS_SUPPORTED + +METHODDEF(void) +process_data_crank_post (j_decompress_ptr cinfo, + JSAMPARRAY output_buf, JDIMENSION *out_row_ctr, + JDIMENSION out_rows_avail) +{ + (*cinfo->post->post_process_data) (cinfo, (JSAMPIMAGE) NULL, + (JDIMENSION *) NULL, (JDIMENSION) 0, + output_buf, out_row_ctr, out_rows_avail); +} + +#endif /* QUANT_2PASS_SUPPORTED */ + + +/* + * Initialize main buffer controller. + */ + +GLOBAL(void) +jinit_d_main_controller (j_decompress_ptr cinfo, boolean need_full_buffer) +{ + my_main_ptr main_ptr; + int ci, rgroup, ngroups; + jpeg_component_info *compptr; + + main_ptr = (my_main_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(my_main_controller)); + cinfo->main = (struct jpeg_d_main_controller *) main_ptr; + main_ptr->pub.start_pass = start_pass_main; + + if (need_full_buffer) /* shouldn't happen */ + ERREXIT(cinfo, JERR_BAD_BUFFER_MODE); + + /* Allocate the workspace. + * ngroups is the number of row groups we need. + */ + if (cinfo->upsample->need_context_rows) { + if (cinfo->_min_DCT_scaled_size < 2) /* unsupported, see comments above */ + ERREXIT(cinfo, JERR_NOTIMPL); + alloc_funny_pointers(cinfo); /* Alloc space for xbuffer[] lists */ + ngroups = cinfo->_min_DCT_scaled_size + 2; + } else { + ngroups = cinfo->_min_DCT_scaled_size; + } + + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) / + cinfo->_min_DCT_scaled_size; /* height of a row group of component */ + main_ptr->buffer[ci] = (*cinfo->mem->alloc_sarray) + ((j_common_ptr) cinfo, JPOOL_IMAGE, + compptr->width_in_blocks * compptr->_DCT_scaled_size, + (JDIMENSION) (rgroup * ngroups)); + } +} diff --git a/app/src/main/jni/libjpeg-turbo/jdmarker.c b/app/src/main/jni/libjpeg-turbo/jdmarker.c new file mode 100644 index 000000000..d1357af4c --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdmarker.c @@ -0,0 +1,1376 @@ +/* + * jdmarker.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1998, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright (C) 2012, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains routines to decode JPEG datastream markers. + * Most of the complexity arises from our desire to support input + * suspension: if not all of the data for a marker is available, + * we must exit back to the application. On resumption, we reprocess + * the marker. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" + + +typedef enum { /* JPEG marker codes */ + M_SOF0 = 0xc0, + M_SOF1 = 0xc1, + M_SOF2 = 0xc2, + M_SOF3 = 0xc3, + + M_SOF5 = 0xc5, + M_SOF6 = 0xc6, + M_SOF7 = 0xc7, + + M_JPG = 0xc8, + M_SOF9 = 0xc9, + M_SOF10 = 0xca, + M_SOF11 = 0xcb, + + M_SOF13 = 0xcd, + M_SOF14 = 0xce, + M_SOF15 = 0xcf, + + M_DHT = 0xc4, + + M_DAC = 0xcc, + + M_RST0 = 0xd0, + M_RST1 = 0xd1, + M_RST2 = 0xd2, + M_RST3 = 0xd3, + M_RST4 = 0xd4, + M_RST5 = 0xd5, + M_RST6 = 0xd6, + M_RST7 = 0xd7, + + M_SOI = 0xd8, + M_EOI = 0xd9, + M_SOS = 0xda, + M_DQT = 0xdb, + M_DNL = 0xdc, + M_DRI = 0xdd, + M_DHP = 0xde, + M_EXP = 0xdf, + + M_APP0 = 0xe0, + M_APP1 = 0xe1, + M_APP2 = 0xe2, + M_APP3 = 0xe3, + M_APP4 = 0xe4, + M_APP5 = 0xe5, + M_APP6 = 0xe6, + M_APP7 = 0xe7, + M_APP8 = 0xe8, + M_APP9 = 0xe9, + M_APP10 = 0xea, + M_APP11 = 0xeb, + M_APP12 = 0xec, + M_APP13 = 0xed, + M_APP14 = 0xee, + M_APP15 = 0xef, + + M_JPG0 = 0xf0, + M_JPG13 = 0xfd, + M_COM = 0xfe, + + M_TEM = 0x01, + + M_ERROR = 0x100 +} JPEG_MARKER; + + +/* Private state */ + +typedef struct { + struct jpeg_marker_reader pub; /* public fields */ + + /* Application-overridable marker processing methods */ + jpeg_marker_parser_method process_COM; + jpeg_marker_parser_method process_APPn[16]; + + /* Limit on marker data length to save for each marker type */ + unsigned int length_limit_COM; + unsigned int length_limit_APPn[16]; + + /* Status of COM/APPn marker saving */ + jpeg_saved_marker_ptr cur_marker; /* NULL if not processing a marker */ + unsigned int bytes_read; /* data bytes read so far in marker */ + /* Note: cur_marker is not linked into marker_list until it's all read. */ +} my_marker_reader; + +typedef my_marker_reader * my_marker_ptr; + + +/* + * Macros for fetching data from the data source module. + * + * At all times, cinfo->src->next_input_byte and ->bytes_in_buffer reflect + * the current restart point; we update them only when we have reached a + * suitable place to restart if a suspension occurs. + */ + +/* Declare and initialize local copies of input pointer/count */ +#define INPUT_VARS(cinfo) \ + struct jpeg_source_mgr * datasrc = (cinfo)->src; \ + const JOCTET * next_input_byte = datasrc->next_input_byte; \ + size_t bytes_in_buffer = datasrc->bytes_in_buffer + +/* Unload the local copies --- do this only at a restart boundary */ +#define INPUT_SYNC(cinfo) \ + ( datasrc->next_input_byte = next_input_byte, \ + datasrc->bytes_in_buffer = bytes_in_buffer ) + +/* Reload the local copies --- used only in MAKE_BYTE_AVAIL */ +#define INPUT_RELOAD(cinfo) \ + ( next_input_byte = datasrc->next_input_byte, \ + bytes_in_buffer = datasrc->bytes_in_buffer ) + +/* Internal macro for INPUT_BYTE and INPUT_2BYTES: make a byte available. + * Note we do *not* do INPUT_SYNC before calling fill_input_buffer, + * but we must reload the local copies after a successful fill. + */ +#define MAKE_BYTE_AVAIL(cinfo,action) \ + if (bytes_in_buffer == 0) { \ + if (! (*datasrc->fill_input_buffer) (cinfo)) \ + { action; } \ + INPUT_RELOAD(cinfo); \ + } + +/* Read a byte into variable V. + * If must suspend, take the specified action (typically "return FALSE"). + */ +#define INPUT_BYTE(cinfo,V,action) \ + MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \ + bytes_in_buffer--; \ + V = GETJOCTET(*next_input_byte++); ) + +/* As above, but read two bytes interpreted as an unsigned 16-bit integer. + * V should be declared unsigned int or perhaps INT32. + */ +#define INPUT_2BYTES(cinfo,V,action) \ + MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \ + bytes_in_buffer--; \ + V = ((unsigned int) GETJOCTET(*next_input_byte++)) << 8; \ + MAKE_BYTE_AVAIL(cinfo,action); \ + bytes_in_buffer--; \ + V += GETJOCTET(*next_input_byte++); ) + + +/* + * Routines to process JPEG markers. + * + * Entry condition: JPEG marker itself has been read and its code saved + * in cinfo->unread_marker; input restart point is just after the marker. + * + * Exit: if return TRUE, have read and processed any parameters, and have + * updated the restart point to point after the parameters. + * If return FALSE, was forced to suspend before reaching end of + * marker parameters; restart point has not been moved. Same routine + * will be called again after application supplies more input data. + * + * This approach to suspension assumes that all of a marker's parameters + * can fit into a single input bufferload. This should hold for "normal" + * markers. Some COM/APPn markers might have large parameter segments + * that might not fit. If we are simply dropping such a marker, we use + * skip_input_data to get past it, and thereby put the problem on the + * source manager's shoulders. If we are saving the marker's contents + * into memory, we use a slightly different convention: when forced to + * suspend, the marker processor updates the restart point to the end of + * what it's consumed (ie, the end of the buffer) before returning FALSE. + * On resumption, cinfo->unread_marker still contains the marker code, + * but the data source will point to the next chunk of marker data. + * The marker processor must retain internal state to deal with this. + * + * Note that we don't bother to avoid duplicate trace messages if a + * suspension occurs within marker parameters. Other side effects + * require more care. + */ + + +LOCAL(boolean) +get_soi (j_decompress_ptr cinfo) +/* Process an SOI marker */ +{ + int i; + + TRACEMS(cinfo, 1, JTRC_SOI); + + if (cinfo->marker->saw_SOI) + ERREXIT(cinfo, JERR_SOI_DUPLICATE); + + /* Reset all parameters that are defined to be reset by SOI */ + + for (i = 0; i < NUM_ARITH_TBLS; i++) { + cinfo->arith_dc_L[i] = 0; + cinfo->arith_dc_U[i] = 1; + cinfo->arith_ac_K[i] = 5; + } + cinfo->restart_interval = 0; + + /* Set initial assumptions for colorspace etc */ + + cinfo->jpeg_color_space = JCS_UNKNOWN; + cinfo->CCIR601_sampling = FALSE; /* Assume non-CCIR sampling??? */ + + cinfo->saw_JFIF_marker = FALSE; + cinfo->JFIF_major_version = 1; /* set default JFIF APP0 values */ + cinfo->JFIF_minor_version = 1; + cinfo->density_unit = 0; + cinfo->X_density = 1; + cinfo->Y_density = 1; + cinfo->saw_Adobe_marker = FALSE; + cinfo->Adobe_transform = 0; + + cinfo->marker->saw_SOI = TRUE; + + return TRUE; +} + + +LOCAL(boolean) +get_sof (j_decompress_ptr cinfo, boolean is_prog, boolean is_arith) +/* Process a SOFn marker */ +{ + INT32 length; + int c, ci; + jpeg_component_info * compptr; + INPUT_VARS(cinfo); + + cinfo->progressive_mode = is_prog; + cinfo->arith_code = is_arith; + + INPUT_2BYTES(cinfo, length, return FALSE); + + INPUT_BYTE(cinfo, cinfo->data_precision, return FALSE); + INPUT_2BYTES(cinfo, cinfo->image_height, return FALSE); + INPUT_2BYTES(cinfo, cinfo->image_width, return FALSE); + INPUT_BYTE(cinfo, cinfo->num_components, return FALSE); + + length -= 8; + + TRACEMS4(cinfo, 1, JTRC_SOF, cinfo->unread_marker, + (int) cinfo->image_width, (int) cinfo->image_height, + cinfo->num_components); + + if (cinfo->marker->saw_SOF) + ERREXIT(cinfo, JERR_SOF_DUPLICATE); + + /* We don't support files in which the image height is initially specified */ + /* as 0 and is later redefined by DNL. As long as we have to check that, */ + /* might as well have a general sanity check. */ + if (cinfo->image_height <= 0 || cinfo->image_width <= 0 + || cinfo->num_components <= 0) + ERREXIT(cinfo, JERR_EMPTY_IMAGE); + + if (length != (cinfo->num_components * 3)) + ERREXIT(cinfo, JERR_BAD_LENGTH); + + if (cinfo->comp_info == NULL) /* do only once, even if suspend */ + cinfo->comp_info = (jpeg_component_info *) (*cinfo->mem->alloc_small) + ((j_common_ptr) cinfo, JPOOL_IMAGE, + cinfo->num_components * sizeof(jpeg_component_info)); + + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + compptr->component_index = ci; + INPUT_BYTE(cinfo, compptr->component_id, return FALSE); + INPUT_BYTE(cinfo, c, return FALSE); + compptr->h_samp_factor = (c >> 4) & 15; + compptr->v_samp_factor = (c ) & 15; + INPUT_BYTE(cinfo, compptr->quant_tbl_no, return FALSE); + + TRACEMS4(cinfo, 1, JTRC_SOF_COMPONENT, + compptr->component_id, compptr->h_samp_factor, + compptr->v_samp_factor, compptr->quant_tbl_no); + } + + cinfo->marker->saw_SOF = TRUE; + + INPUT_SYNC(cinfo); + return TRUE; +} + + +LOCAL(boolean) +get_sos (j_decompress_ptr cinfo) +/* Process a SOS marker */ +{ + INT32 length; + int i, ci, n, c, cc, pi; + jpeg_component_info * compptr; + INPUT_VARS(cinfo); + + if (! cinfo->marker->saw_SOF) + ERREXIT(cinfo, JERR_SOS_NO_SOF); + + INPUT_2BYTES(cinfo, length, return FALSE); + + INPUT_BYTE(cinfo, n, return FALSE); /* Number of components */ + + TRACEMS1(cinfo, 1, JTRC_SOS, n); + + if (length != (n * 2 + 6) || n < 1 || n > MAX_COMPS_IN_SCAN) + ERREXIT(cinfo, JERR_BAD_LENGTH); + + cinfo->comps_in_scan = n; + + /* Collect the component-spec parameters */ + + for (i = 0; i < MAX_COMPS_IN_SCAN; i++) + cinfo->cur_comp_info[i] = NULL; + + for (i = 0; i < n; i++) { + INPUT_BYTE(cinfo, cc, return FALSE); + INPUT_BYTE(cinfo, c, return FALSE); + + for (ci = 0, compptr = cinfo->comp_info; + ci < cinfo->num_components && ci < MAX_COMPS_IN_SCAN; + ci++, compptr++) { + if (cc == compptr->component_id && !cinfo->cur_comp_info[ci]) + goto id_found; + } + + ERREXIT1(cinfo, JERR_BAD_COMPONENT_ID, cc); + + id_found: + + cinfo->cur_comp_info[i] = compptr; + compptr->dc_tbl_no = (c >> 4) & 15; + compptr->ac_tbl_no = (c ) & 15; + + TRACEMS3(cinfo, 1, JTRC_SOS_COMPONENT, cc, + compptr->dc_tbl_no, compptr->ac_tbl_no); + + /* This CSi (cc) should differ from the previous CSi */ + for (pi = 0; pi < i; pi++) { + if (cinfo->cur_comp_info[pi] == compptr) { + ERREXIT1(cinfo, JERR_BAD_COMPONENT_ID, cc); + } + } + } + + /* Collect the additional scan parameters Ss, Se, Ah/Al. */ + INPUT_BYTE(cinfo, c, return FALSE); + cinfo->Ss = c; + INPUT_BYTE(cinfo, c, return FALSE); + cinfo->Se = c; + INPUT_BYTE(cinfo, c, return FALSE); + cinfo->Ah = (c >> 4) & 15; + cinfo->Al = (c ) & 15; + + TRACEMS4(cinfo, 1, JTRC_SOS_PARAMS, cinfo->Ss, cinfo->Se, + cinfo->Ah, cinfo->Al); + + /* Prepare to scan data & restart markers */ + cinfo->marker->next_restart_num = 0; + + /* Count another SOS marker */ + cinfo->input_scan_number++; + + INPUT_SYNC(cinfo); + return TRUE; +} + + +#ifdef D_ARITH_CODING_SUPPORTED + +LOCAL(boolean) +get_dac (j_decompress_ptr cinfo) +/* Process a DAC marker */ +{ + INT32 length; + int index, val; + INPUT_VARS(cinfo); + + INPUT_2BYTES(cinfo, length, return FALSE); + length -= 2; + + while (length > 0) { + INPUT_BYTE(cinfo, index, return FALSE); + INPUT_BYTE(cinfo, val, return FALSE); + + length -= 2; + + TRACEMS2(cinfo, 1, JTRC_DAC, index, val); + + if (index < 0 || index >= (2*NUM_ARITH_TBLS)) + ERREXIT1(cinfo, JERR_DAC_INDEX, index); + + if (index >= NUM_ARITH_TBLS) { /* define AC table */ + cinfo->arith_ac_K[index-NUM_ARITH_TBLS] = (UINT8) val; + } else { /* define DC table */ + cinfo->arith_dc_L[index] = (UINT8) (val & 0x0F); + cinfo->arith_dc_U[index] = (UINT8) (val >> 4); + if (cinfo->arith_dc_L[index] > cinfo->arith_dc_U[index]) + ERREXIT1(cinfo, JERR_DAC_VALUE, val); + } + } + + if (length != 0) + ERREXIT(cinfo, JERR_BAD_LENGTH); + + INPUT_SYNC(cinfo); + return TRUE; +} + +#else /* ! D_ARITH_CODING_SUPPORTED */ + +#define get_dac(cinfo) skip_variable(cinfo) + +#endif /* D_ARITH_CODING_SUPPORTED */ + + +LOCAL(boolean) +get_dht (j_decompress_ptr cinfo) +/* Process a DHT marker */ +{ + INT32 length; + UINT8 bits[17]; + UINT8 huffval[256]; + int i, index, count; + JHUFF_TBL **htblptr; + INPUT_VARS(cinfo); + + INPUT_2BYTES(cinfo, length, return FALSE); + length -= 2; + + while (length > 16) { + INPUT_BYTE(cinfo, index, return FALSE); + + TRACEMS1(cinfo, 1, JTRC_DHT, index); + + bits[0] = 0; + count = 0; + for (i = 1; i <= 16; i++) { + INPUT_BYTE(cinfo, bits[i], return FALSE); + count += bits[i]; + } + + length -= 1 + 16; + + TRACEMS8(cinfo, 2, JTRC_HUFFBITS, + bits[1], bits[2], bits[3], bits[4], + bits[5], bits[6], bits[7], bits[8]); + TRACEMS8(cinfo, 2, JTRC_HUFFBITS, + bits[9], bits[10], bits[11], bits[12], + bits[13], bits[14], bits[15], bits[16]); + + /* Here we just do minimal validation of the counts to avoid walking + * off the end of our table space. jdhuff.c will check more carefully. + */ + if (count > 256 || ((INT32) count) > length) + ERREXIT(cinfo, JERR_BAD_HUFF_TABLE); + + for (i = 0; i < count; i++) + INPUT_BYTE(cinfo, huffval[i], return FALSE); + + MEMZERO(&huffval[count], (256 - count) * sizeof(UINT8)); + + length -= count; + + if (index & 0x10) { /* AC table definition */ + index -= 0x10; + if (index < 0 || index >= NUM_HUFF_TBLS) + ERREXIT1(cinfo, JERR_DHT_INDEX, index); + htblptr = &cinfo->ac_huff_tbl_ptrs[index]; + } else { /* DC table definition */ + if (index < 0 || index >= NUM_HUFF_TBLS) + ERREXIT1(cinfo, JERR_DHT_INDEX, index); + htblptr = &cinfo->dc_huff_tbl_ptrs[index]; + } + + if (*htblptr == NULL) + *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo); + + MEMCOPY((*htblptr)->bits, bits, sizeof((*htblptr)->bits)); + MEMCOPY((*htblptr)->huffval, huffval, sizeof((*htblptr)->huffval)); + } + + if (length != 0) + ERREXIT(cinfo, JERR_BAD_LENGTH); + + INPUT_SYNC(cinfo); + return TRUE; +} + + +LOCAL(boolean) +get_dqt (j_decompress_ptr cinfo) +/* Process a DQT marker */ +{ + INT32 length; + int n, i, prec; + unsigned int tmp; + JQUANT_TBL *quant_ptr; + INPUT_VARS(cinfo); + + INPUT_2BYTES(cinfo, length, return FALSE); + length -= 2; + + while (length > 0) { + INPUT_BYTE(cinfo, n, return FALSE); + prec = n >> 4; + n &= 0x0F; + + TRACEMS2(cinfo, 1, JTRC_DQT, n, prec); + + if (n >= NUM_QUANT_TBLS) + ERREXIT1(cinfo, JERR_DQT_INDEX, n); + + if (cinfo->quant_tbl_ptrs[n] == NULL) + cinfo->quant_tbl_ptrs[n] = jpeg_alloc_quant_table((j_common_ptr) cinfo); + quant_ptr = cinfo->quant_tbl_ptrs[n]; + + for (i = 0; i < DCTSIZE2; i++) { + if (prec) + INPUT_2BYTES(cinfo, tmp, return FALSE); + else + INPUT_BYTE(cinfo, tmp, return FALSE); + /* We convert the zigzag-order table to natural array order. */ + quant_ptr->quantval[jpeg_natural_order[i]] = (UINT16) tmp; + } + + if (cinfo->err->trace_level >= 2) { + for (i = 0; i < DCTSIZE2; i += 8) { + TRACEMS8(cinfo, 2, JTRC_QUANTVALS, + quant_ptr->quantval[i], quant_ptr->quantval[i+1], + quant_ptr->quantval[i+2], quant_ptr->quantval[i+3], + quant_ptr->quantval[i+4], quant_ptr->quantval[i+5], + quant_ptr->quantval[i+6], quant_ptr->quantval[i+7]); + } + } + + length -= DCTSIZE2+1; + if (prec) length -= DCTSIZE2; + } + + if (length != 0) + ERREXIT(cinfo, JERR_BAD_LENGTH); + + INPUT_SYNC(cinfo); + return TRUE; +} + + +LOCAL(boolean) +get_dri (j_decompress_ptr cinfo) +/* Process a DRI marker */ +{ + INT32 length; + unsigned int tmp; + INPUT_VARS(cinfo); + + INPUT_2BYTES(cinfo, length, return FALSE); + + if (length != 4) + ERREXIT(cinfo, JERR_BAD_LENGTH); + + INPUT_2BYTES(cinfo, tmp, return FALSE); + + TRACEMS1(cinfo, 1, JTRC_DRI, tmp); + + cinfo->restart_interval = tmp; + + INPUT_SYNC(cinfo); + return TRUE; +} + + +/* + * Routines for processing APPn and COM markers. + * These are either saved in memory or discarded, per application request. + * APP0 and APP14 are specially checked to see if they are + * JFIF and Adobe markers, respectively. + */ + +#define APP0_DATA_LEN 14 /* Length of interesting data in APP0 */ +#define APP14_DATA_LEN 12 /* Length of interesting data in APP14 */ +#define APPN_DATA_LEN 14 /* Must be the largest of the above!! */ + + +LOCAL(void) +examine_app0 (j_decompress_ptr cinfo, JOCTET * data, + unsigned int datalen, INT32 remaining) +/* Examine first few bytes from an APP0. + * Take appropriate action if it is a JFIF marker. + * datalen is # of bytes at data[], remaining is length of rest of marker data. + */ +{ + INT32 totallen = (INT32) datalen + remaining; + + if (datalen >= APP0_DATA_LEN && + GETJOCTET(data[0]) == 0x4A && + GETJOCTET(data[1]) == 0x46 && + GETJOCTET(data[2]) == 0x49 && + GETJOCTET(data[3]) == 0x46 && + GETJOCTET(data[4]) == 0) { + /* Found JFIF APP0 marker: save info */ + cinfo->saw_JFIF_marker = TRUE; + cinfo->JFIF_major_version = GETJOCTET(data[5]); + cinfo->JFIF_minor_version = GETJOCTET(data[6]); + cinfo->density_unit = GETJOCTET(data[7]); + cinfo->X_density = (GETJOCTET(data[8]) << 8) + GETJOCTET(data[9]); + cinfo->Y_density = (GETJOCTET(data[10]) << 8) + GETJOCTET(data[11]); + /* Check version. + * Major version must be 1, anything else signals an incompatible change. + * (We used to treat this as an error, but now it's a nonfatal warning, + * because some bozo at Hijaak couldn't read the spec.) + * Minor version should be 0..2, but process anyway if newer. + */ + if (cinfo->JFIF_major_version != 1) + WARNMS2(cinfo, JWRN_JFIF_MAJOR, + cinfo->JFIF_major_version, cinfo->JFIF_minor_version); + /* Generate trace messages */ + TRACEMS5(cinfo, 1, JTRC_JFIF, + cinfo->JFIF_major_version, cinfo->JFIF_minor_version, + cinfo->X_density, cinfo->Y_density, cinfo->density_unit); + /* Validate thumbnail dimensions and issue appropriate messages */ + if (GETJOCTET(data[12]) | GETJOCTET(data[13])) + TRACEMS2(cinfo, 1, JTRC_JFIF_THUMBNAIL, + GETJOCTET(data[12]), GETJOCTET(data[13])); + totallen -= APP0_DATA_LEN; + if (totallen != + ((INT32)GETJOCTET(data[12]) * (INT32)GETJOCTET(data[13]) * (INT32) 3)) + TRACEMS1(cinfo, 1, JTRC_JFIF_BADTHUMBNAILSIZE, (int) totallen); + } else if (datalen >= 6 && + GETJOCTET(data[0]) == 0x4A && + GETJOCTET(data[1]) == 0x46 && + GETJOCTET(data[2]) == 0x58 && + GETJOCTET(data[3]) == 0x58 && + GETJOCTET(data[4]) == 0) { + /* Found JFIF "JFXX" extension APP0 marker */ + /* The library doesn't actually do anything with these, + * but we try to produce a helpful trace message. + */ + switch (GETJOCTET(data[5])) { + case 0x10: + TRACEMS1(cinfo, 1, JTRC_THUMB_JPEG, (int) totallen); + break; + case 0x11: + TRACEMS1(cinfo, 1, JTRC_THUMB_PALETTE, (int) totallen); + break; + case 0x13: + TRACEMS1(cinfo, 1, JTRC_THUMB_RGB, (int) totallen); + break; + default: + TRACEMS2(cinfo, 1, JTRC_JFIF_EXTENSION, + GETJOCTET(data[5]), (int) totallen); + break; + } + } else { + /* Start of APP0 does not match "JFIF" or "JFXX", or too short */ + TRACEMS1(cinfo, 1, JTRC_APP0, (int) totallen); + } +} + + +LOCAL(void) +examine_app14 (j_decompress_ptr cinfo, JOCTET * data, + unsigned int datalen, INT32 remaining) +/* Examine first few bytes from an APP14. + * Take appropriate action if it is an Adobe marker. + * datalen is # of bytes at data[], remaining is length of rest of marker data. + */ +{ + unsigned int version, flags0, flags1, transform; + + if (datalen >= APP14_DATA_LEN && + GETJOCTET(data[0]) == 0x41 && + GETJOCTET(data[1]) == 0x64 && + GETJOCTET(data[2]) == 0x6F && + GETJOCTET(data[3]) == 0x62 && + GETJOCTET(data[4]) == 0x65) { + /* Found Adobe APP14 marker */ + version = (GETJOCTET(data[5]) << 8) + GETJOCTET(data[6]); + flags0 = (GETJOCTET(data[7]) << 8) + GETJOCTET(data[8]); + flags1 = (GETJOCTET(data[9]) << 8) + GETJOCTET(data[10]); + transform = GETJOCTET(data[11]); + TRACEMS4(cinfo, 1, JTRC_ADOBE, version, flags0, flags1, transform); + cinfo->saw_Adobe_marker = TRUE; + cinfo->Adobe_transform = (UINT8) transform; + } else { + /* Start of APP14 does not match "Adobe", or too short */ + TRACEMS1(cinfo, 1, JTRC_APP14, (int) (datalen + remaining)); + } +} + + +METHODDEF(boolean) +get_interesting_appn (j_decompress_ptr cinfo) +/* Process an APP0 or APP14 marker without saving it */ +{ + INT32 length; + JOCTET b[APPN_DATA_LEN]; + unsigned int i, numtoread; + INPUT_VARS(cinfo); + + INPUT_2BYTES(cinfo, length, return FALSE); + length -= 2; + + /* get the interesting part of the marker data */ + if (length >= APPN_DATA_LEN) + numtoread = APPN_DATA_LEN; + else if (length > 0) + numtoread = (unsigned int) length; + else + numtoread = 0; + for (i = 0; i < numtoread; i++) + INPUT_BYTE(cinfo, b[i], return FALSE); + length -= numtoread; + + /* process it */ + switch (cinfo->unread_marker) { + case M_APP0: + examine_app0(cinfo, (JOCTET *) b, numtoread, length); + break; + case M_APP14: + examine_app14(cinfo, (JOCTET *) b, numtoread, length); + break; + default: + /* can't get here unless jpeg_save_markers chooses wrong processor */ + ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, cinfo->unread_marker); + break; + } + + /* skip any remaining data -- could be lots */ + INPUT_SYNC(cinfo); + if (length > 0) + (*cinfo->src->skip_input_data) (cinfo, (long) length); + + return TRUE; +} + + +#ifdef SAVE_MARKERS_SUPPORTED + +METHODDEF(boolean) +save_marker (j_decompress_ptr cinfo) +/* Save an APPn or COM marker into the marker list */ +{ + my_marker_ptr marker = (my_marker_ptr) cinfo->marker; + jpeg_saved_marker_ptr cur_marker = marker->cur_marker; + unsigned int bytes_read, data_length; + JOCTET * data; + INT32 length = 0; + INPUT_VARS(cinfo); + + if (cur_marker == NULL) { + /* begin reading a marker */ + INPUT_2BYTES(cinfo, length, return FALSE); + length -= 2; + if (length >= 0) { /* watch out for bogus length word */ + /* figure out how much we want to save */ + unsigned int limit; + if (cinfo->unread_marker == (int) M_COM) + limit = marker->length_limit_COM; + else + limit = marker->length_limit_APPn[cinfo->unread_marker - (int) M_APP0]; + if ((unsigned int) length < limit) + limit = (unsigned int) length; + /* allocate and initialize the marker item */ + cur_marker = (jpeg_saved_marker_ptr) + (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(struct jpeg_marker_struct) + limit); + cur_marker->next = NULL; + cur_marker->marker = (UINT8) cinfo->unread_marker; + cur_marker->original_length = (unsigned int) length; + cur_marker->data_length = limit; + /* data area is just beyond the jpeg_marker_struct */ + data = cur_marker->data = (JOCTET *) (cur_marker + 1); + marker->cur_marker = cur_marker; + marker->bytes_read = 0; + bytes_read = 0; + data_length = limit; + } else { + /* deal with bogus length word */ + bytes_read = data_length = 0; + data = NULL; + } + } else { + /* resume reading a marker */ + bytes_read = marker->bytes_read; + data_length = cur_marker->data_length; + data = cur_marker->data + bytes_read; + } + + while (bytes_read < data_length) { + INPUT_SYNC(cinfo); /* move the restart point to here */ + marker->bytes_read = bytes_read; + /* If there's not at least one byte in buffer, suspend */ + MAKE_BYTE_AVAIL(cinfo, return FALSE); + /* Copy bytes with reasonable rapidity */ + while (bytes_read < data_length && bytes_in_buffer > 0) { + *data++ = *next_input_byte++; + bytes_in_buffer--; + bytes_read++; + } + } + + /* Done reading what we want to read */ + if (cur_marker != NULL) { /* will be NULL if bogus length word */ + /* Add new marker to end of list */ + if (cinfo->marker_list == NULL) { + cinfo->marker_list = cur_marker; + } else { + jpeg_saved_marker_ptr prev = cinfo->marker_list; + while (prev->next != NULL) + prev = prev->next; + prev->next = cur_marker; + } + /* Reset pointer & calc remaining data length */ + data = cur_marker->data; + length = cur_marker->original_length - data_length; + } + /* Reset to initial state for next marker */ + marker->cur_marker = NULL; + + /* Process the marker if interesting; else just make a generic trace msg */ + switch (cinfo->unread_marker) { + case M_APP0: + examine_app0(cinfo, data, data_length, length); + break; + case M_APP14: + examine_app14(cinfo, data, data_length, length); + break; + default: + TRACEMS2(cinfo, 1, JTRC_MISC_MARKER, cinfo->unread_marker, + (int) (data_length + length)); + break; + } + + /* skip any remaining data -- could be lots */ + INPUT_SYNC(cinfo); /* do before skip_input_data */ + if (length > 0) + (*cinfo->src->skip_input_data) (cinfo, (long) length); + + return TRUE; +} + +#endif /* SAVE_MARKERS_SUPPORTED */ + + +METHODDEF(boolean) +skip_variable (j_decompress_ptr cinfo) +/* Skip over an unknown or uninteresting variable-length marker */ +{ + INT32 length; + INPUT_VARS(cinfo); + + INPUT_2BYTES(cinfo, length, return FALSE); + length -= 2; + + TRACEMS2(cinfo, 1, JTRC_MISC_MARKER, cinfo->unread_marker, (int) length); + + INPUT_SYNC(cinfo); /* do before skip_input_data */ + if (length > 0) + (*cinfo->src->skip_input_data) (cinfo, (long) length); + + return TRUE; +} + + +/* + * Find the next JPEG marker, save it in cinfo->unread_marker. + * Returns FALSE if had to suspend before reaching a marker; + * in that case cinfo->unread_marker is unchanged. + * + * Note that the result might not be a valid marker code, + * but it will never be 0 or FF. + */ + +LOCAL(boolean) +next_marker (j_decompress_ptr cinfo) +{ + int c; + INPUT_VARS(cinfo); + + for (;;) { + INPUT_BYTE(cinfo, c, return FALSE); + /* Skip any non-FF bytes. + * This may look a bit inefficient, but it will not occur in a valid file. + * We sync after each discarded byte so that a suspending data source + * can discard the byte from its buffer. + */ + while (c != 0xFF) { + cinfo->marker->discarded_bytes++; + INPUT_SYNC(cinfo); + INPUT_BYTE(cinfo, c, return FALSE); + } + /* This loop swallows any duplicate FF bytes. Extra FFs are legal as + * pad bytes, so don't count them in discarded_bytes. We assume there + * will not be so many consecutive FF bytes as to overflow a suspending + * data source's input buffer. + */ + do { + INPUT_BYTE(cinfo, c, return FALSE); + } while (c == 0xFF); + if (c != 0) + break; /* found a valid marker, exit loop */ + /* Reach here if we found a stuffed-zero data sequence (FF/00). + * Discard it and loop back to try again. + */ + cinfo->marker->discarded_bytes += 2; + INPUT_SYNC(cinfo); + } + + if (cinfo->marker->discarded_bytes != 0) { + WARNMS2(cinfo, JWRN_EXTRANEOUS_DATA, cinfo->marker->discarded_bytes, c); + cinfo->marker->discarded_bytes = 0; + } + + cinfo->unread_marker = c; + + INPUT_SYNC(cinfo); + return TRUE; +} + + +LOCAL(boolean) +first_marker (j_decompress_ptr cinfo) +/* Like next_marker, but used to obtain the initial SOI marker. */ +/* For this marker, we do not allow preceding garbage or fill; otherwise, + * we might well scan an entire input file before realizing it ain't JPEG. + * If an application wants to process non-JFIF files, it must seek to the + * SOI before calling the JPEG library. + */ +{ + int c, c2; + INPUT_VARS(cinfo); + + INPUT_BYTE(cinfo, c, return FALSE); + INPUT_BYTE(cinfo, c2, return FALSE); + if (c != 0xFF || c2 != (int) M_SOI) + ERREXIT2(cinfo, JERR_NO_SOI, c, c2); + + cinfo->unread_marker = c2; + + INPUT_SYNC(cinfo); + return TRUE; +} + + +/* + * Read markers until SOS or EOI. + * + * Returns same codes as are defined for jpeg_consume_input: + * JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI. + */ + +METHODDEF(int) +read_markers (j_decompress_ptr cinfo) +{ + /* Outer loop repeats once for each marker. */ + for (;;) { + /* Collect the marker proper, unless we already did. */ + /* NB: first_marker() enforces the requirement that SOI appear first. */ + if (cinfo->unread_marker == 0) { + if (! cinfo->marker->saw_SOI) { + if (! first_marker(cinfo)) + return JPEG_SUSPENDED; + } else { + if (! next_marker(cinfo)) + return JPEG_SUSPENDED; + } + } + /* At this point cinfo->unread_marker contains the marker code and the + * input point is just past the marker proper, but before any parameters. + * A suspension will cause us to return with this state still true. + */ + switch (cinfo->unread_marker) { + case M_SOI: + if (! get_soi(cinfo)) + return JPEG_SUSPENDED; + break; + + case M_SOF0: /* Baseline */ + case M_SOF1: /* Extended sequential, Huffman */ + if (! get_sof(cinfo, FALSE, FALSE)) + return JPEG_SUSPENDED; + break; + + case M_SOF2: /* Progressive, Huffman */ + if (! get_sof(cinfo, TRUE, FALSE)) + return JPEG_SUSPENDED; + break; + + case M_SOF9: /* Extended sequential, arithmetic */ + if (! get_sof(cinfo, FALSE, TRUE)) + return JPEG_SUSPENDED; + break; + + case M_SOF10: /* Progressive, arithmetic */ + if (! get_sof(cinfo, TRUE, TRUE)) + return JPEG_SUSPENDED; + break; + + /* Currently unsupported SOFn types */ + case M_SOF3: /* Lossless, Huffman */ + case M_SOF5: /* Differential sequential, Huffman */ + case M_SOF6: /* Differential progressive, Huffman */ + case M_SOF7: /* Differential lossless, Huffman */ + case M_JPG: /* Reserved for JPEG extensions */ + case M_SOF11: /* Lossless, arithmetic */ + case M_SOF13: /* Differential sequential, arithmetic */ + case M_SOF14: /* Differential progressive, arithmetic */ + case M_SOF15: /* Differential lossless, arithmetic */ + ERREXIT1(cinfo, JERR_SOF_UNSUPPORTED, cinfo->unread_marker); + break; + + case M_SOS: + if (! get_sos(cinfo)) + return JPEG_SUSPENDED; + cinfo->unread_marker = 0; /* processed the marker */ + return JPEG_REACHED_SOS; + + case M_EOI: + TRACEMS(cinfo, 1, JTRC_EOI); + cinfo->unread_marker = 0; /* processed the marker */ + return JPEG_REACHED_EOI; + + case M_DAC: + if (! get_dac(cinfo)) + return JPEG_SUSPENDED; + break; + + case M_DHT: + if (! get_dht(cinfo)) + return JPEG_SUSPENDED; + break; + + case M_DQT: + if (! get_dqt(cinfo)) + return JPEG_SUSPENDED; + break; + + case M_DRI: + if (! get_dri(cinfo)) + return JPEG_SUSPENDED; + break; + + case M_APP0: + case M_APP1: + case M_APP2: + case M_APP3: + case M_APP4: + case M_APP5: + case M_APP6: + case M_APP7: + case M_APP8: + case M_APP9: + case M_APP10: + case M_APP11: + case M_APP12: + case M_APP13: + case M_APP14: + case M_APP15: + if (! (*((my_marker_ptr) cinfo->marker)->process_APPn[ + cinfo->unread_marker - (int) M_APP0]) (cinfo)) + return JPEG_SUSPENDED; + break; + + case M_COM: + if (! (*((my_marker_ptr) cinfo->marker)->process_COM) (cinfo)) + return JPEG_SUSPENDED; + break; + + case M_RST0: /* these are all parameterless */ + case M_RST1: + case M_RST2: + case M_RST3: + case M_RST4: + case M_RST5: + case M_RST6: + case M_RST7: + case M_TEM: + TRACEMS1(cinfo, 1, JTRC_PARMLESS_MARKER, cinfo->unread_marker); + break; + + case M_DNL: /* Ignore DNL ... perhaps the wrong thing */ + if (! skip_variable(cinfo)) + return JPEG_SUSPENDED; + break; + + default: /* must be DHP, EXP, JPGn, or RESn */ + /* For now, we treat the reserved markers as fatal errors since they are + * likely to be used to signal incompatible JPEG Part 3 extensions. + * Once the JPEG 3 version-number marker is well defined, this code + * ought to change! + */ + ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, cinfo->unread_marker); + break; + } + /* Successfully processed marker, so reset state variable */ + cinfo->unread_marker = 0; + } /* end loop */ +} + + +/* + * Read a restart marker, which is expected to appear next in the datastream; + * if the marker is not there, take appropriate recovery action. + * Returns FALSE if suspension is required. + * + * This is called by the entropy decoder after it has read an appropriate + * number of MCUs. cinfo->unread_marker may be nonzero if the entropy decoder + * has already read a marker from the data source. Under normal conditions + * cinfo->unread_marker will be reset to 0 before returning; if not reset, + * it holds a marker which the decoder will be unable to read past. + */ + +METHODDEF(boolean) +read_restart_marker (j_decompress_ptr cinfo) +{ + /* Obtain a marker unless we already did. */ + /* Note that next_marker will complain if it skips any data. */ + if (cinfo->unread_marker == 0) { + if (! next_marker(cinfo)) + return FALSE; + } + + if (cinfo->unread_marker == + ((int) M_RST0 + cinfo->marker->next_restart_num)) { + /* Normal case --- swallow the marker and let entropy decoder continue */ + TRACEMS1(cinfo, 3, JTRC_RST, cinfo->marker->next_restart_num); + cinfo->unread_marker = 0; + } else { + /* Uh-oh, the restart markers have been messed up. */ + /* Let the data source manager determine how to resync. */ + if (! (*cinfo->src->resync_to_restart) (cinfo, + cinfo->marker->next_restart_num)) + return FALSE; + } + + /* Update next-restart state */ + cinfo->marker->next_restart_num = (cinfo->marker->next_restart_num + 1) & 7; + + return TRUE; +} + + +/* + * This is the default resync_to_restart method for data source managers + * to use if they don't have any better approach. Some data source managers + * may be able to back up, or may have additional knowledge about the data + * which permits a more intelligent recovery strategy; such managers would + * presumably supply their own resync method. + * + * read_restart_marker calls resync_to_restart if it finds a marker other than + * the restart marker it was expecting. (This code is *not* used unless + * a nonzero restart interval has been declared.) cinfo->unread_marker is + * the marker code actually found (might be anything, except 0 or FF). + * The desired restart marker number (0..7) is passed as a parameter. + * This routine is supposed to apply whatever error recovery strategy seems + * appropriate in order to position the input stream to the next data segment. + * Note that cinfo->unread_marker is treated as a marker appearing before + * the current data-source input point; usually it should be reset to zero + * before returning. + * Returns FALSE if suspension is required. + * + * This implementation is substantially constrained by wanting to treat the + * input as a data stream; this means we can't back up. Therefore, we have + * only the following actions to work with: + * 1. Simply discard the marker and let the entropy decoder resume at next + * byte of file. + * 2. Read forward until we find another marker, discarding intervening + * data. (In theory we could look ahead within the current bufferload, + * without having to discard data if we don't find the desired marker. + * This idea is not implemented here, in part because it makes behavior + * dependent on buffer size and chance buffer-boundary positions.) + * 3. Leave the marker unread (by failing to zero cinfo->unread_marker). + * This will cause the entropy decoder to process an empty data segment, + * inserting dummy zeroes, and then we will reprocess the marker. + * + * #2 is appropriate if we think the desired marker lies ahead, while #3 is + * appropriate if the found marker is a future restart marker (indicating + * that we have missed the desired restart marker, probably because it got + * corrupted). + * We apply #2 or #3 if the found marker is a restart marker no more than + * two counts behind or ahead of the expected one. We also apply #2 if the + * found marker is not a legal JPEG marker code (it's certainly bogus data). + * If the found marker is a restart marker more than 2 counts away, we do #1 + * (too much risk that the marker is erroneous; with luck we will be able to + * resync at some future point). + * For any valid non-restart JPEG marker, we apply #3. This keeps us from + * overrunning the end of a scan. An implementation limited to single-scan + * files might find it better to apply #2 for markers other than EOI, since + * any other marker would have to be bogus data in that case. + */ + +GLOBAL(boolean) +jpeg_resync_to_restart (j_decompress_ptr cinfo, int desired) +{ + int marker = cinfo->unread_marker; + int action = 1; + + /* Always put up a warning. */ + WARNMS2(cinfo, JWRN_MUST_RESYNC, marker, desired); + + /* Outer loop handles repeated decision after scanning forward. */ + for (;;) { + if (marker < (int) M_SOF0) + action = 2; /* invalid marker */ + else if (marker < (int) M_RST0 || marker > (int) M_RST7) + action = 3; /* valid non-restart marker */ + else { + if (marker == ((int) M_RST0 + ((desired+1) & 7)) || + marker == ((int) M_RST0 + ((desired+2) & 7))) + action = 3; /* one of the next two expected restarts */ + else if (marker == ((int) M_RST0 + ((desired-1) & 7)) || + marker == ((int) M_RST0 + ((desired-2) & 7))) + action = 2; /* a prior restart, so advance */ + else + action = 1; /* desired restart or too far away */ + } + TRACEMS2(cinfo, 4, JTRC_RECOVERY_ACTION, marker, action); + switch (action) { + case 1: + /* Discard marker and let entropy decoder resume processing. */ + cinfo->unread_marker = 0; + return TRUE; + case 2: + /* Scan to the next marker, and repeat the decision loop. */ + if (! next_marker(cinfo)) + return FALSE; + marker = cinfo->unread_marker; + break; + case 3: + /* Return without advancing past this marker. */ + /* Entropy decoder will be forced to process an empty segment. */ + return TRUE; + } + } /* end loop */ +} + + +/* + * Reset marker processing state to begin a fresh datastream. + */ + +METHODDEF(void) +reset_marker_reader (j_decompress_ptr cinfo) +{ + my_marker_ptr marker = (my_marker_ptr) cinfo->marker; + + cinfo->comp_info = NULL; /* until allocated by get_sof */ + cinfo->input_scan_number = 0; /* no SOS seen yet */ + cinfo->unread_marker = 0; /* no pending marker */ + marker->pub.saw_SOI = FALSE; /* set internal state too */ + marker->pub.saw_SOF = FALSE; + marker->pub.discarded_bytes = 0; + marker->cur_marker = NULL; +} + + +/* + * Initialize the marker reader module. + * This is called only once, when the decompression object is created. + */ + +GLOBAL(void) +jinit_marker_reader (j_decompress_ptr cinfo) +{ + my_marker_ptr marker; + int i; + + /* Create subobject in permanent pool */ + marker = (my_marker_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT, + sizeof(my_marker_reader)); + cinfo->marker = (struct jpeg_marker_reader *) marker; + /* Initialize public method pointers */ + marker->pub.reset_marker_reader = reset_marker_reader; + marker->pub.read_markers = read_markers; + marker->pub.read_restart_marker = read_restart_marker; + /* Initialize COM/APPn processing. + * By default, we examine and then discard APP0 and APP14, + * but simply discard COM and all other APPn. + */ + marker->process_COM = skip_variable; + marker->length_limit_COM = 0; + for (i = 0; i < 16; i++) { + marker->process_APPn[i] = skip_variable; + marker->length_limit_APPn[i] = 0; + } + marker->process_APPn[0] = get_interesting_appn; + marker->process_APPn[14] = get_interesting_appn; + /* Reset marker processing state */ + reset_marker_reader(cinfo); +} + + +/* + * Control saving of COM and APPn markers into marker_list. + */ + +#ifdef SAVE_MARKERS_SUPPORTED + +GLOBAL(void) +jpeg_save_markers (j_decompress_ptr cinfo, int marker_code, + unsigned int length_limit) +{ + my_marker_ptr marker = (my_marker_ptr) cinfo->marker; + long maxlength; + jpeg_marker_parser_method processor; + + /* Length limit mustn't be larger than what we can allocate + * (should only be a concern in a 16-bit environment). + */ + maxlength = cinfo->mem->max_alloc_chunk - sizeof(struct jpeg_marker_struct); + if (((long) length_limit) > maxlength) + length_limit = (unsigned int) maxlength; + + /* Choose processor routine to use. + * APP0/APP14 have special requirements. + */ + if (length_limit) { + processor = save_marker; + /* If saving APP0/APP14, save at least enough for our internal use. */ + if (marker_code == (int) M_APP0 && length_limit < APP0_DATA_LEN) + length_limit = APP0_DATA_LEN; + else if (marker_code == (int) M_APP14 && length_limit < APP14_DATA_LEN) + length_limit = APP14_DATA_LEN; + } else { + processor = skip_variable; + /* If discarding APP0/APP14, use our regular on-the-fly processor. */ + if (marker_code == (int) M_APP0 || marker_code == (int) M_APP14) + processor = get_interesting_appn; + } + + if (marker_code == (int) M_COM) { + marker->process_COM = processor; + marker->length_limit_COM = length_limit; + } else if (marker_code >= (int) M_APP0 && marker_code <= (int) M_APP15) { + marker->process_APPn[marker_code - (int) M_APP0] = processor; + marker->length_limit_APPn[marker_code - (int) M_APP0] = length_limit; + } else + ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, marker_code); +} + +#endif /* SAVE_MARKERS_SUPPORTED */ + + +/* + * Install a special processing method for COM or APPn markers. + */ + +GLOBAL(void) +jpeg_set_marker_processor (j_decompress_ptr cinfo, int marker_code, + jpeg_marker_parser_method routine) +{ + my_marker_ptr marker = (my_marker_ptr) cinfo->marker; + + if (marker_code == (int) M_COM) + marker->process_COM = routine; + else if (marker_code >= (int) M_APP0 && marker_code <= (int) M_APP15) + marker->process_APPn[marker_code - (int) M_APP0] = routine; + else + ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, marker_code); +} diff --git a/app/src/main/jni/libjpeg-turbo/jdmaster.c b/app/src/main/jni/libjpeg-turbo/jdmaster.c new file mode 100644 index 000000000..604e29161 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdmaster.c @@ -0,0 +1,737 @@ +/* + * jdmaster.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1997, Thomas G. Lane. + * Modified 2002-2009 by Guido Vollbeding. + * libjpeg-turbo Modifications: + * Copyright (C) 2009-2011, D. R. Commander. + * Copyright (C) 2013, Linaro Limited. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains master control logic for the JPEG decompressor. + * These routines are concerned with selecting the modules to be executed + * and with determining the number of passes and the work to be done in each + * pass. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jpegcomp.h" + + +/* Private state */ + +typedef struct { + struct jpeg_decomp_master pub; /* public fields */ + + int pass_number; /* # of passes completed */ + + boolean using_merged_upsample; /* TRUE if using merged upsample/cconvert */ + + /* Saved references to initialized quantizer modules, + * in case we need to switch modes. + */ + struct jpeg_color_quantizer * quantizer_1pass; + struct jpeg_color_quantizer * quantizer_2pass; +} my_decomp_master; + +typedef my_decomp_master * my_master_ptr; + + +/* + * Determine whether merged upsample/color conversion should be used. + * CRUCIAL: this must match the actual capabilities of jdmerge.c! + */ + +LOCAL(boolean) +use_merged_upsample (j_decompress_ptr cinfo) +{ +#ifdef UPSAMPLE_MERGING_SUPPORTED + /* Merging is the equivalent of plain box-filter upsampling */ + if (cinfo->do_fancy_upsampling || cinfo->CCIR601_sampling) + return FALSE; + /* jdmerge.c only supports YCC=>RGB and YCC=>RGB565 color conversion */ + if (cinfo->jpeg_color_space != JCS_YCbCr || cinfo->num_components != 3 || + (cinfo->out_color_space != JCS_RGB && + cinfo->out_color_space != JCS_RGB565 && + cinfo->out_color_space != JCS_EXT_RGB && + cinfo->out_color_space != JCS_EXT_RGBX && + cinfo->out_color_space != JCS_EXT_BGR && + cinfo->out_color_space != JCS_EXT_BGRX && + cinfo->out_color_space != JCS_EXT_XBGR && + cinfo->out_color_space != JCS_EXT_XRGB && + cinfo->out_color_space != JCS_EXT_RGBA && + cinfo->out_color_space != JCS_EXT_BGRA && + cinfo->out_color_space != JCS_EXT_ABGR && + cinfo->out_color_space != JCS_EXT_ARGB)) + return FALSE; + if ((cinfo->out_color_space == JCS_RGB565 && + cinfo->out_color_components != 3) || + (cinfo->out_color_space != JCS_RGB565 && + cinfo->out_color_components != rgb_pixelsize[cinfo->out_color_space])) + return FALSE; + /* and it only handles 2h1v or 2h2v sampling ratios */ + if (cinfo->comp_info[0].h_samp_factor != 2 || + cinfo->comp_info[1].h_samp_factor != 1 || + cinfo->comp_info[2].h_samp_factor != 1 || + cinfo->comp_info[0].v_samp_factor > 2 || + cinfo->comp_info[1].v_samp_factor != 1 || + cinfo->comp_info[2].v_samp_factor != 1) + return FALSE; + /* furthermore, it doesn't work if we've scaled the IDCTs differently */ + if (cinfo->comp_info[0]._DCT_scaled_size != cinfo->_min_DCT_scaled_size || + cinfo->comp_info[1]._DCT_scaled_size != cinfo->_min_DCT_scaled_size || + cinfo->comp_info[2]._DCT_scaled_size != cinfo->_min_DCT_scaled_size) + return FALSE; + /* ??? also need to test for upsample-time rescaling, when & if supported */ + return TRUE; /* by golly, it'll work... */ +#else + return FALSE; +#endif +} + + +/* + * Compute output image dimensions and related values. + * NOTE: this is exported for possible use by application. + * Hence it mustn't do anything that can't be done twice. + */ + +#if JPEG_LIB_VERSION >= 80 +GLOBAL(void) +#else +LOCAL(void) +#endif +jpeg_core_output_dimensions (j_decompress_ptr cinfo) +/* Do computations that are needed before master selection phase. + * This function is used for transcoding and full decompression. + */ +{ +#ifdef IDCT_SCALING_SUPPORTED + int ci; + jpeg_component_info *compptr; + + /* Compute actual output image dimensions and DCT scaling choices. */ + if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom) { + /* Provide 1/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 1; + cinfo->_min_DCT_v_scaled_size = 1; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 2) { + /* Provide 2/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 2L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 2L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 2; + cinfo->_min_DCT_v_scaled_size = 2; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 3) { + /* Provide 3/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 3L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 3L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 3; + cinfo->_min_DCT_v_scaled_size = 3; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 4) { + /* Provide 4/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 4L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 4L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 4; + cinfo->_min_DCT_v_scaled_size = 4; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 5) { + /* Provide 5/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 5L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 5L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 5; + cinfo->_min_DCT_v_scaled_size = 5; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 6) { + /* Provide 6/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 6L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 6L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 6; + cinfo->_min_DCT_v_scaled_size = 6; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 7) { + /* Provide 7/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 7L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 7L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 7; + cinfo->_min_DCT_v_scaled_size = 7; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 8) { + /* Provide 8/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 8L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 8L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 8; + cinfo->_min_DCT_v_scaled_size = 8; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 9) { + /* Provide 9/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 9L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 9L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 9; + cinfo->_min_DCT_v_scaled_size = 9; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 10) { + /* Provide 10/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 10L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 10L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 10; + cinfo->_min_DCT_v_scaled_size = 10; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 11) { + /* Provide 11/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 11L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 11L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 11; + cinfo->_min_DCT_v_scaled_size = 11; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 12) { + /* Provide 12/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 12L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 12L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 12; + cinfo->_min_DCT_v_scaled_size = 12; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 13) { + /* Provide 13/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 13L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 13L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 13; + cinfo->_min_DCT_v_scaled_size = 13; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 14) { + /* Provide 14/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 14L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 14L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 14; + cinfo->_min_DCT_v_scaled_size = 14; + } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 15) { + /* Provide 15/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 15L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 15L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 15; + cinfo->_min_DCT_v_scaled_size = 15; + } else { + /* Provide 16/block_size scaling */ + cinfo->output_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * 16L, (long) DCTSIZE); + cinfo->output_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * 16L, (long) DCTSIZE); + cinfo->_min_DCT_h_scaled_size = 16; + cinfo->_min_DCT_v_scaled_size = 16; + } + + /* Recompute dimensions of components */ + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + compptr->_DCT_h_scaled_size = cinfo->_min_DCT_h_scaled_size; + compptr->_DCT_v_scaled_size = cinfo->_min_DCT_v_scaled_size; + } + +#else /* !IDCT_SCALING_SUPPORTED */ + + /* Hardwire it to "no scaling" */ + cinfo->output_width = cinfo->image_width; + cinfo->output_height = cinfo->image_height; + /* jdinput.c has already initialized DCT_scaled_size, + * and has computed unscaled downsampled_width and downsampled_height. + */ + +#endif /* IDCT_SCALING_SUPPORTED */ +} + + +/* + * Compute output image dimensions and related values. + * NOTE: this is exported for possible use by application. + * Hence it mustn't do anything that can't be done twice. + * Also note that it may be called before the master module is initialized! + */ + +GLOBAL(void) +jpeg_calc_output_dimensions (j_decompress_ptr cinfo) +/* Do computations that are needed before master selection phase */ +{ +#ifdef IDCT_SCALING_SUPPORTED + int ci; + jpeg_component_info *compptr; +#endif + + /* Prevent application from calling me at wrong times */ + if (cinfo->global_state != DSTATE_READY) + ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state); + + /* Compute core output image dimensions and DCT scaling choices. */ + jpeg_core_output_dimensions(cinfo); + +#ifdef IDCT_SCALING_SUPPORTED + + /* In selecting the actual DCT scaling for each component, we try to + * scale up the chroma components via IDCT scaling rather than upsampling. + * This saves time if the upsampler gets to use 1:1 scaling. + * Note this code adapts subsampling ratios which are powers of 2. + */ + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + int ssize = cinfo->_min_DCT_scaled_size; + while (ssize < DCTSIZE && + ((cinfo->max_h_samp_factor * cinfo->_min_DCT_scaled_size) % + (compptr->h_samp_factor * ssize * 2) == 0) && + ((cinfo->max_v_samp_factor * cinfo->_min_DCT_scaled_size) % + (compptr->v_samp_factor * ssize * 2) == 0)) { + ssize = ssize * 2; + } +#if JPEG_LIB_VERSION >= 70 + compptr->DCT_h_scaled_size = compptr->DCT_v_scaled_size = ssize; +#else + compptr->DCT_scaled_size = ssize; +#endif + } + + /* Recompute downsampled dimensions of components; + * application needs to know these if using raw downsampled data. + */ + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + /* Size in samples, after IDCT scaling */ + compptr->downsampled_width = (JDIMENSION) + jdiv_round_up((long) cinfo->image_width * + (long) (compptr->h_samp_factor * compptr->_DCT_scaled_size), + (long) (cinfo->max_h_samp_factor * DCTSIZE)); + compptr->downsampled_height = (JDIMENSION) + jdiv_round_up((long) cinfo->image_height * + (long) (compptr->v_samp_factor * compptr->_DCT_scaled_size), + (long) (cinfo->max_v_samp_factor * DCTSIZE)); + } + +#else /* !IDCT_SCALING_SUPPORTED */ + + /* Hardwire it to "no scaling" */ + cinfo->output_width = cinfo->image_width; + cinfo->output_height = cinfo->image_height; + /* jdinput.c has already initialized DCT_scaled_size to DCTSIZE, + * and has computed unscaled downsampled_width and downsampled_height. + */ + +#endif /* IDCT_SCALING_SUPPORTED */ + + /* Report number of components in selected colorspace. */ + /* Probably this should be in the color conversion module... */ + switch (cinfo->out_color_space) { + case JCS_GRAYSCALE: + cinfo->out_color_components = 1; + break; + case JCS_RGB: + case JCS_EXT_RGB: + case JCS_EXT_RGBX: + case JCS_EXT_BGR: + case JCS_EXT_BGRX: + case JCS_EXT_XBGR: + case JCS_EXT_XRGB: + case JCS_EXT_RGBA: + case JCS_EXT_BGRA: + case JCS_EXT_ABGR: + case JCS_EXT_ARGB: + cinfo->out_color_components = rgb_pixelsize[cinfo->out_color_space]; + break; + case JCS_YCbCr: + case JCS_RGB565: + cinfo->out_color_components = 3; + break; + case JCS_CMYK: + case JCS_YCCK: + cinfo->out_color_components = 4; + break; + default: /* else must be same colorspace as in file */ + cinfo->out_color_components = cinfo->num_components; + break; + } + cinfo->output_components = (cinfo->quantize_colors ? 1 : + cinfo->out_color_components); + + /* See if upsampler will want to emit more than one row at a time */ + if (use_merged_upsample(cinfo)) + cinfo->rec_outbuf_height = cinfo->max_v_samp_factor; + else + cinfo->rec_outbuf_height = 1; +} + + +/* + * Several decompression processes need to range-limit values to the range + * 0..MAXJSAMPLE; the input value may fall somewhat outside this range + * due to noise introduced by quantization, roundoff error, etc. These + * processes are inner loops and need to be as fast as possible. On most + * machines, particularly CPUs with pipelines or instruction prefetch, + * a (subscript-check-less) C table lookup + * x = sample_range_limit[x]; + * is faster than explicit tests + * if (x < 0) x = 0; + * else if (x > MAXJSAMPLE) x = MAXJSAMPLE; + * These processes all use a common table prepared by the routine below. + * + * For most steps we can mathematically guarantee that the initial value + * of x is within MAXJSAMPLE+1 of the legal range, so a table running from + * -(MAXJSAMPLE+1) to 2*MAXJSAMPLE+1 is sufficient. But for the initial + * limiting step (just after the IDCT), a wildly out-of-range value is + * possible if the input data is corrupt. To avoid any chance of indexing + * off the end of memory and getting a bad-pointer trap, we perform the + * post-IDCT limiting thus: + * x = range_limit[x & MASK]; + * where MASK is 2 bits wider than legal sample data, ie 10 bits for 8-bit + * samples. Under normal circumstances this is more than enough range and + * a correct output will be generated; with bogus input data the mask will + * cause wraparound, and we will safely generate a bogus-but-in-range output. + * For the post-IDCT step, we want to convert the data from signed to unsigned + * representation by adding CENTERJSAMPLE at the same time that we limit it. + * So the post-IDCT limiting table ends up looking like this: + * CENTERJSAMPLE,CENTERJSAMPLE+1,...,MAXJSAMPLE, + * MAXJSAMPLE (repeat 2*(MAXJSAMPLE+1)-CENTERJSAMPLE times), + * 0 (repeat 2*(MAXJSAMPLE+1)-CENTERJSAMPLE times), + * 0,1,...,CENTERJSAMPLE-1 + * Negative inputs select values from the upper half of the table after + * masking. + * + * We can save some space by overlapping the start of the post-IDCT table + * with the simpler range limiting table. The post-IDCT table begins at + * sample_range_limit + CENTERJSAMPLE. + */ + +LOCAL(void) +prepare_range_limit_table (j_decompress_ptr cinfo) +/* Allocate and fill in the sample_range_limit table */ +{ + JSAMPLE * table; + int i; + + table = (JSAMPLE *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + (5 * (MAXJSAMPLE+1) + CENTERJSAMPLE) * sizeof(JSAMPLE)); + table += (MAXJSAMPLE+1); /* allow negative subscripts of simple table */ + cinfo->sample_range_limit = table; + /* First segment of "simple" table: limit[x] = 0 for x < 0 */ + MEMZERO(table - (MAXJSAMPLE+1), (MAXJSAMPLE+1) * sizeof(JSAMPLE)); + /* Main part of "simple" table: limit[x] = x */ + for (i = 0; i <= MAXJSAMPLE; i++) + table[i] = (JSAMPLE) i; + table += CENTERJSAMPLE; /* Point to where post-IDCT table starts */ + /* End of simple table, rest of first half of post-IDCT table */ + for (i = CENTERJSAMPLE; i < 2*(MAXJSAMPLE+1); i++) + table[i] = MAXJSAMPLE; + /* Second half of post-IDCT table */ + MEMZERO(table + (2 * (MAXJSAMPLE+1)), + (2 * (MAXJSAMPLE+1) - CENTERJSAMPLE) * sizeof(JSAMPLE)); + MEMCOPY(table + (4 * (MAXJSAMPLE+1) - CENTERJSAMPLE), + cinfo->sample_range_limit, CENTERJSAMPLE * sizeof(JSAMPLE)); +} + + +/* + * Master selection of decompression modules. + * This is done once at jpeg_start_decompress time. We determine + * which modules will be used and give them appropriate initialization calls. + * We also initialize the decompressor input side to begin consuming data. + * + * Since jpeg_read_header has finished, we know what is in the SOF + * and (first) SOS markers. We also have all the application parameter + * settings. + */ + +LOCAL(void) +master_selection (j_decompress_ptr cinfo) +{ + my_master_ptr master = (my_master_ptr) cinfo->master; + boolean use_c_buffer; + long samplesperrow; + JDIMENSION jd_samplesperrow; + + /* Initialize dimensions and other stuff */ + jpeg_calc_output_dimensions(cinfo); + prepare_range_limit_table(cinfo); + + /* Width of an output scanline must be representable as JDIMENSION. */ + samplesperrow = (long) cinfo->output_width * (long) cinfo->out_color_components; + jd_samplesperrow = (JDIMENSION) samplesperrow; + if ((long) jd_samplesperrow != samplesperrow) + ERREXIT(cinfo, JERR_WIDTH_OVERFLOW); + + /* Initialize my private state */ + master->pass_number = 0; + master->using_merged_upsample = use_merged_upsample(cinfo); + + /* Color quantizer selection */ + master->quantizer_1pass = NULL; + master->quantizer_2pass = NULL; + /* No mode changes if not using buffered-image mode. */ + if (! cinfo->quantize_colors || ! cinfo->buffered_image) { + cinfo->enable_1pass_quant = FALSE; + cinfo->enable_external_quant = FALSE; + cinfo->enable_2pass_quant = FALSE; + } + if (cinfo->quantize_colors) { + if (cinfo->raw_data_out) + ERREXIT(cinfo, JERR_NOTIMPL); + /* 2-pass quantizer only works in 3-component color space. */ + if (cinfo->out_color_components != 3) { + cinfo->enable_1pass_quant = TRUE; + cinfo->enable_external_quant = FALSE; + cinfo->enable_2pass_quant = FALSE; + cinfo->colormap = NULL; + } else if (cinfo->colormap != NULL) { + cinfo->enable_external_quant = TRUE; + } else if (cinfo->two_pass_quantize) { + cinfo->enable_2pass_quant = TRUE; + } else { + cinfo->enable_1pass_quant = TRUE; + } + + if (cinfo->enable_1pass_quant) { +#ifdef QUANT_1PASS_SUPPORTED + jinit_1pass_quantizer(cinfo); + master->quantizer_1pass = cinfo->cquantize; +#else + ERREXIT(cinfo, JERR_NOT_COMPILED); +#endif + } + + /* We use the 2-pass code to map to external colormaps. */ + if (cinfo->enable_2pass_quant || cinfo->enable_external_quant) { +#ifdef QUANT_2PASS_SUPPORTED + jinit_2pass_quantizer(cinfo); + master->quantizer_2pass = cinfo->cquantize; +#else + ERREXIT(cinfo, JERR_NOT_COMPILED); +#endif + } + /* If both quantizers are initialized, the 2-pass one is left active; + * this is necessary for starting with quantization to an external map. + */ + } + + /* Post-processing: in particular, color conversion first */ + if (! cinfo->raw_data_out) { + if (master->using_merged_upsample) { +#ifdef UPSAMPLE_MERGING_SUPPORTED + jinit_merged_upsampler(cinfo); /* does color conversion too */ +#else + ERREXIT(cinfo, JERR_NOT_COMPILED); +#endif + } else { + jinit_color_deconverter(cinfo); + jinit_upsampler(cinfo); + } + jinit_d_post_controller(cinfo, cinfo->enable_2pass_quant); + } + /* Inverse DCT */ + jinit_inverse_dct(cinfo); + /* Entropy decoding: either Huffman or arithmetic coding. */ + if (cinfo->arith_code) { +#ifdef D_ARITH_CODING_SUPPORTED + jinit_arith_decoder(cinfo); +#else + ERREXIT(cinfo, JERR_ARITH_NOTIMPL); +#endif + } else { + if (cinfo->progressive_mode) { +#ifdef D_PROGRESSIVE_SUPPORTED + jinit_phuff_decoder(cinfo); +#else + ERREXIT(cinfo, JERR_NOT_COMPILED); +#endif + } else + jinit_huff_decoder(cinfo); + } + + /* Initialize principal buffer controllers. */ + use_c_buffer = cinfo->inputctl->has_multiple_scans || cinfo->buffered_image; + jinit_d_coef_controller(cinfo, use_c_buffer); + + if (! cinfo->raw_data_out) + jinit_d_main_controller(cinfo, FALSE /* never need full buffer here */); + + /* We can now tell the memory manager to allocate virtual arrays. */ + (*cinfo->mem->realize_virt_arrays) ((j_common_ptr) cinfo); + + /* Initialize input side of decompressor to consume first scan. */ + (*cinfo->inputctl->start_input_pass) (cinfo); + +#ifdef D_MULTISCAN_FILES_SUPPORTED + /* If jpeg_start_decompress will read the whole file, initialize + * progress monitoring appropriately. The input step is counted + * as one pass. + */ + if (cinfo->progress != NULL && ! cinfo->buffered_image && + cinfo->inputctl->has_multiple_scans) { + int nscans; + /* Estimate number of scans to set pass_limit. */ + if (cinfo->progressive_mode) { + /* Arbitrarily estimate 2 interleaved DC scans + 3 AC scans/component. */ + nscans = 2 + 3 * cinfo->num_components; + } else { + /* For a nonprogressive multiscan file, estimate 1 scan per component. */ + nscans = cinfo->num_components; + } + cinfo->progress->pass_counter = 0L; + cinfo->progress->pass_limit = (long) cinfo->total_iMCU_rows * nscans; + cinfo->progress->completed_passes = 0; + cinfo->progress->total_passes = (cinfo->enable_2pass_quant ? 3 : 2); + /* Count the input pass as done */ + master->pass_number++; + } +#endif /* D_MULTISCAN_FILES_SUPPORTED */ +} + + +/* + * Per-pass setup. + * This is called at the beginning of each output pass. We determine which + * modules will be active during this pass and give them appropriate + * start_pass calls. We also set is_dummy_pass to indicate whether this + * is a "real" output pass or a dummy pass for color quantization. + * (In the latter case, jdapistd.c will crank the pass to completion.) + */ + +METHODDEF(void) +prepare_for_output_pass (j_decompress_ptr cinfo) +{ + my_master_ptr master = (my_master_ptr) cinfo->master; + + if (master->pub.is_dummy_pass) { +#ifdef QUANT_2PASS_SUPPORTED + /* Final pass of 2-pass quantization */ + master->pub.is_dummy_pass = FALSE; + (*cinfo->cquantize->start_pass) (cinfo, FALSE); + (*cinfo->post->start_pass) (cinfo, JBUF_CRANK_DEST); + (*cinfo->main->start_pass) (cinfo, JBUF_CRANK_DEST); +#else + ERREXIT(cinfo, JERR_NOT_COMPILED); +#endif /* QUANT_2PASS_SUPPORTED */ + } else { + if (cinfo->quantize_colors && cinfo->colormap == NULL) { + /* Select new quantization method */ + if (cinfo->two_pass_quantize && cinfo->enable_2pass_quant) { + cinfo->cquantize = master->quantizer_2pass; + master->pub.is_dummy_pass = TRUE; + } else if (cinfo->enable_1pass_quant) { + cinfo->cquantize = master->quantizer_1pass; + } else { + ERREXIT(cinfo, JERR_MODE_CHANGE); + } + } + (*cinfo->idct->start_pass) (cinfo); + (*cinfo->coef->start_output_pass) (cinfo); + if (! cinfo->raw_data_out) { + if (! master->using_merged_upsample) + (*cinfo->cconvert->start_pass) (cinfo); + (*cinfo->upsample->start_pass) (cinfo); + if (cinfo->quantize_colors) + (*cinfo->cquantize->start_pass) (cinfo, master->pub.is_dummy_pass); + (*cinfo->post->start_pass) (cinfo, + (master->pub.is_dummy_pass ? JBUF_SAVE_AND_PASS : JBUF_PASS_THRU)); + (*cinfo->main->start_pass) (cinfo, JBUF_PASS_THRU); + } + } + + /* Set up progress monitor's pass info if present */ + if (cinfo->progress != NULL) { + cinfo->progress->completed_passes = master->pass_number; + cinfo->progress->total_passes = master->pass_number + + (master->pub.is_dummy_pass ? 2 : 1); + /* In buffered-image mode, we assume one more output pass if EOI not + * yet reached, but no more passes if EOI has been reached. + */ + if (cinfo->buffered_image && ! cinfo->inputctl->eoi_reached) { + cinfo->progress->total_passes += (cinfo->enable_2pass_quant ? 2 : 1); + } + } +} + + +/* + * Finish up at end of an output pass. + */ + +METHODDEF(void) +finish_output_pass (j_decompress_ptr cinfo) +{ + my_master_ptr master = (my_master_ptr) cinfo->master; + + if (cinfo->quantize_colors) + (*cinfo->cquantize->finish_pass) (cinfo); + master->pass_number++; +} + + +#ifdef D_MULTISCAN_FILES_SUPPORTED + +/* + * Switch to a new external colormap between output passes. + */ + +GLOBAL(void) +jpeg_new_colormap (j_decompress_ptr cinfo) +{ + my_master_ptr master = (my_master_ptr) cinfo->master; + + /* Prevent application from calling me at wrong times */ + if (cinfo->global_state != DSTATE_BUFIMAGE) + ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state); + + if (cinfo->quantize_colors && cinfo->enable_external_quant && + cinfo->colormap != NULL) { + /* Select 2-pass quantizer for external colormap use */ + cinfo->cquantize = master->quantizer_2pass; + /* Notify quantizer of colormap change */ + (*cinfo->cquantize->new_color_map) (cinfo); + master->pub.is_dummy_pass = FALSE; /* just in case */ + } else + ERREXIT(cinfo, JERR_MODE_CHANGE); +} + +#endif /* D_MULTISCAN_FILES_SUPPORTED */ + + +/* + * Initialize master decompression control and select active modules. + * This is performed at the start of jpeg_start_decompress. + */ + +GLOBAL(void) +jinit_master_decompress (j_decompress_ptr cinfo) +{ + my_master_ptr master; + + master = (my_master_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(my_decomp_master)); + cinfo->master = (struct jpeg_decomp_master *) master; + master->pub.prepare_for_output_pass = prepare_for_output_pass; + master->pub.finish_output_pass = finish_output_pass; + + master->pub.is_dummy_pass = FALSE; + + master_selection(cinfo); +} diff --git a/app/src/main/jni/libjpeg-turbo/jdmerge.c b/app/src/main/jni/libjpeg-turbo/jdmerge.c new file mode 100644 index 000000000..e13adb9f2 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdmerge.c @@ -0,0 +1,626 @@ +/* + * jdmerge.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1996, Thomas G. Lane. + * Copyright 2009 Pierre Ossman for Cendio AB + * libjpeg-turbo Modifications: + * Copyright (C) 2009, 2011, 2014 D. R. Commander. + * Copyright (C) 2013, Linaro Limited. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains code for merged upsampling/color conversion. + * + * This file combines functions from jdsample.c and jdcolor.c; + * read those files first to understand what's going on. + * + * When the chroma components are to be upsampled by simple replication + * (ie, box filtering), we can save some work in color conversion by + * calculating all the output pixels corresponding to a pair of chroma + * samples at one time. In the conversion equations + * R = Y + K1 * Cr + * G = Y + K2 * Cb + K3 * Cr + * B = Y + K4 * Cb + * only the Y term varies among the group of pixels corresponding to a pair + * of chroma samples, so the rest of the terms can be calculated just once. + * At typical sampling ratios, this eliminates half or three-quarters of the + * multiplications needed for color conversion. + * + * This file currently provides implementations for the following cases: + * YCbCr => RGB color conversion only. + * Sampling ratios of 2h1v or 2h2v. + * No scaling needed at upsample time. + * Corner-aligned (non-CCIR601) sampling alignment. + * Other special cases could be added, but in most applications these are + * the only common cases. (For uncommon cases we fall back on the more + * general code in jdsample.c and jdcolor.c.) + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jsimd.h" +#include "jconfigint.h" + +#ifdef UPSAMPLE_MERGING_SUPPORTED + + +/* Private subobject */ + +typedef struct { + struct jpeg_upsampler pub; /* public fields */ + + /* Pointer to routine to do actual upsampling/conversion of one row group */ + void (*upmethod) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); + + /* Private state for YCC->RGB conversion */ + int * Cr_r_tab; /* => table for Cr to R conversion */ + int * Cb_b_tab; /* => table for Cb to B conversion */ + INT32 * Cr_g_tab; /* => table for Cr to G conversion */ + INT32 * Cb_g_tab; /* => table for Cb to G conversion */ + + /* For 2:1 vertical sampling, we produce two output rows at a time. + * We need a "spare" row buffer to hold the second output row if the + * application provides just a one-row buffer; we also use the spare + * to discard the dummy last row if the image height is odd. + */ + JSAMPROW spare_row; + boolean spare_full; /* T if spare buffer is occupied */ + + JDIMENSION out_row_width; /* samples per output row */ + JDIMENSION rows_to_go; /* counts rows remaining in image */ +} my_upsampler; + +typedef my_upsampler * my_upsample_ptr; + +#define SCALEBITS 16 /* speediest right-shift on some machines */ +#define ONE_HALF ((INT32) 1 << (SCALEBITS-1)) +#define FIX(x) ((INT32) ((x) * (1L<RGB colorspace conversion. + * This is taken directly from jdcolor.c; see that file for more info. + */ + +LOCAL(void) +build_ycc_rgb_table (j_decompress_ptr cinfo) +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + int i; + INT32 x; + SHIFT_TEMPS + + upsample->Cr_r_tab = (int *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + (MAXJSAMPLE+1) * sizeof(int)); + upsample->Cb_b_tab = (int *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + (MAXJSAMPLE+1) * sizeof(int)); + upsample->Cr_g_tab = (INT32 *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + (MAXJSAMPLE+1) * sizeof(INT32)); + upsample->Cb_g_tab = (INT32 *) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + (MAXJSAMPLE+1) * sizeof(INT32)); + + for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) { + /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */ + /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */ + /* Cr=>R value is nearest int to 1.40200 * x */ + upsample->Cr_r_tab[i] = (int) + RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS); + /* Cb=>B value is nearest int to 1.77200 * x */ + upsample->Cb_b_tab[i] = (int) + RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS); + /* Cr=>G value is scaled-up -0.71414 * x */ + upsample->Cr_g_tab[i] = (- FIX(0.71414)) * x; + /* Cb=>G value is scaled-up -0.34414 * x */ + /* We also add in ONE_HALF so that need not do it in inner loop */ + upsample->Cb_g_tab[i] = (- FIX(0.34414)) * x + ONE_HALF; + } +} + + +/* + * Initialize for an upsampling pass. + */ + +METHODDEF(void) +start_pass_merged_upsample (j_decompress_ptr cinfo) +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + + /* Mark the spare buffer empty */ + upsample->spare_full = FALSE; + /* Initialize total-height counter for detecting bottom of image */ + upsample->rows_to_go = cinfo->output_height; +} + + +/* + * Control routine to do upsampling (and color conversion). + * + * The control routine just handles the row buffering considerations. + */ + +METHODDEF(void) +merged_2v_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr, + JDIMENSION in_row_groups_avail, + JSAMPARRAY output_buf, JDIMENSION *out_row_ctr, + JDIMENSION out_rows_avail) +/* 2:1 vertical sampling case: may need a spare row. */ +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + JSAMPROW work_ptrs[2]; + JDIMENSION num_rows; /* number of rows returned to caller */ + + if (upsample->spare_full) { + /* If we have a spare row saved from a previous cycle, just return it. */ + JDIMENSION size = upsample->out_row_width; + if (cinfo->out_color_space == JCS_RGB565) + size = cinfo->output_width * 2; + jcopy_sample_rows(& upsample->spare_row, 0, output_buf + *out_row_ctr, 0, + 1, size); + num_rows = 1; + upsample->spare_full = FALSE; + } else { + /* Figure number of rows to return to caller. */ + num_rows = 2; + /* Not more than the distance to the end of the image. */ + if (num_rows > upsample->rows_to_go) + num_rows = upsample->rows_to_go; + /* And not more than what the client can accept: */ + out_rows_avail -= *out_row_ctr; + if (num_rows > out_rows_avail) + num_rows = out_rows_avail; + /* Create output pointer array for upsampler. */ + work_ptrs[0] = output_buf[*out_row_ctr]; + if (num_rows > 1) { + work_ptrs[1] = output_buf[*out_row_ctr + 1]; + } else { + work_ptrs[1] = upsample->spare_row; + upsample->spare_full = TRUE; + } + /* Now do the upsampling. */ + (*upsample->upmethod) (cinfo, input_buf, *in_row_group_ctr, work_ptrs); + } + + /* Adjust counts */ + *out_row_ctr += num_rows; + upsample->rows_to_go -= num_rows; + /* When the buffer is emptied, declare this input row group consumed */ + if (! upsample->spare_full) + (*in_row_group_ctr)++; +} + + +METHODDEF(void) +merged_1v_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr, + JDIMENSION in_row_groups_avail, + JSAMPARRAY output_buf, JDIMENSION *out_row_ctr, + JDIMENSION out_rows_avail) +/* 1:1 vertical sampling case: much easier, never need a spare row. */ +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + + /* Just do the upsampling. */ + (*upsample->upmethod) (cinfo, input_buf, *in_row_group_ctr, + output_buf + *out_row_ctr); + /* Adjust counts */ + (*out_row_ctr)++; + (*in_row_group_ctr)++; +} + + +/* + * These are the routines invoked by the control routines to do + * the actual upsampling/conversion. One row group is processed per call. + * + * Note: since we may be writing directly into application-supplied buffers, + * we have to be honest about the output width; we can't assume the buffer + * has been rounded up to an even width. + */ + + +/* + * Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical. + */ + +METHODDEF(void) +h2v1_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + switch (cinfo->out_color_space) { + case JCS_EXT_RGB: + extrgb_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + extrgbx_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + case JCS_EXT_BGR: + extbgr_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + extbgrx_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + extxbgr_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + extxrgb_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + default: + h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + } +} + + +/* + * Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical. + */ + +METHODDEF(void) +h2v2_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + switch (cinfo->out_color_space) { + case JCS_EXT_RGB: + extrgb_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + extrgbx_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + case JCS_EXT_BGR: + extbgr_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + extbgrx_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + extxbgr_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + extxrgb_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + default: + h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr, + output_buf); + break; + } +} + + +/* + * RGB565 conversion + */ + +#define PACK_SHORT_565_LE(r, g, b) ((((r) << 8) & 0xF800) | \ + (((g) << 3) & 0x7E0) | ((b) >> 3)) +#define PACK_SHORT_565_BE(r, g, b) (((r) & 0xF8) | ((g) >> 5) | \ + (((g) << 11) & 0xE000) | \ + (((b) << 5) & 0x1F00)) + +#define PACK_TWO_PIXELS_LE(l, r) ((r << 16) | l) +#define PACK_TWO_PIXELS_BE(l, r) ((l << 16) | r) + +#define PACK_NEED_ALIGNMENT(ptr) (((size_t)(ptr)) & 3) + +#define WRITE_TWO_PIXELS_LE(addr, pixels) { \ + ((INT16*)(addr))[0] = (pixels); \ + ((INT16*)(addr))[1] = (pixels) >> 16; \ +} +#define WRITE_TWO_PIXELS_BE(addr, pixels) { \ + ((INT16*)(addr))[1] = (pixels); \ + ((INT16*)(addr))[0] = (pixels) >> 16; \ +} + +#define DITHER_565_R(r, dither) ((r) + ((dither) & 0xFF)) +#define DITHER_565_G(g, dither) ((g) + (((dither) & 0xFF) >> 1)) +#define DITHER_565_B(b, dither) ((b) + ((dither) & 0xFF)) + + +/* Declarations for ordered dithering + * + * We use a 4x4 ordered dither array packed into 32 bits. This array is + * sufficent for dithering RGB888 to RGB565. + */ + +#define DITHER_MASK 0x3 +#define DITHER_ROTATE(x) (((x) << 24) | (((x) >> 8) & 0x00FFFFFF)) +static const INT32 dither_matrix[4] = { + 0x0008020A, + 0x0C040E06, + 0x030B0109, + 0x0F070D05 +}; + + +/* Include inline routines for RGB565 conversion */ + +#define PACK_SHORT_565 PACK_SHORT_565_LE +#define PACK_TWO_PIXELS PACK_TWO_PIXELS_LE +#define WRITE_TWO_PIXELS WRITE_TWO_PIXELS_LE +#define h2v1_merged_upsample_565_internal h2v1_merged_upsample_565_le +#define h2v1_merged_upsample_565D_internal h2v1_merged_upsample_565D_le +#define h2v2_merged_upsample_565_internal h2v2_merged_upsample_565_le +#define h2v2_merged_upsample_565D_internal h2v2_merged_upsample_565D_le +#include "jdmrg565.c" +#undef PACK_SHORT_565 +#undef PACK_TWO_PIXELS +#undef WRITE_TWO_PIXELS +#undef h2v1_merged_upsample_565_internal +#undef h2v1_merged_upsample_565D_internal +#undef h2v2_merged_upsample_565_internal +#undef h2v2_merged_upsample_565D_internal + +#define PACK_SHORT_565 PACK_SHORT_565_BE +#define PACK_TWO_PIXELS PACK_TWO_PIXELS_BE +#define WRITE_TWO_PIXELS WRITE_TWO_PIXELS_BE +#define h2v1_merged_upsample_565_internal h2v1_merged_upsample_565_be +#define h2v1_merged_upsample_565D_internal h2v1_merged_upsample_565D_be +#define h2v2_merged_upsample_565_internal h2v2_merged_upsample_565_be +#define h2v2_merged_upsample_565D_internal h2v2_merged_upsample_565D_be +#include "jdmrg565.c" +#undef PACK_SHORT_565 +#undef PACK_TWO_PIXELS +#undef WRITE_TWO_PIXELS +#undef h2v1_merged_upsample_565_internal +#undef h2v1_merged_upsample_565D_internal +#undef h2v2_merged_upsample_565_internal +#undef h2v2_merged_upsample_565D_internal + + +static INLINE boolean is_big_endian(void) +{ + int test_value = 1; + if(*(char *)&test_value != 1) + return TRUE; + return FALSE; +} + + +METHODDEF(void) +h2v1_merged_upsample_565 (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + if (is_big_endian()) + h2v1_merged_upsample_565_be(cinfo, input_buf, in_row_group_ctr, + output_buf); + else + h2v1_merged_upsample_565_le(cinfo, input_buf, in_row_group_ctr, + output_buf); + } + + +METHODDEF(void) +h2v1_merged_upsample_565D (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + if (is_big_endian()) + h2v1_merged_upsample_565D_be(cinfo, input_buf, in_row_group_ctr, + output_buf); + else + h2v1_merged_upsample_565D_le(cinfo, input_buf, in_row_group_ctr, + output_buf); +} + + +METHODDEF(void) +h2v2_merged_upsample_565 (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + if (is_big_endian()) + h2v2_merged_upsample_565_be(cinfo, input_buf, in_row_group_ctr, + output_buf); + else + h2v2_merged_upsample_565_le(cinfo, input_buf, in_row_group_ctr, + output_buf); +} + + +METHODDEF(void) +h2v2_merged_upsample_565D (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + if (is_big_endian()) + h2v2_merged_upsample_565D_be(cinfo, input_buf, in_row_group_ctr, + output_buf); + else + h2v2_merged_upsample_565D_le(cinfo, input_buf, in_row_group_ctr, + output_buf); +} + + +/* + * Module initialization routine for merged upsampling/color conversion. + * + * NB: this is called under the conditions determined by use_merged_upsample() + * in jdmaster.c. That routine MUST correspond to the actual capabilities + * of this module; no safety checks are made here. + */ + +GLOBAL(void) +jinit_merged_upsampler (j_decompress_ptr cinfo) +{ + my_upsample_ptr upsample; + + upsample = (my_upsample_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(my_upsampler)); + cinfo->upsample = (struct jpeg_upsampler *) upsample; + upsample->pub.start_pass = start_pass_merged_upsample; + upsample->pub.need_context_rows = FALSE; + + upsample->out_row_width = cinfo->output_width * cinfo->out_color_components; + + if (cinfo->max_v_samp_factor == 2) { + upsample->pub.upsample = merged_2v_upsample; + if (jsimd_can_h2v2_merged_upsample()) + upsample->upmethod = jsimd_h2v2_merged_upsample; + else + upsample->upmethod = h2v2_merged_upsample; + if (cinfo->out_color_space == JCS_RGB565) { + if (cinfo->dither_mode != JDITHER_NONE) { + upsample->upmethod = h2v2_merged_upsample_565D; + } else { + upsample->upmethod = h2v2_merged_upsample_565; + } + } + /* Allocate a spare row buffer */ + upsample->spare_row = (JSAMPROW) + (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE, + (size_t) (upsample->out_row_width * sizeof(JSAMPLE))); + } else { + upsample->pub.upsample = merged_1v_upsample; + if (jsimd_can_h2v1_merged_upsample()) + upsample->upmethod = jsimd_h2v1_merged_upsample; + else + upsample->upmethod = h2v1_merged_upsample; + if (cinfo->out_color_space == JCS_RGB565) { + if (cinfo->dither_mode != JDITHER_NONE) { + upsample->upmethod = h2v1_merged_upsample_565D; + } else { + upsample->upmethod = h2v1_merged_upsample_565; + } + } + /* No spare row needed */ + upsample->spare_row = NULL; + } + + build_ycc_rgb_table(cinfo); +} + +#endif /* UPSAMPLE_MERGING_SUPPORTED */ diff --git a/app/src/main/jni/libjpeg-turbo/jdmrg565.c b/app/src/main/jni/libjpeg-turbo/jdmrg565.c new file mode 100644 index 000000000..0a10bccb4 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdmrg565.c @@ -0,0 +1,355 @@ +/* + * jdmrg565.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1996, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright (C) 2013, Linaro Limited. + * Copyright (C) 2014, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains code for merged upsampling/color conversion. + */ + + +INLINE +LOCAL(void) +h2v1_merged_upsample_565_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + register int y, cred, cgreen, cblue; + int cb, cr; + register JSAMPROW outptr; + JSAMPROW inptr0, inptr1, inptr2; + JDIMENSION col; + /* copy these pointers into registers if possible */ + register JSAMPLE * range_limit = cinfo->sample_range_limit; + int * Crrtab = upsample->Cr_r_tab; + int * Cbbtab = upsample->Cb_b_tab; + INT32 * Crgtab = upsample->Cr_g_tab; + INT32 * Cbgtab = upsample->Cb_g_tab; + unsigned int r, g, b; + INT32 rgb; + SHIFT_TEMPS + + inptr0 = input_buf[0][in_row_group_ctr]; + inptr1 = input_buf[1][in_row_group_ctr]; + inptr2 = input_buf[2][in_row_group_ctr]; + outptr = output_buf[0]; + + /* Loop for each pair of output pixels */ + for (col = cinfo->output_width >> 1; col > 0; col--) { + /* Do the chroma part of the calculation */ + cb = GETJSAMPLE(*inptr1++); + cr = GETJSAMPLE(*inptr2++); + cred = Crrtab[cr]; + cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS); + cblue = Cbbtab[cb]; + + /* Fetch 2 Y values and emit 2 pixels */ + y = GETJSAMPLE(*inptr0++); + r = range_limit[y + cred]; + g = range_limit[y + cgreen]; + b = range_limit[y + cblue]; + rgb = PACK_SHORT_565(r, g, b); + + y = GETJSAMPLE(*inptr0++); + r = range_limit[y + cred]; + g = range_limit[y + cgreen]; + b = range_limit[y + cblue]; + rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b)); + + WRITE_TWO_PIXELS(outptr, rgb); + outptr += 4; + } + + /* If image width is odd, do the last output column separately */ + if (cinfo->output_width & 1) { + cb = GETJSAMPLE(*inptr1); + cr = GETJSAMPLE(*inptr2); + cred = Crrtab[cr]; + cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS); + cblue = Cbbtab[cb]; + y = GETJSAMPLE(*inptr0); + r = range_limit[y + cred]; + g = range_limit[y + cgreen]; + b = range_limit[y + cblue]; + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr = rgb; + } + } + + +INLINE +LOCAL(void) +h2v1_merged_upsample_565D_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + register int y, cred, cgreen, cblue; + int cb, cr; + register JSAMPROW outptr; + JSAMPROW inptr0, inptr1, inptr2; + JDIMENSION col; + /* copy these pointers into registers if possible */ + register JSAMPLE * range_limit = cinfo->sample_range_limit; + int * Crrtab = upsample->Cr_r_tab; + int * Cbbtab = upsample->Cb_b_tab; + INT32 * Crgtab = upsample->Cr_g_tab; + INT32 * Cbgtab = upsample->Cb_g_tab; + INT32 d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK]; + unsigned int r, g, b; + INT32 rgb; + SHIFT_TEMPS + + inptr0 = input_buf[0][in_row_group_ctr]; + inptr1 = input_buf[1][in_row_group_ctr]; + inptr2 = input_buf[2][in_row_group_ctr]; + outptr = output_buf[0]; + + /* Loop for each pair of output pixels */ + for (col = cinfo->output_width >> 1; col > 0; col--) { + /* Do the chroma part of the calculation */ + cb = GETJSAMPLE(*inptr1++); + cr = GETJSAMPLE(*inptr2++); + cred = Crrtab[cr]; + cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS); + cblue = Cbbtab[cb]; + + /* Fetch 2 Y values and emit 2 pixels */ + y = GETJSAMPLE(*inptr0++); + r = range_limit[DITHER_565_R(y + cred, d0)]; + g = range_limit[DITHER_565_G(y + cgreen, d0)]; + b = range_limit[DITHER_565_B(y + cblue, d0)]; + d0 = DITHER_ROTATE(d0); + rgb = PACK_SHORT_565(r, g, b); + + y = GETJSAMPLE(*inptr0++); + r = range_limit[DITHER_565_R(y + cred, d0)]; + g = range_limit[DITHER_565_G(y + cgreen, d0)]; + b = range_limit[DITHER_565_B(y + cblue, d0)]; + d0 = DITHER_ROTATE(d0); + rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b)); + + WRITE_TWO_PIXELS(outptr, rgb); + outptr += 4; + } + + /* If image width is odd, do the last output column separately */ + if (cinfo->output_width & 1) { + cb = GETJSAMPLE(*inptr1); + cr = GETJSAMPLE(*inptr2); + cred = Crrtab[cr]; + cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS); + cblue = Cbbtab[cb]; + y = GETJSAMPLE(*inptr0); + r = range_limit[DITHER_565_R(y + cred, d0)]; + g = range_limit[DITHER_565_G(y + cgreen, d0)]; + b = range_limit[DITHER_565_B(y + cblue, d0)]; + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr = rgb; + } +} + + +INLINE +LOCAL(void) +h2v2_merged_upsample_565_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + register int y, cred, cgreen, cblue; + int cb, cr; + register JSAMPROW outptr0, outptr1; + JSAMPROW inptr00, inptr01, inptr1, inptr2; + JDIMENSION col; + /* copy these pointers into registers if possible */ + register JSAMPLE * range_limit = cinfo->sample_range_limit; + int * Crrtab = upsample->Cr_r_tab; + int * Cbbtab = upsample->Cb_b_tab; + INT32 * Crgtab = upsample->Cr_g_tab; + INT32 * Cbgtab = upsample->Cb_g_tab; + unsigned int r, g, b; + INT32 rgb; + SHIFT_TEMPS + + inptr00 = input_buf[0][in_row_group_ctr * 2]; + inptr01 = input_buf[0][in_row_group_ctr * 2 + 1]; + inptr1 = input_buf[1][in_row_group_ctr]; + inptr2 = input_buf[2][in_row_group_ctr]; + outptr0 = output_buf[0]; + outptr1 = output_buf[1]; + + /* Loop for each group of output pixels */ + for (col = cinfo->output_width >> 1; col > 0; col--) { + /* Do the chroma part of the calculation */ + cb = GETJSAMPLE(*inptr1++); + cr = GETJSAMPLE(*inptr2++); + cred = Crrtab[cr]; + cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS); + cblue = Cbbtab[cb]; + + /* Fetch 4 Y values and emit 4 pixels */ + y = GETJSAMPLE(*inptr00++); + r = range_limit[y + cred]; + g = range_limit[y + cgreen]; + b = range_limit[y + cblue]; + rgb = PACK_SHORT_565(r, g, b); + + y = GETJSAMPLE(*inptr00++); + r = range_limit[y + cred]; + g = range_limit[y + cgreen]; + b = range_limit[y + cblue]; + rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b)); + + WRITE_TWO_PIXELS(outptr0, rgb); + outptr0 += 4; + + y = GETJSAMPLE(*inptr01++); + r = range_limit[y + cred]; + g = range_limit[y + cgreen]; + b = range_limit[y + cblue]; + rgb = PACK_SHORT_565(r, g, b); + + y = GETJSAMPLE(*inptr01++); + r = range_limit[y + cred]; + g = range_limit[y + cgreen]; + b = range_limit[y + cblue]; + rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b)); + + WRITE_TWO_PIXELS(outptr1, rgb); + outptr1 += 4; + } + + /* If image width is odd, do the last output column separately */ + if (cinfo->output_width & 1) { + cb = GETJSAMPLE(*inptr1); + cr = GETJSAMPLE(*inptr2); + cred = Crrtab[cr]; + cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS); + cblue = Cbbtab[cb]; + + y = GETJSAMPLE(*inptr00); + r = range_limit[y + cred]; + g = range_limit[y + cgreen]; + b = range_limit[y + cblue]; + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr0 = rgb; + + y = GETJSAMPLE(*inptr01); + r = range_limit[y + cred]; + g = range_limit[y + cgreen]; + b = range_limit[y + cblue]; + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr1 = rgb; + } +} + + +INLINE +LOCAL(void) +h2v2_merged_upsample_565D_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + register int y, cred, cgreen, cblue; + int cb, cr; + register JSAMPROW outptr0, outptr1; + JSAMPROW inptr00, inptr01, inptr1, inptr2; + JDIMENSION col; + /* copy these pointers into registers if possible */ + register JSAMPLE * range_limit = cinfo->sample_range_limit; + int * Crrtab = upsample->Cr_r_tab; + int * Cbbtab = upsample->Cb_b_tab; + INT32 * Crgtab = upsample->Cr_g_tab; + INT32 * Cbgtab = upsample->Cb_g_tab; + INT32 d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK]; + INT32 d1 = dither_matrix[(cinfo->output_scanline+1) & DITHER_MASK]; + unsigned int r, g, b; + INT32 rgb; + SHIFT_TEMPS + + inptr00 = input_buf[0][in_row_group_ctr*2]; + inptr01 = input_buf[0][in_row_group_ctr*2 + 1]; + inptr1 = input_buf[1][in_row_group_ctr]; + inptr2 = input_buf[2][in_row_group_ctr]; + outptr0 = output_buf[0]; + outptr1 = output_buf[1]; + + /* Loop for each group of output pixels */ + for (col = cinfo->output_width >> 1; col > 0; col--) { + /* Do the chroma part of the calculation */ + cb = GETJSAMPLE(*inptr1++); + cr = GETJSAMPLE(*inptr2++); + cred = Crrtab[cr]; + cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS); + cblue = Cbbtab[cb]; + + /* Fetch 4 Y values and emit 4 pixels */ + y = GETJSAMPLE(*inptr00++); + r = range_limit[DITHER_565_R(y + cred, d0)]; + g = range_limit[DITHER_565_G(y + cgreen, d0)]; + b = range_limit[DITHER_565_B(y + cblue, d0)]; + d0 = DITHER_ROTATE(d0); + rgb = PACK_SHORT_565(r, g, b); + + y = GETJSAMPLE(*inptr00++); + r = range_limit[DITHER_565_R(y + cred, d1)]; + g = range_limit[DITHER_565_G(y + cgreen, d1)]; + b = range_limit[DITHER_565_B(y + cblue, d1)]; + d1 = DITHER_ROTATE(d1); + rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b)); + + WRITE_TWO_PIXELS(outptr0, rgb); + outptr0 += 4; + + y = GETJSAMPLE(*inptr01++); + r = range_limit[DITHER_565_R(y + cred, d0)]; + g = range_limit[DITHER_565_G(y + cgreen, d0)]; + b = range_limit[DITHER_565_B(y + cblue, d0)]; + d0 = DITHER_ROTATE(d0); + rgb = PACK_SHORT_565(r, g, b); + + y = GETJSAMPLE(*inptr01++); + r = range_limit[DITHER_565_R(y + cred, d1)]; + g = range_limit[DITHER_565_G(y + cgreen, d1)]; + b = range_limit[DITHER_565_B(y + cblue, d1)]; + d1 = DITHER_ROTATE(d1); + rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b)); + + WRITE_TWO_PIXELS(outptr1, rgb); + outptr1 += 4; + } + + /* If image width is odd, do the last output column separately */ + if (cinfo->output_width & 1) { + cb = GETJSAMPLE(*inptr1); + cr = GETJSAMPLE(*inptr2); + cred = Crrtab[cr]; + cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS); + cblue = Cbbtab[cb]; + + y = GETJSAMPLE(*inptr00); + r = range_limit[DITHER_565_R(y + cred, d0)]; + g = range_limit[DITHER_565_G(y + cgreen, d0)]; + b = range_limit[DITHER_565_B(y + cblue, d0)]; + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr0 = rgb; + + y = GETJSAMPLE(*inptr01); + r = range_limit[DITHER_565_R(y + cred, d1)]; + g = range_limit[DITHER_565_G(y + cgreen, d1)]; + b = range_limit[DITHER_565_B(y + cblue, d1)]; + rgb = PACK_SHORT_565(r, g, b); + *(INT16*)outptr1 = rgb; + } +} diff --git a/app/src/main/jni/libjpeg-turbo/jdmrgext.c b/app/src/main/jni/libjpeg-turbo/jdmrgext.c new file mode 100644 index 000000000..1f0a5507a --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdmrgext.c @@ -0,0 +1,185 @@ +/* + * jdmrgext.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1996, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright (C) 2011, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains code for merged upsampling/color conversion. + */ + + +/* This file is included by jdmerge.c */ + + +/* + * Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical. + */ + +INLINE +LOCAL(void) +h2v1_merged_upsample_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + register int y, cred, cgreen, cblue; + int cb, cr; + register JSAMPROW outptr; + JSAMPROW inptr0, inptr1, inptr2; + JDIMENSION col; + /* copy these pointers into registers if possible */ + register JSAMPLE * range_limit = cinfo->sample_range_limit; + int * Crrtab = upsample->Cr_r_tab; + int * Cbbtab = upsample->Cb_b_tab; + INT32 * Crgtab = upsample->Cr_g_tab; + INT32 * Cbgtab = upsample->Cb_g_tab; + SHIFT_TEMPS + + inptr0 = input_buf[0][in_row_group_ctr]; + inptr1 = input_buf[1][in_row_group_ctr]; + inptr2 = input_buf[2][in_row_group_ctr]; + outptr = output_buf[0]; + /* Loop for each pair of output pixels */ + for (col = cinfo->output_width >> 1; col > 0; col--) { + /* Do the chroma part of the calculation */ + cb = GETJSAMPLE(*inptr1++); + cr = GETJSAMPLE(*inptr2++); + cred = Crrtab[cr]; + cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS); + cblue = Cbbtab[cb]; + /* Fetch 2 Y values and emit 2 pixels */ + y = GETJSAMPLE(*inptr0++); + outptr[RGB_RED] = range_limit[y + cred]; + outptr[RGB_GREEN] = range_limit[y + cgreen]; + outptr[RGB_BLUE] = range_limit[y + cblue]; +#ifdef RGB_ALPHA + outptr[RGB_ALPHA] = 0xFF; +#endif + outptr += RGB_PIXELSIZE; + y = GETJSAMPLE(*inptr0++); + outptr[RGB_RED] = range_limit[y + cred]; + outptr[RGB_GREEN] = range_limit[y + cgreen]; + outptr[RGB_BLUE] = range_limit[y + cblue]; +#ifdef RGB_ALPHA + outptr[RGB_ALPHA] = 0xFF; +#endif + outptr += RGB_PIXELSIZE; + } + /* If image width is odd, do the last output column separately */ + if (cinfo->output_width & 1) { + cb = GETJSAMPLE(*inptr1); + cr = GETJSAMPLE(*inptr2); + cred = Crrtab[cr]; + cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS); + cblue = Cbbtab[cb]; + y = GETJSAMPLE(*inptr0); + outptr[RGB_RED] = range_limit[y + cred]; + outptr[RGB_GREEN] = range_limit[y + cgreen]; + outptr[RGB_BLUE] = range_limit[y + cblue]; +#ifdef RGB_ALPHA + outptr[RGB_ALPHA] = 0xFF; +#endif + } +} + + +/* + * Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical. + */ + +INLINE +LOCAL(void) +h2v2_merged_upsample_internal (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + register int y, cred, cgreen, cblue; + int cb, cr; + register JSAMPROW outptr0, outptr1; + JSAMPROW inptr00, inptr01, inptr1, inptr2; + JDIMENSION col; + /* copy these pointers into registers if possible */ + register JSAMPLE * range_limit = cinfo->sample_range_limit; + int * Crrtab = upsample->Cr_r_tab; + int * Cbbtab = upsample->Cb_b_tab; + INT32 * Crgtab = upsample->Cr_g_tab; + INT32 * Cbgtab = upsample->Cb_g_tab; + SHIFT_TEMPS + + inptr00 = input_buf[0][in_row_group_ctr*2]; + inptr01 = input_buf[0][in_row_group_ctr*2 + 1]; + inptr1 = input_buf[1][in_row_group_ctr]; + inptr2 = input_buf[2][in_row_group_ctr]; + outptr0 = output_buf[0]; + outptr1 = output_buf[1]; + /* Loop for each group of output pixels */ + for (col = cinfo->output_width >> 1; col > 0; col--) { + /* Do the chroma part of the calculation */ + cb = GETJSAMPLE(*inptr1++); + cr = GETJSAMPLE(*inptr2++); + cred = Crrtab[cr]; + cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS); + cblue = Cbbtab[cb]; + /* Fetch 4 Y values and emit 4 pixels */ + y = GETJSAMPLE(*inptr00++); + outptr0[RGB_RED] = range_limit[y + cred]; + outptr0[RGB_GREEN] = range_limit[y + cgreen]; + outptr0[RGB_BLUE] = range_limit[y + cblue]; +#ifdef RGB_ALPHA + outptr0[RGB_ALPHA] = 0xFF; +#endif + outptr0 += RGB_PIXELSIZE; + y = GETJSAMPLE(*inptr00++); + outptr0[RGB_RED] = range_limit[y + cred]; + outptr0[RGB_GREEN] = range_limit[y + cgreen]; + outptr0[RGB_BLUE] = range_limit[y + cblue]; +#ifdef RGB_ALPHA + outptr0[RGB_ALPHA] = 0xFF; +#endif + outptr0 += RGB_PIXELSIZE; + y = GETJSAMPLE(*inptr01++); + outptr1[RGB_RED] = range_limit[y + cred]; + outptr1[RGB_GREEN] = range_limit[y + cgreen]; + outptr1[RGB_BLUE] = range_limit[y + cblue]; +#ifdef RGB_ALPHA + outptr1[RGB_ALPHA] = 0xFF; +#endif + outptr1 += RGB_PIXELSIZE; + y = GETJSAMPLE(*inptr01++); + outptr1[RGB_RED] = range_limit[y + cred]; + outptr1[RGB_GREEN] = range_limit[y + cgreen]; + outptr1[RGB_BLUE] = range_limit[y + cblue]; +#ifdef RGB_ALPHA + outptr1[RGB_ALPHA] = 0xFF; +#endif + outptr1 += RGB_PIXELSIZE; + } + /* If image width is odd, do the last output column separately */ + if (cinfo->output_width & 1) { + cb = GETJSAMPLE(*inptr1); + cr = GETJSAMPLE(*inptr2); + cred = Crrtab[cr]; + cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS); + cblue = Cbbtab[cb]; + y = GETJSAMPLE(*inptr00); + outptr0[RGB_RED] = range_limit[y + cred]; + outptr0[RGB_GREEN] = range_limit[y + cgreen]; + outptr0[RGB_BLUE] = range_limit[y + cblue]; +#ifdef RGB_ALPHA + outptr0[RGB_ALPHA] = 0xFF; +#endif + y = GETJSAMPLE(*inptr01); + outptr1[RGB_RED] = range_limit[y + cred]; + outptr1[RGB_GREEN] = range_limit[y + cgreen]; + outptr1[RGB_BLUE] = range_limit[y + cblue]; +#ifdef RGB_ALPHA + outptr1[RGB_ALPHA] = 0xFF; +#endif + } +} diff --git a/app/src/main/jni/libjpeg-turbo/jdphuff.c b/app/src/main/jni/libjpeg-turbo/jdphuff.c new file mode 100644 index 000000000..eae153836 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdphuff.c @@ -0,0 +1,671 @@ +/* + * jdphuff.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1995-1997, Thomas G. Lane. + * It was modified by The libjpeg-turbo Project to include only code relevant + * to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains Huffman entropy decoding routines for progressive JPEG. + * + * Much of the complexity here has to do with supporting input suspension. + * If the data source module demands suspension, we want to be able to back + * up to the start of the current MCU. To do this, we copy state variables + * into local working storage, and update them back to the permanent + * storage only upon successful completion of an MCU. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jdhuff.h" /* Declarations shared with jdhuff.c */ + + +#ifdef D_PROGRESSIVE_SUPPORTED + +/* + * Expanded entropy decoder object for progressive Huffman decoding. + * + * The savable_state subrecord contains fields that change within an MCU, + * but must not be updated permanently until we complete the MCU. + */ + +typedef struct { + unsigned int EOBRUN; /* remaining EOBs in EOBRUN */ + int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */ +} savable_state; + +/* This macro is to work around compilers with missing or broken + * structure assignment. You'll need to fix this code if you have + * such a compiler and you change MAX_COMPS_IN_SCAN. + */ + +#ifndef NO_STRUCT_ASSIGN +#define ASSIGN_STATE(dest,src) ((dest) = (src)) +#else +#if MAX_COMPS_IN_SCAN == 4 +#define ASSIGN_STATE(dest,src) \ + ((dest).EOBRUN = (src).EOBRUN, \ + (dest).last_dc_val[0] = (src).last_dc_val[0], \ + (dest).last_dc_val[1] = (src).last_dc_val[1], \ + (dest).last_dc_val[2] = (src).last_dc_val[2], \ + (dest).last_dc_val[3] = (src).last_dc_val[3]) +#endif +#endif + + +typedef struct { + struct jpeg_entropy_decoder pub; /* public fields */ + + /* These fields are loaded into local variables at start of each MCU. + * In case of suspension, we exit WITHOUT updating them. + */ + bitread_perm_state bitstate; /* Bit buffer at start of MCU */ + savable_state saved; /* Other state at start of MCU */ + + /* These fields are NOT loaded into local working state. */ + unsigned int restarts_to_go; /* MCUs left in this restart interval */ + + /* Pointers to derived tables (these workspaces have image lifespan) */ + d_derived_tbl * derived_tbls[NUM_HUFF_TBLS]; + + d_derived_tbl * ac_derived_tbl; /* active table during an AC scan */ +} phuff_entropy_decoder; + +typedef phuff_entropy_decoder * phuff_entropy_ptr; + +/* Forward declarations */ +METHODDEF(boolean) decode_mcu_DC_first (j_decompress_ptr cinfo, + JBLOCKROW *MCU_data); +METHODDEF(boolean) decode_mcu_AC_first (j_decompress_ptr cinfo, + JBLOCKROW *MCU_data); +METHODDEF(boolean) decode_mcu_DC_refine (j_decompress_ptr cinfo, + JBLOCKROW *MCU_data); +METHODDEF(boolean) decode_mcu_AC_refine (j_decompress_ptr cinfo, + JBLOCKROW *MCU_data); + + +/* + * Initialize for a Huffman-compressed scan. + */ + +METHODDEF(void) +start_pass_phuff_decoder (j_decompress_ptr cinfo) +{ + phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy; + boolean is_DC_band, bad; + int ci, coefi, tbl; + int *coef_bit_ptr; + jpeg_component_info * compptr; + + is_DC_band = (cinfo->Ss == 0); + + /* Validate scan parameters */ + bad = FALSE; + if (is_DC_band) { + if (cinfo->Se != 0) + bad = TRUE; + } else { + /* need not check Ss/Se < 0 since they came from unsigned bytes */ + if (cinfo->Ss > cinfo->Se || cinfo->Se >= DCTSIZE2) + bad = TRUE; + /* AC scans may have only one component */ + if (cinfo->comps_in_scan != 1) + bad = TRUE; + } + if (cinfo->Ah != 0) { + /* Successive approximation refinement scan: must have Al = Ah-1. */ + if (cinfo->Al != cinfo->Ah-1) + bad = TRUE; + } + if (cinfo->Al > 13) /* need not check for < 0 */ + bad = TRUE; + /* Arguably the maximum Al value should be less than 13 for 8-bit precision, + * but the spec doesn't say so, and we try to be liberal about what we + * accept. Note: large Al values could result in out-of-range DC + * coefficients during early scans, leading to bizarre displays due to + * overflows in the IDCT math. But we won't crash. + */ + if (bad) + ERREXIT4(cinfo, JERR_BAD_PROGRESSION, + cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al); + /* Update progression status, and verify that scan order is legal. + * Note that inter-scan inconsistencies are treated as warnings + * not fatal errors ... not clear if this is right way to behave. + */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + int cindex = cinfo->cur_comp_info[ci]->component_index; + coef_bit_ptr = & cinfo->coef_bits[cindex][0]; + if (!is_DC_band && coef_bit_ptr[0] < 0) /* AC without prior DC scan */ + WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0); + for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) { + int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi]; + if (cinfo->Ah != expected) + WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi); + coef_bit_ptr[coefi] = cinfo->Al; + } + } + + /* Select MCU decoding routine */ + if (cinfo->Ah == 0) { + if (is_DC_band) + entropy->pub.decode_mcu = decode_mcu_DC_first; + else + entropy->pub.decode_mcu = decode_mcu_AC_first; + } else { + if (is_DC_band) + entropy->pub.decode_mcu = decode_mcu_DC_refine; + else + entropy->pub.decode_mcu = decode_mcu_AC_refine; + } + + for (ci = 0; ci < cinfo->comps_in_scan; ci++) { + compptr = cinfo->cur_comp_info[ci]; + /* Make sure requested tables are present, and compute derived tables. + * We may build same derived table more than once, but it's not expensive. + */ + if (is_DC_band) { + if (cinfo->Ah == 0) { /* DC refinement needs no table */ + tbl = compptr->dc_tbl_no; + jpeg_make_d_derived_tbl(cinfo, TRUE, tbl, + & entropy->derived_tbls[tbl]); + } + } else { + tbl = compptr->ac_tbl_no; + jpeg_make_d_derived_tbl(cinfo, FALSE, tbl, + & entropy->derived_tbls[tbl]); + /* remember the single active table */ + entropy->ac_derived_tbl = entropy->derived_tbls[tbl]; + } + /* Initialize DC predictions to 0 */ + entropy->saved.last_dc_val[ci] = 0; + } + + /* Initialize bitread state variables */ + entropy->bitstate.bits_left = 0; + entropy->bitstate.get_buffer = 0; /* unnecessary, but keeps Purify quiet */ + entropy->pub.insufficient_data = FALSE; + + /* Initialize private state variables */ + entropy->saved.EOBRUN = 0; + + /* Initialize restart counter */ + entropy->restarts_to_go = cinfo->restart_interval; +} + + +/* + * Figure F.12: extend sign bit. + * On some machines, a shift and add will be faster than a table lookup. + */ + +#define AVOID_TABLES +#ifdef AVOID_TABLES + +#define HUFF_EXTEND(x,s) ((x) < (1<<((s)-1)) ? (x) + (((-1)<<(s)) + 1) : (x)) + +#else + +#define HUFF_EXTEND(x,s) ((x) < extend_test[s] ? (x) + extend_offset[s] : (x)) + +static const int extend_test[16] = /* entry n is 2**(n-1) */ + { 0, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080, + 0x0100, 0x0200, 0x0400, 0x0800, 0x1000, 0x2000, 0x4000 }; + +static const int extend_offset[16] = /* entry n is (-1 << n) + 1 */ + { 0, ((-1)<<1) + 1, ((-1)<<2) + 1, ((-1)<<3) + 1, ((-1)<<4) + 1, + ((-1)<<5) + 1, ((-1)<<6) + 1, ((-1)<<7) + 1, ((-1)<<8) + 1, + ((-1)<<9) + 1, ((-1)<<10) + 1, ((-1)<<11) + 1, ((-1)<<12) + 1, + ((-1)<<13) + 1, ((-1)<<14) + 1, ((-1)<<15) + 1 }; + +#endif /* AVOID_TABLES */ + + +/* + * Check for a restart marker & resynchronize decoder. + * Returns FALSE if must suspend. + */ + +LOCAL(boolean) +process_restart (j_decompress_ptr cinfo) +{ + phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy; + int ci; + + /* Throw away any unused bits remaining in bit buffer; */ + /* include any full bytes in next_marker's count of discarded bytes */ + cinfo->marker->discarded_bytes += entropy->bitstate.bits_left / 8; + entropy->bitstate.bits_left = 0; + + /* Advance past the RSTn marker */ + if (! (*cinfo->marker->read_restart_marker) (cinfo)) + return FALSE; + + /* Re-initialize DC predictions to 0 */ + for (ci = 0; ci < cinfo->comps_in_scan; ci++) + entropy->saved.last_dc_val[ci] = 0; + /* Re-init EOB run count, too */ + entropy->saved.EOBRUN = 0; + + /* Reset restart counter */ + entropy->restarts_to_go = cinfo->restart_interval; + + /* Reset out-of-data flag, unless read_restart_marker left us smack up + * against a marker. In that case we will end up treating the next data + * segment as empty, and we can avoid producing bogus output pixels by + * leaving the flag set. + */ + if (cinfo->unread_marker == 0) + entropy->pub.insufficient_data = FALSE; + + return TRUE; +} + + +/* + * Huffman MCU decoding. + * Each of these routines decodes and returns one MCU's worth of + * Huffman-compressed coefficients. + * The coefficients are reordered from zigzag order into natural array order, + * but are not dequantized. + * + * The i'th block of the MCU is stored into the block pointed to by + * MCU_data[i]. WE ASSUME THIS AREA IS INITIALLY ZEROED BY THE CALLER. + * + * We return FALSE if data source requested suspension. In that case no + * changes have been made to permanent state. (Exception: some output + * coefficients may already have been assigned. This is harmless for + * spectral selection, since we'll just re-assign them on the next call. + * Successive approximation AC refinement has to be more careful, however.) + */ + +/* + * MCU decoding for DC initial scan (either spectral selection, + * or first pass of successive approximation). + */ + +METHODDEF(boolean) +decode_mcu_DC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data) +{ + phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy; + int Al = cinfo->Al; + register int s, r; + int blkn, ci; + JBLOCKROW block; + BITREAD_STATE_VARS; + savable_state state; + d_derived_tbl * tbl; + jpeg_component_info * compptr; + + /* Process restart marker if needed; may have to suspend */ + if (cinfo->restart_interval) { + if (entropy->restarts_to_go == 0) + if (! process_restart(cinfo)) + return FALSE; + } + + /* If we've run out of data, just leave the MCU set to zeroes. + * This way, we return uniform gray for the remainder of the segment. + */ + if (! entropy->pub.insufficient_data) { + + /* Load up working state */ + BITREAD_LOAD_STATE(cinfo,entropy->bitstate); + ASSIGN_STATE(state, entropy->saved); + + /* Outer loop handles each block in the MCU */ + + for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) { + block = MCU_data[blkn]; + ci = cinfo->MCU_membership[blkn]; + compptr = cinfo->cur_comp_info[ci]; + tbl = entropy->derived_tbls[compptr->dc_tbl_no]; + + /* Decode a single block's worth of coefficients */ + + /* Section F.2.2.1: decode the DC coefficient difference */ + HUFF_DECODE(s, br_state, tbl, return FALSE, label1); + if (s) { + CHECK_BIT_BUFFER(br_state, s, return FALSE); + r = GET_BITS(s); + s = HUFF_EXTEND(r, s); + } + + /* Convert DC difference to actual value, update last_dc_val */ + s += state.last_dc_val[ci]; + state.last_dc_val[ci] = s; + /* Scale and output the coefficient (assumes jpeg_natural_order[0]=0) */ + (*block)[0] = (JCOEF) (s << Al); + } + + /* Completed MCU, so update state */ + BITREAD_SAVE_STATE(cinfo,entropy->bitstate); + ASSIGN_STATE(entropy->saved, state); + } + + /* Account for restart interval (no-op if not using restarts) */ + entropy->restarts_to_go--; + + return TRUE; +} + + +/* + * MCU decoding for AC initial scan (either spectral selection, + * or first pass of successive approximation). + */ + +METHODDEF(boolean) +decode_mcu_AC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data) +{ + phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy; + int Se = cinfo->Se; + int Al = cinfo->Al; + register int s, k, r; + unsigned int EOBRUN; + JBLOCKROW block; + BITREAD_STATE_VARS; + d_derived_tbl * tbl; + + /* Process restart marker if needed; may have to suspend */ + if (cinfo->restart_interval) { + if (entropy->restarts_to_go == 0) + if (! process_restart(cinfo)) + return FALSE; + } + + /* If we've run out of data, just leave the MCU set to zeroes. + * This way, we return uniform gray for the remainder of the segment. + */ + if (! entropy->pub.insufficient_data) { + + /* Load up working state. + * We can avoid loading/saving bitread state if in an EOB run. + */ + EOBRUN = entropy->saved.EOBRUN; /* only part of saved state we need */ + + /* There is always only one block per MCU */ + + if (EOBRUN > 0) /* if it's a band of zeroes... */ + EOBRUN--; /* ...process it now (we do nothing) */ + else { + BITREAD_LOAD_STATE(cinfo,entropy->bitstate); + block = MCU_data[0]; + tbl = entropy->ac_derived_tbl; + + for (k = cinfo->Ss; k <= Se; k++) { + HUFF_DECODE(s, br_state, tbl, return FALSE, label2); + r = s >> 4; + s &= 15; + if (s) { + k += r; + CHECK_BIT_BUFFER(br_state, s, return FALSE); + r = GET_BITS(s); + s = HUFF_EXTEND(r, s); + /* Scale and output coefficient in natural (dezigzagged) order */ + (*block)[jpeg_natural_order[k]] = (JCOEF) (s << Al); + } else { + if (r == 15) { /* ZRL */ + k += 15; /* skip 15 zeroes in band */ + } else { /* EOBr, run length is 2^r + appended bits */ + EOBRUN = 1 << r; + if (r) { /* EOBr, r > 0 */ + CHECK_BIT_BUFFER(br_state, r, return FALSE); + r = GET_BITS(r); + EOBRUN += r; + } + EOBRUN--; /* this band is processed at this moment */ + break; /* force end-of-band */ + } + } + } + + BITREAD_SAVE_STATE(cinfo,entropy->bitstate); + } + + /* Completed MCU, so update state */ + entropy->saved.EOBRUN = EOBRUN; /* only part of saved state we need */ + } + + /* Account for restart interval (no-op if not using restarts) */ + entropy->restarts_to_go--; + + return TRUE; +} + + +/* + * MCU decoding for DC successive approximation refinement scan. + * Note: we assume such scans can be multi-component, although the spec + * is not very clear on the point. + */ + +METHODDEF(boolean) +decode_mcu_DC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data) +{ + phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy; + int p1 = 1 << cinfo->Al; /* 1 in the bit position being coded */ + int blkn; + JBLOCKROW block; + BITREAD_STATE_VARS; + + /* Process restart marker if needed; may have to suspend */ + if (cinfo->restart_interval) { + if (entropy->restarts_to_go == 0) + if (! process_restart(cinfo)) + return FALSE; + } + + /* Not worth the cycles to check insufficient_data here, + * since we will not change the data anyway if we read zeroes. + */ + + /* Load up working state */ + BITREAD_LOAD_STATE(cinfo,entropy->bitstate); + + /* Outer loop handles each block in the MCU */ + + for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) { + block = MCU_data[blkn]; + + /* Encoded data is simply the next bit of the two's-complement DC value */ + CHECK_BIT_BUFFER(br_state, 1, return FALSE); + if (GET_BITS(1)) + (*block)[0] |= p1; + /* Note: since we use |=, repeating the assignment later is safe */ + } + + /* Completed MCU, so update state */ + BITREAD_SAVE_STATE(cinfo,entropy->bitstate); + + /* Account for restart interval (no-op if not using restarts) */ + entropy->restarts_to_go--; + + return TRUE; +} + + +/* + * MCU decoding for AC successive approximation refinement scan. + */ + +METHODDEF(boolean) +decode_mcu_AC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data) +{ + phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy; + int Se = cinfo->Se; + int p1 = 1 << cinfo->Al; /* 1 in the bit position being coded */ + int m1 = (-1) << cinfo->Al; /* -1 in the bit position being coded */ + register int s, k, r; + unsigned int EOBRUN; + JBLOCKROW block; + JCOEFPTR thiscoef; + BITREAD_STATE_VARS; + d_derived_tbl * tbl; + int num_newnz; + int newnz_pos[DCTSIZE2]; + + /* Process restart marker if needed; may have to suspend */ + if (cinfo->restart_interval) { + if (entropy->restarts_to_go == 0) + if (! process_restart(cinfo)) + return FALSE; + } + + /* If we've run out of data, don't modify the MCU. + */ + if (! entropy->pub.insufficient_data) { + + /* Load up working state */ + BITREAD_LOAD_STATE(cinfo,entropy->bitstate); + EOBRUN = entropy->saved.EOBRUN; /* only part of saved state we need */ + + /* There is always only one block per MCU */ + block = MCU_data[0]; + tbl = entropy->ac_derived_tbl; + + /* If we are forced to suspend, we must undo the assignments to any newly + * nonzero coefficients in the block, because otherwise we'd get confused + * next time about which coefficients were already nonzero. + * But we need not undo addition of bits to already-nonzero coefficients; + * instead, we can test the current bit to see if we already did it. + */ + num_newnz = 0; + + /* initialize coefficient loop counter to start of band */ + k = cinfo->Ss; + + if (EOBRUN == 0) { + for (; k <= Se; k++) { + HUFF_DECODE(s, br_state, tbl, goto undoit, label3); + r = s >> 4; + s &= 15; + if (s) { + if (s != 1) /* size of new coef should always be 1 */ + WARNMS(cinfo, JWRN_HUFF_BAD_CODE); + CHECK_BIT_BUFFER(br_state, 1, goto undoit); + if (GET_BITS(1)) + s = p1; /* newly nonzero coef is positive */ + else + s = m1; /* newly nonzero coef is negative */ + } else { + if (r != 15) { + EOBRUN = 1 << r; /* EOBr, run length is 2^r + appended bits */ + if (r) { + CHECK_BIT_BUFFER(br_state, r, goto undoit); + r = GET_BITS(r); + EOBRUN += r; + } + break; /* rest of block is handled by EOB logic */ + } + /* note s = 0 for processing ZRL */ + } + /* Advance over already-nonzero coefs and r still-zero coefs, + * appending correction bits to the nonzeroes. A correction bit is 1 + * if the absolute value of the coefficient must be increased. + */ + do { + thiscoef = *block + jpeg_natural_order[k]; + if (*thiscoef != 0) { + CHECK_BIT_BUFFER(br_state, 1, goto undoit); + if (GET_BITS(1)) { + if ((*thiscoef & p1) == 0) { /* do nothing if already set it */ + if (*thiscoef >= 0) + *thiscoef += p1; + else + *thiscoef += m1; + } + } + } else { + if (--r < 0) + break; /* reached target zero coefficient */ + } + k++; + } while (k <= Se); + if (s) { + int pos = jpeg_natural_order[k]; + /* Output newly nonzero coefficient */ + (*block)[pos] = (JCOEF) s; + /* Remember its position in case we have to suspend */ + newnz_pos[num_newnz++] = pos; + } + } + } + + if (EOBRUN > 0) { + /* Scan any remaining coefficient positions after the end-of-band + * (the last newly nonzero coefficient, if any). Append a correction + * bit to each already-nonzero coefficient. A correction bit is 1 + * if the absolute value of the coefficient must be increased. + */ + for (; k <= Se; k++) { + thiscoef = *block + jpeg_natural_order[k]; + if (*thiscoef != 0) { + CHECK_BIT_BUFFER(br_state, 1, goto undoit); + if (GET_BITS(1)) { + if ((*thiscoef & p1) == 0) { /* do nothing if already changed it */ + if (*thiscoef >= 0) + *thiscoef += p1; + else + *thiscoef += m1; + } + } + } + } + /* Count one block completed in EOB run */ + EOBRUN--; + } + + /* Completed MCU, so update state */ + BITREAD_SAVE_STATE(cinfo,entropy->bitstate); + entropy->saved.EOBRUN = EOBRUN; /* only part of saved state we need */ + } + + /* Account for restart interval (no-op if not using restarts) */ + entropy->restarts_to_go--; + + return TRUE; + +undoit: + /* Re-zero any output coefficients that we made newly nonzero */ + while (num_newnz > 0) + (*block)[newnz_pos[--num_newnz]] = 0; + + return FALSE; +} + + +/* + * Module initialization routine for progressive Huffman entropy decoding. + */ + +GLOBAL(void) +jinit_phuff_decoder (j_decompress_ptr cinfo) +{ + phuff_entropy_ptr entropy; + int *coef_bit_ptr; + int ci, i; + + entropy = (phuff_entropy_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(phuff_entropy_decoder)); + cinfo->entropy = (struct jpeg_entropy_decoder *) entropy; + entropy->pub.start_pass = start_pass_phuff_decoder; + + /* Mark derived tables unallocated */ + for (i = 0; i < NUM_HUFF_TBLS; i++) { + entropy->derived_tbls[i] = NULL; + } + + /* Create progression status table */ + cinfo->coef_bits = (int (*)[DCTSIZE2]) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + cinfo->num_components*DCTSIZE2*sizeof(int)); + coef_bit_ptr = & cinfo->coef_bits[0][0]; + for (ci = 0; ci < cinfo->num_components; ci++) + for (i = 0; i < DCTSIZE2; i++) + *coef_bit_ptr++ = -1; +} + +#endif /* D_PROGRESSIVE_SUPPORTED */ diff --git a/app/src/main/jni/libjpeg-turbo/jdpostct.c b/app/src/main/jni/libjpeg-turbo/jdpostct.c new file mode 100644 index 000000000..9eef9ee0e --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdpostct.c @@ -0,0 +1,289 @@ +/* + * jdpostct.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1996, Thomas G. Lane. + * It was modified by The libjpeg-turbo Project to include only code relevant + * to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains the decompression postprocessing controller. + * This controller manages the upsampling, color conversion, and color + * quantization/reduction steps; specifically, it controls the buffering + * between upsample/color conversion and color quantization/reduction. + * + * If no color quantization/reduction is required, then this module has no + * work to do, and it just hands off to the upsample/color conversion code. + * An integrated upsample/convert/quantize process would replace this module + * entirely. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" + + +/* Private buffer controller object */ + +typedef struct { + struct jpeg_d_post_controller pub; /* public fields */ + + /* Color quantization source buffer: this holds output data from + * the upsample/color conversion step to be passed to the quantizer. + * For two-pass color quantization, we need a full-image buffer; + * for one-pass operation, a strip buffer is sufficient. + */ + jvirt_sarray_ptr whole_image; /* virtual array, or NULL if one-pass */ + JSAMPARRAY buffer; /* strip buffer, or current strip of virtual */ + JDIMENSION strip_height; /* buffer size in rows */ + /* for two-pass mode only: */ + JDIMENSION starting_row; /* row # of first row in current strip */ + JDIMENSION next_row; /* index of next row to fill/empty in strip */ +} my_post_controller; + +typedef my_post_controller * my_post_ptr; + + +/* Forward declarations */ +METHODDEF(void) post_process_1pass + (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, + JDIMENSION *in_row_group_ctr, JDIMENSION in_row_groups_avail, + JSAMPARRAY output_buf, JDIMENSION *out_row_ctr, + JDIMENSION out_rows_avail); +#ifdef QUANT_2PASS_SUPPORTED +METHODDEF(void) post_process_prepass + (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, + JDIMENSION *in_row_group_ctr, JDIMENSION in_row_groups_avail, + JSAMPARRAY output_buf, JDIMENSION *out_row_ctr, + JDIMENSION out_rows_avail); +METHODDEF(void) post_process_2pass + (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, + JDIMENSION *in_row_group_ctr, JDIMENSION in_row_groups_avail, + JSAMPARRAY output_buf, JDIMENSION *out_row_ctr, + JDIMENSION out_rows_avail); +#endif + + +/* + * Initialize for a processing pass. + */ + +METHODDEF(void) +start_pass_dpost (j_decompress_ptr cinfo, J_BUF_MODE pass_mode) +{ + my_post_ptr post = (my_post_ptr) cinfo->post; + + switch (pass_mode) { + case JBUF_PASS_THRU: + if (cinfo->quantize_colors) { + /* Single-pass processing with color quantization. */ + post->pub.post_process_data = post_process_1pass; + /* We could be doing buffered-image output before starting a 2-pass + * color quantization; in that case, jinit_d_post_controller did not + * allocate a strip buffer. Use the virtual-array buffer as workspace. + */ + if (post->buffer == NULL) { + post->buffer = (*cinfo->mem->access_virt_sarray) + ((j_common_ptr) cinfo, post->whole_image, + (JDIMENSION) 0, post->strip_height, TRUE); + } + } else { + /* For single-pass processing without color quantization, + * I have no work to do; just call the upsampler directly. + */ + post->pub.post_process_data = cinfo->upsample->upsample; + } + break; +#ifdef QUANT_2PASS_SUPPORTED + case JBUF_SAVE_AND_PASS: + /* First pass of 2-pass quantization */ + if (post->whole_image == NULL) + ERREXIT(cinfo, JERR_BAD_BUFFER_MODE); + post->pub.post_process_data = post_process_prepass; + break; + case JBUF_CRANK_DEST: + /* Second pass of 2-pass quantization */ + if (post->whole_image == NULL) + ERREXIT(cinfo, JERR_BAD_BUFFER_MODE); + post->pub.post_process_data = post_process_2pass; + break; +#endif /* QUANT_2PASS_SUPPORTED */ + default: + ERREXIT(cinfo, JERR_BAD_BUFFER_MODE); + break; + } + post->starting_row = post->next_row = 0; +} + + +/* + * Process some data in the one-pass (strip buffer) case. + * This is used for color precision reduction as well as one-pass quantization. + */ + +METHODDEF(void) +post_process_1pass (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr, + JDIMENSION in_row_groups_avail, + JSAMPARRAY output_buf, JDIMENSION *out_row_ctr, + JDIMENSION out_rows_avail) +{ + my_post_ptr post = (my_post_ptr) cinfo->post; + JDIMENSION num_rows, max_rows; + + /* Fill the buffer, but not more than what we can dump out in one go. */ + /* Note we rely on the upsampler to detect bottom of image. */ + max_rows = out_rows_avail - *out_row_ctr; + if (max_rows > post->strip_height) + max_rows = post->strip_height; + num_rows = 0; + (*cinfo->upsample->upsample) (cinfo, + input_buf, in_row_group_ctr, in_row_groups_avail, + post->buffer, &num_rows, max_rows); + /* Quantize and emit data. */ + (*cinfo->cquantize->color_quantize) (cinfo, + post->buffer, output_buf + *out_row_ctr, (int) num_rows); + *out_row_ctr += num_rows; +} + + +#ifdef QUANT_2PASS_SUPPORTED + +/* + * Process some data in the first pass of 2-pass quantization. + */ + +METHODDEF(void) +post_process_prepass (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr, + JDIMENSION in_row_groups_avail, + JSAMPARRAY output_buf, JDIMENSION *out_row_ctr, + JDIMENSION out_rows_avail) +{ + my_post_ptr post = (my_post_ptr) cinfo->post; + JDIMENSION old_next_row, num_rows; + + /* Reposition virtual buffer if at start of strip. */ + if (post->next_row == 0) { + post->buffer = (*cinfo->mem->access_virt_sarray) + ((j_common_ptr) cinfo, post->whole_image, + post->starting_row, post->strip_height, TRUE); + } + + /* Upsample some data (up to a strip height's worth). */ + old_next_row = post->next_row; + (*cinfo->upsample->upsample) (cinfo, + input_buf, in_row_group_ctr, in_row_groups_avail, + post->buffer, &post->next_row, post->strip_height); + + /* Allow quantizer to scan new data. No data is emitted, */ + /* but we advance out_row_ctr so outer loop can tell when we're done. */ + if (post->next_row > old_next_row) { + num_rows = post->next_row - old_next_row; + (*cinfo->cquantize->color_quantize) (cinfo, post->buffer + old_next_row, + (JSAMPARRAY) NULL, (int) num_rows); + *out_row_ctr += num_rows; + } + + /* Advance if we filled the strip. */ + if (post->next_row >= post->strip_height) { + post->starting_row += post->strip_height; + post->next_row = 0; + } +} + + +/* + * Process some data in the second pass of 2-pass quantization. + */ + +METHODDEF(void) +post_process_2pass (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr, + JDIMENSION in_row_groups_avail, + JSAMPARRAY output_buf, JDIMENSION *out_row_ctr, + JDIMENSION out_rows_avail) +{ + my_post_ptr post = (my_post_ptr) cinfo->post; + JDIMENSION num_rows, max_rows; + + /* Reposition virtual buffer if at start of strip. */ + if (post->next_row == 0) { + post->buffer = (*cinfo->mem->access_virt_sarray) + ((j_common_ptr) cinfo, post->whole_image, + post->starting_row, post->strip_height, FALSE); + } + + /* Determine number of rows to emit. */ + num_rows = post->strip_height - post->next_row; /* available in strip */ + max_rows = out_rows_avail - *out_row_ctr; /* available in output area */ + if (num_rows > max_rows) + num_rows = max_rows; + /* We have to check bottom of image here, can't depend on upsampler. */ + max_rows = cinfo->output_height - post->starting_row; + if (num_rows > max_rows) + num_rows = max_rows; + + /* Quantize and emit data. */ + (*cinfo->cquantize->color_quantize) (cinfo, + post->buffer + post->next_row, output_buf + *out_row_ctr, + (int) num_rows); + *out_row_ctr += num_rows; + + /* Advance if we filled the strip. */ + post->next_row += num_rows; + if (post->next_row >= post->strip_height) { + post->starting_row += post->strip_height; + post->next_row = 0; + } +} + +#endif /* QUANT_2PASS_SUPPORTED */ + + +/* + * Initialize postprocessing controller. + */ + +GLOBAL(void) +jinit_d_post_controller (j_decompress_ptr cinfo, boolean need_full_buffer) +{ + my_post_ptr post; + + post = (my_post_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(my_post_controller)); + cinfo->post = (struct jpeg_d_post_controller *) post; + post->pub.start_pass = start_pass_dpost; + post->whole_image = NULL; /* flag for no virtual arrays */ + post->buffer = NULL; /* flag for no strip buffer */ + + /* Create the quantization buffer, if needed */ + if (cinfo->quantize_colors) { + /* The buffer strip height is max_v_samp_factor, which is typically + * an efficient number of rows for upsampling to return. + * (In the presence of output rescaling, we might want to be smarter?) + */ + post->strip_height = (JDIMENSION) cinfo->max_v_samp_factor; + if (need_full_buffer) { + /* Two-pass color quantization: need full-image storage. */ + /* We round up the number of rows to a multiple of the strip height. */ +#ifdef QUANT_2PASS_SUPPORTED + post->whole_image = (*cinfo->mem->request_virt_sarray) + ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE, + cinfo->output_width * cinfo->out_color_components, + (JDIMENSION) jround_up((long) cinfo->output_height, + (long) post->strip_height), + post->strip_height); +#else + ERREXIT(cinfo, JERR_BAD_BUFFER_MODE); +#endif /* QUANT_2PASS_SUPPORTED */ + } else { + /* One-pass color quantization: just make a strip buffer. */ + post->buffer = (*cinfo->mem->alloc_sarray) + ((j_common_ptr) cinfo, JPOOL_IMAGE, + cinfo->output_width * cinfo->out_color_components, + post->strip_height); + } + } +} diff --git a/app/src/main/jni/libjpeg-turbo/jdsample.c b/app/src/main/jni/libjpeg-turbo/jdsample.c new file mode 100644 index 000000000..275296670 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdsample.c @@ -0,0 +1,504 @@ +/* + * jdsample.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1996, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright 2009 Pierre Ossman for Cendio AB + * Copyright (C) 2010, D. R. Commander. + * Copyright (C) 2014, MIPS Technologies, Inc., California + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains upsampling routines. + * + * Upsampling input data is counted in "row groups". A row group + * is defined to be (v_samp_factor * DCT_scaled_size / min_DCT_scaled_size) + * sample rows of each component. Upsampling will normally produce + * max_v_samp_factor pixel rows from each row group (but this could vary + * if the upsampler is applying a scale factor of its own). + * + * An excellent reference for image resampling is + * Digital Image Warping, George Wolberg, 1990. + * Pub. by IEEE Computer Society Press, Los Alamitos, CA. ISBN 0-8186-8944-7. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jsimd.h" +#include "jpegcomp.h" + + +/* Pointer to routine to upsample a single component */ +typedef void (*upsample1_ptr) (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr); + +/* Private subobject */ + +typedef struct { + struct jpeg_upsampler pub; /* public fields */ + + /* Color conversion buffer. When using separate upsampling and color + * conversion steps, this buffer holds one upsampled row group until it + * has been color converted and output. + * Note: we do not allocate any storage for component(s) which are full-size, + * ie do not need rescaling. The corresponding entry of color_buf[] is + * simply set to point to the input data array, thereby avoiding copying. + */ + JSAMPARRAY color_buf[MAX_COMPONENTS]; + + /* Per-component upsampling method pointers */ + upsample1_ptr methods[MAX_COMPONENTS]; + + int next_row_out; /* counts rows emitted from color_buf */ + JDIMENSION rows_to_go; /* counts rows remaining in image */ + + /* Height of an input row group for each component. */ + int rowgroup_height[MAX_COMPONENTS]; + + /* These arrays save pixel expansion factors so that int_expand need not + * recompute them each time. They are unused for other upsampling methods. + */ + UINT8 h_expand[MAX_COMPONENTS]; + UINT8 v_expand[MAX_COMPONENTS]; +} my_upsampler; + +typedef my_upsampler * my_upsample_ptr; + + +/* + * Initialize for an upsampling pass. + */ + +METHODDEF(void) +start_pass_upsample (j_decompress_ptr cinfo) +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + + /* Mark the conversion buffer empty */ + upsample->next_row_out = cinfo->max_v_samp_factor; + /* Initialize total-height counter for detecting bottom of image */ + upsample->rows_to_go = cinfo->output_height; +} + + +/* + * Control routine to do upsampling (and color conversion). + * + * In this version we upsample each component independently. + * We upsample one row group into the conversion buffer, then apply + * color conversion a row at a time. + */ + +METHODDEF(void) +sep_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr, + JDIMENSION in_row_groups_avail, + JSAMPARRAY output_buf, JDIMENSION *out_row_ctr, + JDIMENSION out_rows_avail) +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + int ci; + jpeg_component_info * compptr; + JDIMENSION num_rows; + + /* Fill the conversion buffer, if it's empty */ + if (upsample->next_row_out >= cinfo->max_v_samp_factor) { + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + /* Invoke per-component upsample method. Notice we pass a POINTER + * to color_buf[ci], so that fullsize_upsample can change it. + */ + (*upsample->methods[ci]) (cinfo, compptr, + input_buf[ci] + (*in_row_group_ctr * upsample->rowgroup_height[ci]), + upsample->color_buf + ci); + } + upsample->next_row_out = 0; + } + + /* Color-convert and emit rows */ + + /* How many we have in the buffer: */ + num_rows = (JDIMENSION) (cinfo->max_v_samp_factor - upsample->next_row_out); + /* Not more than the distance to the end of the image. Need this test + * in case the image height is not a multiple of max_v_samp_factor: + */ + if (num_rows > upsample->rows_to_go) + num_rows = upsample->rows_to_go; + /* And not more than what the client can accept: */ + out_rows_avail -= *out_row_ctr; + if (num_rows > out_rows_avail) + num_rows = out_rows_avail; + + (*cinfo->cconvert->color_convert) (cinfo, upsample->color_buf, + (JDIMENSION) upsample->next_row_out, + output_buf + *out_row_ctr, + (int) num_rows); + + /* Adjust counts */ + *out_row_ctr += num_rows; + upsample->rows_to_go -= num_rows; + upsample->next_row_out += num_rows; + /* When the buffer is emptied, declare this input row group consumed */ + if (upsample->next_row_out >= cinfo->max_v_samp_factor) + (*in_row_group_ctr)++; +} + + +/* + * These are the routines invoked by sep_upsample to upsample pixel values + * of a single component. One row group is processed per call. + */ + + +/* + * For full-size components, we just make color_buf[ci] point at the + * input buffer, and thus avoid copying any data. Note that this is + * safe only because sep_upsample doesn't declare the input row group + * "consumed" until we are done color converting and emitting it. + */ + +METHODDEF(void) +fullsize_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr) +{ + *output_data_ptr = input_data; +} + + +/* + * This is a no-op version used for "uninteresting" components. + * These components will not be referenced by color conversion. + */ + +METHODDEF(void) +noop_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr) +{ + *output_data_ptr = NULL; /* safety check */ +} + + +/* + * This version handles any integral sampling ratios. + * This is not used for typical JPEG files, so it need not be fast. + * Nor, for that matter, is it particularly accurate: the algorithm is + * simple replication of the input pixel onto the corresponding output + * pixels. The hi-falutin sampling literature refers to this as a + * "box filter". A box filter tends to introduce visible artifacts, + * so if you are actually going to use 3:1 or 4:1 sampling ratios + * you would be well advised to improve this code. + */ + +METHODDEF(void) +int_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr) +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + JSAMPARRAY output_data = *output_data_ptr; + register JSAMPROW inptr, outptr; + register JSAMPLE invalue; + register int h; + JSAMPROW outend; + int h_expand, v_expand; + int inrow, outrow; + + h_expand = upsample->h_expand[compptr->component_index]; + v_expand = upsample->v_expand[compptr->component_index]; + + inrow = outrow = 0; + while (outrow < cinfo->max_v_samp_factor) { + /* Generate one output row with proper horizontal expansion */ + inptr = input_data[inrow]; + outptr = output_data[outrow]; + outend = outptr + cinfo->output_width; + while (outptr < outend) { + invalue = *inptr++; /* don't need GETJSAMPLE() here */ + for (h = h_expand; h > 0; h--) { + *outptr++ = invalue; + } + } + /* Generate any additional output rows by duplicating the first one */ + if (v_expand > 1) { + jcopy_sample_rows(output_data, outrow, output_data, outrow+1, + v_expand-1, cinfo->output_width); + } + inrow++; + outrow += v_expand; + } +} + + +/* + * Fast processing for the common case of 2:1 horizontal and 1:1 vertical. + * It's still a box filter. + */ + +METHODDEF(void) +h2v1_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr) +{ + JSAMPARRAY output_data = *output_data_ptr; + register JSAMPROW inptr, outptr; + register JSAMPLE invalue; + JSAMPROW outend; + int inrow; + + for (inrow = 0; inrow < cinfo->max_v_samp_factor; inrow++) { + inptr = input_data[inrow]; + outptr = output_data[inrow]; + outend = outptr + cinfo->output_width; + while (outptr < outend) { + invalue = *inptr++; /* don't need GETJSAMPLE() here */ + *outptr++ = invalue; + *outptr++ = invalue; + } + } +} + + +/* + * Fast processing for the common case of 2:1 horizontal and 2:1 vertical. + * It's still a box filter. + */ + +METHODDEF(void) +h2v2_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr) +{ + JSAMPARRAY output_data = *output_data_ptr; + register JSAMPROW inptr, outptr; + register JSAMPLE invalue; + JSAMPROW outend; + int inrow, outrow; + + inrow = outrow = 0; + while (outrow < cinfo->max_v_samp_factor) { + inptr = input_data[inrow]; + outptr = output_data[outrow]; + outend = outptr + cinfo->output_width; + while (outptr < outend) { + invalue = *inptr++; /* don't need GETJSAMPLE() here */ + *outptr++ = invalue; + *outptr++ = invalue; + } + jcopy_sample_rows(output_data, outrow, output_data, outrow+1, + 1, cinfo->output_width); + inrow++; + outrow += 2; + } +} + + +/* + * Fancy processing for the common case of 2:1 horizontal and 1:1 vertical. + * + * The upsampling algorithm is linear interpolation between pixel centers, + * also known as a "triangle filter". This is a good compromise between + * speed and visual quality. The centers of the output pixels are 1/4 and 3/4 + * of the way between input pixel centers. + * + * A note about the "bias" calculations: when rounding fractional values to + * integer, we do not want to always round 0.5 up to the next integer. + * If we did that, we'd introduce a noticeable bias towards larger values. + * Instead, this code is arranged so that 0.5 will be rounded up or down at + * alternate pixel locations (a simple ordered dither pattern). + */ + +METHODDEF(void) +h2v1_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr) +{ + JSAMPARRAY output_data = *output_data_ptr; + register JSAMPROW inptr, outptr; + register int invalue; + register JDIMENSION colctr; + int inrow; + + for (inrow = 0; inrow < cinfo->max_v_samp_factor; inrow++) { + inptr = input_data[inrow]; + outptr = output_data[inrow]; + /* Special case for first column */ + invalue = GETJSAMPLE(*inptr++); + *outptr++ = (JSAMPLE) invalue; + *outptr++ = (JSAMPLE) ((invalue * 3 + GETJSAMPLE(*inptr) + 2) >> 2); + + for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) { + /* General case: 3/4 * nearer pixel + 1/4 * further pixel */ + invalue = GETJSAMPLE(*inptr++) * 3; + *outptr++ = (JSAMPLE) ((invalue + GETJSAMPLE(inptr[-2]) + 1) >> 2); + *outptr++ = (JSAMPLE) ((invalue + GETJSAMPLE(*inptr) + 2) >> 2); + } + + /* Special case for last column */ + invalue = GETJSAMPLE(*inptr); + *outptr++ = (JSAMPLE) ((invalue * 3 + GETJSAMPLE(inptr[-1]) + 1) >> 2); + *outptr++ = (JSAMPLE) invalue; + } +} + + +/* + * Fancy processing for the common case of 2:1 horizontal and 2:1 vertical. + * Again a triangle filter; see comments for h2v1 case, above. + * + * It is OK for us to reference the adjacent input rows because we demanded + * context from the main buffer controller (see initialization code). + */ + +METHODDEF(void) +h2v2_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr) +{ + JSAMPARRAY output_data = *output_data_ptr; + register JSAMPROW inptr0, inptr1, outptr; +#if BITS_IN_JSAMPLE == 8 + register int thiscolsum, lastcolsum, nextcolsum; +#else + register INT32 thiscolsum, lastcolsum, nextcolsum; +#endif + register JDIMENSION colctr; + int inrow, outrow, v; + + inrow = outrow = 0; + while (outrow < cinfo->max_v_samp_factor) { + for (v = 0; v < 2; v++) { + /* inptr0 points to nearest input row, inptr1 points to next nearest */ + inptr0 = input_data[inrow]; + if (v == 0) /* next nearest is row above */ + inptr1 = input_data[inrow-1]; + else /* next nearest is row below */ + inptr1 = input_data[inrow+1]; + outptr = output_data[outrow++]; + + /* Special case for first column */ + thiscolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++); + nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++); + *outptr++ = (JSAMPLE) ((thiscolsum * 4 + 8) >> 4); + *outptr++ = (JSAMPLE) ((thiscolsum * 3 + nextcolsum + 7) >> 4); + lastcolsum = thiscolsum; thiscolsum = nextcolsum; + + for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) { + /* General case: 3/4 * nearer pixel + 1/4 * further pixel in each */ + /* dimension, thus 9/16, 3/16, 3/16, 1/16 overall */ + nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++); + *outptr++ = (JSAMPLE) ((thiscolsum * 3 + lastcolsum + 8) >> 4); + *outptr++ = (JSAMPLE) ((thiscolsum * 3 + nextcolsum + 7) >> 4); + lastcolsum = thiscolsum; thiscolsum = nextcolsum; + } + + /* Special case for last column */ + *outptr++ = (JSAMPLE) ((thiscolsum * 3 + lastcolsum + 8) >> 4); + *outptr++ = (JSAMPLE) ((thiscolsum * 4 + 7) >> 4); + } + inrow++; + } +} + + +/* + * Module initialization routine for upsampling. + */ + +GLOBAL(void) +jinit_upsampler (j_decompress_ptr cinfo) +{ + my_upsample_ptr upsample; + int ci; + jpeg_component_info * compptr; + boolean need_buffer, do_fancy; + int h_in_group, v_in_group, h_out_group, v_out_group; + + upsample = (my_upsample_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(my_upsampler)); + cinfo->upsample = (struct jpeg_upsampler *) upsample; + upsample->pub.start_pass = start_pass_upsample; + upsample->pub.upsample = sep_upsample; + upsample->pub.need_context_rows = FALSE; /* until we find out differently */ + + if (cinfo->CCIR601_sampling) /* this isn't supported */ + ERREXIT(cinfo, JERR_CCIR601_NOTIMPL); + + /* jdmainct.c doesn't support context rows when min_DCT_scaled_size = 1, + * so don't ask for it. + */ + do_fancy = cinfo->do_fancy_upsampling && cinfo->_min_DCT_scaled_size > 1; + + /* Verify we can handle the sampling factors, select per-component methods, + * and create storage as needed. + */ + for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components; + ci++, compptr++) { + /* Compute size of an "input group" after IDCT scaling. This many samples + * are to be converted to max_h_samp_factor * max_v_samp_factor pixels. + */ + h_in_group = (compptr->h_samp_factor * compptr->_DCT_scaled_size) / + cinfo->_min_DCT_scaled_size; + v_in_group = (compptr->v_samp_factor * compptr->_DCT_scaled_size) / + cinfo->_min_DCT_scaled_size; + h_out_group = cinfo->max_h_samp_factor; + v_out_group = cinfo->max_v_samp_factor; + upsample->rowgroup_height[ci] = v_in_group; /* save for use later */ + need_buffer = TRUE; + if (! compptr->component_needed) { + /* Don't bother to upsample an uninteresting component. */ + upsample->methods[ci] = noop_upsample; + need_buffer = FALSE; + } else if (h_in_group == h_out_group && v_in_group == v_out_group) { + /* Fullsize components can be processed without any work. */ + upsample->methods[ci] = fullsize_upsample; + need_buffer = FALSE; + } else if (h_in_group * 2 == h_out_group && + v_in_group == v_out_group) { + /* Special cases for 2h1v upsampling */ + if (do_fancy && compptr->downsampled_width > 2) { + if (jsimd_can_h2v1_fancy_upsample()) + upsample->methods[ci] = jsimd_h2v1_fancy_upsample; + else + upsample->methods[ci] = h2v1_fancy_upsample; + } else { + if (jsimd_can_h2v1_upsample()) + upsample->methods[ci] = jsimd_h2v1_upsample; + else + upsample->methods[ci] = h2v1_upsample; + } + } else if (h_in_group * 2 == h_out_group && + v_in_group * 2 == v_out_group) { + /* Special cases for 2h2v upsampling */ + if (do_fancy && compptr->downsampled_width > 2) { + if (jsimd_can_h2v2_fancy_upsample()) + upsample->methods[ci] = jsimd_h2v2_fancy_upsample; + else + upsample->methods[ci] = h2v2_fancy_upsample; + upsample->pub.need_context_rows = TRUE; + } else { + if (jsimd_can_h2v2_upsample()) + upsample->methods[ci] = jsimd_h2v2_upsample; + else + upsample->methods[ci] = h2v2_upsample; + } + } else if ((h_out_group % h_in_group) == 0 && + (v_out_group % v_in_group) == 0) { + /* Generic integral-factors upsampling method */ +#if defined(__mips__) + if (jsimd_can_int_upsample()) + upsample->methods[ci] = jsimd_int_upsample; + else +#endif + upsample->methods[ci] = int_upsample; + upsample->h_expand[ci] = (UINT8) (h_out_group / h_in_group); + upsample->v_expand[ci] = (UINT8) (v_out_group / v_in_group); + } else + ERREXIT(cinfo, JERR_FRACT_SAMPLE_NOTIMPL); + if (need_buffer) { + upsample->color_buf[ci] = (*cinfo->mem->alloc_sarray) + ((j_common_ptr) cinfo, JPOOL_IMAGE, + (JDIMENSION) jround_up((long) cinfo->output_width, + (long) cinfo->max_h_samp_factor), + (JDIMENSION) cinfo->max_v_samp_factor); + } + } +} diff --git a/app/src/main/jni/libjpeg-turbo/jdtrans.c b/app/src/main/jni/libjpeg-turbo/jdtrans.c new file mode 100644 index 000000000..86fda3b8b --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jdtrans.c @@ -0,0 +1,154 @@ +/* + * jdtrans.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1995-1997, Thomas G. Lane. + * It was modified by The libjpeg-turbo Project to include only code relevant + * to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains library routines for transcoding decompression, + * that is, reading raw DCT coefficient arrays from an input JPEG file. + * The routines in jdapimin.c will also be needed by a transcoder. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" + + +/* Forward declarations */ +LOCAL(void) transdecode_master_selection (j_decompress_ptr cinfo); + + +/* + * Read the coefficient arrays from a JPEG file. + * jpeg_read_header must be completed before calling this. + * + * The entire image is read into a set of virtual coefficient-block arrays, + * one per component. The return value is a pointer to the array of + * virtual-array descriptors. These can be manipulated directly via the + * JPEG memory manager, or handed off to jpeg_write_coefficients(). + * To release the memory occupied by the virtual arrays, call + * jpeg_finish_decompress() when done with the data. + * + * An alternative usage is to simply obtain access to the coefficient arrays + * during a buffered-image-mode decompression operation. This is allowed + * after any jpeg_finish_output() call. The arrays can be accessed until + * jpeg_finish_decompress() is called. (Note that any call to the library + * may reposition the arrays, so don't rely on access_virt_barray() results + * to stay valid across library calls.) + * + * Returns NULL if suspended. This case need be checked only if + * a suspending data source is used. + */ + +GLOBAL(jvirt_barray_ptr *) +jpeg_read_coefficients (j_decompress_ptr cinfo) +{ + if (cinfo->global_state == DSTATE_READY) { + /* First call: initialize active modules */ + transdecode_master_selection(cinfo); + cinfo->global_state = DSTATE_RDCOEFS; + } + if (cinfo->global_state == DSTATE_RDCOEFS) { + /* Absorb whole file into the coef buffer */ + for (;;) { + int retcode; + /* Call progress monitor hook if present */ + if (cinfo->progress != NULL) + (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo); + /* Absorb some more input */ + retcode = (*cinfo->inputctl->consume_input) (cinfo); + if (retcode == JPEG_SUSPENDED) + return NULL; + if (retcode == JPEG_REACHED_EOI) + break; + /* Advance progress counter if appropriate */ + if (cinfo->progress != NULL && + (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) { + if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) { + /* startup underestimated number of scans; ratchet up one scan */ + cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows; + } + } + } + /* Set state so that jpeg_finish_decompress does the right thing */ + cinfo->global_state = DSTATE_STOPPING; + } + /* At this point we should be in state DSTATE_STOPPING if being used + * standalone, or in state DSTATE_BUFIMAGE if being invoked to get access + * to the coefficients during a full buffered-image-mode decompression. + */ + if ((cinfo->global_state == DSTATE_STOPPING || + cinfo->global_state == DSTATE_BUFIMAGE) && cinfo->buffered_image) { + return cinfo->coef->coef_arrays; + } + /* Oops, improper usage */ + ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state); + return NULL; /* keep compiler happy */ +} + + +/* + * Master selection of decompression modules for transcoding. + * This substitutes for jdmaster.c's initialization of the full decompressor. + */ + +LOCAL(void) +transdecode_master_selection (j_decompress_ptr cinfo) +{ + /* This is effectively a buffered-image operation. */ + cinfo->buffered_image = TRUE; + +#if JPEG_LIB_VERSION >= 80 + /* Compute output image dimensions and related values. */ + jpeg_core_output_dimensions(cinfo); +#endif + + /* Entropy decoding: either Huffman or arithmetic coding. */ + if (cinfo->arith_code) { +#ifdef D_ARITH_CODING_SUPPORTED + jinit_arith_decoder(cinfo); +#else + ERREXIT(cinfo, JERR_ARITH_NOTIMPL); +#endif + } else { + if (cinfo->progressive_mode) { +#ifdef D_PROGRESSIVE_SUPPORTED + jinit_phuff_decoder(cinfo); +#else + ERREXIT(cinfo, JERR_NOT_COMPILED); +#endif + } else + jinit_huff_decoder(cinfo); + } + + /* Always get a full-image coefficient buffer. */ + jinit_d_coef_controller(cinfo, TRUE); + + /* We can now tell the memory manager to allocate virtual arrays. */ + (*cinfo->mem->realize_virt_arrays) ((j_common_ptr) cinfo); + + /* Initialize input side of decompressor to consume first scan. */ + (*cinfo->inputctl->start_input_pass) (cinfo); + + /* Initialize progress monitoring. */ + if (cinfo->progress != NULL) { + int nscans; + /* Estimate number of scans to set pass_limit. */ + if (cinfo->progressive_mode) { + /* Arbitrarily estimate 2 interleaved DC scans + 3 AC scans/component. */ + nscans = 2 + 3 * cinfo->num_components; + } else if (cinfo->inputctl->has_multiple_scans) { + /* For a nonprogressive multiscan file, estimate 1 scan per component. */ + nscans = cinfo->num_components; + } else { + nscans = 1; + } + cinfo->progress->pass_counter = 0L; + cinfo->progress->pass_limit = (long) cinfo->total_iMCU_rows * nscans; + cinfo->progress->completed_passes = 0; + cinfo->progress->total_passes = 1; + } +} diff --git a/app/src/main/jni/libjpeg-turbo/jerror.c b/app/src/main/jni/libjpeg-turbo/jerror.c new file mode 100644 index 000000000..cd3098da7 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jerror.c @@ -0,0 +1,250 @@ +/* + * jerror.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1998, Thomas G. Lane. + * It was modified by The libjpeg-turbo Project to include only code relevant + * to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains simple error-reporting and trace-message routines. + * These are suitable for Unix-like systems and others where writing to + * stderr is the right thing to do. Many applications will want to replace + * some or all of these routines. + * + * If you define USE_WINDOWS_MESSAGEBOX in jconfig.h or in the makefile, + * you get a Windows-specific hack to display error messages in a dialog box. + * It ain't much, but it beats dropping error messages into the bit bucket, + * which is what happens to output to stderr under most Windows C compilers. + * + * These routines are used by both the compression and decompression code. + */ + +/* this is not a core library module, so it doesn't define JPEG_INTERNALS */ +#include "jinclude.h" +#include "jpeglib.h" +#include "jversion.h" +#include "jerror.h" + +#ifdef USE_WINDOWS_MESSAGEBOX +#include +#endif + +#ifndef EXIT_FAILURE /* define exit() codes if not provided */ +#define EXIT_FAILURE 1 +#endif + + +/* + * Create the message string table. + * We do this from the master message list in jerror.h by re-reading + * jerror.h with a suitable definition for macro JMESSAGE. + * The message table is made an external symbol just in case any applications + * want to refer to it directly. + */ + +#define JMESSAGE(code,string) string , + +const char * const jpeg_std_message_table[] = { +#include "jerror.h" + NULL +}; + + +/* + * Error exit handler: must not return to caller. + * + * Applications may override this if they want to get control back after + * an error. Typically one would longjmp somewhere instead of exiting. + * The setjmp buffer can be made a private field within an expanded error + * handler object. Note that the info needed to generate an error message + * is stored in the error object, so you can generate the message now or + * later, at your convenience. + * You should make sure that the JPEG object is cleaned up (with jpeg_abort + * or jpeg_destroy) at some point. + */ + +METHODDEF(void) +error_exit (j_common_ptr cinfo) +{ + /* Always display the message */ + (*cinfo->err->output_message) (cinfo); + + /* Let the memory manager delete any temp files before we die */ + jpeg_destroy(cinfo); + + exit(EXIT_FAILURE); +} + + +/* + * Actual output of an error or trace message. + * Applications may override this method to send JPEG messages somewhere + * other than stderr. + * + * On Windows, printing to stderr is generally completely useless, + * so we provide optional code to produce an error-dialog popup. + * Most Windows applications will still prefer to override this routine, + * but if they don't, it'll do something at least marginally useful. + * + * NOTE: to use the library in an environment that doesn't support the + * C stdio library, you may have to delete the call to fprintf() entirely, + * not just not use this routine. + */ + +METHODDEF(void) +output_message (j_common_ptr cinfo) +{ + char buffer[JMSG_LENGTH_MAX]; + + /* Create the message */ + (*cinfo->err->format_message) (cinfo, buffer); + +#ifdef USE_WINDOWS_MESSAGEBOX + /* Display it in a message dialog box */ + MessageBox(GetActiveWindow(), buffer, "JPEG Library Error", + MB_OK | MB_ICONERROR); +#else + /* Send it to stderr, adding a newline */ + fprintf(stderr, "%s\n", buffer); +#endif +} + + +/* + * Decide whether to emit a trace or warning message. + * msg_level is one of: + * -1: recoverable corrupt-data warning, may want to abort. + * 0: important advisory messages (always display to user). + * 1: first level of tracing detail. + * 2,3,...: successively more detailed tracing messages. + * An application might override this method if it wanted to abort on warnings + * or change the policy about which messages to display. + */ + +METHODDEF(void) +emit_message (j_common_ptr cinfo, int msg_level) +{ + struct jpeg_error_mgr * err = cinfo->err; + + if (msg_level < 0) { + /* It's a warning message. Since corrupt files may generate many warnings, + * the policy implemented here is to show only the first warning, + * unless trace_level >= 3. + */ + if (err->num_warnings == 0 || err->trace_level >= 3) + (*err->output_message) (cinfo); + /* Always count warnings in num_warnings. */ + err->num_warnings++; + } else { + /* It's a trace message. Show it if trace_level >= msg_level. */ + if (err->trace_level >= msg_level) + (*err->output_message) (cinfo); + } +} + + +/* + * Format a message string for the most recent JPEG error or message. + * The message is stored into buffer, which should be at least JMSG_LENGTH_MAX + * characters. Note that no '\n' character is added to the string. + * Few applications should need to override this method. + */ + +METHODDEF(void) +format_message (j_common_ptr cinfo, char * buffer) +{ + struct jpeg_error_mgr * err = cinfo->err; + int msg_code = err->msg_code; + const char * msgtext = NULL; + const char * msgptr; + char ch; + boolean isstring; + + /* Look up message string in proper table */ + if (msg_code > 0 && msg_code <= err->last_jpeg_message) { + msgtext = err->jpeg_message_table[msg_code]; + } else if (err->addon_message_table != NULL && + msg_code >= err->first_addon_message && + msg_code <= err->last_addon_message) { + msgtext = err->addon_message_table[msg_code - err->first_addon_message]; + } + + /* Defend against bogus message number */ + if (msgtext == NULL) { + err->msg_parm.i[0] = msg_code; + msgtext = err->jpeg_message_table[0]; + } + + /* Check for string parameter, as indicated by %s in the message text */ + isstring = FALSE; + msgptr = msgtext; + while ((ch = *msgptr++) != '\0') { + if (ch == '%') { + if (*msgptr == 's') isstring = TRUE; + break; + } + } + + /* Format the message into the passed buffer */ + if (isstring) + sprintf(buffer, msgtext, err->msg_parm.s); + else + sprintf(buffer, msgtext, + err->msg_parm.i[0], err->msg_parm.i[1], + err->msg_parm.i[2], err->msg_parm.i[3], + err->msg_parm.i[4], err->msg_parm.i[5], + err->msg_parm.i[6], err->msg_parm.i[7]); +} + + +/* + * Reset error state variables at start of a new image. + * This is called during compression startup to reset trace/error + * processing to default state, without losing any application-specific + * method pointers. An application might possibly want to override + * this method if it has additional error processing state. + */ + +METHODDEF(void) +reset_error_mgr (j_common_ptr cinfo) +{ + cinfo->err->num_warnings = 0; + /* trace_level is not reset since it is an application-supplied parameter */ + cinfo->err->msg_code = 0; /* may be useful as a flag for "no error" */ +} + + +/* + * Fill in the standard error-handling methods in a jpeg_error_mgr object. + * Typical call is: + * struct jpeg_compress_struct cinfo; + * struct jpeg_error_mgr err; + * + * cinfo.err = jpeg_std_error(&err); + * after which the application may override some of the methods. + */ + +GLOBAL(struct jpeg_error_mgr *) +jpeg_std_error (struct jpeg_error_mgr * err) +{ + err->error_exit = error_exit; + err->emit_message = emit_message; + err->output_message = output_message; + err->format_message = format_message; + err->reset_error_mgr = reset_error_mgr; + + err->trace_level = 0; /* default = no tracing */ + err->num_warnings = 0; /* no warnings emitted yet */ + err->msg_code = 0; /* may be useful as a flag for "no error" */ + + /* Initialize message table pointers */ + err->jpeg_message_table = jpeg_std_message_table; + err->last_jpeg_message = (int) JMSG_LASTMSGCODE - 1; + + err->addon_message_table = NULL; + err->first_addon_message = 0; /* for safety */ + err->last_addon_message = 0; + + return err; +} diff --git a/app/src/main/jni/libjpeg-turbo/jerror.h b/app/src/main/jni/libjpeg-turbo/jerror.h new file mode 100644 index 000000000..402613e00 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jerror.h @@ -0,0 +1,316 @@ +/* + * jerror.h + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1997, Thomas G. Lane. + * Modified 1997-2009 by Guido Vollbeding. + * libjpeg-turbo Modifications: + * Copyright (C) 2014, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file defines the error and message codes for the JPEG library. + * Edit this file to add new codes, or to translate the message strings to + * some other language. + * A set of error-reporting macros are defined too. Some applications using + * the JPEG library may wish to include this file to get the error codes + * and/or the macros. + */ + +/* + * To define the enum list of message codes, include this file without + * defining macro JMESSAGE. To create a message string table, include it + * again with a suitable JMESSAGE definition (see jerror.c for an example). + */ +#ifndef JMESSAGE +#ifndef JERROR_H +/* First time through, define the enum list */ +#define JMAKE_ENUM_LIST +#else +/* Repeated inclusions of this file are no-ops unless JMESSAGE is defined */ +#define JMESSAGE(code,string) +#endif /* JERROR_H */ +#endif /* JMESSAGE */ + +#ifdef JMAKE_ENUM_LIST + +typedef enum { + +#define JMESSAGE(code,string) code , + +#endif /* JMAKE_ENUM_LIST */ + +JMESSAGE(JMSG_NOMESSAGE, "Bogus message code %d") /* Must be first entry! */ + +/* For maintenance convenience, list is alphabetical by message code name */ +#if JPEG_LIB_VERSION < 70 +JMESSAGE(JERR_ARITH_NOTIMPL, + "Sorry, arithmetic coding is not implemented") +#endif +JMESSAGE(JERR_BAD_ALIGN_TYPE, "ALIGN_TYPE is wrong, please fix") +JMESSAGE(JERR_BAD_ALLOC_CHUNK, "MAX_ALLOC_CHUNK is wrong, please fix") +JMESSAGE(JERR_BAD_BUFFER_MODE, "Bogus buffer control mode") +JMESSAGE(JERR_BAD_COMPONENT_ID, "Invalid component ID %d in SOS") +#if JPEG_LIB_VERSION >= 70 +JMESSAGE(JERR_BAD_CROP_SPEC, "Invalid crop request") +#endif +JMESSAGE(JERR_BAD_DCT_COEF, "DCT coefficient out of range") +JMESSAGE(JERR_BAD_DCTSIZE, "IDCT output block size %d not supported") +#if JPEG_LIB_VERSION >= 70 +JMESSAGE(JERR_BAD_DROP_SAMPLING, + "Component index %d: mismatching sampling ratio %d:%d, %d:%d, %c") +#endif +JMESSAGE(JERR_BAD_HUFF_TABLE, "Bogus Huffman table definition") +JMESSAGE(JERR_BAD_IN_COLORSPACE, "Bogus input colorspace") +JMESSAGE(JERR_BAD_J_COLORSPACE, "Bogus JPEG colorspace") +JMESSAGE(JERR_BAD_LENGTH, "Bogus marker length") +JMESSAGE(JERR_BAD_LIB_VERSION, + "Wrong JPEG library version: library is %d, caller expects %d") +JMESSAGE(JERR_BAD_MCU_SIZE, "Sampling factors too large for interleaved scan") +JMESSAGE(JERR_BAD_POOL_ID, "Invalid memory pool code %d") +JMESSAGE(JERR_BAD_PRECISION, "Unsupported JPEG data precision %d") +JMESSAGE(JERR_BAD_PROGRESSION, + "Invalid progressive parameters Ss=%d Se=%d Ah=%d Al=%d") +JMESSAGE(JERR_BAD_PROG_SCRIPT, + "Invalid progressive parameters at scan script entry %d") +JMESSAGE(JERR_BAD_SAMPLING, "Bogus sampling factors") +JMESSAGE(JERR_BAD_SCAN_SCRIPT, "Invalid scan script at entry %d") +JMESSAGE(JERR_BAD_STATE, "Improper call to JPEG library in state %d") +JMESSAGE(JERR_BAD_STRUCT_SIZE, + "JPEG parameter struct mismatch: library thinks size is %u, caller expects %u") +JMESSAGE(JERR_BAD_VIRTUAL_ACCESS, "Bogus virtual array access") +JMESSAGE(JERR_BUFFER_SIZE, "Buffer passed to JPEG library is too small") +JMESSAGE(JERR_CANT_SUSPEND, "Suspension not allowed here") +JMESSAGE(JERR_CCIR601_NOTIMPL, "CCIR601 sampling not implemented yet") +JMESSAGE(JERR_COMPONENT_COUNT, "Too many color components: %d, max %d") +JMESSAGE(JERR_CONVERSION_NOTIMPL, "Unsupported color conversion request") +JMESSAGE(JERR_DAC_INDEX, "Bogus DAC index %d") +JMESSAGE(JERR_DAC_VALUE, "Bogus DAC value 0x%x") +JMESSAGE(JERR_DHT_INDEX, "Bogus DHT index %d") +JMESSAGE(JERR_DQT_INDEX, "Bogus DQT index %d") +JMESSAGE(JERR_EMPTY_IMAGE, "Empty JPEG image (DNL not supported)") +JMESSAGE(JERR_EMS_READ, "Read from EMS failed") +JMESSAGE(JERR_EMS_WRITE, "Write to EMS failed") +JMESSAGE(JERR_EOI_EXPECTED, "Didn't expect more than one scan") +JMESSAGE(JERR_FILE_READ, "Input file read error") +JMESSAGE(JERR_FILE_WRITE, "Output file write error --- out of disk space?") +JMESSAGE(JERR_FRACT_SAMPLE_NOTIMPL, "Fractional sampling not implemented yet") +JMESSAGE(JERR_HUFF_CLEN_OVERFLOW, "Huffman code size table overflow") +JMESSAGE(JERR_HUFF_MISSING_CODE, "Missing Huffman code table entry") +JMESSAGE(JERR_IMAGE_TOO_BIG, "Maximum supported image dimension is %u pixels") +JMESSAGE(JERR_INPUT_EMPTY, "Empty input file") +JMESSAGE(JERR_INPUT_EOF, "Premature end of input file") +JMESSAGE(JERR_MISMATCHED_QUANT_TABLE, + "Cannot transcode due to multiple use of quantization table %d") +JMESSAGE(JERR_MISSING_DATA, "Scan script does not transmit all data") +JMESSAGE(JERR_MODE_CHANGE, "Invalid color quantization mode change") +JMESSAGE(JERR_NOTIMPL, "Not implemented yet") +JMESSAGE(JERR_NOT_COMPILED, "Requested feature was omitted at compile time") +#if JPEG_LIB_VERSION >= 70 +JMESSAGE(JERR_NO_ARITH_TABLE, "Arithmetic table 0x%02x was not defined") +#endif +JMESSAGE(JERR_NO_BACKING_STORE, "Backing store not supported") +JMESSAGE(JERR_NO_HUFF_TABLE, "Huffman table 0x%02x was not defined") +JMESSAGE(JERR_NO_IMAGE, "JPEG datastream contains no image") +JMESSAGE(JERR_NO_QUANT_TABLE, "Quantization table 0x%02x was not defined") +JMESSAGE(JERR_NO_SOI, "Not a JPEG file: starts with 0x%02x 0x%02x") +JMESSAGE(JERR_OUT_OF_MEMORY, "Insufficient memory (case %d)") +JMESSAGE(JERR_QUANT_COMPONENTS, + "Cannot quantize more than %d color components") +JMESSAGE(JERR_QUANT_FEW_COLORS, "Cannot quantize to fewer than %d colors") +JMESSAGE(JERR_QUANT_MANY_COLORS, "Cannot quantize to more than %d colors") +JMESSAGE(JERR_SOF_DUPLICATE, "Invalid JPEG file structure: two SOF markers") +JMESSAGE(JERR_SOF_NO_SOS, "Invalid JPEG file structure: missing SOS marker") +JMESSAGE(JERR_SOF_UNSUPPORTED, "Unsupported JPEG process: SOF type 0x%02x") +JMESSAGE(JERR_SOI_DUPLICATE, "Invalid JPEG file structure: two SOI markers") +JMESSAGE(JERR_SOS_NO_SOF, "Invalid JPEG file structure: SOS before SOF") +JMESSAGE(JERR_TFILE_CREATE, "Failed to create temporary file %s") +JMESSAGE(JERR_TFILE_READ, "Read failed on temporary file") +JMESSAGE(JERR_TFILE_SEEK, "Seek failed on temporary file") +JMESSAGE(JERR_TFILE_WRITE, + "Write failed on temporary file --- out of disk space?") +JMESSAGE(JERR_TOO_LITTLE_DATA, "Application transferred too few scanlines") +JMESSAGE(JERR_UNKNOWN_MARKER, "Unsupported marker type 0x%02x") +JMESSAGE(JERR_VIRTUAL_BUG, "Virtual array controller messed up") +JMESSAGE(JERR_WIDTH_OVERFLOW, "Image too wide for this implementation") +JMESSAGE(JERR_XMS_READ, "Read from XMS failed") +JMESSAGE(JERR_XMS_WRITE, "Write to XMS failed") +JMESSAGE(JMSG_COPYRIGHT, JCOPYRIGHT_SHORT) +JMESSAGE(JMSG_VERSION, JVERSION) +JMESSAGE(JTRC_16BIT_TABLES, + "Caution: quantization tables are too coarse for baseline JPEG") +JMESSAGE(JTRC_ADOBE, + "Adobe APP14 marker: version %d, flags 0x%04x 0x%04x, transform %d") +JMESSAGE(JTRC_APP0, "Unknown APP0 marker (not JFIF), length %u") +JMESSAGE(JTRC_APP14, "Unknown APP14 marker (not Adobe), length %u") +JMESSAGE(JTRC_DAC, "Define Arithmetic Table 0x%02x: 0x%02x") +JMESSAGE(JTRC_DHT, "Define Huffman Table 0x%02x") +JMESSAGE(JTRC_DQT, "Define Quantization Table %d precision %d") +JMESSAGE(JTRC_DRI, "Define Restart Interval %u") +JMESSAGE(JTRC_EMS_CLOSE, "Freed EMS handle %u") +JMESSAGE(JTRC_EMS_OPEN, "Obtained EMS handle %u") +JMESSAGE(JTRC_EOI, "End Of Image") +JMESSAGE(JTRC_HUFFBITS, " %3d %3d %3d %3d %3d %3d %3d %3d") +JMESSAGE(JTRC_JFIF, "JFIF APP0 marker: version %d.%02d, density %dx%d %d") +JMESSAGE(JTRC_JFIF_BADTHUMBNAILSIZE, + "Warning: thumbnail image size does not match data length %u") +JMESSAGE(JTRC_JFIF_EXTENSION, + "JFIF extension marker: type 0x%02x, length %u") +JMESSAGE(JTRC_JFIF_THUMBNAIL, " with %d x %d thumbnail image") +JMESSAGE(JTRC_MISC_MARKER, "Miscellaneous marker 0x%02x, length %u") +JMESSAGE(JTRC_PARMLESS_MARKER, "Unexpected marker 0x%02x") +JMESSAGE(JTRC_QUANTVALS, " %4u %4u %4u %4u %4u %4u %4u %4u") +JMESSAGE(JTRC_QUANT_3_NCOLORS, "Quantizing to %d = %d*%d*%d colors") +JMESSAGE(JTRC_QUANT_NCOLORS, "Quantizing to %d colors") +JMESSAGE(JTRC_QUANT_SELECTED, "Selected %d colors for quantization") +JMESSAGE(JTRC_RECOVERY_ACTION, "At marker 0x%02x, recovery action %d") +JMESSAGE(JTRC_RST, "RST%d") +JMESSAGE(JTRC_SMOOTH_NOTIMPL, + "Smoothing not supported with nonstandard sampling ratios") +JMESSAGE(JTRC_SOF, "Start Of Frame 0x%02x: width=%u, height=%u, components=%d") +JMESSAGE(JTRC_SOF_COMPONENT, " Component %d: %dhx%dv q=%d") +JMESSAGE(JTRC_SOI, "Start of Image") +JMESSAGE(JTRC_SOS, "Start Of Scan: %d components") +JMESSAGE(JTRC_SOS_COMPONENT, " Component %d: dc=%d ac=%d") +JMESSAGE(JTRC_SOS_PARAMS, " Ss=%d, Se=%d, Ah=%d, Al=%d") +JMESSAGE(JTRC_TFILE_CLOSE, "Closed temporary file %s") +JMESSAGE(JTRC_TFILE_OPEN, "Opened temporary file %s") +JMESSAGE(JTRC_THUMB_JPEG, + "JFIF extension marker: JPEG-compressed thumbnail image, length %u") +JMESSAGE(JTRC_THUMB_PALETTE, + "JFIF extension marker: palette thumbnail image, length %u") +JMESSAGE(JTRC_THUMB_RGB, + "JFIF extension marker: RGB thumbnail image, length %u") +JMESSAGE(JTRC_UNKNOWN_IDS, + "Unrecognized component IDs %d %d %d, assuming YCbCr") +JMESSAGE(JTRC_XMS_CLOSE, "Freed XMS handle %u") +JMESSAGE(JTRC_XMS_OPEN, "Obtained XMS handle %u") +JMESSAGE(JWRN_ADOBE_XFORM, "Unknown Adobe color transform code %d") +#if JPEG_LIB_VERSION >= 70 +JMESSAGE(JWRN_ARITH_BAD_CODE, "Corrupt JPEG data: bad arithmetic code") +#endif +JMESSAGE(JWRN_BOGUS_PROGRESSION, + "Inconsistent progression sequence for component %d coefficient %d") +JMESSAGE(JWRN_EXTRANEOUS_DATA, + "Corrupt JPEG data: %u extraneous bytes before marker 0x%02x") +JMESSAGE(JWRN_HIT_MARKER, "Corrupt JPEG data: premature end of data segment") +JMESSAGE(JWRN_HUFF_BAD_CODE, "Corrupt JPEG data: bad Huffman code") +JMESSAGE(JWRN_JFIF_MAJOR, "Warning: unknown JFIF revision number %d.%02d") +JMESSAGE(JWRN_JPEG_EOF, "Premature end of JPEG file") +JMESSAGE(JWRN_MUST_RESYNC, + "Corrupt JPEG data: found marker 0x%02x instead of RST%d") +JMESSAGE(JWRN_NOT_SEQUENTIAL, "Invalid SOS parameters for sequential JPEG") +JMESSAGE(JWRN_TOO_MUCH_DATA, "Application transferred too many scanlines") +#if JPEG_LIB_VERSION < 70 +JMESSAGE(JERR_BAD_CROP_SPEC, "Invalid crop request") +#if defined(C_ARITH_CODING_SUPPORTED) || defined(D_ARITH_CODING_SUPPORTED) +JMESSAGE(JERR_NO_ARITH_TABLE, "Arithmetic table 0x%02x was not defined") +JMESSAGE(JWRN_ARITH_BAD_CODE, "Corrupt JPEG data: bad arithmetic code") +#endif +#endif + +#ifdef JMAKE_ENUM_LIST + + JMSG_LASTMSGCODE +} J_MESSAGE_CODE; + +#undef JMAKE_ENUM_LIST +#endif /* JMAKE_ENUM_LIST */ + +/* Zap JMESSAGE macro so that future re-inclusions do nothing by default */ +#undef JMESSAGE + + +#ifndef JERROR_H +#define JERROR_H + +/* Macros to simplify using the error and trace message stuff */ +/* The first parameter is either type of cinfo pointer */ + +/* Fatal errors (print message and exit) */ +#define ERREXIT(cinfo,code) \ + ((cinfo)->err->msg_code = (code), \ + (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo))) +#define ERREXIT1(cinfo,code,p1) \ + ((cinfo)->err->msg_code = (code), \ + (cinfo)->err->msg_parm.i[0] = (p1), \ + (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo))) +#define ERREXIT2(cinfo,code,p1,p2) \ + ((cinfo)->err->msg_code = (code), \ + (cinfo)->err->msg_parm.i[0] = (p1), \ + (cinfo)->err->msg_parm.i[1] = (p2), \ + (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo))) +#define ERREXIT3(cinfo,code,p1,p2,p3) \ + ((cinfo)->err->msg_code = (code), \ + (cinfo)->err->msg_parm.i[0] = (p1), \ + (cinfo)->err->msg_parm.i[1] = (p2), \ + (cinfo)->err->msg_parm.i[2] = (p3), \ + (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo))) +#define ERREXIT4(cinfo,code,p1,p2,p3,p4) \ + ((cinfo)->err->msg_code = (code), \ + (cinfo)->err->msg_parm.i[0] = (p1), \ + (cinfo)->err->msg_parm.i[1] = (p2), \ + (cinfo)->err->msg_parm.i[2] = (p3), \ + (cinfo)->err->msg_parm.i[3] = (p4), \ + (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo))) +#define ERREXITS(cinfo,code,str) \ + ((cinfo)->err->msg_code = (code), \ + strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \ + (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo))) + +#define MAKESTMT(stuff) do { stuff } while (0) + +/* Nonfatal errors (we can keep going, but the data is probably corrupt) */ +#define WARNMS(cinfo,code) \ + ((cinfo)->err->msg_code = (code), \ + (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), -1)) +#define WARNMS1(cinfo,code,p1) \ + ((cinfo)->err->msg_code = (code), \ + (cinfo)->err->msg_parm.i[0] = (p1), \ + (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), -1)) +#define WARNMS2(cinfo,code,p1,p2) \ + ((cinfo)->err->msg_code = (code), \ + (cinfo)->err->msg_parm.i[0] = (p1), \ + (cinfo)->err->msg_parm.i[1] = (p2), \ + (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), -1)) + +/* Informational/debugging messages */ +#define TRACEMS(cinfo,lvl,code) \ + ((cinfo)->err->msg_code = (code), \ + (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl))) +#define TRACEMS1(cinfo,lvl,code,p1) \ + ((cinfo)->err->msg_code = (code), \ + (cinfo)->err->msg_parm.i[0] = (p1), \ + (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl))) +#define TRACEMS2(cinfo,lvl,code,p1,p2) \ + ((cinfo)->err->msg_code = (code), \ + (cinfo)->err->msg_parm.i[0] = (p1), \ + (cinfo)->err->msg_parm.i[1] = (p2), \ + (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl))) +#define TRACEMS3(cinfo,lvl,code,p1,p2,p3) \ + MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \ + _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); \ + (cinfo)->err->msg_code = (code); \ + (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); ) +#define TRACEMS4(cinfo,lvl,code,p1,p2,p3,p4) \ + MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \ + _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \ + (cinfo)->err->msg_code = (code); \ + (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); ) +#define TRACEMS5(cinfo,lvl,code,p1,p2,p3,p4,p5) \ + MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \ + _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \ + _mp[4] = (p5); \ + (cinfo)->err->msg_code = (code); \ + (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); ) +#define TRACEMS8(cinfo,lvl,code,p1,p2,p3,p4,p5,p6,p7,p8) \ + MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \ + _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \ + _mp[4] = (p5); _mp[5] = (p6); _mp[6] = (p7); _mp[7] = (p8); \ + (cinfo)->err->msg_code = (code); \ + (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); ) +#define TRACEMSS(cinfo,lvl,code,str) \ + ((cinfo)->err->msg_code = (code), \ + strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \ + (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl))) + +#endif /* JERROR_H */ diff --git a/app/src/main/jni/libjpeg-turbo/jfdctflt.c b/app/src/main/jni/libjpeg-turbo/jfdctflt.c new file mode 100644 index 000000000..a8367c68c --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jfdctflt.c @@ -0,0 +1,168 @@ +/* + * jfdctflt.c + * + * Copyright (C) 1994-1996, Thomas G. Lane. + * This file is part of the Independent JPEG Group's software. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains a floating-point implementation of the + * forward DCT (Discrete Cosine Transform). + * + * This implementation should be more accurate than either of the integer + * DCT implementations. However, it may not give the same results on all + * machines because of differences in roundoff behavior. Speed will depend + * on the hardware's floating point capacity. + * + * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT + * on each column. Direct algorithms are also available, but they are + * much more complex and seem not to be any faster when reduced to code. + * + * This implementation is based on Arai, Agui, and Nakajima's algorithm for + * scaled DCT. Their original paper (Trans. IEICE E-71(11):1095) is in + * Japanese, but the algorithm is described in the Pennebaker & Mitchell + * JPEG textbook (see REFERENCES section in file README). The following code + * is based directly on figure 4-8 in P&M. + * While an 8-point DCT cannot be done in less than 11 multiplies, it is + * possible to arrange the computation so that many of the multiplies are + * simple scalings of the final outputs. These multiplies can then be + * folded into the multiplications or divisions by the JPEG quantization + * table entries. The AA&N method leaves only 5 multiplies and 29 adds + * to be done in the DCT itself. + * The primary disadvantage of this method is that with a fixed-point + * implementation, accuracy is lost due to imprecise representation of the + * scaled quantization values. However, that problem does not arise if + * we use floating point arithmetic. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jdct.h" /* Private declarations for DCT subsystem */ + +#ifdef DCT_FLOAT_SUPPORTED + + +/* + * This module is specialized to the case DCTSIZE = 8. + */ + +#if DCTSIZE != 8 + Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */ +#endif + + +/* + * Perform the forward DCT on one block of samples. + */ + +GLOBAL(void) +jpeg_fdct_float (FAST_FLOAT * data) +{ + FAST_FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7; + FAST_FLOAT tmp10, tmp11, tmp12, tmp13; + FAST_FLOAT z1, z2, z3, z4, z5, z11, z13; + FAST_FLOAT *dataptr; + int ctr; + + /* Pass 1: process rows. */ + + dataptr = data; + for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { + tmp0 = dataptr[0] + dataptr[7]; + tmp7 = dataptr[0] - dataptr[7]; + tmp1 = dataptr[1] + dataptr[6]; + tmp6 = dataptr[1] - dataptr[6]; + tmp2 = dataptr[2] + dataptr[5]; + tmp5 = dataptr[2] - dataptr[5]; + tmp3 = dataptr[3] + dataptr[4]; + tmp4 = dataptr[3] - dataptr[4]; + + /* Even part */ + + tmp10 = tmp0 + tmp3; /* phase 2 */ + tmp13 = tmp0 - tmp3; + tmp11 = tmp1 + tmp2; + tmp12 = tmp1 - tmp2; + + dataptr[0] = tmp10 + tmp11; /* phase 3 */ + dataptr[4] = tmp10 - tmp11; + + z1 = (tmp12 + tmp13) * ((FAST_FLOAT) 0.707106781); /* c4 */ + dataptr[2] = tmp13 + z1; /* phase 5 */ + dataptr[6] = tmp13 - z1; + + /* Odd part */ + + tmp10 = tmp4 + tmp5; /* phase 2 */ + tmp11 = tmp5 + tmp6; + tmp12 = tmp6 + tmp7; + + /* The rotator is modified from fig 4-8 to avoid extra negations. */ + z5 = (tmp10 - tmp12) * ((FAST_FLOAT) 0.382683433); /* c6 */ + z2 = ((FAST_FLOAT) 0.541196100) * tmp10 + z5; /* c2-c6 */ + z4 = ((FAST_FLOAT) 1.306562965) * tmp12 + z5; /* c2+c6 */ + z3 = tmp11 * ((FAST_FLOAT) 0.707106781); /* c4 */ + + z11 = tmp7 + z3; /* phase 5 */ + z13 = tmp7 - z3; + + dataptr[5] = z13 + z2; /* phase 6 */ + dataptr[3] = z13 - z2; + dataptr[1] = z11 + z4; + dataptr[7] = z11 - z4; + + dataptr += DCTSIZE; /* advance pointer to next row */ + } + + /* Pass 2: process columns. */ + + dataptr = data; + for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { + tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7]; + tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7]; + tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6]; + tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6]; + tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5]; + tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5]; + tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4]; + tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4]; + + /* Even part */ + + tmp10 = tmp0 + tmp3; /* phase 2 */ + tmp13 = tmp0 - tmp3; + tmp11 = tmp1 + tmp2; + tmp12 = tmp1 - tmp2; + + dataptr[DCTSIZE*0] = tmp10 + tmp11; /* phase 3 */ + dataptr[DCTSIZE*4] = tmp10 - tmp11; + + z1 = (tmp12 + tmp13) * ((FAST_FLOAT) 0.707106781); /* c4 */ + dataptr[DCTSIZE*2] = tmp13 + z1; /* phase 5 */ + dataptr[DCTSIZE*6] = tmp13 - z1; + + /* Odd part */ + + tmp10 = tmp4 + tmp5; /* phase 2 */ + tmp11 = tmp5 + tmp6; + tmp12 = tmp6 + tmp7; + + /* The rotator is modified from fig 4-8 to avoid extra negations. */ + z5 = (tmp10 - tmp12) * ((FAST_FLOAT) 0.382683433); /* c6 */ + z2 = ((FAST_FLOAT) 0.541196100) * tmp10 + z5; /* c2-c6 */ + z4 = ((FAST_FLOAT) 1.306562965) * tmp12 + z5; /* c2+c6 */ + z3 = tmp11 * ((FAST_FLOAT) 0.707106781); /* c4 */ + + z11 = tmp7 + z3; /* phase 5 */ + z13 = tmp7 - z3; + + dataptr[DCTSIZE*5] = z13 + z2; /* phase 6 */ + dataptr[DCTSIZE*3] = z13 - z2; + dataptr[DCTSIZE*1] = z11 + z4; + dataptr[DCTSIZE*7] = z11 - z4; + + dataptr++; /* advance pointer to next column */ + } +} + +#endif /* DCT_FLOAT_SUPPORTED */ diff --git a/app/src/main/jni/libjpeg-turbo/jfdctfst.c b/app/src/main/jni/libjpeg-turbo/jfdctfst.c new file mode 100644 index 000000000..4936d4763 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jfdctfst.c @@ -0,0 +1,224 @@ +/* + * jfdctfst.c + * + * Copyright (C) 1994-1996, Thomas G. Lane. + * This file is part of the Independent JPEG Group's software. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains a fast, not so accurate integer implementation of the + * forward DCT (Discrete Cosine Transform). + * + * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT + * on each column. Direct algorithms are also available, but they are + * much more complex and seem not to be any faster when reduced to code. + * + * This implementation is based on Arai, Agui, and Nakajima's algorithm for + * scaled DCT. Their original paper (Trans. IEICE E-71(11):1095) is in + * Japanese, but the algorithm is described in the Pennebaker & Mitchell + * JPEG textbook (see REFERENCES section in file README). The following code + * is based directly on figure 4-8 in P&M. + * While an 8-point DCT cannot be done in less than 11 multiplies, it is + * possible to arrange the computation so that many of the multiplies are + * simple scalings of the final outputs. These multiplies can then be + * folded into the multiplications or divisions by the JPEG quantization + * table entries. The AA&N method leaves only 5 multiplies and 29 adds + * to be done in the DCT itself. + * The primary disadvantage of this method is that with fixed-point math, + * accuracy is lost due to imprecise representation of the scaled + * quantization values. The smaller the quantization table entry, the less + * precise the scaled value, so this implementation does worse with high- + * quality-setting files than with low-quality ones. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jdct.h" /* Private declarations for DCT subsystem */ + +#ifdef DCT_IFAST_SUPPORTED + + +/* + * This module is specialized to the case DCTSIZE = 8. + */ + +#if DCTSIZE != 8 + Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */ +#endif + + +/* Scaling decisions are generally the same as in the LL&M algorithm; + * see jfdctint.c for more details. However, we choose to descale + * (right shift) multiplication products as soon as they are formed, + * rather than carrying additional fractional bits into subsequent additions. + * This compromises accuracy slightly, but it lets us save a few shifts. + * More importantly, 16-bit arithmetic is then adequate (for 8-bit samples) + * everywhere except in the multiplications proper; this saves a good deal + * of work on 16-bit-int machines. + * + * Again to save a few shifts, the intermediate results between pass 1 and + * pass 2 are not upscaled, but are represented only to integral precision. + * + * A final compromise is to represent the multiplicative constants to only + * 8 fractional bits, rather than 13. This saves some shifting work on some + * machines, and may also reduce the cost of multiplication (since there + * are fewer one-bits in the constants). + */ + +#define CONST_BITS 8 + + +/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus + * causing a lot of useless floating-point operations at run time. + * To get around this we use the following pre-calculated constants. + * If you change CONST_BITS you may want to add appropriate values. + * (With a reasonable C compiler, you can just rely on the FIX() macro...) + */ + +#if CONST_BITS == 8 +#define FIX_0_382683433 ((INT32) 98) /* FIX(0.382683433) */ +#define FIX_0_541196100 ((INT32) 139) /* FIX(0.541196100) */ +#define FIX_0_707106781 ((INT32) 181) /* FIX(0.707106781) */ +#define FIX_1_306562965 ((INT32) 334) /* FIX(1.306562965) */ +#else +#define FIX_0_382683433 FIX(0.382683433) +#define FIX_0_541196100 FIX(0.541196100) +#define FIX_0_707106781 FIX(0.707106781) +#define FIX_1_306562965 FIX(1.306562965) +#endif + + +/* We can gain a little more speed, with a further compromise in accuracy, + * by omitting the addition in a descaling shift. This yields an incorrectly + * rounded result half the time... + */ + +#ifndef USE_ACCURATE_ROUNDING +#undef DESCALE +#define DESCALE(x,n) RIGHT_SHIFT(x, n) +#endif + + +/* Multiply a DCTELEM variable by an INT32 constant, and immediately + * descale to yield a DCTELEM result. + */ + +#define MULTIPLY(var,const) ((DCTELEM) DESCALE((var) * (const), CONST_BITS)) + + +/* + * Perform the forward DCT on one block of samples. + */ + +GLOBAL(void) +jpeg_fdct_ifast (DCTELEM * data) +{ + DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7; + DCTELEM tmp10, tmp11, tmp12, tmp13; + DCTELEM z1, z2, z3, z4, z5, z11, z13; + DCTELEM *dataptr; + int ctr; + SHIFT_TEMPS + + /* Pass 1: process rows. */ + + dataptr = data; + for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { + tmp0 = dataptr[0] + dataptr[7]; + tmp7 = dataptr[0] - dataptr[7]; + tmp1 = dataptr[1] + dataptr[6]; + tmp6 = dataptr[1] - dataptr[6]; + tmp2 = dataptr[2] + dataptr[5]; + tmp5 = dataptr[2] - dataptr[5]; + tmp3 = dataptr[3] + dataptr[4]; + tmp4 = dataptr[3] - dataptr[4]; + + /* Even part */ + + tmp10 = tmp0 + tmp3; /* phase 2 */ + tmp13 = tmp0 - tmp3; + tmp11 = tmp1 + tmp2; + tmp12 = tmp1 - tmp2; + + dataptr[0] = tmp10 + tmp11; /* phase 3 */ + dataptr[4] = tmp10 - tmp11; + + z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */ + dataptr[2] = tmp13 + z1; /* phase 5 */ + dataptr[6] = tmp13 - z1; + + /* Odd part */ + + tmp10 = tmp4 + tmp5; /* phase 2 */ + tmp11 = tmp5 + tmp6; + tmp12 = tmp6 + tmp7; + + /* The rotator is modified from fig 4-8 to avoid extra negations. */ + z5 = MULTIPLY(tmp10 - tmp12, FIX_0_382683433); /* c6 */ + z2 = MULTIPLY(tmp10, FIX_0_541196100) + z5; /* c2-c6 */ + z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */ + z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */ + + z11 = tmp7 + z3; /* phase 5 */ + z13 = tmp7 - z3; + + dataptr[5] = z13 + z2; /* phase 6 */ + dataptr[3] = z13 - z2; + dataptr[1] = z11 + z4; + dataptr[7] = z11 - z4; + + dataptr += DCTSIZE; /* advance pointer to next row */ + } + + /* Pass 2: process columns. */ + + dataptr = data; + for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { + tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7]; + tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7]; + tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6]; + tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6]; + tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5]; + tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5]; + tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4]; + tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4]; + + /* Even part */ + + tmp10 = tmp0 + tmp3; /* phase 2 */ + tmp13 = tmp0 - tmp3; + tmp11 = tmp1 + tmp2; + tmp12 = tmp1 - tmp2; + + dataptr[DCTSIZE*0] = tmp10 + tmp11; /* phase 3 */ + dataptr[DCTSIZE*4] = tmp10 - tmp11; + + z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */ + dataptr[DCTSIZE*2] = tmp13 + z1; /* phase 5 */ + dataptr[DCTSIZE*6] = tmp13 - z1; + + /* Odd part */ + + tmp10 = tmp4 + tmp5; /* phase 2 */ + tmp11 = tmp5 + tmp6; + tmp12 = tmp6 + tmp7; + + /* The rotator is modified from fig 4-8 to avoid extra negations. */ + z5 = MULTIPLY(tmp10 - tmp12, FIX_0_382683433); /* c6 */ + z2 = MULTIPLY(tmp10, FIX_0_541196100) + z5; /* c2-c6 */ + z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */ + z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */ + + z11 = tmp7 + z3; /* phase 5 */ + z13 = tmp7 - z3; + + dataptr[DCTSIZE*5] = z13 + z2; /* phase 6 */ + dataptr[DCTSIZE*3] = z13 - z2; + dataptr[DCTSIZE*1] = z11 + z4; + dataptr[DCTSIZE*7] = z11 - z4; + + dataptr++; /* advance pointer to next column */ + } +} + +#endif /* DCT_IFAST_SUPPORTED */ diff --git a/app/src/main/jni/libjpeg-turbo/jfdctint.c b/app/src/main/jni/libjpeg-turbo/jfdctint.c new file mode 100644 index 000000000..14f486cdc --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jfdctint.c @@ -0,0 +1,283 @@ +/* + * jfdctint.c + * + * Copyright (C) 1991-1996, Thomas G. Lane. + * This file is part of the Independent JPEG Group's software. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains a slow-but-accurate integer implementation of the + * forward DCT (Discrete Cosine Transform). + * + * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT + * on each column. Direct algorithms are also available, but they are + * much more complex and seem not to be any faster when reduced to code. + * + * This implementation is based on an algorithm described in + * C. Loeffler, A. Ligtenberg and G. Moschytz, "Practical Fast 1-D DCT + * Algorithms with 11 Multiplications", Proc. Int'l. Conf. on Acoustics, + * Speech, and Signal Processing 1989 (ICASSP '89), pp. 988-991. + * The primary algorithm described there uses 11 multiplies and 29 adds. + * We use their alternate method with 12 multiplies and 32 adds. + * The advantage of this method is that no data path contains more than one + * multiplication; this allows a very simple and accurate implementation in + * scaled fixed-point arithmetic, with a minimal number of shifts. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jdct.h" /* Private declarations for DCT subsystem */ + +#ifdef DCT_ISLOW_SUPPORTED + + +/* + * This module is specialized to the case DCTSIZE = 8. + */ + +#if DCTSIZE != 8 + Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */ +#endif + + +/* + * The poop on this scaling stuff is as follows: + * + * Each 1-D DCT step produces outputs which are a factor of sqrt(N) + * larger than the true DCT outputs. The final outputs are therefore + * a factor of N larger than desired; since N=8 this can be cured by + * a simple right shift at the end of the algorithm. The advantage of + * this arrangement is that we save two multiplications per 1-D DCT, + * because the y0 and y4 outputs need not be divided by sqrt(N). + * In the IJG code, this factor of 8 is removed by the quantization step + * (in jcdctmgr.c), NOT in this module. + * + * We have to do addition and subtraction of the integer inputs, which + * is no problem, and multiplication by fractional constants, which is + * a problem to do in integer arithmetic. We multiply all the constants + * by CONST_SCALE and convert them to integer constants (thus retaining + * CONST_BITS bits of precision in the constants). After doing a + * multiplication we have to divide the product by CONST_SCALE, with proper + * rounding, to produce the correct output. This division can be done + * cheaply as a right shift of CONST_BITS bits. We postpone shifting + * as long as possible so that partial sums can be added together with + * full fractional precision. + * + * The outputs of the first pass are scaled up by PASS1_BITS bits so that + * they are represented to better-than-integral precision. These outputs + * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word + * with the recommended scaling. (For 12-bit sample data, the intermediate + * array is INT32 anyway.) + * + * To avoid overflow of the 32-bit intermediate results in pass 2, we must + * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26. Error analysis + * shows that the values given below are the most effective. + */ + +#if BITS_IN_JSAMPLE == 8 +#define CONST_BITS 13 +#define PASS1_BITS 2 +#else +#define CONST_BITS 13 +#define PASS1_BITS 1 /* lose a little precision to avoid overflow */ +#endif + +/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus + * causing a lot of useless floating-point operations at run time. + * To get around this we use the following pre-calculated constants. + * If you change CONST_BITS you may want to add appropriate values. + * (With a reasonable C compiler, you can just rely on the FIX() macro...) + */ + +#if CONST_BITS == 13 +#define FIX_0_298631336 ((INT32) 2446) /* FIX(0.298631336) */ +#define FIX_0_390180644 ((INT32) 3196) /* FIX(0.390180644) */ +#define FIX_0_541196100 ((INT32) 4433) /* FIX(0.541196100) */ +#define FIX_0_765366865 ((INT32) 6270) /* FIX(0.765366865) */ +#define FIX_0_899976223 ((INT32) 7373) /* FIX(0.899976223) */ +#define FIX_1_175875602 ((INT32) 9633) /* FIX(1.175875602) */ +#define FIX_1_501321110 ((INT32) 12299) /* FIX(1.501321110) */ +#define FIX_1_847759065 ((INT32) 15137) /* FIX(1.847759065) */ +#define FIX_1_961570560 ((INT32) 16069) /* FIX(1.961570560) */ +#define FIX_2_053119869 ((INT32) 16819) /* FIX(2.053119869) */ +#define FIX_2_562915447 ((INT32) 20995) /* FIX(2.562915447) */ +#define FIX_3_072711026 ((INT32) 25172) /* FIX(3.072711026) */ +#else +#define FIX_0_298631336 FIX(0.298631336) +#define FIX_0_390180644 FIX(0.390180644) +#define FIX_0_541196100 FIX(0.541196100) +#define FIX_0_765366865 FIX(0.765366865) +#define FIX_0_899976223 FIX(0.899976223) +#define FIX_1_175875602 FIX(1.175875602) +#define FIX_1_501321110 FIX(1.501321110) +#define FIX_1_847759065 FIX(1.847759065) +#define FIX_1_961570560 FIX(1.961570560) +#define FIX_2_053119869 FIX(2.053119869) +#define FIX_2_562915447 FIX(2.562915447) +#define FIX_3_072711026 FIX(3.072711026) +#endif + + +/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result. + * For 8-bit samples with the recommended scaling, all the variable + * and constant values involved are no more than 16 bits wide, so a + * 16x16->32 bit multiply can be used instead of a full 32x32 multiply. + * For 12-bit samples, a full 32-bit multiplication will be needed. + */ + +#if BITS_IN_JSAMPLE == 8 +#define MULTIPLY(var,const) MULTIPLY16C16(var,const) +#else +#define MULTIPLY(var,const) ((var) * (const)) +#endif + + +/* + * Perform the forward DCT on one block of samples. + */ + +GLOBAL(void) +jpeg_fdct_islow (DCTELEM * data) +{ + INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7; + INT32 tmp10, tmp11, tmp12, tmp13; + INT32 z1, z2, z3, z4, z5; + DCTELEM *dataptr; + int ctr; + SHIFT_TEMPS + + /* Pass 1: process rows. */ + /* Note results are scaled up by sqrt(8) compared to a true DCT; */ + /* furthermore, we scale the results by 2**PASS1_BITS. */ + + dataptr = data; + for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { + tmp0 = dataptr[0] + dataptr[7]; + tmp7 = dataptr[0] - dataptr[7]; + tmp1 = dataptr[1] + dataptr[6]; + tmp6 = dataptr[1] - dataptr[6]; + tmp2 = dataptr[2] + dataptr[5]; + tmp5 = dataptr[2] - dataptr[5]; + tmp3 = dataptr[3] + dataptr[4]; + tmp4 = dataptr[3] - dataptr[4]; + + /* Even part per LL&M figure 1 --- note that published figure is faulty; + * rotator "sqrt(2)*c1" should be "sqrt(2)*c6". + */ + + tmp10 = tmp0 + tmp3; + tmp13 = tmp0 - tmp3; + tmp11 = tmp1 + tmp2; + tmp12 = tmp1 - tmp2; + + dataptr[0] = (DCTELEM) ((tmp10 + tmp11) << PASS1_BITS); + dataptr[4] = (DCTELEM) ((tmp10 - tmp11) << PASS1_BITS); + + z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100); + dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865), + CONST_BITS-PASS1_BITS); + dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065), + CONST_BITS-PASS1_BITS); + + /* Odd part per figure 8 --- note paper omits factor of sqrt(2). + * cK represents cos(K*pi/16). + * i0..i3 in the paper are tmp4..tmp7 here. + */ + + z1 = tmp4 + tmp7; + z2 = tmp5 + tmp6; + z3 = tmp4 + tmp6; + z4 = tmp5 + tmp7; + z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */ + + tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */ + tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */ + tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */ + tmp7 = MULTIPLY(tmp7, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */ + z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */ + z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */ + z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */ + z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */ + + z3 += z5; + z4 += z5; + + dataptr[7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS); + dataptr[5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS); + dataptr[3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS); + dataptr[1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS); + + dataptr += DCTSIZE; /* advance pointer to next row */ + } + + /* Pass 2: process columns. + * We remove the PASS1_BITS scaling, but leave the results scaled up + * by an overall factor of 8. + */ + + dataptr = data; + for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { + tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7]; + tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7]; + tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6]; + tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6]; + tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5]; + tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5]; + tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4]; + tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4]; + + /* Even part per LL&M figure 1 --- note that published figure is faulty; + * rotator "sqrt(2)*c1" should be "sqrt(2)*c6". + */ + + tmp10 = tmp0 + tmp3; + tmp13 = tmp0 - tmp3; + tmp11 = tmp1 + tmp2; + tmp12 = tmp1 - tmp2; + + dataptr[DCTSIZE*0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS); + dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS); + + z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100); + dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865), + CONST_BITS+PASS1_BITS); + dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065), + CONST_BITS+PASS1_BITS); + + /* Odd part per figure 8 --- note paper omits factor of sqrt(2). + * cK represents cos(K*pi/16). + * i0..i3 in the paper are tmp4..tmp7 here. + */ + + z1 = tmp4 + tmp7; + z2 = tmp5 + tmp6; + z3 = tmp4 + tmp6; + z4 = tmp5 + tmp7; + z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */ + + tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */ + tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */ + tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */ + tmp7 = MULTIPLY(tmp7, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */ + z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */ + z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */ + z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */ + z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */ + + z3 += z5; + z4 += z5; + + dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, + CONST_BITS+PASS1_BITS); + dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, + CONST_BITS+PASS1_BITS); + dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, + CONST_BITS+PASS1_BITS); + dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, + CONST_BITS+PASS1_BITS); + + dataptr++; /* advance pointer to next column */ + } +} + +#endif /* DCT_ISLOW_SUPPORTED */ diff --git a/app/src/main/jni/libjpeg-turbo/jidctflt.c b/app/src/main/jni/libjpeg-turbo/jidctflt.c new file mode 100644 index 000000000..2b2e2282a --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jidctflt.c @@ -0,0 +1,239 @@ +/* + * jidctflt.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1994-1998, Thomas G. Lane. + * Modified 2010 by Guido Vollbeding. + * libjpeg-turbo Modifications: + * Copyright (C) 2014, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains a floating-point implementation of the + * inverse DCT (Discrete Cosine Transform). In the IJG code, this routine + * must also perform dequantization of the input coefficients. + * + * This implementation should be more accurate than either of the integer + * IDCT implementations. However, it may not give the same results on all + * machines because of differences in roundoff behavior. Speed will depend + * on the hardware's floating point capacity. + * + * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT + * on each row (or vice versa, but it's more convenient to emit a row at + * a time). Direct algorithms are also available, but they are much more + * complex and seem not to be any faster when reduced to code. + * + * This implementation is based on Arai, Agui, and Nakajima's algorithm for + * scaled DCT. Their original paper (Trans. IEICE E-71(11):1095) is in + * Japanese, but the algorithm is described in the Pennebaker & Mitchell + * JPEG textbook (see REFERENCES section in file README). The following code + * is based directly on figure 4-8 in P&M. + * While an 8-point DCT cannot be done in less than 11 multiplies, it is + * possible to arrange the computation so that many of the multiplies are + * simple scalings of the final outputs. These multiplies can then be + * folded into the multiplications or divisions by the JPEG quantization + * table entries. The AA&N method leaves only 5 multiplies and 29 adds + * to be done in the DCT itself. + * The primary disadvantage of this method is that with a fixed-point + * implementation, accuracy is lost due to imprecise representation of the + * scaled quantization values. However, that problem does not arise if + * we use floating point arithmetic. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jdct.h" /* Private declarations for DCT subsystem */ + +#ifdef DCT_FLOAT_SUPPORTED + + +/* + * This module is specialized to the case DCTSIZE = 8. + */ + +#if DCTSIZE != 8 + Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */ +#endif + + +/* Dequantize a coefficient by multiplying it by the multiplier-table + * entry; produce a float result. + */ + +#define DEQUANTIZE(coef,quantval) (((FAST_FLOAT) (coef)) * (quantval)) + + +/* + * Perform dequantization and inverse DCT on one block of coefficients. + */ + +GLOBAL(void) +jpeg_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + FAST_FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7; + FAST_FLOAT tmp10, tmp11, tmp12, tmp13; + FAST_FLOAT z5, z10, z11, z12, z13; + JCOEFPTR inptr; + FLOAT_MULT_TYPE * quantptr; + FAST_FLOAT * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = cinfo->sample_range_limit; + int ctr; + FAST_FLOAT workspace[DCTSIZE2]; /* buffers data between passes */ + #define _0_125 ((FLOAT_MULT_TYPE)0.125) + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (FLOAT_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = DCTSIZE; ctr > 0; ctr--) { + /* Due to quantization, we will usually find that many of the input + * coefficients are zero, especially the AC terms. We can exploit this + * by short-circuiting the IDCT calculation for any column in which all + * the AC terms are zero. In that case each output is equal to the + * DC coefficient (with scale factor as needed). + * With typical images and quantization tables, half or more of the + * column DCT calculations can be simplified this way. + */ + + if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 && + inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 && + inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 && + inptr[DCTSIZE*7] == 0) { + /* AC terms all zero */ + FAST_FLOAT dcval = DEQUANTIZE(inptr[DCTSIZE*0], + quantptr[DCTSIZE*0] * _0_125); + + wsptr[DCTSIZE*0] = dcval; + wsptr[DCTSIZE*1] = dcval; + wsptr[DCTSIZE*2] = dcval; + wsptr[DCTSIZE*3] = dcval; + wsptr[DCTSIZE*4] = dcval; + wsptr[DCTSIZE*5] = dcval; + wsptr[DCTSIZE*6] = dcval; + wsptr[DCTSIZE*7] = dcval; + + inptr++; /* advance pointers to next column */ + quantptr++; + wsptr++; + continue; + } + + /* Even part */ + + tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0] * _0_125); + tmp1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2] * _0_125); + tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4] * _0_125); + tmp3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6] * _0_125); + + tmp10 = tmp0 + tmp2; /* phase 3 */ + tmp11 = tmp0 - tmp2; + + tmp13 = tmp1 + tmp3; /* phases 5-3 */ + tmp12 = (tmp1 - tmp3) * ((FAST_FLOAT) 1.414213562) - tmp13; /* 2*c4 */ + + tmp0 = tmp10 + tmp13; /* phase 2 */ + tmp3 = tmp10 - tmp13; + tmp1 = tmp11 + tmp12; + tmp2 = tmp11 - tmp12; + + /* Odd part */ + + tmp4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1] * _0_125); + tmp5 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3] * _0_125); + tmp6 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5] * _0_125); + tmp7 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7] * _0_125); + + z13 = tmp6 + tmp5; /* phase 6 */ + z10 = tmp6 - tmp5; + z11 = tmp4 + tmp7; + z12 = tmp4 - tmp7; + + tmp7 = z11 + z13; /* phase 5 */ + tmp11 = (z11 - z13) * ((FAST_FLOAT) 1.414213562); /* 2*c4 */ + + z5 = (z10 + z12) * ((FAST_FLOAT) 1.847759065); /* 2*c2 */ + tmp10 = z5 - z12 * ((FAST_FLOAT) 1.082392200); /* 2*(c2-c6) */ + tmp12 = z5 - z10 * ((FAST_FLOAT) 2.613125930); /* 2*(c2+c6) */ + + tmp6 = tmp12 - tmp7; /* phase 2 */ + tmp5 = tmp11 - tmp6; + tmp4 = tmp10 - tmp5; + + wsptr[DCTSIZE*0] = tmp0 + tmp7; + wsptr[DCTSIZE*7] = tmp0 - tmp7; + wsptr[DCTSIZE*1] = tmp1 + tmp6; + wsptr[DCTSIZE*6] = tmp1 - tmp6; + wsptr[DCTSIZE*2] = tmp2 + tmp5; + wsptr[DCTSIZE*5] = tmp2 - tmp5; + wsptr[DCTSIZE*3] = tmp3 + tmp4; + wsptr[DCTSIZE*4] = tmp3 - tmp4; + + inptr++; /* advance pointers to next column */ + quantptr++; + wsptr++; + } + + /* Pass 2: process rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < DCTSIZE; ctr++) { + outptr = output_buf[ctr] + output_col; + /* Rows of zeroes can be exploited in the same way as we did with columns. + * However, the column calculation has created many nonzero AC terms, so + * the simplification applies less often (typically 5% to 10% of the time). + * And testing floats for zero is relatively expensive, so we don't bother. + */ + + /* Even part */ + + /* Apply signed->unsigned and prepare float->int conversion */ + z5 = wsptr[0] + ((FAST_FLOAT) CENTERJSAMPLE + (FAST_FLOAT) 0.5); + tmp10 = z5 + wsptr[4]; + tmp11 = z5 - wsptr[4]; + + tmp13 = wsptr[2] + wsptr[6]; + tmp12 = (wsptr[2] - wsptr[6]) * ((FAST_FLOAT) 1.414213562) - tmp13; + + tmp0 = tmp10 + tmp13; + tmp3 = tmp10 - tmp13; + tmp1 = tmp11 + tmp12; + tmp2 = tmp11 - tmp12; + + /* Odd part */ + + z13 = wsptr[5] + wsptr[3]; + z10 = wsptr[5] - wsptr[3]; + z11 = wsptr[1] + wsptr[7]; + z12 = wsptr[1] - wsptr[7]; + + tmp7 = z11 + z13; + tmp11 = (z11 - z13) * ((FAST_FLOAT) 1.414213562); + + z5 = (z10 + z12) * ((FAST_FLOAT) 1.847759065); /* 2*c2 */ + tmp10 = z5 - z12 * ((FAST_FLOAT) 1.082392200); /* 2*(c2-c6) */ + tmp12 = z5 - z10 * ((FAST_FLOAT) 2.613125930); /* 2*(c2+c6) */ + + tmp6 = tmp12 - tmp7; + tmp5 = tmp11 - tmp6; + tmp4 = tmp10 - tmp5; + + /* Final output stage: float->int conversion and range-limit */ + + outptr[0] = range_limit[((int) (tmp0 + tmp7)) & RANGE_MASK]; + outptr[7] = range_limit[((int) (tmp0 - tmp7)) & RANGE_MASK]; + outptr[1] = range_limit[((int) (tmp1 + tmp6)) & RANGE_MASK]; + outptr[6] = range_limit[((int) (tmp1 - tmp6)) & RANGE_MASK]; + outptr[2] = range_limit[((int) (tmp2 + tmp5)) & RANGE_MASK]; + outptr[5] = range_limit[((int) (tmp2 - tmp5)) & RANGE_MASK]; + outptr[3] = range_limit[((int) (tmp3 + tmp4)) & RANGE_MASK]; + outptr[4] = range_limit[((int) (tmp3 - tmp4)) & RANGE_MASK]; + + wsptr += DCTSIZE; /* advance pointer to next row */ + } +} + +#endif /* DCT_FLOAT_SUPPORTED */ diff --git a/app/src/main/jni/libjpeg-turbo/jidctfst.c b/app/src/main/jni/libjpeg-turbo/jidctfst.c new file mode 100644 index 000000000..cae22b9a8 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jidctfst.c @@ -0,0 +1,368 @@ +/* + * jidctfst.c + * + * Copyright (C) 1994-1998, Thomas G. Lane. + * This file is part of the Independent JPEG Group's software. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains a fast, not so accurate integer implementation of the + * inverse DCT (Discrete Cosine Transform). In the IJG code, this routine + * must also perform dequantization of the input coefficients. + * + * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT + * on each row (or vice versa, but it's more convenient to emit a row at + * a time). Direct algorithms are also available, but they are much more + * complex and seem not to be any faster when reduced to code. + * + * This implementation is based on Arai, Agui, and Nakajima's algorithm for + * scaled DCT. Their original paper (Trans. IEICE E-71(11):1095) is in + * Japanese, but the algorithm is described in the Pennebaker & Mitchell + * JPEG textbook (see REFERENCES section in file README). The following code + * is based directly on figure 4-8 in P&M. + * While an 8-point DCT cannot be done in less than 11 multiplies, it is + * possible to arrange the computation so that many of the multiplies are + * simple scalings of the final outputs. These multiplies can then be + * folded into the multiplications or divisions by the JPEG quantization + * table entries. The AA&N method leaves only 5 multiplies and 29 adds + * to be done in the DCT itself. + * The primary disadvantage of this method is that with fixed-point math, + * accuracy is lost due to imprecise representation of the scaled + * quantization values. The smaller the quantization table entry, the less + * precise the scaled value, so this implementation does worse with high- + * quality-setting files than with low-quality ones. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jdct.h" /* Private declarations for DCT subsystem */ + +#ifdef DCT_IFAST_SUPPORTED + + +/* + * This module is specialized to the case DCTSIZE = 8. + */ + +#if DCTSIZE != 8 + Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */ +#endif + + +/* Scaling decisions are generally the same as in the LL&M algorithm; + * see jidctint.c for more details. However, we choose to descale + * (right shift) multiplication products as soon as they are formed, + * rather than carrying additional fractional bits into subsequent additions. + * This compromises accuracy slightly, but it lets us save a few shifts. + * More importantly, 16-bit arithmetic is then adequate (for 8-bit samples) + * everywhere except in the multiplications proper; this saves a good deal + * of work on 16-bit-int machines. + * + * The dequantized coefficients are not integers because the AA&N scaling + * factors have been incorporated. We represent them scaled up by PASS1_BITS, + * so that the first and second IDCT rounds have the same input scaling. + * For 8-bit JSAMPLEs, we choose IFAST_SCALE_BITS = PASS1_BITS so as to + * avoid a descaling shift; this compromises accuracy rather drastically + * for small quantization table entries, but it saves a lot of shifts. + * For 12-bit JSAMPLEs, there's no hope of using 16x16 multiplies anyway, + * so we use a much larger scaling factor to preserve accuracy. + * + * A final compromise is to represent the multiplicative constants to only + * 8 fractional bits, rather than 13. This saves some shifting work on some + * machines, and may also reduce the cost of multiplication (since there + * are fewer one-bits in the constants). + */ + +#if BITS_IN_JSAMPLE == 8 +#define CONST_BITS 8 +#define PASS1_BITS 2 +#else +#define CONST_BITS 8 +#define PASS1_BITS 1 /* lose a little precision to avoid overflow */ +#endif + +/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus + * causing a lot of useless floating-point operations at run time. + * To get around this we use the following pre-calculated constants. + * If you change CONST_BITS you may want to add appropriate values. + * (With a reasonable C compiler, you can just rely on the FIX() macro...) + */ + +#if CONST_BITS == 8 +#define FIX_1_082392200 ((INT32) 277) /* FIX(1.082392200) */ +#define FIX_1_414213562 ((INT32) 362) /* FIX(1.414213562) */ +#define FIX_1_847759065 ((INT32) 473) /* FIX(1.847759065) */ +#define FIX_2_613125930 ((INT32) 669) /* FIX(2.613125930) */ +#else +#define FIX_1_082392200 FIX(1.082392200) +#define FIX_1_414213562 FIX(1.414213562) +#define FIX_1_847759065 FIX(1.847759065) +#define FIX_2_613125930 FIX(2.613125930) +#endif + + +/* We can gain a little more speed, with a further compromise in accuracy, + * by omitting the addition in a descaling shift. This yields an incorrectly + * rounded result half the time... + */ + +#ifndef USE_ACCURATE_ROUNDING +#undef DESCALE +#define DESCALE(x,n) RIGHT_SHIFT(x, n) +#endif + + +/* Multiply a DCTELEM variable by an INT32 constant, and immediately + * descale to yield a DCTELEM result. + */ + +#define MULTIPLY(var,const) ((DCTELEM) DESCALE((var) * (const), CONST_BITS)) + + +/* Dequantize a coefficient by multiplying it by the multiplier-table + * entry; produce a DCTELEM result. For 8-bit data a 16x16->16 + * multiplication will do. For 12-bit data, the multiplier table is + * declared INT32, so a 32-bit multiply will be used. + */ + +#if BITS_IN_JSAMPLE == 8 +#define DEQUANTIZE(coef,quantval) (((IFAST_MULT_TYPE) (coef)) * (quantval)) +#else +#define DEQUANTIZE(coef,quantval) \ + DESCALE((coef)*(quantval), IFAST_SCALE_BITS-PASS1_BITS) +#endif + + +/* Like DESCALE, but applies to a DCTELEM and produces an int. + * We assume that int right shift is unsigned if INT32 right shift is. + */ + +#ifdef RIGHT_SHIFT_IS_UNSIGNED +#define ISHIFT_TEMPS DCTELEM ishift_temp; +#if BITS_IN_JSAMPLE == 8 +#define DCTELEMBITS 16 /* DCTELEM may be 16 or 32 bits */ +#else +#define DCTELEMBITS 32 /* DCTELEM must be 32 bits */ +#endif +#define IRIGHT_SHIFT(x,shft) \ + ((ishift_temp = (x)) < 0 ? \ + (ishift_temp >> (shft)) | ((~((DCTELEM) 0)) << (DCTELEMBITS-(shft))) : \ + (ishift_temp >> (shft))) +#else +#define ISHIFT_TEMPS +#define IRIGHT_SHIFT(x,shft) ((x) >> (shft)) +#endif + +#ifdef USE_ACCURATE_ROUNDING +#define IDESCALE(x,n) ((int) IRIGHT_SHIFT((x) + (1 << ((n)-1)), n)) +#else +#define IDESCALE(x,n) ((int) IRIGHT_SHIFT(x, n)) +#endif + + +/* + * Perform dequantization and inverse DCT on one block of coefficients. + */ + +GLOBAL(void) +jpeg_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7; + DCTELEM tmp10, tmp11, tmp12, tmp13; + DCTELEM z5, z10, z11, z12, z13; + JCOEFPTR inptr; + IFAST_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[DCTSIZE2]; /* buffers data between passes */ + SHIFT_TEMPS /* for DESCALE */ + ISHIFT_TEMPS /* for IDESCALE */ + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (IFAST_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = DCTSIZE; ctr > 0; ctr--) { + /* Due to quantization, we will usually find that many of the input + * coefficients are zero, especially the AC terms. We can exploit this + * by short-circuiting the IDCT calculation for any column in which all + * the AC terms are zero. In that case each output is equal to the + * DC coefficient (with scale factor as needed). + * With typical images and quantization tables, half or more of the + * column DCT calculations can be simplified this way. + */ + + if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 && + inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 && + inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 && + inptr[DCTSIZE*7] == 0) { + /* AC terms all zero */ + int dcval = (int) DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + + wsptr[DCTSIZE*0] = dcval; + wsptr[DCTSIZE*1] = dcval; + wsptr[DCTSIZE*2] = dcval; + wsptr[DCTSIZE*3] = dcval; + wsptr[DCTSIZE*4] = dcval; + wsptr[DCTSIZE*5] = dcval; + wsptr[DCTSIZE*6] = dcval; + wsptr[DCTSIZE*7] = dcval; + + inptr++; /* advance pointers to next column */ + quantptr++; + wsptr++; + continue; + } + + /* Even part */ + + tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + tmp1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]); + tmp3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]); + + tmp10 = tmp0 + tmp2; /* phase 3 */ + tmp11 = tmp0 - tmp2; + + tmp13 = tmp1 + tmp3; /* phases 5-3 */ + tmp12 = MULTIPLY(tmp1 - tmp3, FIX_1_414213562) - tmp13; /* 2*c4 */ + + tmp0 = tmp10 + tmp13; /* phase 2 */ + tmp3 = tmp10 - tmp13; + tmp1 = tmp11 + tmp12; + tmp2 = tmp11 - tmp12; + + /* Odd part */ + + tmp4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + tmp5 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + tmp6 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + tmp7 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]); + + z13 = tmp6 + tmp5; /* phase 6 */ + z10 = tmp6 - tmp5; + z11 = tmp4 + tmp7; + z12 = tmp4 - tmp7; + + tmp7 = z11 + z13; /* phase 5 */ + tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562); /* 2*c4 */ + + z5 = MULTIPLY(z10 + z12, FIX_1_847759065); /* 2*c2 */ + tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5; /* 2*(c2-c6) */ + tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5; /* -2*(c2+c6) */ + + tmp6 = tmp12 - tmp7; /* phase 2 */ + tmp5 = tmp11 - tmp6; + tmp4 = tmp10 + tmp5; + + wsptr[DCTSIZE*0] = (int) (tmp0 + tmp7); + wsptr[DCTSIZE*7] = (int) (tmp0 - tmp7); + wsptr[DCTSIZE*1] = (int) (tmp1 + tmp6); + wsptr[DCTSIZE*6] = (int) (tmp1 - tmp6); + wsptr[DCTSIZE*2] = (int) (tmp2 + tmp5); + wsptr[DCTSIZE*5] = (int) (tmp2 - tmp5); + wsptr[DCTSIZE*4] = (int) (tmp3 + tmp4); + wsptr[DCTSIZE*3] = (int) (tmp3 - tmp4); + + inptr++; /* advance pointers to next column */ + quantptr++; + wsptr++; + } + + /* Pass 2: process rows from work array, store into output array. */ + /* Note that we must descale the results by a factor of 8 == 2**3, */ + /* and also undo the PASS1_BITS scaling. */ + + wsptr = workspace; + for (ctr = 0; ctr < DCTSIZE; ctr++) { + outptr = output_buf[ctr] + output_col; + /* Rows of zeroes can be exploited in the same way as we did with columns. + * However, the column calculation has created many nonzero AC terms, so + * the simplification applies less often (typically 5% to 10% of the time). + * On machines with very fast multiplication, it's possible that the + * test takes more time than it's worth. In that case this section + * may be commented out. + */ + +#ifndef NO_ZERO_ROW_TEST + if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 && wsptr[4] == 0 && + wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) { + /* AC terms all zero */ + JSAMPLE dcval = range_limit[IDESCALE(wsptr[0], PASS1_BITS+3) + & RANGE_MASK]; + + outptr[0] = dcval; + outptr[1] = dcval; + outptr[2] = dcval; + outptr[3] = dcval; + outptr[4] = dcval; + outptr[5] = dcval; + outptr[6] = dcval; + outptr[7] = dcval; + + wsptr += DCTSIZE; /* advance pointer to next row */ + continue; + } +#endif + + /* Even part */ + + tmp10 = ((DCTELEM) wsptr[0] + (DCTELEM) wsptr[4]); + tmp11 = ((DCTELEM) wsptr[0] - (DCTELEM) wsptr[4]); + + tmp13 = ((DCTELEM) wsptr[2] + (DCTELEM) wsptr[6]); + tmp12 = MULTIPLY((DCTELEM) wsptr[2] - (DCTELEM) wsptr[6], FIX_1_414213562) + - tmp13; + + tmp0 = tmp10 + tmp13; + tmp3 = tmp10 - tmp13; + tmp1 = tmp11 + tmp12; + tmp2 = tmp11 - tmp12; + + /* Odd part */ + + z13 = (DCTELEM) wsptr[5] + (DCTELEM) wsptr[3]; + z10 = (DCTELEM) wsptr[5] - (DCTELEM) wsptr[3]; + z11 = (DCTELEM) wsptr[1] + (DCTELEM) wsptr[7]; + z12 = (DCTELEM) wsptr[1] - (DCTELEM) wsptr[7]; + + tmp7 = z11 + z13; /* phase 5 */ + tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562); /* 2*c4 */ + + z5 = MULTIPLY(z10 + z12, FIX_1_847759065); /* 2*c2 */ + tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5; /* 2*(c2-c6) */ + tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5; /* -2*(c2+c6) */ + + tmp6 = tmp12 - tmp7; /* phase 2 */ + tmp5 = tmp11 - tmp6; + tmp4 = tmp10 + tmp5; + + /* Final output stage: scale down by a factor of 8 and range-limit */ + + outptr[0] = range_limit[IDESCALE(tmp0 + tmp7, PASS1_BITS+3) + & RANGE_MASK]; + outptr[7] = range_limit[IDESCALE(tmp0 - tmp7, PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[IDESCALE(tmp1 + tmp6, PASS1_BITS+3) + & RANGE_MASK]; + outptr[6] = range_limit[IDESCALE(tmp1 - tmp6, PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[IDESCALE(tmp2 + tmp5, PASS1_BITS+3) + & RANGE_MASK]; + outptr[5] = range_limit[IDESCALE(tmp2 - tmp5, PASS1_BITS+3) + & RANGE_MASK]; + outptr[4] = range_limit[IDESCALE(tmp3 + tmp4, PASS1_BITS+3) + & RANGE_MASK]; + outptr[3] = range_limit[IDESCALE(tmp3 - tmp4, PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += DCTSIZE; /* advance pointer to next row */ + } +} + +#endif /* DCT_IFAST_SUPPORTED */ diff --git a/app/src/main/jni/libjpeg-turbo/jidctint.c b/app/src/main/jni/libjpeg-turbo/jidctint.c new file mode 100644 index 000000000..688fd2243 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jidctint.c @@ -0,0 +1,2623 @@ +/* + * jidctint.c + * + * Copyright (C) 1991-1998, Thomas G. Lane. + * Modification developed 2002-2009 by Guido Vollbeding. + * This file is part of the Independent JPEG Group's software. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains a slow-but-accurate integer implementation of the + * inverse DCT (Discrete Cosine Transform). In the IJG code, this routine + * must also perform dequantization of the input coefficients. + * + * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT + * on each row (or vice versa, but it's more convenient to emit a row at + * a time). Direct algorithms are also available, but they are much more + * complex and seem not to be any faster when reduced to code. + * + * This implementation is based on an algorithm described in + * C. Loeffler, A. Ligtenberg and G. Moschytz, "Practical Fast 1-D DCT + * Algorithms with 11 Multiplications", Proc. Int'l. Conf. on Acoustics, + * Speech, and Signal Processing 1989 (ICASSP '89), pp. 988-991. + * The primary algorithm described there uses 11 multiplies and 29 adds. + * We use their alternate method with 12 multiplies and 32 adds. + * The advantage of this method is that no data path contains more than one + * multiplication; this allows a very simple and accurate implementation in + * scaled fixed-point arithmetic, with a minimal number of shifts. + * + * We also provide IDCT routines with various output sample block sizes for + * direct resolution reduction or enlargement without additional resampling: + * NxN (N=1...16) pixels for one 8x8 input DCT block. + * + * For N<8 we simply take the corresponding low-frequency coefficients of + * the 8x8 input DCT block and apply an NxN point IDCT on the sub-block + * to yield the downscaled outputs. + * This can be seen as direct low-pass downsampling from the DCT domain + * point of view rather than the usual spatial domain point of view, + * yielding significant computational savings and results at least + * as good as common bilinear (averaging) spatial downsampling. + * + * For N>8 we apply a partial NxN IDCT on the 8 input coefficients as + * lower frequencies and higher frequencies assumed to be zero. + * It turns out that the computational effort is similar to the 8x8 IDCT + * regarding the output size. + * Furthermore, the scaling and descaling is the same for all IDCT sizes. + * + * CAUTION: We rely on the FIX() macro except for the N=1,2,4,8 cases + * since there would be too many additional constants to pre-calculate. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jdct.h" /* Private declarations for DCT subsystem */ + +#ifdef DCT_ISLOW_SUPPORTED + + +/* + * This module is specialized to the case DCTSIZE = 8. + */ + +#if DCTSIZE != 8 + Sorry, this code only copes with 8x8 DCT blocks. /* deliberate syntax err */ +#endif + + +/* + * The poop on this scaling stuff is as follows: + * + * Each 1-D IDCT step produces outputs which are a factor of sqrt(N) + * larger than the true IDCT outputs. The final outputs are therefore + * a factor of N larger than desired; since N=8 this can be cured by + * a simple right shift at the end of the algorithm. The advantage of + * this arrangement is that we save two multiplications per 1-D IDCT, + * because the y0 and y4 inputs need not be divided by sqrt(N). + * + * We have to do addition and subtraction of the integer inputs, which + * is no problem, and multiplication by fractional constants, which is + * a problem to do in integer arithmetic. We multiply all the constants + * by CONST_SCALE and convert them to integer constants (thus retaining + * CONST_BITS bits of precision in the constants). After doing a + * multiplication we have to divide the product by CONST_SCALE, with proper + * rounding, to produce the correct output. This division can be done + * cheaply as a right shift of CONST_BITS bits. We postpone shifting + * as long as possible so that partial sums can be added together with + * full fractional precision. + * + * The outputs of the first pass are scaled up by PASS1_BITS bits so that + * they are represented to better-than-integral precision. These outputs + * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word + * with the recommended scaling. (To scale up 12-bit sample data further, an + * intermediate INT32 array would be needed.) + * + * To avoid overflow of the 32-bit intermediate results in pass 2, we must + * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26. Error analysis + * shows that the values given below are the most effective. + */ + +#if BITS_IN_JSAMPLE == 8 +#define CONST_BITS 13 +#define PASS1_BITS 2 +#else +#define CONST_BITS 13 +#define PASS1_BITS 1 /* lose a little precision to avoid overflow */ +#endif + +/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus + * causing a lot of useless floating-point operations at run time. + * To get around this we use the following pre-calculated constants. + * If you change CONST_BITS you may want to add appropriate values. + * (With a reasonable C compiler, you can just rely on the FIX() macro...) + */ + +#if CONST_BITS == 13 +#define FIX_0_298631336 ((INT32) 2446) /* FIX(0.298631336) */ +#define FIX_0_390180644 ((INT32) 3196) /* FIX(0.390180644) */ +#define FIX_0_541196100 ((INT32) 4433) /* FIX(0.541196100) */ +#define FIX_0_765366865 ((INT32) 6270) /* FIX(0.765366865) */ +#define FIX_0_899976223 ((INT32) 7373) /* FIX(0.899976223) */ +#define FIX_1_175875602 ((INT32) 9633) /* FIX(1.175875602) */ +#define FIX_1_501321110 ((INT32) 12299) /* FIX(1.501321110) */ +#define FIX_1_847759065 ((INT32) 15137) /* FIX(1.847759065) */ +#define FIX_1_961570560 ((INT32) 16069) /* FIX(1.961570560) */ +#define FIX_2_053119869 ((INT32) 16819) /* FIX(2.053119869) */ +#define FIX_2_562915447 ((INT32) 20995) /* FIX(2.562915447) */ +#define FIX_3_072711026 ((INT32) 25172) /* FIX(3.072711026) */ +#else +#define FIX_0_298631336 FIX(0.298631336) +#define FIX_0_390180644 FIX(0.390180644) +#define FIX_0_541196100 FIX(0.541196100) +#define FIX_0_765366865 FIX(0.765366865) +#define FIX_0_899976223 FIX(0.899976223) +#define FIX_1_175875602 FIX(1.175875602) +#define FIX_1_501321110 FIX(1.501321110) +#define FIX_1_847759065 FIX(1.847759065) +#define FIX_1_961570560 FIX(1.961570560) +#define FIX_2_053119869 FIX(2.053119869) +#define FIX_2_562915447 FIX(2.562915447) +#define FIX_3_072711026 FIX(3.072711026) +#endif + + +/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result. + * For 8-bit samples with the recommended scaling, all the variable + * and constant values involved are no more than 16 bits wide, so a + * 16x16->32 bit multiply can be used instead of a full 32x32 multiply. + * For 12-bit samples, a full 32-bit multiplication will be needed. + */ + +#if BITS_IN_JSAMPLE == 8 +#define MULTIPLY(var,const) MULTIPLY16C16(var,const) +#else +#define MULTIPLY(var,const) ((var) * (const)) +#endif + + +/* Dequantize a coefficient by multiplying it by the multiplier-table + * entry; produce an int result. In this module, both inputs and result + * are 16 bits or less, so either int or short multiply will work. + */ + +#define DEQUANTIZE(coef,quantval) (((ISLOW_MULT_TYPE) (coef)) * (quantval)) + + +/* + * Perform dequantization and inverse DCT on one block of coefficients. + */ + +GLOBAL(void) +jpeg_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp0, tmp1, tmp2, tmp3; + INT32 tmp10, tmp11, tmp12, tmp13; + INT32 z1, z2, z3, z4, z5; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[DCTSIZE2]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + /* Note results are scaled up by sqrt(8) compared to a true IDCT; */ + /* furthermore, we scale the results by 2**PASS1_BITS. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = DCTSIZE; ctr > 0; ctr--) { + /* Due to quantization, we will usually find that many of the input + * coefficients are zero, especially the AC terms. We can exploit this + * by short-circuiting the IDCT calculation for any column in which all + * the AC terms are zero. In that case each output is equal to the + * DC coefficient (with scale factor as needed). + * With typical images and quantization tables, half or more of the + * column DCT calculations can be simplified this way. + */ + + if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 && + inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 && + inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 && + inptr[DCTSIZE*7] == 0) { + /* AC terms all zero */ + int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS; + + wsptr[DCTSIZE*0] = dcval; + wsptr[DCTSIZE*1] = dcval; + wsptr[DCTSIZE*2] = dcval; + wsptr[DCTSIZE*3] = dcval; + wsptr[DCTSIZE*4] = dcval; + wsptr[DCTSIZE*5] = dcval; + wsptr[DCTSIZE*6] = dcval; + wsptr[DCTSIZE*7] = dcval; + + inptr++; /* advance pointers to next column */ + quantptr++; + wsptr++; + continue; + } + + /* Even part: reverse the even part of the forward DCT. */ + /* The rotator is sqrt(2)*c(-6). */ + + z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]); + + z1 = MULTIPLY(z2 + z3, FIX_0_541196100); + tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); + tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); + + z2 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + z3 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]); + + tmp0 = (z2 + z3) << CONST_BITS; + tmp1 = (z2 - z3) << CONST_BITS; + + tmp10 = tmp0 + tmp3; + tmp13 = tmp0 - tmp3; + tmp11 = tmp1 + tmp2; + tmp12 = tmp1 - tmp2; + + /* Odd part per figure 8; the matrix is unitary and hence its + * transpose is its inverse. i0..i3 are y7,y5,y3,y1 respectively. + */ + + tmp0 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]); + tmp1 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + tmp2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + tmp3 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + + z1 = tmp0 + tmp3; + z2 = tmp1 + tmp2; + z3 = tmp0 + tmp2; + z4 = tmp1 + tmp3; + z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */ + + tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */ + tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */ + tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */ + tmp3 = MULTIPLY(tmp3, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */ + z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */ + z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */ + z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */ + z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */ + + z3 += z5; + z4 += z5; + + tmp0 += z1 + z3; + tmp1 += z2 + z4; + tmp2 += z2 + z3; + tmp3 += z1 + z4; + + /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */ + + wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS); + wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS); + wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS); + wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS); + wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS); + wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS); + wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS); + wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS); + + inptr++; /* advance pointers to next column */ + quantptr++; + wsptr++; + } + + /* Pass 2: process rows from work array, store into output array. */ + /* Note that we must descale the results by a factor of 8 == 2**3, */ + /* and also undo the PASS1_BITS scaling. */ + + wsptr = workspace; + for (ctr = 0; ctr < DCTSIZE; ctr++) { + outptr = output_buf[ctr] + output_col; + /* Rows of zeroes can be exploited in the same way as we did with columns. + * However, the column calculation has created many nonzero AC terms, so + * the simplification applies less often (typically 5% to 10% of the time). + * On machines with very fast multiplication, it's possible that the + * test takes more time than it's worth. In that case this section + * may be commented out. + */ + +#ifndef NO_ZERO_ROW_TEST + if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 && wsptr[4] == 0 && + wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) { + /* AC terms all zero */ + JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3) + & RANGE_MASK]; + + outptr[0] = dcval; + outptr[1] = dcval; + outptr[2] = dcval; + outptr[3] = dcval; + outptr[4] = dcval; + outptr[5] = dcval; + outptr[6] = dcval; + outptr[7] = dcval; + + wsptr += DCTSIZE; /* advance pointer to next row */ + continue; + } +#endif + + /* Even part: reverse the even part of the forward DCT. */ + /* The rotator is sqrt(2)*c(-6). */ + + z2 = (INT32) wsptr[2]; + z3 = (INT32) wsptr[6]; + + z1 = MULTIPLY(z2 + z3, FIX_0_541196100); + tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); + tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); + + tmp0 = ((INT32) wsptr[0] + (INT32) wsptr[4]) << CONST_BITS; + tmp1 = ((INT32) wsptr[0] - (INT32) wsptr[4]) << CONST_BITS; + + tmp10 = tmp0 + tmp3; + tmp13 = tmp0 - tmp3; + tmp11 = tmp1 + tmp2; + tmp12 = tmp1 - tmp2; + + /* Odd part per figure 8; the matrix is unitary and hence its + * transpose is its inverse. i0..i3 are y7,y5,y3,y1 respectively. + */ + + tmp0 = (INT32) wsptr[7]; + tmp1 = (INT32) wsptr[5]; + tmp2 = (INT32) wsptr[3]; + tmp3 = (INT32) wsptr[1]; + + z1 = tmp0 + tmp3; + z2 = tmp1 + tmp2; + z3 = tmp0 + tmp2; + z4 = tmp1 + tmp3; + z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */ + + tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */ + tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */ + tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */ + tmp3 = MULTIPLY(tmp3, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */ + z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */ + z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */ + z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */ + z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */ + + z3 += z5; + z4 += z5; + + tmp0 += z1 + z3; + tmp1 += z2 + z4; + tmp2 += z2 + z3; + tmp3 += z1 + z4; + + /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */ + + outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp3, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[7] = range_limit[(int) DESCALE(tmp10 - tmp3, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[(int) DESCALE(tmp11 + tmp2, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[6] = range_limit[(int) DESCALE(tmp11 - tmp2, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[(int) DESCALE(tmp12 + tmp1, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[5] = range_limit[(int) DESCALE(tmp12 - tmp1, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[3] = range_limit[(int) DESCALE(tmp13 + tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[4] = range_limit[(int) DESCALE(tmp13 - tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += DCTSIZE; /* advance pointer to next row */ + } +} + +#ifdef IDCT_SCALING_SUPPORTED + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a 7x7 output block. + * + * Optimized algorithm with 12 multiplications in the 1-D kernel. + * cK represents sqrt(2) * cos(K*pi/14). + */ + +GLOBAL(void) +jpeg_idct_7x7 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp0, tmp1, tmp2, tmp10, tmp11, tmp12, tmp13; + INT32 z1, z2, z3; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[7*7]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = 0; ctr < 7; ctr++, inptr++, quantptr++, wsptr++) { + /* Even part */ + + tmp13 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + tmp13 <<= CONST_BITS; + /* Add fudge factor here for final descale. */ + tmp13 += ONE << (CONST_BITS-PASS1_BITS-1); + + z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + z2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]); + z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]); + + tmp10 = MULTIPLY(z2 - z3, FIX(0.881747734)); /* c4 */ + tmp12 = MULTIPLY(z1 - z2, FIX(0.314692123)); /* c6 */ + tmp11 = tmp10 + tmp12 + tmp13 - MULTIPLY(z2, FIX(1.841218003)); /* c2+c4-c6 */ + tmp0 = z1 + z3; + z2 -= tmp0; + tmp0 = MULTIPLY(tmp0, FIX(1.274162392)) + tmp13; /* c2 */ + tmp10 += tmp0 - MULTIPLY(z3, FIX(0.077722536)); /* c2-c4-c6 */ + tmp12 += tmp0 - MULTIPLY(z1, FIX(2.470602249)); /* c2+c4+c6 */ + tmp13 += MULTIPLY(z2, FIX(1.414213562)); /* c0 */ + + /* Odd part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + + tmp1 = MULTIPLY(z1 + z2, FIX(0.935414347)); /* (c3+c1-c5)/2 */ + tmp2 = MULTIPLY(z1 - z2, FIX(0.170262339)); /* (c3+c5-c1)/2 */ + tmp0 = tmp1 - tmp2; + tmp1 += tmp2; + tmp2 = MULTIPLY(z2 + z3, - FIX(1.378756276)); /* -c1 */ + tmp1 += tmp2; + z2 = MULTIPLY(z1 + z3, FIX(0.613604268)); /* c5 */ + tmp0 += z2; + tmp2 += z2 + MULTIPLY(z3, FIX(1.870828693)); /* c3+c1-c5 */ + + /* Final output stage */ + + wsptr[7*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS); + wsptr[7*6] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS); + wsptr[7*1] = (int) RIGHT_SHIFT(tmp11 + tmp1, CONST_BITS-PASS1_BITS); + wsptr[7*5] = (int) RIGHT_SHIFT(tmp11 - tmp1, CONST_BITS-PASS1_BITS); + wsptr[7*2] = (int) RIGHT_SHIFT(tmp12 + tmp2, CONST_BITS-PASS1_BITS); + wsptr[7*4] = (int) RIGHT_SHIFT(tmp12 - tmp2, CONST_BITS-PASS1_BITS); + wsptr[7*3] = (int) RIGHT_SHIFT(tmp13, CONST_BITS-PASS1_BITS); + } + + /* Pass 2: process 7 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 7; ctr++) { + outptr = output_buf[ctr] + output_col; + + /* Even part */ + + /* Add fudge factor here for final descale. */ + tmp13 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2)); + tmp13 <<= CONST_BITS; + + z1 = (INT32) wsptr[2]; + z2 = (INT32) wsptr[4]; + z3 = (INT32) wsptr[6]; + + tmp10 = MULTIPLY(z2 - z3, FIX(0.881747734)); /* c4 */ + tmp12 = MULTIPLY(z1 - z2, FIX(0.314692123)); /* c6 */ + tmp11 = tmp10 + tmp12 + tmp13 - MULTIPLY(z2, FIX(1.841218003)); /* c2+c4-c6 */ + tmp0 = z1 + z3; + z2 -= tmp0; + tmp0 = MULTIPLY(tmp0, FIX(1.274162392)) + tmp13; /* c2 */ + tmp10 += tmp0 - MULTIPLY(z3, FIX(0.077722536)); /* c2-c4-c6 */ + tmp12 += tmp0 - MULTIPLY(z1, FIX(2.470602249)); /* c2+c4+c6 */ + tmp13 += MULTIPLY(z2, FIX(1.414213562)); /* c0 */ + + /* Odd part */ + + z1 = (INT32) wsptr[1]; + z2 = (INT32) wsptr[3]; + z3 = (INT32) wsptr[5]; + + tmp1 = MULTIPLY(z1 + z2, FIX(0.935414347)); /* (c3+c1-c5)/2 */ + tmp2 = MULTIPLY(z1 - z2, FIX(0.170262339)); /* (c3+c5-c1)/2 */ + tmp0 = tmp1 - tmp2; + tmp1 += tmp2; + tmp2 = MULTIPLY(z2 + z3, - FIX(1.378756276)); /* -c1 */ + tmp1 += tmp2; + z2 = MULTIPLY(z1 + z3, FIX(0.613604268)); /* c5 */ + tmp0 += z2; + tmp2 += z2 + MULTIPLY(z3, FIX(1.870828693)); /* c3+c1-c5 */ + + /* Final output stage */ + + outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp11 + tmp1, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp11 - tmp1, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp12 + tmp2, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp12 - tmp2, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += 7; /* advance pointer to next row */ + } +} + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a reduced-size 6x6 output block. + * + * Optimized algorithm with 3 multiplications in the 1-D kernel. + * cK represents sqrt(2) * cos(K*pi/12). + */ + +GLOBAL(void) +jpeg_idct_6x6 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp0, tmp1, tmp2, tmp10, tmp11, tmp12; + INT32 z1, z2, z3; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[6*6]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = 0; ctr < 6; ctr++, inptr++, quantptr++, wsptr++) { + /* Even part */ + + tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + tmp0 <<= CONST_BITS; + /* Add fudge factor here for final descale. */ + tmp0 += ONE << (CONST_BITS-PASS1_BITS-1); + tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]); + tmp10 = MULTIPLY(tmp2, FIX(0.707106781)); /* c4 */ + tmp1 = tmp0 + tmp10; + tmp11 = RIGHT_SHIFT(tmp0 - tmp10 - tmp10, CONST_BITS-PASS1_BITS); + tmp10 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + tmp0 = MULTIPLY(tmp10, FIX(1.224744871)); /* c2 */ + tmp10 = tmp1 + tmp0; + tmp12 = tmp1 - tmp0; + + /* Odd part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + tmp1 = MULTIPLY(z1 + z3, FIX(0.366025404)); /* c5 */ + tmp0 = tmp1 + ((z1 + z2) << CONST_BITS); + tmp2 = tmp1 + ((z3 - z2) << CONST_BITS); + tmp1 = (z1 - z2 - z3) << PASS1_BITS; + + /* Final output stage */ + + wsptr[6*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS); + wsptr[6*5] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS); + wsptr[6*1] = (int) (tmp11 + tmp1); + wsptr[6*4] = (int) (tmp11 - tmp1); + wsptr[6*2] = (int) RIGHT_SHIFT(tmp12 + tmp2, CONST_BITS-PASS1_BITS); + wsptr[6*3] = (int) RIGHT_SHIFT(tmp12 - tmp2, CONST_BITS-PASS1_BITS); + } + + /* Pass 2: process 6 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 6; ctr++) { + outptr = output_buf[ctr] + output_col; + + /* Even part */ + + /* Add fudge factor here for final descale. */ + tmp0 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2)); + tmp0 <<= CONST_BITS; + tmp2 = (INT32) wsptr[4]; + tmp10 = MULTIPLY(tmp2, FIX(0.707106781)); /* c4 */ + tmp1 = tmp0 + tmp10; + tmp11 = tmp0 - tmp10 - tmp10; + tmp10 = (INT32) wsptr[2]; + tmp0 = MULTIPLY(tmp10, FIX(1.224744871)); /* c2 */ + tmp10 = tmp1 + tmp0; + tmp12 = tmp1 - tmp0; + + /* Odd part */ + + z1 = (INT32) wsptr[1]; + z2 = (INT32) wsptr[3]; + z3 = (INT32) wsptr[5]; + tmp1 = MULTIPLY(z1 + z3, FIX(0.366025404)); /* c5 */ + tmp0 = tmp1 + ((z1 + z2) << CONST_BITS); + tmp2 = tmp1 + ((z3 - z2) << CONST_BITS); + tmp1 = (z1 - z2 - z3) << CONST_BITS; + + /* Final output stage */ + + outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp11 + tmp1, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp11 - tmp1, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp12 + tmp2, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp12 - tmp2, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += 6; /* advance pointer to next row */ + } +} + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a reduced-size 5x5 output block. + * + * Optimized algorithm with 5 multiplications in the 1-D kernel. + * cK represents sqrt(2) * cos(K*pi/10). + */ + +GLOBAL(void) +jpeg_idct_5x5 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp0, tmp1, tmp10, tmp11, tmp12; + INT32 z1, z2, z3; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[5*5]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = 0; ctr < 5; ctr++, inptr++, quantptr++, wsptr++) { + /* Even part */ + + tmp12 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + tmp12 <<= CONST_BITS; + /* Add fudge factor here for final descale. */ + tmp12 += ONE << (CONST_BITS-PASS1_BITS-1); + tmp0 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + tmp1 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]); + z1 = MULTIPLY(tmp0 + tmp1, FIX(0.790569415)); /* (c2+c4)/2 */ + z2 = MULTIPLY(tmp0 - tmp1, FIX(0.353553391)); /* (c2-c4)/2 */ + z3 = tmp12 + z2; + tmp10 = z3 + z1; + tmp11 = z3 - z1; + tmp12 -= z2 << 2; + + /* Odd part */ + + z2 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + z3 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + + z1 = MULTIPLY(z2 + z3, FIX(0.831253876)); /* c3 */ + tmp0 = z1 + MULTIPLY(z2, FIX(0.513743148)); /* c1-c3 */ + tmp1 = z1 - MULTIPLY(z3, FIX(2.176250899)); /* c1+c3 */ + + /* Final output stage */ + + wsptr[5*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS); + wsptr[5*4] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS); + wsptr[5*1] = (int) RIGHT_SHIFT(tmp11 + tmp1, CONST_BITS-PASS1_BITS); + wsptr[5*3] = (int) RIGHT_SHIFT(tmp11 - tmp1, CONST_BITS-PASS1_BITS); + wsptr[5*2] = (int) RIGHT_SHIFT(tmp12, CONST_BITS-PASS1_BITS); + } + + /* Pass 2: process 5 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 5; ctr++) { + outptr = output_buf[ctr] + output_col; + + /* Even part */ + + /* Add fudge factor here for final descale. */ + tmp12 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2)); + tmp12 <<= CONST_BITS; + tmp0 = (INT32) wsptr[2]; + tmp1 = (INT32) wsptr[4]; + z1 = MULTIPLY(tmp0 + tmp1, FIX(0.790569415)); /* (c2+c4)/2 */ + z2 = MULTIPLY(tmp0 - tmp1, FIX(0.353553391)); /* (c2-c4)/2 */ + z3 = tmp12 + z2; + tmp10 = z3 + z1; + tmp11 = z3 - z1; + tmp12 -= z2 << 2; + + /* Odd part */ + + z2 = (INT32) wsptr[1]; + z3 = (INT32) wsptr[3]; + + z1 = MULTIPLY(z2 + z3, FIX(0.831253876)); /* c3 */ + tmp0 = z1 + MULTIPLY(z2, FIX(0.513743148)); /* c1-c3 */ + tmp1 = z1 - MULTIPLY(z3, FIX(2.176250899)); /* c1+c3 */ + + /* Final output stage */ + + outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp11 + tmp1, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp11 - tmp1, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += 5; /* advance pointer to next row */ + } +} + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a reduced-size 3x3 output block. + * + * Optimized algorithm with 2 multiplications in the 1-D kernel. + * cK represents sqrt(2) * cos(K*pi/6). + */ + +GLOBAL(void) +jpeg_idct_3x3 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp0, tmp2, tmp10, tmp12; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[3*3]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = 0; ctr < 3; ctr++, inptr++, quantptr++, wsptr++) { + /* Even part */ + + tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + tmp0 <<= CONST_BITS; + /* Add fudge factor here for final descale. */ + tmp0 += ONE << (CONST_BITS-PASS1_BITS-1); + tmp2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + tmp12 = MULTIPLY(tmp2, FIX(0.707106781)); /* c2 */ + tmp10 = tmp0 + tmp12; + tmp2 = tmp0 - tmp12 - tmp12; + + /* Odd part */ + + tmp12 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + tmp0 = MULTIPLY(tmp12, FIX(1.224744871)); /* c1 */ + + /* Final output stage */ + + wsptr[3*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS); + wsptr[3*2] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS); + wsptr[3*1] = (int) RIGHT_SHIFT(tmp2, CONST_BITS-PASS1_BITS); + } + + /* Pass 2: process 3 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 3; ctr++) { + outptr = output_buf[ctr] + output_col; + + /* Even part */ + + /* Add fudge factor here for final descale. */ + tmp0 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2)); + tmp0 <<= CONST_BITS; + tmp2 = (INT32) wsptr[2]; + tmp12 = MULTIPLY(tmp2, FIX(0.707106781)); /* c2 */ + tmp10 = tmp0 + tmp12; + tmp2 = tmp0 - tmp12 - tmp12; + + /* Odd part */ + + tmp12 = (INT32) wsptr[1]; + tmp0 = MULTIPLY(tmp12, FIX(1.224744871)); /* c1 */ + + /* Final output stage */ + + outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp2, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += 3; /* advance pointer to next row */ + } +} + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a 9x9 output block. + * + * Optimized algorithm with 10 multiplications in the 1-D kernel. + * cK represents sqrt(2) * cos(K*pi/18). + */ + +GLOBAL(void) +jpeg_idct_9x9 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp0, tmp1, tmp2, tmp3, tmp10, tmp11, tmp12, tmp13, tmp14; + INT32 z1, z2, z3, z4; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[8*9]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) { + /* Even part */ + + tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + tmp0 <<= CONST_BITS; + /* Add fudge factor here for final descale. */ + tmp0 += ONE << (CONST_BITS-PASS1_BITS-1); + + z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + z2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]); + z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]); + + tmp3 = MULTIPLY(z3, FIX(0.707106781)); /* c6 */ + tmp1 = tmp0 + tmp3; + tmp2 = tmp0 - tmp3 - tmp3; + + tmp0 = MULTIPLY(z1 - z2, FIX(0.707106781)); /* c6 */ + tmp11 = tmp2 + tmp0; + tmp14 = tmp2 - tmp0 - tmp0; + + tmp0 = MULTIPLY(z1 + z2, FIX(1.328926049)); /* c2 */ + tmp2 = MULTIPLY(z1, FIX(1.083350441)); /* c4 */ + tmp3 = MULTIPLY(z2, FIX(0.245575608)); /* c8 */ + + tmp10 = tmp1 + tmp0 - tmp3; + tmp12 = tmp1 - tmp0 + tmp2; + tmp13 = tmp1 - tmp2 + tmp3; + + /* Odd part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]); + + z2 = MULTIPLY(z2, - FIX(1.224744871)); /* -c3 */ + + tmp2 = MULTIPLY(z1 + z3, FIX(0.909038955)); /* c5 */ + tmp3 = MULTIPLY(z1 + z4, FIX(0.483689525)); /* c7 */ + tmp0 = tmp2 + tmp3 - z2; + tmp1 = MULTIPLY(z3 - z4, FIX(1.392728481)); /* c1 */ + tmp2 += z2 - tmp1; + tmp3 += z2 + tmp1; + tmp1 = MULTIPLY(z1 - z3 - z4, FIX(1.224744871)); /* c3 */ + + /* Final output stage */ + + wsptr[8*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS); + wsptr[8*8] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS); + wsptr[8*1] = (int) RIGHT_SHIFT(tmp11 + tmp1, CONST_BITS-PASS1_BITS); + wsptr[8*7] = (int) RIGHT_SHIFT(tmp11 - tmp1, CONST_BITS-PASS1_BITS); + wsptr[8*2] = (int) RIGHT_SHIFT(tmp12 + tmp2, CONST_BITS-PASS1_BITS); + wsptr[8*6] = (int) RIGHT_SHIFT(tmp12 - tmp2, CONST_BITS-PASS1_BITS); + wsptr[8*3] = (int) RIGHT_SHIFT(tmp13 + tmp3, CONST_BITS-PASS1_BITS); + wsptr[8*5] = (int) RIGHT_SHIFT(tmp13 - tmp3, CONST_BITS-PASS1_BITS); + wsptr[8*4] = (int) RIGHT_SHIFT(tmp14, CONST_BITS-PASS1_BITS); + } + + /* Pass 2: process 9 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 9; ctr++) { + outptr = output_buf[ctr] + output_col; + + /* Even part */ + + /* Add fudge factor here for final descale. */ + tmp0 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2)); + tmp0 <<= CONST_BITS; + + z1 = (INT32) wsptr[2]; + z2 = (INT32) wsptr[4]; + z3 = (INT32) wsptr[6]; + + tmp3 = MULTIPLY(z3, FIX(0.707106781)); /* c6 */ + tmp1 = tmp0 + tmp3; + tmp2 = tmp0 - tmp3 - tmp3; + + tmp0 = MULTIPLY(z1 - z2, FIX(0.707106781)); /* c6 */ + tmp11 = tmp2 + tmp0; + tmp14 = tmp2 - tmp0 - tmp0; + + tmp0 = MULTIPLY(z1 + z2, FIX(1.328926049)); /* c2 */ + tmp2 = MULTIPLY(z1, FIX(1.083350441)); /* c4 */ + tmp3 = MULTIPLY(z2, FIX(0.245575608)); /* c8 */ + + tmp10 = tmp1 + tmp0 - tmp3; + tmp12 = tmp1 - tmp0 + tmp2; + tmp13 = tmp1 - tmp2 + tmp3; + + /* Odd part */ + + z1 = (INT32) wsptr[1]; + z2 = (INT32) wsptr[3]; + z3 = (INT32) wsptr[5]; + z4 = (INT32) wsptr[7]; + + z2 = MULTIPLY(z2, - FIX(1.224744871)); /* -c3 */ + + tmp2 = MULTIPLY(z1 + z3, FIX(0.909038955)); /* c5 */ + tmp3 = MULTIPLY(z1 + z4, FIX(0.483689525)); /* c7 */ + tmp0 = tmp2 + tmp3 - z2; + tmp1 = MULTIPLY(z3 - z4, FIX(1.392728481)); /* c1 */ + tmp2 += z2 - tmp1; + tmp3 += z2 + tmp1; + tmp1 = MULTIPLY(z1 - z3 - z4, FIX(1.224744871)); /* c3 */ + + /* Final output stage */ + + outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[8] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp11 + tmp1, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[7] = range_limit[(int) RIGHT_SHIFT(tmp11 - tmp1, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp12 + tmp2, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp12 - tmp2, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp13 + tmp3, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp13 - tmp3, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp14, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += 8; /* advance pointer to next row */ + } +} + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a 10x10 output block. + * + * Optimized algorithm with 12 multiplications in the 1-D kernel. + * cK represents sqrt(2) * cos(K*pi/20). + */ + +GLOBAL(void) +jpeg_idct_10x10 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp10, tmp11, tmp12, tmp13, tmp14; + INT32 tmp20, tmp21, tmp22, tmp23, tmp24; + INT32 z1, z2, z3, z4, z5; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[8*10]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) { + /* Even part */ + + z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + z3 <<= CONST_BITS; + /* Add fudge factor here for final descale. */ + z3 += ONE << (CONST_BITS-PASS1_BITS-1); + z4 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]); + z1 = MULTIPLY(z4, FIX(1.144122806)); /* c4 */ + z2 = MULTIPLY(z4, FIX(0.437016024)); /* c8 */ + tmp10 = z3 + z1; + tmp11 = z3 - z2; + + tmp22 = RIGHT_SHIFT(z3 - ((z1 - z2) << 1), /* c0 = (c4-c8)*2 */ + CONST_BITS-PASS1_BITS); + + z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]); + + z1 = MULTIPLY(z2 + z3, FIX(0.831253876)); /* c6 */ + tmp12 = z1 + MULTIPLY(z2, FIX(0.513743148)); /* c2-c6 */ + tmp13 = z1 - MULTIPLY(z3, FIX(2.176250899)); /* c2+c6 */ + + tmp20 = tmp10 + tmp12; + tmp24 = tmp10 - tmp12; + tmp21 = tmp11 + tmp13; + tmp23 = tmp11 - tmp13; + + /* Odd part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]); + + tmp11 = z2 + z4; + tmp13 = z2 - z4; + + tmp12 = MULTIPLY(tmp13, FIX(0.309016994)); /* (c3-c7)/2 */ + z5 = z3 << CONST_BITS; + + z2 = MULTIPLY(tmp11, FIX(0.951056516)); /* (c3+c7)/2 */ + z4 = z5 + tmp12; + + tmp10 = MULTIPLY(z1, FIX(1.396802247)) + z2 + z4; /* c1 */ + tmp14 = MULTIPLY(z1, FIX(0.221231742)) - z2 + z4; /* c9 */ + + z2 = MULTIPLY(tmp11, FIX(0.587785252)); /* (c1-c9)/2 */ + z4 = z5 - tmp12 - (tmp13 << (CONST_BITS - 1)); + + tmp12 = (z1 - tmp13 - z3) << PASS1_BITS; + + tmp11 = MULTIPLY(z1, FIX(1.260073511)) - z2 - z4; /* c3 */ + tmp13 = MULTIPLY(z1, FIX(0.642039522)) - z2 + z4; /* c7 */ + + /* Final output stage */ + + wsptr[8*0] = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*9] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*1] = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*8] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*2] = (int) (tmp22 + tmp12); + wsptr[8*7] = (int) (tmp22 - tmp12); + wsptr[8*3] = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS); + wsptr[8*6] = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS); + wsptr[8*4] = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS); + wsptr[8*5] = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS); + } + + /* Pass 2: process 10 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 10; ctr++) { + outptr = output_buf[ctr] + output_col; + + /* Even part */ + + /* Add fudge factor here for final descale. */ + z3 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2)); + z3 <<= CONST_BITS; + z4 = (INT32) wsptr[4]; + z1 = MULTIPLY(z4, FIX(1.144122806)); /* c4 */ + z2 = MULTIPLY(z4, FIX(0.437016024)); /* c8 */ + tmp10 = z3 + z1; + tmp11 = z3 - z2; + + tmp22 = z3 - ((z1 - z2) << 1); /* c0 = (c4-c8)*2 */ + + z2 = (INT32) wsptr[2]; + z3 = (INT32) wsptr[6]; + + z1 = MULTIPLY(z2 + z3, FIX(0.831253876)); /* c6 */ + tmp12 = z1 + MULTIPLY(z2, FIX(0.513743148)); /* c2-c6 */ + tmp13 = z1 - MULTIPLY(z3, FIX(2.176250899)); /* c2+c6 */ + + tmp20 = tmp10 + tmp12; + tmp24 = tmp10 - tmp12; + tmp21 = tmp11 + tmp13; + tmp23 = tmp11 - tmp13; + + /* Odd part */ + + z1 = (INT32) wsptr[1]; + z2 = (INT32) wsptr[3]; + z3 = (INT32) wsptr[5]; + z3 <<= CONST_BITS; + z4 = (INT32) wsptr[7]; + + tmp11 = z2 + z4; + tmp13 = z2 - z4; + + tmp12 = MULTIPLY(tmp13, FIX(0.309016994)); /* (c3-c7)/2 */ + + z2 = MULTIPLY(tmp11, FIX(0.951056516)); /* (c3+c7)/2 */ + z4 = z3 + tmp12; + + tmp10 = MULTIPLY(z1, FIX(1.396802247)) + z2 + z4; /* c1 */ + tmp14 = MULTIPLY(z1, FIX(0.221231742)) - z2 + z4; /* c9 */ + + z2 = MULTIPLY(tmp11, FIX(0.587785252)); /* (c1-c9)/2 */ + z4 = z3 - tmp12 - (tmp13 << (CONST_BITS - 1)); + + tmp12 = ((z1 - tmp13) << CONST_BITS) - z3; + + tmp11 = MULTIPLY(z1, FIX(1.260073511)) - z2 - z4; /* c3 */ + tmp13 = MULTIPLY(z1, FIX(0.642039522)) - z2 + z4; /* c7 */ + + /* Final output stage */ + + outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[9] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[8] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[7] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += 8; /* advance pointer to next row */ + } +} + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a 11x11 output block. + * + * Optimized algorithm with 24 multiplications in the 1-D kernel. + * cK represents sqrt(2) * cos(K*pi/22). + */ + +GLOBAL(void) +jpeg_idct_11x11 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp10, tmp11, tmp12, tmp13, tmp14; + INT32 tmp20, tmp21, tmp22, tmp23, tmp24, tmp25; + INT32 z1, z2, z3, z4; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[8*11]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) { + /* Even part */ + + tmp10 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + tmp10 <<= CONST_BITS; + /* Add fudge factor here for final descale. */ + tmp10 += ONE << (CONST_BITS-PASS1_BITS-1); + + z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + z2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]); + z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]); + + tmp20 = MULTIPLY(z2 - z3, FIX(2.546640132)); /* c2+c4 */ + tmp23 = MULTIPLY(z2 - z1, FIX(0.430815045)); /* c2-c6 */ + z4 = z1 + z3; + tmp24 = MULTIPLY(z4, - FIX(1.155664402)); /* -(c2-c10) */ + z4 -= z2; + tmp25 = tmp10 + MULTIPLY(z4, FIX(1.356927976)); /* c2 */ + tmp21 = tmp20 + tmp23 + tmp25 - + MULTIPLY(z2, FIX(1.821790775)); /* c2+c4+c10-c6 */ + tmp20 += tmp25 + MULTIPLY(z3, FIX(2.115825087)); /* c4+c6 */ + tmp23 += tmp25 - MULTIPLY(z1, FIX(1.513598477)); /* c6+c8 */ + tmp24 += tmp25; + tmp22 = tmp24 - MULTIPLY(z3, FIX(0.788749120)); /* c8+c10 */ + tmp24 += MULTIPLY(z2, FIX(1.944413522)) - /* c2+c8 */ + MULTIPLY(z1, FIX(1.390975730)); /* c4+c10 */ + tmp25 = tmp10 - MULTIPLY(z4, FIX(1.414213562)); /* c0 */ + + /* Odd part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]); + + tmp11 = z1 + z2; + tmp14 = MULTIPLY(tmp11 + z3 + z4, FIX(0.398430003)); /* c9 */ + tmp11 = MULTIPLY(tmp11, FIX(0.887983902)); /* c3-c9 */ + tmp12 = MULTIPLY(z1 + z3, FIX(0.670361295)); /* c5-c9 */ + tmp13 = tmp14 + MULTIPLY(z1 + z4, FIX(0.366151574)); /* c7-c9 */ + tmp10 = tmp11 + tmp12 + tmp13 - + MULTIPLY(z1, FIX(0.923107866)); /* c7+c5+c3-c1-2*c9 */ + z1 = tmp14 - MULTIPLY(z2 + z3, FIX(1.163011579)); /* c7+c9 */ + tmp11 += z1 + MULTIPLY(z2, FIX(2.073276588)); /* c1+c7+3*c9-c3 */ + tmp12 += z1 - MULTIPLY(z3, FIX(1.192193623)); /* c3+c5-c7-c9 */ + z1 = MULTIPLY(z2 + z4, - FIX(1.798248910)); /* -(c1+c9) */ + tmp11 += z1; + tmp13 += z1 + MULTIPLY(z4, FIX(2.102458632)); /* c1+c5+c9-c7 */ + tmp14 += MULTIPLY(z2, - FIX(1.467221301)) + /* -(c5+c9) */ + MULTIPLY(z3, FIX(1.001388905)) - /* c1-c9 */ + MULTIPLY(z4, FIX(1.684843907)); /* c3+c9 */ + + /* Final output stage */ + + wsptr[8*0] = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*10] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*1] = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*9] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*2] = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS); + wsptr[8*8] = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS); + wsptr[8*3] = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS); + wsptr[8*7] = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS); + wsptr[8*4] = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS); + wsptr[8*6] = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS); + wsptr[8*5] = (int) RIGHT_SHIFT(tmp25, CONST_BITS-PASS1_BITS); + } + + /* Pass 2: process 11 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 11; ctr++) { + outptr = output_buf[ctr] + output_col; + + /* Even part */ + + /* Add fudge factor here for final descale. */ + tmp10 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2)); + tmp10 <<= CONST_BITS; + + z1 = (INT32) wsptr[2]; + z2 = (INT32) wsptr[4]; + z3 = (INT32) wsptr[6]; + + tmp20 = MULTIPLY(z2 - z3, FIX(2.546640132)); /* c2+c4 */ + tmp23 = MULTIPLY(z2 - z1, FIX(0.430815045)); /* c2-c6 */ + z4 = z1 + z3; + tmp24 = MULTIPLY(z4, - FIX(1.155664402)); /* -(c2-c10) */ + z4 -= z2; + tmp25 = tmp10 + MULTIPLY(z4, FIX(1.356927976)); /* c2 */ + tmp21 = tmp20 + tmp23 + tmp25 - + MULTIPLY(z2, FIX(1.821790775)); /* c2+c4+c10-c6 */ + tmp20 += tmp25 + MULTIPLY(z3, FIX(2.115825087)); /* c4+c6 */ + tmp23 += tmp25 - MULTIPLY(z1, FIX(1.513598477)); /* c6+c8 */ + tmp24 += tmp25; + tmp22 = tmp24 - MULTIPLY(z3, FIX(0.788749120)); /* c8+c10 */ + tmp24 += MULTIPLY(z2, FIX(1.944413522)) - /* c2+c8 */ + MULTIPLY(z1, FIX(1.390975730)); /* c4+c10 */ + tmp25 = tmp10 - MULTIPLY(z4, FIX(1.414213562)); /* c0 */ + + /* Odd part */ + + z1 = (INT32) wsptr[1]; + z2 = (INT32) wsptr[3]; + z3 = (INT32) wsptr[5]; + z4 = (INT32) wsptr[7]; + + tmp11 = z1 + z2; + tmp14 = MULTIPLY(tmp11 + z3 + z4, FIX(0.398430003)); /* c9 */ + tmp11 = MULTIPLY(tmp11, FIX(0.887983902)); /* c3-c9 */ + tmp12 = MULTIPLY(z1 + z3, FIX(0.670361295)); /* c5-c9 */ + tmp13 = tmp14 + MULTIPLY(z1 + z4, FIX(0.366151574)); /* c7-c9 */ + tmp10 = tmp11 + tmp12 + tmp13 - + MULTIPLY(z1, FIX(0.923107866)); /* c7+c5+c3-c1-2*c9 */ + z1 = tmp14 - MULTIPLY(z2 + z3, FIX(1.163011579)); /* c7+c9 */ + tmp11 += z1 + MULTIPLY(z2, FIX(2.073276588)); /* c1+c7+3*c9-c3 */ + tmp12 += z1 - MULTIPLY(z3, FIX(1.192193623)); /* c3+c5-c7-c9 */ + z1 = MULTIPLY(z2 + z4, - FIX(1.798248910)); /* -(c1+c9) */ + tmp11 += z1; + tmp13 += z1 + MULTIPLY(z4, FIX(2.102458632)); /* c1+c5+c9-c7 */ + tmp14 += MULTIPLY(z2, - FIX(1.467221301)) + /* -(c5+c9) */ + MULTIPLY(z3, FIX(1.001388905)) - /* c1-c9 */ + MULTIPLY(z4, FIX(1.684843907)); /* c3+c9 */ + + /* Final output stage */ + + outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[9] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[8] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[7] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp25, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += 8; /* advance pointer to next row */ + } +} + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a 12x12 output block. + * + * Optimized algorithm with 15 multiplications in the 1-D kernel. + * cK represents sqrt(2) * cos(K*pi/24). + */ + +GLOBAL(void) +jpeg_idct_12x12 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15; + INT32 tmp20, tmp21, tmp22, tmp23, tmp24, tmp25; + INT32 z1, z2, z3, z4; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[8*12]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) { + /* Even part */ + + z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + z3 <<= CONST_BITS; + /* Add fudge factor here for final descale. */ + z3 += ONE << (CONST_BITS-PASS1_BITS-1); + + z4 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]); + z4 = MULTIPLY(z4, FIX(1.224744871)); /* c4 */ + + tmp10 = z3 + z4; + tmp11 = z3 - z4; + + z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + z4 = MULTIPLY(z1, FIX(1.366025404)); /* c2 */ + z1 <<= CONST_BITS; + z2 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]); + z2 <<= CONST_BITS; + + tmp12 = z1 - z2; + + tmp21 = z3 + tmp12; + tmp24 = z3 - tmp12; + + tmp12 = z4 + z2; + + tmp20 = tmp10 + tmp12; + tmp25 = tmp10 - tmp12; + + tmp12 = z4 - z1 - z2; + + tmp22 = tmp11 + tmp12; + tmp23 = tmp11 - tmp12; + + /* Odd part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]); + + tmp11 = MULTIPLY(z2, FIX(1.306562965)); /* c3 */ + tmp14 = MULTIPLY(z2, - FIX_0_541196100); /* -c9 */ + + tmp10 = z1 + z3; + tmp15 = MULTIPLY(tmp10 + z4, FIX(0.860918669)); /* c7 */ + tmp12 = tmp15 + MULTIPLY(tmp10, FIX(0.261052384)); /* c5-c7 */ + tmp10 = tmp12 + tmp11 + MULTIPLY(z1, FIX(0.280143716)); /* c1-c5 */ + tmp13 = MULTIPLY(z3 + z4, - FIX(1.045510580)); /* -(c7+c11) */ + tmp12 += tmp13 + tmp14 - MULTIPLY(z3, FIX(1.478575242)); /* c1+c5-c7-c11 */ + tmp13 += tmp15 - tmp11 + MULTIPLY(z4, FIX(1.586706681)); /* c1+c11 */ + tmp15 += tmp14 - MULTIPLY(z1, FIX(0.676326758)) - /* c7-c11 */ + MULTIPLY(z4, FIX(1.982889723)); /* c5+c7 */ + + z1 -= z4; + z2 -= z3; + z3 = MULTIPLY(z1 + z2, FIX_0_541196100); /* c9 */ + tmp11 = z3 + MULTIPLY(z1, FIX_0_765366865); /* c3-c9 */ + tmp14 = z3 - MULTIPLY(z2, FIX_1_847759065); /* c3+c9 */ + + /* Final output stage */ + + wsptr[8*0] = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*11] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*1] = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*10] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*2] = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS); + wsptr[8*9] = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS); + wsptr[8*3] = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS); + wsptr[8*8] = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS); + wsptr[8*4] = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS); + wsptr[8*7] = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS); + wsptr[8*5] = (int) RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS-PASS1_BITS); + wsptr[8*6] = (int) RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS-PASS1_BITS); + } + + /* Pass 2: process 12 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 12; ctr++) { + outptr = output_buf[ctr] + output_col; + + /* Even part */ + + /* Add fudge factor here for final descale. */ + z3 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2)); + z3 <<= CONST_BITS; + + z4 = (INT32) wsptr[4]; + z4 = MULTIPLY(z4, FIX(1.224744871)); /* c4 */ + + tmp10 = z3 + z4; + tmp11 = z3 - z4; + + z1 = (INT32) wsptr[2]; + z4 = MULTIPLY(z1, FIX(1.366025404)); /* c2 */ + z1 <<= CONST_BITS; + z2 = (INT32) wsptr[6]; + z2 <<= CONST_BITS; + + tmp12 = z1 - z2; + + tmp21 = z3 + tmp12; + tmp24 = z3 - tmp12; + + tmp12 = z4 + z2; + + tmp20 = tmp10 + tmp12; + tmp25 = tmp10 - tmp12; + + tmp12 = z4 - z1 - z2; + + tmp22 = tmp11 + tmp12; + tmp23 = tmp11 - tmp12; + + /* Odd part */ + + z1 = (INT32) wsptr[1]; + z2 = (INT32) wsptr[3]; + z3 = (INT32) wsptr[5]; + z4 = (INT32) wsptr[7]; + + tmp11 = MULTIPLY(z2, FIX(1.306562965)); /* c3 */ + tmp14 = MULTIPLY(z2, - FIX_0_541196100); /* -c9 */ + + tmp10 = z1 + z3; + tmp15 = MULTIPLY(tmp10 + z4, FIX(0.860918669)); /* c7 */ + tmp12 = tmp15 + MULTIPLY(tmp10, FIX(0.261052384)); /* c5-c7 */ + tmp10 = tmp12 + tmp11 + MULTIPLY(z1, FIX(0.280143716)); /* c1-c5 */ + tmp13 = MULTIPLY(z3 + z4, - FIX(1.045510580)); /* -(c7+c11) */ + tmp12 += tmp13 + tmp14 - MULTIPLY(z3, FIX(1.478575242)); /* c1+c5-c7-c11 */ + tmp13 += tmp15 - tmp11 + MULTIPLY(z4, FIX(1.586706681)); /* c1+c11 */ + tmp15 += tmp14 - MULTIPLY(z1, FIX(0.676326758)) - /* c7-c11 */ + MULTIPLY(z4, FIX(1.982889723)); /* c5+c7 */ + + z1 -= z4; + z2 -= z3; + z3 = MULTIPLY(z1 + z2, FIX_0_541196100); /* c9 */ + tmp11 = z3 + MULTIPLY(z1, FIX_0_765366865); /* c3-c9 */ + tmp14 = z3 - MULTIPLY(z2, FIX_1_847759065); /* c3+c9 */ + + /* Final output stage */ + + outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[9] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[8] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[7] = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp15, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp15, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += 8; /* advance pointer to next row */ + } +} + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a 13x13 output block. + * + * Optimized algorithm with 29 multiplications in the 1-D kernel. + * cK represents sqrt(2) * cos(K*pi/26). + */ + +GLOBAL(void) +jpeg_idct_13x13 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15; + INT32 tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26; + INT32 z1, z2, z3, z4; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[8*13]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) { + /* Even part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + z1 <<= CONST_BITS; + /* Add fudge factor here for final descale. */ + z1 += ONE << (CONST_BITS-PASS1_BITS-1); + + z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + z3 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]); + z4 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]); + + tmp10 = z3 + z4; + tmp11 = z3 - z4; + + tmp12 = MULTIPLY(tmp10, FIX(1.155388986)); /* (c4+c6)/2 */ + tmp13 = MULTIPLY(tmp11, FIX(0.096834934)) + z1; /* (c4-c6)/2 */ + + tmp20 = MULTIPLY(z2, FIX(1.373119086)) + tmp12 + tmp13; /* c2 */ + tmp22 = MULTIPLY(z2, FIX(0.501487041)) - tmp12 + tmp13; /* c10 */ + + tmp12 = MULTIPLY(tmp10, FIX(0.316450131)); /* (c8-c12)/2 */ + tmp13 = MULTIPLY(tmp11, FIX(0.486914739)) + z1; /* (c8+c12)/2 */ + + tmp21 = MULTIPLY(z2, FIX(1.058554052)) - tmp12 + tmp13; /* c6 */ + tmp25 = MULTIPLY(z2, - FIX(1.252223920)) + tmp12 + tmp13; /* c4 */ + + tmp12 = MULTIPLY(tmp10, FIX(0.435816023)); /* (c2-c10)/2 */ + tmp13 = MULTIPLY(tmp11, FIX(0.937303064)) - z1; /* (c2+c10)/2 */ + + tmp23 = MULTIPLY(z2, - FIX(0.170464608)) - tmp12 - tmp13; /* c12 */ + tmp24 = MULTIPLY(z2, - FIX(0.803364869)) + tmp12 - tmp13; /* c8 */ + + tmp26 = MULTIPLY(tmp11 - z2, FIX(1.414213562)) + z1; /* c0 */ + + /* Odd part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]); + + tmp11 = MULTIPLY(z1 + z2, FIX(1.322312651)); /* c3 */ + tmp12 = MULTIPLY(z1 + z3, FIX(1.163874945)); /* c5 */ + tmp15 = z1 + z4; + tmp13 = MULTIPLY(tmp15, FIX(0.937797057)); /* c7 */ + tmp10 = tmp11 + tmp12 + tmp13 - + MULTIPLY(z1, FIX(2.020082300)); /* c7+c5+c3-c1 */ + tmp14 = MULTIPLY(z2 + z3, - FIX(0.338443458)); /* -c11 */ + tmp11 += tmp14 + MULTIPLY(z2, FIX(0.837223564)); /* c5+c9+c11-c3 */ + tmp12 += tmp14 - MULTIPLY(z3, FIX(1.572116027)); /* c1+c5-c9-c11 */ + tmp14 = MULTIPLY(z2 + z4, - FIX(1.163874945)); /* -c5 */ + tmp11 += tmp14; + tmp13 += tmp14 + MULTIPLY(z4, FIX(2.205608352)); /* c3+c5+c9-c7 */ + tmp14 = MULTIPLY(z3 + z4, - FIX(0.657217813)); /* -c9 */ + tmp12 += tmp14; + tmp13 += tmp14; + tmp15 = MULTIPLY(tmp15, FIX(0.338443458)); /* c11 */ + tmp14 = tmp15 + MULTIPLY(z1, FIX(0.318774355)) - /* c9-c11 */ + MULTIPLY(z2, FIX(0.466105296)); /* c1-c7 */ + z1 = MULTIPLY(z3 - z2, FIX(0.937797057)); /* c7 */ + tmp14 += z1; + tmp15 += z1 + MULTIPLY(z3, FIX(0.384515595)) - /* c3-c7 */ + MULTIPLY(z4, FIX(1.742345811)); /* c1+c11 */ + + /* Final output stage */ + + wsptr[8*0] = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*12] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*1] = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*11] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*2] = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS); + wsptr[8*10] = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS); + wsptr[8*3] = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS); + wsptr[8*9] = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS); + wsptr[8*4] = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS); + wsptr[8*8] = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS); + wsptr[8*5] = (int) RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS-PASS1_BITS); + wsptr[8*7] = (int) RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS-PASS1_BITS); + wsptr[8*6] = (int) RIGHT_SHIFT(tmp26, CONST_BITS-PASS1_BITS); + } + + /* Pass 2: process 13 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 13; ctr++) { + outptr = output_buf[ctr] + output_col; + + /* Even part */ + + /* Add fudge factor here for final descale. */ + z1 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2)); + z1 <<= CONST_BITS; + + z2 = (INT32) wsptr[2]; + z3 = (INT32) wsptr[4]; + z4 = (INT32) wsptr[6]; + + tmp10 = z3 + z4; + tmp11 = z3 - z4; + + tmp12 = MULTIPLY(tmp10, FIX(1.155388986)); /* (c4+c6)/2 */ + tmp13 = MULTIPLY(tmp11, FIX(0.096834934)) + z1; /* (c4-c6)/2 */ + + tmp20 = MULTIPLY(z2, FIX(1.373119086)) + tmp12 + tmp13; /* c2 */ + tmp22 = MULTIPLY(z2, FIX(0.501487041)) - tmp12 + tmp13; /* c10 */ + + tmp12 = MULTIPLY(tmp10, FIX(0.316450131)); /* (c8-c12)/2 */ + tmp13 = MULTIPLY(tmp11, FIX(0.486914739)) + z1; /* (c8+c12)/2 */ + + tmp21 = MULTIPLY(z2, FIX(1.058554052)) - tmp12 + tmp13; /* c6 */ + tmp25 = MULTIPLY(z2, - FIX(1.252223920)) + tmp12 + tmp13; /* c4 */ + + tmp12 = MULTIPLY(tmp10, FIX(0.435816023)); /* (c2-c10)/2 */ + tmp13 = MULTIPLY(tmp11, FIX(0.937303064)) - z1; /* (c2+c10)/2 */ + + tmp23 = MULTIPLY(z2, - FIX(0.170464608)) - tmp12 - tmp13; /* c12 */ + tmp24 = MULTIPLY(z2, - FIX(0.803364869)) + tmp12 - tmp13; /* c8 */ + + tmp26 = MULTIPLY(tmp11 - z2, FIX(1.414213562)) + z1; /* c0 */ + + /* Odd part */ + + z1 = (INT32) wsptr[1]; + z2 = (INT32) wsptr[3]; + z3 = (INT32) wsptr[5]; + z4 = (INT32) wsptr[7]; + + tmp11 = MULTIPLY(z1 + z2, FIX(1.322312651)); /* c3 */ + tmp12 = MULTIPLY(z1 + z3, FIX(1.163874945)); /* c5 */ + tmp15 = z1 + z4; + tmp13 = MULTIPLY(tmp15, FIX(0.937797057)); /* c7 */ + tmp10 = tmp11 + tmp12 + tmp13 - + MULTIPLY(z1, FIX(2.020082300)); /* c7+c5+c3-c1 */ + tmp14 = MULTIPLY(z2 + z3, - FIX(0.338443458)); /* -c11 */ + tmp11 += tmp14 + MULTIPLY(z2, FIX(0.837223564)); /* c5+c9+c11-c3 */ + tmp12 += tmp14 - MULTIPLY(z3, FIX(1.572116027)); /* c1+c5-c9-c11 */ + tmp14 = MULTIPLY(z2 + z4, - FIX(1.163874945)); /* -c5 */ + tmp11 += tmp14; + tmp13 += tmp14 + MULTIPLY(z4, FIX(2.205608352)); /* c3+c5+c9-c7 */ + tmp14 = MULTIPLY(z3 + z4, - FIX(0.657217813)); /* -c9 */ + tmp12 += tmp14; + tmp13 += tmp14; + tmp15 = MULTIPLY(tmp15, FIX(0.338443458)); /* c11 */ + tmp14 = tmp15 + MULTIPLY(z1, FIX(0.318774355)) - /* c9-c11 */ + MULTIPLY(z2, FIX(0.466105296)); /* c1-c7 */ + z1 = MULTIPLY(z3 - z2, FIX(0.937797057)); /* c7 */ + tmp14 += z1; + tmp15 += z1 + MULTIPLY(z3, FIX(0.384515595)) - /* c3-c7 */ + MULTIPLY(z4, FIX(1.742345811)); /* c1+c11 */ + + /* Final output stage */ + + outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[12] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[9] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[8] = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp15, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[7] = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp15, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp26, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += 8; /* advance pointer to next row */ + } +} + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a 14x14 output block. + * + * Optimized algorithm with 20 multiplications in the 1-D kernel. + * cK represents sqrt(2) * cos(K*pi/28). + */ + +GLOBAL(void) +jpeg_idct_14x14 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16; + INT32 tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26; + INT32 z1, z2, z3, z4; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[8*14]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) { + /* Even part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + z1 <<= CONST_BITS; + /* Add fudge factor here for final descale. */ + z1 += ONE << (CONST_BITS-PASS1_BITS-1); + z4 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]); + z2 = MULTIPLY(z4, FIX(1.274162392)); /* c4 */ + z3 = MULTIPLY(z4, FIX(0.314692123)); /* c12 */ + z4 = MULTIPLY(z4, FIX(0.881747734)); /* c8 */ + + tmp10 = z1 + z2; + tmp11 = z1 + z3; + tmp12 = z1 - z4; + + tmp23 = RIGHT_SHIFT(z1 - ((z2 + z3 - z4) << 1), /* c0 = (c4+c12-c8)*2 */ + CONST_BITS-PASS1_BITS); + + z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + z2 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]); + + z3 = MULTIPLY(z1 + z2, FIX(1.105676686)); /* c6 */ + + tmp13 = z3 + MULTIPLY(z1, FIX(0.273079590)); /* c2-c6 */ + tmp14 = z3 - MULTIPLY(z2, FIX(1.719280954)); /* c6+c10 */ + tmp15 = MULTIPLY(z1, FIX(0.613604268)) - /* c10 */ + MULTIPLY(z2, FIX(1.378756276)); /* c2 */ + + tmp20 = tmp10 + tmp13; + tmp26 = tmp10 - tmp13; + tmp21 = tmp11 + tmp14; + tmp25 = tmp11 - tmp14; + tmp22 = tmp12 + tmp15; + tmp24 = tmp12 - tmp15; + + /* Odd part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]); + tmp13 = z4 << CONST_BITS; + + tmp14 = z1 + z3; + tmp11 = MULTIPLY(z1 + z2, FIX(1.334852607)); /* c3 */ + tmp12 = MULTIPLY(tmp14, FIX(1.197448846)); /* c5 */ + tmp10 = tmp11 + tmp12 + tmp13 - MULTIPLY(z1, FIX(1.126980169)); /* c3+c5-c1 */ + tmp14 = MULTIPLY(tmp14, FIX(0.752406978)); /* c9 */ + tmp16 = tmp14 - MULTIPLY(z1, FIX(1.061150426)); /* c9+c11-c13 */ + z1 -= z2; + tmp15 = MULTIPLY(z1, FIX(0.467085129)) - tmp13; /* c11 */ + tmp16 += tmp15; + z1 += z4; + z4 = MULTIPLY(z2 + z3, - FIX(0.158341681)) - tmp13; /* -c13 */ + tmp11 += z4 - MULTIPLY(z2, FIX(0.424103948)); /* c3-c9-c13 */ + tmp12 += z4 - MULTIPLY(z3, FIX(2.373959773)); /* c3+c5-c13 */ + z4 = MULTIPLY(z3 - z2, FIX(1.405321284)); /* c1 */ + tmp14 += z4 + tmp13 - MULTIPLY(z3, FIX(1.6906431334)); /* c1+c9-c11 */ + tmp15 += z4 + MULTIPLY(z2, FIX(0.674957567)); /* c1+c11-c5 */ + + tmp13 = (z1 - z3) << PASS1_BITS; + + /* Final output stage */ + + wsptr[8*0] = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*13] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*1] = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*12] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*2] = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS); + wsptr[8*11] = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS); + wsptr[8*3] = (int) (tmp23 + tmp13); + wsptr[8*10] = (int) (tmp23 - tmp13); + wsptr[8*4] = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS); + wsptr[8*9] = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS); + wsptr[8*5] = (int) RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS-PASS1_BITS); + wsptr[8*8] = (int) RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS-PASS1_BITS); + wsptr[8*6] = (int) RIGHT_SHIFT(tmp26 + tmp16, CONST_BITS-PASS1_BITS); + wsptr[8*7] = (int) RIGHT_SHIFT(tmp26 - tmp16, CONST_BITS-PASS1_BITS); + } + + /* Pass 2: process 14 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 14; ctr++) { + outptr = output_buf[ctr] + output_col; + + /* Even part */ + + /* Add fudge factor here for final descale. */ + z1 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2)); + z1 <<= CONST_BITS; + z4 = (INT32) wsptr[4]; + z2 = MULTIPLY(z4, FIX(1.274162392)); /* c4 */ + z3 = MULTIPLY(z4, FIX(0.314692123)); /* c12 */ + z4 = MULTIPLY(z4, FIX(0.881747734)); /* c8 */ + + tmp10 = z1 + z2; + tmp11 = z1 + z3; + tmp12 = z1 - z4; + + tmp23 = z1 - ((z2 + z3 - z4) << 1); /* c0 = (c4+c12-c8)*2 */ + + z1 = (INT32) wsptr[2]; + z2 = (INT32) wsptr[6]; + + z3 = MULTIPLY(z1 + z2, FIX(1.105676686)); /* c6 */ + + tmp13 = z3 + MULTIPLY(z1, FIX(0.273079590)); /* c2-c6 */ + tmp14 = z3 - MULTIPLY(z2, FIX(1.719280954)); /* c6+c10 */ + tmp15 = MULTIPLY(z1, FIX(0.613604268)) - /* c10 */ + MULTIPLY(z2, FIX(1.378756276)); /* c2 */ + + tmp20 = tmp10 + tmp13; + tmp26 = tmp10 - tmp13; + tmp21 = tmp11 + tmp14; + tmp25 = tmp11 - tmp14; + tmp22 = tmp12 + tmp15; + tmp24 = tmp12 - tmp15; + + /* Odd part */ + + z1 = (INT32) wsptr[1]; + z2 = (INT32) wsptr[3]; + z3 = (INT32) wsptr[5]; + z4 = (INT32) wsptr[7]; + z4 <<= CONST_BITS; + + tmp14 = z1 + z3; + tmp11 = MULTIPLY(z1 + z2, FIX(1.334852607)); /* c3 */ + tmp12 = MULTIPLY(tmp14, FIX(1.197448846)); /* c5 */ + tmp10 = tmp11 + tmp12 + z4 - MULTIPLY(z1, FIX(1.126980169)); /* c3+c5-c1 */ + tmp14 = MULTIPLY(tmp14, FIX(0.752406978)); /* c9 */ + tmp16 = tmp14 - MULTIPLY(z1, FIX(1.061150426)); /* c9+c11-c13 */ + z1 -= z2; + tmp15 = MULTIPLY(z1, FIX(0.467085129)) - z4; /* c11 */ + tmp16 += tmp15; + tmp13 = MULTIPLY(z2 + z3, - FIX(0.158341681)) - z4; /* -c13 */ + tmp11 += tmp13 - MULTIPLY(z2, FIX(0.424103948)); /* c3-c9-c13 */ + tmp12 += tmp13 - MULTIPLY(z3, FIX(2.373959773)); /* c3+c5-c13 */ + tmp13 = MULTIPLY(z3 - z2, FIX(1.405321284)); /* c1 */ + tmp14 += tmp13 + z4 - MULTIPLY(z3, FIX(1.6906431334)); /* c1+c9-c11 */ + tmp15 += tmp13 + MULTIPLY(z2, FIX(0.674957567)); /* c1+c11-c5 */ + + tmp13 = ((z1 - z3) << CONST_BITS) + z4; + + /* Final output stage */ + + outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[13] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[12] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[9] = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp15, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[8] = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp15, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp26 + tmp16, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[7] = range_limit[(int) RIGHT_SHIFT(tmp26 - tmp16, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += 8; /* advance pointer to next row */ + } +} + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a 15x15 output block. + * + * Optimized algorithm with 22 multiplications in the 1-D kernel. + * cK represents sqrt(2) * cos(K*pi/30). + */ + +GLOBAL(void) +jpeg_idct_15x15 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16; + INT32 tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26, tmp27; + INT32 z1, z2, z3, z4; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[8*15]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) { + /* Even part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + z1 <<= CONST_BITS; + /* Add fudge factor here for final descale. */ + z1 += ONE << (CONST_BITS-PASS1_BITS-1); + + z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + z3 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]); + z4 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]); + + tmp10 = MULTIPLY(z4, FIX(0.437016024)); /* c12 */ + tmp11 = MULTIPLY(z4, FIX(1.144122806)); /* c6 */ + + tmp12 = z1 - tmp10; + tmp13 = z1 + tmp11; + z1 -= (tmp11 - tmp10) << 1; /* c0 = (c6-c12)*2 */ + + z4 = z2 - z3; + z3 += z2; + tmp10 = MULTIPLY(z3, FIX(1.337628990)); /* (c2+c4)/2 */ + tmp11 = MULTIPLY(z4, FIX(0.045680613)); /* (c2-c4)/2 */ + z2 = MULTIPLY(z2, FIX(1.439773946)); /* c4+c14 */ + + tmp20 = tmp13 + tmp10 + tmp11; + tmp23 = tmp12 - tmp10 + tmp11 + z2; + + tmp10 = MULTIPLY(z3, FIX(0.547059574)); /* (c8+c14)/2 */ + tmp11 = MULTIPLY(z4, FIX(0.399234004)); /* (c8-c14)/2 */ + + tmp25 = tmp13 - tmp10 - tmp11; + tmp26 = tmp12 + tmp10 - tmp11 - z2; + + tmp10 = MULTIPLY(z3, FIX(0.790569415)); /* (c6+c12)/2 */ + tmp11 = MULTIPLY(z4, FIX(0.353553391)); /* (c6-c12)/2 */ + + tmp21 = tmp12 + tmp10 + tmp11; + tmp24 = tmp13 - tmp10 + tmp11; + tmp11 += tmp11; + tmp22 = z1 + tmp11; /* c10 = c6-c12 */ + tmp27 = z1 - tmp11 - tmp11; /* c0 = (c6-c12)*2 */ + + /* Odd part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + z4 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + z3 = MULTIPLY(z4, FIX(1.224744871)); /* c5 */ + z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]); + + tmp13 = z2 - z4; + tmp15 = MULTIPLY(z1 + tmp13, FIX(0.831253876)); /* c9 */ + tmp11 = tmp15 + MULTIPLY(z1, FIX(0.513743148)); /* c3-c9 */ + tmp14 = tmp15 - MULTIPLY(tmp13, FIX(2.176250899)); /* c3+c9 */ + + tmp13 = MULTIPLY(z2, - FIX(0.831253876)); /* -c9 */ + tmp15 = MULTIPLY(z2, - FIX(1.344997024)); /* -c3 */ + z2 = z1 - z4; + tmp12 = z3 + MULTIPLY(z2, FIX(1.406466353)); /* c1 */ + + tmp10 = tmp12 + MULTIPLY(z4, FIX(2.457431844)) - tmp15; /* c1+c7 */ + tmp16 = tmp12 - MULTIPLY(z1, FIX(1.112434820)) + tmp13; /* c1-c13 */ + tmp12 = MULTIPLY(z2, FIX(1.224744871)) - z3; /* c5 */ + z2 = MULTIPLY(z1 + z4, FIX(0.575212477)); /* c11 */ + tmp13 += z2 + MULTIPLY(z1, FIX(0.475753014)) - z3; /* c7-c11 */ + tmp15 += z2 - MULTIPLY(z4, FIX(0.869244010)) + z3; /* c11+c13 */ + + /* Final output stage */ + + wsptr[8*0] = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*14] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*1] = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*13] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*2] = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS); + wsptr[8*12] = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS); + wsptr[8*3] = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS); + wsptr[8*11] = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS); + wsptr[8*4] = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS); + wsptr[8*10] = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS); + wsptr[8*5] = (int) RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS-PASS1_BITS); + wsptr[8*9] = (int) RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS-PASS1_BITS); + wsptr[8*6] = (int) RIGHT_SHIFT(tmp26 + tmp16, CONST_BITS-PASS1_BITS); + wsptr[8*8] = (int) RIGHT_SHIFT(tmp26 - tmp16, CONST_BITS-PASS1_BITS); + wsptr[8*7] = (int) RIGHT_SHIFT(tmp27, CONST_BITS-PASS1_BITS); + } + + /* Pass 2: process 15 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 15; ctr++) { + outptr = output_buf[ctr] + output_col; + + /* Even part */ + + /* Add fudge factor here for final descale. */ + z1 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2)); + z1 <<= CONST_BITS; + + z2 = (INT32) wsptr[2]; + z3 = (INT32) wsptr[4]; + z4 = (INT32) wsptr[6]; + + tmp10 = MULTIPLY(z4, FIX(0.437016024)); /* c12 */ + tmp11 = MULTIPLY(z4, FIX(1.144122806)); /* c6 */ + + tmp12 = z1 - tmp10; + tmp13 = z1 + tmp11; + z1 -= (tmp11 - tmp10) << 1; /* c0 = (c6-c12)*2 */ + + z4 = z2 - z3; + z3 += z2; + tmp10 = MULTIPLY(z3, FIX(1.337628990)); /* (c2+c4)/2 */ + tmp11 = MULTIPLY(z4, FIX(0.045680613)); /* (c2-c4)/2 */ + z2 = MULTIPLY(z2, FIX(1.439773946)); /* c4+c14 */ + + tmp20 = tmp13 + tmp10 + tmp11; + tmp23 = tmp12 - tmp10 + tmp11 + z2; + + tmp10 = MULTIPLY(z3, FIX(0.547059574)); /* (c8+c14)/2 */ + tmp11 = MULTIPLY(z4, FIX(0.399234004)); /* (c8-c14)/2 */ + + tmp25 = tmp13 - tmp10 - tmp11; + tmp26 = tmp12 + tmp10 - tmp11 - z2; + + tmp10 = MULTIPLY(z3, FIX(0.790569415)); /* (c6+c12)/2 */ + tmp11 = MULTIPLY(z4, FIX(0.353553391)); /* (c6-c12)/2 */ + + tmp21 = tmp12 + tmp10 + tmp11; + tmp24 = tmp13 - tmp10 + tmp11; + tmp11 += tmp11; + tmp22 = z1 + tmp11; /* c10 = c6-c12 */ + tmp27 = z1 - tmp11 - tmp11; /* c0 = (c6-c12)*2 */ + + /* Odd part */ + + z1 = (INT32) wsptr[1]; + z2 = (INT32) wsptr[3]; + z4 = (INT32) wsptr[5]; + z3 = MULTIPLY(z4, FIX(1.224744871)); /* c5 */ + z4 = (INT32) wsptr[7]; + + tmp13 = z2 - z4; + tmp15 = MULTIPLY(z1 + tmp13, FIX(0.831253876)); /* c9 */ + tmp11 = tmp15 + MULTIPLY(z1, FIX(0.513743148)); /* c3-c9 */ + tmp14 = tmp15 - MULTIPLY(tmp13, FIX(2.176250899)); /* c3+c9 */ + + tmp13 = MULTIPLY(z2, - FIX(0.831253876)); /* -c9 */ + tmp15 = MULTIPLY(z2, - FIX(1.344997024)); /* -c3 */ + z2 = z1 - z4; + tmp12 = z3 + MULTIPLY(z2, FIX(1.406466353)); /* c1 */ + + tmp10 = tmp12 + MULTIPLY(z4, FIX(2.457431844)) - tmp15; /* c1+c7 */ + tmp16 = tmp12 - MULTIPLY(z1, FIX(1.112434820)) + tmp13; /* c1-c13 */ + tmp12 = MULTIPLY(z2, FIX(1.224744871)) - z3; /* c5 */ + z2 = MULTIPLY(z1 + z4, FIX(0.575212477)); /* c11 */ + tmp13 += z2 + MULTIPLY(z1, FIX(0.475753014)) - z3; /* c7-c11 */ + tmp15 += z2 - MULTIPLY(z4, FIX(0.869244010)) + z3; /* c11+c13 */ + + /* Final output stage */ + + outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[14] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[13] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[12] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp15, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[9] = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp15, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp26 + tmp16, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[8] = range_limit[(int) RIGHT_SHIFT(tmp26 - tmp16, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[7] = range_limit[(int) RIGHT_SHIFT(tmp27, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += 8; /* advance pointer to next row */ + } +} + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a 16x16 output block. + * + * Optimized algorithm with 28 multiplications in the 1-D kernel. + * cK represents sqrt(2) * cos(K*pi/32). + */ + +GLOBAL(void) +jpeg_idct_16x16 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp0, tmp1, tmp2, tmp3, tmp10, tmp11, tmp12, tmp13; + INT32 tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26, tmp27; + INT32 z1, z2, z3, z4; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[8*16]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) { + /* Even part */ + + tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + tmp0 <<= CONST_BITS; + /* Add fudge factor here for final descale. */ + tmp0 += 1 << (CONST_BITS-PASS1_BITS-1); + + z1 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]); + tmp1 = MULTIPLY(z1, FIX(1.306562965)); /* c4[16] = c2[8] */ + tmp2 = MULTIPLY(z1, FIX_0_541196100); /* c12[16] = c6[8] */ + + tmp10 = tmp0 + tmp1; + tmp11 = tmp0 - tmp1; + tmp12 = tmp0 + tmp2; + tmp13 = tmp0 - tmp2; + + z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + z2 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]); + z3 = z1 - z2; + z4 = MULTIPLY(z3, FIX(0.275899379)); /* c14[16] = c7[8] */ + z3 = MULTIPLY(z3, FIX(1.387039845)); /* c2[16] = c1[8] */ + + tmp0 = z3 + MULTIPLY(z2, FIX_2_562915447); /* (c6+c2)[16] = (c3+c1)[8] */ + tmp1 = z4 + MULTIPLY(z1, FIX_0_899976223); /* (c6-c14)[16] = (c3-c7)[8] */ + tmp2 = z3 - MULTIPLY(z1, FIX(0.601344887)); /* (c2-c10)[16] = (c1-c5)[8] */ + tmp3 = z4 - MULTIPLY(z2, FIX(0.509795579)); /* (c10-c14)[16] = (c5-c7)[8] */ + + tmp20 = tmp10 + tmp0; + tmp27 = tmp10 - tmp0; + tmp21 = tmp12 + tmp1; + tmp26 = tmp12 - tmp1; + tmp22 = tmp13 + tmp2; + tmp25 = tmp13 - tmp2; + tmp23 = tmp11 + tmp3; + tmp24 = tmp11 - tmp3; + + /* Odd part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]); + + tmp11 = z1 + z3; + + tmp1 = MULTIPLY(z1 + z2, FIX(1.353318001)); /* c3 */ + tmp2 = MULTIPLY(tmp11, FIX(1.247225013)); /* c5 */ + tmp3 = MULTIPLY(z1 + z4, FIX(1.093201867)); /* c7 */ + tmp10 = MULTIPLY(z1 - z4, FIX(0.897167586)); /* c9 */ + tmp11 = MULTIPLY(tmp11, FIX(0.666655658)); /* c11 */ + tmp12 = MULTIPLY(z1 - z2, FIX(0.410524528)); /* c13 */ + tmp0 = tmp1 + tmp2 + tmp3 - + MULTIPLY(z1, FIX(2.286341144)); /* c7+c5+c3-c1 */ + tmp13 = tmp10 + tmp11 + tmp12 - + MULTIPLY(z1, FIX(1.835730603)); /* c9+c11+c13-c15 */ + z1 = MULTIPLY(z2 + z3, FIX(0.138617169)); /* c15 */ + tmp1 += z1 + MULTIPLY(z2, FIX(0.071888074)); /* c9+c11-c3-c15 */ + tmp2 += z1 - MULTIPLY(z3, FIX(1.125726048)); /* c5+c7+c15-c3 */ + z1 = MULTIPLY(z3 - z2, FIX(1.407403738)); /* c1 */ + tmp11 += z1 - MULTIPLY(z3, FIX(0.766367282)); /* c1+c11-c9-c13 */ + tmp12 += z1 + MULTIPLY(z2, FIX(1.971951411)); /* c1+c5+c13-c7 */ + z2 += z4; + z1 = MULTIPLY(z2, - FIX(0.666655658)); /* -c11 */ + tmp1 += z1; + tmp3 += z1 + MULTIPLY(z4, FIX(1.065388962)); /* c3+c11+c15-c7 */ + z2 = MULTIPLY(z2, - FIX(1.247225013)); /* -c5 */ + tmp10 += z2 + MULTIPLY(z4, FIX(3.141271809)); /* c1+c5+c9-c13 */ + tmp12 += z2; + z2 = MULTIPLY(z3 + z4, - FIX(1.353318001)); /* -c3 */ + tmp2 += z2; + tmp3 += z2; + z2 = MULTIPLY(z4 - z3, FIX(0.410524528)); /* c13 */ + tmp10 += z2; + tmp11 += z2; + + /* Final output stage */ + + wsptr[8*0] = (int) RIGHT_SHIFT(tmp20 + tmp0, CONST_BITS-PASS1_BITS); + wsptr[8*15] = (int) RIGHT_SHIFT(tmp20 - tmp0, CONST_BITS-PASS1_BITS); + wsptr[8*1] = (int) RIGHT_SHIFT(tmp21 + tmp1, CONST_BITS-PASS1_BITS); + wsptr[8*14] = (int) RIGHT_SHIFT(tmp21 - tmp1, CONST_BITS-PASS1_BITS); + wsptr[8*2] = (int) RIGHT_SHIFT(tmp22 + tmp2, CONST_BITS-PASS1_BITS); + wsptr[8*13] = (int) RIGHT_SHIFT(tmp22 - tmp2, CONST_BITS-PASS1_BITS); + wsptr[8*3] = (int) RIGHT_SHIFT(tmp23 + tmp3, CONST_BITS-PASS1_BITS); + wsptr[8*12] = (int) RIGHT_SHIFT(tmp23 - tmp3, CONST_BITS-PASS1_BITS); + wsptr[8*4] = (int) RIGHT_SHIFT(tmp24 + tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*11] = (int) RIGHT_SHIFT(tmp24 - tmp10, CONST_BITS-PASS1_BITS); + wsptr[8*5] = (int) RIGHT_SHIFT(tmp25 + tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*10] = (int) RIGHT_SHIFT(tmp25 - tmp11, CONST_BITS-PASS1_BITS); + wsptr[8*6] = (int) RIGHT_SHIFT(tmp26 + tmp12, CONST_BITS-PASS1_BITS); + wsptr[8*9] = (int) RIGHT_SHIFT(tmp26 - tmp12, CONST_BITS-PASS1_BITS); + wsptr[8*7] = (int) RIGHT_SHIFT(tmp27 + tmp13, CONST_BITS-PASS1_BITS); + wsptr[8*8] = (int) RIGHT_SHIFT(tmp27 - tmp13, CONST_BITS-PASS1_BITS); + } + + /* Pass 2: process 16 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 16; ctr++) { + outptr = output_buf[ctr] + output_col; + + /* Even part */ + + /* Add fudge factor here for final descale. */ + tmp0 = (INT32) wsptr[0] + (ONE << (PASS1_BITS+2)); + tmp0 <<= CONST_BITS; + + z1 = (INT32) wsptr[4]; + tmp1 = MULTIPLY(z1, FIX(1.306562965)); /* c4[16] = c2[8] */ + tmp2 = MULTIPLY(z1, FIX_0_541196100); /* c12[16] = c6[8] */ + + tmp10 = tmp0 + tmp1; + tmp11 = tmp0 - tmp1; + tmp12 = tmp0 + tmp2; + tmp13 = tmp0 - tmp2; + + z1 = (INT32) wsptr[2]; + z2 = (INT32) wsptr[6]; + z3 = z1 - z2; + z4 = MULTIPLY(z3, FIX(0.275899379)); /* c14[16] = c7[8] */ + z3 = MULTIPLY(z3, FIX(1.387039845)); /* c2[16] = c1[8] */ + + tmp0 = z3 + MULTIPLY(z2, FIX_2_562915447); /* (c6+c2)[16] = (c3+c1)[8] */ + tmp1 = z4 + MULTIPLY(z1, FIX_0_899976223); /* (c6-c14)[16] = (c3-c7)[8] */ + tmp2 = z3 - MULTIPLY(z1, FIX(0.601344887)); /* (c2-c10)[16] = (c1-c5)[8] */ + tmp3 = z4 - MULTIPLY(z2, FIX(0.509795579)); /* (c10-c14)[16] = (c5-c7)[8] */ + + tmp20 = tmp10 + tmp0; + tmp27 = tmp10 - tmp0; + tmp21 = tmp12 + tmp1; + tmp26 = tmp12 - tmp1; + tmp22 = tmp13 + tmp2; + tmp25 = tmp13 - tmp2; + tmp23 = tmp11 + tmp3; + tmp24 = tmp11 - tmp3; + + /* Odd part */ + + z1 = (INT32) wsptr[1]; + z2 = (INT32) wsptr[3]; + z3 = (INT32) wsptr[5]; + z4 = (INT32) wsptr[7]; + + tmp11 = z1 + z3; + + tmp1 = MULTIPLY(z1 + z2, FIX(1.353318001)); /* c3 */ + tmp2 = MULTIPLY(tmp11, FIX(1.247225013)); /* c5 */ + tmp3 = MULTIPLY(z1 + z4, FIX(1.093201867)); /* c7 */ + tmp10 = MULTIPLY(z1 - z4, FIX(0.897167586)); /* c9 */ + tmp11 = MULTIPLY(tmp11, FIX(0.666655658)); /* c11 */ + tmp12 = MULTIPLY(z1 - z2, FIX(0.410524528)); /* c13 */ + tmp0 = tmp1 + tmp2 + tmp3 - + MULTIPLY(z1, FIX(2.286341144)); /* c7+c5+c3-c1 */ + tmp13 = tmp10 + tmp11 + tmp12 - + MULTIPLY(z1, FIX(1.835730603)); /* c9+c11+c13-c15 */ + z1 = MULTIPLY(z2 + z3, FIX(0.138617169)); /* c15 */ + tmp1 += z1 + MULTIPLY(z2, FIX(0.071888074)); /* c9+c11-c3-c15 */ + tmp2 += z1 - MULTIPLY(z3, FIX(1.125726048)); /* c5+c7+c15-c3 */ + z1 = MULTIPLY(z3 - z2, FIX(1.407403738)); /* c1 */ + tmp11 += z1 - MULTIPLY(z3, FIX(0.766367282)); /* c1+c11-c9-c13 */ + tmp12 += z1 + MULTIPLY(z2, FIX(1.971951411)); /* c1+c5+c13-c7 */ + z2 += z4; + z1 = MULTIPLY(z2, - FIX(0.666655658)); /* -c11 */ + tmp1 += z1; + tmp3 += z1 + MULTIPLY(z4, FIX(1.065388962)); /* c3+c11+c15-c7 */ + z2 = MULTIPLY(z2, - FIX(1.247225013)); /* -c5 */ + tmp10 += z2 + MULTIPLY(z4, FIX(3.141271809)); /* c1+c5+c9-c13 */ + tmp12 += z2; + z2 = MULTIPLY(z3 + z4, - FIX(1.353318001)); /* -c3 */ + tmp2 += z2; + tmp3 += z2; + z2 = MULTIPLY(z4 - z3, FIX(0.410524528)); /* c13 */ + tmp10 += z2; + tmp11 += z2; + + /* Final output stage */ + + outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[15] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp0, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp1, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[14] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp1, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp2, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[13] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp2, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp3, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[12] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp3, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp10, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp11, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp26 + tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[9] = range_limit[(int) RIGHT_SHIFT(tmp26 - tmp12, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[7] = range_limit[(int) RIGHT_SHIFT(tmp27 + tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + outptr[8] = range_limit[(int) RIGHT_SHIFT(tmp27 - tmp13, + CONST_BITS+PASS1_BITS+3) + & RANGE_MASK]; + + wsptr += 8; /* advance pointer to next row */ + } +} + +#endif /* IDCT_SCALING_SUPPORTED */ +#endif /* DCT_ISLOW_SUPPORTED */ diff --git a/app/src/main/jni/libjpeg-turbo/jidctred.c b/app/src/main/jni/libjpeg-turbo/jidctred.c new file mode 100644 index 000000000..2b385f86c --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jidctred.c @@ -0,0 +1,398 @@ +/* + * jidctred.c + * + * Copyright (C) 1994-1998, Thomas G. Lane. + * This file is part of the Independent JPEG Group's software. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains inverse-DCT routines that produce reduced-size output: + * either 4x4, 2x2, or 1x1 pixels from an 8x8 DCT block. + * + * The implementation is based on the Loeffler, Ligtenberg and Moschytz (LL&M) + * algorithm used in jidctint.c. We simply replace each 8-to-8 1-D IDCT step + * with an 8-to-4 step that produces the four averages of two adjacent outputs + * (or an 8-to-2 step producing two averages of four outputs, for 2x2 output). + * These steps were derived by computing the corresponding values at the end + * of the normal LL&M code, then simplifying as much as possible. + * + * 1x1 is trivial: just take the DC coefficient divided by 8. + * + * See jidctint.c for additional comments. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jdct.h" /* Private declarations for DCT subsystem */ + +#ifdef IDCT_SCALING_SUPPORTED + + +/* + * This module is specialized to the case DCTSIZE = 8. + */ + +#if DCTSIZE != 8 + Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */ +#endif + + +/* Scaling is the same as in jidctint.c. */ + +#if BITS_IN_JSAMPLE == 8 +#define CONST_BITS 13 +#define PASS1_BITS 2 +#else +#define CONST_BITS 13 +#define PASS1_BITS 1 /* lose a little precision to avoid overflow */ +#endif + +/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus + * causing a lot of useless floating-point operations at run time. + * To get around this we use the following pre-calculated constants. + * If you change CONST_BITS you may want to add appropriate values. + * (With a reasonable C compiler, you can just rely on the FIX() macro...) + */ + +#if CONST_BITS == 13 +#define FIX_0_211164243 ((INT32) 1730) /* FIX(0.211164243) */ +#define FIX_0_509795579 ((INT32) 4176) /* FIX(0.509795579) */ +#define FIX_0_601344887 ((INT32) 4926) /* FIX(0.601344887) */ +#define FIX_0_720959822 ((INT32) 5906) /* FIX(0.720959822) */ +#define FIX_0_765366865 ((INT32) 6270) /* FIX(0.765366865) */ +#define FIX_0_850430095 ((INT32) 6967) /* FIX(0.850430095) */ +#define FIX_0_899976223 ((INT32) 7373) /* FIX(0.899976223) */ +#define FIX_1_061594337 ((INT32) 8697) /* FIX(1.061594337) */ +#define FIX_1_272758580 ((INT32) 10426) /* FIX(1.272758580) */ +#define FIX_1_451774981 ((INT32) 11893) /* FIX(1.451774981) */ +#define FIX_1_847759065 ((INT32) 15137) /* FIX(1.847759065) */ +#define FIX_2_172734803 ((INT32) 17799) /* FIX(2.172734803) */ +#define FIX_2_562915447 ((INT32) 20995) /* FIX(2.562915447) */ +#define FIX_3_624509785 ((INT32) 29692) /* FIX(3.624509785) */ +#else +#define FIX_0_211164243 FIX(0.211164243) +#define FIX_0_509795579 FIX(0.509795579) +#define FIX_0_601344887 FIX(0.601344887) +#define FIX_0_720959822 FIX(0.720959822) +#define FIX_0_765366865 FIX(0.765366865) +#define FIX_0_850430095 FIX(0.850430095) +#define FIX_0_899976223 FIX(0.899976223) +#define FIX_1_061594337 FIX(1.061594337) +#define FIX_1_272758580 FIX(1.272758580) +#define FIX_1_451774981 FIX(1.451774981) +#define FIX_1_847759065 FIX(1.847759065) +#define FIX_2_172734803 FIX(2.172734803) +#define FIX_2_562915447 FIX(2.562915447) +#define FIX_3_624509785 FIX(3.624509785) +#endif + + +/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result. + * For 8-bit samples with the recommended scaling, all the variable + * and constant values involved are no more than 16 bits wide, so a + * 16x16->32 bit multiply can be used instead of a full 32x32 multiply. + * For 12-bit samples, a full 32-bit multiplication will be needed. + */ + +#if BITS_IN_JSAMPLE == 8 +#define MULTIPLY(var,const) MULTIPLY16C16(var,const) +#else +#define MULTIPLY(var,const) ((var) * (const)) +#endif + + +/* Dequantize a coefficient by multiplying it by the multiplier-table + * entry; produce an int result. In this module, both inputs and result + * are 16 bits or less, so either int or short multiply will work. + */ + +#define DEQUANTIZE(coef,quantval) (((ISLOW_MULT_TYPE) (coef)) * (quantval)) + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a reduced-size 4x4 output block. + */ + +GLOBAL(void) +jpeg_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp0, tmp2, tmp10, tmp12; + INT32 z1, z2, z3, z4; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[DCTSIZE*4]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = DCTSIZE; ctr > 0; inptr++, quantptr++, wsptr++, ctr--) { + /* Don't bother to process column 4, because second pass won't use it */ + if (ctr == DCTSIZE-4) + continue; + if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 && + inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*5] == 0 && + inptr[DCTSIZE*6] == 0 && inptr[DCTSIZE*7] == 0) { + /* AC terms all zero; we need not examine term 4 for 4x4 output */ + int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS; + + wsptr[DCTSIZE*0] = dcval; + wsptr[DCTSIZE*1] = dcval; + wsptr[DCTSIZE*2] = dcval; + wsptr[DCTSIZE*3] = dcval; + + continue; + } + + /* Even part */ + + tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + tmp0 <<= (CONST_BITS+1); + + z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]); + z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]); + + tmp2 = MULTIPLY(z2, FIX_1_847759065) + MULTIPLY(z3, - FIX_0_765366865); + + tmp10 = tmp0 + tmp2; + tmp12 = tmp0 - tmp2; + + /* Odd part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]); + z2 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + z3 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + z4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + + tmp0 = MULTIPLY(z1, - FIX_0_211164243) /* sqrt(2) * (c3-c1) */ + + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */ + + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */ + + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */ + + tmp2 = MULTIPLY(z1, - FIX_0_509795579) /* sqrt(2) * (c7-c5) */ + + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */ + + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */ + + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */ + + /* Final output stage */ + + wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp2, CONST_BITS-PASS1_BITS+1); + wsptr[DCTSIZE*3] = (int) DESCALE(tmp10 - tmp2, CONST_BITS-PASS1_BITS+1); + wsptr[DCTSIZE*1] = (int) DESCALE(tmp12 + tmp0, CONST_BITS-PASS1_BITS+1); + wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 - tmp0, CONST_BITS-PASS1_BITS+1); + } + + /* Pass 2: process 4 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 4; ctr++) { + outptr = output_buf[ctr] + output_col; + /* It's not clear whether a zero row test is worthwhile here ... */ + +#ifndef NO_ZERO_ROW_TEST + if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 && + wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) { + /* AC terms all zero */ + JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3) + & RANGE_MASK]; + + outptr[0] = dcval; + outptr[1] = dcval; + outptr[2] = dcval; + outptr[3] = dcval; + + wsptr += DCTSIZE; /* advance pointer to next row */ + continue; + } +#endif + + /* Even part */ + + tmp0 = ((INT32) wsptr[0]) << (CONST_BITS+1); + + tmp2 = MULTIPLY((INT32) wsptr[2], FIX_1_847759065) + + MULTIPLY((INT32) wsptr[6], - FIX_0_765366865); + + tmp10 = tmp0 + tmp2; + tmp12 = tmp0 - tmp2; + + /* Odd part */ + + z1 = (INT32) wsptr[7]; + z2 = (INT32) wsptr[5]; + z3 = (INT32) wsptr[3]; + z4 = (INT32) wsptr[1]; + + tmp0 = MULTIPLY(z1, - FIX_0_211164243) /* sqrt(2) * (c3-c1) */ + + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */ + + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */ + + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */ + + tmp2 = MULTIPLY(z1, - FIX_0_509795579) /* sqrt(2) * (c7-c5) */ + + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */ + + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */ + + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */ + + /* Final output stage */ + + outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp2, + CONST_BITS+PASS1_BITS+3+1) + & RANGE_MASK]; + outptr[3] = range_limit[(int) DESCALE(tmp10 - tmp2, + CONST_BITS+PASS1_BITS+3+1) + & RANGE_MASK]; + outptr[1] = range_limit[(int) DESCALE(tmp12 + tmp0, + CONST_BITS+PASS1_BITS+3+1) + & RANGE_MASK]; + outptr[2] = range_limit[(int) DESCALE(tmp12 - tmp0, + CONST_BITS+PASS1_BITS+3+1) + & RANGE_MASK]; + + wsptr += DCTSIZE; /* advance pointer to next row */ + } +} + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a reduced-size 2x2 output block. + */ + +GLOBAL(void) +jpeg_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + INT32 tmp0, tmp10, z1; + JCOEFPTR inptr; + ISLOW_MULT_TYPE * quantptr; + int * wsptr; + JSAMPROW outptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + int ctr; + int workspace[DCTSIZE*2]; /* buffers data between passes */ + SHIFT_TEMPS + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + wsptr = workspace; + for (ctr = DCTSIZE; ctr > 0; inptr++, quantptr++, wsptr++, ctr--) { + /* Don't bother to process columns 2,4,6 */ + if (ctr == DCTSIZE-2 || ctr == DCTSIZE-4 || ctr == DCTSIZE-6) + continue; + if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*3] == 0 && + inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*7] == 0) { + /* AC terms all zero; we need not examine terms 2,4,6 for 2x2 output */ + int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS; + + wsptr[DCTSIZE*0] = dcval; + wsptr[DCTSIZE*1] = dcval; + + continue; + } + + /* Even part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]); + tmp10 = z1 << (CONST_BITS+2); + + /* Odd part */ + + z1 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]); + tmp0 = MULTIPLY(z1, - FIX_0_720959822); /* sqrt(2) * (c7-c5+c3-c1) */ + z1 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]); + tmp0 += MULTIPLY(z1, FIX_0_850430095); /* sqrt(2) * (-c1+c3+c5+c7) */ + z1 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]); + tmp0 += MULTIPLY(z1, - FIX_1_272758580); /* sqrt(2) * (-c1+c3-c5-c7) */ + z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]); + tmp0 += MULTIPLY(z1, FIX_3_624509785); /* sqrt(2) * (c1+c3+c5+c7) */ + + /* Final output stage */ + + wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp0, CONST_BITS-PASS1_BITS+2); + wsptr[DCTSIZE*1] = (int) DESCALE(tmp10 - tmp0, CONST_BITS-PASS1_BITS+2); + } + + /* Pass 2: process 2 rows from work array, store into output array. */ + + wsptr = workspace; + for (ctr = 0; ctr < 2; ctr++) { + outptr = output_buf[ctr] + output_col; + /* It's not clear whether a zero row test is worthwhile here ... */ + +#ifndef NO_ZERO_ROW_TEST + if (wsptr[1] == 0 && wsptr[3] == 0 && wsptr[5] == 0 && wsptr[7] == 0) { + /* AC terms all zero */ + JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3) + & RANGE_MASK]; + + outptr[0] = dcval; + outptr[1] = dcval; + + wsptr += DCTSIZE; /* advance pointer to next row */ + continue; + } +#endif + + /* Even part */ + + tmp10 = ((INT32) wsptr[0]) << (CONST_BITS+2); + + /* Odd part */ + + tmp0 = MULTIPLY((INT32) wsptr[7], - FIX_0_720959822) /* sqrt(2) * (c7-c5+c3-c1) */ + + MULTIPLY((INT32) wsptr[5], FIX_0_850430095) /* sqrt(2) * (-c1+c3+c5+c7) */ + + MULTIPLY((INT32) wsptr[3], - FIX_1_272758580) /* sqrt(2) * (-c1+c3-c5-c7) */ + + MULTIPLY((INT32) wsptr[1], FIX_3_624509785); /* sqrt(2) * (c1+c3+c5+c7) */ + + /* Final output stage */ + + outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp0, + CONST_BITS+PASS1_BITS+3+2) + & RANGE_MASK]; + outptr[1] = range_limit[(int) DESCALE(tmp10 - tmp0, + CONST_BITS+PASS1_BITS+3+2) + & RANGE_MASK]; + + wsptr += DCTSIZE; /* advance pointer to next row */ + } +} + + +/* + * Perform dequantization and inverse DCT on one block of coefficients, + * producing a reduced-size 1x1 output block. + */ + +GLOBAL(void) +jpeg_idct_1x1 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + int dcval; + ISLOW_MULT_TYPE * quantptr; + JSAMPLE *range_limit = IDCT_range_limit(cinfo); + SHIFT_TEMPS + + /* We hardly need an inverse DCT routine for this: just take the + * average pixel value, which is one-eighth of the DC coefficient. + */ + quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table; + dcval = DEQUANTIZE(coef_block[0], quantptr[0]); + dcval = (int) DESCALE((INT32) dcval, 3); + + output_buf[0][output_col] = range_limit[dcval & RANGE_MASK]; +} + +#endif /* IDCT_SCALING_SUPPORTED */ diff --git a/app/src/main/jni/libjpeg-turbo/jinclude.h b/app/src/main/jni/libjpeg-turbo/jinclude.h new file mode 100644 index 000000000..4dced6e2f --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jinclude.h @@ -0,0 +1,83 @@ +/* + * jinclude.h + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1994, Thomas G. Lane. + * It was modified by The libjpeg-turbo Project to include only code relevant + * to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This file exists to provide a single place to fix any problems with + * including the wrong system include files. (Common problems are taken + * care of by the standard jconfig symbols, but on really weird systems + * you may have to edit this file.) + * + * NOTE: this file is NOT intended to be included by applications using the + * JPEG library. Most applications need only include jpeglib.h. + */ + + +/* Include auto-config file to find out which system include files we need. */ + +#include "jconfig.h" /* auto configuration options */ +#define JCONFIG_INCLUDED /* so that jpeglib.h doesn't do it again */ + +/* + * We need the NULL macro and size_t typedef. + * On an ANSI-conforming system it is sufficient to include . + * Otherwise, we get them from or ; we may have to + * pull in as well. + * Note that the core JPEG library does not require ; + * only the default error handler and data source/destination modules do. + * But we must pull it in because of the references to FILE in jpeglib.h. + * You can remove those references if you want to compile without . + */ + +#ifdef HAVE_STDDEF_H +#include +#endif + +#ifdef HAVE_STDLIB_H +#include +#endif + +#ifdef NEED_SYS_TYPES_H +#include +#endif + +#include + +/* + * We need memory copying and zeroing functions, plus strncpy(). + * ANSI and System V implementations declare these in . + * BSD doesn't have the mem() functions, but it does have bcopy()/bzero(). + * Some systems may declare memset and memcpy in . + * + * NOTE: we assume the size parameters to these functions are of type size_t. + * Change the casts in these macros if not! + */ + +#ifdef NEED_BSD_STRINGS + +#include +#define MEMZERO(target,size) bzero((void *)(target), (size_t)(size)) +#define MEMCOPY(dest,src,size) bcopy((const void *)(src), (void *)(dest), (size_t)(size)) + +#else /* not BSD, assume ANSI/SysV string lib */ + +#include +#define MEMZERO(target,size) memset((void *)(target), 0, (size_t)(size)) +#define MEMCOPY(dest,src,size) memcpy((void *)(dest), (const void *)(src), (size_t)(size)) + +#endif + +/* + * The modules that use fread() and fwrite() always invoke them through + * these macros. On some systems you may need to twiddle the argument casts. + * CAUTION: argument order is different from underlying functions! + */ + +#define JFREAD(file,buf,sizeofbuf) \ + ((size_t) fread((void *) (buf), (size_t) 1, (size_t) (sizeofbuf), (file))) +#define JFWRITE(file,buf,sizeofbuf) \ + ((size_t) fwrite((const void *) (buf), (size_t) 1, (size_t) (sizeofbuf), (file))) diff --git a/app/src/main/jni/libjpeg-turbo/jmemmgr.c b/app/src/main/jni/libjpeg-turbo/jmemmgr.c new file mode 100644 index 000000000..f7219d265 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jmemmgr.c @@ -0,0 +1,1150 @@ +/* + * jmemmgr.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1997, Thomas G. Lane. + * It was modified by The libjpeg-turbo Project to include only code and + * information relevant to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains the JPEG system-independent memory management + * routines. This code is usable across a wide variety of machines; most + * of the system dependencies have been isolated in a separate file. + * The major functions provided here are: + * * pool-based allocation and freeing of memory; + * * policy decisions about how to divide available memory among the + * virtual arrays; + * * control logic for swapping virtual arrays between main memory and + * backing storage. + * The separate system-dependent file provides the actual backing-storage + * access code, and it contains the policy decision about how much total + * main memory to use. + * This file is system-dependent in the sense that some of its functions + * are unnecessary in some systems. For example, if there is enough virtual + * memory so that backing storage will never be used, much of the virtual + * array control logic could be removed. (Of course, if you have that much + * memory then you shouldn't care about a little bit of unused code...) + */ + +#define JPEG_INTERNALS +#define AM_MEMORY_MANAGER /* we define jvirt_Xarray_control structs */ +#include "jinclude.h" +#include "jpeglib.h" +#include "jmemsys.h" /* import the system-dependent declarations */ + +#ifndef NO_GETENV +#ifndef HAVE_STDLIB_H /* should declare getenv() */ +extern char * getenv (const char * name); +#endif +#endif + + +LOCAL(size_t) +round_up_pow2 (size_t a, size_t b) +/* a rounded up to the next multiple of b, i.e. ceil(a/b)*b */ +/* Assumes a >= 0, b > 0, and b is a power of 2 */ +{ + return ((a + b - 1) & (~(b - 1))); +} + + +/* + * Some important notes: + * The allocation routines provided here must never return NULL. + * They should exit to error_exit if unsuccessful. + * + * It's not a good idea to try to merge the sarray and barray routines, + * even though they are textually almost the same, because samples are + * usually stored as bytes while coefficients are shorts or ints. Thus, + * in machines where byte pointers have a different representation from + * word pointers, the resulting machine code could not be the same. + */ + + +/* + * Many machines require storage alignment: longs must start on 4-byte + * boundaries, doubles on 8-byte boundaries, etc. On such machines, malloc() + * always returns pointers that are multiples of the worst-case alignment + * requirement, and we had better do so too. + * There isn't any really portable way to determine the worst-case alignment + * requirement. This module assumes that the alignment requirement is + * multiples of ALIGN_SIZE. + * By default, we define ALIGN_SIZE as sizeof(double). This is necessary on some + * workstations (where doubles really do need 8-byte alignment) and will work + * fine on nearly everything. If your machine has lesser alignment needs, + * you can save a few bytes by making ALIGN_SIZE smaller. + * The only place I know of where this will NOT work is certain Macintosh + * 680x0 compilers that define double as a 10-byte IEEE extended float. + * Doing 10-byte alignment is counterproductive because longwords won't be + * aligned well. Put "#define ALIGN_SIZE 4" in jconfig.h if you have + * such a compiler. + */ + +#ifndef ALIGN_SIZE /* so can override from jconfig.h */ +#ifndef WITH_SIMD +#define ALIGN_SIZE sizeof(double) +#else +#define ALIGN_SIZE 16 /* Most SIMD implementations require this */ +#endif +#endif + +/* + * We allocate objects from "pools", where each pool is gotten with a single + * request to jpeg_get_small() or jpeg_get_large(). There is no per-object + * overhead within a pool, except for alignment padding. Each pool has a + * header with a link to the next pool of the same class. + * Small and large pool headers are identical. + */ + +typedef struct small_pool_struct * small_pool_ptr; + +typedef struct small_pool_struct { + small_pool_ptr next; /* next in list of pools */ + size_t bytes_used; /* how many bytes already used within pool */ + size_t bytes_left; /* bytes still available in this pool */ +} small_pool_hdr; + +typedef struct large_pool_struct * large_pool_ptr; + +typedef struct large_pool_struct { + large_pool_ptr next; /* next in list of pools */ + size_t bytes_used; /* how many bytes already used within pool */ + size_t bytes_left; /* bytes still available in this pool */ +} large_pool_hdr; + +/* + * Here is the full definition of a memory manager object. + */ + +typedef struct { + struct jpeg_memory_mgr pub; /* public fields */ + + /* Each pool identifier (lifetime class) names a linked list of pools. */ + small_pool_ptr small_list[JPOOL_NUMPOOLS]; + large_pool_ptr large_list[JPOOL_NUMPOOLS]; + + /* Since we only have one lifetime class of virtual arrays, only one + * linked list is necessary (for each datatype). Note that the virtual + * array control blocks being linked together are actually stored somewhere + * in the small-pool list. + */ + jvirt_sarray_ptr virt_sarray_list; + jvirt_barray_ptr virt_barray_list; + + /* This counts total space obtained from jpeg_get_small/large */ + size_t total_space_allocated; + + /* alloc_sarray and alloc_barray set this value for use by virtual + * array routines. + */ + JDIMENSION last_rowsperchunk; /* from most recent alloc_sarray/barray */ +} my_memory_mgr; + +typedef my_memory_mgr * my_mem_ptr; + + +/* + * The control blocks for virtual arrays. + * Note that these blocks are allocated in the "small" pool area. + * System-dependent info for the associated backing store (if any) is hidden + * inside the backing_store_info struct. + */ + +struct jvirt_sarray_control { + JSAMPARRAY mem_buffer; /* => the in-memory buffer */ + JDIMENSION rows_in_array; /* total virtual array height */ + JDIMENSION samplesperrow; /* width of array (and of memory buffer) */ + JDIMENSION maxaccess; /* max rows accessed by access_virt_sarray */ + JDIMENSION rows_in_mem; /* height of memory buffer */ + JDIMENSION rowsperchunk; /* allocation chunk size in mem_buffer */ + JDIMENSION cur_start_row; /* first logical row # in the buffer */ + JDIMENSION first_undef_row; /* row # of first uninitialized row */ + boolean pre_zero; /* pre-zero mode requested? */ + boolean dirty; /* do current buffer contents need written? */ + boolean b_s_open; /* is backing-store data valid? */ + jvirt_sarray_ptr next; /* link to next virtual sarray control block */ + backing_store_info b_s_info; /* System-dependent control info */ +}; + +struct jvirt_barray_control { + JBLOCKARRAY mem_buffer; /* => the in-memory buffer */ + JDIMENSION rows_in_array; /* total virtual array height */ + JDIMENSION blocksperrow; /* width of array (and of memory buffer) */ + JDIMENSION maxaccess; /* max rows accessed by access_virt_barray */ + JDIMENSION rows_in_mem; /* height of memory buffer */ + JDIMENSION rowsperchunk; /* allocation chunk size in mem_buffer */ + JDIMENSION cur_start_row; /* first logical row # in the buffer */ + JDIMENSION first_undef_row; /* row # of first uninitialized row */ + boolean pre_zero; /* pre-zero mode requested? */ + boolean dirty; /* do current buffer contents need written? */ + boolean b_s_open; /* is backing-store data valid? */ + jvirt_barray_ptr next; /* link to next virtual barray control block */ + backing_store_info b_s_info; /* System-dependent control info */ +}; + + +#ifdef MEM_STATS /* optional extra stuff for statistics */ + +LOCAL(void) +print_mem_stats (j_common_ptr cinfo, int pool_id) +{ + my_mem_ptr mem = (my_mem_ptr) cinfo->mem; + small_pool_ptr shdr_ptr; + large_pool_ptr lhdr_ptr; + + /* Since this is only a debugging stub, we can cheat a little by using + * fprintf directly rather than going through the trace message code. + * This is helpful because message parm array can't handle longs. + */ + fprintf(stderr, "Freeing pool %d, total space = %ld\n", + pool_id, mem->total_space_allocated); + + for (lhdr_ptr = mem->large_list[pool_id]; lhdr_ptr != NULL; + lhdr_ptr = lhdr_ptr->next) { + fprintf(stderr, " Large chunk used %ld\n", + (long) lhdr_ptr->bytes_used); + } + + for (shdr_ptr = mem->small_list[pool_id]; shdr_ptr != NULL; + shdr_ptr = shdr_ptr->next) { + fprintf(stderr, " Small chunk used %ld free %ld\n", + (long) shdr_ptr->bytes_used, + (long) shdr_ptr->bytes_left); + } +} + +#endif /* MEM_STATS */ + + +LOCAL(void) +out_of_memory (j_common_ptr cinfo, int which) +/* Report an out-of-memory error and stop execution */ +/* If we compiled MEM_STATS support, report alloc requests before dying */ +{ +#ifdef MEM_STATS + cinfo->err->trace_level = 2; /* force self_destruct to report stats */ +#endif + ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, which); +} + + +/* + * Allocation of "small" objects. + * + * For these, we use pooled storage. When a new pool must be created, + * we try to get enough space for the current request plus a "slop" factor, + * where the slop will be the amount of leftover space in the new pool. + * The speed vs. space tradeoff is largely determined by the slop values. + * A different slop value is provided for each pool class (lifetime), + * and we also distinguish the first pool of a class from later ones. + * NOTE: the values given work fairly well on both 16- and 32-bit-int + * machines, but may be too small if longs are 64 bits or more. + * + * Since we do not know what alignment malloc() gives us, we have to + * allocate ALIGN_SIZE-1 extra space per pool to have room for alignment + * adjustment. + */ + +static const size_t first_pool_slop[JPOOL_NUMPOOLS] = +{ + 1600, /* first PERMANENT pool */ + 16000 /* first IMAGE pool */ +}; + +static const size_t extra_pool_slop[JPOOL_NUMPOOLS] = +{ + 0, /* additional PERMANENT pools */ + 5000 /* additional IMAGE pools */ +}; + +#define MIN_SLOP 50 /* greater than 0 to avoid futile looping */ + + +METHODDEF(void *) +alloc_small (j_common_ptr cinfo, int pool_id, size_t sizeofobject) +/* Allocate a "small" object */ +{ + my_mem_ptr mem = (my_mem_ptr) cinfo->mem; + small_pool_ptr hdr_ptr, prev_hdr_ptr; + char * data_ptr; + size_t min_request, slop; + + /* + * Round up the requested size to a multiple of ALIGN_SIZE in order + * to assure alignment for the next object allocated in the same pool + * and so that algorithms can straddle outside the proper area up + * to the next alignment. + */ + sizeofobject = round_up_pow2(sizeofobject, ALIGN_SIZE); + + /* Check for unsatisfiable request (do now to ensure no overflow below) */ + if ((sizeof(small_pool_hdr) + sizeofobject + ALIGN_SIZE - 1) > MAX_ALLOC_CHUNK) + out_of_memory(cinfo, 1); /* request exceeds malloc's ability */ + + /* See if space is available in any existing pool */ + if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS) + ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */ + prev_hdr_ptr = NULL; + hdr_ptr = mem->small_list[pool_id]; + while (hdr_ptr != NULL) { + if (hdr_ptr->bytes_left >= sizeofobject) + break; /* found pool with enough space */ + prev_hdr_ptr = hdr_ptr; + hdr_ptr = hdr_ptr->next; + } + + /* Time to make a new pool? */ + if (hdr_ptr == NULL) { + /* min_request is what we need now, slop is what will be leftover */ + min_request = sizeof(small_pool_hdr) + sizeofobject + ALIGN_SIZE - 1; + if (prev_hdr_ptr == NULL) /* first pool in class? */ + slop = first_pool_slop[pool_id]; + else + slop = extra_pool_slop[pool_id]; + /* Don't ask for more than MAX_ALLOC_CHUNK */ + if (slop > (size_t) (MAX_ALLOC_CHUNK-min_request)) + slop = (size_t) (MAX_ALLOC_CHUNK-min_request); + /* Try to get space, if fail reduce slop and try again */ + for (;;) { + hdr_ptr = (small_pool_ptr) jpeg_get_small(cinfo, min_request + slop); + if (hdr_ptr != NULL) + break; + slop /= 2; + if (slop < MIN_SLOP) /* give up when it gets real small */ + out_of_memory(cinfo, 2); /* jpeg_get_small failed */ + } + mem->total_space_allocated += min_request + slop; + /* Success, initialize the new pool header and add to end of list */ + hdr_ptr->next = NULL; + hdr_ptr->bytes_used = 0; + hdr_ptr->bytes_left = sizeofobject + slop; + if (prev_hdr_ptr == NULL) /* first pool in class? */ + mem->small_list[pool_id] = hdr_ptr; + else + prev_hdr_ptr->next = hdr_ptr; + } + + /* OK, allocate the object from the current pool */ + data_ptr = (char *) hdr_ptr; /* point to first data byte in pool... */ + data_ptr += sizeof(small_pool_hdr); /* ...by skipping the header... */ + if ((size_t)data_ptr % ALIGN_SIZE) /* ...and adjust for alignment */ + data_ptr += ALIGN_SIZE - (size_t)data_ptr % ALIGN_SIZE; + data_ptr += hdr_ptr->bytes_used; /* point to place for object */ + hdr_ptr->bytes_used += sizeofobject; + hdr_ptr->bytes_left -= sizeofobject; + + return (void *) data_ptr; +} + + +/* + * Allocation of "large" objects. + * + * The external semantics of these are the same as "small" objects. However, + * the pool management heuristics are quite different. We assume that each + * request is large enough that it may as well be passed directly to + * jpeg_get_large; the pool management just links everything together + * so that we can free it all on demand. + * Note: the major use of "large" objects is in JSAMPARRAY and JBLOCKARRAY + * structures. The routines that create these structures (see below) + * deliberately bunch rows together to ensure a large request size. + */ + +METHODDEF(void *) +alloc_large (j_common_ptr cinfo, int pool_id, size_t sizeofobject) +/* Allocate a "large" object */ +{ + my_mem_ptr mem = (my_mem_ptr) cinfo->mem; + large_pool_ptr hdr_ptr; + char * data_ptr; + + /* + * Round up the requested size to a multiple of ALIGN_SIZE so that + * algorithms can straddle outside the proper area up to the next + * alignment. + */ + sizeofobject = round_up_pow2(sizeofobject, ALIGN_SIZE); + + /* Check for unsatisfiable request (do now to ensure no overflow below) */ + if ((sizeof(large_pool_hdr) + sizeofobject + ALIGN_SIZE - 1) > MAX_ALLOC_CHUNK) + out_of_memory(cinfo, 3); /* request exceeds malloc's ability */ + + /* Always make a new pool */ + if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS) + ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */ + + hdr_ptr = (large_pool_ptr) jpeg_get_large(cinfo, sizeofobject + + sizeof(large_pool_hdr) + + ALIGN_SIZE - 1); + if (hdr_ptr == NULL) + out_of_memory(cinfo, 4); /* jpeg_get_large failed */ + mem->total_space_allocated += sizeofobject + sizeof(large_pool_hdr) + ALIGN_SIZE - 1; + + /* Success, initialize the new pool header and add to list */ + hdr_ptr->next = mem->large_list[pool_id]; + /* We maintain space counts in each pool header for statistical purposes, + * even though they are not needed for allocation. + */ + hdr_ptr->bytes_used = sizeofobject; + hdr_ptr->bytes_left = 0; + mem->large_list[pool_id] = hdr_ptr; + + data_ptr = (char *) hdr_ptr; /* point to first data byte in pool... */ + data_ptr += sizeof(small_pool_hdr); /* ...by skipping the header... */ + if ((size_t)data_ptr % ALIGN_SIZE) /* ...and adjust for alignment */ + data_ptr += ALIGN_SIZE - (size_t)data_ptr % ALIGN_SIZE; + + return (void *) data_ptr; +} + + +/* + * Creation of 2-D sample arrays. + * + * To minimize allocation overhead and to allow I/O of large contiguous + * blocks, we allocate the sample rows in groups of as many rows as possible + * without exceeding MAX_ALLOC_CHUNK total bytes per allocation request. + * NB: the virtual array control routines, later in this file, know about + * this chunking of rows. The rowsperchunk value is left in the mem manager + * object so that it can be saved away if this sarray is the workspace for + * a virtual array. + * + * Since we are often upsampling with a factor 2, we align the size (not + * the start) to 2 * ALIGN_SIZE so that the upsampling routines don't have + * to be as careful about size. + */ + +METHODDEF(JSAMPARRAY) +alloc_sarray (j_common_ptr cinfo, int pool_id, + JDIMENSION samplesperrow, JDIMENSION numrows) +/* Allocate a 2-D sample array */ +{ + my_mem_ptr mem = (my_mem_ptr) cinfo->mem; + JSAMPARRAY result; + JSAMPROW workspace; + JDIMENSION rowsperchunk, currow, i; + long ltemp; + + /* Make sure each row is properly aligned */ + if ((ALIGN_SIZE % sizeof(JSAMPLE)) != 0) + out_of_memory(cinfo, 5); /* safety check */ + samplesperrow = (JDIMENSION)round_up_pow2(samplesperrow, (2 * ALIGN_SIZE) / sizeof(JSAMPLE)); + + /* Calculate max # of rows allowed in one allocation chunk */ + ltemp = (MAX_ALLOC_CHUNK-sizeof(large_pool_hdr)) / + ((long) samplesperrow * sizeof(JSAMPLE)); + if (ltemp <= 0) + ERREXIT(cinfo, JERR_WIDTH_OVERFLOW); + if (ltemp < (long) numrows) + rowsperchunk = (JDIMENSION) ltemp; + else + rowsperchunk = numrows; + mem->last_rowsperchunk = rowsperchunk; + + /* Get space for row pointers (small object) */ + result = (JSAMPARRAY) alloc_small(cinfo, pool_id, + (size_t) (numrows * sizeof(JSAMPROW))); + + /* Get the rows themselves (large objects) */ + currow = 0; + while (currow < numrows) { + rowsperchunk = MIN(rowsperchunk, numrows - currow); + workspace = (JSAMPROW) alloc_large(cinfo, pool_id, + (size_t) ((size_t) rowsperchunk * (size_t) samplesperrow + * sizeof(JSAMPLE))); + for (i = rowsperchunk; i > 0; i--) { + result[currow++] = workspace; + workspace += samplesperrow; + } + } + + return result; +} + + +/* + * Creation of 2-D coefficient-block arrays. + * This is essentially the same as the code for sample arrays, above. + */ + +METHODDEF(JBLOCKARRAY) +alloc_barray (j_common_ptr cinfo, int pool_id, + JDIMENSION blocksperrow, JDIMENSION numrows) +/* Allocate a 2-D coefficient-block array */ +{ + my_mem_ptr mem = (my_mem_ptr) cinfo->mem; + JBLOCKARRAY result; + JBLOCKROW workspace; + JDIMENSION rowsperchunk, currow, i; + long ltemp; + + /* Make sure each row is properly aligned */ + if ((sizeof(JBLOCK) % ALIGN_SIZE) != 0) + out_of_memory(cinfo, 6); /* safety check */ + + /* Calculate max # of rows allowed in one allocation chunk */ + ltemp = (MAX_ALLOC_CHUNK-sizeof(large_pool_hdr)) / + ((long) blocksperrow * sizeof(JBLOCK)); + if (ltemp <= 0) + ERREXIT(cinfo, JERR_WIDTH_OVERFLOW); + if (ltemp < (long) numrows) + rowsperchunk = (JDIMENSION) ltemp; + else + rowsperchunk = numrows; + mem->last_rowsperchunk = rowsperchunk; + + /* Get space for row pointers (small object) */ + result = (JBLOCKARRAY) alloc_small(cinfo, pool_id, + (size_t) (numrows * sizeof(JBLOCKROW))); + + /* Get the rows themselves (large objects) */ + currow = 0; + while (currow < numrows) { + rowsperchunk = MIN(rowsperchunk, numrows - currow); + workspace = (JBLOCKROW) alloc_large(cinfo, pool_id, + (size_t) ((size_t) rowsperchunk * (size_t) blocksperrow + * sizeof(JBLOCK))); + for (i = rowsperchunk; i > 0; i--) { + result[currow++] = workspace; + workspace += blocksperrow; + } + } + + return result; +} + + +/* + * About virtual array management: + * + * The above "normal" array routines are only used to allocate strip buffers + * (as wide as the image, but just a few rows high). Full-image-sized buffers + * are handled as "virtual" arrays. The array is still accessed a strip at a + * time, but the memory manager must save the whole array for repeated + * accesses. The intended implementation is that there is a strip buffer in + * memory (as high as is possible given the desired memory limit), plus a + * backing file that holds the rest of the array. + * + * The request_virt_array routines are told the total size of the image and + * the maximum number of rows that will be accessed at once. The in-memory + * buffer must be at least as large as the maxaccess value. + * + * The request routines create control blocks but not the in-memory buffers. + * That is postponed until realize_virt_arrays is called. At that time the + * total amount of space needed is known (approximately, anyway), so free + * memory can be divided up fairly. + * + * The access_virt_array routines are responsible for making a specific strip + * area accessible (after reading or writing the backing file, if necessary). + * Note that the access routines are told whether the caller intends to modify + * the accessed strip; during a read-only pass this saves having to rewrite + * data to disk. The access routines are also responsible for pre-zeroing + * any newly accessed rows, if pre-zeroing was requested. + * + * In current usage, the access requests are usually for nonoverlapping + * strips; that is, successive access start_row numbers differ by exactly + * num_rows = maxaccess. This means we can get good performance with simple + * buffer dump/reload logic, by making the in-memory buffer be a multiple + * of the access height; then there will never be accesses across bufferload + * boundaries. The code will still work with overlapping access requests, + * but it doesn't handle bufferload overlaps very efficiently. + */ + + +METHODDEF(jvirt_sarray_ptr) +request_virt_sarray (j_common_ptr cinfo, int pool_id, boolean pre_zero, + JDIMENSION samplesperrow, JDIMENSION numrows, + JDIMENSION maxaccess) +/* Request a virtual 2-D sample array */ +{ + my_mem_ptr mem = (my_mem_ptr) cinfo->mem; + jvirt_sarray_ptr result; + + /* Only IMAGE-lifetime virtual arrays are currently supported */ + if (pool_id != JPOOL_IMAGE) + ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */ + + /* get control block */ + result = (jvirt_sarray_ptr) alloc_small(cinfo, pool_id, + sizeof(struct jvirt_sarray_control)); + + result->mem_buffer = NULL; /* marks array not yet realized */ + result->rows_in_array = numrows; + result->samplesperrow = samplesperrow; + result->maxaccess = maxaccess; + result->pre_zero = pre_zero; + result->b_s_open = FALSE; /* no associated backing-store object */ + result->next = mem->virt_sarray_list; /* add to list of virtual arrays */ + mem->virt_sarray_list = result; + + return result; +} + + +METHODDEF(jvirt_barray_ptr) +request_virt_barray (j_common_ptr cinfo, int pool_id, boolean pre_zero, + JDIMENSION blocksperrow, JDIMENSION numrows, + JDIMENSION maxaccess) +/* Request a virtual 2-D coefficient-block array */ +{ + my_mem_ptr mem = (my_mem_ptr) cinfo->mem; + jvirt_barray_ptr result; + + /* Only IMAGE-lifetime virtual arrays are currently supported */ + if (pool_id != JPOOL_IMAGE) + ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */ + + /* get control block */ + result = (jvirt_barray_ptr) alloc_small(cinfo, pool_id, + sizeof(struct jvirt_barray_control)); + + result->mem_buffer = NULL; /* marks array not yet realized */ + result->rows_in_array = numrows; + result->blocksperrow = blocksperrow; + result->maxaccess = maxaccess; + result->pre_zero = pre_zero; + result->b_s_open = FALSE; /* no associated backing-store object */ + result->next = mem->virt_barray_list; /* add to list of virtual arrays */ + mem->virt_barray_list = result; + + return result; +} + + +METHODDEF(void) +realize_virt_arrays (j_common_ptr cinfo) +/* Allocate the in-memory buffers for any unrealized virtual arrays */ +{ + my_mem_ptr mem = (my_mem_ptr) cinfo->mem; + size_t space_per_minheight, maximum_space, avail_mem; + size_t minheights, max_minheights; + jvirt_sarray_ptr sptr; + jvirt_barray_ptr bptr; + + /* Compute the minimum space needed (maxaccess rows in each buffer) + * and the maximum space needed (full image height in each buffer). + * These may be of use to the system-dependent jpeg_mem_available routine. + */ + space_per_minheight = 0; + maximum_space = 0; + for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) { + if (sptr->mem_buffer == NULL) { /* if not realized yet */ + space_per_minheight += (long) sptr->maxaccess * + (long) sptr->samplesperrow * sizeof(JSAMPLE); + maximum_space += (long) sptr->rows_in_array * + (long) sptr->samplesperrow * sizeof(JSAMPLE); + } + } + for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) { + if (bptr->mem_buffer == NULL) { /* if not realized yet */ + space_per_minheight += (long) bptr->maxaccess * + (long) bptr->blocksperrow * sizeof(JBLOCK); + maximum_space += (long) bptr->rows_in_array * + (long) bptr->blocksperrow * sizeof(JBLOCK); + } + } + + if (space_per_minheight <= 0) + return; /* no unrealized arrays, no work */ + + /* Determine amount of memory to actually use; this is system-dependent. */ + avail_mem = jpeg_mem_available(cinfo, space_per_minheight, maximum_space, + mem->total_space_allocated); + + /* If the maximum space needed is available, make all the buffers full + * height; otherwise parcel it out with the same number of minheights + * in each buffer. + */ + if (avail_mem >= maximum_space) + max_minheights = 1000000000L; + else { + max_minheights = avail_mem / space_per_minheight; + /* If there doesn't seem to be enough space, try to get the minimum + * anyway. This allows a "stub" implementation of jpeg_mem_available(). + */ + if (max_minheights <= 0) + max_minheights = 1; + } + + /* Allocate the in-memory buffers and initialize backing store as needed. */ + + for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) { + if (sptr->mem_buffer == NULL) { /* if not realized yet */ + minheights = ((long) sptr->rows_in_array - 1L) / sptr->maxaccess + 1L; + if (minheights <= max_minheights) { + /* This buffer fits in memory */ + sptr->rows_in_mem = sptr->rows_in_array; + } else { + /* It doesn't fit in memory, create backing store. */ + sptr->rows_in_mem = (JDIMENSION) (max_minheights * sptr->maxaccess); + jpeg_open_backing_store(cinfo, & sptr->b_s_info, + (long) sptr->rows_in_array * + (long) sptr->samplesperrow * + (long) sizeof(JSAMPLE)); + sptr->b_s_open = TRUE; + } + sptr->mem_buffer = alloc_sarray(cinfo, JPOOL_IMAGE, + sptr->samplesperrow, sptr->rows_in_mem); + sptr->rowsperchunk = mem->last_rowsperchunk; + sptr->cur_start_row = 0; + sptr->first_undef_row = 0; + sptr->dirty = FALSE; + } + } + + for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) { + if (bptr->mem_buffer == NULL) { /* if not realized yet */ + minheights = ((long) bptr->rows_in_array - 1L) / bptr->maxaccess + 1L; + if (minheights <= max_minheights) { + /* This buffer fits in memory */ + bptr->rows_in_mem = bptr->rows_in_array; + } else { + /* It doesn't fit in memory, create backing store. */ + bptr->rows_in_mem = (JDIMENSION) (max_minheights * bptr->maxaccess); + jpeg_open_backing_store(cinfo, & bptr->b_s_info, + (long) bptr->rows_in_array * + (long) bptr->blocksperrow * + (long) sizeof(JBLOCK)); + bptr->b_s_open = TRUE; + } + bptr->mem_buffer = alloc_barray(cinfo, JPOOL_IMAGE, + bptr->blocksperrow, bptr->rows_in_mem); + bptr->rowsperchunk = mem->last_rowsperchunk; + bptr->cur_start_row = 0; + bptr->first_undef_row = 0; + bptr->dirty = FALSE; + } + } +} + + +LOCAL(void) +do_sarray_io (j_common_ptr cinfo, jvirt_sarray_ptr ptr, boolean writing) +/* Do backing store read or write of a virtual sample array */ +{ + long bytesperrow, file_offset, byte_count, rows, thisrow, i; + + bytesperrow = (long) ptr->samplesperrow * sizeof(JSAMPLE); + file_offset = ptr->cur_start_row * bytesperrow; + /* Loop to read or write each allocation chunk in mem_buffer */ + for (i = 0; i < (long) ptr->rows_in_mem; i += ptr->rowsperchunk) { + /* One chunk, but check for short chunk at end of buffer */ + rows = MIN((long) ptr->rowsperchunk, (long) ptr->rows_in_mem - i); + /* Transfer no more than is currently defined */ + thisrow = (long) ptr->cur_start_row + i; + rows = MIN(rows, (long) ptr->first_undef_row - thisrow); + /* Transfer no more than fits in file */ + rows = MIN(rows, (long) ptr->rows_in_array - thisrow); + if (rows <= 0) /* this chunk might be past end of file! */ + break; + byte_count = rows * bytesperrow; + if (writing) + (*ptr->b_s_info.write_backing_store) (cinfo, & ptr->b_s_info, + (void *) ptr->mem_buffer[i], + file_offset, byte_count); + else + (*ptr->b_s_info.read_backing_store) (cinfo, & ptr->b_s_info, + (void *) ptr->mem_buffer[i], + file_offset, byte_count); + file_offset += byte_count; + } +} + + +LOCAL(void) +do_barray_io (j_common_ptr cinfo, jvirt_barray_ptr ptr, boolean writing) +/* Do backing store read or write of a virtual coefficient-block array */ +{ + long bytesperrow, file_offset, byte_count, rows, thisrow, i; + + bytesperrow = (long) ptr->blocksperrow * sizeof(JBLOCK); + file_offset = ptr->cur_start_row * bytesperrow; + /* Loop to read or write each allocation chunk in mem_buffer */ + for (i = 0; i < (long) ptr->rows_in_mem; i += ptr->rowsperchunk) { + /* One chunk, but check for short chunk at end of buffer */ + rows = MIN((long) ptr->rowsperchunk, (long) ptr->rows_in_mem - i); + /* Transfer no more than is currently defined */ + thisrow = (long) ptr->cur_start_row + i; + rows = MIN(rows, (long) ptr->first_undef_row - thisrow); + /* Transfer no more than fits in file */ + rows = MIN(rows, (long) ptr->rows_in_array - thisrow); + if (rows <= 0) /* this chunk might be past end of file! */ + break; + byte_count = rows * bytesperrow; + if (writing) + (*ptr->b_s_info.write_backing_store) (cinfo, & ptr->b_s_info, + (void *) ptr->mem_buffer[i], + file_offset, byte_count); + else + (*ptr->b_s_info.read_backing_store) (cinfo, & ptr->b_s_info, + (void *) ptr->mem_buffer[i], + file_offset, byte_count); + file_offset += byte_count; + } +} + + +METHODDEF(JSAMPARRAY) +access_virt_sarray (j_common_ptr cinfo, jvirt_sarray_ptr ptr, + JDIMENSION start_row, JDIMENSION num_rows, + boolean writable) +/* Access the part of a virtual sample array starting at start_row */ +/* and extending for num_rows rows. writable is true if */ +/* caller intends to modify the accessed area. */ +{ + JDIMENSION end_row = start_row + num_rows; + JDIMENSION undef_row; + + /* debugging check */ + if (end_row > ptr->rows_in_array || num_rows > ptr->maxaccess || + ptr->mem_buffer == NULL) + ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS); + + /* Make the desired part of the virtual array accessible */ + if (start_row < ptr->cur_start_row || + end_row > ptr->cur_start_row+ptr->rows_in_mem) { + if (! ptr->b_s_open) + ERREXIT(cinfo, JERR_VIRTUAL_BUG); + /* Flush old buffer contents if necessary */ + if (ptr->dirty) { + do_sarray_io(cinfo, ptr, TRUE); + ptr->dirty = FALSE; + } + /* Decide what part of virtual array to access. + * Algorithm: if target address > current window, assume forward scan, + * load starting at target address. If target address < current window, + * assume backward scan, load so that target area is top of window. + * Note that when switching from forward write to forward read, will have + * start_row = 0, so the limiting case applies and we load from 0 anyway. + */ + if (start_row > ptr->cur_start_row) { + ptr->cur_start_row = start_row; + } else { + /* use long arithmetic here to avoid overflow & unsigned problems */ + long ltemp; + + ltemp = (long) end_row - (long) ptr->rows_in_mem; + if (ltemp < 0) + ltemp = 0; /* don't fall off front end of file */ + ptr->cur_start_row = (JDIMENSION) ltemp; + } + /* Read in the selected part of the array. + * During the initial write pass, we will do no actual read + * because the selected part is all undefined. + */ + do_sarray_io(cinfo, ptr, FALSE); + } + /* Ensure the accessed part of the array is defined; prezero if needed. + * To improve locality of access, we only prezero the part of the array + * that the caller is about to access, not the entire in-memory array. + */ + if (ptr->first_undef_row < end_row) { + if (ptr->first_undef_row < start_row) { + if (writable) /* writer skipped over a section of array */ + ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS); + undef_row = start_row; /* but reader is allowed to read ahead */ + } else { + undef_row = ptr->first_undef_row; + } + if (writable) + ptr->first_undef_row = end_row; + if (ptr->pre_zero) { + size_t bytesperrow = (size_t) ptr->samplesperrow * sizeof(JSAMPLE); + undef_row -= ptr->cur_start_row; /* make indexes relative to buffer */ + end_row -= ptr->cur_start_row; + while (undef_row < end_row) { + jzero_far((void *) ptr->mem_buffer[undef_row], bytesperrow); + undef_row++; + } + } else { + if (! writable) /* reader looking at undefined data */ + ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS); + } + } + /* Flag the buffer dirty if caller will write in it */ + if (writable) + ptr->dirty = TRUE; + /* Return address of proper part of the buffer */ + return ptr->mem_buffer + (start_row - ptr->cur_start_row); +} + + +METHODDEF(JBLOCKARRAY) +access_virt_barray (j_common_ptr cinfo, jvirt_barray_ptr ptr, + JDIMENSION start_row, JDIMENSION num_rows, + boolean writable) +/* Access the part of a virtual block array starting at start_row */ +/* and extending for num_rows rows. writable is true if */ +/* caller intends to modify the accessed area. */ +{ + JDIMENSION end_row = start_row + num_rows; + JDIMENSION undef_row; + + /* debugging check */ + if (end_row > ptr->rows_in_array || num_rows > ptr->maxaccess || + ptr->mem_buffer == NULL) + ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS); + + /* Make the desired part of the virtual array accessible */ + if (start_row < ptr->cur_start_row || + end_row > ptr->cur_start_row+ptr->rows_in_mem) { + if (! ptr->b_s_open) + ERREXIT(cinfo, JERR_VIRTUAL_BUG); + /* Flush old buffer contents if necessary */ + if (ptr->dirty) { + do_barray_io(cinfo, ptr, TRUE); + ptr->dirty = FALSE; + } + /* Decide what part of virtual array to access. + * Algorithm: if target address > current window, assume forward scan, + * load starting at target address. If target address < current window, + * assume backward scan, load so that target area is top of window. + * Note that when switching from forward write to forward read, will have + * start_row = 0, so the limiting case applies and we load from 0 anyway. + */ + if (start_row > ptr->cur_start_row) { + ptr->cur_start_row = start_row; + } else { + /* use long arithmetic here to avoid overflow & unsigned problems */ + long ltemp; + + ltemp = (long) end_row - (long) ptr->rows_in_mem; + if (ltemp < 0) + ltemp = 0; /* don't fall off front end of file */ + ptr->cur_start_row = (JDIMENSION) ltemp; + } + /* Read in the selected part of the array. + * During the initial write pass, we will do no actual read + * because the selected part is all undefined. + */ + do_barray_io(cinfo, ptr, FALSE); + } + /* Ensure the accessed part of the array is defined; prezero if needed. + * To improve locality of access, we only prezero the part of the array + * that the caller is about to access, not the entire in-memory array. + */ + if (ptr->first_undef_row < end_row) { + if (ptr->first_undef_row < start_row) { + if (writable) /* writer skipped over a section of array */ + ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS); + undef_row = start_row; /* but reader is allowed to read ahead */ + } else { + undef_row = ptr->first_undef_row; + } + if (writable) + ptr->first_undef_row = end_row; + if (ptr->pre_zero) { + size_t bytesperrow = (size_t) ptr->blocksperrow * sizeof(JBLOCK); + undef_row -= ptr->cur_start_row; /* make indexes relative to buffer */ + end_row -= ptr->cur_start_row; + while (undef_row < end_row) { + jzero_far((void *) ptr->mem_buffer[undef_row], bytesperrow); + undef_row++; + } + } else { + if (! writable) /* reader looking at undefined data */ + ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS); + } + } + /* Flag the buffer dirty if caller will write in it */ + if (writable) + ptr->dirty = TRUE; + /* Return address of proper part of the buffer */ + return ptr->mem_buffer + (start_row - ptr->cur_start_row); +} + + +/* + * Release all objects belonging to a specified pool. + */ + +METHODDEF(void) +free_pool (j_common_ptr cinfo, int pool_id) +{ + my_mem_ptr mem = (my_mem_ptr) cinfo->mem; + small_pool_ptr shdr_ptr; + large_pool_ptr lhdr_ptr; + size_t space_freed; + + if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS) + ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */ + +#ifdef MEM_STATS + if (cinfo->err->trace_level > 1) + print_mem_stats(cinfo, pool_id); /* print pool's memory usage statistics */ +#endif + + /* If freeing IMAGE pool, close any virtual arrays first */ + if (pool_id == JPOOL_IMAGE) { + jvirt_sarray_ptr sptr; + jvirt_barray_ptr bptr; + + for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) { + if (sptr->b_s_open) { /* there may be no backing store */ + sptr->b_s_open = FALSE; /* prevent recursive close if error */ + (*sptr->b_s_info.close_backing_store) (cinfo, & sptr->b_s_info); + } + } + mem->virt_sarray_list = NULL; + for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) { + if (bptr->b_s_open) { /* there may be no backing store */ + bptr->b_s_open = FALSE; /* prevent recursive close if error */ + (*bptr->b_s_info.close_backing_store) (cinfo, & bptr->b_s_info); + } + } + mem->virt_barray_list = NULL; + } + + /* Release large objects */ + lhdr_ptr = mem->large_list[pool_id]; + mem->large_list[pool_id] = NULL; + + while (lhdr_ptr != NULL) { + large_pool_ptr next_lhdr_ptr = lhdr_ptr->next; + space_freed = lhdr_ptr->bytes_used + + lhdr_ptr->bytes_left + + sizeof(large_pool_hdr); + jpeg_free_large(cinfo, (void *) lhdr_ptr, space_freed); + mem->total_space_allocated -= space_freed; + lhdr_ptr = next_lhdr_ptr; + } + + /* Release small objects */ + shdr_ptr = mem->small_list[pool_id]; + mem->small_list[pool_id] = NULL; + + while (shdr_ptr != NULL) { + small_pool_ptr next_shdr_ptr = shdr_ptr->next; + space_freed = shdr_ptr->bytes_used + + shdr_ptr->bytes_left + + sizeof(small_pool_hdr); + jpeg_free_small(cinfo, (void *) shdr_ptr, space_freed); + mem->total_space_allocated -= space_freed; + shdr_ptr = next_shdr_ptr; + } +} + + +/* + * Close up shop entirely. + * Note that this cannot be called unless cinfo->mem is non-NULL. + */ + +METHODDEF(void) +self_destruct (j_common_ptr cinfo) +{ + int pool; + + /* Close all backing store, release all memory. + * Releasing pools in reverse order might help avoid fragmentation + * with some (brain-damaged) malloc libraries. + */ + for (pool = JPOOL_NUMPOOLS-1; pool >= JPOOL_PERMANENT; pool--) { + free_pool(cinfo, pool); + } + + /* Release the memory manager control block too. */ + jpeg_free_small(cinfo, (void *) cinfo->mem, sizeof(my_memory_mgr)); + cinfo->mem = NULL; /* ensures I will be called only once */ + + jpeg_mem_term(cinfo); /* system-dependent cleanup */ +} + + +/* + * Memory manager initialization. + * When this is called, only the error manager pointer is valid in cinfo! + */ + +GLOBAL(void) +jinit_memory_mgr (j_common_ptr cinfo) +{ + my_mem_ptr mem; + long max_to_use; + int pool; + size_t test_mac; + + cinfo->mem = NULL; /* for safety if init fails */ + + /* Check for configuration errors. + * sizeof(ALIGN_TYPE) should be a power of 2; otherwise, it probably + * doesn't reflect any real hardware alignment requirement. + * The test is a little tricky: for X>0, X and X-1 have no one-bits + * in common if and only if X is a power of 2, ie has only one one-bit. + * Some compilers may give an "unreachable code" warning here; ignore it. + */ + if ((ALIGN_SIZE & (ALIGN_SIZE-1)) != 0) + ERREXIT(cinfo, JERR_BAD_ALIGN_TYPE); + /* MAX_ALLOC_CHUNK must be representable as type size_t, and must be + * a multiple of ALIGN_SIZE. + * Again, an "unreachable code" warning may be ignored here. + * But a "constant too large" warning means you need to fix MAX_ALLOC_CHUNK. + */ + test_mac = (size_t) MAX_ALLOC_CHUNK; + if ((long) test_mac != MAX_ALLOC_CHUNK || + (MAX_ALLOC_CHUNK % ALIGN_SIZE) != 0) + ERREXIT(cinfo, JERR_BAD_ALLOC_CHUNK); + + max_to_use = jpeg_mem_init(cinfo); /* system-dependent initialization */ + + /* Attempt to allocate memory manager's control block */ + mem = (my_mem_ptr) jpeg_get_small(cinfo, sizeof(my_memory_mgr)); + + if (mem == NULL) { + jpeg_mem_term(cinfo); /* system-dependent cleanup */ + ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 0); + } + + /* OK, fill in the method pointers */ + mem->pub.alloc_small = alloc_small; + mem->pub.alloc_large = alloc_large; + mem->pub.alloc_sarray = alloc_sarray; + mem->pub.alloc_barray = alloc_barray; + mem->pub.request_virt_sarray = request_virt_sarray; + mem->pub.request_virt_barray = request_virt_barray; + mem->pub.realize_virt_arrays = realize_virt_arrays; + mem->pub.access_virt_sarray = access_virt_sarray; + mem->pub.access_virt_barray = access_virt_barray; + mem->pub.free_pool = free_pool; + mem->pub.self_destruct = self_destruct; + + /* Make MAX_ALLOC_CHUNK accessible to other modules */ + mem->pub.max_alloc_chunk = MAX_ALLOC_CHUNK; + + /* Initialize working state */ + mem->pub.max_memory_to_use = max_to_use; + + for (pool = JPOOL_NUMPOOLS-1; pool >= JPOOL_PERMANENT; pool--) { + mem->small_list[pool] = NULL; + mem->large_list[pool] = NULL; + } + mem->virt_sarray_list = NULL; + mem->virt_barray_list = NULL; + + mem->total_space_allocated = sizeof(my_memory_mgr); + + /* Declare ourselves open for business */ + cinfo->mem = & mem->pub; + + /* Check for an environment variable JPEGMEM; if found, override the + * default max_memory setting from jpeg_mem_init. Note that the + * surrounding application may again override this value. + * If your system doesn't support getenv(), define NO_GETENV to disable + * this feature. + */ +#ifndef NO_GETENV + { char * memenv; + + if ((memenv = getenv("JPEGMEM")) != NULL) { + char ch = 'x'; + + if (sscanf(memenv, "%ld%c", &max_to_use, &ch) > 0) { + if (ch == 'm' || ch == 'M') + max_to_use *= 1000L; + mem->pub.max_memory_to_use = max_to_use * 1000L; + } + } + } +#endif + +} diff --git a/app/src/main/jni/libjpeg-turbo/jmemnobs.c b/app/src/main/jni/libjpeg-turbo/jmemnobs.c new file mode 100644 index 000000000..628283276 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jmemnobs.c @@ -0,0 +1,108 @@ +/* + * jmemnobs.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1992-1996, Thomas G. Lane. + * It was modified by The libjpeg-turbo Project to include only code and + * information relevant to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This file provides a really simple implementation of the system- + * dependent portion of the JPEG memory manager. This implementation + * assumes that no backing-store files are needed: all required space + * can be obtained from malloc(). + * This is very portable in the sense that it'll compile on almost anything, + * but you'd better have lots of main memory (or virtual memory) if you want + * to process big images. + * Note that the max_memory_to_use option is ignored by this implementation. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jmemsys.h" /* import the system-dependent declarations */ + +#ifndef HAVE_STDLIB_H /* should declare malloc(),free() */ +extern void * malloc (size_t size); +extern void free (void *ptr); +#endif + + +/* + * Memory allocation and freeing are controlled by the regular library + * routines malloc() and free(). + */ + +GLOBAL(void *) +jpeg_get_small (j_common_ptr cinfo, size_t sizeofobject) +{ + return (void *) malloc(sizeofobject); +} + +GLOBAL(void) +jpeg_free_small (j_common_ptr cinfo, void * object, size_t sizeofobject) +{ + free(object); +} + + +/* + * "Large" objects are treated the same as "small" ones. + */ + +GLOBAL(void *) +jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject) +{ + return (void *) malloc(sizeofobject); +} + +GLOBAL(void) +jpeg_free_large (j_common_ptr cinfo, void * object, size_t sizeofobject) +{ + free(object); +} + + +/* + * This routine computes the total memory space available for allocation. + * Here we always say, "we got all you want bud!" + */ + +GLOBAL(size_t) +jpeg_mem_available (j_common_ptr cinfo, size_t min_bytes_needed, + size_t max_bytes_needed, size_t already_allocated) +{ + return max_bytes_needed; +} + + +/* + * Backing store (temporary file) management. + * Since jpeg_mem_available always promised the moon, + * this should never be called and we can just error out. + */ + +GLOBAL(void) +jpeg_open_backing_store (j_common_ptr cinfo, backing_store_ptr info, + long total_bytes_needed) +{ + ERREXIT(cinfo, JERR_NO_BACKING_STORE); +} + + +/* + * These routines take care of any system-dependent initialization and + * cleanup required. Here, there isn't any. + */ + +GLOBAL(long) +jpeg_mem_init (j_common_ptr cinfo) +{ + return 0; /* just set max_memory_to_use to 0 */ +} + +GLOBAL(void) +jpeg_mem_term (j_common_ptr cinfo) +{ + /* no work */ +} diff --git a/app/src/main/jni/libjpeg-turbo/jmemsys.h b/app/src/main/jni/libjpeg-turbo/jmemsys.h new file mode 100644 index 000000000..5026c7c78 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jmemsys.h @@ -0,0 +1,177 @@ +/* + * jmemsys.h + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1992-1997, Thomas G. Lane. + * It was modified by The libjpeg-turbo Project to include only code and + * information relevant to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This include file defines the interface between the system-independent + * and system-dependent portions of the JPEG memory manager. No other + * modules need include it. (The system-independent portion is jmemmgr.c; + * there are several different versions of the system-dependent portion.) + * + * This file works as-is for the system-dependent memory managers supplied + * in the IJG distribution. You may need to modify it if you write a + * custom memory manager. If system-dependent changes are needed in + * this file, the best method is to #ifdef them based on a configuration + * symbol supplied in jconfig.h. + */ + + +/* + * These two functions are used to allocate and release small chunks of + * memory. (Typically the total amount requested through jpeg_get_small is + * no more than 20K or so; this will be requested in chunks of a few K each.) + * Behavior should be the same as for the standard library functions malloc + * and free; in particular, jpeg_get_small must return NULL on failure. + * On most systems, these ARE malloc and free. jpeg_free_small is passed the + * size of the object being freed, just in case it's needed. + */ + +EXTERN(void *) jpeg_get_small (j_common_ptr cinfo, size_t sizeofobject); +EXTERN(void) jpeg_free_small (j_common_ptr cinfo, void * object, + size_t sizeofobject); + +/* + * These two functions are used to allocate and release large chunks of + * memory (up to the total free space designated by jpeg_mem_available). + * These are identical to the jpeg_get/free_small routines; but we keep them + * separate anyway, in case a different allocation strategy is desirable for + * large chunks. + */ + +EXTERN(void *) jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject); +EXTERN(void) jpeg_free_large (j_common_ptr cinfo, void * object, + size_t sizeofobject); + +/* + * The macro MAX_ALLOC_CHUNK designates the maximum number of bytes that may + * be requested in a single call to jpeg_get_large (and jpeg_get_small for that + * matter, but that case should never come into play). This macro was needed + * to model the 64Kb-segment-size limit of far addressing on 80x86 machines. + * On machines with flat address spaces, any large constant may be used. + * + * NB: jmemmgr.c expects that MAX_ALLOC_CHUNK will be representable as type + * size_t and will be a multiple of sizeof(align_type). + */ + +#ifndef MAX_ALLOC_CHUNK /* may be overridden in jconfig.h */ +#define MAX_ALLOC_CHUNK 1000000000L +#endif + +/* + * This routine computes the total space still available for allocation by + * jpeg_get_large. If more space than this is needed, backing store will be + * used. NOTE: any memory already allocated must not be counted. + * + * There is a minimum space requirement, corresponding to the minimum + * feasible buffer sizes; jmemmgr.c will request that much space even if + * jpeg_mem_available returns zero. The maximum space needed, enough to hold + * all working storage in memory, is also passed in case it is useful. + * Finally, the total space already allocated is passed. If no better + * method is available, cinfo->mem->max_memory_to_use - already_allocated + * is often a suitable calculation. + * + * It is OK for jpeg_mem_available to underestimate the space available + * (that'll just lead to more backing-store access than is really necessary). + * However, an overestimate will lead to failure. Hence it's wise to subtract + * a slop factor from the true available space. 5% should be enough. + * + * On machines with lots of virtual memory, any large constant may be returned. + * Conversely, zero may be returned to always use the minimum amount of memory. + */ + +EXTERN(size_t) jpeg_mem_available (j_common_ptr cinfo, size_t min_bytes_needed, + size_t max_bytes_needed, + size_t already_allocated); + + +/* + * This structure holds whatever state is needed to access a single + * backing-store object. The read/write/close method pointers are called + * by jmemmgr.c to manipulate the backing-store object; all other fields + * are private to the system-dependent backing store routines. + */ + +#define TEMP_NAME_LENGTH 64 /* max length of a temporary file's name */ + + +#ifdef USE_MSDOS_MEMMGR /* DOS-specific junk */ + +typedef unsigned short XMSH; /* type of extended-memory handles */ +typedef unsigned short EMSH; /* type of expanded-memory handles */ + +typedef union { + short file_handle; /* DOS file handle if it's a temp file */ + XMSH xms_handle; /* handle if it's a chunk of XMS */ + EMSH ems_handle; /* handle if it's a chunk of EMS */ +} handle_union; + +#endif /* USE_MSDOS_MEMMGR */ + +#ifdef USE_MAC_MEMMGR /* Mac-specific junk */ +#include +#endif /* USE_MAC_MEMMGR */ + + +typedef struct backing_store_struct * backing_store_ptr; + +typedef struct backing_store_struct { + /* Methods for reading/writing/closing this backing-store object */ + void (*read_backing_store) (j_common_ptr cinfo, backing_store_ptr info, + void * buffer_address, long file_offset, + long byte_count); + void (*write_backing_store) (j_common_ptr cinfo, backing_store_ptr info, + void * buffer_address, long file_offset, + long byte_count); + void (*close_backing_store) (j_common_ptr cinfo, backing_store_ptr info); + + /* Private fields for system-dependent backing-store management */ +#ifdef USE_MSDOS_MEMMGR + /* For the MS-DOS manager (jmemdos.c), we need: */ + handle_union handle; /* reference to backing-store storage object */ + char temp_name[TEMP_NAME_LENGTH]; /* name if it's a file */ +#else +#ifdef USE_MAC_MEMMGR + /* For the Mac manager (jmemmac.c), we need: */ + short temp_file; /* file reference number to temp file */ + FSSpec tempSpec; /* the FSSpec for the temp file */ + char temp_name[TEMP_NAME_LENGTH]; /* name if it's a file */ +#else + /* For a typical implementation with temp files, we need: */ + FILE * temp_file; /* stdio reference to temp file */ + char temp_name[TEMP_NAME_LENGTH]; /* name of temp file */ +#endif +#endif +} backing_store_info; + + +/* + * Initial opening of a backing-store object. This must fill in the + * read/write/close pointers in the object. The read/write routines + * may take an error exit if the specified maximum file size is exceeded. + * (If jpeg_mem_available always returns a large value, this routine can + * just take an error exit.) + */ + +EXTERN(void) jpeg_open_backing_store (j_common_ptr cinfo, + backing_store_ptr info, + long total_bytes_needed); + + +/* + * These routines take care of any system-dependent initialization and + * cleanup required. jpeg_mem_init will be called before anything is + * allocated (and, therefore, nothing in cinfo is of use except the error + * manager pointer). It should return a suitable default value for + * max_memory_to_use; this may subsequently be overridden by the surrounding + * application. (Note that max_memory_to_use is only important if + * jpeg_mem_available chooses to consult it ... no one else will.) + * jpeg_mem_term may assume that all requested memory has been freed and that + * all opened backing-store objects have been closed. + */ + +EXTERN(long) jpeg_mem_init (j_common_ptr cinfo); +EXTERN(void) jpeg_mem_term (j_common_ptr cinfo); diff --git a/app/src/main/jni/libjpeg-turbo/jmorecfg.h b/app/src/main/jni/libjpeg-turbo/jmorecfg.h new file mode 100644 index 000000000..108e7de98 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jmorecfg.h @@ -0,0 +1,388 @@ +/* + * jmorecfg.h + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1997, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright (C) 2009, 2011, 2014, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains additional configuration options that customize the + * JPEG software for special applications or support machine-dependent + * optimizations. Most users will not need to touch this file. + */ + + +/* + * Maximum number of components (color channels) allowed in JPEG image. + * To meet the letter of the JPEG spec, set this to 255. However, darn + * few applications need more than 4 channels (maybe 5 for CMYK + alpha + * mask). We recommend 10 as a reasonable compromise; use 4 if you are + * really short on memory. (Each allowed component costs a hundred or so + * bytes of storage, whether actually used in an image or not.) + */ + +#define MAX_COMPONENTS 10 /* maximum number of image components */ + + +/* + * Basic data types. + * You may need to change these if you have a machine with unusual data + * type sizes; for example, "char" not 8 bits, "short" not 16 bits, + * or "long" not 32 bits. We don't care whether "int" is 16 or 32 bits, + * but it had better be at least 16. + */ + +/* Representation of a single sample (pixel element value). + * We frequently allocate large arrays of these, so it's important to keep + * them small. But if you have memory to burn and access to char or short + * arrays is very slow on your hardware, you might want to change these. + */ + +#if BITS_IN_JSAMPLE == 8 +/* JSAMPLE should be the smallest type that will hold the values 0..255. + * You can use a signed char by having GETJSAMPLE mask it with 0xFF. + */ + +#ifdef HAVE_UNSIGNED_CHAR + +typedef unsigned char JSAMPLE; +#define GETJSAMPLE(value) ((int) (value)) + +#else /* not HAVE_UNSIGNED_CHAR */ + +typedef char JSAMPLE; +#ifdef __CHAR_UNSIGNED__ +#define GETJSAMPLE(value) ((int) (value)) +#else +#define GETJSAMPLE(value) ((int) (value) & 0xFF) +#endif /* __CHAR_UNSIGNED__ */ + +#endif /* HAVE_UNSIGNED_CHAR */ + +#define MAXJSAMPLE 255 +#define CENTERJSAMPLE 128 + +#endif /* BITS_IN_JSAMPLE == 8 */ + + +#if BITS_IN_JSAMPLE == 12 +/* JSAMPLE should be the smallest type that will hold the values 0..4095. + * On nearly all machines "short" will do nicely. + */ + +typedef short JSAMPLE; +#define GETJSAMPLE(value) ((int) (value)) + +#define MAXJSAMPLE 4095 +#define CENTERJSAMPLE 2048 + +#endif /* BITS_IN_JSAMPLE == 12 */ + + +/* Representation of a DCT frequency coefficient. + * This should be a signed value of at least 16 bits; "short" is usually OK. + * Again, we allocate large arrays of these, but you can change to int + * if you have memory to burn and "short" is really slow. + */ + +typedef short JCOEF; + + +/* Compressed datastreams are represented as arrays of JOCTET. + * These must be EXACTLY 8 bits wide, at least once they are written to + * external storage. Note that when using the stdio data source/destination + * managers, this is also the data type passed to fread/fwrite. + */ + +#ifdef HAVE_UNSIGNED_CHAR + +typedef unsigned char JOCTET; +#define GETJOCTET(value) (value) + +#else /* not HAVE_UNSIGNED_CHAR */ + +typedef char JOCTET; +#ifdef __CHAR_UNSIGNED__ +#define GETJOCTET(value) (value) +#else +#define GETJOCTET(value) ((value) & 0xFF) +#endif /* __CHAR_UNSIGNED__ */ + +#endif /* HAVE_UNSIGNED_CHAR */ + + +/* These typedefs are used for various table entries and so forth. + * They must be at least as wide as specified; but making them too big + * won't cost a huge amount of memory, so we don't provide special + * extraction code like we did for JSAMPLE. (In other words, these + * typedefs live at a different point on the speed/space tradeoff curve.) + */ + +/* UINT8 must hold at least the values 0..255. */ + +#ifdef HAVE_UNSIGNED_CHAR +typedef unsigned char UINT8; +#else /* not HAVE_UNSIGNED_CHAR */ +#ifdef __CHAR_UNSIGNED__ +typedef char UINT8; +#else /* not __CHAR_UNSIGNED__ */ +typedef short UINT8; +#endif /* __CHAR_UNSIGNED__ */ +#endif /* HAVE_UNSIGNED_CHAR */ + +/* UINT16 must hold at least the values 0..65535. */ + +#ifdef HAVE_UNSIGNED_SHORT +typedef unsigned short UINT16; +#else /* not HAVE_UNSIGNED_SHORT */ +typedef unsigned int UINT16; +#endif /* HAVE_UNSIGNED_SHORT */ + +/* INT16 must hold at least the values -32768..32767. */ + +#ifndef XMD_H /* X11/xmd.h correctly defines INT16 */ +typedef short INT16; +#endif + +/* INT32 must hold at least signed 32-bit values. */ + +#ifndef XMD_H /* X11/xmd.h correctly defines INT32 */ +typedef long INT32; +#endif + +/* Datatype used for image dimensions. The JPEG standard only supports + * images up to 64K*64K due to 16-bit fields in SOF markers. Therefore + * "unsigned int" is sufficient on all machines. However, if you need to + * handle larger images and you don't mind deviating from the spec, you + * can change this datatype. + */ + +typedef unsigned int JDIMENSION; + +#define JPEG_MAX_DIMENSION 65500L /* a tad under 64K to prevent overflows */ + + +/* These macros are used in all function definitions and extern declarations. + * You could modify them if you need to change function linkage conventions; + * in particular, you'll need to do that to make the library a Windows DLL. + * Another application is to make all functions global for use with debuggers + * or code profilers that require it. + */ + +/* a function called through method pointers: */ +#define METHODDEF(type) static type +/* a function used only in its module: */ +#define LOCAL(type) static type +/* a function referenced thru EXTERNs: */ +#define GLOBAL(type) type +/* a reference to a GLOBAL function: */ +#define EXTERN(type) extern type + + +/* Originally, this macro was used as a way of defining function prototypes + * for both modern compilers as well as older compilers that did not support + * prototype parameters. libjpeg-turbo has never supported these older, + * non-ANSI compilers, but the macro is still included because there is some + * software out there that uses it. + */ + +#define JMETHOD(type,methodname,arglist) type (*methodname) arglist + + +/* libjpeg-turbo no longer supports platforms that have far symbols (MS-DOS), + * but again, some software relies on this macro. + */ + +#undef FAR +#define FAR + + +/* + * On a few systems, type boolean and/or its values FALSE, TRUE may appear + * in standard header files. Or you may have conflicts with application- + * specific header files that you want to include together with these files. + * Defining HAVE_BOOLEAN before including jpeglib.h should make it work. + */ + +#ifndef HAVE_BOOLEAN +typedef int boolean; +#endif +#ifndef FALSE /* in case these macros already exist */ +#define FALSE 0 /* values of boolean */ +#endif +#ifndef TRUE +#define TRUE 1 +#endif + + +/* + * The remaining options affect code selection within the JPEG library, + * but they don't need to be visible to most applications using the library. + * To minimize application namespace pollution, the symbols won't be + * defined unless JPEG_INTERNALS or JPEG_INTERNAL_OPTIONS has been defined. + */ + +#ifdef JPEG_INTERNALS +#define JPEG_INTERNAL_OPTIONS +#endif + +#ifdef JPEG_INTERNAL_OPTIONS + + +/* + * These defines indicate whether to include various optional functions. + * Undefining some of these symbols will produce a smaller but less capable + * library. Note that you can leave certain source files out of the + * compilation/linking process if you've #undef'd the corresponding symbols. + * (You may HAVE to do that if your compiler doesn't like null source files.) + */ + +/* Capability options common to encoder and decoder: */ + +#define DCT_ISLOW_SUPPORTED /* slow but accurate integer algorithm */ +#define DCT_IFAST_SUPPORTED /* faster, less accurate integer method */ +#define DCT_FLOAT_SUPPORTED /* floating-point: accurate, fast on fast HW */ + +/* Encoder capability options: */ + +#define C_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */ +#define C_PROGRESSIVE_SUPPORTED /* Progressive JPEG? (Requires MULTISCAN)*/ +#define ENTROPY_OPT_SUPPORTED /* Optimization of entropy coding parms? */ +/* Note: if you selected 12-bit data precision, it is dangerous to turn off + * ENTROPY_OPT_SUPPORTED. The standard Huffman tables are only good for 8-bit + * precision, so jchuff.c normally uses entropy optimization to compute + * usable tables for higher precision. If you don't want to do optimization, + * you'll have to supply different default Huffman tables. + * The exact same statements apply for progressive JPEG: the default tables + * don't work for progressive mode. (This may get fixed, however.) + */ +#define INPUT_SMOOTHING_SUPPORTED /* Input image smoothing option? */ + +/* Decoder capability options: */ + +#define D_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */ +#define D_PROGRESSIVE_SUPPORTED /* Progressive JPEG? (Requires MULTISCAN)*/ +#define SAVE_MARKERS_SUPPORTED /* jpeg_save_markers() needed? */ +#define BLOCK_SMOOTHING_SUPPORTED /* Block smoothing? (Progressive only) */ +#define IDCT_SCALING_SUPPORTED /* Output rescaling via IDCT? */ +#undef UPSAMPLE_SCALING_SUPPORTED /* Output rescaling at upsample stage? */ +#define UPSAMPLE_MERGING_SUPPORTED /* Fast path for sloppy upsampling? */ +#define QUANT_1PASS_SUPPORTED /* 1-pass color quantization? */ +#define QUANT_2PASS_SUPPORTED /* 2-pass color quantization? */ + +/* more capability options later, no doubt */ + + +/* + * The RGB_RED, RGB_GREEN, RGB_BLUE, and RGB_PIXELSIZE macros are a vestigial + * feature of libjpeg. The idea was that, if an application developer needed + * to compress from/decompress to a BGR/BGRX/RGBX/XBGR/XRGB buffer, they could + * change these macros, rebuild libjpeg, and link their application statically + * with it. In reality, few people ever did this, because there were some + * severe restrictions involved (cjpeg and djpeg no longer worked properly, + * compressing/decompressing RGB JPEGs no longer worked properly, and the color + * quantizer wouldn't work with pixel sizes other than 3.) Further, since all + * of the O/S-supplied versions of libjpeg were built with the default values + * of RGB_RED, RGB_GREEN, RGB_BLUE, and RGB_PIXELSIZE, many applications have + * come to regard these values as immutable. + * + * The libjpeg-turbo colorspace extensions provide a much cleaner way of + * compressing from/decompressing to buffers with arbitrary component orders + * and pixel sizes. Thus, we do not support changing the values of RGB_RED, + * RGB_GREEN, RGB_BLUE, or RGB_PIXELSIZE. In addition to the restrictions + * listed above, changing these values will also break the SIMD extensions and + * the regression tests. + */ + +#define RGB_RED 0 /* Offset of Red in an RGB scanline element */ +#define RGB_GREEN 1 /* Offset of Green */ +#define RGB_BLUE 2 /* Offset of Blue */ +#define RGB_PIXELSIZE 3 /* JSAMPLEs per RGB scanline element */ + +#define JPEG_NUMCS 17 + +#define EXT_RGB_RED 0 +#define EXT_RGB_GREEN 1 +#define EXT_RGB_BLUE 2 +#define EXT_RGB_PIXELSIZE 3 + +#define EXT_RGBX_RED 0 +#define EXT_RGBX_GREEN 1 +#define EXT_RGBX_BLUE 2 +#define EXT_RGBX_PIXELSIZE 4 + +#define EXT_BGR_RED 2 +#define EXT_BGR_GREEN 1 +#define EXT_BGR_BLUE 0 +#define EXT_BGR_PIXELSIZE 3 + +#define EXT_BGRX_RED 2 +#define EXT_BGRX_GREEN 1 +#define EXT_BGRX_BLUE 0 +#define EXT_BGRX_PIXELSIZE 4 + +#define EXT_XBGR_RED 3 +#define EXT_XBGR_GREEN 2 +#define EXT_XBGR_BLUE 1 +#define EXT_XBGR_PIXELSIZE 4 + +#define EXT_XRGB_RED 1 +#define EXT_XRGB_GREEN 2 +#define EXT_XRGB_BLUE 3 +#define EXT_XRGB_PIXELSIZE 4 + +static const int rgb_red[JPEG_NUMCS] = { + -1, -1, RGB_RED, -1, -1, -1, EXT_RGB_RED, EXT_RGBX_RED, + EXT_BGR_RED, EXT_BGRX_RED, EXT_XBGR_RED, EXT_XRGB_RED, + EXT_RGBX_RED, EXT_BGRX_RED, EXT_XBGR_RED, EXT_XRGB_RED, + -1 +}; + +static const int rgb_green[JPEG_NUMCS] = { + -1, -1, RGB_GREEN, -1, -1, -1, EXT_RGB_GREEN, EXT_RGBX_GREEN, + EXT_BGR_GREEN, EXT_BGRX_GREEN, EXT_XBGR_GREEN, EXT_XRGB_GREEN, + EXT_RGBX_GREEN, EXT_BGRX_GREEN, EXT_XBGR_GREEN, EXT_XRGB_GREEN, + -1 +}; + +static const int rgb_blue[JPEG_NUMCS] = { + -1, -1, RGB_BLUE, -1, -1, -1, EXT_RGB_BLUE, EXT_RGBX_BLUE, + EXT_BGR_BLUE, EXT_BGRX_BLUE, EXT_XBGR_BLUE, EXT_XRGB_BLUE, + EXT_RGBX_BLUE, EXT_BGRX_BLUE, EXT_XBGR_BLUE, EXT_XRGB_BLUE, + -1 +}; + +static const int rgb_pixelsize[JPEG_NUMCS] = { + -1, -1, RGB_PIXELSIZE, -1, -1, -1, EXT_RGB_PIXELSIZE, EXT_RGBX_PIXELSIZE, + EXT_BGR_PIXELSIZE, EXT_BGRX_PIXELSIZE, EXT_XBGR_PIXELSIZE, EXT_XRGB_PIXELSIZE, + EXT_RGBX_PIXELSIZE, EXT_BGRX_PIXELSIZE, EXT_XBGR_PIXELSIZE, EXT_XRGB_PIXELSIZE, + -1 +}; + +/* Definitions for speed-related optimizations. */ + +/* On some machines (notably 68000 series) "int" is 32 bits, but multiplying + * two 16-bit shorts is faster than multiplying two ints. Define MULTIPLIER + * as short on such a machine. MULTIPLIER must be at least 16 bits wide. + */ + +#ifndef MULTIPLIER +#ifndef WITH_SIMD +#define MULTIPLIER int /* type for fastest integer multiply */ +#else +#define MULTIPLIER short /* prefer 16-bit with SIMD for parellelism */ +#endif +#endif + + +/* FAST_FLOAT should be either float or double, whichever is done faster + * by your compiler. (Note that this type is only used in the floating point + * DCT routines, so it only matters if you've defined DCT_FLOAT_SUPPORTED.) + */ + +#ifndef FAST_FLOAT +#define FAST_FLOAT float +#endif + +#endif /* JPEG_INTERNAL_OPTIONS */ diff --git a/app/src/main/jni/libjpeg-turbo/jpegcomp.h b/app/src/main/jni/libjpeg-turbo/jpegcomp.h new file mode 100644 index 000000000..ed9eeab8a --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jpegcomp.h @@ -0,0 +1,30 @@ +/* + * jpegcomp.h + * + * Copyright (C) 2010, D. R. Commander + * For conditions of distribution and use, see the accompanying README file. + * + * JPEG compatibility macros + * These declarations are considered internal to the JPEG library; most + * applications using the library shouldn't need to include this file. + */ + +#if JPEG_LIB_VERSION >= 70 +#define _DCT_scaled_size DCT_h_scaled_size +#define _DCT_h_scaled_size DCT_h_scaled_size +#define _DCT_v_scaled_size DCT_v_scaled_size +#define _min_DCT_scaled_size min_DCT_h_scaled_size +#define _min_DCT_h_scaled_size min_DCT_h_scaled_size +#define _min_DCT_v_scaled_size min_DCT_v_scaled_size +#define _jpeg_width jpeg_width +#define _jpeg_height jpeg_height +#else +#define _DCT_scaled_size DCT_scaled_size +#define _DCT_h_scaled_size DCT_scaled_size +#define _DCT_v_scaled_size DCT_scaled_size +#define _min_DCT_scaled_size min_DCT_scaled_size +#define _min_DCT_h_scaled_size min_DCT_scaled_size +#define _min_DCT_v_scaled_size min_DCT_scaled_size +#define _jpeg_width image_width +#define _jpeg_height image_height +#endif diff --git a/app/src/main/jni/libjpeg-turbo/jpegint.h b/app/src/main/jni/libjpeg-turbo/jpegint.h new file mode 100644 index 000000000..025accd91 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jpegint.h @@ -0,0 +1,347 @@ +/* + * jpegint.h + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1997, Thomas G. Lane. + * Modified 1997-2009 by Guido Vollbeding. + * It was modified by The libjpeg-turbo Project to include only code relevant + * to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This file provides common declarations for the various JPEG modules. + * These declarations are considered internal to the JPEG library; most + * applications using the library shouldn't need to include this file. + */ + + +/* Declarations for both compression & decompression */ + +typedef enum { /* Operating modes for buffer controllers */ + JBUF_PASS_THRU, /* Plain stripwise operation */ + /* Remaining modes require a full-image buffer to have been created */ + JBUF_SAVE_SOURCE, /* Run source subobject only, save output */ + JBUF_CRANK_DEST, /* Run dest subobject only, using saved data */ + JBUF_SAVE_AND_PASS /* Run both subobjects, save output */ +} J_BUF_MODE; + +/* Values of global_state field (jdapi.c has some dependencies on ordering!) */ +#define CSTATE_START 100 /* after create_compress */ +#define CSTATE_SCANNING 101 /* start_compress done, write_scanlines OK */ +#define CSTATE_RAW_OK 102 /* start_compress done, write_raw_data OK */ +#define CSTATE_WRCOEFS 103 /* jpeg_write_coefficients done */ +#define DSTATE_START 200 /* after create_decompress */ +#define DSTATE_INHEADER 201 /* reading header markers, no SOS yet */ +#define DSTATE_READY 202 /* found SOS, ready for start_decompress */ +#define DSTATE_PRELOAD 203 /* reading multiscan file in start_decompress*/ +#define DSTATE_PRESCAN 204 /* performing dummy pass for 2-pass quant */ +#define DSTATE_SCANNING 205 /* start_decompress done, read_scanlines OK */ +#define DSTATE_RAW_OK 206 /* start_decompress done, read_raw_data OK */ +#define DSTATE_BUFIMAGE 207 /* expecting jpeg_start_output */ +#define DSTATE_BUFPOST 208 /* looking for SOS/EOI in jpeg_finish_output */ +#define DSTATE_RDCOEFS 209 /* reading file in jpeg_read_coefficients */ +#define DSTATE_STOPPING 210 /* looking for EOI in jpeg_finish_decompress */ + + +/* Declarations for compression modules */ + +/* Master control module */ +struct jpeg_comp_master { + void (*prepare_for_pass) (j_compress_ptr cinfo); + void (*pass_startup) (j_compress_ptr cinfo); + void (*finish_pass) (j_compress_ptr cinfo); + + /* State variables made visible to other modules */ + boolean call_pass_startup; /* True if pass_startup must be called */ + boolean is_last_pass; /* True during last pass */ +}; + +/* Main buffer control (downsampled-data buffer) */ +struct jpeg_c_main_controller { + void (*start_pass) (j_compress_ptr cinfo, J_BUF_MODE pass_mode); + void (*process_data) (j_compress_ptr cinfo, JSAMPARRAY input_buf, + JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail); +}; + +/* Compression preprocessing (downsampling input buffer control) */ +struct jpeg_c_prep_controller { + void (*start_pass) (j_compress_ptr cinfo, J_BUF_MODE pass_mode); + void (*pre_process_data) (j_compress_ptr cinfo, JSAMPARRAY input_buf, + JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail, + JSAMPIMAGE output_buf, + JDIMENSION *out_row_group_ctr, + JDIMENSION out_row_groups_avail); +}; + +/* Coefficient buffer control */ +struct jpeg_c_coef_controller { + void (*start_pass) (j_compress_ptr cinfo, J_BUF_MODE pass_mode); + boolean (*compress_data) (j_compress_ptr cinfo, JSAMPIMAGE input_buf); +}; + +/* Colorspace conversion */ +struct jpeg_color_converter { + void (*start_pass) (j_compress_ptr cinfo); + void (*color_convert) (j_compress_ptr cinfo, JSAMPARRAY input_buf, + JSAMPIMAGE output_buf, JDIMENSION output_row, + int num_rows); +}; + +/* Downsampling */ +struct jpeg_downsampler { + void (*start_pass) (j_compress_ptr cinfo); + void (*downsample) (j_compress_ptr cinfo, JSAMPIMAGE input_buf, + JDIMENSION in_row_index, JSAMPIMAGE output_buf, + JDIMENSION out_row_group_index); + + boolean need_context_rows; /* TRUE if need rows above & below */ +}; + +/* Forward DCT (also controls coefficient quantization) */ +struct jpeg_forward_dct { + void (*start_pass) (j_compress_ptr cinfo); + /* perhaps this should be an array??? */ + void (*forward_DCT) (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY sample_data, JBLOCKROW coef_blocks, + JDIMENSION start_row, JDIMENSION start_col, + JDIMENSION num_blocks); +}; + +/* Entropy encoding */ +struct jpeg_entropy_encoder { + void (*start_pass) (j_compress_ptr cinfo, boolean gather_statistics); + boolean (*encode_mcu) (j_compress_ptr cinfo, JBLOCKROW *MCU_data); + void (*finish_pass) (j_compress_ptr cinfo); +}; + +/* Marker writing */ +struct jpeg_marker_writer { + void (*write_file_header) (j_compress_ptr cinfo); + void (*write_frame_header) (j_compress_ptr cinfo); + void (*write_scan_header) (j_compress_ptr cinfo); + void (*write_file_trailer) (j_compress_ptr cinfo); + void (*write_tables_only) (j_compress_ptr cinfo); + /* These routines are exported to allow insertion of extra markers */ + /* Probably only COM and APPn markers should be written this way */ + void (*write_marker_header) (j_compress_ptr cinfo, int marker, + unsigned int datalen); + void (*write_marker_byte) (j_compress_ptr cinfo, int val); +}; + + +/* Declarations for decompression modules */ + +/* Master control module */ +struct jpeg_decomp_master { + void (*prepare_for_output_pass) (j_decompress_ptr cinfo); + void (*finish_output_pass) (j_decompress_ptr cinfo); + + /* State variables made visible to other modules */ + boolean is_dummy_pass; /* True during 1st pass for 2-pass quant */ +}; + +/* Input control module */ +struct jpeg_input_controller { + int (*consume_input) (j_decompress_ptr cinfo); + void (*reset_input_controller) (j_decompress_ptr cinfo); + void (*start_input_pass) (j_decompress_ptr cinfo); + void (*finish_input_pass) (j_decompress_ptr cinfo); + + /* State variables made visible to other modules */ + boolean has_multiple_scans; /* True if file has multiple scans */ + boolean eoi_reached; /* True when EOI has been consumed */ +}; + +/* Main buffer control (downsampled-data buffer) */ +struct jpeg_d_main_controller { + void (*start_pass) (j_decompress_ptr cinfo, J_BUF_MODE pass_mode); + void (*process_data) (j_decompress_ptr cinfo, JSAMPARRAY output_buf, + JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail); +}; + +/* Coefficient buffer control */ +struct jpeg_d_coef_controller { + void (*start_input_pass) (j_decompress_ptr cinfo); + int (*consume_data) (j_decompress_ptr cinfo); + void (*start_output_pass) (j_decompress_ptr cinfo); + int (*decompress_data) (j_decompress_ptr cinfo, JSAMPIMAGE output_buf); + /* Pointer to array of coefficient virtual arrays, or NULL if none */ + jvirt_barray_ptr *coef_arrays; +}; + +/* Decompression postprocessing (color quantization buffer control) */ +struct jpeg_d_post_controller { + void (*start_pass) (j_decompress_ptr cinfo, J_BUF_MODE pass_mode); + void (*post_process_data) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, + JDIMENSION *in_row_group_ctr, + JDIMENSION in_row_groups_avail, + JSAMPARRAY output_buf, JDIMENSION *out_row_ctr, + JDIMENSION out_rows_avail); +}; + +/* Marker reading & parsing */ +struct jpeg_marker_reader { + void (*reset_marker_reader) (j_decompress_ptr cinfo); + /* Read markers until SOS or EOI. + * Returns same codes as are defined for jpeg_consume_input: + * JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI. + */ + int (*read_markers) (j_decompress_ptr cinfo); + /* Read a restart marker --- exported for use by entropy decoder only */ + jpeg_marker_parser_method read_restart_marker; + + /* State of marker reader --- nominally internal, but applications + * supplying COM or APPn handlers might like to know the state. + */ + boolean saw_SOI; /* found SOI? */ + boolean saw_SOF; /* found SOF? */ + int next_restart_num; /* next restart number expected (0-7) */ + unsigned int discarded_bytes; /* # of bytes skipped looking for a marker */ +}; + +/* Entropy decoding */ +struct jpeg_entropy_decoder { + void (*start_pass) (j_decompress_ptr cinfo); + boolean (*decode_mcu) (j_decompress_ptr cinfo, JBLOCKROW *MCU_data); + + /* This is here to share code between baseline and progressive decoders; */ + /* other modules probably should not use it */ + boolean insufficient_data; /* set TRUE after emitting warning */ +}; + +/* Inverse DCT (also performs dequantization) */ +typedef void (*inverse_DCT_method_ptr) (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, + JDIMENSION output_col); + +struct jpeg_inverse_dct { + void (*start_pass) (j_decompress_ptr cinfo); + /* It is useful to allow each component to have a separate IDCT method. */ + inverse_DCT_method_ptr inverse_DCT[MAX_COMPONENTS]; +}; + +/* Upsampling (note that upsampler must also call color converter) */ +struct jpeg_upsampler { + void (*start_pass) (j_decompress_ptr cinfo); + void (*upsample) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, + JDIMENSION *in_row_group_ctr, + JDIMENSION in_row_groups_avail, JSAMPARRAY output_buf, + JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail); + + boolean need_context_rows; /* TRUE if need rows above & below */ +}; + +/* Colorspace conversion */ +struct jpeg_color_deconverter { + void (*start_pass) (j_decompress_ptr cinfo); + void (*color_convert) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, + JDIMENSION input_row, JSAMPARRAY output_buf, + int num_rows); +}; + +/* Color quantization or color precision reduction */ +struct jpeg_color_quantizer { + void (*start_pass) (j_decompress_ptr cinfo, boolean is_pre_scan); + void (*color_quantize) (j_decompress_ptr cinfo, JSAMPARRAY input_buf, + JSAMPARRAY output_buf, int num_rows); + void (*finish_pass) (j_decompress_ptr cinfo); + void (*new_color_map) (j_decompress_ptr cinfo); +}; + + +/* Miscellaneous useful macros */ + +#undef MAX +#define MAX(a,b) ((a) > (b) ? (a) : (b)) +#undef MIN +#define MIN(a,b) ((a) < (b) ? (a) : (b)) + + +/* We assume that right shift corresponds to signed division by 2 with + * rounding towards minus infinity. This is correct for typical "arithmetic + * shift" instructions that shift in copies of the sign bit. But some + * C compilers implement >> with an unsigned shift. For these machines you + * must define RIGHT_SHIFT_IS_UNSIGNED. + * RIGHT_SHIFT provides a proper signed right shift of an INT32 quantity. + * It is only applied with constant shift counts. SHIFT_TEMPS must be + * included in the variables of any routine using RIGHT_SHIFT. + */ + +#ifdef RIGHT_SHIFT_IS_UNSIGNED +#define SHIFT_TEMPS INT32 shift_temp; +#define RIGHT_SHIFT(x,shft) \ + ((shift_temp = (x)) < 0 ? \ + (shift_temp >> (shft)) | ((~((INT32) 0)) << (32-(shft))) : \ + (shift_temp >> (shft))) +#else +#define SHIFT_TEMPS +#define RIGHT_SHIFT(x,shft) ((x) >> (shft)) +#endif + + +/* Compression module initialization routines */ +EXTERN(void) jinit_compress_master (j_compress_ptr cinfo); +EXTERN(void) jinit_c_master_control (j_compress_ptr cinfo, + boolean transcode_only); +EXTERN(void) jinit_c_main_controller (j_compress_ptr cinfo, + boolean need_full_buffer); +EXTERN(void) jinit_c_prep_controller (j_compress_ptr cinfo, + boolean need_full_buffer); +EXTERN(void) jinit_c_coef_controller (j_compress_ptr cinfo, + boolean need_full_buffer); +EXTERN(void) jinit_color_converter (j_compress_ptr cinfo); +EXTERN(void) jinit_downsampler (j_compress_ptr cinfo); +EXTERN(void) jinit_forward_dct (j_compress_ptr cinfo); +EXTERN(void) jinit_huff_encoder (j_compress_ptr cinfo); +EXTERN(void) jinit_phuff_encoder (j_compress_ptr cinfo); +EXTERN(void) jinit_arith_encoder (j_compress_ptr cinfo); +EXTERN(void) jinit_marker_writer (j_compress_ptr cinfo); +/* Decompression module initialization routines */ +EXTERN(void) jinit_master_decompress (j_decompress_ptr cinfo); +EXTERN(void) jinit_d_main_controller (j_decompress_ptr cinfo, + boolean need_full_buffer); +EXTERN(void) jinit_d_coef_controller (j_decompress_ptr cinfo, + boolean need_full_buffer); +EXTERN(void) jinit_d_post_controller (j_decompress_ptr cinfo, + boolean need_full_buffer); +EXTERN(void) jinit_input_controller (j_decompress_ptr cinfo); +EXTERN(void) jinit_marker_reader (j_decompress_ptr cinfo); +EXTERN(void) jinit_huff_decoder (j_decompress_ptr cinfo); +EXTERN(void) jinit_phuff_decoder (j_decompress_ptr cinfo); +EXTERN(void) jinit_arith_decoder (j_decompress_ptr cinfo); +EXTERN(void) jinit_inverse_dct (j_decompress_ptr cinfo); +EXTERN(void) jinit_upsampler (j_decompress_ptr cinfo); +EXTERN(void) jinit_color_deconverter (j_decompress_ptr cinfo); +EXTERN(void) jinit_1pass_quantizer (j_decompress_ptr cinfo); +EXTERN(void) jinit_2pass_quantizer (j_decompress_ptr cinfo); +EXTERN(void) jinit_merged_upsampler (j_decompress_ptr cinfo); +/* Memory manager initialization */ +EXTERN(void) jinit_memory_mgr (j_common_ptr cinfo); + +/* Utility routines in jutils.c */ +EXTERN(long) jdiv_round_up (long a, long b); +EXTERN(long) jround_up (long a, long b); +EXTERN(void) jcopy_sample_rows (JSAMPARRAY input_array, int source_row, + JSAMPARRAY output_array, int dest_row, + int num_rows, JDIMENSION num_cols); +EXTERN(void) jcopy_block_row (JBLOCKROW input_row, JBLOCKROW output_row, + JDIMENSION num_blocks); +EXTERN(void) jzero_far (void * target, size_t bytestozero); +/* Constant tables in jutils.c */ +#if 0 /* This table is not actually needed in v6a */ +extern const int jpeg_zigzag_order[]; /* natural coef order to zigzag order */ +#endif +extern const int jpeg_natural_order[]; /* zigzag coef order to natural order */ + +/* Arithmetic coding probability estimation tables in jaricom.c */ +extern const INT32 jpeg_aritab[]; + +/* Suppress undefined-structure complaints if necessary. */ + +#ifdef INCOMPLETE_TYPES_BROKEN +#ifndef AM_MEMORY_MANAGER /* only jmemmgr.c defines these */ +struct jvirt_sarray_control { long dummy; }; +struct jvirt_barray_control { long dummy; }; +#endif +#endif /* INCOMPLETE_TYPES_BROKEN */ diff --git a/app/src/main/jni/libjpeg-turbo/jpeglib.h b/app/src/main/jni/libjpeg-turbo/jpeglib.h new file mode 100644 index 000000000..9615c5d07 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jpeglib.h @@ -0,0 +1,1115 @@ +/* + * jpeglib.h + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1998, Thomas G. Lane. + * Modified 2002-2009 by Guido Vollbeding. + * libjpeg-turbo Modifications: + * Copyright (C) 2009-2011, 2013-2014, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file defines the application interface for the JPEG library. + * Most applications using the library need only include this file, + * and perhaps jerror.h if they want to know the exact error codes. + */ + +#ifndef JPEGLIB_H +#define JPEGLIB_H + +/* + * First we include the configuration files that record how this + * installation of the JPEG library is set up. jconfig.h can be + * generated automatically for many systems. jmorecfg.h contains + * manual configuration options that most people need not worry about. + */ + +#ifndef JCONFIG_INCLUDED /* in case jinclude.h already did */ +#include "jconfig.h" /* widely used configuration options */ +#endif +#include "jmorecfg.h" /* seldom changed options */ + + +#ifdef __cplusplus +#ifndef DONT_USE_EXTERN_C +extern "C" { +#endif +#endif + + +/* Various constants determining the sizes of things. + * All of these are specified by the JPEG standard, so don't change them + * if you want to be compatible. + */ + +#define DCTSIZE 8 /* The basic DCT block is 8x8 samples */ +#define DCTSIZE2 64 /* DCTSIZE squared; # of elements in a block */ +#define NUM_QUANT_TBLS 4 /* Quantization tables are numbered 0..3 */ +#define NUM_HUFF_TBLS 4 /* Huffman tables are numbered 0..3 */ +#define NUM_ARITH_TBLS 16 /* Arith-coding tables are numbered 0..15 */ +#define MAX_COMPS_IN_SCAN 4 /* JPEG limit on # of components in one scan */ +#define MAX_SAMP_FACTOR 4 /* JPEG limit on sampling factors */ +/* Unfortunately, some bozo at Adobe saw no reason to be bound by the standard; + * the PostScript DCT filter can emit files with many more than 10 blocks/MCU. + * If you happen to run across such a file, you can up D_MAX_BLOCKS_IN_MCU + * to handle it. We even let you do this from the jconfig.h file. However, + * we strongly discourage changing C_MAX_BLOCKS_IN_MCU; just because Adobe + * sometimes emits noncompliant files doesn't mean you should too. + */ +#define C_MAX_BLOCKS_IN_MCU 10 /* compressor's limit on blocks per MCU */ +#ifndef D_MAX_BLOCKS_IN_MCU +#define D_MAX_BLOCKS_IN_MCU 10 /* decompressor's limit on blocks per MCU */ +#endif + + +/* Data structures for images (arrays of samples and of DCT coefficients). + */ + +typedef JSAMPLE *JSAMPROW; /* ptr to one image row of pixel samples. */ +typedef JSAMPROW *JSAMPARRAY; /* ptr to some rows (a 2-D sample array) */ +typedef JSAMPARRAY *JSAMPIMAGE; /* a 3-D sample array: top index is color */ + +typedef JCOEF JBLOCK[DCTSIZE2]; /* one block of coefficients */ +typedef JBLOCK *JBLOCKROW; /* pointer to one row of coefficient blocks */ +typedef JBLOCKROW *JBLOCKARRAY; /* a 2-D array of coefficient blocks */ +typedef JBLOCKARRAY *JBLOCKIMAGE; /* a 3-D array of coefficient blocks */ + +typedef JCOEF *JCOEFPTR; /* useful in a couple of places */ + + +/* Types for JPEG compression parameters and working tables. */ + + +/* DCT coefficient quantization tables. */ + +typedef struct { + /* This array gives the coefficient quantizers in natural array order + * (not the zigzag order in which they are stored in a JPEG DQT marker). + * CAUTION: IJG versions prior to v6a kept this array in zigzag order. + */ + UINT16 quantval[DCTSIZE2]; /* quantization step for each coefficient */ + /* This field is used only during compression. It's initialized FALSE when + * the table is created, and set TRUE when it's been output to the file. + * You could suppress output of a table by setting this to TRUE. + * (See jpeg_suppress_tables for an example.) + */ + boolean sent_table; /* TRUE when table has been output */ +} JQUANT_TBL; + + +/* Huffman coding tables. */ + +typedef struct { + /* These two fields directly represent the contents of a JPEG DHT marker */ + UINT8 bits[17]; /* bits[k] = # of symbols with codes of */ + /* length k bits; bits[0] is unused */ + UINT8 huffval[256]; /* The symbols, in order of incr code length */ + /* This field is used only during compression. It's initialized FALSE when + * the table is created, and set TRUE when it's been output to the file. + * You could suppress output of a table by setting this to TRUE. + * (See jpeg_suppress_tables for an example.) + */ + boolean sent_table; /* TRUE when table has been output */ +} JHUFF_TBL; + + +/* Basic info about one component (color channel). */ + +typedef struct { + /* These values are fixed over the whole image. */ + /* For compression, they must be supplied by parameter setup; */ + /* for decompression, they are read from the SOF marker. */ + int component_id; /* identifier for this component (0..255) */ + int component_index; /* its index in SOF or cinfo->comp_info[] */ + int h_samp_factor; /* horizontal sampling factor (1..4) */ + int v_samp_factor; /* vertical sampling factor (1..4) */ + int quant_tbl_no; /* quantization table selector (0..3) */ + /* These values may vary between scans. */ + /* For compression, they must be supplied by parameter setup; */ + /* for decompression, they are read from the SOS marker. */ + /* The decompressor output side may not use these variables. */ + int dc_tbl_no; /* DC entropy table selector (0..3) */ + int ac_tbl_no; /* AC entropy table selector (0..3) */ + + /* Remaining fields should be treated as private by applications. */ + + /* These values are computed during compression or decompression startup: */ + /* Component's size in DCT blocks. + * Any dummy blocks added to complete an MCU are not counted; therefore + * these values do not depend on whether a scan is interleaved or not. + */ + JDIMENSION width_in_blocks; + JDIMENSION height_in_blocks; + /* Size of a DCT block in samples. Always DCTSIZE for compression. + * For decompression this is the size of the output from one DCT block, + * reflecting any scaling we choose to apply during the IDCT step. + * Values from 1 to 16 are supported. + * Note that different components may receive different IDCT scalings. + */ +#if JPEG_LIB_VERSION >= 70 + int DCT_h_scaled_size; + int DCT_v_scaled_size; +#else + int DCT_scaled_size; +#endif + /* The downsampled dimensions are the component's actual, unpadded number + * of samples at the main buffer (preprocessing/compression interface), thus + * downsampled_width = ceil(image_width * Hi/Hmax) + * and similarly for height. For decompression, IDCT scaling is included, so + * downsampled_width = ceil(image_width * Hi/Hmax * DCT_[h_]scaled_size/DCTSIZE) + */ + JDIMENSION downsampled_width; /* actual width in samples */ + JDIMENSION downsampled_height; /* actual height in samples */ + /* This flag is used only for decompression. In cases where some of the + * components will be ignored (eg grayscale output from YCbCr image), + * we can skip most computations for the unused components. + */ + boolean component_needed; /* do we need the value of this component? */ + + /* These values are computed before starting a scan of the component. */ + /* The decompressor output side may not use these variables. */ + int MCU_width; /* number of blocks per MCU, horizontally */ + int MCU_height; /* number of blocks per MCU, vertically */ + int MCU_blocks; /* MCU_width * MCU_height */ + int MCU_sample_width; /* MCU width in samples, MCU_width*DCT_[h_]scaled_size */ + int last_col_width; /* # of non-dummy blocks across in last MCU */ + int last_row_height; /* # of non-dummy blocks down in last MCU */ + + /* Saved quantization table for component; NULL if none yet saved. + * See jdinput.c comments about the need for this information. + * This field is currently used only for decompression. + */ + JQUANT_TBL * quant_table; + + /* Private per-component storage for DCT or IDCT subsystem. */ + void * dct_table; +} jpeg_component_info; + + +/* The script for encoding a multiple-scan file is an array of these: */ + +typedef struct { + int comps_in_scan; /* number of components encoded in this scan */ + int component_index[MAX_COMPS_IN_SCAN]; /* their SOF/comp_info[] indexes */ + int Ss, Se; /* progressive JPEG spectral selection parms */ + int Ah, Al; /* progressive JPEG successive approx. parms */ +} jpeg_scan_info; + +/* The decompressor can save APPn and COM markers in a list of these: */ + +typedef struct jpeg_marker_struct * jpeg_saved_marker_ptr; + +struct jpeg_marker_struct { + jpeg_saved_marker_ptr next; /* next in list, or NULL */ + UINT8 marker; /* marker code: JPEG_COM, or JPEG_APP0+n */ + unsigned int original_length; /* # bytes of data in the file */ + unsigned int data_length; /* # bytes of data saved at data[] */ + JOCTET * data; /* the data contained in the marker */ + /* the marker length word is not counted in data_length or original_length */ +}; + +/* Known color spaces. */ + +#define JCS_EXTENSIONS 1 +#define JCS_ALPHA_EXTENSIONS 1 + +typedef enum { + JCS_UNKNOWN, /* error/unspecified */ + JCS_GRAYSCALE, /* monochrome */ + JCS_RGB, /* red/green/blue as specified by the RGB_RED, + RGB_GREEN, RGB_BLUE, and RGB_PIXELSIZE macros */ + JCS_YCbCr, /* Y/Cb/Cr (also known as YUV) */ + JCS_CMYK, /* C/M/Y/K */ + JCS_YCCK, /* Y/Cb/Cr/K */ + JCS_EXT_RGB, /* red/green/blue */ + JCS_EXT_RGBX, /* red/green/blue/x */ + JCS_EXT_BGR, /* blue/green/red */ + JCS_EXT_BGRX, /* blue/green/red/x */ + JCS_EXT_XBGR, /* x/blue/green/red */ + JCS_EXT_XRGB, /* x/red/green/blue */ + /* When out_color_space it set to JCS_EXT_RGBX, JCS_EXT_BGRX, JCS_EXT_XBGR, + or JCS_EXT_XRGB during decompression, the X byte is undefined, and in + order to ensure the best performance, libjpeg-turbo can set that byte to + whatever value it wishes. Use the following colorspace constants to + ensure that the X byte is set to 0xFF, so that it can be interpreted as an + opaque alpha channel. */ + JCS_EXT_RGBA, /* red/green/blue/alpha */ + JCS_EXT_BGRA, /* blue/green/red/alpha */ + JCS_EXT_ABGR, /* alpha/blue/green/red */ + JCS_EXT_ARGB, /* alpha/red/green/blue */ + JCS_RGB565 /* 5-bit red/6-bit green/5-bit blue */ +} J_COLOR_SPACE; + +/* DCT/IDCT algorithm options. */ + +typedef enum { + JDCT_ISLOW, /* slow but accurate integer algorithm */ + JDCT_IFAST, /* faster, less accurate integer method */ + JDCT_FLOAT /* floating-point: accurate, fast on fast HW */ +} J_DCT_METHOD; + +#ifndef JDCT_DEFAULT /* may be overridden in jconfig.h */ +#define JDCT_DEFAULT JDCT_ISLOW +#endif +#ifndef JDCT_FASTEST /* may be overridden in jconfig.h */ +#define JDCT_FASTEST JDCT_IFAST +#endif + +/* Dithering options for decompression. */ + +typedef enum { + JDITHER_NONE, /* no dithering */ + JDITHER_ORDERED, /* simple ordered dither */ + JDITHER_FS /* Floyd-Steinberg error diffusion dither */ +} J_DITHER_MODE; + + +/* Common fields between JPEG compression and decompression master structs. */ + +#define jpeg_common_fields \ + struct jpeg_error_mgr * err; /* Error handler module */\ + struct jpeg_memory_mgr * mem; /* Memory manager module */\ + struct jpeg_progress_mgr * progress; /* Progress monitor, or NULL if none */\ + void * client_data; /* Available for use by application */\ + boolean is_decompressor; /* So common code can tell which is which */\ + int global_state /* For checking call sequence validity */ + +/* Routines that are to be used by both halves of the library are declared + * to receive a pointer to this structure. There are no actual instances of + * jpeg_common_struct, only of jpeg_compress_struct and jpeg_decompress_struct. + */ +struct jpeg_common_struct { + jpeg_common_fields; /* Fields common to both master struct types */ + /* Additional fields follow in an actual jpeg_compress_struct or + * jpeg_decompress_struct. All three structs must agree on these + * initial fields! (This would be a lot cleaner in C++.) + */ +}; + +typedef struct jpeg_common_struct * j_common_ptr; +typedef struct jpeg_compress_struct * j_compress_ptr; +typedef struct jpeg_decompress_struct * j_decompress_ptr; + + +/* Master record for a compression instance */ + +struct jpeg_compress_struct { + jpeg_common_fields; /* Fields shared with jpeg_decompress_struct */ + + /* Destination for compressed data */ + struct jpeg_destination_mgr * dest; + + /* Description of source image --- these fields must be filled in by + * outer application before starting compression. in_color_space must + * be correct before you can even call jpeg_set_defaults(). + */ + + JDIMENSION image_width; /* input image width */ + JDIMENSION image_height; /* input image height */ + int input_components; /* # of color components in input image */ + J_COLOR_SPACE in_color_space; /* colorspace of input image */ + + double input_gamma; /* image gamma of input image */ + + /* Compression parameters --- these fields must be set before calling + * jpeg_start_compress(). We recommend calling jpeg_set_defaults() to + * initialize everything to reasonable defaults, then changing anything + * the application specifically wants to change. That way you won't get + * burnt when new parameters are added. Also note that there are several + * helper routines to simplify changing parameters. + */ + +#if JPEG_LIB_VERSION >= 70 + unsigned int scale_num, scale_denom; /* fraction by which to scale image */ + + JDIMENSION jpeg_width; /* scaled JPEG image width */ + JDIMENSION jpeg_height; /* scaled JPEG image height */ + /* Dimensions of actual JPEG image that will be written to file, + * derived from input dimensions by scaling factors above. + * These fields are computed by jpeg_start_compress(). + * You can also use jpeg_calc_jpeg_dimensions() to determine these values + * in advance of calling jpeg_start_compress(). + */ +#endif + + int data_precision; /* bits of precision in image data */ + + int num_components; /* # of color components in JPEG image */ + J_COLOR_SPACE jpeg_color_space; /* colorspace of JPEG image */ + + jpeg_component_info * comp_info; + /* comp_info[i] describes component that appears i'th in SOF */ + + JQUANT_TBL * quant_tbl_ptrs[NUM_QUANT_TBLS]; +#if JPEG_LIB_VERSION >= 70 + int q_scale_factor[NUM_QUANT_TBLS]; +#endif + /* ptrs to coefficient quantization tables, or NULL if not defined, + * and corresponding scale factors (percentage, initialized 100). + */ + + JHUFF_TBL * dc_huff_tbl_ptrs[NUM_HUFF_TBLS]; + JHUFF_TBL * ac_huff_tbl_ptrs[NUM_HUFF_TBLS]; + /* ptrs to Huffman coding tables, or NULL if not defined */ + + UINT8 arith_dc_L[NUM_ARITH_TBLS]; /* L values for DC arith-coding tables */ + UINT8 arith_dc_U[NUM_ARITH_TBLS]; /* U values for DC arith-coding tables */ + UINT8 arith_ac_K[NUM_ARITH_TBLS]; /* Kx values for AC arith-coding tables */ + + int num_scans; /* # of entries in scan_info array */ + const jpeg_scan_info * scan_info; /* script for multi-scan file, or NULL */ + /* The default value of scan_info is NULL, which causes a single-scan + * sequential JPEG file to be emitted. To create a multi-scan file, + * set num_scans and scan_info to point to an array of scan definitions. + */ + + boolean raw_data_in; /* TRUE=caller supplies downsampled data */ + boolean arith_code; /* TRUE=arithmetic coding, FALSE=Huffman */ + boolean optimize_coding; /* TRUE=optimize entropy encoding parms */ + boolean CCIR601_sampling; /* TRUE=first samples are cosited */ +#if JPEG_LIB_VERSION >= 70 + boolean do_fancy_downsampling; /* TRUE=apply fancy downsampling */ +#endif + int smoothing_factor; /* 1..100, or 0 for no input smoothing */ + J_DCT_METHOD dct_method; /* DCT algorithm selector */ + + /* The restart interval can be specified in absolute MCUs by setting + * restart_interval, or in MCU rows by setting restart_in_rows + * (in which case the correct restart_interval will be figured + * for each scan). + */ + unsigned int restart_interval; /* MCUs per restart, or 0 for no restart */ + int restart_in_rows; /* if > 0, MCU rows per restart interval */ + + /* Parameters controlling emission of special markers. */ + + boolean write_JFIF_header; /* should a JFIF marker be written? */ + UINT8 JFIF_major_version; /* What to write for the JFIF version number */ + UINT8 JFIF_minor_version; + /* These three values are not used by the JPEG code, merely copied */ + /* into the JFIF APP0 marker. density_unit can be 0 for unknown, */ + /* 1 for dots/inch, or 2 for dots/cm. Note that the pixel aspect */ + /* ratio is defined by X_density/Y_density even when density_unit=0. */ + UINT8 density_unit; /* JFIF code for pixel size units */ + UINT16 X_density; /* Horizontal pixel density */ + UINT16 Y_density; /* Vertical pixel density */ + boolean write_Adobe_marker; /* should an Adobe marker be written? */ + + /* State variable: index of next scanline to be written to + * jpeg_write_scanlines(). Application may use this to control its + * processing loop, e.g., "while (next_scanline < image_height)". + */ + + JDIMENSION next_scanline; /* 0 .. image_height-1 */ + + /* Remaining fields are known throughout compressor, but generally + * should not be touched by a surrounding application. + */ + + /* + * These fields are computed during compression startup + */ + boolean progressive_mode; /* TRUE if scan script uses progressive mode */ + int max_h_samp_factor; /* largest h_samp_factor */ + int max_v_samp_factor; /* largest v_samp_factor */ + +#if JPEG_LIB_VERSION >= 70 + int min_DCT_h_scaled_size; /* smallest DCT_h_scaled_size of any component */ + int min_DCT_v_scaled_size; /* smallest DCT_v_scaled_size of any component */ +#endif + + JDIMENSION total_iMCU_rows; /* # of iMCU rows to be input to coef ctlr */ + /* The coefficient controller receives data in units of MCU rows as defined + * for fully interleaved scans (whether the JPEG file is interleaved or not). + * There are v_samp_factor * DCTSIZE sample rows of each component in an + * "iMCU" (interleaved MCU) row. + */ + + /* + * These fields are valid during any one scan. + * They describe the components and MCUs actually appearing in the scan. + */ + int comps_in_scan; /* # of JPEG components in this scan */ + jpeg_component_info * cur_comp_info[MAX_COMPS_IN_SCAN]; + /* *cur_comp_info[i] describes component that appears i'th in SOS */ + + JDIMENSION MCUs_per_row; /* # of MCUs across the image */ + JDIMENSION MCU_rows_in_scan; /* # of MCU rows in the image */ + + int blocks_in_MCU; /* # of DCT blocks per MCU */ + int MCU_membership[C_MAX_BLOCKS_IN_MCU]; + /* MCU_membership[i] is index in cur_comp_info of component owning */ + /* i'th block in an MCU */ + + int Ss, Se, Ah, Al; /* progressive JPEG parameters for scan */ + +#if JPEG_LIB_VERSION >= 80 + int block_size; /* the basic DCT block size: 1..16 */ + const int * natural_order; /* natural-order position array */ + int lim_Se; /* min( Se, DCTSIZE2-1 ) */ +#endif + + /* + * Links to compression subobjects (methods and private variables of modules) + */ + struct jpeg_comp_master * master; + struct jpeg_c_main_controller * main; + struct jpeg_c_prep_controller * prep; + struct jpeg_c_coef_controller * coef; + struct jpeg_marker_writer * marker; + struct jpeg_color_converter * cconvert; + struct jpeg_downsampler * downsample; + struct jpeg_forward_dct * fdct; + struct jpeg_entropy_encoder * entropy; + jpeg_scan_info * script_space; /* workspace for jpeg_simple_progression */ + int script_space_size; +}; + + +/* Master record for a decompression instance */ + +struct jpeg_decompress_struct { + jpeg_common_fields; /* Fields shared with jpeg_compress_struct */ + + /* Source of compressed data */ + struct jpeg_source_mgr * src; + + /* Basic description of image --- filled in by jpeg_read_header(). */ + /* Application may inspect these values to decide how to process image. */ + + JDIMENSION image_width; /* nominal image width (from SOF marker) */ + JDIMENSION image_height; /* nominal image height */ + int num_components; /* # of color components in JPEG image */ + J_COLOR_SPACE jpeg_color_space; /* colorspace of JPEG image */ + + /* Decompression processing parameters --- these fields must be set before + * calling jpeg_start_decompress(). Note that jpeg_read_header() initializes + * them to default values. + */ + + J_COLOR_SPACE out_color_space; /* colorspace for output */ + + unsigned int scale_num, scale_denom; /* fraction by which to scale image */ + + double output_gamma; /* image gamma wanted in output */ + + boolean buffered_image; /* TRUE=multiple output passes */ + boolean raw_data_out; /* TRUE=downsampled data wanted */ + + J_DCT_METHOD dct_method; /* IDCT algorithm selector */ + boolean do_fancy_upsampling; /* TRUE=apply fancy upsampling */ + boolean do_block_smoothing; /* TRUE=apply interblock smoothing */ + + boolean quantize_colors; /* TRUE=colormapped output wanted */ + /* the following are ignored if not quantize_colors: */ + J_DITHER_MODE dither_mode; /* type of color dithering to use */ + boolean two_pass_quantize; /* TRUE=use two-pass color quantization */ + int desired_number_of_colors; /* max # colors to use in created colormap */ + /* these are significant only in buffered-image mode: */ + boolean enable_1pass_quant; /* enable future use of 1-pass quantizer */ + boolean enable_external_quant;/* enable future use of external colormap */ + boolean enable_2pass_quant; /* enable future use of 2-pass quantizer */ + + /* Description of actual output image that will be returned to application. + * These fields are computed by jpeg_start_decompress(). + * You can also use jpeg_calc_output_dimensions() to determine these values + * in advance of calling jpeg_start_decompress(). + */ + + JDIMENSION output_width; /* scaled image width */ + JDIMENSION output_height; /* scaled image height */ + int out_color_components; /* # of color components in out_color_space */ + int output_components; /* # of color components returned */ + /* output_components is 1 (a colormap index) when quantizing colors; + * otherwise it equals out_color_components. + */ + int rec_outbuf_height; /* min recommended height of scanline buffer */ + /* If the buffer passed to jpeg_read_scanlines() is less than this many rows + * high, space and time will be wasted due to unnecessary data copying. + * Usually rec_outbuf_height will be 1 or 2, at most 4. + */ + + /* When quantizing colors, the output colormap is described by these fields. + * The application can supply a colormap by setting colormap non-NULL before + * calling jpeg_start_decompress; otherwise a colormap is created during + * jpeg_start_decompress or jpeg_start_output. + * The map has out_color_components rows and actual_number_of_colors columns. + */ + int actual_number_of_colors; /* number of entries in use */ + JSAMPARRAY colormap; /* The color map as a 2-D pixel array */ + + /* State variables: these variables indicate the progress of decompression. + * The application may examine these but must not modify them. + */ + + /* Row index of next scanline to be read from jpeg_read_scanlines(). + * Application may use this to control its processing loop, e.g., + * "while (output_scanline < output_height)". + */ + JDIMENSION output_scanline; /* 0 .. output_height-1 */ + + /* Current input scan number and number of iMCU rows completed in scan. + * These indicate the progress of the decompressor input side. + */ + int input_scan_number; /* Number of SOS markers seen so far */ + JDIMENSION input_iMCU_row; /* Number of iMCU rows completed */ + + /* The "output scan number" is the notional scan being displayed by the + * output side. The decompressor will not allow output scan/row number + * to get ahead of input scan/row, but it can fall arbitrarily far behind. + */ + int output_scan_number; /* Nominal scan number being displayed */ + JDIMENSION output_iMCU_row; /* Number of iMCU rows read */ + + /* Current progression status. coef_bits[c][i] indicates the precision + * with which component c's DCT coefficient i (in zigzag order) is known. + * It is -1 when no data has yet been received, otherwise it is the point + * transform (shift) value for the most recent scan of the coefficient + * (thus, 0 at completion of the progression). + * This pointer is NULL when reading a non-progressive file. + */ + int (*coef_bits)[DCTSIZE2]; /* -1 or current Al value for each coef */ + + /* Internal JPEG parameters --- the application usually need not look at + * these fields. Note that the decompressor output side may not use + * any parameters that can change between scans. + */ + + /* Quantization and Huffman tables are carried forward across input + * datastreams when processing abbreviated JPEG datastreams. + */ + + JQUANT_TBL * quant_tbl_ptrs[NUM_QUANT_TBLS]; + /* ptrs to coefficient quantization tables, or NULL if not defined */ + + JHUFF_TBL * dc_huff_tbl_ptrs[NUM_HUFF_TBLS]; + JHUFF_TBL * ac_huff_tbl_ptrs[NUM_HUFF_TBLS]; + /* ptrs to Huffman coding tables, or NULL if not defined */ + + /* These parameters are never carried across datastreams, since they + * are given in SOF/SOS markers or defined to be reset by SOI. + */ + + int data_precision; /* bits of precision in image data */ + + jpeg_component_info * comp_info; + /* comp_info[i] describes component that appears i'th in SOF */ + +#if JPEG_LIB_VERSION >= 80 + boolean is_baseline; /* TRUE if Baseline SOF0 encountered */ +#endif + boolean progressive_mode; /* TRUE if SOFn specifies progressive mode */ + boolean arith_code; /* TRUE=arithmetic coding, FALSE=Huffman */ + + UINT8 arith_dc_L[NUM_ARITH_TBLS]; /* L values for DC arith-coding tables */ + UINT8 arith_dc_U[NUM_ARITH_TBLS]; /* U values for DC arith-coding tables */ + UINT8 arith_ac_K[NUM_ARITH_TBLS]; /* Kx values for AC arith-coding tables */ + + unsigned int restart_interval; /* MCUs per restart interval, or 0 for no restart */ + + /* These fields record data obtained from optional markers recognized by + * the JPEG library. + */ + boolean saw_JFIF_marker; /* TRUE iff a JFIF APP0 marker was found */ + /* Data copied from JFIF marker; only valid if saw_JFIF_marker is TRUE: */ + UINT8 JFIF_major_version; /* JFIF version number */ + UINT8 JFIF_minor_version; + UINT8 density_unit; /* JFIF code for pixel size units */ + UINT16 X_density; /* Horizontal pixel density */ + UINT16 Y_density; /* Vertical pixel density */ + boolean saw_Adobe_marker; /* TRUE iff an Adobe APP14 marker was found */ + UINT8 Adobe_transform; /* Color transform code from Adobe marker */ + + boolean CCIR601_sampling; /* TRUE=first samples are cosited */ + + /* Aside from the specific data retained from APPn markers known to the + * library, the uninterpreted contents of any or all APPn and COM markers + * can be saved in a list for examination by the application. + */ + jpeg_saved_marker_ptr marker_list; /* Head of list of saved markers */ + + /* Remaining fields are known throughout decompressor, but generally + * should not be touched by a surrounding application. + */ + + /* + * These fields are computed during decompression startup + */ + int max_h_samp_factor; /* largest h_samp_factor */ + int max_v_samp_factor; /* largest v_samp_factor */ + +#if JPEG_LIB_VERSION >= 70 + int min_DCT_h_scaled_size; /* smallest DCT_h_scaled_size of any component */ + int min_DCT_v_scaled_size; /* smallest DCT_v_scaled_size of any component */ +#else + int min_DCT_scaled_size; /* smallest DCT_scaled_size of any component */ +#endif + + JDIMENSION total_iMCU_rows; /* # of iMCU rows in image */ + /* The coefficient controller's input and output progress is measured in + * units of "iMCU" (interleaved MCU) rows. These are the same as MCU rows + * in fully interleaved JPEG scans, but are used whether the scan is + * interleaved or not. We define an iMCU row as v_samp_factor DCT block + * rows of each component. Therefore, the IDCT output contains + * v_samp_factor*DCT_[v_]scaled_size sample rows of a component per iMCU row. + */ + + JSAMPLE * sample_range_limit; /* table for fast range-limiting */ + + /* + * These fields are valid during any one scan. + * They describe the components and MCUs actually appearing in the scan. + * Note that the decompressor output side must not use these fields. + */ + int comps_in_scan; /* # of JPEG components in this scan */ + jpeg_component_info * cur_comp_info[MAX_COMPS_IN_SCAN]; + /* *cur_comp_info[i] describes component that appears i'th in SOS */ + + JDIMENSION MCUs_per_row; /* # of MCUs across the image */ + JDIMENSION MCU_rows_in_scan; /* # of MCU rows in the image */ + + int blocks_in_MCU; /* # of DCT blocks per MCU */ + int MCU_membership[D_MAX_BLOCKS_IN_MCU]; + /* MCU_membership[i] is index in cur_comp_info of component owning */ + /* i'th block in an MCU */ + + int Ss, Se, Ah, Al; /* progressive JPEG parameters for scan */ + +#if JPEG_LIB_VERSION >= 80 + /* These fields are derived from Se of first SOS marker. + */ + int block_size; /* the basic DCT block size: 1..16 */ + const int * natural_order; /* natural-order position array for entropy decode */ + int lim_Se; /* min( Se, DCTSIZE2-1 ) for entropy decode */ +#endif + + /* This field is shared between entropy decoder and marker parser. + * It is either zero or the code of a JPEG marker that has been + * read from the data source, but has not yet been processed. + */ + int unread_marker; + + /* + * Links to decompression subobjects (methods, private variables of modules) + */ + struct jpeg_decomp_master * master; + struct jpeg_d_main_controller * main; + struct jpeg_d_coef_controller * coef; + struct jpeg_d_post_controller * post; + struct jpeg_input_controller * inputctl; + struct jpeg_marker_reader * marker; + struct jpeg_entropy_decoder * entropy; + struct jpeg_inverse_dct * idct; + struct jpeg_upsampler * upsample; + struct jpeg_color_deconverter * cconvert; + struct jpeg_color_quantizer * cquantize; +}; + + +/* "Object" declarations for JPEG modules that may be supplied or called + * directly by the surrounding application. + * As with all objects in the JPEG library, these structs only define the + * publicly visible methods and state variables of a module. Additional + * private fields may exist after the public ones. + */ + + +/* Error handler object */ + +struct jpeg_error_mgr { + /* Error exit handler: does not return to caller */ + void (*error_exit) (j_common_ptr cinfo); + /* Conditionally emit a trace or warning message */ + void (*emit_message) (j_common_ptr cinfo, int msg_level); + /* Routine that actually outputs a trace or error message */ + void (*output_message) (j_common_ptr cinfo); + /* Format a message string for the most recent JPEG error or message */ + void (*format_message) (j_common_ptr cinfo, char * buffer); +#define JMSG_LENGTH_MAX 200 /* recommended size of format_message buffer */ + /* Reset error state variables at start of a new image */ + void (*reset_error_mgr) (j_common_ptr cinfo); + + /* The message ID code and any parameters are saved here. + * A message can have one string parameter or up to 8 int parameters. + */ + int msg_code; +#define JMSG_STR_PARM_MAX 80 + union { + int i[8]; + char s[JMSG_STR_PARM_MAX]; + } msg_parm; + + /* Standard state variables for error facility */ + + int trace_level; /* max msg_level that will be displayed */ + + /* For recoverable corrupt-data errors, we emit a warning message, + * but keep going unless emit_message chooses to abort. emit_message + * should count warnings in num_warnings. The surrounding application + * can check for bad data by seeing if num_warnings is nonzero at the + * end of processing. + */ + long num_warnings; /* number of corrupt-data warnings */ + + /* These fields point to the table(s) of error message strings. + * An application can change the table pointer to switch to a different + * message list (typically, to change the language in which errors are + * reported). Some applications may wish to add additional error codes + * that will be handled by the JPEG library error mechanism; the second + * table pointer is used for this purpose. + * + * First table includes all errors generated by JPEG library itself. + * Error code 0 is reserved for a "no such error string" message. + */ + const char * const * jpeg_message_table; /* Library errors */ + int last_jpeg_message; /* Table contains strings 0..last_jpeg_message */ + /* Second table can be added by application (see cjpeg/djpeg for example). + * It contains strings numbered first_addon_message..last_addon_message. + */ + const char * const * addon_message_table; /* Non-library errors */ + int first_addon_message; /* code for first string in addon table */ + int last_addon_message; /* code for last string in addon table */ +}; + + +/* Progress monitor object */ + +struct jpeg_progress_mgr { + void (*progress_monitor) (j_common_ptr cinfo); + + long pass_counter; /* work units completed in this pass */ + long pass_limit; /* total number of work units in this pass */ + int completed_passes; /* passes completed so far */ + int total_passes; /* total number of passes expected */ +}; + + +/* Data destination object for compression */ + +struct jpeg_destination_mgr { + JOCTET * next_output_byte; /* => next byte to write in buffer */ + size_t free_in_buffer; /* # of byte spaces remaining in buffer */ + + void (*init_destination) (j_compress_ptr cinfo); + boolean (*empty_output_buffer) (j_compress_ptr cinfo); + void (*term_destination) (j_compress_ptr cinfo); +}; + + +/* Data source object for decompression */ + +struct jpeg_source_mgr { + const JOCTET * next_input_byte; /* => next byte to read from buffer */ + size_t bytes_in_buffer; /* # of bytes remaining in buffer */ + + void (*init_source) (j_decompress_ptr cinfo); + boolean (*fill_input_buffer) (j_decompress_ptr cinfo); + void (*skip_input_data) (j_decompress_ptr cinfo, long num_bytes); + boolean (*resync_to_restart) (j_decompress_ptr cinfo, int desired); + void (*term_source) (j_decompress_ptr cinfo); +}; + + +/* Memory manager object. + * Allocates "small" objects (a few K total), "large" objects (tens of K), + * and "really big" objects (virtual arrays with backing store if needed). + * The memory manager does not allow individual objects to be freed; rather, + * each created object is assigned to a pool, and whole pools can be freed + * at once. This is faster and more convenient than remembering exactly what + * to free, especially where malloc()/free() are not too speedy. + * NB: alloc routines never return NULL. They exit to error_exit if not + * successful. + */ + +#define JPOOL_PERMANENT 0 /* lasts until master record is destroyed */ +#define JPOOL_IMAGE 1 /* lasts until done with image/datastream */ +#define JPOOL_NUMPOOLS 2 + +typedef struct jvirt_sarray_control * jvirt_sarray_ptr; +typedef struct jvirt_barray_control * jvirt_barray_ptr; + + +struct jpeg_memory_mgr { + /* Method pointers */ + void * (*alloc_small) (j_common_ptr cinfo, int pool_id, size_t sizeofobject); + void * (*alloc_large) (j_common_ptr cinfo, int pool_id, + size_t sizeofobject); + JSAMPARRAY (*alloc_sarray) (j_common_ptr cinfo, int pool_id, + JDIMENSION samplesperrow, JDIMENSION numrows); + JBLOCKARRAY (*alloc_barray) (j_common_ptr cinfo, int pool_id, + JDIMENSION blocksperrow, JDIMENSION numrows); + jvirt_sarray_ptr (*request_virt_sarray) (j_common_ptr cinfo, int pool_id, + boolean pre_zero, + JDIMENSION samplesperrow, + JDIMENSION numrows, + JDIMENSION maxaccess); + jvirt_barray_ptr (*request_virt_barray) (j_common_ptr cinfo, int pool_id, + boolean pre_zero, + JDIMENSION blocksperrow, + JDIMENSION numrows, + JDIMENSION maxaccess); + void (*realize_virt_arrays) (j_common_ptr cinfo); + JSAMPARRAY (*access_virt_sarray) (j_common_ptr cinfo, jvirt_sarray_ptr ptr, + JDIMENSION start_row, JDIMENSION num_rows, + boolean writable); + JBLOCKARRAY (*access_virt_barray) (j_common_ptr cinfo, jvirt_barray_ptr ptr, + JDIMENSION start_row, JDIMENSION num_rows, + boolean writable); + void (*free_pool) (j_common_ptr cinfo, int pool_id); + void (*self_destruct) (j_common_ptr cinfo); + + /* Limit on memory allocation for this JPEG object. (Note that this is + * merely advisory, not a guaranteed maximum; it only affects the space + * used for virtual-array buffers.) May be changed by outer application + * after creating the JPEG object. + */ + long max_memory_to_use; + + /* Maximum allocation request accepted by alloc_large. */ + long max_alloc_chunk; +}; + + +/* Routine signature for application-supplied marker processing methods. + * Need not pass marker code since it is stored in cinfo->unread_marker. + */ +typedef boolean (*jpeg_marker_parser_method) (j_decompress_ptr cinfo); + + +/* Originally, this macro was used as a way of defining function prototypes + * for both modern compilers as well as older compilers that did not support + * prototype parameters. libjpeg-turbo has never supported these older, + * non-ANSI compilers, but the macro is still included because there is some + * software out there that uses it. + */ + +#define JPP(arglist) arglist + + +/* Default error-management setup */ +EXTERN(struct jpeg_error_mgr *) jpeg_std_error (struct jpeg_error_mgr * err); + +/* Initialization of JPEG compression objects. + * jpeg_create_compress() and jpeg_create_decompress() are the exported + * names that applications should call. These expand to calls on + * jpeg_CreateCompress and jpeg_CreateDecompress with additional information + * passed for version mismatch checking. + * NB: you must set up the error-manager BEFORE calling jpeg_create_xxx. + */ +#define jpeg_create_compress(cinfo) \ + jpeg_CreateCompress((cinfo), JPEG_LIB_VERSION, \ + (size_t) sizeof(struct jpeg_compress_struct)) +#define jpeg_create_decompress(cinfo) \ + jpeg_CreateDecompress((cinfo), JPEG_LIB_VERSION, \ + (size_t) sizeof(struct jpeg_decompress_struct)) +EXTERN(void) jpeg_CreateCompress (j_compress_ptr cinfo, int version, + size_t structsize); +EXTERN(void) jpeg_CreateDecompress (j_decompress_ptr cinfo, int version, + size_t structsize); +/* Destruction of JPEG compression objects */ +EXTERN(void) jpeg_destroy_compress (j_compress_ptr cinfo); +EXTERN(void) jpeg_destroy_decompress (j_decompress_ptr cinfo); + +/* Standard data source and destination managers: stdio streams. */ +/* Caller is responsible for opening the file before and closing after. */ +EXTERN(void) jpeg_stdio_dest (j_compress_ptr cinfo, FILE * outfile); +EXTERN(void) jpeg_stdio_src (j_decompress_ptr cinfo, FILE * infile); + +#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED) +/* Data source and destination managers: memory buffers. */ +EXTERN(void) jpeg_mem_dest (j_compress_ptr cinfo, unsigned char ** outbuffer, + unsigned long * outsize); +EXTERN(void) jpeg_mem_src (j_decompress_ptr cinfo, unsigned char * inbuffer, + unsigned long insize); +#endif + +/* Default parameter setup for compression */ +EXTERN(void) jpeg_set_defaults (j_compress_ptr cinfo); +/* Compression parameter setup aids */ +EXTERN(void) jpeg_set_colorspace (j_compress_ptr cinfo, + J_COLOR_SPACE colorspace); +EXTERN(void) jpeg_default_colorspace (j_compress_ptr cinfo); +EXTERN(void) jpeg_set_quality (j_compress_ptr cinfo, int quality, + boolean force_baseline); +EXTERN(void) jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor, + boolean force_baseline); +#if JPEG_LIB_VERSION >= 70 +EXTERN(void) jpeg_default_qtables (j_compress_ptr cinfo, + boolean force_baseline); +#endif +EXTERN(void) jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl, + const unsigned int *basic_table, + int scale_factor, boolean force_baseline); +EXTERN(int) jpeg_quality_scaling (int quality); +EXTERN(void) jpeg_simple_progression (j_compress_ptr cinfo); +EXTERN(void) jpeg_suppress_tables (j_compress_ptr cinfo, boolean suppress); +EXTERN(JQUANT_TBL *) jpeg_alloc_quant_table (j_common_ptr cinfo); +EXTERN(JHUFF_TBL *) jpeg_alloc_huff_table (j_common_ptr cinfo); + +/* Main entry points for compression */ +EXTERN(void) jpeg_start_compress (j_compress_ptr cinfo, + boolean write_all_tables); +EXTERN(JDIMENSION) jpeg_write_scanlines (j_compress_ptr cinfo, + JSAMPARRAY scanlines, + JDIMENSION num_lines); +EXTERN(void) jpeg_finish_compress (j_compress_ptr cinfo); + +#if JPEG_LIB_VERSION >= 70 +/* Precalculate JPEG dimensions for current compression parameters. */ +EXTERN(void) jpeg_calc_jpeg_dimensions (j_compress_ptr cinfo); +#endif + +/* Replaces jpeg_write_scanlines when writing raw downsampled data. */ +EXTERN(JDIMENSION) jpeg_write_raw_data (j_compress_ptr cinfo, JSAMPIMAGE data, + JDIMENSION num_lines); + +/* Write a special marker. See libjpeg.txt concerning safe usage. */ +EXTERN(void) jpeg_write_marker (j_compress_ptr cinfo, int marker, + const JOCTET * dataptr, unsigned int datalen); +/* Same, but piecemeal. */ +EXTERN(void) jpeg_write_m_header (j_compress_ptr cinfo, int marker, + unsigned int datalen); +EXTERN(void) jpeg_write_m_byte (j_compress_ptr cinfo, int val); + +/* Alternate compression function: just write an abbreviated table file */ +EXTERN(void) jpeg_write_tables (j_compress_ptr cinfo); + +/* Decompression startup: read start of JPEG datastream to see what's there */ +EXTERN(int) jpeg_read_header (j_decompress_ptr cinfo, boolean require_image); +/* Return value is one of: */ +#define JPEG_SUSPENDED 0 /* Suspended due to lack of input data */ +#define JPEG_HEADER_OK 1 /* Found valid image datastream */ +#define JPEG_HEADER_TABLES_ONLY 2 /* Found valid table-specs-only datastream */ +/* If you pass require_image = TRUE (normal case), you need not check for + * a TABLES_ONLY return code; an abbreviated file will cause an error exit. + * JPEG_SUSPENDED is only possible if you use a data source module that can + * give a suspension return (the stdio source module doesn't). + */ + +/* Main entry points for decompression */ +EXTERN(boolean) jpeg_start_decompress (j_decompress_ptr cinfo); +EXTERN(JDIMENSION) jpeg_read_scanlines (j_decompress_ptr cinfo, + JSAMPARRAY scanlines, + JDIMENSION max_lines); +EXTERN(boolean) jpeg_finish_decompress (j_decompress_ptr cinfo); + +/* Replaces jpeg_read_scanlines when reading raw downsampled data. */ +EXTERN(JDIMENSION) jpeg_read_raw_data (j_decompress_ptr cinfo, JSAMPIMAGE data, + JDIMENSION max_lines); + +/* Additional entry points for buffered-image mode. */ +EXTERN(boolean) jpeg_has_multiple_scans (j_decompress_ptr cinfo); +EXTERN(boolean) jpeg_start_output (j_decompress_ptr cinfo, int scan_number); +EXTERN(boolean) jpeg_finish_output (j_decompress_ptr cinfo); +EXTERN(boolean) jpeg_input_complete (j_decompress_ptr cinfo); +EXTERN(void) jpeg_new_colormap (j_decompress_ptr cinfo); +EXTERN(int) jpeg_consume_input (j_decompress_ptr cinfo); +/* Return value is one of: */ +/* #define JPEG_SUSPENDED 0 Suspended due to lack of input data */ +#define JPEG_REACHED_SOS 1 /* Reached start of new scan */ +#define JPEG_REACHED_EOI 2 /* Reached end of image */ +#define JPEG_ROW_COMPLETED 3 /* Completed one iMCU row */ +#define JPEG_SCAN_COMPLETED 4 /* Completed last iMCU row of a scan */ + +/* Precalculate output dimensions for current decompression parameters. */ +#if JPEG_LIB_VERSION >= 80 +EXTERN(void) jpeg_core_output_dimensions (j_decompress_ptr cinfo); +#endif +EXTERN(void) jpeg_calc_output_dimensions (j_decompress_ptr cinfo); + +/* Control saving of COM and APPn markers into marker_list. */ +EXTERN(void) jpeg_save_markers (j_decompress_ptr cinfo, int marker_code, + unsigned int length_limit); + +/* Install a special processing method for COM or APPn markers. */ +EXTERN(void) jpeg_set_marker_processor (j_decompress_ptr cinfo, + int marker_code, + jpeg_marker_parser_method routine); + +/* Read or write raw DCT coefficients --- useful for lossless transcoding. */ +EXTERN(jvirt_barray_ptr *) jpeg_read_coefficients (j_decompress_ptr cinfo); +EXTERN(void) jpeg_write_coefficients (j_compress_ptr cinfo, + jvirt_barray_ptr * coef_arrays); +EXTERN(void) jpeg_copy_critical_parameters (j_decompress_ptr srcinfo, + j_compress_ptr dstinfo); + +/* If you choose to abort compression or decompression before completing + * jpeg_finish_(de)compress, then you need to clean up to release memory, + * temporary files, etc. You can just call jpeg_destroy_(de)compress + * if you're done with the JPEG object, but if you want to clean it up and + * reuse it, call this: + */ +EXTERN(void) jpeg_abort_compress (j_compress_ptr cinfo); +EXTERN(void) jpeg_abort_decompress (j_decompress_ptr cinfo); + +/* Generic versions of jpeg_abort and jpeg_destroy that work on either + * flavor of JPEG object. These may be more convenient in some places. + */ +EXTERN(void) jpeg_abort (j_common_ptr cinfo); +EXTERN(void) jpeg_destroy (j_common_ptr cinfo); + +/* Default restart-marker-resync procedure for use by data source modules */ +EXTERN(boolean) jpeg_resync_to_restart (j_decompress_ptr cinfo, int desired); + + +/* These marker codes are exported since applications and data source modules + * are likely to want to use them. + */ + +#define JPEG_RST0 0xD0 /* RST0 marker code */ +#define JPEG_EOI 0xD9 /* EOI marker code */ +#define JPEG_APP0 0xE0 /* APP0 marker code */ +#define JPEG_COM 0xFE /* COM marker code */ + + +/* If we have a brain-damaged compiler that emits warnings (or worse, errors) + * for structure definitions that are never filled in, keep it quiet by + * supplying dummy definitions for the various substructures. + */ + +#ifdef INCOMPLETE_TYPES_BROKEN +#ifndef JPEG_INTERNALS /* will be defined in jpegint.h */ +struct jvirt_sarray_control { long dummy; }; +struct jvirt_barray_control { long dummy; }; +struct jpeg_comp_master { long dummy; }; +struct jpeg_c_main_controller { long dummy; }; +struct jpeg_c_prep_controller { long dummy; }; +struct jpeg_c_coef_controller { long dummy; }; +struct jpeg_marker_writer { long dummy; }; +struct jpeg_color_converter { long dummy; }; +struct jpeg_downsampler { long dummy; }; +struct jpeg_forward_dct { long dummy; }; +struct jpeg_entropy_encoder { long dummy; }; +struct jpeg_decomp_master { long dummy; }; +struct jpeg_d_main_controller { long dummy; }; +struct jpeg_d_coef_controller { long dummy; }; +struct jpeg_d_post_controller { long dummy; }; +struct jpeg_input_controller { long dummy; }; +struct jpeg_marker_reader { long dummy; }; +struct jpeg_entropy_decoder { long dummy; }; +struct jpeg_inverse_dct { long dummy; }; +struct jpeg_upsampler { long dummy; }; +struct jpeg_color_deconverter { long dummy; }; +struct jpeg_color_quantizer { long dummy; }; +#endif /* JPEG_INTERNALS */ +#endif /* INCOMPLETE_TYPES_BROKEN */ + + +/* + * The JPEG library modules define JPEG_INTERNALS before including this file. + * The internal structure declarations are read only when that is true. + * Applications using the library should not include jpegint.h, but may wish + * to include jerror.h. + */ + +#ifdef JPEG_INTERNALS +#include "jpegint.h" /* fetch private declarations */ +#include "jerror.h" /* fetch error codes too */ +#endif + +#ifdef __cplusplus +#ifndef DONT_USE_EXTERN_C +} +#endif +#endif + +#endif /* JPEGLIB_H */ diff --git a/app/src/main/jni/libjpeg-turbo/jquant1.c b/app/src/main/jni/libjpeg-turbo/jquant1.c new file mode 100644 index 000000000..0e2535477 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jquant1.c @@ -0,0 +1,856 @@ +/* + * jquant1.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1996, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright (C) 2009, D. R. Commander + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains 1-pass color quantization (color mapping) routines. + * These routines provide mapping to a fixed color map using equally spaced + * color values. Optional Floyd-Steinberg or ordered dithering is available. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" + +#ifdef QUANT_1PASS_SUPPORTED + + +/* + * The main purpose of 1-pass quantization is to provide a fast, if not very + * high quality, colormapped output capability. A 2-pass quantizer usually + * gives better visual quality; however, for quantized grayscale output this + * quantizer is perfectly adequate. Dithering is highly recommended with this + * quantizer, though you can turn it off if you really want to. + * + * In 1-pass quantization the colormap must be chosen in advance of seeing the + * image. We use a map consisting of all combinations of Ncolors[i] color + * values for the i'th component. The Ncolors[] values are chosen so that + * their product, the total number of colors, is no more than that requested. + * (In most cases, the product will be somewhat less.) + * + * Since the colormap is orthogonal, the representative value for each color + * component can be determined without considering the other components; + * then these indexes can be combined into a colormap index by a standard + * N-dimensional-array-subscript calculation. Most of the arithmetic involved + * can be precalculated and stored in the lookup table colorindex[]. + * colorindex[i][j] maps pixel value j in component i to the nearest + * representative value (grid plane) for that component; this index is + * multiplied by the array stride for component i, so that the + * index of the colormap entry closest to a given pixel value is just + * sum( colorindex[component-number][pixel-component-value] ) + * Aside from being fast, this scheme allows for variable spacing between + * representative values with no additional lookup cost. + * + * If gamma correction has been applied in color conversion, it might be wise + * to adjust the color grid spacing so that the representative colors are + * equidistant in linear space. At this writing, gamma correction is not + * implemented by jdcolor, so nothing is done here. + */ + + +/* Declarations for ordered dithering. + * + * We use a standard 16x16 ordered dither array. The basic concept of ordered + * dithering is described in many references, for instance Dale Schumacher's + * chapter II.2 of Graphics Gems II (James Arvo, ed. Academic Press, 1991). + * In place of Schumacher's comparisons against a "threshold" value, we add a + * "dither" value to the input pixel and then round the result to the nearest + * output value. The dither value is equivalent to (0.5 - threshold) times + * the distance between output values. For ordered dithering, we assume that + * the output colors are equally spaced; if not, results will probably be + * worse, since the dither may be too much or too little at a given point. + * + * The normal calculation would be to form pixel value + dither, range-limit + * this to 0..MAXJSAMPLE, and then index into the colorindex table as usual. + * We can skip the separate range-limiting step by extending the colorindex + * table in both directions. + */ + +#define ODITHER_SIZE 16 /* dimension of dither matrix */ +/* NB: if ODITHER_SIZE is not a power of 2, ODITHER_MASK uses will break */ +#define ODITHER_CELLS (ODITHER_SIZE*ODITHER_SIZE) /* # cells in matrix */ +#define ODITHER_MASK (ODITHER_SIZE-1) /* mask for wrapping around counters */ + +typedef int ODITHER_MATRIX[ODITHER_SIZE][ODITHER_SIZE]; +typedef int (*ODITHER_MATRIX_PTR)[ODITHER_SIZE]; + +static const UINT8 base_dither_matrix[ODITHER_SIZE][ODITHER_SIZE] = { + /* Bayer's order-4 dither array. Generated by the code given in + * Stephen Hawley's article "Ordered Dithering" in Graphics Gems I. + * The values in this array must range from 0 to ODITHER_CELLS-1. + */ + { 0,192, 48,240, 12,204, 60,252, 3,195, 51,243, 15,207, 63,255 }, + { 128, 64,176,112,140, 76,188,124,131, 67,179,115,143, 79,191,127 }, + { 32,224, 16,208, 44,236, 28,220, 35,227, 19,211, 47,239, 31,223 }, + { 160, 96,144, 80,172,108,156, 92,163, 99,147, 83,175,111,159, 95 }, + { 8,200, 56,248, 4,196, 52,244, 11,203, 59,251, 7,199, 55,247 }, + { 136, 72,184,120,132, 68,180,116,139, 75,187,123,135, 71,183,119 }, + { 40,232, 24,216, 36,228, 20,212, 43,235, 27,219, 39,231, 23,215 }, + { 168,104,152, 88,164,100,148, 84,171,107,155, 91,167,103,151, 87 }, + { 2,194, 50,242, 14,206, 62,254, 1,193, 49,241, 13,205, 61,253 }, + { 130, 66,178,114,142, 78,190,126,129, 65,177,113,141, 77,189,125 }, + { 34,226, 18,210, 46,238, 30,222, 33,225, 17,209, 45,237, 29,221 }, + { 162, 98,146, 82,174,110,158, 94,161, 97,145, 81,173,109,157, 93 }, + { 10,202, 58,250, 6,198, 54,246, 9,201, 57,249, 5,197, 53,245 }, + { 138, 74,186,122,134, 70,182,118,137, 73,185,121,133, 69,181,117 }, + { 42,234, 26,218, 38,230, 22,214, 41,233, 25,217, 37,229, 21,213 }, + { 170,106,154, 90,166,102,150, 86,169,105,153, 89,165,101,149, 85 } +}; + + +/* Declarations for Floyd-Steinberg dithering. + * + * Errors are accumulated into the array fserrors[], at a resolution of + * 1/16th of a pixel count. The error at a given pixel is propagated + * to its not-yet-processed neighbors using the standard F-S fractions, + * ... (here) 7/16 + * 3/16 5/16 1/16 + * We work left-to-right on even rows, right-to-left on odd rows. + * + * We can get away with a single array (holding one row's worth of errors) + * by using it to store the current row's errors at pixel columns not yet + * processed, but the next row's errors at columns already processed. We + * need only a few extra variables to hold the errors immediately around the + * current column. (If we are lucky, those variables are in registers, but + * even if not, they're probably cheaper to access than array elements are.) + * + * The fserrors[] array is indexed [component#][position]. + * We provide (#columns + 2) entries per component; the extra entry at each + * end saves us from special-casing the first and last pixels. + */ + +#if BITS_IN_JSAMPLE == 8 +typedef INT16 FSERROR; /* 16 bits should be enough */ +typedef int LOCFSERROR; /* use 'int' for calculation temps */ +#else +typedef INT32 FSERROR; /* may need more than 16 bits */ +typedef INT32 LOCFSERROR; /* be sure calculation temps are big enough */ +#endif + +typedef FSERROR *FSERRPTR; /* pointer to error array */ + + +/* Private subobject */ + +#define MAX_Q_COMPS 4 /* max components I can handle */ + +typedef struct { + struct jpeg_color_quantizer pub; /* public fields */ + + /* Initially allocated colormap is saved here */ + JSAMPARRAY sv_colormap; /* The color map as a 2-D pixel array */ + int sv_actual; /* number of entries in use */ + + JSAMPARRAY colorindex; /* Precomputed mapping for speed */ + /* colorindex[i][j] = index of color closest to pixel value j in component i, + * premultiplied as described above. Since colormap indexes must fit into + * JSAMPLEs, the entries of this array will too. + */ + boolean is_padded; /* is the colorindex padded for odither? */ + + int Ncolors[MAX_Q_COMPS]; /* # of values alloced to each component */ + + /* Variables for ordered dithering */ + int row_index; /* cur row's vertical index in dither matrix */ + ODITHER_MATRIX_PTR odither[MAX_Q_COMPS]; /* one dither array per component */ + + /* Variables for Floyd-Steinberg dithering */ + FSERRPTR fserrors[MAX_Q_COMPS]; /* accumulated errors */ + boolean on_odd_row; /* flag to remember which row we are on */ +} my_cquantizer; + +typedef my_cquantizer * my_cquantize_ptr; + + +/* + * Policy-making subroutines for create_colormap and create_colorindex. + * These routines determine the colormap to be used. The rest of the module + * only assumes that the colormap is orthogonal. + * + * * select_ncolors decides how to divvy up the available colors + * among the components. + * * output_value defines the set of representative values for a component. + * * largest_input_value defines the mapping from input values to + * representative values for a component. + * Note that the latter two routines may impose different policies for + * different components, though this is not currently done. + */ + + +LOCAL(int) +select_ncolors (j_decompress_ptr cinfo, int Ncolors[]) +/* Determine allocation of desired colors to components, */ +/* and fill in Ncolors[] array to indicate choice. */ +/* Return value is total number of colors (product of Ncolors[] values). */ +{ + int nc = cinfo->out_color_components; /* number of color components */ + int max_colors = cinfo->desired_number_of_colors; + int total_colors, iroot, i, j; + boolean changed; + long temp; + int RGB_order[3] = { RGB_GREEN, RGB_RED, RGB_BLUE }; + RGB_order[0] = rgb_green[cinfo->out_color_space]; + RGB_order[1] = rgb_red[cinfo->out_color_space]; + RGB_order[2] = rgb_blue[cinfo->out_color_space]; + + /* We can allocate at least the nc'th root of max_colors per component. */ + /* Compute floor(nc'th root of max_colors). */ + iroot = 1; + do { + iroot++; + temp = iroot; /* set temp = iroot ** nc */ + for (i = 1; i < nc; i++) + temp *= iroot; + } while (temp <= (long) max_colors); /* repeat till iroot exceeds root */ + iroot--; /* now iroot = floor(root) */ + + /* Must have at least 2 color values per component */ + if (iroot < 2) + ERREXIT1(cinfo, JERR_QUANT_FEW_COLORS, (int) temp); + + /* Initialize to iroot color values for each component */ + total_colors = 1; + for (i = 0; i < nc; i++) { + Ncolors[i] = iroot; + total_colors *= iroot; + } + /* We may be able to increment the count for one or more components without + * exceeding max_colors, though we know not all can be incremented. + * Sometimes, the first component can be incremented more than once! + * (Example: for 16 colors, we start at 2*2*2, go to 3*2*2, then 4*2*2.) + * In RGB colorspace, try to increment G first, then R, then B. + */ + do { + changed = FALSE; + for (i = 0; i < nc; i++) { + j = (cinfo->out_color_space == JCS_RGB ? RGB_order[i] : i); + /* calculate new total_colors if Ncolors[j] is incremented */ + temp = total_colors / Ncolors[j]; + temp *= Ncolors[j]+1; /* done in long arith to avoid oflo */ + if (temp > (long) max_colors) + break; /* won't fit, done with this pass */ + Ncolors[j]++; /* OK, apply the increment */ + total_colors = (int) temp; + changed = TRUE; + } + } while (changed); + + return total_colors; +} + + +LOCAL(int) +output_value (j_decompress_ptr cinfo, int ci, int j, int maxj) +/* Return j'th output value, where j will range from 0 to maxj */ +/* The output values must fall in 0..MAXJSAMPLE in increasing order */ +{ + /* We always provide values 0 and MAXJSAMPLE for each component; + * any additional values are equally spaced between these limits. + * (Forcing the upper and lower values to the limits ensures that + * dithering can't produce a color outside the selected gamut.) + */ + return (int) (((INT32) j * MAXJSAMPLE + maxj/2) / maxj); +} + + +LOCAL(int) +largest_input_value (j_decompress_ptr cinfo, int ci, int j, int maxj) +/* Return largest input value that should map to j'th output value */ +/* Must have largest(j=0) >= 0, and largest(j=maxj) >= MAXJSAMPLE */ +{ + /* Breakpoints are halfway between values returned by output_value */ + return (int) (((INT32) (2*j + 1) * MAXJSAMPLE + maxj) / (2*maxj)); +} + + +/* + * Create the colormap. + */ + +LOCAL(void) +create_colormap (j_decompress_ptr cinfo) +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + JSAMPARRAY colormap; /* Created colormap */ + int total_colors; /* Number of distinct output colors */ + int i,j,k, nci, blksize, blkdist, ptr, val; + + /* Select number of colors for each component */ + total_colors = select_ncolors(cinfo, cquantize->Ncolors); + + /* Report selected color counts */ + if (cinfo->out_color_components == 3) + TRACEMS4(cinfo, 1, JTRC_QUANT_3_NCOLORS, + total_colors, cquantize->Ncolors[0], + cquantize->Ncolors[1], cquantize->Ncolors[2]); + else + TRACEMS1(cinfo, 1, JTRC_QUANT_NCOLORS, total_colors); + + /* Allocate and fill in the colormap. */ + /* The colors are ordered in the map in standard row-major order, */ + /* i.e. rightmost (highest-indexed) color changes most rapidly. */ + + colormap = (*cinfo->mem->alloc_sarray) + ((j_common_ptr) cinfo, JPOOL_IMAGE, + (JDIMENSION) total_colors, (JDIMENSION) cinfo->out_color_components); + + /* blksize is number of adjacent repeated entries for a component */ + /* blkdist is distance between groups of identical entries for a component */ + blkdist = total_colors; + + for (i = 0; i < cinfo->out_color_components; i++) { + /* fill in colormap entries for i'th color component */ + nci = cquantize->Ncolors[i]; /* # of distinct values for this color */ + blksize = blkdist / nci; + for (j = 0; j < nci; j++) { + /* Compute j'th output value (out of nci) for component */ + val = output_value(cinfo, i, j, nci-1); + /* Fill in all colormap entries that have this value of this component */ + for (ptr = j * blksize; ptr < total_colors; ptr += blkdist) { + /* fill in blksize entries beginning at ptr */ + for (k = 0; k < blksize; k++) + colormap[i][ptr+k] = (JSAMPLE) val; + } + } + blkdist = blksize; /* blksize of this color is blkdist of next */ + } + + /* Save the colormap in private storage, + * where it will survive color quantization mode changes. + */ + cquantize->sv_colormap = colormap; + cquantize->sv_actual = total_colors; +} + + +/* + * Create the color index table. + */ + +LOCAL(void) +create_colorindex (j_decompress_ptr cinfo) +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + JSAMPROW indexptr; + int i,j,k, nci, blksize, val, pad; + + /* For ordered dither, we pad the color index tables by MAXJSAMPLE in + * each direction (input index values can be -MAXJSAMPLE .. 2*MAXJSAMPLE). + * This is not necessary in the other dithering modes. However, we + * flag whether it was done in case user changes dithering mode. + */ + if (cinfo->dither_mode == JDITHER_ORDERED) { + pad = MAXJSAMPLE*2; + cquantize->is_padded = TRUE; + } else { + pad = 0; + cquantize->is_padded = FALSE; + } + + cquantize->colorindex = (*cinfo->mem->alloc_sarray) + ((j_common_ptr) cinfo, JPOOL_IMAGE, + (JDIMENSION) (MAXJSAMPLE+1 + pad), + (JDIMENSION) cinfo->out_color_components); + + /* blksize is number of adjacent repeated entries for a component */ + blksize = cquantize->sv_actual; + + for (i = 0; i < cinfo->out_color_components; i++) { + /* fill in colorindex entries for i'th color component */ + nci = cquantize->Ncolors[i]; /* # of distinct values for this color */ + blksize = blksize / nci; + + /* adjust colorindex pointers to provide padding at negative indexes. */ + if (pad) + cquantize->colorindex[i] += MAXJSAMPLE; + + /* in loop, val = index of current output value, */ + /* and k = largest j that maps to current val */ + indexptr = cquantize->colorindex[i]; + val = 0; + k = largest_input_value(cinfo, i, 0, nci-1); + for (j = 0; j <= MAXJSAMPLE; j++) { + while (j > k) /* advance val if past boundary */ + k = largest_input_value(cinfo, i, ++val, nci-1); + /* premultiply so that no multiplication needed in main processing */ + indexptr[j] = (JSAMPLE) (val * blksize); + } + /* Pad at both ends if necessary */ + if (pad) + for (j = 1; j <= MAXJSAMPLE; j++) { + indexptr[-j] = indexptr[0]; + indexptr[MAXJSAMPLE+j] = indexptr[MAXJSAMPLE]; + } + } +} + + +/* + * Create an ordered-dither array for a component having ncolors + * distinct output values. + */ + +LOCAL(ODITHER_MATRIX_PTR) +make_odither_array (j_decompress_ptr cinfo, int ncolors) +{ + ODITHER_MATRIX_PTR odither; + int j,k; + INT32 num,den; + + odither = (ODITHER_MATRIX_PTR) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(ODITHER_MATRIX)); + /* The inter-value distance for this color is MAXJSAMPLE/(ncolors-1). + * Hence the dither value for the matrix cell with fill order f + * (f=0..N-1) should be (N-1-2*f)/(2*N) * MAXJSAMPLE/(ncolors-1). + * On 16-bit-int machine, be careful to avoid overflow. + */ + den = 2 * ODITHER_CELLS * ((INT32) (ncolors - 1)); + for (j = 0; j < ODITHER_SIZE; j++) { + for (k = 0; k < ODITHER_SIZE; k++) { + num = ((INT32) (ODITHER_CELLS-1 - 2*((int)base_dither_matrix[j][k]))) + * MAXJSAMPLE; + /* Ensure round towards zero despite C's lack of consistency + * about rounding negative values in integer division... + */ + odither[j][k] = (int) (num<0 ? -((-num)/den) : num/den); + } + } + return odither; +} + + +/* + * Create the ordered-dither tables. + * Components having the same number of representative colors may + * share a dither table. + */ + +LOCAL(void) +create_odither_tables (j_decompress_ptr cinfo) +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + ODITHER_MATRIX_PTR odither; + int i, j, nci; + + for (i = 0; i < cinfo->out_color_components; i++) { + nci = cquantize->Ncolors[i]; /* # of distinct values for this color */ + odither = NULL; /* search for matching prior component */ + for (j = 0; j < i; j++) { + if (nci == cquantize->Ncolors[j]) { + odither = cquantize->odither[j]; + break; + } + } + if (odither == NULL) /* need a new table? */ + odither = make_odither_array(cinfo, nci); + cquantize->odither[i] = odither; + } +} + + +/* + * Map some rows of pixels to the output colormapped representation. + */ + +METHODDEF(void) +color_quantize (j_decompress_ptr cinfo, JSAMPARRAY input_buf, + JSAMPARRAY output_buf, int num_rows) +/* General case, no dithering */ +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + JSAMPARRAY colorindex = cquantize->colorindex; + register int pixcode, ci; + register JSAMPROW ptrin, ptrout; + int row; + JDIMENSION col; + JDIMENSION width = cinfo->output_width; + register int nc = cinfo->out_color_components; + + for (row = 0; row < num_rows; row++) { + ptrin = input_buf[row]; + ptrout = output_buf[row]; + for (col = width; col > 0; col--) { + pixcode = 0; + for (ci = 0; ci < nc; ci++) { + pixcode += GETJSAMPLE(colorindex[ci][GETJSAMPLE(*ptrin++)]); + } + *ptrout++ = (JSAMPLE) pixcode; + } + } +} + + +METHODDEF(void) +color_quantize3 (j_decompress_ptr cinfo, JSAMPARRAY input_buf, + JSAMPARRAY output_buf, int num_rows) +/* Fast path for out_color_components==3, no dithering */ +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + register int pixcode; + register JSAMPROW ptrin, ptrout; + JSAMPROW colorindex0 = cquantize->colorindex[0]; + JSAMPROW colorindex1 = cquantize->colorindex[1]; + JSAMPROW colorindex2 = cquantize->colorindex[2]; + int row; + JDIMENSION col; + JDIMENSION width = cinfo->output_width; + + for (row = 0; row < num_rows; row++) { + ptrin = input_buf[row]; + ptrout = output_buf[row]; + for (col = width; col > 0; col--) { + pixcode = GETJSAMPLE(colorindex0[GETJSAMPLE(*ptrin++)]); + pixcode += GETJSAMPLE(colorindex1[GETJSAMPLE(*ptrin++)]); + pixcode += GETJSAMPLE(colorindex2[GETJSAMPLE(*ptrin++)]); + *ptrout++ = (JSAMPLE) pixcode; + } + } +} + + +METHODDEF(void) +quantize_ord_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf, + JSAMPARRAY output_buf, int num_rows) +/* General case, with ordered dithering */ +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + register JSAMPROW input_ptr; + register JSAMPROW output_ptr; + JSAMPROW colorindex_ci; + int * dither; /* points to active row of dither matrix */ + int row_index, col_index; /* current indexes into dither matrix */ + int nc = cinfo->out_color_components; + int ci; + int row; + JDIMENSION col; + JDIMENSION width = cinfo->output_width; + + for (row = 0; row < num_rows; row++) { + /* Initialize output values to 0 so can process components separately */ + jzero_far((void *) output_buf[row], (size_t) (width * sizeof(JSAMPLE))); + row_index = cquantize->row_index; + for (ci = 0; ci < nc; ci++) { + input_ptr = input_buf[row] + ci; + output_ptr = output_buf[row]; + colorindex_ci = cquantize->colorindex[ci]; + dither = cquantize->odither[ci][row_index]; + col_index = 0; + + for (col = width; col > 0; col--) { + /* Form pixel value + dither, range-limit to 0..MAXJSAMPLE, + * select output value, accumulate into output code for this pixel. + * Range-limiting need not be done explicitly, as we have extended + * the colorindex table to produce the right answers for out-of-range + * inputs. The maximum dither is +- MAXJSAMPLE; this sets the + * required amount of padding. + */ + *output_ptr += colorindex_ci[GETJSAMPLE(*input_ptr)+dither[col_index]]; + input_ptr += nc; + output_ptr++; + col_index = (col_index + 1) & ODITHER_MASK; + } + } + /* Advance row index for next row */ + row_index = (row_index + 1) & ODITHER_MASK; + cquantize->row_index = row_index; + } +} + + +METHODDEF(void) +quantize3_ord_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf, + JSAMPARRAY output_buf, int num_rows) +/* Fast path for out_color_components==3, with ordered dithering */ +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + register int pixcode; + register JSAMPROW input_ptr; + register JSAMPROW output_ptr; + JSAMPROW colorindex0 = cquantize->colorindex[0]; + JSAMPROW colorindex1 = cquantize->colorindex[1]; + JSAMPROW colorindex2 = cquantize->colorindex[2]; + int * dither0; /* points to active row of dither matrix */ + int * dither1; + int * dither2; + int row_index, col_index; /* current indexes into dither matrix */ + int row; + JDIMENSION col; + JDIMENSION width = cinfo->output_width; + + for (row = 0; row < num_rows; row++) { + row_index = cquantize->row_index; + input_ptr = input_buf[row]; + output_ptr = output_buf[row]; + dither0 = cquantize->odither[0][row_index]; + dither1 = cquantize->odither[1][row_index]; + dither2 = cquantize->odither[2][row_index]; + col_index = 0; + + for (col = width; col > 0; col--) { + pixcode = GETJSAMPLE(colorindex0[GETJSAMPLE(*input_ptr++) + + dither0[col_index]]); + pixcode += GETJSAMPLE(colorindex1[GETJSAMPLE(*input_ptr++) + + dither1[col_index]]); + pixcode += GETJSAMPLE(colorindex2[GETJSAMPLE(*input_ptr++) + + dither2[col_index]]); + *output_ptr++ = (JSAMPLE) pixcode; + col_index = (col_index + 1) & ODITHER_MASK; + } + row_index = (row_index + 1) & ODITHER_MASK; + cquantize->row_index = row_index; + } +} + + +METHODDEF(void) +quantize_fs_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf, + JSAMPARRAY output_buf, int num_rows) +/* General case, with Floyd-Steinberg dithering */ +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + register LOCFSERROR cur; /* current error or pixel value */ + LOCFSERROR belowerr; /* error for pixel below cur */ + LOCFSERROR bpreverr; /* error for below/prev col */ + LOCFSERROR bnexterr; /* error for below/next col */ + LOCFSERROR delta; + register FSERRPTR errorptr; /* => fserrors[] at column before current */ + register JSAMPROW input_ptr; + register JSAMPROW output_ptr; + JSAMPROW colorindex_ci; + JSAMPROW colormap_ci; + int pixcode; + int nc = cinfo->out_color_components; + int dir; /* 1 for left-to-right, -1 for right-to-left */ + int dirnc; /* dir * nc */ + int ci; + int row; + JDIMENSION col; + JDIMENSION width = cinfo->output_width; + JSAMPLE *range_limit = cinfo->sample_range_limit; + SHIFT_TEMPS + + for (row = 0; row < num_rows; row++) { + /* Initialize output values to 0 so can process components separately */ + jzero_far((void *) output_buf[row], (size_t) (width * sizeof(JSAMPLE))); + for (ci = 0; ci < nc; ci++) { + input_ptr = input_buf[row] + ci; + output_ptr = output_buf[row]; + if (cquantize->on_odd_row) { + /* work right to left in this row */ + input_ptr += (width-1) * nc; /* so point to rightmost pixel */ + output_ptr += width-1; + dir = -1; + dirnc = -nc; + errorptr = cquantize->fserrors[ci] + (width+1); /* => entry after last column */ + } else { + /* work left to right in this row */ + dir = 1; + dirnc = nc; + errorptr = cquantize->fserrors[ci]; /* => entry before first column */ + } + colorindex_ci = cquantize->colorindex[ci]; + colormap_ci = cquantize->sv_colormap[ci]; + /* Preset error values: no error propagated to first pixel from left */ + cur = 0; + /* and no error propagated to row below yet */ + belowerr = bpreverr = 0; + + for (col = width; col > 0; col--) { + /* cur holds the error propagated from the previous pixel on the + * current line. Add the error propagated from the previous line + * to form the complete error correction term for this pixel, and + * round the error term (which is expressed * 16) to an integer. + * RIGHT_SHIFT rounds towards minus infinity, so adding 8 is correct + * for either sign of the error value. + * Note: errorptr points to *previous* column's array entry. + */ + cur = RIGHT_SHIFT(cur + errorptr[dir] + 8, 4); + /* Form pixel value + error, and range-limit to 0..MAXJSAMPLE. + * The maximum error is +- MAXJSAMPLE; this sets the required size + * of the range_limit array. + */ + cur += GETJSAMPLE(*input_ptr); + cur = GETJSAMPLE(range_limit[cur]); + /* Select output value, accumulate into output code for this pixel */ + pixcode = GETJSAMPLE(colorindex_ci[cur]); + *output_ptr += (JSAMPLE) pixcode; + /* Compute actual representation error at this pixel */ + /* Note: we can do this even though we don't have the final */ + /* pixel code, because the colormap is orthogonal. */ + cur -= GETJSAMPLE(colormap_ci[pixcode]); + /* Compute error fractions to be propagated to adjacent pixels. + * Add these into the running sums, and simultaneously shift the + * next-line error sums left by 1 column. + */ + bnexterr = cur; + delta = cur * 2; + cur += delta; /* form error * 3 */ + errorptr[0] = (FSERROR) (bpreverr + cur); + cur += delta; /* form error * 5 */ + bpreverr = belowerr + cur; + belowerr = bnexterr; + cur += delta; /* form error * 7 */ + /* At this point cur contains the 7/16 error value to be propagated + * to the next pixel on the current line, and all the errors for the + * next line have been shifted over. We are therefore ready to move on. + */ + input_ptr += dirnc; /* advance input ptr to next column */ + output_ptr += dir; /* advance output ptr to next column */ + errorptr += dir; /* advance errorptr to current column */ + } + /* Post-loop cleanup: we must unload the final error value into the + * final fserrors[] entry. Note we need not unload belowerr because + * it is for the dummy column before or after the actual array. + */ + errorptr[0] = (FSERROR) bpreverr; /* unload prev err into array */ + } + cquantize->on_odd_row = (cquantize->on_odd_row ? FALSE : TRUE); + } +} + + +/* + * Allocate workspace for Floyd-Steinberg errors. + */ + +LOCAL(void) +alloc_fs_workspace (j_decompress_ptr cinfo) +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + size_t arraysize; + int i; + + arraysize = (size_t) ((cinfo->output_width + 2) * sizeof(FSERROR)); + for (i = 0; i < cinfo->out_color_components; i++) { + cquantize->fserrors[i] = (FSERRPTR) + (*cinfo->mem->alloc_large)((j_common_ptr) cinfo, JPOOL_IMAGE, arraysize); + } +} + + +/* + * Initialize for one-pass color quantization. + */ + +METHODDEF(void) +start_pass_1_quant (j_decompress_ptr cinfo, boolean is_pre_scan) +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + size_t arraysize; + int i; + + /* Install my colormap. */ + cinfo->colormap = cquantize->sv_colormap; + cinfo->actual_number_of_colors = cquantize->sv_actual; + + /* Initialize for desired dithering mode. */ + switch (cinfo->dither_mode) { + case JDITHER_NONE: + if (cinfo->out_color_components == 3) + cquantize->pub.color_quantize = color_quantize3; + else + cquantize->pub.color_quantize = color_quantize; + break; + case JDITHER_ORDERED: + if (cinfo->out_color_components == 3) + cquantize->pub.color_quantize = quantize3_ord_dither; + else + cquantize->pub.color_quantize = quantize_ord_dither; + cquantize->row_index = 0; /* initialize state for ordered dither */ + /* If user changed to ordered dither from another mode, + * we must recreate the color index table with padding. + * This will cost extra space, but probably isn't very likely. + */ + if (! cquantize->is_padded) + create_colorindex(cinfo); + /* Create ordered-dither tables if we didn't already. */ + if (cquantize->odither[0] == NULL) + create_odither_tables(cinfo); + break; + case JDITHER_FS: + cquantize->pub.color_quantize = quantize_fs_dither; + cquantize->on_odd_row = FALSE; /* initialize state for F-S dither */ + /* Allocate Floyd-Steinberg workspace if didn't already. */ + if (cquantize->fserrors[0] == NULL) + alloc_fs_workspace(cinfo); + /* Initialize the propagated errors to zero. */ + arraysize = (size_t) ((cinfo->output_width + 2) * sizeof(FSERROR)); + for (i = 0; i < cinfo->out_color_components; i++) + jzero_far((void *) cquantize->fserrors[i], arraysize); + break; + default: + ERREXIT(cinfo, JERR_NOT_COMPILED); + break; + } +} + + +/* + * Finish up at the end of the pass. + */ + +METHODDEF(void) +finish_pass_1_quant (j_decompress_ptr cinfo) +{ + /* no work in 1-pass case */ +} + + +/* + * Switch to a new external colormap between output passes. + * Shouldn't get to this module! + */ + +METHODDEF(void) +new_color_map_1_quant (j_decompress_ptr cinfo) +{ + ERREXIT(cinfo, JERR_MODE_CHANGE); +} + + +/* + * Module initialization routine for 1-pass color quantization. + */ + +GLOBAL(void) +jinit_1pass_quantizer (j_decompress_ptr cinfo) +{ + my_cquantize_ptr cquantize; + + cquantize = (my_cquantize_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(my_cquantizer)); + cinfo->cquantize = (struct jpeg_color_quantizer *) cquantize; + cquantize->pub.start_pass = start_pass_1_quant; + cquantize->pub.finish_pass = finish_pass_1_quant; + cquantize->pub.new_color_map = new_color_map_1_quant; + cquantize->fserrors[0] = NULL; /* Flag FS workspace not allocated */ + cquantize->odither[0] = NULL; /* Also flag odither arrays not allocated */ + + /* Make sure my internal arrays won't overflow */ + if (cinfo->out_color_components > MAX_Q_COMPS) + ERREXIT1(cinfo, JERR_QUANT_COMPONENTS, MAX_Q_COMPS); + /* Make sure colormap indexes can be represented by JSAMPLEs */ + if (cinfo->desired_number_of_colors > (MAXJSAMPLE+1)) + ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, MAXJSAMPLE+1); + + /* Create the colormap and color index table. */ + create_colormap(cinfo); + create_colorindex(cinfo); + + /* Allocate Floyd-Steinberg workspace now if requested. + * We do this now since it may affect the memory manager's space + * calculations. If the user changes to FS dither mode in a later pass, we + * will allocate the space then, and will possibly overrun the + * max_memory_to_use setting. + */ + if (cinfo->dither_mode == JDITHER_FS) + alloc_fs_workspace(cinfo); +} + +#endif /* QUANT_1PASS_SUPPORTED */ diff --git a/app/src/main/jni/libjpeg-turbo/jquant2.c b/app/src/main/jni/libjpeg-turbo/jquant2.c new file mode 100644 index 000000000..291b4f1b8 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jquant2.c @@ -0,0 +1,1281 @@ +/* + * jquant2.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1996, Thomas G. Lane. + * libjpeg-turbo Modifications: + * Copyright (C) 2009, 2014, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains 2-pass color quantization (color mapping) routines. + * These routines provide selection of a custom color map for an image, + * followed by mapping of the image to that color map, with optional + * Floyd-Steinberg dithering. + * It is also possible to use just the second pass to map to an arbitrary + * externally-given color map. + * + * Note: ordered dithering is not supported, since there isn't any fast + * way to compute intercolor distances; it's unclear that ordered dither's + * fundamental assumptions even hold with an irregularly spaced color map. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" + +#ifdef QUANT_2PASS_SUPPORTED + + +/* + * This module implements the well-known Heckbert paradigm for color + * quantization. Most of the ideas used here can be traced back to + * Heckbert's seminal paper + * Heckbert, Paul. "Color Image Quantization for Frame Buffer Display", + * Proc. SIGGRAPH '82, Computer Graphics v.16 #3 (July 1982), pp 297-304. + * + * In the first pass over the image, we accumulate a histogram showing the + * usage count of each possible color. To keep the histogram to a reasonable + * size, we reduce the precision of the input; typical practice is to retain + * 5 or 6 bits per color, so that 8 or 4 different input values are counted + * in the same histogram cell. + * + * Next, the color-selection step begins with a box representing the whole + * color space, and repeatedly splits the "largest" remaining box until we + * have as many boxes as desired colors. Then the mean color in each + * remaining box becomes one of the possible output colors. + * + * The second pass over the image maps each input pixel to the closest output + * color (optionally after applying a Floyd-Steinberg dithering correction). + * This mapping is logically trivial, but making it go fast enough requires + * considerable care. + * + * Heckbert-style quantizers vary a good deal in their policies for choosing + * the "largest" box and deciding where to cut it. The particular policies + * used here have proved out well in experimental comparisons, but better ones + * may yet be found. + * + * In earlier versions of the IJG code, this module quantized in YCbCr color + * space, processing the raw upsampled data without a color conversion step. + * This allowed the color conversion math to be done only once per colormap + * entry, not once per pixel. However, that optimization precluded other + * useful optimizations (such as merging color conversion with upsampling) + * and it also interfered with desired capabilities such as quantizing to an + * externally-supplied colormap. We have therefore abandoned that approach. + * The present code works in the post-conversion color space, typically RGB. + * + * To improve the visual quality of the results, we actually work in scaled + * RGB space, giving G distances more weight than R, and R in turn more than + * B. To do everything in integer math, we must use integer scale factors. + * The 2/3/1 scale factors used here correspond loosely to the relative + * weights of the colors in the NTSC grayscale equation. + * If you want to use this code to quantize a non-RGB color space, you'll + * probably need to change these scale factors. + */ + +#define R_SCALE 2 /* scale R distances by this much */ +#define G_SCALE 3 /* scale G distances by this much */ +#define B_SCALE 1 /* and B by this much */ + +static const int c_scales[3]={R_SCALE, G_SCALE, B_SCALE}; +#define C0_SCALE c_scales[rgb_red[cinfo->out_color_space]] +#define C1_SCALE c_scales[rgb_green[cinfo->out_color_space]] +#define C2_SCALE c_scales[rgb_blue[cinfo->out_color_space]] + +/* + * First we have the histogram data structure and routines for creating it. + * + * The number of bits of precision can be adjusted by changing these symbols. + * We recommend keeping 6 bits for G and 5 each for R and B. + * If you have plenty of memory and cycles, 6 bits all around gives marginally + * better results; if you are short of memory, 5 bits all around will save + * some space but degrade the results. + * To maintain a fully accurate histogram, we'd need to allocate a "long" + * (preferably unsigned long) for each cell. In practice this is overkill; + * we can get by with 16 bits per cell. Few of the cell counts will overflow, + * and clamping those that do overflow to the maximum value will give close- + * enough results. This reduces the recommended histogram size from 256Kb + * to 128Kb, which is a useful savings on PC-class machines. + * (In the second pass the histogram space is re-used for pixel mapping data; + * in that capacity, each cell must be able to store zero to the number of + * desired colors. 16 bits/cell is plenty for that too.) + * Since the JPEG code is intended to run in small memory model on 80x86 + * machines, we can't just allocate the histogram in one chunk. Instead + * of a true 3-D array, we use a row of pointers to 2-D arrays. Each + * pointer corresponds to a C0 value (typically 2^5 = 32 pointers) and + * each 2-D array has 2^6*2^5 = 2048 or 2^6*2^6 = 4096 entries. + */ + +#define MAXNUMCOLORS (MAXJSAMPLE+1) /* maximum size of colormap */ + +/* These will do the right thing for either R,G,B or B,G,R color order, + * but you may not like the results for other color orders. + */ +#define HIST_C0_BITS 5 /* bits of precision in R/B histogram */ +#define HIST_C1_BITS 6 /* bits of precision in G histogram */ +#define HIST_C2_BITS 5 /* bits of precision in B/R histogram */ + +/* Number of elements along histogram axes. */ +#define HIST_C0_ELEMS (1<cquantize; + register JSAMPROW ptr; + register histptr histp; + register hist3d histogram = cquantize->histogram; + int row; + JDIMENSION col; + JDIMENSION width = cinfo->output_width; + + for (row = 0; row < num_rows; row++) { + ptr = input_buf[row]; + for (col = width; col > 0; col--) { + /* get pixel value and index into the histogram */ + histp = & histogram[GETJSAMPLE(ptr[0]) >> C0_SHIFT] + [GETJSAMPLE(ptr[1]) >> C1_SHIFT] + [GETJSAMPLE(ptr[2]) >> C2_SHIFT]; + /* increment, check for overflow and undo increment if so. */ + if (++(*histp) <= 0) + (*histp)--; + ptr += 3; + } + } +} + + +/* + * Next we have the really interesting routines: selection of a colormap + * given the completed histogram. + * These routines work with a list of "boxes", each representing a rectangular + * subset of the input color space (to histogram precision). + */ + +typedef struct { + /* The bounds of the box (inclusive); expressed as histogram indexes */ + int c0min, c0max; + int c1min, c1max; + int c2min, c2max; + /* The volume (actually 2-norm) of the box */ + INT32 volume; + /* The number of nonzero histogram cells within this box */ + long colorcount; +} box; + +typedef box * boxptr; + + +LOCAL(boxptr) +find_biggest_color_pop (boxptr boxlist, int numboxes) +/* Find the splittable box with the largest color population */ +/* Returns NULL if no splittable boxes remain */ +{ + register boxptr boxp; + register int i; + register long maxc = 0; + boxptr which = NULL; + + for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) { + if (boxp->colorcount > maxc && boxp->volume > 0) { + which = boxp; + maxc = boxp->colorcount; + } + } + return which; +} + + +LOCAL(boxptr) +find_biggest_volume (boxptr boxlist, int numboxes) +/* Find the splittable box with the largest (scaled) volume */ +/* Returns NULL if no splittable boxes remain */ +{ + register boxptr boxp; + register int i; + register INT32 maxv = 0; + boxptr which = NULL; + + for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) { + if (boxp->volume > maxv) { + which = boxp; + maxv = boxp->volume; + } + } + return which; +} + + +LOCAL(void) +update_box (j_decompress_ptr cinfo, boxptr boxp) +/* Shrink the min/max bounds of a box to enclose only nonzero elements, */ +/* and recompute its volume and population */ +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + hist3d histogram = cquantize->histogram; + histptr histp; + int c0,c1,c2; + int c0min,c0max,c1min,c1max,c2min,c2max; + INT32 dist0,dist1,dist2; + long ccount; + + c0min = boxp->c0min; c0max = boxp->c0max; + c1min = boxp->c1min; c1max = boxp->c1max; + c2min = boxp->c2min; c2max = boxp->c2max; + + if (c0max > c0min) + for (c0 = c0min; c0 <= c0max; c0++) + for (c1 = c1min; c1 <= c1max; c1++) { + histp = & histogram[c0][c1][c2min]; + for (c2 = c2min; c2 <= c2max; c2++) + if (*histp++ != 0) { + boxp->c0min = c0min = c0; + goto have_c0min; + } + } + have_c0min: + if (c0max > c0min) + for (c0 = c0max; c0 >= c0min; c0--) + for (c1 = c1min; c1 <= c1max; c1++) { + histp = & histogram[c0][c1][c2min]; + for (c2 = c2min; c2 <= c2max; c2++) + if (*histp++ != 0) { + boxp->c0max = c0max = c0; + goto have_c0max; + } + } + have_c0max: + if (c1max > c1min) + for (c1 = c1min; c1 <= c1max; c1++) + for (c0 = c0min; c0 <= c0max; c0++) { + histp = & histogram[c0][c1][c2min]; + for (c2 = c2min; c2 <= c2max; c2++) + if (*histp++ != 0) { + boxp->c1min = c1min = c1; + goto have_c1min; + } + } + have_c1min: + if (c1max > c1min) + for (c1 = c1max; c1 >= c1min; c1--) + for (c0 = c0min; c0 <= c0max; c0++) { + histp = & histogram[c0][c1][c2min]; + for (c2 = c2min; c2 <= c2max; c2++) + if (*histp++ != 0) { + boxp->c1max = c1max = c1; + goto have_c1max; + } + } + have_c1max: + if (c2max > c2min) + for (c2 = c2min; c2 <= c2max; c2++) + for (c0 = c0min; c0 <= c0max; c0++) { + histp = & histogram[c0][c1min][c2]; + for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS) + if (*histp != 0) { + boxp->c2min = c2min = c2; + goto have_c2min; + } + } + have_c2min: + if (c2max > c2min) + for (c2 = c2max; c2 >= c2min; c2--) + for (c0 = c0min; c0 <= c0max; c0++) { + histp = & histogram[c0][c1min][c2]; + for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS) + if (*histp != 0) { + boxp->c2max = c2max = c2; + goto have_c2max; + } + } + have_c2max: + + /* Update box volume. + * We use 2-norm rather than real volume here; this biases the method + * against making long narrow boxes, and it has the side benefit that + * a box is splittable iff norm > 0. + * Since the differences are expressed in histogram-cell units, + * we have to shift back to JSAMPLE units to get consistent distances; + * after which, we scale according to the selected distance scale factors. + */ + dist0 = ((c0max - c0min) << C0_SHIFT) * C0_SCALE; + dist1 = ((c1max - c1min) << C1_SHIFT) * C1_SCALE; + dist2 = ((c2max - c2min) << C2_SHIFT) * C2_SCALE; + boxp->volume = dist0*dist0 + dist1*dist1 + dist2*dist2; + + /* Now scan remaining volume of box and compute population */ + ccount = 0; + for (c0 = c0min; c0 <= c0max; c0++) + for (c1 = c1min; c1 <= c1max; c1++) { + histp = & histogram[c0][c1][c2min]; + for (c2 = c2min; c2 <= c2max; c2++, histp++) + if (*histp != 0) { + ccount++; + } + } + boxp->colorcount = ccount; +} + + +LOCAL(int) +median_cut (j_decompress_ptr cinfo, boxptr boxlist, int numboxes, + int desired_colors) +/* Repeatedly select and split the largest box until we have enough boxes */ +{ + int n,lb; + int c0,c1,c2,cmax; + register boxptr b1,b2; + + while (numboxes < desired_colors) { + /* Select box to split. + * Current algorithm: by population for first half, then by volume. + */ + if (numboxes*2 <= desired_colors) { + b1 = find_biggest_color_pop(boxlist, numboxes); + } else { + b1 = find_biggest_volume(boxlist, numboxes); + } + if (b1 == NULL) /* no splittable boxes left! */ + break; + b2 = &boxlist[numboxes]; /* where new box will go */ + /* Copy the color bounds to the new box. */ + b2->c0max = b1->c0max; b2->c1max = b1->c1max; b2->c2max = b1->c2max; + b2->c0min = b1->c0min; b2->c1min = b1->c1min; b2->c2min = b1->c2min; + /* Choose which axis to split the box on. + * Current algorithm: longest scaled axis. + * See notes in update_box about scaling distances. + */ + c0 = ((b1->c0max - b1->c0min) << C0_SHIFT) * C0_SCALE; + c1 = ((b1->c1max - b1->c1min) << C1_SHIFT) * C1_SCALE; + c2 = ((b1->c2max - b1->c2min) << C2_SHIFT) * C2_SCALE; + /* We want to break any ties in favor of green, then red, blue last. + * This code does the right thing for R,G,B or B,G,R color orders only. + */ + if (rgb_red[cinfo->out_color_space] == 0) { + cmax = c1; n = 1; + if (c0 > cmax) { cmax = c0; n = 0; } + if (c2 > cmax) { n = 2; } + } + else { + cmax = c1; n = 1; + if (c2 > cmax) { cmax = c2; n = 2; } + if (c0 > cmax) { n = 0; } + } + /* Choose split point along selected axis, and update box bounds. + * Current algorithm: split at halfway point. + * (Since the box has been shrunk to minimum volume, + * any split will produce two nonempty subboxes.) + * Note that lb value is max for lower box, so must be < old max. + */ + switch (n) { + case 0: + lb = (b1->c0max + b1->c0min) / 2; + b1->c0max = lb; + b2->c0min = lb+1; + break; + case 1: + lb = (b1->c1max + b1->c1min) / 2; + b1->c1max = lb; + b2->c1min = lb+1; + break; + case 2: + lb = (b1->c2max + b1->c2min) / 2; + b1->c2max = lb; + b2->c2min = lb+1; + break; + } + /* Update stats for boxes */ + update_box(cinfo, b1); + update_box(cinfo, b2); + numboxes++; + } + return numboxes; +} + + +LOCAL(void) +compute_color (j_decompress_ptr cinfo, boxptr boxp, int icolor) +/* Compute representative color for a box, put it in colormap[icolor] */ +{ + /* Current algorithm: mean weighted by pixels (not colors) */ + /* Note it is important to get the rounding correct! */ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + hist3d histogram = cquantize->histogram; + histptr histp; + int c0,c1,c2; + int c0min,c0max,c1min,c1max,c2min,c2max; + long count; + long total = 0; + long c0total = 0; + long c1total = 0; + long c2total = 0; + + c0min = boxp->c0min; c0max = boxp->c0max; + c1min = boxp->c1min; c1max = boxp->c1max; + c2min = boxp->c2min; c2max = boxp->c2max; + + for (c0 = c0min; c0 <= c0max; c0++) + for (c1 = c1min; c1 <= c1max; c1++) { + histp = & histogram[c0][c1][c2min]; + for (c2 = c2min; c2 <= c2max; c2++) { + if ((count = *histp++) != 0) { + total += count; + c0total += ((c0 << C0_SHIFT) + ((1<>1)) * count; + c1total += ((c1 << C1_SHIFT) + ((1<>1)) * count; + c2total += ((c2 << C2_SHIFT) + ((1<>1)) * count; + } + } + } + + cinfo->colormap[0][icolor] = (JSAMPLE) ((c0total + (total>>1)) / total); + cinfo->colormap[1][icolor] = (JSAMPLE) ((c1total + (total>>1)) / total); + cinfo->colormap[2][icolor] = (JSAMPLE) ((c2total + (total>>1)) / total); +} + + +LOCAL(void) +select_colors (j_decompress_ptr cinfo, int desired_colors) +/* Master routine for color selection */ +{ + boxptr boxlist; + int numboxes; + int i; + + /* Allocate workspace for box list */ + boxlist = (boxptr) (*cinfo->mem->alloc_small) + ((j_common_ptr) cinfo, JPOOL_IMAGE, desired_colors * sizeof(box)); + /* Initialize one box containing whole space */ + numboxes = 1; + boxlist[0].c0min = 0; + boxlist[0].c0max = MAXJSAMPLE >> C0_SHIFT; + boxlist[0].c1min = 0; + boxlist[0].c1max = MAXJSAMPLE >> C1_SHIFT; + boxlist[0].c2min = 0; + boxlist[0].c2max = MAXJSAMPLE >> C2_SHIFT; + /* Shrink it to actually-used volume and set its statistics */ + update_box(cinfo, & boxlist[0]); + /* Perform median-cut to produce final box list */ + numboxes = median_cut(cinfo, boxlist, numboxes, desired_colors); + /* Compute the representative color for each box, fill colormap */ + for (i = 0; i < numboxes; i++) + compute_color(cinfo, & boxlist[i], i); + cinfo->actual_number_of_colors = numboxes; + TRACEMS1(cinfo, 1, JTRC_QUANT_SELECTED, numboxes); +} + + +/* + * These routines are concerned with the time-critical task of mapping input + * colors to the nearest color in the selected colormap. + * + * We re-use the histogram space as an "inverse color map", essentially a + * cache for the results of nearest-color searches. All colors within a + * histogram cell will be mapped to the same colormap entry, namely the one + * closest to the cell's center. This may not be quite the closest entry to + * the actual input color, but it's almost as good. A zero in the cache + * indicates we haven't found the nearest color for that cell yet; the array + * is cleared to zeroes before starting the mapping pass. When we find the + * nearest color for a cell, its colormap index plus one is recorded in the + * cache for future use. The pass2 scanning routines call fill_inverse_cmap + * when they need to use an unfilled entry in the cache. + * + * Our method of efficiently finding nearest colors is based on the "locally + * sorted search" idea described by Heckbert and on the incremental distance + * calculation described by Spencer W. Thomas in chapter III.1 of Graphics + * Gems II (James Arvo, ed. Academic Press, 1991). Thomas points out that + * the distances from a given colormap entry to each cell of the histogram can + * be computed quickly using an incremental method: the differences between + * distances to adjacent cells themselves differ by a constant. This allows a + * fairly fast implementation of the "brute force" approach of computing the + * distance from every colormap entry to every histogram cell. Unfortunately, + * it needs a work array to hold the best-distance-so-far for each histogram + * cell (because the inner loop has to be over cells, not colormap entries). + * The work array elements have to be INT32s, so the work array would need + * 256Kb at our recommended precision. This is not feasible in DOS machines. + * + * To get around these problems, we apply Thomas' method to compute the + * nearest colors for only the cells within a small subbox of the histogram. + * The work array need be only as big as the subbox, so the memory usage + * problem is solved. Furthermore, we need not fill subboxes that are never + * referenced in pass2; many images use only part of the color gamut, so a + * fair amount of work is saved. An additional advantage of this + * approach is that we can apply Heckbert's locality criterion to quickly + * eliminate colormap entries that are far away from the subbox; typically + * three-fourths of the colormap entries are rejected by Heckbert's criterion, + * and we need not compute their distances to individual cells in the subbox. + * The speed of this approach is heavily influenced by the subbox size: too + * small means too much overhead, too big loses because Heckbert's criterion + * can't eliminate as many colormap entries. Empirically the best subbox + * size seems to be about 1/512th of the histogram (1/8th in each direction). + * + * Thomas' article also describes a refined method which is asymptotically + * faster than the brute-force method, but it is also far more complex and + * cannot efficiently be applied to small subboxes. It is therefore not + * useful for programs intended to be portable to DOS machines. On machines + * with plenty of memory, filling the whole histogram in one shot with Thomas' + * refined method might be faster than the present code --- but then again, + * it might not be any faster, and it's certainly more complicated. + */ + + +/* log2(histogram cells in update box) for each axis; this can be adjusted */ +#define BOX_C0_LOG (HIST_C0_BITS-3) +#define BOX_C1_LOG (HIST_C1_BITS-3) +#define BOX_C2_LOG (HIST_C2_BITS-3) + +#define BOX_C0_ELEMS (1<actual_number_of_colors; + int maxc0, maxc1, maxc2; + int centerc0, centerc1, centerc2; + int i, x, ncolors; + INT32 minmaxdist, min_dist, max_dist, tdist; + INT32 mindist[MAXNUMCOLORS]; /* min distance to colormap entry i */ + + /* Compute true coordinates of update box's upper corner and center. + * Actually we compute the coordinates of the center of the upper-corner + * histogram cell, which are the upper bounds of the volume we care about. + * Note that since ">>" rounds down, the "center" values may be closer to + * min than to max; hence comparisons to them must be "<=", not "<". + */ + maxc0 = minc0 + ((1 << BOX_C0_SHIFT) - (1 << C0_SHIFT)); + centerc0 = (minc0 + maxc0) >> 1; + maxc1 = minc1 + ((1 << BOX_C1_SHIFT) - (1 << C1_SHIFT)); + centerc1 = (minc1 + maxc1) >> 1; + maxc2 = minc2 + ((1 << BOX_C2_SHIFT) - (1 << C2_SHIFT)); + centerc2 = (minc2 + maxc2) >> 1; + + /* For each color in colormap, find: + * 1. its minimum squared-distance to any point in the update box + * (zero if color is within update box); + * 2. its maximum squared-distance to any point in the update box. + * Both of these can be found by considering only the corners of the box. + * We save the minimum distance for each color in mindist[]; + * only the smallest maximum distance is of interest. + */ + minmaxdist = 0x7FFFFFFFL; + + for (i = 0; i < numcolors; i++) { + /* We compute the squared-c0-distance term, then add in the other two. */ + x = GETJSAMPLE(cinfo->colormap[0][i]); + if (x < minc0) { + tdist = (x - minc0) * C0_SCALE; + min_dist = tdist*tdist; + tdist = (x - maxc0) * C0_SCALE; + max_dist = tdist*tdist; + } else if (x > maxc0) { + tdist = (x - maxc0) * C0_SCALE; + min_dist = tdist*tdist; + tdist = (x - minc0) * C0_SCALE; + max_dist = tdist*tdist; + } else { + /* within cell range so no contribution to min_dist */ + min_dist = 0; + if (x <= centerc0) { + tdist = (x - maxc0) * C0_SCALE; + max_dist = tdist*tdist; + } else { + tdist = (x - minc0) * C0_SCALE; + max_dist = tdist*tdist; + } + } + + x = GETJSAMPLE(cinfo->colormap[1][i]); + if (x < minc1) { + tdist = (x - minc1) * C1_SCALE; + min_dist += tdist*tdist; + tdist = (x - maxc1) * C1_SCALE; + max_dist += tdist*tdist; + } else if (x > maxc1) { + tdist = (x - maxc1) * C1_SCALE; + min_dist += tdist*tdist; + tdist = (x - minc1) * C1_SCALE; + max_dist += tdist*tdist; + } else { + /* within cell range so no contribution to min_dist */ + if (x <= centerc1) { + tdist = (x - maxc1) * C1_SCALE; + max_dist += tdist*tdist; + } else { + tdist = (x - minc1) * C1_SCALE; + max_dist += tdist*tdist; + } + } + + x = GETJSAMPLE(cinfo->colormap[2][i]); + if (x < minc2) { + tdist = (x - minc2) * C2_SCALE; + min_dist += tdist*tdist; + tdist = (x - maxc2) * C2_SCALE; + max_dist += tdist*tdist; + } else if (x > maxc2) { + tdist = (x - maxc2) * C2_SCALE; + min_dist += tdist*tdist; + tdist = (x - minc2) * C2_SCALE; + max_dist += tdist*tdist; + } else { + /* within cell range so no contribution to min_dist */ + if (x <= centerc2) { + tdist = (x - maxc2) * C2_SCALE; + max_dist += tdist*tdist; + } else { + tdist = (x - minc2) * C2_SCALE; + max_dist += tdist*tdist; + } + } + + mindist[i] = min_dist; /* save away the results */ + if (max_dist < minmaxdist) + minmaxdist = max_dist; + } + + /* Now we know that no cell in the update box is more than minmaxdist + * away from some colormap entry. Therefore, only colors that are + * within minmaxdist of some part of the box need be considered. + */ + ncolors = 0; + for (i = 0; i < numcolors; i++) { + if (mindist[i] <= minmaxdist) + colorlist[ncolors++] = (JSAMPLE) i; + } + return ncolors; +} + + +LOCAL(void) +find_best_colors (j_decompress_ptr cinfo, int minc0, int minc1, int minc2, + int numcolors, JSAMPLE colorlist[], JSAMPLE bestcolor[]) +/* Find the closest colormap entry for each cell in the update box, + * given the list of candidate colors prepared by find_nearby_colors. + * Return the indexes of the closest entries in the bestcolor[] array. + * This routine uses Thomas' incremental distance calculation method to + * find the distance from a colormap entry to successive cells in the box. + */ +{ + int ic0, ic1, ic2; + int i, icolor; + register INT32 * bptr; /* pointer into bestdist[] array */ + JSAMPLE * cptr; /* pointer into bestcolor[] array */ + INT32 dist0, dist1; /* initial distance values */ + register INT32 dist2; /* current distance in inner loop */ + INT32 xx0, xx1; /* distance increments */ + register INT32 xx2; + INT32 inc0, inc1, inc2; /* initial values for increments */ + /* This array holds the distance to the nearest-so-far color for each cell */ + INT32 bestdist[BOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS]; + + /* Initialize best-distance for each cell of the update box */ + bptr = bestdist; + for (i = BOX_C0_ELEMS*BOX_C1_ELEMS*BOX_C2_ELEMS-1; i >= 0; i--) + *bptr++ = 0x7FFFFFFFL; + + /* For each color selected by find_nearby_colors, + * compute its distance to the center of each cell in the box. + * If that's less than best-so-far, update best distance and color number. + */ + + /* Nominal steps between cell centers ("x" in Thomas article) */ +#define STEP_C0 ((1 << C0_SHIFT) * C0_SCALE) +#define STEP_C1 ((1 << C1_SHIFT) * C1_SCALE) +#define STEP_C2 ((1 << C2_SHIFT) * C2_SCALE) + + for (i = 0; i < numcolors; i++) { + icolor = GETJSAMPLE(colorlist[i]); + /* Compute (square of) distance from minc0/c1/c2 to this color */ + inc0 = (minc0 - GETJSAMPLE(cinfo->colormap[0][icolor])) * C0_SCALE; + dist0 = inc0*inc0; + inc1 = (minc1 - GETJSAMPLE(cinfo->colormap[1][icolor])) * C1_SCALE; + dist0 += inc1*inc1; + inc2 = (minc2 - GETJSAMPLE(cinfo->colormap[2][icolor])) * C2_SCALE; + dist0 += inc2*inc2; + /* Form the initial difference increments */ + inc0 = inc0 * (2 * STEP_C0) + STEP_C0 * STEP_C0; + inc1 = inc1 * (2 * STEP_C1) + STEP_C1 * STEP_C1; + inc2 = inc2 * (2 * STEP_C2) + STEP_C2 * STEP_C2; + /* Now loop over all cells in box, updating distance per Thomas method */ + bptr = bestdist; + cptr = bestcolor; + xx0 = inc0; + for (ic0 = BOX_C0_ELEMS-1; ic0 >= 0; ic0--) { + dist1 = dist0; + xx1 = inc1; + for (ic1 = BOX_C1_ELEMS-1; ic1 >= 0; ic1--) { + dist2 = dist1; + xx2 = inc2; + for (ic2 = BOX_C2_ELEMS-1; ic2 >= 0; ic2--) { + if (dist2 < *bptr) { + *bptr = dist2; + *cptr = (JSAMPLE) icolor; + } + dist2 += xx2; + xx2 += 2 * STEP_C2 * STEP_C2; + bptr++; + cptr++; + } + dist1 += xx1; + xx1 += 2 * STEP_C1 * STEP_C1; + } + dist0 += xx0; + xx0 += 2 * STEP_C0 * STEP_C0; + } + } +} + + +LOCAL(void) +fill_inverse_cmap (j_decompress_ptr cinfo, int c0, int c1, int c2) +/* Fill the inverse-colormap entries in the update box that contains */ +/* histogram cell c0/c1/c2. (Only that one cell MUST be filled, but */ +/* we can fill as many others as we wish.) */ +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + hist3d histogram = cquantize->histogram; + int minc0, minc1, minc2; /* lower left corner of update box */ + int ic0, ic1, ic2; + register JSAMPLE * cptr; /* pointer into bestcolor[] array */ + register histptr cachep; /* pointer into main cache array */ + /* This array lists the candidate colormap indexes. */ + JSAMPLE colorlist[MAXNUMCOLORS]; + int numcolors; /* number of candidate colors */ + /* This array holds the actually closest colormap index for each cell. */ + JSAMPLE bestcolor[BOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS]; + + /* Convert cell coordinates to update box ID */ + c0 >>= BOX_C0_LOG; + c1 >>= BOX_C1_LOG; + c2 >>= BOX_C2_LOG; + + /* Compute true coordinates of update box's origin corner. + * Actually we compute the coordinates of the center of the corner + * histogram cell, which are the lower bounds of the volume we care about. + */ + minc0 = (c0 << BOX_C0_SHIFT) + ((1 << C0_SHIFT) >> 1); + minc1 = (c1 << BOX_C1_SHIFT) + ((1 << C1_SHIFT) >> 1); + minc2 = (c2 << BOX_C2_SHIFT) + ((1 << C2_SHIFT) >> 1); + + /* Determine which colormap entries are close enough to be candidates + * for the nearest entry to some cell in the update box. + */ + numcolors = find_nearby_colors(cinfo, minc0, minc1, minc2, colorlist); + + /* Determine the actually nearest colors. */ + find_best_colors(cinfo, minc0, minc1, minc2, numcolors, colorlist, + bestcolor); + + /* Save the best color numbers (plus 1) in the main cache array */ + c0 <<= BOX_C0_LOG; /* convert ID back to base cell indexes */ + c1 <<= BOX_C1_LOG; + c2 <<= BOX_C2_LOG; + cptr = bestcolor; + for (ic0 = 0; ic0 < BOX_C0_ELEMS; ic0++) { + for (ic1 = 0; ic1 < BOX_C1_ELEMS; ic1++) { + cachep = & histogram[c0+ic0][c1+ic1][c2]; + for (ic2 = 0; ic2 < BOX_C2_ELEMS; ic2++) { + *cachep++ = (histcell) (GETJSAMPLE(*cptr++) + 1); + } + } + } +} + + +/* + * Map some rows of pixels to the output colormapped representation. + */ + +METHODDEF(void) +pass2_no_dither (j_decompress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows) +/* This version performs no dithering */ +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + hist3d histogram = cquantize->histogram; + register JSAMPROW inptr, outptr; + register histptr cachep; + register int c0, c1, c2; + int row; + JDIMENSION col; + JDIMENSION width = cinfo->output_width; + + for (row = 0; row < num_rows; row++) { + inptr = input_buf[row]; + outptr = output_buf[row]; + for (col = width; col > 0; col--) { + /* get pixel value and index into the cache */ + c0 = GETJSAMPLE(*inptr++) >> C0_SHIFT; + c1 = GETJSAMPLE(*inptr++) >> C1_SHIFT; + c2 = GETJSAMPLE(*inptr++) >> C2_SHIFT; + cachep = & histogram[c0][c1][c2]; + /* If we have not seen this color before, find nearest colormap entry */ + /* and update the cache */ + if (*cachep == 0) + fill_inverse_cmap(cinfo, c0,c1,c2); + /* Now emit the colormap index for this cell */ + *outptr++ = (JSAMPLE) (*cachep - 1); + } + } +} + + +METHODDEF(void) +pass2_fs_dither (j_decompress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows) +/* This version performs Floyd-Steinberg dithering */ +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + hist3d histogram = cquantize->histogram; + register LOCFSERROR cur0, cur1, cur2; /* current error or pixel value */ + LOCFSERROR belowerr0, belowerr1, belowerr2; /* error for pixel below cur */ + LOCFSERROR bpreverr0, bpreverr1, bpreverr2; /* error for below/prev col */ + register FSERRPTR errorptr; /* => fserrors[] at column before current */ + JSAMPROW inptr; /* => current input pixel */ + JSAMPROW outptr; /* => current output pixel */ + histptr cachep; + int dir; /* +1 or -1 depending on direction */ + int dir3; /* 3*dir, for advancing inptr & errorptr */ + int row; + JDIMENSION col; + JDIMENSION width = cinfo->output_width; + JSAMPLE *range_limit = cinfo->sample_range_limit; + int *error_limit = cquantize->error_limiter; + JSAMPROW colormap0 = cinfo->colormap[0]; + JSAMPROW colormap1 = cinfo->colormap[1]; + JSAMPROW colormap2 = cinfo->colormap[2]; + SHIFT_TEMPS + + for (row = 0; row < num_rows; row++) { + inptr = input_buf[row]; + outptr = output_buf[row]; + if (cquantize->on_odd_row) { + /* work right to left in this row */ + inptr += (width-1) * 3; /* so point to rightmost pixel */ + outptr += width-1; + dir = -1; + dir3 = -3; + errorptr = cquantize->fserrors + (width+1)*3; /* => entry after last column */ + cquantize->on_odd_row = FALSE; /* flip for next time */ + } else { + /* work left to right in this row */ + dir = 1; + dir3 = 3; + errorptr = cquantize->fserrors; /* => entry before first real column */ + cquantize->on_odd_row = TRUE; /* flip for next time */ + } + /* Preset error values: no error propagated to first pixel from left */ + cur0 = cur1 = cur2 = 0; + /* and no error propagated to row below yet */ + belowerr0 = belowerr1 = belowerr2 = 0; + bpreverr0 = bpreverr1 = bpreverr2 = 0; + + for (col = width; col > 0; col--) { + /* curN holds the error propagated from the previous pixel on the + * current line. Add the error propagated from the previous line + * to form the complete error correction term for this pixel, and + * round the error term (which is expressed * 16) to an integer. + * RIGHT_SHIFT rounds towards minus infinity, so adding 8 is correct + * for either sign of the error value. + * Note: errorptr points to *previous* column's array entry. + */ + cur0 = RIGHT_SHIFT(cur0 + errorptr[dir3+0] + 8, 4); + cur1 = RIGHT_SHIFT(cur1 + errorptr[dir3+1] + 8, 4); + cur2 = RIGHT_SHIFT(cur2 + errorptr[dir3+2] + 8, 4); + /* Limit the error using transfer function set by init_error_limit. + * See comments with init_error_limit for rationale. + */ + cur0 = error_limit[cur0]; + cur1 = error_limit[cur1]; + cur2 = error_limit[cur2]; + /* Form pixel value + error, and range-limit to 0..MAXJSAMPLE. + * The maximum error is +- MAXJSAMPLE (or less with error limiting); + * this sets the required size of the range_limit array. + */ + cur0 += GETJSAMPLE(inptr[0]); + cur1 += GETJSAMPLE(inptr[1]); + cur2 += GETJSAMPLE(inptr[2]); + cur0 = GETJSAMPLE(range_limit[cur0]); + cur1 = GETJSAMPLE(range_limit[cur1]); + cur2 = GETJSAMPLE(range_limit[cur2]); + /* Index into the cache with adjusted pixel value */ + cachep = & histogram[cur0>>C0_SHIFT][cur1>>C1_SHIFT][cur2>>C2_SHIFT]; + /* If we have not seen this color before, find nearest colormap */ + /* entry and update the cache */ + if (*cachep == 0) + fill_inverse_cmap(cinfo, cur0>>C0_SHIFT,cur1>>C1_SHIFT,cur2>>C2_SHIFT); + /* Now emit the colormap index for this cell */ + { register int pixcode = *cachep - 1; + *outptr = (JSAMPLE) pixcode; + /* Compute representation error for this pixel */ + cur0 -= GETJSAMPLE(colormap0[pixcode]); + cur1 -= GETJSAMPLE(colormap1[pixcode]); + cur2 -= GETJSAMPLE(colormap2[pixcode]); + } + /* Compute error fractions to be propagated to adjacent pixels. + * Add these into the running sums, and simultaneously shift the + * next-line error sums left by 1 column. + */ + { register LOCFSERROR bnexterr; + + bnexterr = cur0; /* Process component 0 */ + errorptr[0] = (FSERROR) (bpreverr0 + cur0 * 3); + bpreverr0 = belowerr0 + cur0 * 5; + belowerr0 = bnexterr; + cur0 *= 7; + bnexterr = cur1; /* Process component 1 */ + errorptr[1] = (FSERROR) (bpreverr1 + cur1 * 3); + bpreverr1 = belowerr1 + cur1 * 5; + belowerr1 = bnexterr; + cur1 *= 7; + bnexterr = cur2; /* Process component 2 */ + errorptr[2] = (FSERROR) (bpreverr2 + cur2 * 3); + bpreverr2 = belowerr2 + cur2 * 5; + belowerr2 = bnexterr; + cur2 *= 7; + } + /* At this point curN contains the 7/16 error value to be propagated + * to the next pixel on the current line, and all the errors for the + * next line have been shifted over. We are therefore ready to move on. + */ + inptr += dir3; /* Advance pixel pointers to next column */ + outptr += dir; + errorptr += dir3; /* advance errorptr to current column */ + } + /* Post-loop cleanup: we must unload the final error values into the + * final fserrors[] entry. Note we need not unload belowerrN because + * it is for the dummy column before or after the actual array. + */ + errorptr[0] = (FSERROR) bpreverr0; /* unload prev errs into array */ + errorptr[1] = (FSERROR) bpreverr1; + errorptr[2] = (FSERROR) bpreverr2; + } +} + + +/* + * Initialize the error-limiting transfer function (lookup table). + * The raw F-S error computation can potentially compute error values of up to + * +- MAXJSAMPLE. But we want the maximum correction applied to a pixel to be + * much less, otherwise obviously wrong pixels will be created. (Typical + * effects include weird fringes at color-area boundaries, isolated bright + * pixels in a dark area, etc.) The standard advice for avoiding this problem + * is to ensure that the "corners" of the color cube are allocated as output + * colors; then repeated errors in the same direction cannot cause cascading + * error buildup. However, that only prevents the error from getting + * completely out of hand; Aaron Giles reports that error limiting improves + * the results even with corner colors allocated. + * A simple clamping of the error values to about +- MAXJSAMPLE/8 works pretty + * well, but the smoother transfer function used below is even better. Thanks + * to Aaron Giles for this idea. + */ + +LOCAL(void) +init_error_limit (j_decompress_ptr cinfo) +/* Allocate and fill in the error_limiter table */ +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + int * table; + int in, out; + + table = (int *) (*cinfo->mem->alloc_small) + ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE*2+1) * sizeof(int)); + table += MAXJSAMPLE; /* so can index -MAXJSAMPLE .. +MAXJSAMPLE */ + cquantize->error_limiter = table; + +#define STEPSIZE ((MAXJSAMPLE+1)/16) + /* Map errors 1:1 up to +- MAXJSAMPLE/16 */ + out = 0; + for (in = 0; in < STEPSIZE; in++, out++) { + table[in] = out; table[-in] = -out; + } + /* Map errors 1:2 up to +- 3*MAXJSAMPLE/16 */ + for (; in < STEPSIZE*3; in++, out += (in&1) ? 0 : 1) { + table[in] = out; table[-in] = -out; + } + /* Clamp the rest to final out value (which is (MAXJSAMPLE+1)/8) */ + for (; in <= MAXJSAMPLE; in++) { + table[in] = out; table[-in] = -out; + } +#undef STEPSIZE +} + + +/* + * Finish up at the end of each pass. + */ + +METHODDEF(void) +finish_pass1 (j_decompress_ptr cinfo) +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + + /* Select the representative colors and fill in cinfo->colormap */ + cinfo->colormap = cquantize->sv_colormap; + select_colors(cinfo, cquantize->desired); + /* Force next pass to zero the color index table */ + cquantize->needs_zeroed = TRUE; +} + + +METHODDEF(void) +finish_pass2 (j_decompress_ptr cinfo) +{ + /* no work */ +} + + +/* + * Initialize for each processing pass. + */ + +METHODDEF(void) +start_pass_2_quant (j_decompress_ptr cinfo, boolean is_pre_scan) +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + hist3d histogram = cquantize->histogram; + int i; + + /* Only F-S dithering or no dithering is supported. */ + /* If user asks for ordered dither, give him F-S. */ + if (cinfo->dither_mode != JDITHER_NONE) + cinfo->dither_mode = JDITHER_FS; + + if (is_pre_scan) { + /* Set up method pointers */ + cquantize->pub.color_quantize = prescan_quantize; + cquantize->pub.finish_pass = finish_pass1; + cquantize->needs_zeroed = TRUE; /* Always zero histogram */ + } else { + /* Set up method pointers */ + if (cinfo->dither_mode == JDITHER_FS) + cquantize->pub.color_quantize = pass2_fs_dither; + else + cquantize->pub.color_quantize = pass2_no_dither; + cquantize->pub.finish_pass = finish_pass2; + + /* Make sure color count is acceptable */ + i = cinfo->actual_number_of_colors; + if (i < 1) + ERREXIT1(cinfo, JERR_QUANT_FEW_COLORS, 1); + if (i > MAXNUMCOLORS) + ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, MAXNUMCOLORS); + + if (cinfo->dither_mode == JDITHER_FS) { + size_t arraysize = (size_t) ((cinfo->output_width + 2) * + (3 * sizeof(FSERROR))); + /* Allocate Floyd-Steinberg workspace if we didn't already. */ + if (cquantize->fserrors == NULL) + cquantize->fserrors = (FSERRPTR) (*cinfo->mem->alloc_large) + ((j_common_ptr) cinfo, JPOOL_IMAGE, arraysize); + /* Initialize the propagated errors to zero. */ + jzero_far((void *) cquantize->fserrors, arraysize); + /* Make the error-limit table if we didn't already. */ + if (cquantize->error_limiter == NULL) + init_error_limit(cinfo); + cquantize->on_odd_row = FALSE; + } + + } + /* Zero the histogram or inverse color map, if necessary */ + if (cquantize->needs_zeroed) { + for (i = 0; i < HIST_C0_ELEMS; i++) { + jzero_far((void *) histogram[i], + HIST_C1_ELEMS*HIST_C2_ELEMS * sizeof(histcell)); + } + cquantize->needs_zeroed = FALSE; + } +} + + +/* + * Switch to a new external colormap between output passes. + */ + +METHODDEF(void) +new_color_map_2_quant (j_decompress_ptr cinfo) +{ + my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize; + + /* Reset the inverse color map */ + cquantize->needs_zeroed = TRUE; +} + + +/* + * Module initialization routine for 2-pass color quantization. + */ + +GLOBAL(void) +jinit_2pass_quantizer (j_decompress_ptr cinfo) +{ + my_cquantize_ptr cquantize; + int i; + + cquantize = (my_cquantize_ptr) + (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, + sizeof(my_cquantizer)); + cinfo->cquantize = (struct jpeg_color_quantizer *) cquantize; + cquantize->pub.start_pass = start_pass_2_quant; + cquantize->pub.new_color_map = new_color_map_2_quant; + cquantize->fserrors = NULL; /* flag optional arrays not allocated */ + cquantize->error_limiter = NULL; + + /* Make sure jdmaster didn't give me a case I can't handle */ + if (cinfo->out_color_components != 3) + ERREXIT(cinfo, JERR_NOTIMPL); + + /* Allocate the histogram/inverse colormap storage */ + cquantize->histogram = (hist3d) (*cinfo->mem->alloc_small) + ((j_common_ptr) cinfo, JPOOL_IMAGE, HIST_C0_ELEMS * sizeof(hist2d)); + for (i = 0; i < HIST_C0_ELEMS; i++) { + cquantize->histogram[i] = (hist2d) (*cinfo->mem->alloc_large) + ((j_common_ptr) cinfo, JPOOL_IMAGE, + HIST_C1_ELEMS*HIST_C2_ELEMS * sizeof(histcell)); + } + cquantize->needs_zeroed = TRUE; /* histogram is garbage now */ + + /* Allocate storage for the completed colormap, if required. + * We do this now since it may affect the memory manager's space + * calculations. + */ + if (cinfo->enable_2pass_quant) { + /* Make sure color count is acceptable */ + int desired = cinfo->desired_number_of_colors; + /* Lower bound on # of colors ... somewhat arbitrary as long as > 0 */ + if (desired < 8) + ERREXIT1(cinfo, JERR_QUANT_FEW_COLORS, 8); + /* Make sure colormap indexes can be represented by JSAMPLEs */ + if (desired > MAXNUMCOLORS) + ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, MAXNUMCOLORS); + cquantize->sv_colormap = (*cinfo->mem->alloc_sarray) + ((j_common_ptr) cinfo,JPOOL_IMAGE, (JDIMENSION) desired, (JDIMENSION) 3); + cquantize->desired = desired; + } else + cquantize->sv_colormap = NULL; + + /* Only F-S dithering or no dithering is supported. */ + /* If user asks for ordered dither, give him F-S. */ + if (cinfo->dither_mode != JDITHER_NONE) + cinfo->dither_mode = JDITHER_FS; + + /* Allocate Floyd-Steinberg workspace if necessary. + * This isn't really needed until pass 2, but again it may affect the memory + * manager's space calculations. Although we will cope with a later change + * in dither_mode, we do not promise to honor max_memory_to_use if + * dither_mode changes. + */ + if (cinfo->dither_mode == JDITHER_FS) { + cquantize->fserrors = (FSERRPTR) (*cinfo->mem->alloc_large) + ((j_common_ptr) cinfo, JPOOL_IMAGE, + (size_t) ((cinfo->output_width + 2) * (3 * sizeof(FSERROR)))); + /* Might as well create the error-limiting table too. */ + init_error_limit(cinfo); + } +} + +#endif /* QUANT_2PASS_SUPPORTED */ diff --git a/app/src/main/jni/libjpeg-turbo/jsimd.h b/app/src/main/jni/libjpeg-turbo/jsimd.h new file mode 100644 index 000000000..f1f584b8b --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jsimd.h @@ -0,0 +1,84 @@ +/* + * jsimd.h + * + * Copyright 2009 Pierre Ossman for Cendio AB + * Copyright 2011, 2014 D. R. Commander + * + * Based on the x86 SIMD extension for IJG JPEG library, + * Copyright (C) 1999-2006, MIYASAKA Masaru. + * For conditions of distribution and use, see copyright notice in jsimdext.inc + * + */ + +EXTERN(int) jsimd_can_rgb_ycc (void); +EXTERN(int) jsimd_can_rgb_gray (void); +EXTERN(int) jsimd_can_ycc_rgb (void); +EXTERN(int) jsimd_can_ycc_rgb565 (void); +EXTERN(int) jsimd_c_can_null_convert (void); + +EXTERN(void) jsimd_rgb_ycc_convert + (j_compress_ptr cinfo, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_rgb_gray_convert + (j_compress_ptr cinfo, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_ycc_rgb_convert + (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_rgb565_convert + (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_c_null_convert + (j_compress_ptr cinfo, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); + +EXTERN(int) jsimd_can_h2v2_downsample (void); +EXTERN(int) jsimd_can_h2v1_downsample (void); + +EXTERN(void) jsimd_h2v2_downsample + (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data); + +EXTERN(int) jsimd_can_h2v2_smooth_downsample (void); + +EXTERN(void) jsimd_h2v2_smooth_downsample + (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data); + +EXTERN(void) jsimd_h2v1_downsample + (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data); + +EXTERN(int) jsimd_can_h2v2_upsample (void); +EXTERN(int) jsimd_can_h2v1_upsample (void); +EXTERN(int) jsimd_can_int_upsample (void); + +EXTERN(void) jsimd_h2v2_upsample + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr); +EXTERN(void) jsimd_h2v1_upsample + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr); +EXTERN(void) jsimd_int_upsample + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr); + +EXTERN(int) jsimd_can_h2v2_fancy_upsample (void); +EXTERN(int) jsimd_can_h2v1_fancy_upsample (void); + +EXTERN(void) jsimd_h2v2_fancy_upsample + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr); +EXTERN(void) jsimd_h2v1_fancy_upsample + (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr); + +EXTERN(int) jsimd_can_h2v2_merged_upsample (void); +EXTERN(int) jsimd_can_h2v1_merged_upsample (void); + +EXTERN(void) jsimd_h2v2_merged_upsample + (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v1_merged_upsample + (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); diff --git a/app/src/main/jni/libjpeg-turbo/jsimd_none.c b/app/src/main/jni/libjpeg-turbo/jsimd_none.c new file mode 100644 index 000000000..34aefc9fa --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jsimd_none.c @@ -0,0 +1,389 @@ +/* + * jsimd_none.c + * + * Copyright 2009 Pierre Ossman for Cendio AB + * Copyright 2009-2011, 2014 D. R. Commander + * + * Based on the x86 SIMD extension for IJG JPEG library, + * Copyright (C) 1999-2006, MIYASAKA Masaru. + * For conditions of distribution and use, see copyright notice in jsimdext.inc + * + * This file contains stubs for when there is no SIMD support available. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" +#include "jsimd.h" +#include "jdct.h" +#include "jsimddct.h" + +GLOBAL(int) +jsimd_can_rgb_ycc (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_rgb_gray (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_ycc_rgb (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_ycc_rgb565 (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_c_can_null_convert (void) +{ + return 0; +} + +GLOBAL(void) +jsimd_rgb_ycc_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ +} + +GLOBAL(void) +jsimd_rgb_gray_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ +} + +GLOBAL(void) +jsimd_ycc_rgb_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ +} + +GLOBAL(void) +jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ +} + +GLOBAL(void) +jsimd_c_null_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ +} + +GLOBAL(int) +jsimd_can_h2v2_downsample (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_downsample (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_h2v2_smooth_downsample (void) +{ + return 0; +} + +GLOBAL(void) +jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ +} + +GLOBAL(void) +jsimd_h2v2_smooth_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ +} + +GLOBAL(void) +jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ +} + +GLOBAL(int) +jsimd_can_h2v2_upsample (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_upsample (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_int_upsample (void) +{ + return 0; +} + +GLOBAL(void) +jsimd_int_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr) +{ +} + +GLOBAL(void) +jsimd_h2v2_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ +} + +GLOBAL(void) +jsimd_h2v1_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ +} + +GLOBAL(int) +jsimd_can_h2v2_fancy_upsample (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_fancy_upsample (void) +{ + return 0; +} + +GLOBAL(void) +jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ +} + +GLOBAL(void) +jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ +} + +GLOBAL(int) +jsimd_can_h2v2_merged_upsample (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_merged_upsample (void) +{ + return 0; +} + +GLOBAL(void) +jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ +} + +GLOBAL(void) +jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ +} + +GLOBAL(int) +jsimd_can_convsamp (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_convsamp_float (void) +{ + return 0; +} + +GLOBAL(void) +jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col, + DCTELEM * workspace) +{ +} + +GLOBAL(void) +jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col, + FAST_FLOAT * workspace) +{ +} + +GLOBAL(int) +jsimd_can_fdct_islow (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_fdct_ifast (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_fdct_float (void) +{ + return 0; +} + +GLOBAL(void) +jsimd_fdct_islow (DCTELEM * data) +{ +} + +GLOBAL(void) +jsimd_fdct_ifast (DCTELEM * data) +{ +} + +GLOBAL(void) +jsimd_fdct_float (FAST_FLOAT * data) +{ +} + +GLOBAL(int) +jsimd_can_quantize (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_quantize_float (void) +{ + return 0; +} + +GLOBAL(void) +jsimd_quantize (JCOEFPTR coef_block, DCTELEM * divisors, + DCTELEM * workspace) +{ +} + +GLOBAL(void) +jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT * divisors, + FAST_FLOAT * workspace) +{ +} + +GLOBAL(int) +jsimd_can_idct_2x2 (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_idct_4x4 (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_idct_6x6 (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_idct_12x12 (void) +{ + return 0; +} + +GLOBAL(void) +jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ +} + +GLOBAL(void) +jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ +} + +GLOBAL(void) +jsimd_idct_6x6 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ +} + +GLOBAL(void) +jsimd_idct_12x12 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ +} + +GLOBAL(int) +jsimd_can_idct_islow (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_idct_ifast (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_can_idct_float (void) +{ + return 0; +} + +GLOBAL(void) +jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ +} + +GLOBAL(void) +jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ +} + +GLOBAL(void) +jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ +} + diff --git a/app/src/main/jni/libjpeg-turbo/jsimddct.h b/app/src/main/jni/libjpeg-turbo/jsimddct.h new file mode 100644 index 000000000..aa421fb1b --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jsimddct.h @@ -0,0 +1,74 @@ +/* + * jsimddct.h + * + * Copyright 2009 Pierre Ossman for Cendio AB + * + * Based on the x86 SIMD extension for IJG JPEG library, + * Copyright (C) 1999-2006, MIYASAKA Masaru. + * For conditions of distribution and use, see copyright notice in jsimdext.inc + * + */ + +EXTERN(int) jsimd_can_convsamp (void); +EXTERN(int) jsimd_can_convsamp_float (void); + +EXTERN(void) jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col, + DCTELEM * workspace); +EXTERN(void) jsimd_convsamp_float (JSAMPARRAY sample_data, + JDIMENSION start_col, + FAST_FLOAT * workspace); + +EXTERN(int) jsimd_can_fdct_islow (void); +EXTERN(int) jsimd_can_fdct_ifast (void); +EXTERN(int) jsimd_can_fdct_float (void); + +EXTERN(void) jsimd_fdct_islow (DCTELEM * data); +EXTERN(void) jsimd_fdct_ifast (DCTELEM * data); +EXTERN(void) jsimd_fdct_float (FAST_FLOAT * data); + +EXTERN(int) jsimd_can_quantize (void); +EXTERN(int) jsimd_can_quantize_float (void); + +EXTERN(void) jsimd_quantize (JCOEFPTR coef_block, DCTELEM * divisors, + DCTELEM * workspace); +EXTERN(void) jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT * divisors, + FAST_FLOAT * workspace); + +EXTERN(int) jsimd_can_idct_2x2 (void); +EXTERN(int) jsimd_can_idct_4x4 (void); +EXTERN(int) jsimd_can_idct_6x6 (void); +EXTERN(int) jsimd_can_idct_12x12 (void); + +EXTERN(void) jsimd_idct_2x2 (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); +EXTERN(void) jsimd_idct_4x4 (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); +EXTERN(void) jsimd_idct_6x6 (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); +EXTERN(void) jsimd_idct_12x12 (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); + +EXTERN(int) jsimd_can_idct_islow (void); +EXTERN(int) jsimd_can_idct_ifast (void); +EXTERN(int) jsimd_can_idct_float (void); + +EXTERN(void) jsimd_idct_islow (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); +EXTERN(void) jsimd_idct_ifast (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); +EXTERN(void) jsimd_idct_float (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); diff --git a/app/src/main/jni/libjpeg-turbo/jstdhuff.c b/app/src/main/jni/libjpeg-turbo/jstdhuff.c new file mode 100644 index 000000000..b29e5ea47 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jstdhuff.c @@ -0,0 +1,133 @@ +/* +* jstdhuff.c +* +* This file was part of the Independent JPEG Group's software: +* Copyright (C) 1991-1998, Thomas G. Lane. +* libjpeg-turbo Modifications: +* Copyright (C) 2013, D. R. Commander. +* For conditions of distribution and use, see the accompanying README file. +* +* This file contains routines to set the default Huffman tables, if they are +* not already set. +*/ + +/* + * Huffman table setup routines + */ + +LOCAL(void) +add_huff_table (j_common_ptr cinfo, + JHUFF_TBL **htblptr, const UINT8 *bits, const UINT8 *val) +/* Define a Huffman table */ +{ + int nsymbols, len; + + if (*htblptr == NULL) + *htblptr = jpeg_alloc_huff_table(cinfo); + else + return; + + /* Copy the number-of-symbols-of-each-code-length counts */ + MEMCOPY((*htblptr)->bits, bits, sizeof((*htblptr)->bits)); + + /* Validate the counts. We do this here mainly so we can copy the right + * number of symbols from the val[] array, without risking marching off + * the end of memory. jchuff.c will do a more thorough test later. + */ + nsymbols = 0; + for (len = 1; len <= 16; len++) + nsymbols += bits[len]; + if (nsymbols < 1 || nsymbols > 256) + ERREXIT(cinfo, JERR_BAD_HUFF_TABLE); + + MEMCOPY((*htblptr)->huffval, val, nsymbols * sizeof(UINT8)); + + /* Initialize sent_table FALSE so table will be written to JPEG file. */ + (*htblptr)->sent_table = FALSE; +} + + +LOCAL(void) +std_huff_tables (j_common_ptr cinfo) +/* Set up the standard Huffman tables (cf. JPEG standard section K.3) */ +/* IMPORTANT: these are only valid for 8-bit data precision! */ +{ + JHUFF_TBL **dc_huff_tbl_ptrs, **ac_huff_tbl_ptrs; + + static const UINT8 bits_dc_luminance[17] = + { /* 0-base */ 0, 0, 1, 5, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0 }; + static const UINT8 val_dc_luminance[] = + { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 }; + + static const UINT8 bits_dc_chrominance[17] = + { /* 0-base */ 0, 0, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0 }; + static const UINT8 val_dc_chrominance[] = + { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 }; + + static const UINT8 bits_ac_luminance[17] = + { /* 0-base */ 0, 0, 2, 1, 3, 3, 2, 4, 3, 5, 5, 4, 4, 0, 0, 1, 0x7d }; + static const UINT8 val_ac_luminance[] = + { 0x01, 0x02, 0x03, 0x00, 0x04, 0x11, 0x05, 0x12, + 0x21, 0x31, 0x41, 0x06, 0x13, 0x51, 0x61, 0x07, + 0x22, 0x71, 0x14, 0x32, 0x81, 0x91, 0xa1, 0x08, + 0x23, 0x42, 0xb1, 0xc1, 0x15, 0x52, 0xd1, 0xf0, + 0x24, 0x33, 0x62, 0x72, 0x82, 0x09, 0x0a, 0x16, + 0x17, 0x18, 0x19, 0x1a, 0x25, 0x26, 0x27, 0x28, + 0x29, 0x2a, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, + 0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49, + 0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59, + 0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, + 0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, + 0x7a, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, + 0x8a, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, + 0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, + 0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6, + 0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3, 0xc4, 0xc5, + 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2, 0xd3, 0xd4, + 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xe1, 0xe2, + 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea, + 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, + 0xf9, 0xfa }; + + static const UINT8 bits_ac_chrominance[17] = + { /* 0-base */ 0, 0, 2, 1, 2, 4, 4, 3, 4, 7, 5, 4, 4, 0, 1, 2, 0x77 }; + static const UINT8 val_ac_chrominance[] = + { 0x00, 0x01, 0x02, 0x03, 0x11, 0x04, 0x05, 0x21, + 0x31, 0x06, 0x12, 0x41, 0x51, 0x07, 0x61, 0x71, + 0x13, 0x22, 0x32, 0x81, 0x08, 0x14, 0x42, 0x91, + 0xa1, 0xb1, 0xc1, 0x09, 0x23, 0x33, 0x52, 0xf0, + 0x15, 0x62, 0x72, 0xd1, 0x0a, 0x16, 0x24, 0x34, + 0xe1, 0x25, 0xf1, 0x17, 0x18, 0x19, 0x1a, 0x26, + 0x27, 0x28, 0x29, 0x2a, 0x35, 0x36, 0x37, 0x38, + 0x39, 0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, + 0x49, 0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, + 0x59, 0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, + 0x69, 0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, + 0x79, 0x7a, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, + 0x88, 0x89, 0x8a, 0x92, 0x93, 0x94, 0x95, 0x96, + 0x97, 0x98, 0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5, + 0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4, + 0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3, + 0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2, + 0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, + 0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, + 0xea, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, + 0xf9, 0xfa }; + + if (cinfo->is_decompressor) { + dc_huff_tbl_ptrs = ((j_decompress_ptr)cinfo)->dc_huff_tbl_ptrs; + ac_huff_tbl_ptrs = ((j_decompress_ptr)cinfo)->ac_huff_tbl_ptrs; + } else { + dc_huff_tbl_ptrs = ((j_compress_ptr)cinfo)->dc_huff_tbl_ptrs; + ac_huff_tbl_ptrs = ((j_compress_ptr)cinfo)->ac_huff_tbl_ptrs; + } + + add_huff_table(cinfo, &dc_huff_tbl_ptrs[0], bits_dc_luminance, + val_dc_luminance); + add_huff_table(cinfo, &ac_huff_tbl_ptrs[0], bits_ac_luminance, + val_ac_luminance); + add_huff_table(cinfo, &dc_huff_tbl_ptrs[1], bits_dc_chrominance, + val_dc_chrominance); + add_huff_table(cinfo, &ac_huff_tbl_ptrs[1], bits_ac_chrominance, + val_ac_chrominance); +} diff --git a/app/src/main/jni/libjpeg-turbo/jutils.c b/app/src/main/jni/libjpeg-turbo/jutils.c new file mode 100644 index 000000000..0e2611c8c --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jutils.c @@ -0,0 +1,132 @@ +/* + * jutils.c + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-1996, Thomas G. Lane. + * It was modified by The libjpeg-turbo Project to include only code + * relevant to libjpeg-turbo. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains tables and miscellaneous utility routines needed + * for both compression and decompression. + * Note we prefix all global names with "j" to minimize conflicts with + * a surrounding application. + */ + +#define JPEG_INTERNALS +#include "jinclude.h" +#include "jpeglib.h" + + +/* + * jpeg_zigzag_order[i] is the zigzag-order position of the i'th element + * of a DCT block read in natural order (left to right, top to bottom). + */ + +#if 0 /* This table is not actually needed in v6a */ + +const int jpeg_zigzag_order[DCTSIZE2] = { + 0, 1, 5, 6, 14, 15, 27, 28, + 2, 4, 7, 13, 16, 26, 29, 42, + 3, 8, 12, 17, 25, 30, 41, 43, + 9, 11, 18, 24, 31, 40, 44, 53, + 10, 19, 23, 32, 39, 45, 52, 54, + 20, 22, 33, 38, 46, 51, 55, 60, + 21, 34, 37, 47, 50, 56, 59, 61, + 35, 36, 48, 49, 57, 58, 62, 63 +}; + +#endif + +/* + * jpeg_natural_order[i] is the natural-order position of the i'th element + * of zigzag order. + * + * When reading corrupted data, the Huffman decoders could attempt + * to reference an entry beyond the end of this array (if the decoded + * zero run length reaches past the end of the block). To prevent + * wild stores without adding an inner-loop test, we put some extra + * "63"s after the real entries. This will cause the extra coefficient + * to be stored in location 63 of the block, not somewhere random. + * The worst case would be a run-length of 15, which means we need 16 + * fake entries. + */ + +const int jpeg_natural_order[DCTSIZE2+16] = { + 0, 1, 8, 16, 9, 2, 3, 10, + 17, 24, 32, 25, 18, 11, 4, 5, + 12, 19, 26, 33, 40, 48, 41, 34, + 27, 20, 13, 6, 7, 14, 21, 28, + 35, 42, 49, 56, 57, 50, 43, 36, + 29, 22, 15, 23, 30, 37, 44, 51, + 58, 59, 52, 45, 38, 31, 39, 46, + 53, 60, 61, 54, 47, 55, 62, 63, + 63, 63, 63, 63, 63, 63, 63, 63, /* extra entries for safety in decoder */ + 63, 63, 63, 63, 63, 63, 63, 63 +}; + + +/* + * Arithmetic utilities + */ + +GLOBAL(long) +jdiv_round_up (long a, long b) +/* Compute a/b rounded up to next integer, ie, ceil(a/b) */ +/* Assumes a >= 0, b > 0 */ +{ + return (a + b - 1L) / b; +} + + +GLOBAL(long) +jround_up (long a, long b) +/* Compute a rounded up to next multiple of b, ie, ceil(a/b)*b */ +/* Assumes a >= 0, b > 0 */ +{ + a += b - 1L; + return a - (a % b); +} + + +GLOBAL(void) +jcopy_sample_rows (JSAMPARRAY input_array, int source_row, + JSAMPARRAY output_array, int dest_row, + int num_rows, JDIMENSION num_cols) +/* Copy some rows of samples from one place to another. + * num_rows rows are copied from input_array[source_row++] + * to output_array[dest_row++]; these areas may overlap for duplication. + * The source and destination arrays must be at least as wide as num_cols. + */ +{ + register JSAMPROW inptr, outptr; + register size_t count = (size_t) (num_cols * sizeof(JSAMPLE)); + register int row; + + input_array += source_row; + output_array += dest_row; + + for (row = num_rows; row > 0; row--) { + inptr = *input_array++; + outptr = *output_array++; + MEMCOPY(outptr, inptr, count); + } +} + + +GLOBAL(void) +jcopy_block_row (JBLOCKROW input_row, JBLOCKROW output_row, + JDIMENSION num_blocks) +/* Copy a row of coefficient blocks from one place to another. */ +{ + MEMCOPY(output_row, input_row, num_blocks * (DCTSIZE2 * sizeof(JCOEF))); +} + + +GLOBAL(void) +jzero_far (void * target, size_t bytestozero) +/* Zero out a chunk of memory. */ +/* This might be sample-array data, block-array data, or alloc_large data. */ +{ + MEMZERO(target, bytestozero); +} diff --git a/app/src/main/jni/libjpeg-turbo/jversion.h b/app/src/main/jni/libjpeg-turbo/jversion.h new file mode 100644 index 000000000..25c3cf0b0 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/jversion.h @@ -0,0 +1,36 @@ +/* + * jversion.h + * + * This file was part of the Independent JPEG Group's software: + * Copyright (C) 1991-2012, Thomas G. Lane, Guido Vollbeding. + * libjpeg-turbo Modifications: + * Copyright (C) 2010, 2012-2014, D. R. Commander. + * For conditions of distribution and use, see the accompanying README file. + * + * This file contains software version identification. + */ + + +#if JPEG_LIB_VERSION >= 80 + +#define JVERSION "8d 15-Jan-2012" + +#elif JPEG_LIB_VERSION >= 70 + +#define JVERSION "7 27-Jun-2009" + +#else + +#define JVERSION "6b 27-Mar-1998" + +#endif + +#define JCOPYRIGHT "Copyright (C) 1991-2012 Thomas G. Lane, Guido Vollbeding\n" \ + "Copyright (C) 1999-2006 MIYASAKA Masaru\n" \ + "Copyright (C) 2009 Pierre Ossman for Cendio AB\n" \ + "Copyright (C) 2009-2014 D. R. Commander\n" \ + "Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies)\n" \ + "Copyright (C) 2013-2014 MIPS Technologies, Inc.\n" \ + "Copyright (C) 2013 Linaro Limited" + +#define JCOPYRIGHT_SHORT "Copyright (C) 1991-2014 The libjpeg-turbo Project and many others" diff --git a/app/src/main/jni/libjpeg-turbo/simd/jccolext-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jccolext-mmx.asm new file mode 100644 index 000000000..d3d47a56c --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jccolext-mmx.asm @@ -0,0 +1,477 @@ +; +; jccolext.asm - colorspace conversion (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jcolsamp.inc" + +; -------------------------------------------------------------------------- +; +; Convert some rows of samples to the output colorspace. +; +; GLOBAL(void) +; jsimd_rgb_ycc_convert_mmx (JDIMENSION img_width, +; JSAMPARRAY input_buf, JSAMPIMAGE output_buf, +; JDIMENSION output_row, int num_rows); +; + +%define img_width(b) (b)+8 ; JDIMENSION img_width +%define input_buf(b) (b)+12 ; JSAMPARRAY input_buf +%define output_buf(b) (b)+16 ; JSAMPIMAGE output_buf +%define output_row(b) (b)+20 ; JDIMENSION output_row +%define num_rows(b) (b)+24 ; int num_rows + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_MMWORD ; mmword wk[WK_NUM] +%define WK_NUM 8 +%define gotptr wk(0)-SIZEOF_POINTER ; void * gotptr + + align 16 + global EXTN(jsimd_rgb_ycc_convert_mmx) + +EXTN(jsimd_rgb_ycc_convert_mmx): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_MMWORD) ; align to 64 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic eax ; make a room for GOT address + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + movpic POINTER [gotptr], ebx ; save GOT address + + mov ecx, JDIMENSION [img_width(eax)] ; num_cols + test ecx,ecx + jz near .return + + push ecx + + mov esi, JSAMPIMAGE [output_buf(eax)] + mov ecx, JDIMENSION [output_row(eax)] + mov edi, JSAMPARRAY [esi+0*SIZEOF_JSAMPARRAY] + mov ebx, JSAMPARRAY [esi+1*SIZEOF_JSAMPARRAY] + mov edx, JSAMPARRAY [esi+2*SIZEOF_JSAMPARRAY] + lea edi, [edi+ecx*SIZEOF_JSAMPROW] + lea ebx, [ebx+ecx*SIZEOF_JSAMPROW] + lea edx, [edx+ecx*SIZEOF_JSAMPROW] + + pop ecx + + mov esi, JSAMPARRAY [input_buf(eax)] + mov eax, INT [num_rows(eax)] + test eax,eax + jle near .return + alignx 16,7 +.rowloop: + pushpic eax + push edx + push ebx + push edi + push esi + push ecx ; col + + mov esi, JSAMPROW [esi] ; inptr + mov edi, JSAMPROW [edi] ; outptr0 + mov ebx, JSAMPROW [ebx] ; outptr1 + mov edx, JSAMPROW [edx] ; outptr2 + movpic eax, POINTER [gotptr] ; load GOT address (eax) + + cmp ecx, byte SIZEOF_MMWORD + jae short .columnloop + alignx 16,7 + +%if RGB_PIXELSIZE == 3 ; --------------- + +.column_ld1: + push eax + push edx + lea ecx,[ecx+ecx*2] ; imul ecx,RGB_PIXELSIZE + test cl, SIZEOF_BYTE + jz short .column_ld2 + sub ecx, byte SIZEOF_BYTE + xor eax,eax + mov al, BYTE [esi+ecx] +.column_ld2: + test cl, SIZEOF_WORD + jz short .column_ld4 + sub ecx, byte SIZEOF_WORD + xor edx,edx + mov dx, WORD [esi+ecx] + shl eax, WORD_BIT + or eax,edx +.column_ld4: + movd mmA,eax + pop edx + pop eax + test cl, SIZEOF_DWORD + jz short .column_ld8 + sub ecx, byte SIZEOF_DWORD + movd mmG, DWORD [esi+ecx] + psllq mmA, DWORD_BIT + por mmA,mmG +.column_ld8: + test cl, SIZEOF_MMWORD + jz short .column_ld16 + movq mmG,mmA + movq mmA, MMWORD [esi+0*SIZEOF_MMWORD] + mov ecx, SIZEOF_MMWORD + jmp short .rgb_ycc_cnv +.column_ld16: + test cl, 2*SIZEOF_MMWORD + mov ecx, SIZEOF_MMWORD + jz short .rgb_ycc_cnv + movq mmF,mmA + movq mmA, MMWORD [esi+0*SIZEOF_MMWORD] + movq mmG, MMWORD [esi+1*SIZEOF_MMWORD] + jmp short .rgb_ycc_cnv + alignx 16,7 + +.columnloop: + movq mmA, MMWORD [esi+0*SIZEOF_MMWORD] + movq mmG, MMWORD [esi+1*SIZEOF_MMWORD] + movq mmF, MMWORD [esi+2*SIZEOF_MMWORD] + +.rgb_ycc_cnv: + ; mmA=(00 10 20 01 11 21 02 12) + ; mmG=(22 03 13 23 04 14 24 05) + ; mmF=(15 25 06 16 26 07 17 27) + + movq mmD,mmA + psllq mmA,4*BYTE_BIT ; mmA=(-- -- -- -- 00 10 20 01) + psrlq mmD,4*BYTE_BIT ; mmD=(11 21 02 12 -- -- -- --) + + punpckhbw mmA,mmG ; mmA=(00 04 10 14 20 24 01 05) + psllq mmG,4*BYTE_BIT ; mmG=(-- -- -- -- 22 03 13 23) + + punpcklbw mmD,mmF ; mmD=(11 15 21 25 02 06 12 16) + punpckhbw mmG,mmF ; mmG=(22 26 03 07 13 17 23 27) + + movq mmE,mmA + psllq mmA,4*BYTE_BIT ; mmA=(-- -- -- -- 00 04 10 14) + psrlq mmE,4*BYTE_BIT ; mmE=(20 24 01 05 -- -- -- --) + + punpckhbw mmA,mmD ; mmA=(00 02 04 06 10 12 14 16) + psllq mmD,4*BYTE_BIT ; mmD=(-- -- -- -- 11 15 21 25) + + punpcklbw mmE,mmG ; mmE=(20 22 24 26 01 03 05 07) + punpckhbw mmD,mmG ; mmD=(11 13 15 17 21 23 25 27) + + pxor mmH,mmH + + movq mmC,mmA + punpcklbw mmA,mmH ; mmA=(00 02 04 06) + punpckhbw mmC,mmH ; mmC=(10 12 14 16) + + movq mmB,mmE + punpcklbw mmE,mmH ; mmE=(20 22 24 26) + punpckhbw mmB,mmH ; mmB=(01 03 05 07) + + movq mmF,mmD + punpcklbw mmD,mmH ; mmD=(11 13 15 17) + punpckhbw mmF,mmH ; mmF=(21 23 25 27) + +%else ; RGB_PIXELSIZE == 4 ; ----------- + +.column_ld1: + test cl, SIZEOF_MMWORD/8 + jz short .column_ld2 + sub ecx, byte SIZEOF_MMWORD/8 + movd mmA, DWORD [esi+ecx*RGB_PIXELSIZE] +.column_ld2: + test cl, SIZEOF_MMWORD/4 + jz short .column_ld4 + sub ecx, byte SIZEOF_MMWORD/4 + movq mmF,mmA + movq mmA, MMWORD [esi+ecx*RGB_PIXELSIZE] +.column_ld4: + test cl, SIZEOF_MMWORD/2 + mov ecx, SIZEOF_MMWORD + jz short .rgb_ycc_cnv + movq mmD,mmA + movq mmC,mmF + movq mmA, MMWORD [esi+0*SIZEOF_MMWORD] + movq mmF, MMWORD [esi+1*SIZEOF_MMWORD] + jmp short .rgb_ycc_cnv + alignx 16,7 + +.columnloop: + movq mmA, MMWORD [esi+0*SIZEOF_MMWORD] + movq mmF, MMWORD [esi+1*SIZEOF_MMWORD] + movq mmD, MMWORD [esi+2*SIZEOF_MMWORD] + movq mmC, MMWORD [esi+3*SIZEOF_MMWORD] + +.rgb_ycc_cnv: + ; mmA=(00 10 20 30 01 11 21 31) + ; mmF=(02 12 22 32 03 13 23 33) + ; mmD=(04 14 24 34 05 15 25 35) + ; mmC=(06 16 26 36 07 17 27 37) + + movq mmB,mmA + punpcklbw mmA,mmF ; mmA=(00 02 10 12 20 22 30 32) + punpckhbw mmB,mmF ; mmB=(01 03 11 13 21 23 31 33) + + movq mmG,mmD + punpcklbw mmD,mmC ; mmD=(04 06 14 16 24 26 34 36) + punpckhbw mmG,mmC ; mmG=(05 07 15 17 25 27 35 37) + + movq mmE,mmA + punpcklwd mmA,mmD ; mmA=(00 02 04 06 10 12 14 16) + punpckhwd mmE,mmD ; mmE=(20 22 24 26 30 32 34 36) + + movq mmH,mmB + punpcklwd mmB,mmG ; mmB=(01 03 05 07 11 13 15 17) + punpckhwd mmH,mmG ; mmH=(21 23 25 27 31 33 35 37) + + pxor mmF,mmF + + movq mmC,mmA + punpcklbw mmA,mmF ; mmA=(00 02 04 06) + punpckhbw mmC,mmF ; mmC=(10 12 14 16) + + movq mmD,mmB + punpcklbw mmB,mmF ; mmB=(01 03 05 07) + punpckhbw mmD,mmF ; mmD=(11 13 15 17) + + movq mmG,mmE + punpcklbw mmE,mmF ; mmE=(20 22 24 26) + punpckhbw mmG,mmF ; mmG=(30 32 34 36) + + punpcklbw mmF,mmH + punpckhbw mmH,mmH + psrlw mmF,BYTE_BIT ; mmF=(21 23 25 27) + psrlw mmH,BYTE_BIT ; mmH=(31 33 35 37) + +%endif ; RGB_PIXELSIZE ; --------------- + + ; mm0=(R0 R2 R4 R6)=RE, mm2=(G0 G2 G4 G6)=GE, mm4=(B0 B2 B4 B6)=BE + ; mm1=(R1 R3 R5 R7)=RO, mm3=(G1 G3 G5 G7)=GO, mm5=(B1 B3 B5 B7)=BO + + ; (Original) + ; Y = 0.29900 * R + 0.58700 * G + 0.11400 * B + ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE + ; Cr = 0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE + ; + ; (This implementation) + ; Y = 0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G + ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE + ; Cr = 0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE + + movq MMWORD [wk(0)], mm0 ; wk(0)=RE + movq MMWORD [wk(1)], mm1 ; wk(1)=RO + movq MMWORD [wk(2)], mm4 ; wk(2)=BE + movq MMWORD [wk(3)], mm5 ; wk(3)=BO + + movq mm6,mm1 + punpcklwd mm1,mm3 + punpckhwd mm6,mm3 + movq mm7,mm1 + movq mm4,mm6 + pmaddwd mm1,[GOTOFF(eax,PW_F0299_F0337)] ; mm1=ROL*FIX(0.299)+GOL*FIX(0.337) + pmaddwd mm6,[GOTOFF(eax,PW_F0299_F0337)] ; mm6=ROH*FIX(0.299)+GOH*FIX(0.337) + pmaddwd mm7,[GOTOFF(eax,PW_MF016_MF033)] ; mm7=ROL*-FIX(0.168)+GOL*-FIX(0.331) + pmaddwd mm4,[GOTOFF(eax,PW_MF016_MF033)] ; mm4=ROH*-FIX(0.168)+GOH*-FIX(0.331) + + movq MMWORD [wk(4)], mm1 ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337) + movq MMWORD [wk(5)], mm6 ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337) + + pxor mm1,mm1 + pxor mm6,mm6 + punpcklwd mm1,mm5 ; mm1=BOL + punpckhwd mm6,mm5 ; mm6=BOH + psrld mm1,1 ; mm1=BOL*FIX(0.500) + psrld mm6,1 ; mm6=BOH*FIX(0.500) + + movq mm5,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; mm5=[PD_ONEHALFM1_CJ] + + paddd mm7,mm1 + paddd mm4,mm6 + paddd mm7,mm5 + paddd mm4,mm5 + psrld mm7,SCALEBITS ; mm7=CbOL + psrld mm4,SCALEBITS ; mm4=CbOH + packssdw mm7,mm4 ; mm7=CbO + + movq mm1, MMWORD [wk(2)] ; mm1=BE + + movq mm6,mm0 + punpcklwd mm0,mm2 + punpckhwd mm6,mm2 + movq mm5,mm0 + movq mm4,mm6 + pmaddwd mm0,[GOTOFF(eax,PW_F0299_F0337)] ; mm0=REL*FIX(0.299)+GEL*FIX(0.337) + pmaddwd mm6,[GOTOFF(eax,PW_F0299_F0337)] ; mm6=REH*FIX(0.299)+GEH*FIX(0.337) + pmaddwd mm5,[GOTOFF(eax,PW_MF016_MF033)] ; mm5=REL*-FIX(0.168)+GEL*-FIX(0.331) + pmaddwd mm4,[GOTOFF(eax,PW_MF016_MF033)] ; mm4=REH*-FIX(0.168)+GEH*-FIX(0.331) + + movq MMWORD [wk(6)], mm0 ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337) + movq MMWORD [wk(7)], mm6 ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337) + + pxor mm0,mm0 + pxor mm6,mm6 + punpcklwd mm0,mm1 ; mm0=BEL + punpckhwd mm6,mm1 ; mm6=BEH + psrld mm0,1 ; mm0=BEL*FIX(0.500) + psrld mm6,1 ; mm6=BEH*FIX(0.500) + + movq mm1,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; mm1=[PD_ONEHALFM1_CJ] + + paddd mm5,mm0 + paddd mm4,mm6 + paddd mm5,mm1 + paddd mm4,mm1 + psrld mm5,SCALEBITS ; mm5=CbEL + psrld mm4,SCALEBITS ; mm4=CbEH + packssdw mm5,mm4 ; mm5=CbE + + psllw mm7,BYTE_BIT + por mm5,mm7 ; mm5=Cb + movq MMWORD [ebx], mm5 ; Save Cb + + movq mm0, MMWORD [wk(3)] ; mm0=BO + movq mm6, MMWORD [wk(2)] ; mm6=BE + movq mm1, MMWORD [wk(1)] ; mm1=RO + + movq mm4,mm0 + punpcklwd mm0,mm3 + punpckhwd mm4,mm3 + movq mm7,mm0 + movq mm5,mm4 + pmaddwd mm0,[GOTOFF(eax,PW_F0114_F0250)] ; mm0=BOL*FIX(0.114)+GOL*FIX(0.250) + pmaddwd mm4,[GOTOFF(eax,PW_F0114_F0250)] ; mm4=BOH*FIX(0.114)+GOH*FIX(0.250) + pmaddwd mm7,[GOTOFF(eax,PW_MF008_MF041)] ; mm7=BOL*-FIX(0.081)+GOL*-FIX(0.418) + pmaddwd mm5,[GOTOFF(eax,PW_MF008_MF041)] ; mm5=BOH*-FIX(0.081)+GOH*-FIX(0.418) + + movq mm3,[GOTOFF(eax,PD_ONEHALF)] ; mm3=[PD_ONEHALF] + + paddd mm0, MMWORD [wk(4)] + paddd mm4, MMWORD [wk(5)] + paddd mm0,mm3 + paddd mm4,mm3 + psrld mm0,SCALEBITS ; mm0=YOL + psrld mm4,SCALEBITS ; mm4=YOH + packssdw mm0,mm4 ; mm0=YO + + pxor mm3,mm3 + pxor mm4,mm4 + punpcklwd mm3,mm1 ; mm3=ROL + punpckhwd mm4,mm1 ; mm4=ROH + psrld mm3,1 ; mm3=ROL*FIX(0.500) + psrld mm4,1 ; mm4=ROH*FIX(0.500) + + movq mm1,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; mm1=[PD_ONEHALFM1_CJ] + + paddd mm7,mm3 + paddd mm5,mm4 + paddd mm7,mm1 + paddd mm5,mm1 + psrld mm7,SCALEBITS ; mm7=CrOL + psrld mm5,SCALEBITS ; mm5=CrOH + packssdw mm7,mm5 ; mm7=CrO + + movq mm3, MMWORD [wk(0)] ; mm3=RE + + movq mm4,mm6 + punpcklwd mm6,mm2 + punpckhwd mm4,mm2 + movq mm1,mm6 + movq mm5,mm4 + pmaddwd mm6,[GOTOFF(eax,PW_F0114_F0250)] ; mm6=BEL*FIX(0.114)+GEL*FIX(0.250) + pmaddwd mm4,[GOTOFF(eax,PW_F0114_F0250)] ; mm4=BEH*FIX(0.114)+GEH*FIX(0.250) + pmaddwd mm1,[GOTOFF(eax,PW_MF008_MF041)] ; mm1=BEL*-FIX(0.081)+GEL*-FIX(0.418) + pmaddwd mm5,[GOTOFF(eax,PW_MF008_MF041)] ; mm5=BEH*-FIX(0.081)+GEH*-FIX(0.418) + + movq mm2,[GOTOFF(eax,PD_ONEHALF)] ; mm2=[PD_ONEHALF] + + paddd mm6, MMWORD [wk(6)] + paddd mm4, MMWORD [wk(7)] + paddd mm6,mm2 + paddd mm4,mm2 + psrld mm6,SCALEBITS ; mm6=YEL + psrld mm4,SCALEBITS ; mm4=YEH + packssdw mm6,mm4 ; mm6=YE + + psllw mm0,BYTE_BIT + por mm6,mm0 ; mm6=Y + movq MMWORD [edi], mm6 ; Save Y + + pxor mm2,mm2 + pxor mm4,mm4 + punpcklwd mm2,mm3 ; mm2=REL + punpckhwd mm4,mm3 ; mm4=REH + psrld mm2,1 ; mm2=REL*FIX(0.500) + psrld mm4,1 ; mm4=REH*FIX(0.500) + + movq mm0,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; mm0=[PD_ONEHALFM1_CJ] + + paddd mm1,mm2 + paddd mm5,mm4 + paddd mm1,mm0 + paddd mm5,mm0 + psrld mm1,SCALEBITS ; mm1=CrEL + psrld mm5,SCALEBITS ; mm5=CrEH + packssdw mm1,mm5 ; mm1=CrE + + psllw mm7,BYTE_BIT + por mm1,mm7 ; mm1=Cr + movq MMWORD [edx], mm1 ; Save Cr + + sub ecx, byte SIZEOF_MMWORD + add esi, byte RGB_PIXELSIZE*SIZEOF_MMWORD ; inptr + add edi, byte SIZEOF_MMWORD ; outptr0 + add ebx, byte SIZEOF_MMWORD ; outptr1 + add edx, byte SIZEOF_MMWORD ; outptr2 + cmp ecx, byte SIZEOF_MMWORD + jae near .columnloop + test ecx,ecx + jnz near .column_ld1 + + pop ecx ; col + pop esi + pop edi + pop ebx + pop edx + poppic eax + + add esi, byte SIZEOF_JSAMPROW ; input_buf + add edi, byte SIZEOF_JSAMPROW + add ebx, byte SIZEOF_JSAMPROW + add edx, byte SIZEOF_JSAMPROW + dec eax ; num_rows + jg near .rowloop + + emms ; empty MMX state + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jccolext-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jccolext-sse2-64.asm new file mode 100644 index 000000000..7bd6d016e --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jccolext-sse2-64.asm @@ -0,0 +1,485 @@ +; +; jccolext.asm - colorspace conversion (64-bit SSE2) +; +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; Copyright (C) 2009, D. R. Commander. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jcolsamp.inc" + +; -------------------------------------------------------------------------- +; +; Convert some rows of samples to the output colorspace. +; +; GLOBAL(void) +; jsimd_rgb_ycc_convert_sse2 (JDIMENSION img_width, +; JSAMPARRAY input_buf, JSAMPIMAGE output_buf, +; JDIMENSION output_row, int num_rows); +; + +; r10 = JDIMENSION img_width +; r11 = JSAMPARRAY input_buf +; r12 = JSAMPIMAGE output_buf +; r13 = JDIMENSION output_row +; r14 = int num_rows + +%define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 8 + + align 16 + + global EXTN(jsimd_rgb_ycc_convert_sse2) + +EXTN(jsimd_rgb_ycc_convert_sse2): + push rbp + mov rax,rsp ; rax = original rbp + sub rsp, byte 4 + and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [rsp],rax + mov rbp,rsp ; rbp = aligned rbp + lea rsp, [wk(0)] + collect_args + push rbx + + mov rcx, r10 + test rcx,rcx + jz near .return + + push rcx + + mov rsi, r12 + mov rcx, r13 + mov rdi, JSAMPARRAY [rsi+0*SIZEOF_JSAMPARRAY] + mov rbx, JSAMPARRAY [rsi+1*SIZEOF_JSAMPARRAY] + mov rdx, JSAMPARRAY [rsi+2*SIZEOF_JSAMPARRAY] + lea rdi, [rdi+rcx*SIZEOF_JSAMPROW] + lea rbx, [rbx+rcx*SIZEOF_JSAMPROW] + lea rdx, [rdx+rcx*SIZEOF_JSAMPROW] + + pop rcx + + mov rsi, r11 + mov eax, r14d + test rax,rax + jle near .return +.rowloop: + push rdx + push rbx + push rdi + push rsi + push rcx ; col + + mov rsi, JSAMPROW [rsi] ; inptr + mov rdi, JSAMPROW [rdi] ; outptr0 + mov rbx, JSAMPROW [rbx] ; outptr1 + mov rdx, JSAMPROW [rdx] ; outptr2 + + cmp rcx, byte SIZEOF_XMMWORD + jae near .columnloop + +%if RGB_PIXELSIZE == 3 ; --------------- + +.column_ld1: + push rax + push rdx + lea rcx,[rcx+rcx*2] ; imul ecx,RGB_PIXELSIZE + test cl, SIZEOF_BYTE + jz short .column_ld2 + sub rcx, byte SIZEOF_BYTE + movzx rax, BYTE [rsi+rcx] +.column_ld2: + test cl, SIZEOF_WORD + jz short .column_ld4 + sub rcx, byte SIZEOF_WORD + movzx rdx, WORD [rsi+rcx] + shl rax, WORD_BIT + or rax,rdx +.column_ld4: + movd xmmA,eax + pop rdx + pop rax + test cl, SIZEOF_DWORD + jz short .column_ld8 + sub rcx, byte SIZEOF_DWORD + movd xmmF, XMM_DWORD [rsi+rcx] + pslldq xmmA, SIZEOF_DWORD + por xmmA,xmmF +.column_ld8: + test cl, SIZEOF_MMWORD + jz short .column_ld16 + sub rcx, byte SIZEOF_MMWORD + movq xmmB, XMM_MMWORD [rsi+rcx] + pslldq xmmA, SIZEOF_MMWORD + por xmmA,xmmB +.column_ld16: + test cl, SIZEOF_XMMWORD + jz short .column_ld32 + movdqa xmmF,xmmA + movdqu xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD] + mov rcx, SIZEOF_XMMWORD + jmp short .rgb_ycc_cnv +.column_ld32: + test cl, 2*SIZEOF_XMMWORD + mov rcx, SIZEOF_XMMWORD + jz short .rgb_ycc_cnv + movdqa xmmB,xmmA + movdqu xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD] + movdqu xmmF, XMMWORD [rsi+1*SIZEOF_XMMWORD] + jmp short .rgb_ycc_cnv + +.columnloop: + movdqu xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD] + movdqu xmmF, XMMWORD [rsi+1*SIZEOF_XMMWORD] + movdqu xmmB, XMMWORD [rsi+2*SIZEOF_XMMWORD] + +.rgb_ycc_cnv: + ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05) + ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A) + ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F) + + movdqa xmmG,xmmA + pslldq xmmA,8 ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12) + psrldq xmmG,8 ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --) + + punpckhbw xmmA,xmmF ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A) + pslldq xmmF,8 ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27) + + punpcklbw xmmG,xmmB ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D) + punpckhbw xmmF,xmmB ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F) + + movdqa xmmD,xmmA + pslldq xmmA,8 ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09) + psrldq xmmD,8 ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --) + + punpckhbw xmmA,xmmG ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D) + pslldq xmmG,8 ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B) + + punpcklbw xmmD,xmmF ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E) + punpckhbw xmmG,xmmF ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F) + + movdqa xmmE,xmmA + pslldq xmmA,8 ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C) + psrldq xmmE,8 ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --) + + punpckhbw xmmA,xmmD ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E) + pslldq xmmD,8 ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D) + + punpcklbw xmmE,xmmG ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F) + punpckhbw xmmD,xmmG ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F) + + pxor xmmH,xmmH + + movdqa xmmC,xmmA + punpcklbw xmmA,xmmH ; xmmA=(00 02 04 06 08 0A 0C 0E) + punpckhbw xmmC,xmmH ; xmmC=(10 12 14 16 18 1A 1C 1E) + + movdqa xmmB,xmmE + punpcklbw xmmE,xmmH ; xmmE=(20 22 24 26 28 2A 2C 2E) + punpckhbw xmmB,xmmH ; xmmB=(01 03 05 07 09 0B 0D 0F) + + movdqa xmmF,xmmD + punpcklbw xmmD,xmmH ; xmmD=(11 13 15 17 19 1B 1D 1F) + punpckhbw xmmF,xmmH ; xmmF=(21 23 25 27 29 2B 2D 2F) + +%else ; RGB_PIXELSIZE == 4 ; ----------- + +.column_ld1: + test cl, SIZEOF_XMMWORD/16 + jz short .column_ld2 + sub rcx, byte SIZEOF_XMMWORD/16 + movd xmmA, XMM_DWORD [rsi+rcx*RGB_PIXELSIZE] +.column_ld2: + test cl, SIZEOF_XMMWORD/8 + jz short .column_ld4 + sub rcx, byte SIZEOF_XMMWORD/8 + movq xmmE, XMM_MMWORD [rsi+rcx*RGB_PIXELSIZE] + pslldq xmmA, SIZEOF_MMWORD + por xmmA,xmmE +.column_ld4: + test cl, SIZEOF_XMMWORD/4 + jz short .column_ld8 + sub rcx, byte SIZEOF_XMMWORD/4 + movdqa xmmE,xmmA + movdqu xmmA, XMMWORD [rsi+rcx*RGB_PIXELSIZE] +.column_ld8: + test cl, SIZEOF_XMMWORD/2 + mov rcx, SIZEOF_XMMWORD + jz short .rgb_ycc_cnv + movdqa xmmF,xmmA + movdqa xmmH,xmmE + movdqu xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD] + movdqu xmmE, XMMWORD [rsi+1*SIZEOF_XMMWORD] + jmp short .rgb_ycc_cnv + +.columnloop: + movdqu xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD] + movdqu xmmE, XMMWORD [rsi+1*SIZEOF_XMMWORD] + movdqu xmmF, XMMWORD [rsi+2*SIZEOF_XMMWORD] + movdqu xmmH, XMMWORD [rsi+3*SIZEOF_XMMWORD] + +.rgb_ycc_cnv: + ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33) + ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37) + ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B) + ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F) + + movdqa xmmD,xmmA + punpcklbw xmmA,xmmE ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35) + punpckhbw xmmD,xmmE ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37) + + movdqa xmmC,xmmF + punpcklbw xmmF,xmmH ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D) + punpckhbw xmmC,xmmH ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F) + + movdqa xmmB,xmmA + punpcklwd xmmA,xmmF ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C) + punpckhwd xmmB,xmmF ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D) + + movdqa xmmG,xmmD + punpcklwd xmmD,xmmC ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E) + punpckhwd xmmG,xmmC ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F) + + movdqa xmmE,xmmA + punpcklbw xmmA,xmmD ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E) + punpckhbw xmmE,xmmD ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E) + + movdqa xmmH,xmmB + punpcklbw xmmB,xmmG ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F) + punpckhbw xmmH,xmmG ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F) + + pxor xmmF,xmmF + + movdqa xmmC,xmmA + punpcklbw xmmA,xmmF ; xmmA=(00 02 04 06 08 0A 0C 0E) + punpckhbw xmmC,xmmF ; xmmC=(10 12 14 16 18 1A 1C 1E) + + movdqa xmmD,xmmB + punpcklbw xmmB,xmmF ; xmmB=(01 03 05 07 09 0B 0D 0F) + punpckhbw xmmD,xmmF ; xmmD=(11 13 15 17 19 1B 1D 1F) + + movdqa xmmG,xmmE + punpcklbw xmmE,xmmF ; xmmE=(20 22 24 26 28 2A 2C 2E) + punpckhbw xmmG,xmmF ; xmmG=(30 32 34 36 38 3A 3C 3E) + + punpcklbw xmmF,xmmH + punpckhbw xmmH,xmmH + psrlw xmmF,BYTE_BIT ; xmmF=(21 23 25 27 29 2B 2D 2F) + psrlw xmmH,BYTE_BIT ; xmmH=(31 33 35 37 39 3B 3D 3F) + +%endif ; RGB_PIXELSIZE ; --------------- + + ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE + ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO + + ; (Original) + ; Y = 0.29900 * R + 0.58700 * G + 0.11400 * B + ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE + ; Cr = 0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE + ; + ; (This implementation) + ; Y = 0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G + ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE + ; Cr = 0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE + + movdqa XMMWORD [wk(0)], xmm0 ; wk(0)=RE + movdqa XMMWORD [wk(1)], xmm1 ; wk(1)=RO + movdqa XMMWORD [wk(2)], xmm4 ; wk(2)=BE + movdqa XMMWORD [wk(3)], xmm5 ; wk(3)=BO + + movdqa xmm6,xmm1 + punpcklwd xmm1,xmm3 + punpckhwd xmm6,xmm3 + movdqa xmm7,xmm1 + movdqa xmm4,xmm6 + pmaddwd xmm1,[rel PW_F0299_F0337] ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337) + pmaddwd xmm6,[rel PW_F0299_F0337] ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337) + pmaddwd xmm7,[rel PW_MF016_MF033] ; xmm7=ROL*-FIX(0.168)+GOL*-FIX(0.331) + pmaddwd xmm4,[rel PW_MF016_MF033] ; xmm4=ROH*-FIX(0.168)+GOH*-FIX(0.331) + + movdqa XMMWORD [wk(4)], xmm1 ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337) + movdqa XMMWORD [wk(5)], xmm6 ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337) + + pxor xmm1,xmm1 + pxor xmm6,xmm6 + punpcklwd xmm1,xmm5 ; xmm1=BOL + punpckhwd xmm6,xmm5 ; xmm6=BOH + psrld xmm1,1 ; xmm1=BOL*FIX(0.500) + psrld xmm6,1 ; xmm6=BOH*FIX(0.500) + + movdqa xmm5,[rel PD_ONEHALFM1_CJ] ; xmm5=[PD_ONEHALFM1_CJ] + + paddd xmm7,xmm1 + paddd xmm4,xmm6 + paddd xmm7,xmm5 + paddd xmm4,xmm5 + psrld xmm7,SCALEBITS ; xmm7=CbOL + psrld xmm4,SCALEBITS ; xmm4=CbOH + packssdw xmm7,xmm4 ; xmm7=CbO + + movdqa xmm1, XMMWORD [wk(2)] ; xmm1=BE + + movdqa xmm6,xmm0 + punpcklwd xmm0,xmm2 + punpckhwd xmm6,xmm2 + movdqa xmm5,xmm0 + movdqa xmm4,xmm6 + pmaddwd xmm0,[rel PW_F0299_F0337] ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337) + pmaddwd xmm6,[rel PW_F0299_F0337] ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337) + pmaddwd xmm5,[rel PW_MF016_MF033] ; xmm5=REL*-FIX(0.168)+GEL*-FIX(0.331) + pmaddwd xmm4,[rel PW_MF016_MF033] ; xmm4=REH*-FIX(0.168)+GEH*-FIX(0.331) + + movdqa XMMWORD [wk(6)], xmm0 ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337) + movdqa XMMWORD [wk(7)], xmm6 ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337) + + pxor xmm0,xmm0 + pxor xmm6,xmm6 + punpcklwd xmm0,xmm1 ; xmm0=BEL + punpckhwd xmm6,xmm1 ; xmm6=BEH + psrld xmm0,1 ; xmm0=BEL*FIX(0.500) + psrld xmm6,1 ; xmm6=BEH*FIX(0.500) + + movdqa xmm1,[rel PD_ONEHALFM1_CJ] ; xmm1=[PD_ONEHALFM1_CJ] + + paddd xmm5,xmm0 + paddd xmm4,xmm6 + paddd xmm5,xmm1 + paddd xmm4,xmm1 + psrld xmm5,SCALEBITS ; xmm5=CbEL + psrld xmm4,SCALEBITS ; xmm4=CbEH + packssdw xmm5,xmm4 ; xmm5=CbE + + psllw xmm7,BYTE_BIT + por xmm5,xmm7 ; xmm5=Cb + movdqa XMMWORD [rbx], xmm5 ; Save Cb + + movdqa xmm0, XMMWORD [wk(3)] ; xmm0=BO + movdqa xmm6, XMMWORD [wk(2)] ; xmm6=BE + movdqa xmm1, XMMWORD [wk(1)] ; xmm1=RO + + movdqa xmm4,xmm0 + punpcklwd xmm0,xmm3 + punpckhwd xmm4,xmm3 + movdqa xmm7,xmm0 + movdqa xmm5,xmm4 + pmaddwd xmm0,[rel PW_F0114_F0250] ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250) + pmaddwd xmm4,[rel PW_F0114_F0250] ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250) + pmaddwd xmm7,[rel PW_MF008_MF041] ; xmm7=BOL*-FIX(0.081)+GOL*-FIX(0.418) + pmaddwd xmm5,[rel PW_MF008_MF041] ; xmm5=BOH*-FIX(0.081)+GOH*-FIX(0.418) + + movdqa xmm3,[rel PD_ONEHALF] ; xmm3=[PD_ONEHALF] + + paddd xmm0, XMMWORD [wk(4)] + paddd xmm4, XMMWORD [wk(5)] + paddd xmm0,xmm3 + paddd xmm4,xmm3 + psrld xmm0,SCALEBITS ; xmm0=YOL + psrld xmm4,SCALEBITS ; xmm4=YOH + packssdw xmm0,xmm4 ; xmm0=YO + + pxor xmm3,xmm3 + pxor xmm4,xmm4 + punpcklwd xmm3,xmm1 ; xmm3=ROL + punpckhwd xmm4,xmm1 ; xmm4=ROH + psrld xmm3,1 ; xmm3=ROL*FIX(0.500) + psrld xmm4,1 ; xmm4=ROH*FIX(0.500) + + movdqa xmm1,[rel PD_ONEHALFM1_CJ] ; xmm1=[PD_ONEHALFM1_CJ] + + paddd xmm7,xmm3 + paddd xmm5,xmm4 + paddd xmm7,xmm1 + paddd xmm5,xmm1 + psrld xmm7,SCALEBITS ; xmm7=CrOL + psrld xmm5,SCALEBITS ; xmm5=CrOH + packssdw xmm7,xmm5 ; xmm7=CrO + + movdqa xmm3, XMMWORD [wk(0)] ; xmm3=RE + + movdqa xmm4,xmm6 + punpcklwd xmm6,xmm2 + punpckhwd xmm4,xmm2 + movdqa xmm1,xmm6 + movdqa xmm5,xmm4 + pmaddwd xmm6,[rel PW_F0114_F0250] ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250) + pmaddwd xmm4,[rel PW_F0114_F0250] ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250) + pmaddwd xmm1,[rel PW_MF008_MF041] ; xmm1=BEL*-FIX(0.081)+GEL*-FIX(0.418) + pmaddwd xmm5,[rel PW_MF008_MF041] ; xmm5=BEH*-FIX(0.081)+GEH*-FIX(0.418) + + movdqa xmm2,[rel PD_ONEHALF] ; xmm2=[PD_ONEHALF] + + paddd xmm6, XMMWORD [wk(6)] + paddd xmm4, XMMWORD [wk(7)] + paddd xmm6,xmm2 + paddd xmm4,xmm2 + psrld xmm6,SCALEBITS ; xmm6=YEL + psrld xmm4,SCALEBITS ; xmm4=YEH + packssdw xmm6,xmm4 ; xmm6=YE + + psllw xmm0,BYTE_BIT + por xmm6,xmm0 ; xmm6=Y + movdqa XMMWORD [rdi], xmm6 ; Save Y + + pxor xmm2,xmm2 + pxor xmm4,xmm4 + punpcklwd xmm2,xmm3 ; xmm2=REL + punpckhwd xmm4,xmm3 ; xmm4=REH + psrld xmm2,1 ; xmm2=REL*FIX(0.500) + psrld xmm4,1 ; xmm4=REH*FIX(0.500) + + movdqa xmm0,[rel PD_ONEHALFM1_CJ] ; xmm0=[PD_ONEHALFM1_CJ] + + paddd xmm1,xmm2 + paddd xmm5,xmm4 + paddd xmm1,xmm0 + paddd xmm5,xmm0 + psrld xmm1,SCALEBITS ; xmm1=CrEL + psrld xmm5,SCALEBITS ; xmm5=CrEH + packssdw xmm1,xmm5 ; xmm1=CrE + + psllw xmm7,BYTE_BIT + por xmm1,xmm7 ; xmm1=Cr + movdqa XMMWORD [rdx], xmm1 ; Save Cr + + sub rcx, byte SIZEOF_XMMWORD + add rsi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; inptr + add rdi, byte SIZEOF_XMMWORD ; outptr0 + add rbx, byte SIZEOF_XMMWORD ; outptr1 + add rdx, byte SIZEOF_XMMWORD ; outptr2 + cmp rcx, byte SIZEOF_XMMWORD + jae near .columnloop + test rcx,rcx + jnz near .column_ld1 + + pop rcx ; col + pop rsi + pop rdi + pop rbx + pop rdx + + add rsi, byte SIZEOF_JSAMPROW ; input_buf + add rdi, byte SIZEOF_JSAMPROW + add rbx, byte SIZEOF_JSAMPROW + add rdx, byte SIZEOF_JSAMPROW + dec rax ; num_rows + jg near .rowloop + +.return: + pop rbx + uncollect_args + mov rsp,rbp ; rsp <- aligned rbp + pop rsp ; rsp <- original rbp + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jccolext-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jccolext-sse2.asm new file mode 100644 index 000000000..cc38e98a1 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jccolext-sse2.asm @@ -0,0 +1,503 @@ +; +; jccolext.asm - colorspace conversion (SSE2) +; +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jcolsamp.inc" + +; -------------------------------------------------------------------------- +; +; Convert some rows of samples to the output colorspace. +; +; GLOBAL(void) +; jsimd_rgb_ycc_convert_sse2 (JDIMENSION img_width, +; JSAMPARRAY input_buf, JSAMPIMAGE output_buf, +; JDIMENSION output_row, int num_rows); +; + +%define img_width(b) (b)+8 ; JDIMENSION img_width +%define input_buf(b) (b)+12 ; JSAMPARRAY input_buf +%define output_buf(b) (b)+16 ; JSAMPIMAGE output_buf +%define output_row(b) (b)+20 ; JDIMENSION output_row +%define num_rows(b) (b)+24 ; int num_rows + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 8 +%define gotptr wk(0)-SIZEOF_POINTER ; void * gotptr + + align 16 + + global EXTN(jsimd_rgb_ycc_convert_sse2) + +EXTN(jsimd_rgb_ycc_convert_sse2): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic eax ; make a room for GOT address + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + movpic POINTER [gotptr], ebx ; save GOT address + + mov ecx, JDIMENSION [img_width(eax)] + test ecx,ecx + jz near .return + + push ecx + + mov esi, JSAMPIMAGE [output_buf(eax)] + mov ecx, JDIMENSION [output_row(eax)] + mov edi, JSAMPARRAY [esi+0*SIZEOF_JSAMPARRAY] + mov ebx, JSAMPARRAY [esi+1*SIZEOF_JSAMPARRAY] + mov edx, JSAMPARRAY [esi+2*SIZEOF_JSAMPARRAY] + lea edi, [edi+ecx*SIZEOF_JSAMPROW] + lea ebx, [ebx+ecx*SIZEOF_JSAMPROW] + lea edx, [edx+ecx*SIZEOF_JSAMPROW] + + pop ecx + + mov esi, JSAMPARRAY [input_buf(eax)] + mov eax, INT [num_rows(eax)] + test eax,eax + jle near .return + alignx 16,7 +.rowloop: + pushpic eax + push edx + push ebx + push edi + push esi + push ecx ; col + + mov esi, JSAMPROW [esi] ; inptr + mov edi, JSAMPROW [edi] ; outptr0 + mov ebx, JSAMPROW [ebx] ; outptr1 + mov edx, JSAMPROW [edx] ; outptr2 + movpic eax, POINTER [gotptr] ; load GOT address (eax) + + cmp ecx, byte SIZEOF_XMMWORD + jae near .columnloop + alignx 16,7 + +%if RGB_PIXELSIZE == 3 ; --------------- + +.column_ld1: + push eax + push edx + lea ecx,[ecx+ecx*2] ; imul ecx,RGB_PIXELSIZE + test cl, SIZEOF_BYTE + jz short .column_ld2 + sub ecx, byte SIZEOF_BYTE + movzx eax, BYTE [esi+ecx] +.column_ld2: + test cl, SIZEOF_WORD + jz short .column_ld4 + sub ecx, byte SIZEOF_WORD + movzx edx, WORD [esi+ecx] + shl eax, WORD_BIT + or eax,edx +.column_ld4: + movd xmmA,eax + pop edx + pop eax + test cl, SIZEOF_DWORD + jz short .column_ld8 + sub ecx, byte SIZEOF_DWORD + movd xmmF, XMM_DWORD [esi+ecx] + pslldq xmmA, SIZEOF_DWORD + por xmmA,xmmF +.column_ld8: + test cl, SIZEOF_MMWORD + jz short .column_ld16 + sub ecx, byte SIZEOF_MMWORD + movq xmmB, XMM_MMWORD [esi+ecx] + pslldq xmmA, SIZEOF_MMWORD + por xmmA,xmmB +.column_ld16: + test cl, SIZEOF_XMMWORD + jz short .column_ld32 + movdqa xmmF,xmmA + movdqu xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD] + mov ecx, SIZEOF_XMMWORD + jmp short .rgb_ycc_cnv +.column_ld32: + test cl, 2*SIZEOF_XMMWORD + mov ecx, SIZEOF_XMMWORD + jz short .rgb_ycc_cnv + movdqa xmmB,xmmA + movdqu xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD] + movdqu xmmF, XMMWORD [esi+1*SIZEOF_XMMWORD] + jmp short .rgb_ycc_cnv + alignx 16,7 + +.columnloop: + movdqu xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD] + movdqu xmmF, XMMWORD [esi+1*SIZEOF_XMMWORD] + movdqu xmmB, XMMWORD [esi+2*SIZEOF_XMMWORD] + +.rgb_ycc_cnv: + ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05) + ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A) + ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F) + + movdqa xmmG,xmmA + pslldq xmmA,8 ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12) + psrldq xmmG,8 ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --) + + punpckhbw xmmA,xmmF ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A) + pslldq xmmF,8 ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27) + + punpcklbw xmmG,xmmB ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D) + punpckhbw xmmF,xmmB ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F) + + movdqa xmmD,xmmA + pslldq xmmA,8 ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09) + psrldq xmmD,8 ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --) + + punpckhbw xmmA,xmmG ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D) + pslldq xmmG,8 ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B) + + punpcklbw xmmD,xmmF ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E) + punpckhbw xmmG,xmmF ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F) + + movdqa xmmE,xmmA + pslldq xmmA,8 ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C) + psrldq xmmE,8 ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --) + + punpckhbw xmmA,xmmD ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E) + pslldq xmmD,8 ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D) + + punpcklbw xmmE,xmmG ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F) + punpckhbw xmmD,xmmG ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F) + + pxor xmmH,xmmH + + movdqa xmmC,xmmA + punpcklbw xmmA,xmmH ; xmmA=(00 02 04 06 08 0A 0C 0E) + punpckhbw xmmC,xmmH ; xmmC=(10 12 14 16 18 1A 1C 1E) + + movdqa xmmB,xmmE + punpcklbw xmmE,xmmH ; xmmE=(20 22 24 26 28 2A 2C 2E) + punpckhbw xmmB,xmmH ; xmmB=(01 03 05 07 09 0B 0D 0F) + + movdqa xmmF,xmmD + punpcklbw xmmD,xmmH ; xmmD=(11 13 15 17 19 1B 1D 1F) + punpckhbw xmmF,xmmH ; xmmF=(21 23 25 27 29 2B 2D 2F) + +%else ; RGB_PIXELSIZE == 4 ; ----------- + +.column_ld1: + test cl, SIZEOF_XMMWORD/16 + jz short .column_ld2 + sub ecx, byte SIZEOF_XMMWORD/16 + movd xmmA, XMM_DWORD [esi+ecx*RGB_PIXELSIZE] +.column_ld2: + test cl, SIZEOF_XMMWORD/8 + jz short .column_ld4 + sub ecx, byte SIZEOF_XMMWORD/8 + movq xmmE, XMM_MMWORD [esi+ecx*RGB_PIXELSIZE] + pslldq xmmA, SIZEOF_MMWORD + por xmmA,xmmE +.column_ld4: + test cl, SIZEOF_XMMWORD/4 + jz short .column_ld8 + sub ecx, byte SIZEOF_XMMWORD/4 + movdqa xmmE,xmmA + movdqu xmmA, XMMWORD [esi+ecx*RGB_PIXELSIZE] +.column_ld8: + test cl, SIZEOF_XMMWORD/2 + mov ecx, SIZEOF_XMMWORD + jz short .rgb_ycc_cnv + movdqa xmmF,xmmA + movdqa xmmH,xmmE + movdqu xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD] + movdqu xmmE, XMMWORD [esi+1*SIZEOF_XMMWORD] + jmp short .rgb_ycc_cnv + alignx 16,7 + +.columnloop: + movdqu xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD] + movdqu xmmE, XMMWORD [esi+1*SIZEOF_XMMWORD] + movdqu xmmF, XMMWORD [esi+2*SIZEOF_XMMWORD] + movdqu xmmH, XMMWORD [esi+3*SIZEOF_XMMWORD] + +.rgb_ycc_cnv: + ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33) + ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37) + ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B) + ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F) + + movdqa xmmD,xmmA + punpcklbw xmmA,xmmE ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35) + punpckhbw xmmD,xmmE ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37) + + movdqa xmmC,xmmF + punpcklbw xmmF,xmmH ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D) + punpckhbw xmmC,xmmH ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F) + + movdqa xmmB,xmmA + punpcklwd xmmA,xmmF ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C) + punpckhwd xmmB,xmmF ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D) + + movdqa xmmG,xmmD + punpcklwd xmmD,xmmC ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E) + punpckhwd xmmG,xmmC ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F) + + movdqa xmmE,xmmA + punpcklbw xmmA,xmmD ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E) + punpckhbw xmmE,xmmD ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E) + + movdqa xmmH,xmmB + punpcklbw xmmB,xmmG ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F) + punpckhbw xmmH,xmmG ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F) + + pxor xmmF,xmmF + + movdqa xmmC,xmmA + punpcklbw xmmA,xmmF ; xmmA=(00 02 04 06 08 0A 0C 0E) + punpckhbw xmmC,xmmF ; xmmC=(10 12 14 16 18 1A 1C 1E) + + movdqa xmmD,xmmB + punpcklbw xmmB,xmmF ; xmmB=(01 03 05 07 09 0B 0D 0F) + punpckhbw xmmD,xmmF ; xmmD=(11 13 15 17 19 1B 1D 1F) + + movdqa xmmG,xmmE + punpcklbw xmmE,xmmF ; xmmE=(20 22 24 26 28 2A 2C 2E) + punpckhbw xmmG,xmmF ; xmmG=(30 32 34 36 38 3A 3C 3E) + + punpcklbw xmmF,xmmH + punpckhbw xmmH,xmmH + psrlw xmmF,BYTE_BIT ; xmmF=(21 23 25 27 29 2B 2D 2F) + psrlw xmmH,BYTE_BIT ; xmmH=(31 33 35 37 39 3B 3D 3F) + +%endif ; RGB_PIXELSIZE ; --------------- + + ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE + ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO + + ; (Original) + ; Y = 0.29900 * R + 0.58700 * G + 0.11400 * B + ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE + ; Cr = 0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE + ; + ; (This implementation) + ; Y = 0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G + ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE + ; Cr = 0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE + + movdqa XMMWORD [wk(0)], xmm0 ; wk(0)=RE + movdqa XMMWORD [wk(1)], xmm1 ; wk(1)=RO + movdqa XMMWORD [wk(2)], xmm4 ; wk(2)=BE + movdqa XMMWORD [wk(3)], xmm5 ; wk(3)=BO + + movdqa xmm6,xmm1 + punpcklwd xmm1,xmm3 + punpckhwd xmm6,xmm3 + movdqa xmm7,xmm1 + movdqa xmm4,xmm6 + pmaddwd xmm1,[GOTOFF(eax,PW_F0299_F0337)] ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337) + pmaddwd xmm6,[GOTOFF(eax,PW_F0299_F0337)] ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337) + pmaddwd xmm7,[GOTOFF(eax,PW_MF016_MF033)] ; xmm7=ROL*-FIX(0.168)+GOL*-FIX(0.331) + pmaddwd xmm4,[GOTOFF(eax,PW_MF016_MF033)] ; xmm4=ROH*-FIX(0.168)+GOH*-FIX(0.331) + + movdqa XMMWORD [wk(4)], xmm1 ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337) + movdqa XMMWORD [wk(5)], xmm6 ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337) + + pxor xmm1,xmm1 + pxor xmm6,xmm6 + punpcklwd xmm1,xmm5 ; xmm1=BOL + punpckhwd xmm6,xmm5 ; xmm6=BOH + psrld xmm1,1 ; xmm1=BOL*FIX(0.500) + psrld xmm6,1 ; xmm6=BOH*FIX(0.500) + + movdqa xmm5,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; xmm5=[PD_ONEHALFM1_CJ] + + paddd xmm7,xmm1 + paddd xmm4,xmm6 + paddd xmm7,xmm5 + paddd xmm4,xmm5 + psrld xmm7,SCALEBITS ; xmm7=CbOL + psrld xmm4,SCALEBITS ; xmm4=CbOH + packssdw xmm7,xmm4 ; xmm7=CbO + + movdqa xmm1, XMMWORD [wk(2)] ; xmm1=BE + + movdqa xmm6,xmm0 + punpcklwd xmm0,xmm2 + punpckhwd xmm6,xmm2 + movdqa xmm5,xmm0 + movdqa xmm4,xmm6 + pmaddwd xmm0,[GOTOFF(eax,PW_F0299_F0337)] ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337) + pmaddwd xmm6,[GOTOFF(eax,PW_F0299_F0337)] ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337) + pmaddwd xmm5,[GOTOFF(eax,PW_MF016_MF033)] ; xmm5=REL*-FIX(0.168)+GEL*-FIX(0.331) + pmaddwd xmm4,[GOTOFF(eax,PW_MF016_MF033)] ; xmm4=REH*-FIX(0.168)+GEH*-FIX(0.331) + + movdqa XMMWORD [wk(6)], xmm0 ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337) + movdqa XMMWORD [wk(7)], xmm6 ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337) + + pxor xmm0,xmm0 + pxor xmm6,xmm6 + punpcklwd xmm0,xmm1 ; xmm0=BEL + punpckhwd xmm6,xmm1 ; xmm6=BEH + psrld xmm0,1 ; xmm0=BEL*FIX(0.500) + psrld xmm6,1 ; xmm6=BEH*FIX(0.500) + + movdqa xmm1,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; xmm1=[PD_ONEHALFM1_CJ] + + paddd xmm5,xmm0 + paddd xmm4,xmm6 + paddd xmm5,xmm1 + paddd xmm4,xmm1 + psrld xmm5,SCALEBITS ; xmm5=CbEL + psrld xmm4,SCALEBITS ; xmm4=CbEH + packssdw xmm5,xmm4 ; xmm5=CbE + + psllw xmm7,BYTE_BIT + por xmm5,xmm7 ; xmm5=Cb + movdqa XMMWORD [ebx], xmm5 ; Save Cb + + movdqa xmm0, XMMWORD [wk(3)] ; xmm0=BO + movdqa xmm6, XMMWORD [wk(2)] ; xmm6=BE + movdqa xmm1, XMMWORD [wk(1)] ; xmm1=RO + + movdqa xmm4,xmm0 + punpcklwd xmm0,xmm3 + punpckhwd xmm4,xmm3 + movdqa xmm7,xmm0 + movdqa xmm5,xmm4 + pmaddwd xmm0,[GOTOFF(eax,PW_F0114_F0250)] ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250) + pmaddwd xmm4,[GOTOFF(eax,PW_F0114_F0250)] ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250) + pmaddwd xmm7,[GOTOFF(eax,PW_MF008_MF041)] ; xmm7=BOL*-FIX(0.081)+GOL*-FIX(0.418) + pmaddwd xmm5,[GOTOFF(eax,PW_MF008_MF041)] ; xmm5=BOH*-FIX(0.081)+GOH*-FIX(0.418) + + movdqa xmm3,[GOTOFF(eax,PD_ONEHALF)] ; xmm3=[PD_ONEHALF] + + paddd xmm0, XMMWORD [wk(4)] + paddd xmm4, XMMWORD [wk(5)] + paddd xmm0,xmm3 + paddd xmm4,xmm3 + psrld xmm0,SCALEBITS ; xmm0=YOL + psrld xmm4,SCALEBITS ; xmm4=YOH + packssdw xmm0,xmm4 ; xmm0=YO + + pxor xmm3,xmm3 + pxor xmm4,xmm4 + punpcklwd xmm3,xmm1 ; xmm3=ROL + punpckhwd xmm4,xmm1 ; xmm4=ROH + psrld xmm3,1 ; xmm3=ROL*FIX(0.500) + psrld xmm4,1 ; xmm4=ROH*FIX(0.500) + + movdqa xmm1,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; xmm1=[PD_ONEHALFM1_CJ] + + paddd xmm7,xmm3 + paddd xmm5,xmm4 + paddd xmm7,xmm1 + paddd xmm5,xmm1 + psrld xmm7,SCALEBITS ; xmm7=CrOL + psrld xmm5,SCALEBITS ; xmm5=CrOH + packssdw xmm7,xmm5 ; xmm7=CrO + + movdqa xmm3, XMMWORD [wk(0)] ; xmm3=RE + + movdqa xmm4,xmm6 + punpcklwd xmm6,xmm2 + punpckhwd xmm4,xmm2 + movdqa xmm1,xmm6 + movdqa xmm5,xmm4 + pmaddwd xmm6,[GOTOFF(eax,PW_F0114_F0250)] ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250) + pmaddwd xmm4,[GOTOFF(eax,PW_F0114_F0250)] ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250) + pmaddwd xmm1,[GOTOFF(eax,PW_MF008_MF041)] ; xmm1=BEL*-FIX(0.081)+GEL*-FIX(0.418) + pmaddwd xmm5,[GOTOFF(eax,PW_MF008_MF041)] ; xmm5=BEH*-FIX(0.081)+GEH*-FIX(0.418) + + movdqa xmm2,[GOTOFF(eax,PD_ONEHALF)] ; xmm2=[PD_ONEHALF] + + paddd xmm6, XMMWORD [wk(6)] + paddd xmm4, XMMWORD [wk(7)] + paddd xmm6,xmm2 + paddd xmm4,xmm2 + psrld xmm6,SCALEBITS ; xmm6=YEL + psrld xmm4,SCALEBITS ; xmm4=YEH + packssdw xmm6,xmm4 ; xmm6=YE + + psllw xmm0,BYTE_BIT + por xmm6,xmm0 ; xmm6=Y + movdqa XMMWORD [edi], xmm6 ; Save Y + + pxor xmm2,xmm2 + pxor xmm4,xmm4 + punpcklwd xmm2,xmm3 ; xmm2=REL + punpckhwd xmm4,xmm3 ; xmm4=REH + psrld xmm2,1 ; xmm2=REL*FIX(0.500) + psrld xmm4,1 ; xmm4=REH*FIX(0.500) + + movdqa xmm0,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; xmm0=[PD_ONEHALFM1_CJ] + + paddd xmm1,xmm2 + paddd xmm5,xmm4 + paddd xmm1,xmm0 + paddd xmm5,xmm0 + psrld xmm1,SCALEBITS ; xmm1=CrEL + psrld xmm5,SCALEBITS ; xmm5=CrEH + packssdw xmm1,xmm5 ; xmm1=CrE + + psllw xmm7,BYTE_BIT + por xmm1,xmm7 ; xmm1=Cr + movdqa XMMWORD [edx], xmm1 ; Save Cr + + sub ecx, byte SIZEOF_XMMWORD + add esi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; inptr + add edi, byte SIZEOF_XMMWORD ; outptr0 + add ebx, byte SIZEOF_XMMWORD ; outptr1 + add edx, byte SIZEOF_XMMWORD ; outptr2 + cmp ecx, byte SIZEOF_XMMWORD + jae near .columnloop + test ecx,ecx + jnz near .column_ld1 + + pop ecx ; col + pop esi + pop edi + pop ebx + pop edx + poppic eax + + add esi, byte SIZEOF_JSAMPROW ; input_buf + add edi, byte SIZEOF_JSAMPROW + add ebx, byte SIZEOF_JSAMPROW + add edx, byte SIZEOF_JSAMPROW + dec eax ; num_rows + jg near .rowloop + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jccolor-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jccolor-mmx.asm new file mode 100644 index 000000000..c5d376443 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jccolor-mmx.asm @@ -0,0 +1,123 @@ +; +; jccolor.asm - colorspace conversion (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + +%define SCALEBITS 16 + +F_0_081 equ 5329 ; FIX(0.08131) +F_0_114 equ 7471 ; FIX(0.11400) +F_0_168 equ 11059 ; FIX(0.16874) +F_0_250 equ 16384 ; FIX(0.25000) +F_0_299 equ 19595 ; FIX(0.29900) +F_0_331 equ 21709 ; FIX(0.33126) +F_0_418 equ 27439 ; FIX(0.41869) +F_0_587 equ 38470 ; FIX(0.58700) +F_0_337 equ (F_0_587 - F_0_250) ; FIX(0.58700) - FIX(0.25000) + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_rgb_ycc_convert_mmx) + +EXTN(jconst_rgb_ycc_convert_mmx): + +PW_F0299_F0337 times 2 dw F_0_299, F_0_337 +PW_F0114_F0250 times 2 dw F_0_114, F_0_250 +PW_MF016_MF033 times 2 dw -F_0_168,-F_0_331 +PW_MF008_MF041 times 2 dw -F_0_081,-F_0_418 +PD_ONEHALFM1_CJ times 2 dd (1 << (SCALEBITS-1)) - 1 + (CENTERJSAMPLE << SCALEBITS) +PD_ONEHALF times 2 dd (1 << (SCALEBITS-1)) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 + +%include "jccolext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGB_RED +%define RGB_GREEN EXT_RGB_GREEN +%define RGB_BLUE EXT_RGB_BLUE +%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE +%define jsimd_rgb_ycc_convert_mmx jsimd_extrgb_ycc_convert_mmx +%include "jccolext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGBX_RED +%define RGB_GREEN EXT_RGBX_GREEN +%define RGB_BLUE EXT_RGBX_BLUE +%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE +%define jsimd_rgb_ycc_convert_mmx jsimd_extrgbx_ycc_convert_mmx +%include "jccolext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGR_RED +%define RGB_GREEN EXT_BGR_GREEN +%define RGB_BLUE EXT_BGR_BLUE +%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE +%define jsimd_rgb_ycc_convert_mmx jsimd_extbgr_ycc_convert_mmx +%include "jccolext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGRX_RED +%define RGB_GREEN EXT_BGRX_GREEN +%define RGB_BLUE EXT_BGRX_BLUE +%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE +%define jsimd_rgb_ycc_convert_mmx jsimd_extbgrx_ycc_convert_mmx +%include "jccolext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XBGR_RED +%define RGB_GREEN EXT_XBGR_GREEN +%define RGB_BLUE EXT_XBGR_BLUE +%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE +%define jsimd_rgb_ycc_convert_mmx jsimd_extxbgr_ycc_convert_mmx +%include "jccolext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XRGB_RED +%define RGB_GREEN EXT_XRGB_GREEN +%define RGB_BLUE EXT_XRGB_BLUE +%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE +%define jsimd_rgb_ycc_convert_mmx jsimd_extxrgb_ycc_convert_mmx +%include "jccolext-mmx.asm" diff --git a/app/src/main/jni/libjpeg-turbo/simd/jccolor-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jccolor-sse2-64.asm new file mode 100644 index 000000000..55c7e12ec --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jccolor-sse2-64.asm @@ -0,0 +1,120 @@ +; +; jccolor.asm - colorspace conversion (64-bit SSE2) +; +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; Copyright (C) 2009, D. R. Commander. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + +%define SCALEBITS 16 + +F_0_081 equ 5329 ; FIX(0.08131) +F_0_114 equ 7471 ; FIX(0.11400) +F_0_168 equ 11059 ; FIX(0.16874) +F_0_250 equ 16384 ; FIX(0.25000) +F_0_299 equ 19595 ; FIX(0.29900) +F_0_331 equ 21709 ; FIX(0.33126) +F_0_418 equ 27439 ; FIX(0.41869) +F_0_587 equ 38470 ; FIX(0.58700) +F_0_337 equ (F_0_587 - F_0_250) ; FIX(0.58700) - FIX(0.25000) + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_rgb_ycc_convert_sse2) + +EXTN(jconst_rgb_ycc_convert_sse2): + +PW_F0299_F0337 times 4 dw F_0_299, F_0_337 +PW_F0114_F0250 times 4 dw F_0_114, F_0_250 +PW_MF016_MF033 times 4 dw -F_0_168,-F_0_331 +PW_MF008_MF041 times 4 dw -F_0_081,-F_0_418 +PD_ONEHALFM1_CJ times 4 dd (1 << (SCALEBITS-1)) - 1 + (CENTERJSAMPLE << SCALEBITS) +PD_ONEHALF times 4 dd (1 << (SCALEBITS-1)) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 + +%include "jccolext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGB_RED +%define RGB_GREEN EXT_RGB_GREEN +%define RGB_BLUE EXT_RGB_BLUE +%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE +%define jsimd_rgb_ycc_convert_sse2 jsimd_extrgb_ycc_convert_sse2 +%include "jccolext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGBX_RED +%define RGB_GREEN EXT_RGBX_GREEN +%define RGB_BLUE EXT_RGBX_BLUE +%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE +%define jsimd_rgb_ycc_convert_sse2 jsimd_extrgbx_ycc_convert_sse2 +%include "jccolext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGR_RED +%define RGB_GREEN EXT_BGR_GREEN +%define RGB_BLUE EXT_BGR_BLUE +%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE +%define jsimd_rgb_ycc_convert_sse2 jsimd_extbgr_ycc_convert_sse2 +%include "jccolext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGRX_RED +%define RGB_GREEN EXT_BGRX_GREEN +%define RGB_BLUE EXT_BGRX_BLUE +%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE +%define jsimd_rgb_ycc_convert_sse2 jsimd_extbgrx_ycc_convert_sse2 +%include "jccolext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XBGR_RED +%define RGB_GREEN EXT_XBGR_GREEN +%define RGB_BLUE EXT_XBGR_BLUE +%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE +%define jsimd_rgb_ycc_convert_sse2 jsimd_extxbgr_ycc_convert_sse2 +%include "jccolext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XRGB_RED +%define RGB_GREEN EXT_XRGB_GREEN +%define RGB_BLUE EXT_XRGB_BLUE +%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE +%define jsimd_rgb_ycc_convert_sse2 jsimd_extxrgb_ycc_convert_sse2 +%include "jccolext-sse2-64.asm" diff --git a/app/src/main/jni/libjpeg-turbo/simd/jccolor-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jccolor-sse2.asm new file mode 100644 index 000000000..890e2a3b7 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jccolor-sse2.asm @@ -0,0 +1,120 @@ +; +; jccolor.asm - colorspace conversion (SSE2) +; +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; Copyright (C) 2009, D. R. Commander. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + +%define SCALEBITS 16 + +F_0_081 equ 5329 ; FIX(0.08131) +F_0_114 equ 7471 ; FIX(0.11400) +F_0_168 equ 11059 ; FIX(0.16874) +F_0_250 equ 16384 ; FIX(0.25000) +F_0_299 equ 19595 ; FIX(0.29900) +F_0_331 equ 21709 ; FIX(0.33126) +F_0_418 equ 27439 ; FIX(0.41869) +F_0_587 equ 38470 ; FIX(0.58700) +F_0_337 equ (F_0_587 - F_0_250) ; FIX(0.58700) - FIX(0.25000) + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_rgb_ycc_convert_sse2) + +EXTN(jconst_rgb_ycc_convert_sse2): + +PW_F0299_F0337 times 4 dw F_0_299, F_0_337 +PW_F0114_F0250 times 4 dw F_0_114, F_0_250 +PW_MF016_MF033 times 4 dw -F_0_168,-F_0_331 +PW_MF008_MF041 times 4 dw -F_0_081,-F_0_418 +PD_ONEHALFM1_CJ times 4 dd (1 << (SCALEBITS-1)) - 1 + (CENTERJSAMPLE << SCALEBITS) +PD_ONEHALF times 4 dd (1 << (SCALEBITS-1)) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 + +%include "jccolext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGB_RED +%define RGB_GREEN EXT_RGB_GREEN +%define RGB_BLUE EXT_RGB_BLUE +%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE +%define jsimd_rgb_ycc_convert_sse2 jsimd_extrgb_ycc_convert_sse2 +%include "jccolext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGBX_RED +%define RGB_GREEN EXT_RGBX_GREEN +%define RGB_BLUE EXT_RGBX_BLUE +%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE +%define jsimd_rgb_ycc_convert_sse2 jsimd_extrgbx_ycc_convert_sse2 +%include "jccolext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGR_RED +%define RGB_GREEN EXT_BGR_GREEN +%define RGB_BLUE EXT_BGR_BLUE +%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE +%define jsimd_rgb_ycc_convert_sse2 jsimd_extbgr_ycc_convert_sse2 +%include "jccolext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGRX_RED +%define RGB_GREEN EXT_BGRX_GREEN +%define RGB_BLUE EXT_BGRX_BLUE +%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE +%define jsimd_rgb_ycc_convert_sse2 jsimd_extbgrx_ycc_convert_sse2 +%include "jccolext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XBGR_RED +%define RGB_GREEN EXT_XBGR_GREEN +%define RGB_BLUE EXT_XBGR_BLUE +%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE +%define jsimd_rgb_ycc_convert_sse2 jsimd_extxbgr_ycc_convert_sse2 +%include "jccolext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XRGB_RED +%define RGB_GREEN EXT_XRGB_GREEN +%define RGB_BLUE EXT_XRGB_BLUE +%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE +%define jsimd_rgb_ycc_convert_sse2 jsimd_extxrgb_ycc_convert_sse2 +%include "jccolext-sse2.asm" diff --git a/app/src/main/jni/libjpeg-turbo/simd/jcgray-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jcgray-mmx.asm new file mode 100644 index 000000000..b2708ad4d --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jcgray-mmx.asm @@ -0,0 +1,116 @@ +; +; jcgray.asm - grayscale colorspace conversion (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2011 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + +%define SCALEBITS 16 + +F_0_114 equ 7471 ; FIX(0.11400) +F_0_250 equ 16384 ; FIX(0.25000) +F_0_299 equ 19595 ; FIX(0.29900) +F_0_587 equ 38470 ; FIX(0.58700) +F_0_337 equ (F_0_587 - F_0_250) ; FIX(0.58700) - FIX(0.25000) + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_rgb_gray_convert_mmx) + +EXTN(jconst_rgb_gray_convert_mmx): + +PW_F0299_F0337 times 2 dw F_0_299, F_0_337 +PW_F0114_F0250 times 2 dw F_0_114, F_0_250 +PD_ONEHALF times 2 dd (1 << (SCALEBITS-1)) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 + +%include "jcgryext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGB_RED +%define RGB_GREEN EXT_RGB_GREEN +%define RGB_BLUE EXT_RGB_BLUE +%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE +%define jsimd_rgb_gray_convert_mmx jsimd_extrgb_gray_convert_mmx +%include "jcgryext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGBX_RED +%define RGB_GREEN EXT_RGBX_GREEN +%define RGB_BLUE EXT_RGBX_BLUE +%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE +%define jsimd_rgb_gray_convert_mmx jsimd_extrgbx_gray_convert_mmx +%include "jcgryext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGR_RED +%define RGB_GREEN EXT_BGR_GREEN +%define RGB_BLUE EXT_BGR_BLUE +%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE +%define jsimd_rgb_gray_convert_mmx jsimd_extbgr_gray_convert_mmx +%include "jcgryext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGRX_RED +%define RGB_GREEN EXT_BGRX_GREEN +%define RGB_BLUE EXT_BGRX_BLUE +%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE +%define jsimd_rgb_gray_convert_mmx jsimd_extbgrx_gray_convert_mmx +%include "jcgryext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XBGR_RED +%define RGB_GREEN EXT_XBGR_GREEN +%define RGB_BLUE EXT_XBGR_BLUE +%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE +%define jsimd_rgb_gray_convert_mmx jsimd_extxbgr_gray_convert_mmx +%include "jcgryext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XRGB_RED +%define RGB_GREEN EXT_XRGB_GREEN +%define RGB_BLUE EXT_XRGB_BLUE +%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE +%define jsimd_rgb_gray_convert_mmx jsimd_extxrgb_gray_convert_mmx +%include "jcgryext-mmx.asm" diff --git a/app/src/main/jni/libjpeg-turbo/simd/jcgray-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jcgray-sse2-64.asm new file mode 100644 index 000000000..dfc057767 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jcgray-sse2-64.asm @@ -0,0 +1,113 @@ +; +; jcgray.asm - grayscale colorspace conversion (64-bit SSE2) +; +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; Copyright (C) 2011, D. R. Commander. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + +%define SCALEBITS 16 + +F_0_114 equ 7471 ; FIX(0.11400) +F_0_250 equ 16384 ; FIX(0.25000) +F_0_299 equ 19595 ; FIX(0.29900) +F_0_587 equ 38470 ; FIX(0.58700) +F_0_337 equ (F_0_587 - F_0_250) ; FIX(0.58700) - FIX(0.25000) + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_rgb_gray_convert_sse2) + +EXTN(jconst_rgb_gray_convert_sse2): + +PW_F0299_F0337 times 4 dw F_0_299, F_0_337 +PW_F0114_F0250 times 4 dw F_0_114, F_0_250 +PD_ONEHALF times 4 dd (1 << (SCALEBITS-1)) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 + +%include "jcgryext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGB_RED +%define RGB_GREEN EXT_RGB_GREEN +%define RGB_BLUE EXT_RGB_BLUE +%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE +%define jsimd_rgb_gray_convert_sse2 jsimd_extrgb_gray_convert_sse2 +%include "jcgryext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGBX_RED +%define RGB_GREEN EXT_RGBX_GREEN +%define RGB_BLUE EXT_RGBX_BLUE +%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE +%define jsimd_rgb_gray_convert_sse2 jsimd_extrgbx_gray_convert_sse2 +%include "jcgryext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGR_RED +%define RGB_GREEN EXT_BGR_GREEN +%define RGB_BLUE EXT_BGR_BLUE +%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE +%define jsimd_rgb_gray_convert_sse2 jsimd_extbgr_gray_convert_sse2 +%include "jcgryext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGRX_RED +%define RGB_GREEN EXT_BGRX_GREEN +%define RGB_BLUE EXT_BGRX_BLUE +%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE +%define jsimd_rgb_gray_convert_sse2 jsimd_extbgrx_gray_convert_sse2 +%include "jcgryext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XBGR_RED +%define RGB_GREEN EXT_XBGR_GREEN +%define RGB_BLUE EXT_XBGR_BLUE +%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE +%define jsimd_rgb_gray_convert_sse2 jsimd_extxbgr_gray_convert_sse2 +%include "jcgryext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XRGB_RED +%define RGB_GREEN EXT_XRGB_GREEN +%define RGB_BLUE EXT_XRGB_BLUE +%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE +%define jsimd_rgb_gray_convert_sse2 jsimd_extxrgb_gray_convert_sse2 +%include "jcgryext-sse2-64.asm" diff --git a/app/src/main/jni/libjpeg-turbo/simd/jcgray-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jcgray-sse2.asm new file mode 100644 index 000000000..5fa72736a --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jcgray-sse2.asm @@ -0,0 +1,113 @@ +; +; jcgray.asm - grayscale colorspace conversion (SSE2) +; +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; Copyright (C) 2011, D. R. Commander. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + +%define SCALEBITS 16 + +F_0_114 equ 7471 ; FIX(0.11400) +F_0_250 equ 16384 ; FIX(0.25000) +F_0_299 equ 19595 ; FIX(0.29900) +F_0_587 equ 38470 ; FIX(0.58700) +F_0_337 equ (F_0_587 - F_0_250) ; FIX(0.58700) - FIX(0.25000) + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_rgb_gray_convert_sse2) + +EXTN(jconst_rgb_gray_convert_sse2): + +PW_F0299_F0337 times 4 dw F_0_299, F_0_337 +PW_F0114_F0250 times 4 dw F_0_114, F_0_250 +PD_ONEHALF times 4 dd (1 << (SCALEBITS-1)) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 + +%include "jcgryext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGB_RED +%define RGB_GREEN EXT_RGB_GREEN +%define RGB_BLUE EXT_RGB_BLUE +%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE +%define jsimd_rgb_gray_convert_sse2 jsimd_extrgb_gray_convert_sse2 +%include "jcgryext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGBX_RED +%define RGB_GREEN EXT_RGBX_GREEN +%define RGB_BLUE EXT_RGBX_BLUE +%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE +%define jsimd_rgb_gray_convert_sse2 jsimd_extrgbx_gray_convert_sse2 +%include "jcgryext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGR_RED +%define RGB_GREEN EXT_BGR_GREEN +%define RGB_BLUE EXT_BGR_BLUE +%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE +%define jsimd_rgb_gray_convert_sse2 jsimd_extbgr_gray_convert_sse2 +%include "jcgryext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGRX_RED +%define RGB_GREEN EXT_BGRX_GREEN +%define RGB_BLUE EXT_BGRX_BLUE +%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE +%define jsimd_rgb_gray_convert_sse2 jsimd_extbgrx_gray_convert_sse2 +%include "jcgryext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XBGR_RED +%define RGB_GREEN EXT_XBGR_GREEN +%define RGB_BLUE EXT_XBGR_BLUE +%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE +%define jsimd_rgb_gray_convert_sse2 jsimd_extxbgr_gray_convert_sse2 +%include "jcgryext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XRGB_RED +%define RGB_GREEN EXT_XRGB_GREEN +%define RGB_BLUE EXT_XRGB_BLUE +%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE +%define jsimd_rgb_gray_convert_sse2 jsimd_extxrgb_gray_convert_sse2 +%include "jcgryext-sse2.asm" diff --git a/app/src/main/jni/libjpeg-turbo/simd/jcgryext-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jcgryext-mmx.asm new file mode 100644 index 000000000..13b960088 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jcgryext-mmx.asm @@ -0,0 +1,357 @@ +; +; jcgryext.asm - grayscale colorspace conversion (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2011 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jcolsamp.inc" + +; -------------------------------------------------------------------------- +; +; Convert some rows of samples to the output colorspace. +; +; GLOBAL(void) +; jsimd_rgb_gray_convert_mmx (JDIMENSION img_width, +; JSAMPARRAY input_buf, JSAMPIMAGE output_buf, +; JDIMENSION output_row, int num_rows); +; + +%define img_width(b) (b)+8 ; JDIMENSION img_width +%define input_buf(b) (b)+12 ; JSAMPARRAY input_buf +%define output_buf(b) (b)+16 ; JSAMPIMAGE output_buf +%define output_row(b) (b)+20 ; JDIMENSION output_row +%define num_rows(b) (b)+24 ; int num_rows + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_MMWORD ; mmword wk[WK_NUM] +%define WK_NUM 2 +%define gotptr wk(0)-SIZEOF_POINTER ; void * gotptr + + align 16 + global EXTN(jsimd_rgb_gray_convert_mmx) + +EXTN(jsimd_rgb_gray_convert_mmx): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_MMWORD) ; align to 64 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic eax ; make a room for GOT address + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + movpic POINTER [gotptr], ebx ; save GOT address + + mov ecx, JDIMENSION [img_width(eax)] ; num_cols + test ecx,ecx + jz near .return + + push ecx + + mov esi, JSAMPIMAGE [output_buf(eax)] + mov ecx, JDIMENSION [output_row(eax)] + mov edi, JSAMPARRAY [esi+0*SIZEOF_JSAMPARRAY] + lea edi, [edi+ecx*SIZEOF_JSAMPROW] + + pop ecx + + mov esi, JSAMPARRAY [input_buf(eax)] + mov eax, INT [num_rows(eax)] + test eax,eax + jle near .return + alignx 16,7 +.rowloop: + pushpic eax + push edi + push esi + push ecx ; col + + mov esi, JSAMPROW [esi] ; inptr + mov edi, JSAMPROW [edi] ; outptr0 + movpic eax, POINTER [gotptr] ; load GOT address (eax) + + cmp ecx, byte SIZEOF_MMWORD + jae short .columnloop + alignx 16,7 + +%if RGB_PIXELSIZE == 3 ; --------------- + +.column_ld1: + push eax + push edx + lea ecx,[ecx+ecx*2] ; imul ecx,RGB_PIXELSIZE + test cl, SIZEOF_BYTE + jz short .column_ld2 + sub ecx, byte SIZEOF_BYTE + xor eax,eax + mov al, BYTE [esi+ecx] +.column_ld2: + test cl, SIZEOF_WORD + jz short .column_ld4 + sub ecx, byte SIZEOF_WORD + xor edx,edx + mov dx, WORD [esi+ecx] + shl eax, WORD_BIT + or eax,edx +.column_ld4: + movd mmA,eax + pop edx + pop eax + test cl, SIZEOF_DWORD + jz short .column_ld8 + sub ecx, byte SIZEOF_DWORD + movd mmG, DWORD [esi+ecx] + psllq mmA, DWORD_BIT + por mmA,mmG +.column_ld8: + test cl, SIZEOF_MMWORD + jz short .column_ld16 + movq mmG,mmA + movq mmA, MMWORD [esi+0*SIZEOF_MMWORD] + mov ecx, SIZEOF_MMWORD + jmp short .rgb_gray_cnv +.column_ld16: + test cl, 2*SIZEOF_MMWORD + mov ecx, SIZEOF_MMWORD + jz short .rgb_gray_cnv + movq mmF,mmA + movq mmA, MMWORD [esi+0*SIZEOF_MMWORD] + movq mmG, MMWORD [esi+1*SIZEOF_MMWORD] + jmp short .rgb_gray_cnv + alignx 16,7 + +.columnloop: + movq mmA, MMWORD [esi+0*SIZEOF_MMWORD] + movq mmG, MMWORD [esi+1*SIZEOF_MMWORD] + movq mmF, MMWORD [esi+2*SIZEOF_MMWORD] + +.rgb_gray_cnv: + ; mmA=(00 10 20 01 11 21 02 12) + ; mmG=(22 03 13 23 04 14 24 05) + ; mmF=(15 25 06 16 26 07 17 27) + + movq mmD,mmA + psllq mmA,4*BYTE_BIT ; mmA=(-- -- -- -- 00 10 20 01) + psrlq mmD,4*BYTE_BIT ; mmD=(11 21 02 12 -- -- -- --) + + punpckhbw mmA,mmG ; mmA=(00 04 10 14 20 24 01 05) + psllq mmG,4*BYTE_BIT ; mmG=(-- -- -- -- 22 03 13 23) + + punpcklbw mmD,mmF ; mmD=(11 15 21 25 02 06 12 16) + punpckhbw mmG,mmF ; mmG=(22 26 03 07 13 17 23 27) + + movq mmE,mmA + psllq mmA,4*BYTE_BIT ; mmA=(-- -- -- -- 00 04 10 14) + psrlq mmE,4*BYTE_BIT ; mmE=(20 24 01 05 -- -- -- --) + + punpckhbw mmA,mmD ; mmA=(00 02 04 06 10 12 14 16) + psllq mmD,4*BYTE_BIT ; mmD=(-- -- -- -- 11 15 21 25) + + punpcklbw mmE,mmG ; mmE=(20 22 24 26 01 03 05 07) + punpckhbw mmD,mmG ; mmD=(11 13 15 17 21 23 25 27) + + pxor mmH,mmH + + movq mmC,mmA + punpcklbw mmA,mmH ; mmA=(00 02 04 06) + punpckhbw mmC,mmH ; mmC=(10 12 14 16) + + movq mmB,mmE + punpcklbw mmE,mmH ; mmE=(20 22 24 26) + punpckhbw mmB,mmH ; mmB=(01 03 05 07) + + movq mmF,mmD + punpcklbw mmD,mmH ; mmD=(11 13 15 17) + punpckhbw mmF,mmH ; mmF=(21 23 25 27) + +%else ; RGB_PIXELSIZE == 4 ; ----------- + +.column_ld1: + test cl, SIZEOF_MMWORD/8 + jz short .column_ld2 + sub ecx, byte SIZEOF_MMWORD/8 + movd mmA, DWORD [esi+ecx*RGB_PIXELSIZE] +.column_ld2: + test cl, SIZEOF_MMWORD/4 + jz short .column_ld4 + sub ecx, byte SIZEOF_MMWORD/4 + movq mmF,mmA + movq mmA, MMWORD [esi+ecx*RGB_PIXELSIZE] +.column_ld4: + test cl, SIZEOF_MMWORD/2 + mov ecx, SIZEOF_MMWORD + jz short .rgb_gray_cnv + movq mmD,mmA + movq mmC,mmF + movq mmA, MMWORD [esi+0*SIZEOF_MMWORD] + movq mmF, MMWORD [esi+1*SIZEOF_MMWORD] + jmp short .rgb_gray_cnv + alignx 16,7 + +.columnloop: + movq mmA, MMWORD [esi+0*SIZEOF_MMWORD] + movq mmF, MMWORD [esi+1*SIZEOF_MMWORD] + movq mmD, MMWORD [esi+2*SIZEOF_MMWORD] + movq mmC, MMWORD [esi+3*SIZEOF_MMWORD] + +.rgb_gray_cnv: + ; mmA=(00 10 20 30 01 11 21 31) + ; mmF=(02 12 22 32 03 13 23 33) + ; mmD=(04 14 24 34 05 15 25 35) + ; mmC=(06 16 26 36 07 17 27 37) + + movq mmB,mmA + punpcklbw mmA,mmF ; mmA=(00 02 10 12 20 22 30 32) + punpckhbw mmB,mmF ; mmB=(01 03 11 13 21 23 31 33) + + movq mmG,mmD + punpcklbw mmD,mmC ; mmD=(04 06 14 16 24 26 34 36) + punpckhbw mmG,mmC ; mmG=(05 07 15 17 25 27 35 37) + + movq mmE,mmA + punpcklwd mmA,mmD ; mmA=(00 02 04 06 10 12 14 16) + punpckhwd mmE,mmD ; mmE=(20 22 24 26 30 32 34 36) + + movq mmH,mmB + punpcklwd mmB,mmG ; mmB=(01 03 05 07 11 13 15 17) + punpckhwd mmH,mmG ; mmH=(21 23 25 27 31 33 35 37) + + pxor mmF,mmF + + movq mmC,mmA + punpcklbw mmA,mmF ; mmA=(00 02 04 06) + punpckhbw mmC,mmF ; mmC=(10 12 14 16) + + movq mmD,mmB + punpcklbw mmB,mmF ; mmB=(01 03 05 07) + punpckhbw mmD,mmF ; mmD=(11 13 15 17) + + movq mmG,mmE + punpcklbw mmE,mmF ; mmE=(20 22 24 26) + punpckhbw mmG,mmF ; mmG=(30 32 34 36) + + punpcklbw mmF,mmH + punpckhbw mmH,mmH + psrlw mmF,BYTE_BIT ; mmF=(21 23 25 27) + psrlw mmH,BYTE_BIT ; mmH=(31 33 35 37) + +%endif ; RGB_PIXELSIZE ; --------------- + + ; mm0=(R0 R2 R4 R6)=RE, mm2=(G0 G2 G4 G6)=GE, mm4=(B0 B2 B4 B6)=BE + ; mm1=(R1 R3 R5 R7)=RO, mm3=(G1 G3 G5 G7)=GO, mm5=(B1 B3 B5 B7)=BO + + ; (Original) + ; Y = 0.29900 * R + 0.58700 * G + 0.11400 * B + ; + ; (This implementation) + ; Y = 0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G + + movq mm6,mm1 + punpcklwd mm1,mm3 + punpckhwd mm6,mm3 + pmaddwd mm1,[GOTOFF(eax,PW_F0299_F0337)] ; mm1=ROL*FIX(0.299)+GOL*FIX(0.337) + pmaddwd mm6,[GOTOFF(eax,PW_F0299_F0337)] ; mm6=ROH*FIX(0.299)+GOH*FIX(0.337) + + movq mm7, mm6 ; mm7=ROH*FIX(0.299)+GOH*FIX(0.337) + + movq mm6,mm0 + punpcklwd mm0,mm2 + punpckhwd mm6,mm2 + pmaddwd mm0,[GOTOFF(eax,PW_F0299_F0337)] ; mm0=REL*FIX(0.299)+GEL*FIX(0.337) + pmaddwd mm6,[GOTOFF(eax,PW_F0299_F0337)] ; mm6=REH*FIX(0.299)+GEH*FIX(0.337) + + movq MMWORD [wk(0)], mm0 ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337) + movq MMWORD [wk(1)], mm6 ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337) + + movq mm0, mm5 ; mm0=BO + movq mm6, mm4 ; mm6=BE + + movq mm4,mm0 + punpcklwd mm0,mm3 + punpckhwd mm4,mm3 + pmaddwd mm0,[GOTOFF(eax,PW_F0114_F0250)] ; mm0=BOL*FIX(0.114)+GOL*FIX(0.250) + pmaddwd mm4,[GOTOFF(eax,PW_F0114_F0250)] ; mm4=BOH*FIX(0.114)+GOH*FIX(0.250) + + movq mm3,[GOTOFF(eax,PD_ONEHALF)] ; mm3=[PD_ONEHALF] + + paddd mm0, mm1 + paddd mm4, mm7 + paddd mm0,mm3 + paddd mm4,mm3 + psrld mm0,SCALEBITS ; mm0=YOL + psrld mm4,SCALEBITS ; mm4=YOH + packssdw mm0,mm4 ; mm0=YO + + movq mm4,mm6 + punpcklwd mm6,mm2 + punpckhwd mm4,mm2 + pmaddwd mm6,[GOTOFF(eax,PW_F0114_F0250)] ; mm6=BEL*FIX(0.114)+GEL*FIX(0.250) + pmaddwd mm4,[GOTOFF(eax,PW_F0114_F0250)] ; mm4=BEH*FIX(0.114)+GEH*FIX(0.250) + + movq mm2,[GOTOFF(eax,PD_ONEHALF)] ; mm2=[PD_ONEHALF] + + paddd mm6, MMWORD [wk(0)] + paddd mm4, MMWORD [wk(1)] + paddd mm6,mm2 + paddd mm4,mm2 + psrld mm6,SCALEBITS ; mm6=YEL + psrld mm4,SCALEBITS ; mm4=YEH + packssdw mm6,mm4 ; mm6=YE + + psllw mm0,BYTE_BIT + por mm6,mm0 ; mm6=Y + movq MMWORD [edi], mm6 ; Save Y + + sub ecx, byte SIZEOF_MMWORD + add esi, byte RGB_PIXELSIZE*SIZEOF_MMWORD ; inptr + add edi, byte SIZEOF_MMWORD ; outptr0 + cmp ecx, byte SIZEOF_MMWORD + jae near .columnloop + test ecx,ecx + jnz near .column_ld1 + + pop ecx ; col + pop esi + pop edi + poppic eax + + add esi, byte SIZEOF_JSAMPROW ; input_buf + add edi, byte SIZEOF_JSAMPROW + dec eax ; num_rows + jg near .rowloop + + emms ; empty MMX state + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jcgryext-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jcgryext-sse2-64.asm new file mode 100644 index 000000000..5af02e04b --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jcgryext-sse2-64.asm @@ -0,0 +1,364 @@ +; +; jcgryext.asm - grayscale colorspace conversion (64-bit SSE2) +; +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; Copyright (C) 2011, D. R. Commander. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jcolsamp.inc" + +; -------------------------------------------------------------------------- +; +; Convert some rows of samples to the output colorspace. +; +; GLOBAL(void) +; jsimd_rgb_gray_convert_sse2 (JDIMENSION img_width, +; JSAMPARRAY input_buf, JSAMPIMAGE output_buf, +; JDIMENSION output_row, int num_rows); +; + +; r10 = JDIMENSION img_width +; r11 = JSAMPARRAY input_buf +; r12 = JSAMPIMAGE output_buf +; r13 = JDIMENSION output_row +; r14 = int num_rows + +%define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 + + align 16 + + global EXTN(jsimd_rgb_gray_convert_sse2) + +EXTN(jsimd_rgb_gray_convert_sse2): + push rbp + mov rax,rsp ; rax = original rbp + sub rsp, byte 4 + and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [rsp],rax + mov rbp,rsp ; rbp = aligned rbp + lea rsp, [wk(0)] + collect_args + push rbx + + mov rcx, r10 + test rcx,rcx + jz near .return + + push rcx + + mov rsi, r12 + mov rcx, r13 + mov rdi, JSAMPARRAY [rsi+0*SIZEOF_JSAMPARRAY] + lea rdi, [rdi+rcx*SIZEOF_JSAMPROW] + + pop rcx + + mov rsi, r11 + mov eax, r14d + test rax,rax + jle near .return +.rowloop: + push rdi + push rsi + push rcx ; col + + mov rsi, JSAMPROW [rsi] ; inptr + mov rdi, JSAMPROW [rdi] ; outptr0 + + cmp rcx, byte SIZEOF_XMMWORD + jae near .columnloop + +%if RGB_PIXELSIZE == 3 ; --------------- + +.column_ld1: + push rax + push rdx + lea rcx,[rcx+rcx*2] ; imul ecx,RGB_PIXELSIZE + test cl, SIZEOF_BYTE + jz short .column_ld2 + sub rcx, byte SIZEOF_BYTE + movzx rax, BYTE [rsi+rcx] +.column_ld2: + test cl, SIZEOF_WORD + jz short .column_ld4 + sub rcx, byte SIZEOF_WORD + movzx rdx, WORD [rsi+rcx] + shl rax, WORD_BIT + or rax,rdx +.column_ld4: + movd xmmA,eax + pop rdx + pop rax + test cl, SIZEOF_DWORD + jz short .column_ld8 + sub rcx, byte SIZEOF_DWORD + movd xmmF, XMM_DWORD [rsi+rcx] + pslldq xmmA, SIZEOF_DWORD + por xmmA,xmmF +.column_ld8: + test cl, SIZEOF_MMWORD + jz short .column_ld16 + sub rcx, byte SIZEOF_MMWORD + movq xmmB, XMM_MMWORD [rsi+rcx] + pslldq xmmA, SIZEOF_MMWORD + por xmmA,xmmB +.column_ld16: + test cl, SIZEOF_XMMWORD + jz short .column_ld32 + movdqa xmmF,xmmA + movdqu xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD] + mov rcx, SIZEOF_XMMWORD + jmp short .rgb_gray_cnv +.column_ld32: + test cl, 2*SIZEOF_XMMWORD + mov rcx, SIZEOF_XMMWORD + jz short .rgb_gray_cnv + movdqa xmmB,xmmA + movdqu xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD] + movdqu xmmF, XMMWORD [rsi+1*SIZEOF_XMMWORD] + jmp short .rgb_gray_cnv + +.columnloop: + movdqu xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD] + movdqu xmmF, XMMWORD [rsi+1*SIZEOF_XMMWORD] + movdqu xmmB, XMMWORD [rsi+2*SIZEOF_XMMWORD] + +.rgb_gray_cnv: + ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05) + ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A) + ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F) + + movdqa xmmG,xmmA + pslldq xmmA,8 ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12) + psrldq xmmG,8 ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --) + + punpckhbw xmmA,xmmF ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A) + pslldq xmmF,8 ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27) + + punpcklbw xmmG,xmmB ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D) + punpckhbw xmmF,xmmB ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F) + + movdqa xmmD,xmmA + pslldq xmmA,8 ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09) + psrldq xmmD,8 ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --) + + punpckhbw xmmA,xmmG ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D) + pslldq xmmG,8 ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B) + + punpcklbw xmmD,xmmF ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E) + punpckhbw xmmG,xmmF ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F) + + movdqa xmmE,xmmA + pslldq xmmA,8 ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C) + psrldq xmmE,8 ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --) + + punpckhbw xmmA,xmmD ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E) + pslldq xmmD,8 ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D) + + punpcklbw xmmE,xmmG ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F) + punpckhbw xmmD,xmmG ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F) + + pxor xmmH,xmmH + + movdqa xmmC,xmmA + punpcklbw xmmA,xmmH ; xmmA=(00 02 04 06 08 0A 0C 0E) + punpckhbw xmmC,xmmH ; xmmC=(10 12 14 16 18 1A 1C 1E) + + movdqa xmmB,xmmE + punpcklbw xmmE,xmmH ; xmmE=(20 22 24 26 28 2A 2C 2E) + punpckhbw xmmB,xmmH ; xmmB=(01 03 05 07 09 0B 0D 0F) + + movdqa xmmF,xmmD + punpcklbw xmmD,xmmH ; xmmD=(11 13 15 17 19 1B 1D 1F) + punpckhbw xmmF,xmmH ; xmmF=(21 23 25 27 29 2B 2D 2F) + +%else ; RGB_PIXELSIZE == 4 ; ----------- + +.column_ld1: + test cl, SIZEOF_XMMWORD/16 + jz short .column_ld2 + sub rcx, byte SIZEOF_XMMWORD/16 + movd xmmA, XMM_DWORD [rsi+rcx*RGB_PIXELSIZE] +.column_ld2: + test cl, SIZEOF_XMMWORD/8 + jz short .column_ld4 + sub rcx, byte SIZEOF_XMMWORD/8 + movq xmmE, XMM_MMWORD [rsi+rcx*RGB_PIXELSIZE] + pslldq xmmA, SIZEOF_MMWORD + por xmmA,xmmE +.column_ld4: + test cl, SIZEOF_XMMWORD/4 + jz short .column_ld8 + sub rcx, byte SIZEOF_XMMWORD/4 + movdqa xmmE,xmmA + movdqu xmmA, XMMWORD [rsi+rcx*RGB_PIXELSIZE] +.column_ld8: + test cl, SIZEOF_XMMWORD/2 + mov rcx, SIZEOF_XMMWORD + jz short .rgb_gray_cnv + movdqa xmmF,xmmA + movdqa xmmH,xmmE + movdqu xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD] + movdqu xmmE, XMMWORD [rsi+1*SIZEOF_XMMWORD] + jmp short .rgb_gray_cnv + +.columnloop: + movdqu xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD] + movdqu xmmE, XMMWORD [rsi+1*SIZEOF_XMMWORD] + movdqu xmmF, XMMWORD [rsi+2*SIZEOF_XMMWORD] + movdqu xmmH, XMMWORD [rsi+3*SIZEOF_XMMWORD] + +.rgb_gray_cnv: + ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33) + ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37) + ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B) + ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F) + + movdqa xmmD,xmmA + punpcklbw xmmA,xmmE ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35) + punpckhbw xmmD,xmmE ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37) + + movdqa xmmC,xmmF + punpcklbw xmmF,xmmH ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D) + punpckhbw xmmC,xmmH ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F) + + movdqa xmmB,xmmA + punpcklwd xmmA,xmmF ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C) + punpckhwd xmmB,xmmF ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D) + + movdqa xmmG,xmmD + punpcklwd xmmD,xmmC ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E) + punpckhwd xmmG,xmmC ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F) + + movdqa xmmE,xmmA + punpcklbw xmmA,xmmD ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E) + punpckhbw xmmE,xmmD ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E) + + movdqa xmmH,xmmB + punpcklbw xmmB,xmmG ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F) + punpckhbw xmmH,xmmG ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F) + + pxor xmmF,xmmF + + movdqa xmmC,xmmA + punpcklbw xmmA,xmmF ; xmmA=(00 02 04 06 08 0A 0C 0E) + punpckhbw xmmC,xmmF ; xmmC=(10 12 14 16 18 1A 1C 1E) + + movdqa xmmD,xmmB + punpcklbw xmmB,xmmF ; xmmB=(01 03 05 07 09 0B 0D 0F) + punpckhbw xmmD,xmmF ; xmmD=(11 13 15 17 19 1B 1D 1F) + + movdqa xmmG,xmmE + punpcklbw xmmE,xmmF ; xmmE=(20 22 24 26 28 2A 2C 2E) + punpckhbw xmmG,xmmF ; xmmG=(30 32 34 36 38 3A 3C 3E) + + punpcklbw xmmF,xmmH + punpckhbw xmmH,xmmH + psrlw xmmF,BYTE_BIT ; xmmF=(21 23 25 27 29 2B 2D 2F) + psrlw xmmH,BYTE_BIT ; xmmH=(31 33 35 37 39 3B 3D 3F) + +%endif ; RGB_PIXELSIZE ; --------------- + + ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE + ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO + + ; (Original) + ; Y = 0.29900 * R + 0.58700 * G + 0.11400 * B + ; + ; (This implementation) + ; Y = 0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G + + movdqa xmm6,xmm1 + punpcklwd xmm1,xmm3 + punpckhwd xmm6,xmm3 + pmaddwd xmm1,[rel PW_F0299_F0337] ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337) + pmaddwd xmm6,[rel PW_F0299_F0337] ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337) + + movdqa xmm7, xmm6 ; xmm7=ROH*FIX(0.299)+GOH*FIX(0.337) + + movdqa xmm6,xmm0 + punpcklwd xmm0,xmm2 + punpckhwd xmm6,xmm2 + pmaddwd xmm0,[rel PW_F0299_F0337] ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337) + pmaddwd xmm6,[rel PW_F0299_F0337] ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337) + + movdqa XMMWORD [wk(0)], xmm0 ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337) + movdqa XMMWORD [wk(1)], xmm6 ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337) + + movdqa xmm0, xmm5 ; xmm0=BO + movdqa xmm6, xmm4 ; xmm6=BE + + movdqa xmm4,xmm0 + punpcklwd xmm0,xmm3 + punpckhwd xmm4,xmm3 + pmaddwd xmm0,[rel PW_F0114_F0250] ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250) + pmaddwd xmm4,[rel PW_F0114_F0250] ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250) + + movdqa xmm3,[rel PD_ONEHALF] ; xmm3=[PD_ONEHALF] + + paddd xmm0, xmm1 + paddd xmm4, xmm7 + paddd xmm0,xmm3 + paddd xmm4,xmm3 + psrld xmm0,SCALEBITS ; xmm0=YOL + psrld xmm4,SCALEBITS ; xmm4=YOH + packssdw xmm0,xmm4 ; xmm0=YO + + movdqa xmm4,xmm6 + punpcklwd xmm6,xmm2 + punpckhwd xmm4,xmm2 + pmaddwd xmm6,[rel PW_F0114_F0250] ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250) + pmaddwd xmm4,[rel PW_F0114_F0250] ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250) + + movdqa xmm2,[rel PD_ONEHALF] ; xmm2=[PD_ONEHALF] + + paddd xmm6, XMMWORD [wk(0)] + paddd xmm4, XMMWORD [wk(1)] + paddd xmm6,xmm2 + paddd xmm4,xmm2 + psrld xmm6,SCALEBITS ; xmm6=YEL + psrld xmm4,SCALEBITS ; xmm4=YEH + packssdw xmm6,xmm4 ; xmm6=YE + + psllw xmm0,BYTE_BIT + por xmm6,xmm0 ; xmm6=Y + movdqa XMMWORD [rdi], xmm6 ; Save Y + + sub rcx, byte SIZEOF_XMMWORD + add rsi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; inptr + add rdi, byte SIZEOF_XMMWORD ; outptr0 + cmp rcx, byte SIZEOF_XMMWORD + jae near .columnloop + test rcx,rcx + jnz near .column_ld1 + + pop rcx ; col + pop rsi + pop rdi + + add rsi, byte SIZEOF_JSAMPROW ; input_buf + add rdi, byte SIZEOF_JSAMPROW + dec rax ; num_rows + jg near .rowloop + +.return: + pop rbx + uncollect_args + mov rsp,rbp ; rsp <- aligned rbp + pop rsp ; rsp <- original rbp + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jcgryext-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jcgryext-sse2.asm new file mode 100644 index 000000000..1097b29df --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jcgryext-sse2.asm @@ -0,0 +1,383 @@ +; +; jcgryext.asm - grayscale colorspace conversion (SSE2) +; +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; Copyright (C) 2011, D. R. Commander. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jcolsamp.inc" + +; -------------------------------------------------------------------------- +; +; Convert some rows of samples to the output colorspace. +; +; GLOBAL(void) +; jsimd_rgb_gray_convert_sse2 (JDIMENSION img_width, +; JSAMPARRAY input_buf, JSAMPIMAGE output_buf, +; JDIMENSION output_row, int num_rows); +; + +%define img_width(b) (b)+8 ; JDIMENSION img_width +%define input_buf(b) (b)+12 ; JSAMPARRAY input_buf +%define output_buf(b) (b)+16 ; JSAMPIMAGE output_buf +%define output_row(b) (b)+20 ; JDIMENSION output_row +%define num_rows(b) (b)+24 ; int num_rows + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 +%define gotptr wk(0)-SIZEOF_POINTER ; void * gotptr + + align 16 + + global EXTN(jsimd_rgb_gray_convert_sse2) + +EXTN(jsimd_rgb_gray_convert_sse2): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic eax ; make a room for GOT address + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + movpic POINTER [gotptr], ebx ; save GOT address + + mov ecx, JDIMENSION [img_width(eax)] + test ecx,ecx + jz near .return + + push ecx + + mov esi, JSAMPIMAGE [output_buf(eax)] + mov ecx, JDIMENSION [output_row(eax)] + mov edi, JSAMPARRAY [esi+0*SIZEOF_JSAMPARRAY] + lea edi, [edi+ecx*SIZEOF_JSAMPROW] + + pop ecx + + mov esi, JSAMPARRAY [input_buf(eax)] + mov eax, INT [num_rows(eax)] + test eax,eax + jle near .return + alignx 16,7 +.rowloop: + pushpic eax + push edi + push esi + push ecx ; col + + mov esi, JSAMPROW [esi] ; inptr + mov edi, JSAMPROW [edi] ; outptr0 + movpic eax, POINTER [gotptr] ; load GOT address (eax) + + cmp ecx, byte SIZEOF_XMMWORD + jae near .columnloop + alignx 16,7 + +%if RGB_PIXELSIZE == 3 ; --------------- + +.column_ld1: + push eax + push edx + lea ecx,[ecx+ecx*2] ; imul ecx,RGB_PIXELSIZE + test cl, SIZEOF_BYTE + jz short .column_ld2 + sub ecx, byte SIZEOF_BYTE + movzx eax, BYTE [esi+ecx] +.column_ld2: + test cl, SIZEOF_WORD + jz short .column_ld4 + sub ecx, byte SIZEOF_WORD + movzx edx, WORD [esi+ecx] + shl eax, WORD_BIT + or eax,edx +.column_ld4: + movd xmmA,eax + pop edx + pop eax + test cl, SIZEOF_DWORD + jz short .column_ld8 + sub ecx, byte SIZEOF_DWORD + movd xmmF, XMM_DWORD [esi+ecx] + pslldq xmmA, SIZEOF_DWORD + por xmmA,xmmF +.column_ld8: + test cl, SIZEOF_MMWORD + jz short .column_ld16 + sub ecx, byte SIZEOF_MMWORD + movq xmmB, XMM_MMWORD [esi+ecx] + pslldq xmmA, SIZEOF_MMWORD + por xmmA,xmmB +.column_ld16: + test cl, SIZEOF_XMMWORD + jz short .column_ld32 + movdqa xmmF,xmmA + movdqu xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD] + mov ecx, SIZEOF_XMMWORD + jmp short .rgb_gray_cnv +.column_ld32: + test cl, 2*SIZEOF_XMMWORD + mov ecx, SIZEOF_XMMWORD + jz short .rgb_gray_cnv + movdqa xmmB,xmmA + movdqu xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD] + movdqu xmmF, XMMWORD [esi+1*SIZEOF_XMMWORD] + jmp short .rgb_gray_cnv + alignx 16,7 + +.columnloop: + movdqu xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD] + movdqu xmmF, XMMWORD [esi+1*SIZEOF_XMMWORD] + movdqu xmmB, XMMWORD [esi+2*SIZEOF_XMMWORD] + +.rgb_gray_cnv: + ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05) + ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A) + ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F) + + movdqa xmmG,xmmA + pslldq xmmA,8 ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12) + psrldq xmmG,8 ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --) + + punpckhbw xmmA,xmmF ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A) + pslldq xmmF,8 ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27) + + punpcklbw xmmG,xmmB ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D) + punpckhbw xmmF,xmmB ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F) + + movdqa xmmD,xmmA + pslldq xmmA,8 ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09) + psrldq xmmD,8 ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --) + + punpckhbw xmmA,xmmG ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D) + pslldq xmmG,8 ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B) + + punpcklbw xmmD,xmmF ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E) + punpckhbw xmmG,xmmF ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F) + + movdqa xmmE,xmmA + pslldq xmmA,8 ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C) + psrldq xmmE,8 ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --) + + punpckhbw xmmA,xmmD ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E) + pslldq xmmD,8 ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D) + + punpcklbw xmmE,xmmG ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F) + punpckhbw xmmD,xmmG ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F) + + pxor xmmH,xmmH + + movdqa xmmC,xmmA + punpcklbw xmmA,xmmH ; xmmA=(00 02 04 06 08 0A 0C 0E) + punpckhbw xmmC,xmmH ; xmmC=(10 12 14 16 18 1A 1C 1E) + + movdqa xmmB,xmmE + punpcklbw xmmE,xmmH ; xmmE=(20 22 24 26 28 2A 2C 2E) + punpckhbw xmmB,xmmH ; xmmB=(01 03 05 07 09 0B 0D 0F) + + movdqa xmmF,xmmD + punpcklbw xmmD,xmmH ; xmmD=(11 13 15 17 19 1B 1D 1F) + punpckhbw xmmF,xmmH ; xmmF=(21 23 25 27 29 2B 2D 2F) + +%else ; RGB_PIXELSIZE == 4 ; ----------- + +.column_ld1: + test cl, SIZEOF_XMMWORD/16 + jz short .column_ld2 + sub ecx, byte SIZEOF_XMMWORD/16 + movd xmmA, XMM_DWORD [esi+ecx*RGB_PIXELSIZE] +.column_ld2: + test cl, SIZEOF_XMMWORD/8 + jz short .column_ld4 + sub ecx, byte SIZEOF_XMMWORD/8 + movq xmmE, XMM_MMWORD [esi+ecx*RGB_PIXELSIZE] + pslldq xmmA, SIZEOF_MMWORD + por xmmA,xmmE +.column_ld4: + test cl, SIZEOF_XMMWORD/4 + jz short .column_ld8 + sub ecx, byte SIZEOF_XMMWORD/4 + movdqa xmmE,xmmA + movdqu xmmA, XMMWORD [esi+ecx*RGB_PIXELSIZE] +.column_ld8: + test cl, SIZEOF_XMMWORD/2 + mov ecx, SIZEOF_XMMWORD + jz short .rgb_gray_cnv + movdqa xmmF,xmmA + movdqa xmmH,xmmE + movdqu xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD] + movdqu xmmE, XMMWORD [esi+1*SIZEOF_XMMWORD] + jmp short .rgb_gray_cnv + alignx 16,7 + +.columnloop: + movdqu xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD] + movdqu xmmE, XMMWORD [esi+1*SIZEOF_XMMWORD] + movdqu xmmF, XMMWORD [esi+2*SIZEOF_XMMWORD] + movdqu xmmH, XMMWORD [esi+3*SIZEOF_XMMWORD] + +.rgb_gray_cnv: + ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33) + ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37) + ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B) + ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F) + + movdqa xmmD,xmmA + punpcklbw xmmA,xmmE ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35) + punpckhbw xmmD,xmmE ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37) + + movdqa xmmC,xmmF + punpcklbw xmmF,xmmH ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D) + punpckhbw xmmC,xmmH ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F) + + movdqa xmmB,xmmA + punpcklwd xmmA,xmmF ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C) + punpckhwd xmmB,xmmF ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D) + + movdqa xmmG,xmmD + punpcklwd xmmD,xmmC ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E) + punpckhwd xmmG,xmmC ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F) + + movdqa xmmE,xmmA + punpcklbw xmmA,xmmD ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E) + punpckhbw xmmE,xmmD ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E) + + movdqa xmmH,xmmB + punpcklbw xmmB,xmmG ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F) + punpckhbw xmmH,xmmG ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F) + + pxor xmmF,xmmF + + movdqa xmmC,xmmA + punpcklbw xmmA,xmmF ; xmmA=(00 02 04 06 08 0A 0C 0E) + punpckhbw xmmC,xmmF ; xmmC=(10 12 14 16 18 1A 1C 1E) + + movdqa xmmD,xmmB + punpcklbw xmmB,xmmF ; xmmB=(01 03 05 07 09 0B 0D 0F) + punpckhbw xmmD,xmmF ; xmmD=(11 13 15 17 19 1B 1D 1F) + + movdqa xmmG,xmmE + punpcklbw xmmE,xmmF ; xmmE=(20 22 24 26 28 2A 2C 2E) + punpckhbw xmmG,xmmF ; xmmG=(30 32 34 36 38 3A 3C 3E) + + punpcklbw xmmF,xmmH + punpckhbw xmmH,xmmH + psrlw xmmF,BYTE_BIT ; xmmF=(21 23 25 27 29 2B 2D 2F) + psrlw xmmH,BYTE_BIT ; xmmH=(31 33 35 37 39 3B 3D 3F) + +%endif ; RGB_PIXELSIZE ; --------------- + + ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE + ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO + + ; (Original) + ; Y = 0.29900 * R + 0.58700 * G + 0.11400 * B + ; + ; (This implementation) + ; Y = 0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G + + movdqa xmm6,xmm1 + punpcklwd xmm1,xmm3 + punpckhwd xmm6,xmm3 + pmaddwd xmm1,[GOTOFF(eax,PW_F0299_F0337)] ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337) + pmaddwd xmm6,[GOTOFF(eax,PW_F0299_F0337)] ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337) + + movdqa xmm7, xmm6 ; xmm7=ROH*FIX(0.299)+GOH*FIX(0.337) + + movdqa xmm6,xmm0 + punpcklwd xmm0,xmm2 + punpckhwd xmm6,xmm2 + pmaddwd xmm0,[GOTOFF(eax,PW_F0299_F0337)] ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337) + pmaddwd xmm6,[GOTOFF(eax,PW_F0299_F0337)] ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337) + + movdqa XMMWORD [wk(0)], xmm0 ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337) + movdqa XMMWORD [wk(1)], xmm6 ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337) + + movdqa xmm0, xmm5 ; xmm0=BO + movdqa xmm6, xmm4 ; xmm6=BE + + movdqa xmm4,xmm0 + punpcklwd xmm0,xmm3 + punpckhwd xmm4,xmm3 + pmaddwd xmm0,[GOTOFF(eax,PW_F0114_F0250)] ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250) + pmaddwd xmm4,[GOTOFF(eax,PW_F0114_F0250)] ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250) + + movdqa xmm3,[GOTOFF(eax,PD_ONEHALF)] ; xmm3=[PD_ONEHALF] + + paddd xmm0, xmm1 + paddd xmm4, xmm7 + paddd xmm0,xmm3 + paddd xmm4,xmm3 + psrld xmm0,SCALEBITS ; xmm0=YOL + psrld xmm4,SCALEBITS ; xmm4=YOH + packssdw xmm0,xmm4 ; xmm0=YO + + movdqa xmm4,xmm6 + punpcklwd xmm6,xmm2 + punpckhwd xmm4,xmm2 + pmaddwd xmm6,[GOTOFF(eax,PW_F0114_F0250)] ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250) + pmaddwd xmm4,[GOTOFF(eax,PW_F0114_F0250)] ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250) + + movdqa xmm2,[GOTOFF(eax,PD_ONEHALF)] ; xmm2=[PD_ONEHALF] + + paddd xmm6, XMMWORD [wk(0)] + paddd xmm4, XMMWORD [wk(1)] + paddd xmm6,xmm2 + paddd xmm4,xmm2 + psrld xmm6,SCALEBITS ; xmm6=YEL + psrld xmm4,SCALEBITS ; xmm4=YEH + packssdw xmm6,xmm4 ; xmm6=YE + + psllw xmm0,BYTE_BIT + por xmm6,xmm0 ; xmm6=Y + movdqa XMMWORD [edi], xmm6 ; Save Y + + sub ecx, byte SIZEOF_XMMWORD + add esi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; inptr + add edi, byte SIZEOF_XMMWORD ; outptr0 + cmp ecx, byte SIZEOF_XMMWORD + jae near .columnloop + test ecx,ecx + jnz near .column_ld1 + + pop ecx ; col + pop esi + pop edi + poppic eax + + add esi, byte SIZEOF_JSAMPROW ; input_buf + add edi, byte SIZEOF_JSAMPROW + dec eax ; num_rows + jg near .rowloop + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jcolsamp.inc b/app/src/main/jni/libjpeg-turbo/simd/jcolsamp.inc new file mode 100644 index 000000000..79751b7c7 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jcolsamp.inc @@ -0,0 +1,105 @@ +; +; jcolsamp.inc - private declarations for color conversion & up/downsampling +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; [TAB8] + +; -------------------------------------------------------------------------- + +; pseudo-resisters to make ordering of RGB configurable +; +%if RGB_RED == 0 +%define mmA mm0 +%define mmB mm1 +%define xmmA xmm0 +%define xmmB xmm1 +%elif RGB_GREEN == 0 +%define mmA mm2 +%define mmB mm3 +%define xmmA xmm2 +%define xmmB xmm3 +%elif RGB_BLUE == 0 +%define mmA mm4 +%define mmB mm5 +%define xmmA xmm4 +%define xmmB xmm5 +%else +%define mmA mm6 +%define mmB mm7 +%define xmmA xmm6 +%define xmmB xmm7 +%endif + +%if RGB_RED == 1 +%define mmC mm0 +%define mmD mm1 +%define xmmC xmm0 +%define xmmD xmm1 +%elif RGB_GREEN == 1 +%define mmC mm2 +%define mmD mm3 +%define xmmC xmm2 +%define xmmD xmm3 +%elif RGB_BLUE == 1 +%define mmC mm4 +%define mmD mm5 +%define xmmC xmm4 +%define xmmD xmm5 +%else +%define mmC mm6 +%define mmD mm7 +%define xmmC xmm6 +%define xmmD xmm7 +%endif + +%if RGB_RED == 2 +%define mmE mm0 +%define mmF mm1 +%define xmmE xmm0 +%define xmmF xmm1 +%elif RGB_GREEN == 2 +%define mmE mm2 +%define mmF mm3 +%define xmmE xmm2 +%define xmmF xmm3 +%elif RGB_BLUE == 2 +%define mmE mm4 +%define mmF mm5 +%define xmmE xmm4 +%define xmmF xmm5 +%else +%define mmE mm6 +%define mmF mm7 +%define xmmE xmm6 +%define xmmF xmm7 +%endif + +%if RGB_RED == 3 +%define mmG mm0 +%define mmH mm1 +%define xmmG xmm0 +%define xmmH xmm1 +%elif RGB_GREEN == 3 +%define mmG mm2 +%define mmH mm3 +%define xmmG xmm2 +%define xmmH xmm3 +%elif RGB_BLUE == 3 +%define mmG mm4 +%define mmH mm5 +%define xmmG xmm4 +%define xmmH xmm5 +%else +%define mmG mm6 +%define mmH mm7 +%define xmmG xmm6 +%define xmmH xmm7 +%endif + +; -------------------------------------------------------------------------- diff --git a/app/src/main/jni/libjpeg-turbo/simd/jcsample-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jcsample-mmx.asm new file mode 100644 index 000000000..6881a56d7 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jcsample-mmx.asm @@ -0,0 +1,324 @@ +; +; jcsample.asm - downsampling (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Downsample pixel values of a single component. +; This version handles the common case of 2:1 horizontal and 1:1 vertical, +; without smoothing. +; +; GLOBAL(void) +; jsimd_h2v1_downsample_mmx (JDIMENSION image_width, int max_v_samp_factor, +; JDIMENSION v_samp_factor, JDIMENSION width_blocks, +; JSAMPARRAY input_data, JSAMPARRAY output_data); +; + +%define img_width(b) (b)+8 ; JDIMENSION image_width +%define max_v_samp(b) (b)+12 ; int max_v_samp_factor +%define v_samp(b) (b)+16 ; JDIMENSION v_samp_factor +%define width_blks(b) (b)+20 ; JDIMENSION width_blocks +%define input_data(b) (b)+24 ; JSAMPARRAY input_data +%define output_data(b) (b)+28 ; JSAMPARRAY output_data + + align 16 + global EXTN(jsimd_h2v1_downsample_mmx) + +EXTN(jsimd_h2v1_downsample_mmx): + push ebp + mov ebp,esp +; push ebx ; unused +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + mov ecx, JDIMENSION [width_blks(ebp)] + shl ecx,3 ; imul ecx,DCTSIZE (ecx = output_cols) + jz near .return + + mov edx, JDIMENSION [img_width(ebp)] + + ; -- expand_right_edge + + push ecx + shl ecx,1 ; output_cols * 2 + sub ecx,edx + jle short .expand_end + + mov eax, INT [max_v_samp(ebp)] + test eax,eax + jle short .expand_end + + cld + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + alignx 16,7 +.expandloop: + push eax + push ecx + + mov edi, JSAMPROW [esi] + add edi,edx + mov al, JSAMPLE [edi-1] + + rep stosb + + pop ecx + pop eax + + add esi, byte SIZEOF_JSAMPROW + dec eax + jg short .expandloop + +.expand_end: + pop ecx ; output_cols + + ; -- h2v1_downsample + + mov eax, JDIMENSION [v_samp(ebp)] ; rowctr + test eax,eax + jle near .return + + mov edx, 0x00010000 ; bias pattern + movd mm7,edx + pcmpeqw mm6,mm6 + punpckldq mm7,mm7 ; mm7={0, 1, 0, 1} + psrlw mm6,BYTE_BIT ; mm6={0xFF 0x00 0xFF 0x00 ..} + + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + mov edi, JSAMPARRAY [output_data(ebp)] ; output_data + alignx 16,7 +.rowloop: + push ecx + push edi + push esi + + mov esi, JSAMPROW [esi] ; inptr + mov edi, JSAMPROW [edi] ; outptr + alignx 16,7 +.columnloop: + + movq mm0, MMWORD [esi+0*SIZEOF_MMWORD] + movq mm1, MMWORD [esi+1*SIZEOF_MMWORD] + movq mm2,mm0 + movq mm3,mm1 + + pand mm0,mm6 + psrlw mm2,BYTE_BIT + pand mm1,mm6 + psrlw mm3,BYTE_BIT + + paddw mm0,mm2 + paddw mm1,mm3 + paddw mm0,mm7 + paddw mm1,mm7 + psrlw mm0,1 + psrlw mm1,1 + + packuswb mm0,mm1 + + movq MMWORD [edi+0*SIZEOF_MMWORD], mm0 + + add esi, byte 2*SIZEOF_MMWORD ; inptr + add edi, byte 1*SIZEOF_MMWORD ; outptr + sub ecx, byte SIZEOF_MMWORD ; outcol + jnz short .columnloop + + pop esi + pop edi + pop ecx + + add esi, byte SIZEOF_JSAMPROW ; input_data + add edi, byte SIZEOF_JSAMPROW ; output_data + dec eax ; rowctr + jg short .rowloop + + emms ; empty MMX state + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved +; pop ebx ; unused + pop ebp + ret + +; -------------------------------------------------------------------------- +; +; Downsample pixel values of a single component. +; This version handles the standard case of 2:1 horizontal and 2:1 vertical, +; without smoothing. +; +; GLOBAL(void) +; jsimd_h2v2_downsample_mmx (JDIMENSION image_width, int max_v_samp_factor, +; JDIMENSION v_samp_factor, JDIMENSION width_blocks, +; JSAMPARRAY input_data, JSAMPARRAY output_data); +; + +%define img_width(b) (b)+8 ; JDIMENSION image_width +%define max_v_samp(b) (b)+12 ; int max_v_samp_factor +%define v_samp(b) (b)+16 ; JDIMENSION v_samp_factor +%define width_blks(b) (b)+20 ; JDIMENSION width_blocks +%define input_data(b) (b)+24 ; JSAMPARRAY input_data +%define output_data(b) (b)+28 ; JSAMPARRAY output_data + + align 16 + global EXTN(jsimd_h2v2_downsample_mmx) + +EXTN(jsimd_h2v2_downsample_mmx): + push ebp + mov ebp,esp +; push ebx ; unused +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + mov ecx, JDIMENSION [width_blks(ebp)] + shl ecx,3 ; imul ecx,DCTSIZE (ecx = output_cols) + jz near .return + + mov edx, JDIMENSION [img_width(ebp)] + + ; -- expand_right_edge + + push ecx + shl ecx,1 ; output_cols * 2 + sub ecx,edx + jle short .expand_end + + mov eax, INT [max_v_samp(ebp)] + test eax,eax + jle short .expand_end + + cld + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + alignx 16,7 +.expandloop: + push eax + push ecx + + mov edi, JSAMPROW [esi] + add edi,edx + mov al, JSAMPLE [edi-1] + + rep stosb + + pop ecx + pop eax + + add esi, byte SIZEOF_JSAMPROW + dec eax + jg short .expandloop + +.expand_end: + pop ecx ; output_cols + + ; -- h2v2_downsample + + mov eax, JDIMENSION [v_samp(ebp)] ; rowctr + test eax,eax + jle near .return + + mov edx, 0x00020001 ; bias pattern + movd mm7,edx + pcmpeqw mm6,mm6 + punpckldq mm7,mm7 ; mm7={1, 2, 1, 2} + psrlw mm6,BYTE_BIT ; mm6={0xFF 0x00 0xFF 0x00 ..} + + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + mov edi, JSAMPARRAY [output_data(ebp)] ; output_data + alignx 16,7 +.rowloop: + push ecx + push edi + push esi + + mov edx, JSAMPROW [esi+0*SIZEOF_JSAMPROW] ; inptr0 + mov esi, JSAMPROW [esi+1*SIZEOF_JSAMPROW] ; inptr1 + mov edi, JSAMPROW [edi] ; outptr + alignx 16,7 +.columnloop: + + movq mm0, MMWORD [edx+0*SIZEOF_MMWORD] + movq mm1, MMWORD [esi+0*SIZEOF_MMWORD] + movq mm2, MMWORD [edx+1*SIZEOF_MMWORD] + movq mm3, MMWORD [esi+1*SIZEOF_MMWORD] + + movq mm4,mm0 + movq mm5,mm1 + pand mm0,mm6 + psrlw mm4,BYTE_BIT + pand mm1,mm6 + psrlw mm5,BYTE_BIT + paddw mm0,mm4 + paddw mm1,mm5 + + movq mm4,mm2 + movq mm5,mm3 + pand mm2,mm6 + psrlw mm4,BYTE_BIT + pand mm3,mm6 + psrlw mm5,BYTE_BIT + paddw mm2,mm4 + paddw mm3,mm5 + + paddw mm0,mm1 + paddw mm2,mm3 + paddw mm0,mm7 + paddw mm2,mm7 + psrlw mm0,2 + psrlw mm2,2 + + packuswb mm0,mm2 + + movq MMWORD [edi+0*SIZEOF_MMWORD], mm0 + + add edx, byte 2*SIZEOF_MMWORD ; inptr0 + add esi, byte 2*SIZEOF_MMWORD ; inptr1 + add edi, byte 1*SIZEOF_MMWORD ; outptr + sub ecx, byte SIZEOF_MMWORD ; outcol + jnz near .columnloop + + pop esi + pop edi + pop ecx + + add esi, byte 2*SIZEOF_JSAMPROW ; input_data + add edi, byte 1*SIZEOF_JSAMPROW ; output_data + dec eax ; rowctr + jg near .rowloop + + emms ; empty MMX state + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved +; pop ebx ; unused + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jcsample-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jcsample-sse2-64.asm new file mode 100644 index 000000000..f32fb4fcc --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jcsample-sse2-64.asm @@ -0,0 +1,330 @@ +; +; jcsample.asm - downsampling (64-bit SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 +; +; Downsample pixel values of a single component. +; This version handles the common case of 2:1 horizontal and 1:1 vertical, +; without smoothing. +; +; GLOBAL(void) +; jsimd_h2v1_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor, +; JDIMENSION v_samp_factor, JDIMENSION width_blocks, +; JSAMPARRAY input_data, JSAMPARRAY output_data); +; + +; r10 = JDIMENSION image_width +; r11 = int max_v_samp_factor +; r12 = JDIMENSION v_samp_factor +; r13 = JDIMENSION width_blocks +; r14 = JSAMPARRAY input_data +; r15 = JSAMPARRAY output_data + + align 16 + global EXTN(jsimd_h2v1_downsample_sse2) + +EXTN(jsimd_h2v1_downsample_sse2): + push rbp + mov rax,rsp + mov rbp,rsp + collect_args + + mov rcx, r13 + shl rcx,3 ; imul rcx,DCTSIZE (rcx = output_cols) + jz near .return + + mov rdx, r10 + + ; -- expand_right_edge + + push rcx + shl rcx,1 ; output_cols * 2 + sub rcx,rdx + jle short .expand_end + + mov rax, r11 + test rax,rax + jle short .expand_end + + cld + mov rsi, r14 ; input_data +.expandloop: + push rax + push rcx + + mov rdi, JSAMPROW [rsi] + add rdi,rdx + mov al, JSAMPLE [rdi-1] + + rep stosb + + pop rcx + pop rax + + add rsi, byte SIZEOF_JSAMPROW + dec rax + jg short .expandloop + +.expand_end: + pop rcx ; output_cols + + ; -- h2v1_downsample + + mov rax, r12 ; rowctr + test eax,eax + jle near .return + + mov rdx, 0x00010000 ; bias pattern + movd xmm7,edx + pcmpeqw xmm6,xmm6 + pshufd xmm7,xmm7,0x00 ; xmm7={0, 1, 0, 1, 0, 1, 0, 1} + psrlw xmm6,BYTE_BIT ; xmm6={0xFF 0x00 0xFF 0x00 ..} + + mov rsi, r14 ; input_data + mov rdi, r15 ; output_data +.rowloop: + push rcx + push rdi + push rsi + + mov rsi, JSAMPROW [rsi] ; inptr + mov rdi, JSAMPROW [rdi] ; outptr + + cmp rcx, byte SIZEOF_XMMWORD + jae short .columnloop + +.columnloop_r8: + movdqa xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD] + pxor xmm1,xmm1 + mov rcx, SIZEOF_XMMWORD + jmp short .downsample + +.columnloop: + movdqa xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD] + movdqa xmm1, XMMWORD [rsi+1*SIZEOF_XMMWORD] + +.downsample: + movdqa xmm2,xmm0 + movdqa xmm3,xmm1 + + pand xmm0,xmm6 + psrlw xmm2,BYTE_BIT + pand xmm1,xmm6 + psrlw xmm3,BYTE_BIT + + paddw xmm0,xmm2 + paddw xmm1,xmm3 + paddw xmm0,xmm7 + paddw xmm1,xmm7 + psrlw xmm0,1 + psrlw xmm1,1 + + packuswb xmm0,xmm1 + + movdqa XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0 + + sub rcx, byte SIZEOF_XMMWORD ; outcol + add rsi, byte 2*SIZEOF_XMMWORD ; inptr + add rdi, byte 1*SIZEOF_XMMWORD ; outptr + cmp rcx, byte SIZEOF_XMMWORD + jae short .columnloop + test rcx,rcx + jnz short .columnloop_r8 + + pop rsi + pop rdi + pop rcx + + add rsi, byte SIZEOF_JSAMPROW ; input_data + add rdi, byte SIZEOF_JSAMPROW ; output_data + dec rax ; rowctr + jg near .rowloop + +.return: + uncollect_args + pop rbp + ret + +; -------------------------------------------------------------------------- +; +; Downsample pixel values of a single component. +; This version handles the standard case of 2:1 horizontal and 2:1 vertical, +; without smoothing. +; +; GLOBAL(void) +; jsimd_h2v2_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor, +; JDIMENSION v_samp_factor, JDIMENSION width_blocks, +; JSAMPARRAY input_data, JSAMPARRAY output_data); +; + +; r10 = JDIMENSION image_width +; r11 = int max_v_samp_factor +; r12 = JDIMENSION v_samp_factor +; r13 = JDIMENSION width_blocks +; r14 = JSAMPARRAY input_data +; r15 = JSAMPARRAY output_data + + align 16 + global EXTN(jsimd_h2v2_downsample_sse2) + +EXTN(jsimd_h2v2_downsample_sse2): + push rbp + mov rax,rsp + mov rbp,rsp + collect_args + + mov rcx, r13 + shl rcx,3 ; imul rcx,DCTSIZE (rcx = output_cols) + jz near .return + + mov rdx, r10 + + ; -- expand_right_edge + + push rcx + shl rcx,1 ; output_cols * 2 + sub rcx,rdx + jle short .expand_end + + mov rax, r11 + test rax,rax + jle short .expand_end + + cld + mov rsi, r14 ; input_data +.expandloop: + push rax + push rcx + + mov rdi, JSAMPROW [rsi] + add rdi,rdx + mov al, JSAMPLE [rdi-1] + + rep stosb + + pop rcx + pop rax + + add rsi, byte SIZEOF_JSAMPROW + dec rax + jg short .expandloop + +.expand_end: + pop rcx ; output_cols + + ; -- h2v2_downsample + + mov rax, r12 ; rowctr + test rax,rax + jle near .return + + mov rdx, 0x00020001 ; bias pattern + movd xmm7,edx + pcmpeqw xmm6,xmm6 + pshufd xmm7,xmm7,0x00 ; xmm7={1, 2, 1, 2, 1, 2, 1, 2} + psrlw xmm6,BYTE_BIT ; xmm6={0xFF 0x00 0xFF 0x00 ..} + + mov rsi, r14 ; input_data + mov rdi, r15 ; output_data +.rowloop: + push rcx + push rdi + push rsi + + mov rdx, JSAMPROW [rsi+0*SIZEOF_JSAMPROW] ; inptr0 + mov rsi, JSAMPROW [rsi+1*SIZEOF_JSAMPROW] ; inptr1 + mov rdi, JSAMPROW [rdi] ; outptr + + cmp rcx, byte SIZEOF_XMMWORD + jae short .columnloop + +.columnloop_r8: + movdqa xmm0, XMMWORD [rdx+0*SIZEOF_XMMWORD] + movdqa xmm1, XMMWORD [rsi+0*SIZEOF_XMMWORD] + pxor xmm2,xmm2 + pxor xmm3,xmm3 + mov rcx, SIZEOF_XMMWORD + jmp short .downsample + +.columnloop: + movdqa xmm0, XMMWORD [rdx+0*SIZEOF_XMMWORD] + movdqa xmm1, XMMWORD [rsi+0*SIZEOF_XMMWORD] + movdqa xmm2, XMMWORD [rdx+1*SIZEOF_XMMWORD] + movdqa xmm3, XMMWORD [rsi+1*SIZEOF_XMMWORD] + +.downsample: + movdqa xmm4,xmm0 + movdqa xmm5,xmm1 + pand xmm0,xmm6 + psrlw xmm4,BYTE_BIT + pand xmm1,xmm6 + psrlw xmm5,BYTE_BIT + paddw xmm0,xmm4 + paddw xmm1,xmm5 + + movdqa xmm4,xmm2 + movdqa xmm5,xmm3 + pand xmm2,xmm6 + psrlw xmm4,BYTE_BIT + pand xmm3,xmm6 + psrlw xmm5,BYTE_BIT + paddw xmm2,xmm4 + paddw xmm3,xmm5 + + paddw xmm0,xmm1 + paddw xmm2,xmm3 + paddw xmm0,xmm7 + paddw xmm2,xmm7 + psrlw xmm0,2 + psrlw xmm2,2 + + packuswb xmm0,xmm2 + + movdqa XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0 + + sub rcx, byte SIZEOF_XMMWORD ; outcol + add rdx, byte 2*SIZEOF_XMMWORD ; inptr0 + add rsi, byte 2*SIZEOF_XMMWORD ; inptr1 + add rdi, byte 1*SIZEOF_XMMWORD ; outptr + cmp rcx, byte SIZEOF_XMMWORD + jae near .columnloop + test rcx,rcx + jnz near .columnloop_r8 + + pop rsi + pop rdi + pop rcx + + add rsi, byte 2*SIZEOF_JSAMPROW ; input_data + add rdi, byte 1*SIZEOF_JSAMPROW ; output_data + dec rax ; rowctr + jg near .rowloop + +.return: + uncollect_args + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jcsample-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jcsample-sse2.asm new file mode 100644 index 000000000..11202dbf6 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jcsample-sse2.asm @@ -0,0 +1,351 @@ +; +; jcsample.asm - downsampling (SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Downsample pixel values of a single component. +; This version handles the common case of 2:1 horizontal and 1:1 vertical, +; without smoothing. +; +; GLOBAL(void) +; jsimd_h2v1_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor, +; JDIMENSION v_samp_factor, JDIMENSION width_blocks, +; JSAMPARRAY input_data, JSAMPARRAY output_data); +; + +%define img_width(b) (b)+8 ; JDIMENSION image_width +%define max_v_samp(b) (b)+12 ; int max_v_samp_factor +%define v_samp(b) (b)+16 ; JDIMENSION v_samp_factor +%define width_blks(b) (b)+20 ; JDIMENSION width_blocks +%define input_data(b) (b)+24 ; JSAMPARRAY input_data +%define output_data(b) (b)+28 ; JSAMPARRAY output_data + + align 16 + global EXTN(jsimd_h2v1_downsample_sse2) + +EXTN(jsimd_h2v1_downsample_sse2): + push ebp + mov ebp,esp +; push ebx ; unused +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + mov ecx, JDIMENSION [width_blks(ebp)] + shl ecx,3 ; imul ecx,DCTSIZE (ecx = output_cols) + jz near .return + + mov edx, JDIMENSION [img_width(ebp)] + + ; -- expand_right_edge + + push ecx + shl ecx,1 ; output_cols * 2 + sub ecx,edx + jle short .expand_end + + mov eax, INT [max_v_samp(ebp)] + test eax,eax + jle short .expand_end + + cld + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + alignx 16,7 +.expandloop: + push eax + push ecx + + mov edi, JSAMPROW [esi] + add edi,edx + mov al, JSAMPLE [edi-1] + + rep stosb + + pop ecx + pop eax + + add esi, byte SIZEOF_JSAMPROW + dec eax + jg short .expandloop + +.expand_end: + pop ecx ; output_cols + + ; -- h2v1_downsample + + mov eax, JDIMENSION [v_samp(ebp)] ; rowctr + test eax,eax + jle near .return + + mov edx, 0x00010000 ; bias pattern + movd xmm7,edx + pcmpeqw xmm6,xmm6 + pshufd xmm7,xmm7,0x00 ; xmm7={0, 1, 0, 1, 0, 1, 0, 1} + psrlw xmm6,BYTE_BIT ; xmm6={0xFF 0x00 0xFF 0x00 ..} + + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + mov edi, JSAMPARRAY [output_data(ebp)] ; output_data + alignx 16,7 +.rowloop: + push ecx + push edi + push esi + + mov esi, JSAMPROW [esi] ; inptr + mov edi, JSAMPROW [edi] ; outptr + + cmp ecx, byte SIZEOF_XMMWORD + jae short .columnloop + alignx 16,7 + +.columnloop_r8: + movdqa xmm0, XMMWORD [esi+0*SIZEOF_XMMWORD] + pxor xmm1,xmm1 + mov ecx, SIZEOF_XMMWORD + jmp short .downsample + alignx 16,7 + +.columnloop: + movdqa xmm0, XMMWORD [esi+0*SIZEOF_XMMWORD] + movdqa xmm1, XMMWORD [esi+1*SIZEOF_XMMWORD] + +.downsample: + movdqa xmm2,xmm0 + movdqa xmm3,xmm1 + + pand xmm0,xmm6 + psrlw xmm2,BYTE_BIT + pand xmm1,xmm6 + psrlw xmm3,BYTE_BIT + + paddw xmm0,xmm2 + paddw xmm1,xmm3 + paddw xmm0,xmm7 + paddw xmm1,xmm7 + psrlw xmm0,1 + psrlw xmm1,1 + + packuswb xmm0,xmm1 + + movdqa XMMWORD [edi+0*SIZEOF_XMMWORD], xmm0 + + sub ecx, byte SIZEOF_XMMWORD ; outcol + add esi, byte 2*SIZEOF_XMMWORD ; inptr + add edi, byte 1*SIZEOF_XMMWORD ; outptr + cmp ecx, byte SIZEOF_XMMWORD + jae short .columnloop + test ecx,ecx + jnz short .columnloop_r8 + + pop esi + pop edi + pop ecx + + add esi, byte SIZEOF_JSAMPROW ; input_data + add edi, byte SIZEOF_JSAMPROW ; output_data + dec eax ; rowctr + jg near .rowloop + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved +; pop ebx ; unused + pop ebp + ret + +; -------------------------------------------------------------------------- +; +; Downsample pixel values of a single component. +; This version handles the standard case of 2:1 horizontal and 2:1 vertical, +; without smoothing. +; +; GLOBAL(void) +; jsimd_h2v2_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor, +; JDIMENSION v_samp_factor, JDIMENSION width_blocks, +; JSAMPARRAY input_data, JSAMPARRAY output_data); +; + +%define img_width(b) (b)+8 ; JDIMENSION image_width +%define max_v_samp(b) (b)+12 ; int max_v_samp_factor +%define v_samp(b) (b)+16 ; JDIMENSION v_samp_factor +%define width_blks(b) (b)+20 ; JDIMENSION width_blocks +%define input_data(b) (b)+24 ; JSAMPARRAY input_data +%define output_data(b) (b)+28 ; JSAMPARRAY output_data + + align 16 + global EXTN(jsimd_h2v2_downsample_sse2) + +EXTN(jsimd_h2v2_downsample_sse2): + push ebp + mov ebp,esp +; push ebx ; unused +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + mov ecx, JDIMENSION [width_blks(ebp)] + shl ecx,3 ; imul ecx,DCTSIZE (ecx = output_cols) + jz near .return + + mov edx, JDIMENSION [img_width(ebp)] + + ; -- expand_right_edge + + push ecx + shl ecx,1 ; output_cols * 2 + sub ecx,edx + jle short .expand_end + + mov eax, INT [max_v_samp(ebp)] + test eax,eax + jle short .expand_end + + cld + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + alignx 16,7 +.expandloop: + push eax + push ecx + + mov edi, JSAMPROW [esi] + add edi,edx + mov al, JSAMPLE [edi-1] + + rep stosb + + pop ecx + pop eax + + add esi, byte SIZEOF_JSAMPROW + dec eax + jg short .expandloop + +.expand_end: + pop ecx ; output_cols + + ; -- h2v2_downsample + + mov eax, JDIMENSION [v_samp(ebp)] ; rowctr + test eax,eax + jle near .return + + mov edx, 0x00020001 ; bias pattern + movd xmm7,edx + pcmpeqw xmm6,xmm6 + pshufd xmm7,xmm7,0x00 ; xmm7={1, 2, 1, 2, 1, 2, 1, 2} + psrlw xmm6,BYTE_BIT ; xmm6={0xFF 0x00 0xFF 0x00 ..} + + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + mov edi, JSAMPARRAY [output_data(ebp)] ; output_data + alignx 16,7 +.rowloop: + push ecx + push edi + push esi + + mov edx, JSAMPROW [esi+0*SIZEOF_JSAMPROW] ; inptr0 + mov esi, JSAMPROW [esi+1*SIZEOF_JSAMPROW] ; inptr1 + mov edi, JSAMPROW [edi] ; outptr + + cmp ecx, byte SIZEOF_XMMWORD + jae short .columnloop + alignx 16,7 + +.columnloop_r8: + movdqa xmm0, XMMWORD [edx+0*SIZEOF_XMMWORD] + movdqa xmm1, XMMWORD [esi+0*SIZEOF_XMMWORD] + pxor xmm2,xmm2 + pxor xmm3,xmm3 + mov ecx, SIZEOF_XMMWORD + jmp short .downsample + alignx 16,7 + +.columnloop: + movdqa xmm0, XMMWORD [edx+0*SIZEOF_XMMWORD] + movdqa xmm1, XMMWORD [esi+0*SIZEOF_XMMWORD] + movdqa xmm2, XMMWORD [edx+1*SIZEOF_XMMWORD] + movdqa xmm3, XMMWORD [esi+1*SIZEOF_XMMWORD] + +.downsample: + movdqa xmm4,xmm0 + movdqa xmm5,xmm1 + pand xmm0,xmm6 + psrlw xmm4,BYTE_BIT + pand xmm1,xmm6 + psrlw xmm5,BYTE_BIT + paddw xmm0,xmm4 + paddw xmm1,xmm5 + + movdqa xmm4,xmm2 + movdqa xmm5,xmm3 + pand xmm2,xmm6 + psrlw xmm4,BYTE_BIT + pand xmm3,xmm6 + psrlw xmm5,BYTE_BIT + paddw xmm2,xmm4 + paddw xmm3,xmm5 + + paddw xmm0,xmm1 + paddw xmm2,xmm3 + paddw xmm0,xmm7 + paddw xmm2,xmm7 + psrlw xmm0,2 + psrlw xmm2,2 + + packuswb xmm0,xmm2 + + movdqa XMMWORD [edi+0*SIZEOF_XMMWORD], xmm0 + + sub ecx, byte SIZEOF_XMMWORD ; outcol + add edx, byte 2*SIZEOF_XMMWORD ; inptr0 + add esi, byte 2*SIZEOF_XMMWORD ; inptr1 + add edi, byte 1*SIZEOF_XMMWORD ; outptr + cmp ecx, byte SIZEOF_XMMWORD + jae near .columnloop + test ecx,ecx + jnz near .columnloop_r8 + + pop esi + pop edi + pop ecx + + add esi, byte 2*SIZEOF_JSAMPROW ; input_data + add edi, byte 1*SIZEOF_JSAMPROW ; output_data + dec eax ; rowctr + jg near .rowloop + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved +; pop ebx ; unused + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdcolext-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jdcolext-mmx.asm new file mode 100644 index 000000000..de1f00f15 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdcolext-mmx.asm @@ -0,0 +1,405 @@ +; +; jdcolext.asm - colorspace conversion (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jcolsamp.inc" + +; -------------------------------------------------------------------------- +; +; Convert some rows of samples to the output colorspace. +; +; GLOBAL(void) +; jsimd_ycc_rgb_convert_mmx (JDIMENSION out_width, +; JSAMPIMAGE input_buf, JDIMENSION input_row, +; JSAMPARRAY output_buf, int num_rows) +; + +%define out_width(b) (b)+8 ; JDIMENSION out_width +%define input_buf(b) (b)+12 ; JSAMPIMAGE input_buf +%define input_row(b) (b)+16 ; JDIMENSION input_row +%define output_buf(b) (b)+20 ; JSAMPARRAY output_buf +%define num_rows(b) (b)+24 ; int num_rows + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_MMWORD ; mmword wk[WK_NUM] +%define WK_NUM 2 +%define gotptr wk(0)-SIZEOF_POINTER ; void * gotptr + + align 16 + global EXTN(jsimd_ycc_rgb_convert_mmx) + +EXTN(jsimd_ycc_rgb_convert_mmx): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_MMWORD) ; align to 64 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic eax ; make a room for GOT address + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + movpic POINTER [gotptr], ebx ; save GOT address + + mov ecx, JDIMENSION [out_width(eax)] ; num_cols + test ecx,ecx + jz near .return + + push ecx + + mov edi, JSAMPIMAGE [input_buf(eax)] + mov ecx, JDIMENSION [input_row(eax)] + mov esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY] + mov ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY] + mov edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY] + lea esi, [esi+ecx*SIZEOF_JSAMPROW] + lea ebx, [ebx+ecx*SIZEOF_JSAMPROW] + lea edx, [edx+ecx*SIZEOF_JSAMPROW] + + pop ecx + + mov edi, JSAMPARRAY [output_buf(eax)] + mov eax, INT [num_rows(eax)] + test eax,eax + jle near .return + alignx 16,7 +.rowloop: + push eax + push edi + push edx + push ebx + push esi + push ecx ; col + + mov esi, JSAMPROW [esi] ; inptr0 + mov ebx, JSAMPROW [ebx] ; inptr1 + mov edx, JSAMPROW [edx] ; inptr2 + mov edi, JSAMPROW [edi] ; outptr + movpic eax, POINTER [gotptr] ; load GOT address (eax) + alignx 16,7 +.columnloop: + + movq mm5, MMWORD [ebx] ; mm5=Cb(01234567) + movq mm1, MMWORD [edx] ; mm1=Cr(01234567) + + pcmpeqw mm4,mm4 + pcmpeqw mm7,mm7 + psrlw mm4,BYTE_BIT + psllw mm7,7 ; mm7={0xFF80 0xFF80 0xFF80 0xFF80} + movq mm0,mm4 ; mm0=mm4={0xFF 0x00 0xFF 0x00 ..} + + pand mm4,mm5 ; mm4=Cb(0246)=CbE + psrlw mm5,BYTE_BIT ; mm5=Cb(1357)=CbO + pand mm0,mm1 ; mm0=Cr(0246)=CrE + psrlw mm1,BYTE_BIT ; mm1=Cr(1357)=CrO + + paddw mm4,mm7 + paddw mm5,mm7 + paddw mm0,mm7 + paddw mm1,mm7 + + ; (Original) + ; R = Y + 1.40200 * Cr + ; G = Y - 0.34414 * Cb - 0.71414 * Cr + ; B = Y + 1.77200 * Cb + ; + ; (This implementation) + ; R = Y + 0.40200 * Cr + Cr + ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr + ; B = Y - 0.22800 * Cb + Cb + Cb + + movq mm2,mm4 ; mm2=CbE + movq mm3,mm5 ; mm3=CbO + paddw mm4,mm4 ; mm4=2*CbE + paddw mm5,mm5 ; mm5=2*CbO + movq mm6,mm0 ; mm6=CrE + movq mm7,mm1 ; mm7=CrO + paddw mm0,mm0 ; mm0=2*CrE + paddw mm1,mm1 ; mm1=2*CrO + + pmulhw mm4,[GOTOFF(eax,PW_MF0228)] ; mm4=(2*CbE * -FIX(0.22800)) + pmulhw mm5,[GOTOFF(eax,PW_MF0228)] ; mm5=(2*CbO * -FIX(0.22800)) + pmulhw mm0,[GOTOFF(eax,PW_F0402)] ; mm0=(2*CrE * FIX(0.40200)) + pmulhw mm1,[GOTOFF(eax,PW_F0402)] ; mm1=(2*CrO * FIX(0.40200)) + + paddw mm4,[GOTOFF(eax,PW_ONE)] + paddw mm5,[GOTOFF(eax,PW_ONE)] + psraw mm4,1 ; mm4=(CbE * -FIX(0.22800)) + psraw mm5,1 ; mm5=(CbO * -FIX(0.22800)) + paddw mm0,[GOTOFF(eax,PW_ONE)] + paddw mm1,[GOTOFF(eax,PW_ONE)] + psraw mm0,1 ; mm0=(CrE * FIX(0.40200)) + psraw mm1,1 ; mm1=(CrO * FIX(0.40200)) + + paddw mm4,mm2 + paddw mm5,mm3 + paddw mm4,mm2 ; mm4=(CbE * FIX(1.77200))=(B-Y)E + paddw mm5,mm3 ; mm5=(CbO * FIX(1.77200))=(B-Y)O + paddw mm0,mm6 ; mm0=(CrE * FIX(1.40200))=(R-Y)E + paddw mm1,mm7 ; mm1=(CrO * FIX(1.40200))=(R-Y)O + + movq MMWORD [wk(0)], mm4 ; wk(0)=(B-Y)E + movq MMWORD [wk(1)], mm5 ; wk(1)=(B-Y)O + + movq mm4,mm2 + movq mm5,mm3 + punpcklwd mm2,mm6 + punpckhwd mm4,mm6 + pmaddwd mm2,[GOTOFF(eax,PW_MF0344_F0285)] + pmaddwd mm4,[GOTOFF(eax,PW_MF0344_F0285)] + punpcklwd mm3,mm7 + punpckhwd mm5,mm7 + pmaddwd mm3,[GOTOFF(eax,PW_MF0344_F0285)] + pmaddwd mm5,[GOTOFF(eax,PW_MF0344_F0285)] + + paddd mm2,[GOTOFF(eax,PD_ONEHALF)] + paddd mm4,[GOTOFF(eax,PD_ONEHALF)] + psrad mm2,SCALEBITS + psrad mm4,SCALEBITS + paddd mm3,[GOTOFF(eax,PD_ONEHALF)] + paddd mm5,[GOTOFF(eax,PD_ONEHALF)] + psrad mm3,SCALEBITS + psrad mm5,SCALEBITS + + packssdw mm2,mm4 ; mm2=CbE*-FIX(0.344)+CrE*FIX(0.285) + packssdw mm3,mm5 ; mm3=CbO*-FIX(0.344)+CrO*FIX(0.285) + psubw mm2,mm6 ; mm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E + psubw mm3,mm7 ; mm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O + + movq mm5, MMWORD [esi] ; mm5=Y(01234567) + + pcmpeqw mm4,mm4 + psrlw mm4,BYTE_BIT ; mm4={0xFF 0x00 0xFF 0x00 ..} + pand mm4,mm5 ; mm4=Y(0246)=YE + psrlw mm5,BYTE_BIT ; mm5=Y(1357)=YO + + paddw mm0,mm4 ; mm0=((R-Y)E+YE)=RE=(R0 R2 R4 R6) + paddw mm1,mm5 ; mm1=((R-Y)O+YO)=RO=(R1 R3 R5 R7) + packuswb mm0,mm0 ; mm0=(R0 R2 R4 R6 ** ** ** **) + packuswb mm1,mm1 ; mm1=(R1 R3 R5 R7 ** ** ** **) + + paddw mm2,mm4 ; mm2=((G-Y)E+YE)=GE=(G0 G2 G4 G6) + paddw mm3,mm5 ; mm3=((G-Y)O+YO)=GO=(G1 G3 G5 G7) + packuswb mm2,mm2 ; mm2=(G0 G2 G4 G6 ** ** ** **) + packuswb mm3,mm3 ; mm3=(G1 G3 G5 G7 ** ** ** **) + + paddw mm4, MMWORD [wk(0)] ; mm4=(YE+(B-Y)E)=BE=(B0 B2 B4 B6) + paddw mm5, MMWORD [wk(1)] ; mm5=(YO+(B-Y)O)=BO=(B1 B3 B5 B7) + packuswb mm4,mm4 ; mm4=(B0 B2 B4 B6 ** ** ** **) + packuswb mm5,mm5 ; mm5=(B1 B3 B5 B7 ** ** ** **) + +%if RGB_PIXELSIZE == 3 ; --------------- + + ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **) + ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **) + ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **) + ; mmG=(** ** ** ** ** ** ** **), mmH=(** ** ** ** ** ** ** **) + + punpcklbw mmA,mmC ; mmA=(00 10 02 12 04 14 06 16) + punpcklbw mmE,mmB ; mmE=(20 01 22 03 24 05 26 07) + punpcklbw mmD,mmF ; mmD=(11 21 13 23 15 25 17 27) + + movq mmG,mmA + movq mmH,mmA + punpcklwd mmA,mmE ; mmA=(00 10 20 01 02 12 22 03) + punpckhwd mmG,mmE ; mmG=(04 14 24 05 06 16 26 07) + + psrlq mmH,2*BYTE_BIT ; mmH=(02 12 04 14 06 16 -- --) + psrlq mmE,2*BYTE_BIT ; mmE=(22 03 24 05 26 07 -- --) + + movq mmC,mmD + movq mmB,mmD + punpcklwd mmD,mmH ; mmD=(11 21 02 12 13 23 04 14) + punpckhwd mmC,mmH ; mmC=(15 25 06 16 17 27 -- --) + + psrlq mmB,2*BYTE_BIT ; mmB=(13 23 15 25 17 27 -- --) + + movq mmF,mmE + punpcklwd mmE,mmB ; mmE=(22 03 13 23 24 05 15 25) + punpckhwd mmF,mmB ; mmF=(26 07 17 27 -- -- -- --) + + punpckldq mmA,mmD ; mmA=(00 10 20 01 11 21 02 12) + punpckldq mmE,mmG ; mmE=(22 03 13 23 04 14 24 05) + punpckldq mmC,mmF ; mmC=(15 25 06 16 26 07 17 27) + + cmp ecx, byte SIZEOF_MMWORD + jb short .column_st16 + + movq MMWORD [edi+0*SIZEOF_MMWORD], mmA + movq MMWORD [edi+1*SIZEOF_MMWORD], mmE + movq MMWORD [edi+2*SIZEOF_MMWORD], mmC + + sub ecx, byte SIZEOF_MMWORD + jz short .nextrow + + add esi, byte SIZEOF_MMWORD ; inptr0 + add ebx, byte SIZEOF_MMWORD ; inptr1 + add edx, byte SIZEOF_MMWORD ; inptr2 + add edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD ; outptr + jmp near .columnloop + alignx 16,7 + +.column_st16: + lea ecx, [ecx+ecx*2] ; imul ecx, RGB_PIXELSIZE + cmp ecx, byte 2*SIZEOF_MMWORD + jb short .column_st8 + movq MMWORD [edi+0*SIZEOF_MMWORD], mmA + movq MMWORD [edi+1*SIZEOF_MMWORD], mmE + movq mmA,mmC + sub ecx, byte 2*SIZEOF_MMWORD + add edi, byte 2*SIZEOF_MMWORD + jmp short .column_st4 +.column_st8: + cmp ecx, byte SIZEOF_MMWORD + jb short .column_st4 + movq MMWORD [edi+0*SIZEOF_MMWORD], mmA + movq mmA,mmE + sub ecx, byte SIZEOF_MMWORD + add edi, byte SIZEOF_MMWORD +.column_st4: + movd eax,mmA + cmp ecx, byte SIZEOF_DWORD + jb short .column_st2 + mov DWORD [edi+0*SIZEOF_DWORD], eax + psrlq mmA,DWORD_BIT + movd eax,mmA + sub ecx, byte SIZEOF_DWORD + add edi, byte SIZEOF_DWORD +.column_st2: + cmp ecx, byte SIZEOF_WORD + jb short .column_st1 + mov WORD [edi+0*SIZEOF_WORD], ax + shr eax,WORD_BIT + sub ecx, byte SIZEOF_WORD + add edi, byte SIZEOF_WORD +.column_st1: + cmp ecx, byte SIZEOF_BYTE + jb short .nextrow + mov BYTE [edi+0*SIZEOF_BYTE], al + +%else ; RGB_PIXELSIZE == 4 ; ----------- + +%ifdef RGBX_FILLER_0XFF + pcmpeqb mm6,mm6 ; mm6=(X0 X2 X4 X6 ** ** ** **) + pcmpeqb mm7,mm7 ; mm7=(X1 X3 X5 X7 ** ** ** **) +%else + pxor mm6,mm6 ; mm6=(X0 X2 X4 X6 ** ** ** **) + pxor mm7,mm7 ; mm7=(X1 X3 X5 X7 ** ** ** **) +%endif + ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **) + ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **) + ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **) + ; mmG=(30 32 34 36 ** ** ** **), mmH=(31 33 35 37 ** ** ** **) + + punpcklbw mmA,mmC ; mmA=(00 10 02 12 04 14 06 16) + punpcklbw mmE,mmG ; mmE=(20 30 22 32 24 34 26 36) + punpcklbw mmB,mmD ; mmB=(01 11 03 13 05 15 07 17) + punpcklbw mmF,mmH ; mmF=(21 31 23 33 25 35 27 37) + + movq mmC,mmA + punpcklwd mmA,mmE ; mmA=(00 10 20 30 02 12 22 32) + punpckhwd mmC,mmE ; mmC=(04 14 24 34 06 16 26 36) + movq mmG,mmB + punpcklwd mmB,mmF ; mmB=(01 11 21 31 03 13 23 33) + punpckhwd mmG,mmF ; mmG=(05 15 25 35 07 17 27 37) + + movq mmD,mmA + punpckldq mmA,mmB ; mmA=(00 10 20 30 01 11 21 31) + punpckhdq mmD,mmB ; mmD=(02 12 22 32 03 13 23 33) + movq mmH,mmC + punpckldq mmC,mmG ; mmC=(04 14 24 34 05 15 25 35) + punpckhdq mmH,mmG ; mmH=(06 16 26 36 07 17 27 37) + + cmp ecx, byte SIZEOF_MMWORD + jb short .column_st16 + + movq MMWORD [edi+0*SIZEOF_MMWORD], mmA + movq MMWORD [edi+1*SIZEOF_MMWORD], mmD + movq MMWORD [edi+2*SIZEOF_MMWORD], mmC + movq MMWORD [edi+3*SIZEOF_MMWORD], mmH + + sub ecx, byte SIZEOF_MMWORD + jz short .nextrow + + add esi, byte SIZEOF_MMWORD ; inptr0 + add ebx, byte SIZEOF_MMWORD ; inptr1 + add edx, byte SIZEOF_MMWORD ; inptr2 + add edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD ; outptr + jmp near .columnloop + alignx 16,7 + +.column_st16: + cmp ecx, byte SIZEOF_MMWORD/2 + jb short .column_st8 + movq MMWORD [edi+0*SIZEOF_MMWORD], mmA + movq MMWORD [edi+1*SIZEOF_MMWORD], mmD + movq mmA,mmC + movq mmD,mmH + sub ecx, byte SIZEOF_MMWORD/2 + add edi, byte 2*SIZEOF_MMWORD +.column_st8: + cmp ecx, byte SIZEOF_MMWORD/4 + jb short .column_st4 + movq MMWORD [edi+0*SIZEOF_MMWORD], mmA + movq mmA,mmD + sub ecx, byte SIZEOF_MMWORD/4 + add edi, byte 1*SIZEOF_MMWORD +.column_st4: + cmp ecx, byte SIZEOF_MMWORD/8 + jb short .nextrow + movd DWORD [edi+0*SIZEOF_DWORD], mmA + +%endif ; RGB_PIXELSIZE ; --------------- + + alignx 16,7 + +.nextrow: + pop ecx + pop esi + pop ebx + pop edx + pop edi + pop eax + + add esi, byte SIZEOF_JSAMPROW + add ebx, byte SIZEOF_JSAMPROW + add edx, byte SIZEOF_JSAMPROW + add edi, byte SIZEOF_JSAMPROW ; output_buf + dec eax ; num_rows + jg near .rowloop + + emms ; empty MMX state + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdcolext-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jdcolext-sse2-64.asm new file mode 100644 index 000000000..bfd1f35df --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdcolext-sse2-64.asm @@ -0,0 +1,441 @@ +; +; jdcolext.asm - colorspace conversion (64-bit SSE2) +; +; Copyright 2009, 2012 Pierre Ossman for Cendio AB +; Copyright 2009, 2012 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jcolsamp.inc" + +; -------------------------------------------------------------------------- +; +; Convert some rows of samples to the output colorspace. +; +; GLOBAL(void) +; jsimd_ycc_rgb_convert_sse2 (JDIMENSION out_width, +; JSAMPIMAGE input_buf, JDIMENSION input_row, +; JSAMPARRAY output_buf, int num_rows) +; + +; r10 = JDIMENSION out_width +; r11 = JSAMPIMAGE input_buf +; r12 = JDIMENSION input_row +; r13 = JSAMPARRAY output_buf +; r14 = int num_rows + +%define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 + + align 16 + global EXTN(jsimd_ycc_rgb_convert_sse2) + +EXTN(jsimd_ycc_rgb_convert_sse2): + push rbp + mov rax,rsp ; rax = original rbp + sub rsp, byte 4 + and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [rsp],rax + mov rbp,rsp ; rbp = aligned rbp + lea rsp, [wk(0)] + collect_args + push rbx + + mov rcx, r10 ; num_cols + test rcx,rcx + jz near .return + + push rcx + + mov rdi, r11 + mov rcx, r12 + mov rsi, JSAMPARRAY [rdi+0*SIZEOF_JSAMPARRAY] + mov rbx, JSAMPARRAY [rdi+1*SIZEOF_JSAMPARRAY] + mov rdx, JSAMPARRAY [rdi+2*SIZEOF_JSAMPARRAY] + lea rsi, [rsi+rcx*SIZEOF_JSAMPROW] + lea rbx, [rbx+rcx*SIZEOF_JSAMPROW] + lea rdx, [rdx+rcx*SIZEOF_JSAMPROW] + + pop rcx + + mov rdi, r13 + mov eax, r14d + test rax,rax + jle near .return +.rowloop: + push rax + push rdi + push rdx + push rbx + push rsi + push rcx ; col + + mov rsi, JSAMPROW [rsi] ; inptr0 + mov rbx, JSAMPROW [rbx] ; inptr1 + mov rdx, JSAMPROW [rdx] ; inptr2 + mov rdi, JSAMPROW [rdi] ; outptr +.columnloop: + + movdqa xmm5, XMMWORD [rbx] ; xmm5=Cb(0123456789ABCDEF) + movdqa xmm1, XMMWORD [rdx] ; xmm1=Cr(0123456789ABCDEF) + + pcmpeqw xmm4,xmm4 + pcmpeqw xmm7,xmm7 + psrlw xmm4,BYTE_BIT + psllw xmm7,7 ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..} + movdqa xmm0,xmm4 ; xmm0=xmm4={0xFF 0x00 0xFF 0x00 ..} + + pand xmm4,xmm5 ; xmm4=Cb(02468ACE)=CbE + psrlw xmm5,BYTE_BIT ; xmm5=Cb(13579BDF)=CbO + pand xmm0,xmm1 ; xmm0=Cr(02468ACE)=CrE + psrlw xmm1,BYTE_BIT ; xmm1=Cr(13579BDF)=CrO + + paddw xmm4,xmm7 + paddw xmm5,xmm7 + paddw xmm0,xmm7 + paddw xmm1,xmm7 + + ; (Original) + ; R = Y + 1.40200 * Cr + ; G = Y - 0.34414 * Cb - 0.71414 * Cr + ; B = Y + 1.77200 * Cb + ; + ; (This implementation) + ; R = Y + 0.40200 * Cr + Cr + ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr + ; B = Y - 0.22800 * Cb + Cb + Cb + + movdqa xmm2,xmm4 ; xmm2=CbE + movdqa xmm3,xmm5 ; xmm3=CbO + paddw xmm4,xmm4 ; xmm4=2*CbE + paddw xmm5,xmm5 ; xmm5=2*CbO + movdqa xmm6,xmm0 ; xmm6=CrE + movdqa xmm7,xmm1 ; xmm7=CrO + paddw xmm0,xmm0 ; xmm0=2*CrE + paddw xmm1,xmm1 ; xmm1=2*CrO + + pmulhw xmm4,[rel PW_MF0228] ; xmm4=(2*CbE * -FIX(0.22800)) + pmulhw xmm5,[rel PW_MF0228] ; xmm5=(2*CbO * -FIX(0.22800)) + pmulhw xmm0,[rel PW_F0402] ; xmm0=(2*CrE * FIX(0.40200)) + pmulhw xmm1,[rel PW_F0402] ; xmm1=(2*CrO * FIX(0.40200)) + + paddw xmm4,[rel PW_ONE] + paddw xmm5,[rel PW_ONE] + psraw xmm4,1 ; xmm4=(CbE * -FIX(0.22800)) + psraw xmm5,1 ; xmm5=(CbO * -FIX(0.22800)) + paddw xmm0,[rel PW_ONE] + paddw xmm1,[rel PW_ONE] + psraw xmm0,1 ; xmm0=(CrE * FIX(0.40200)) + psraw xmm1,1 ; xmm1=(CrO * FIX(0.40200)) + + paddw xmm4,xmm2 + paddw xmm5,xmm3 + paddw xmm4,xmm2 ; xmm4=(CbE * FIX(1.77200))=(B-Y)E + paddw xmm5,xmm3 ; xmm5=(CbO * FIX(1.77200))=(B-Y)O + paddw xmm0,xmm6 ; xmm0=(CrE * FIX(1.40200))=(R-Y)E + paddw xmm1,xmm7 ; xmm1=(CrO * FIX(1.40200))=(R-Y)O + + movdqa XMMWORD [wk(0)], xmm4 ; wk(0)=(B-Y)E + movdqa XMMWORD [wk(1)], xmm5 ; wk(1)=(B-Y)O + + movdqa xmm4,xmm2 + movdqa xmm5,xmm3 + punpcklwd xmm2,xmm6 + punpckhwd xmm4,xmm6 + pmaddwd xmm2,[rel PW_MF0344_F0285] + pmaddwd xmm4,[rel PW_MF0344_F0285] + punpcklwd xmm3,xmm7 + punpckhwd xmm5,xmm7 + pmaddwd xmm3,[rel PW_MF0344_F0285] + pmaddwd xmm5,[rel PW_MF0344_F0285] + + paddd xmm2,[rel PD_ONEHALF] + paddd xmm4,[rel PD_ONEHALF] + psrad xmm2,SCALEBITS + psrad xmm4,SCALEBITS + paddd xmm3,[rel PD_ONEHALF] + paddd xmm5,[rel PD_ONEHALF] + psrad xmm3,SCALEBITS + psrad xmm5,SCALEBITS + + packssdw xmm2,xmm4 ; xmm2=CbE*-FIX(0.344)+CrE*FIX(0.285) + packssdw xmm3,xmm5 ; xmm3=CbO*-FIX(0.344)+CrO*FIX(0.285) + psubw xmm2,xmm6 ; xmm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E + psubw xmm3,xmm7 ; xmm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O + + movdqa xmm5, XMMWORD [rsi] ; xmm5=Y(0123456789ABCDEF) + + pcmpeqw xmm4,xmm4 + psrlw xmm4,BYTE_BIT ; xmm4={0xFF 0x00 0xFF 0x00 ..} + pand xmm4,xmm5 ; xmm4=Y(02468ACE)=YE + psrlw xmm5,BYTE_BIT ; xmm5=Y(13579BDF)=YO + + paddw xmm0,xmm4 ; xmm0=((R-Y)E+YE)=RE=R(02468ACE) + paddw xmm1,xmm5 ; xmm1=((R-Y)O+YO)=RO=R(13579BDF) + packuswb xmm0,xmm0 ; xmm0=R(02468ACE********) + packuswb xmm1,xmm1 ; xmm1=R(13579BDF********) + + paddw xmm2,xmm4 ; xmm2=((G-Y)E+YE)=GE=G(02468ACE) + paddw xmm3,xmm5 ; xmm3=((G-Y)O+YO)=GO=G(13579BDF) + packuswb xmm2,xmm2 ; xmm2=G(02468ACE********) + packuswb xmm3,xmm3 ; xmm3=G(13579BDF********) + + paddw xmm4, XMMWORD [wk(0)] ; xmm4=(YE+(B-Y)E)=BE=B(02468ACE) + paddw xmm5, XMMWORD [wk(1)] ; xmm5=(YO+(B-Y)O)=BO=B(13579BDF) + packuswb xmm4,xmm4 ; xmm4=B(02468ACE********) + packuswb xmm5,xmm5 ; xmm5=B(13579BDF********) + +%if RGB_PIXELSIZE == 3 ; --------------- + + ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **) + ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **) + ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **) + ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **) + + punpcklbw xmmA,xmmC ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E) + punpcklbw xmmE,xmmB ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F) + punpcklbw xmmD,xmmF ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F) + + movdqa xmmG,xmmA + movdqa xmmH,xmmA + punpcklwd xmmA,xmmE ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07) + punpckhwd xmmG,xmmE ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F) + + psrldq xmmH,2 ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --) + psrldq xmmE,2 ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --) + + movdqa xmmC,xmmD + movdqa xmmB,xmmD + punpcklwd xmmD,xmmH ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18) + punpckhwd xmmC,xmmH ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --) + + psrldq xmmB,2 ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --) + + movdqa xmmF,xmmE + punpcklwd xmmE,xmmB ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29) + punpckhwd xmmF,xmmB ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --) + + pshufd xmmH,xmmA,0x4E; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03) + movdqa xmmB,xmmE + punpckldq xmmA,xmmD ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14) + punpckldq xmmE,xmmH ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07) + punpckhdq xmmD,xmmB ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29) + + pshufd xmmH,xmmG,0x4E; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B) + movdqa xmmB,xmmF + punpckldq xmmG,xmmC ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C) + punpckldq xmmF,xmmH ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F) + punpckhdq xmmC,xmmB ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --) + + punpcklqdq xmmA,xmmE ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05) + punpcklqdq xmmD,xmmG ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A) + punpcklqdq xmmF,xmmC ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F) + + cmp rcx, byte SIZEOF_XMMWORD + jb short .column_st32 + + test rdi, SIZEOF_XMMWORD-1 + jnz short .out1 + ; --(aligned)------------------- + movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD + movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmF + jmp short .out0 +.out1: ; --(unaligned)----------------- + movdqu XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD + movdqu XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmF +.out0: + add rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; outptr + sub rcx, byte SIZEOF_XMMWORD + jz near .nextrow + + add rsi, byte SIZEOF_XMMWORD ; inptr0 + add rbx, byte SIZEOF_XMMWORD ; inptr1 + add rdx, byte SIZEOF_XMMWORD ; inptr2 + jmp near .columnloop + +.column_st32: + lea rcx, [rcx+rcx*2] ; imul ecx, RGB_PIXELSIZE + cmp rcx, byte 2*SIZEOF_XMMWORD + jb short .column_st16 + movdqu XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD + add rdi, byte 2*SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmF + sub rcx, byte 2*SIZEOF_XMMWORD + jmp short .column_st15 +.column_st16: + cmp rcx, byte SIZEOF_XMMWORD + jb short .column_st15 + movdqu XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + add rdi, byte SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmD + sub rcx, byte SIZEOF_XMMWORD +.column_st15: + ; Store the lower 8 bytes of xmmA to the output when it has enough + ; space. + cmp rcx, byte SIZEOF_MMWORD + jb short .column_st7 + movq XMM_MMWORD [rdi], xmmA + add rdi, byte SIZEOF_MMWORD + sub rcx, byte SIZEOF_MMWORD + psrldq xmmA, SIZEOF_MMWORD +.column_st7: + ; Store the lower 4 bytes of xmmA to the output when it has enough + ; space. + cmp rcx, byte SIZEOF_DWORD + jb short .column_st3 + movd XMM_DWORD [rdi], xmmA + add rdi, byte SIZEOF_DWORD + sub rcx, byte SIZEOF_DWORD + psrldq xmmA, SIZEOF_DWORD +.column_st3: + ; Store the lower 2 bytes of rax to the output when it has enough + ; space. + movd eax, xmmA + cmp rcx, byte SIZEOF_WORD + jb short .column_st1 + mov WORD [rdi], ax + add rdi, byte SIZEOF_WORD + sub rcx, byte SIZEOF_WORD + shr rax, 16 +.column_st1: + ; Store the lower 1 byte of rax to the output when it has enough + ; space. + test rcx, rcx + jz short .nextrow + mov BYTE [rdi], al + +%else ; RGB_PIXELSIZE == 4 ; ----------- + +%ifdef RGBX_FILLER_0XFF + pcmpeqb xmm6,xmm6 ; xmm6=XE=X(02468ACE********) + pcmpeqb xmm7,xmm7 ; xmm7=XO=X(13579BDF********) +%else + pxor xmm6,xmm6 ; xmm6=XE=X(02468ACE********) + pxor xmm7,xmm7 ; xmm7=XO=X(13579BDF********) +%endif + ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **) + ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **) + ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **) + ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **) + + punpcklbw xmmA,xmmC ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E) + punpcklbw xmmE,xmmG ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E) + punpcklbw xmmB,xmmD ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F) + punpcklbw xmmF,xmmH ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F) + + movdqa xmmC,xmmA + punpcklwd xmmA,xmmE ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36) + punpckhwd xmmC,xmmE ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E) + movdqa xmmG,xmmB + punpcklwd xmmB,xmmF ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37) + punpckhwd xmmG,xmmF ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F) + + movdqa xmmD,xmmA + punpckldq xmmA,xmmB ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33) + punpckhdq xmmD,xmmB ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37) + movdqa xmmH,xmmC + punpckldq xmmC,xmmG ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B) + punpckhdq xmmH,xmmG ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F) + + cmp rcx, byte SIZEOF_XMMWORD + jb short .column_st32 + + test rdi, SIZEOF_XMMWORD-1 + jnz short .out1 + ; --(aligned)------------------- + movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD + movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmC + movntdq XMMWORD [rdi+3*SIZEOF_XMMWORD], xmmH + jmp short .out0 +.out1: ; --(unaligned)----------------- + movdqu XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD + movdqu XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmC + movdqu XMMWORD [rdi+3*SIZEOF_XMMWORD], xmmH +.out0: + add rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; outptr + sub rcx, byte SIZEOF_XMMWORD + jz near .nextrow + + add rsi, byte SIZEOF_XMMWORD ; inptr0 + add rbx, byte SIZEOF_XMMWORD ; inptr1 + add rdx, byte SIZEOF_XMMWORD ; inptr2 + jmp near .columnloop + +.column_st32: + cmp rcx, byte SIZEOF_XMMWORD/2 + jb short .column_st16 + movdqu XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD + add rdi, byte 2*SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmC + movdqa xmmD,xmmH + sub rcx, byte SIZEOF_XMMWORD/2 +.column_st16: + cmp rcx, byte SIZEOF_XMMWORD/4 + jb short .column_st15 + movdqu XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + add rdi, byte SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmD + sub rcx, byte SIZEOF_XMMWORD/4 +.column_st15: + ; Store two pixels (8 bytes) of xmmA to the output when it has enough + ; space. + cmp rcx, byte SIZEOF_XMMWORD/8 + jb short .column_st7 + movq MMWORD [rdi], xmmA + add rdi, byte SIZEOF_XMMWORD/8*4 + sub rcx, byte SIZEOF_XMMWORD/8 + psrldq xmmA, SIZEOF_XMMWORD/8*4 +.column_st7: + ; Store one pixel (4 bytes) of xmmA to the output when it has enough + ; space. + test rcx, rcx + jz short .nextrow + movd XMM_DWORD [rdi], xmmA + +%endif ; RGB_PIXELSIZE ; --------------- + +.nextrow: + pop rcx + pop rsi + pop rbx + pop rdx + pop rdi + pop rax + + add rsi, byte SIZEOF_JSAMPROW + add rbx, byte SIZEOF_JSAMPROW + add rdx, byte SIZEOF_JSAMPROW + add rdi, byte SIZEOF_JSAMPROW ; output_buf + dec rax ; num_rows + jg near .rowloop + + sfence ; flush the write buffer + +.return: + pop rbx + uncollect_args + mov rsp,rbp ; rsp <- aligned rbp + pop rsp ; rsp <- original rbp + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdcolext-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jdcolext-sse2.asm new file mode 100644 index 000000000..54ae4db53 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdcolext-sse2.asm @@ -0,0 +1,460 @@ +; +; jdcolext.asm - colorspace conversion (SSE2) +; +; Copyright 2009, 2012 Pierre Ossman for Cendio AB +; Copyright 2012 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jcolsamp.inc" + +; -------------------------------------------------------------------------- +; +; Convert some rows of samples to the output colorspace. +; +; GLOBAL(void) +; jsimd_ycc_rgb_convert_sse2 (JDIMENSION out_width, +; JSAMPIMAGE input_buf, JDIMENSION input_row, +; JSAMPARRAY output_buf, int num_rows) +; + +%define out_width(b) (b)+8 ; JDIMENSION out_width +%define input_buf(b) (b)+12 ; JSAMPIMAGE input_buf +%define input_row(b) (b)+16 ; JDIMENSION input_row +%define output_buf(b) (b)+20 ; JSAMPARRAY output_buf +%define num_rows(b) (b)+24 ; int num_rows + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 +%define gotptr wk(0)-SIZEOF_POINTER ; void * gotptr + + align 16 + global EXTN(jsimd_ycc_rgb_convert_sse2) + +EXTN(jsimd_ycc_rgb_convert_sse2): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic eax ; make a room for GOT address + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + movpic POINTER [gotptr], ebx ; save GOT address + + mov ecx, JDIMENSION [out_width(eax)] ; num_cols + test ecx,ecx + jz near .return + + push ecx + + mov edi, JSAMPIMAGE [input_buf(eax)] + mov ecx, JDIMENSION [input_row(eax)] + mov esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY] + mov ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY] + mov edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY] + lea esi, [esi+ecx*SIZEOF_JSAMPROW] + lea ebx, [ebx+ecx*SIZEOF_JSAMPROW] + lea edx, [edx+ecx*SIZEOF_JSAMPROW] + + pop ecx + + mov edi, JSAMPARRAY [output_buf(eax)] + mov eax, INT [num_rows(eax)] + test eax,eax + jle near .return + alignx 16,7 +.rowloop: + push eax + push edi + push edx + push ebx + push esi + push ecx ; col + + mov esi, JSAMPROW [esi] ; inptr0 + mov ebx, JSAMPROW [ebx] ; inptr1 + mov edx, JSAMPROW [edx] ; inptr2 + mov edi, JSAMPROW [edi] ; outptr + movpic eax, POINTER [gotptr] ; load GOT address (eax) + alignx 16,7 +.columnloop: + + movdqa xmm5, XMMWORD [ebx] ; xmm5=Cb(0123456789ABCDEF) + movdqa xmm1, XMMWORD [edx] ; xmm1=Cr(0123456789ABCDEF) + + pcmpeqw xmm4,xmm4 + pcmpeqw xmm7,xmm7 + psrlw xmm4,BYTE_BIT + psllw xmm7,7 ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..} + movdqa xmm0,xmm4 ; xmm0=xmm4={0xFF 0x00 0xFF 0x00 ..} + + pand xmm4,xmm5 ; xmm4=Cb(02468ACE)=CbE + psrlw xmm5,BYTE_BIT ; xmm5=Cb(13579BDF)=CbO + pand xmm0,xmm1 ; xmm0=Cr(02468ACE)=CrE + psrlw xmm1,BYTE_BIT ; xmm1=Cr(13579BDF)=CrO + + paddw xmm4,xmm7 + paddw xmm5,xmm7 + paddw xmm0,xmm7 + paddw xmm1,xmm7 + + ; (Original) + ; R = Y + 1.40200 * Cr + ; G = Y - 0.34414 * Cb - 0.71414 * Cr + ; B = Y + 1.77200 * Cb + ; + ; (This implementation) + ; R = Y + 0.40200 * Cr + Cr + ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr + ; B = Y - 0.22800 * Cb + Cb + Cb + + movdqa xmm2,xmm4 ; xmm2=CbE + movdqa xmm3,xmm5 ; xmm3=CbO + paddw xmm4,xmm4 ; xmm4=2*CbE + paddw xmm5,xmm5 ; xmm5=2*CbO + movdqa xmm6,xmm0 ; xmm6=CrE + movdqa xmm7,xmm1 ; xmm7=CrO + paddw xmm0,xmm0 ; xmm0=2*CrE + paddw xmm1,xmm1 ; xmm1=2*CrO + + pmulhw xmm4,[GOTOFF(eax,PW_MF0228)] ; xmm4=(2*CbE * -FIX(0.22800)) + pmulhw xmm5,[GOTOFF(eax,PW_MF0228)] ; xmm5=(2*CbO * -FIX(0.22800)) + pmulhw xmm0,[GOTOFF(eax,PW_F0402)] ; xmm0=(2*CrE * FIX(0.40200)) + pmulhw xmm1,[GOTOFF(eax,PW_F0402)] ; xmm1=(2*CrO * FIX(0.40200)) + + paddw xmm4,[GOTOFF(eax,PW_ONE)] + paddw xmm5,[GOTOFF(eax,PW_ONE)] + psraw xmm4,1 ; xmm4=(CbE * -FIX(0.22800)) + psraw xmm5,1 ; xmm5=(CbO * -FIX(0.22800)) + paddw xmm0,[GOTOFF(eax,PW_ONE)] + paddw xmm1,[GOTOFF(eax,PW_ONE)] + psraw xmm0,1 ; xmm0=(CrE * FIX(0.40200)) + psraw xmm1,1 ; xmm1=(CrO * FIX(0.40200)) + + paddw xmm4,xmm2 + paddw xmm5,xmm3 + paddw xmm4,xmm2 ; xmm4=(CbE * FIX(1.77200))=(B-Y)E + paddw xmm5,xmm3 ; xmm5=(CbO * FIX(1.77200))=(B-Y)O + paddw xmm0,xmm6 ; xmm0=(CrE * FIX(1.40200))=(R-Y)E + paddw xmm1,xmm7 ; xmm1=(CrO * FIX(1.40200))=(R-Y)O + + movdqa XMMWORD [wk(0)], xmm4 ; wk(0)=(B-Y)E + movdqa XMMWORD [wk(1)], xmm5 ; wk(1)=(B-Y)O + + movdqa xmm4,xmm2 + movdqa xmm5,xmm3 + punpcklwd xmm2,xmm6 + punpckhwd xmm4,xmm6 + pmaddwd xmm2,[GOTOFF(eax,PW_MF0344_F0285)] + pmaddwd xmm4,[GOTOFF(eax,PW_MF0344_F0285)] + punpcklwd xmm3,xmm7 + punpckhwd xmm5,xmm7 + pmaddwd xmm3,[GOTOFF(eax,PW_MF0344_F0285)] + pmaddwd xmm5,[GOTOFF(eax,PW_MF0344_F0285)] + + paddd xmm2,[GOTOFF(eax,PD_ONEHALF)] + paddd xmm4,[GOTOFF(eax,PD_ONEHALF)] + psrad xmm2,SCALEBITS + psrad xmm4,SCALEBITS + paddd xmm3,[GOTOFF(eax,PD_ONEHALF)] + paddd xmm5,[GOTOFF(eax,PD_ONEHALF)] + psrad xmm3,SCALEBITS + psrad xmm5,SCALEBITS + + packssdw xmm2,xmm4 ; xmm2=CbE*-FIX(0.344)+CrE*FIX(0.285) + packssdw xmm3,xmm5 ; xmm3=CbO*-FIX(0.344)+CrO*FIX(0.285) + psubw xmm2,xmm6 ; xmm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E + psubw xmm3,xmm7 ; xmm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O + + movdqa xmm5, XMMWORD [esi] ; xmm5=Y(0123456789ABCDEF) + + pcmpeqw xmm4,xmm4 + psrlw xmm4,BYTE_BIT ; xmm4={0xFF 0x00 0xFF 0x00 ..} + pand xmm4,xmm5 ; xmm4=Y(02468ACE)=YE + psrlw xmm5,BYTE_BIT ; xmm5=Y(13579BDF)=YO + + paddw xmm0,xmm4 ; xmm0=((R-Y)E+YE)=RE=R(02468ACE) + paddw xmm1,xmm5 ; xmm1=((R-Y)O+YO)=RO=R(13579BDF) + packuswb xmm0,xmm0 ; xmm0=R(02468ACE********) + packuswb xmm1,xmm1 ; xmm1=R(13579BDF********) + + paddw xmm2,xmm4 ; xmm2=((G-Y)E+YE)=GE=G(02468ACE) + paddw xmm3,xmm5 ; xmm3=((G-Y)O+YO)=GO=G(13579BDF) + packuswb xmm2,xmm2 ; xmm2=G(02468ACE********) + packuswb xmm3,xmm3 ; xmm3=G(13579BDF********) + + paddw xmm4, XMMWORD [wk(0)] ; xmm4=(YE+(B-Y)E)=BE=B(02468ACE) + paddw xmm5, XMMWORD [wk(1)] ; xmm5=(YO+(B-Y)O)=BO=B(13579BDF) + packuswb xmm4,xmm4 ; xmm4=B(02468ACE********) + packuswb xmm5,xmm5 ; xmm5=B(13579BDF********) + +%if RGB_PIXELSIZE == 3 ; --------------- + + ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **) + ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **) + ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **) + ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **) + + punpcklbw xmmA,xmmC ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E) + punpcklbw xmmE,xmmB ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F) + punpcklbw xmmD,xmmF ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F) + + movdqa xmmG,xmmA + movdqa xmmH,xmmA + punpcklwd xmmA,xmmE ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07) + punpckhwd xmmG,xmmE ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F) + + psrldq xmmH,2 ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --) + psrldq xmmE,2 ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --) + + movdqa xmmC,xmmD + movdqa xmmB,xmmD + punpcklwd xmmD,xmmH ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18) + punpckhwd xmmC,xmmH ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --) + + psrldq xmmB,2 ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --) + + movdqa xmmF,xmmE + punpcklwd xmmE,xmmB ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29) + punpckhwd xmmF,xmmB ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --) + + pshufd xmmH,xmmA,0x4E; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03) + movdqa xmmB,xmmE + punpckldq xmmA,xmmD ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14) + punpckldq xmmE,xmmH ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07) + punpckhdq xmmD,xmmB ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29) + + pshufd xmmH,xmmG,0x4E; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B) + movdqa xmmB,xmmF + punpckldq xmmG,xmmC ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C) + punpckldq xmmF,xmmH ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F) + punpckhdq xmmC,xmmB ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --) + + punpcklqdq xmmA,xmmE ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05) + punpcklqdq xmmD,xmmG ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A) + punpcklqdq xmmF,xmmC ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F) + + cmp ecx, byte SIZEOF_XMMWORD + jb short .column_st32 + + test edi, SIZEOF_XMMWORD-1 + jnz short .out1 + ; --(aligned)------------------- + movntdq XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + movntdq XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD + movntdq XMMWORD [edi+2*SIZEOF_XMMWORD], xmmF + jmp short .out0 +.out1: ; --(unaligned)----------------- + movdqu XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD + movdqu XMMWORD [edi+2*SIZEOF_XMMWORD], xmmF +.out0: + add edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; outptr + sub ecx, byte SIZEOF_XMMWORD + jz near .nextrow + + add esi, byte SIZEOF_XMMWORD ; inptr0 + add ebx, byte SIZEOF_XMMWORD ; inptr1 + add edx, byte SIZEOF_XMMWORD ; inptr2 + jmp near .columnloop + alignx 16,7 + +.column_st32: + lea ecx, [ecx+ecx*2] ; imul ecx, RGB_PIXELSIZE + cmp ecx, byte 2*SIZEOF_XMMWORD + jb short .column_st16 + movdqu XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD + add edi, byte 2*SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmF + sub ecx, byte 2*SIZEOF_XMMWORD + jmp short .column_st15 +.column_st16: + cmp ecx, byte SIZEOF_XMMWORD + jb short .column_st15 + movdqu XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + add edi, byte SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmD + sub ecx, byte SIZEOF_XMMWORD +.column_st15: + ; Store the lower 8 bytes of xmmA to the output when it has enough + ; space. + cmp ecx, byte SIZEOF_MMWORD + jb short .column_st7 + movq XMM_MMWORD [edi], xmmA + add edi, byte SIZEOF_MMWORD + sub ecx, byte SIZEOF_MMWORD + psrldq xmmA, SIZEOF_MMWORD +.column_st7: + ; Store the lower 4 bytes of xmmA to the output when it has enough + ; space. + cmp ecx, byte SIZEOF_DWORD + jb short .column_st3 + movd XMM_DWORD [edi], xmmA + add edi, byte SIZEOF_DWORD + sub ecx, byte SIZEOF_DWORD + psrldq xmmA, SIZEOF_DWORD +.column_st3: + ; Store the lower 2 bytes of eax to the output when it has enough + ; space. + movd eax, xmmA + cmp ecx, byte SIZEOF_WORD + jb short .column_st1 + mov WORD [edi], ax + add edi, byte SIZEOF_WORD + sub ecx, byte SIZEOF_WORD + shr eax, 16 +.column_st1: + ; Store the lower 1 byte of eax to the output when it has enough + ; space. + test ecx, ecx + jz short .nextrow + mov BYTE [edi], al + +%else ; RGB_PIXELSIZE == 4 ; ----------- + +%ifdef RGBX_FILLER_0XFF + pcmpeqb xmm6,xmm6 ; xmm6=XE=X(02468ACE********) + pcmpeqb xmm7,xmm7 ; xmm7=XO=X(13579BDF********) +%else + pxor xmm6,xmm6 ; xmm6=XE=X(02468ACE********) + pxor xmm7,xmm7 ; xmm7=XO=X(13579BDF********) +%endif + ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **) + ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **) + ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **) + ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **) + + punpcklbw xmmA,xmmC ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E) + punpcklbw xmmE,xmmG ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E) + punpcklbw xmmB,xmmD ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F) + punpcklbw xmmF,xmmH ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F) + + movdqa xmmC,xmmA + punpcklwd xmmA,xmmE ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36) + punpckhwd xmmC,xmmE ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E) + movdqa xmmG,xmmB + punpcklwd xmmB,xmmF ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37) + punpckhwd xmmG,xmmF ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F) + + movdqa xmmD,xmmA + punpckldq xmmA,xmmB ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33) + punpckhdq xmmD,xmmB ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37) + movdqa xmmH,xmmC + punpckldq xmmC,xmmG ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B) + punpckhdq xmmH,xmmG ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F) + + cmp ecx, byte SIZEOF_XMMWORD + jb short .column_st32 + + test edi, SIZEOF_XMMWORD-1 + jnz short .out1 + ; --(aligned)------------------- + movntdq XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + movntdq XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD + movntdq XMMWORD [edi+2*SIZEOF_XMMWORD], xmmC + movntdq XMMWORD [edi+3*SIZEOF_XMMWORD], xmmH + jmp short .out0 +.out1: ; --(unaligned)----------------- + movdqu XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD + movdqu XMMWORD [edi+2*SIZEOF_XMMWORD], xmmC + movdqu XMMWORD [edi+3*SIZEOF_XMMWORD], xmmH +.out0: + add edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; outptr + sub ecx, byte SIZEOF_XMMWORD + jz near .nextrow + + add esi, byte SIZEOF_XMMWORD ; inptr0 + add ebx, byte SIZEOF_XMMWORD ; inptr1 + add edx, byte SIZEOF_XMMWORD ; inptr2 + jmp near .columnloop + alignx 16,7 + +.column_st32: + cmp ecx, byte SIZEOF_XMMWORD/2 + jb short .column_st16 + movdqu XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD + add edi, byte 2*SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmC + movdqa xmmD,xmmH + sub ecx, byte SIZEOF_XMMWORD/2 +.column_st16: + cmp ecx, byte SIZEOF_XMMWORD/4 + jb short .column_st15 + movdqu XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + add edi, byte SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmD + sub ecx, byte SIZEOF_XMMWORD/4 +.column_st15: + ; Store two pixels (8 bytes) of xmmA to the output when it has enough + ; space. + cmp ecx, byte SIZEOF_XMMWORD/8 + jb short .column_st7 + movq XMM_MMWORD [edi], xmmA + add edi, byte SIZEOF_XMMWORD/8*4 + sub ecx, byte SIZEOF_XMMWORD/8 + psrldq xmmA, SIZEOF_XMMWORD/8*4 +.column_st7: + ; Store one pixel (4 bytes) of xmmA to the output when it has enough + ; space. + test ecx, ecx + jz short .nextrow + movd XMM_DWORD [edi], xmmA + +%endif ; RGB_PIXELSIZE ; --------------- + + alignx 16,7 + +.nextrow: + pop ecx + pop esi + pop ebx + pop edx + pop edi + pop eax + + add esi, byte SIZEOF_JSAMPROW + add ebx, byte SIZEOF_JSAMPROW + add edx, byte SIZEOF_JSAMPROW + add edi, byte SIZEOF_JSAMPROW ; output_buf + dec eax ; num_rows + jg near .rowloop + + sfence ; flush the write buffer + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdcolor-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jdcolor-mmx.asm new file mode 100644 index 000000000..6730e4859 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdcolor-mmx.asm @@ -0,0 +1,120 @@ +; +; jdcolor.asm - colorspace conversion (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + +%define SCALEBITS 16 + +F_0_344 equ 22554 ; FIX(0.34414) +F_0_714 equ 46802 ; FIX(0.71414) +F_1_402 equ 91881 ; FIX(1.40200) +F_1_772 equ 116130 ; FIX(1.77200) +F_0_402 equ (F_1_402 - 65536) ; FIX(1.40200) - FIX(1) +F_0_285 equ ( 65536 - F_0_714) ; FIX(1) - FIX(0.71414) +F_0_228 equ (131072 - F_1_772) ; FIX(2) - FIX(1.77200) + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_ycc_rgb_convert_mmx) + +EXTN(jconst_ycc_rgb_convert_mmx): + +PW_F0402 times 4 dw F_0_402 +PW_MF0228 times 4 dw -F_0_228 +PW_MF0344_F0285 times 2 dw -F_0_344, F_0_285 +PW_ONE times 4 dw 1 +PD_ONEHALF times 2 dd 1 << (SCALEBITS-1) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 + +%include "jdcolext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGB_RED +%define RGB_GREEN EXT_RGB_GREEN +%define RGB_BLUE EXT_RGB_BLUE +%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE +%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extrgb_convert_mmx +%include "jdcolext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGBX_RED +%define RGB_GREEN EXT_RGBX_GREEN +%define RGB_BLUE EXT_RGBX_BLUE +%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE +%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extrgbx_convert_mmx +%include "jdcolext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGR_RED +%define RGB_GREEN EXT_BGR_GREEN +%define RGB_BLUE EXT_BGR_BLUE +%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE +%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extbgr_convert_mmx +%include "jdcolext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGRX_RED +%define RGB_GREEN EXT_BGRX_GREEN +%define RGB_BLUE EXT_BGRX_BLUE +%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE +%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extbgrx_convert_mmx +%include "jdcolext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XBGR_RED +%define RGB_GREEN EXT_XBGR_GREEN +%define RGB_BLUE EXT_XBGR_BLUE +%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE +%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extxbgr_convert_mmx +%include "jdcolext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XRGB_RED +%define RGB_GREEN EXT_XRGB_GREEN +%define RGB_BLUE EXT_XRGB_BLUE +%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE +%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extxrgb_convert_mmx +%include "jdcolext-mmx.asm" diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdcolor-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jdcolor-sse2-64.asm new file mode 100644 index 000000000..e9277f1d2 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdcolor-sse2-64.asm @@ -0,0 +1,120 @@ +; +; jdcolor.asm - colorspace conversion (64-bit SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + +%define SCALEBITS 16 + +F_0_344 equ 22554 ; FIX(0.34414) +F_0_714 equ 46802 ; FIX(0.71414) +F_1_402 equ 91881 ; FIX(1.40200) +F_1_772 equ 116130 ; FIX(1.77200) +F_0_402 equ (F_1_402 - 65536) ; FIX(1.40200) - FIX(1) +F_0_285 equ ( 65536 - F_0_714) ; FIX(1) - FIX(0.71414) +F_0_228 equ (131072 - F_1_772) ; FIX(2) - FIX(1.77200) + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_ycc_rgb_convert_sse2) + +EXTN(jconst_ycc_rgb_convert_sse2): + +PW_F0402 times 8 dw F_0_402 +PW_MF0228 times 8 dw -F_0_228 +PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285 +PW_ONE times 8 dw 1 +PD_ONEHALF times 4 dd 1 << (SCALEBITS-1) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 + +%include "jdcolext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGB_RED +%define RGB_GREEN EXT_RGB_GREEN +%define RGB_BLUE EXT_RGB_BLUE +%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE +%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgb_convert_sse2 +%include "jdcolext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGBX_RED +%define RGB_GREEN EXT_RGBX_GREEN +%define RGB_BLUE EXT_RGBX_BLUE +%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE +%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgbx_convert_sse2 +%include "jdcolext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGR_RED +%define RGB_GREEN EXT_BGR_GREEN +%define RGB_BLUE EXT_BGR_BLUE +%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE +%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgr_convert_sse2 +%include "jdcolext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGRX_RED +%define RGB_GREEN EXT_BGRX_GREEN +%define RGB_BLUE EXT_BGRX_BLUE +%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE +%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgrx_convert_sse2 +%include "jdcolext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XBGR_RED +%define RGB_GREEN EXT_XBGR_GREEN +%define RGB_BLUE EXT_XBGR_BLUE +%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE +%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxbgr_convert_sse2 +%include "jdcolext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XRGB_RED +%define RGB_GREEN EXT_XRGB_GREEN +%define RGB_BLUE EXT_XRGB_BLUE +%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE +%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxrgb_convert_sse2 +%include "jdcolext-sse2-64.asm" diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdcolor-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jdcolor-sse2.asm new file mode 100644 index 000000000..c122cc753 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdcolor-sse2.asm @@ -0,0 +1,120 @@ +; +; jdcolor.asm - colorspace conversion (SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + +%define SCALEBITS 16 + +F_0_344 equ 22554 ; FIX(0.34414) +F_0_714 equ 46802 ; FIX(0.71414) +F_1_402 equ 91881 ; FIX(1.40200) +F_1_772 equ 116130 ; FIX(1.77200) +F_0_402 equ (F_1_402 - 65536) ; FIX(1.40200) - FIX(1) +F_0_285 equ ( 65536 - F_0_714) ; FIX(1) - FIX(0.71414) +F_0_228 equ (131072 - F_1_772) ; FIX(2) - FIX(1.77200) + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_ycc_rgb_convert_sse2) + +EXTN(jconst_ycc_rgb_convert_sse2): + +PW_F0402 times 8 dw F_0_402 +PW_MF0228 times 8 dw -F_0_228 +PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285 +PW_ONE times 8 dw 1 +PD_ONEHALF times 4 dd 1 << (SCALEBITS-1) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 + +%include "jdcolext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGB_RED +%define RGB_GREEN EXT_RGB_GREEN +%define RGB_BLUE EXT_RGB_BLUE +%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE +%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgb_convert_sse2 +%include "jdcolext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGBX_RED +%define RGB_GREEN EXT_RGBX_GREEN +%define RGB_BLUE EXT_RGBX_BLUE +%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE +%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgbx_convert_sse2 +%include "jdcolext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGR_RED +%define RGB_GREEN EXT_BGR_GREEN +%define RGB_BLUE EXT_BGR_BLUE +%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE +%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgr_convert_sse2 +%include "jdcolext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGRX_RED +%define RGB_GREEN EXT_BGRX_GREEN +%define RGB_BLUE EXT_BGRX_BLUE +%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE +%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgrx_convert_sse2 +%include "jdcolext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XBGR_RED +%define RGB_GREEN EXT_XBGR_GREEN +%define RGB_BLUE EXT_XBGR_BLUE +%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE +%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxbgr_convert_sse2 +%include "jdcolext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XRGB_RED +%define RGB_GREEN EXT_XRGB_GREEN +%define RGB_BLUE EXT_XRGB_BLUE +%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE +%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxrgb_convert_sse2 +%include "jdcolext-sse2.asm" diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdct.inc b/app/src/main/jni/libjpeg-turbo/simd/jdct.inc new file mode 100644 index 000000000..ad5890c6c --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdct.inc @@ -0,0 +1,28 @@ +; +; jdct.inc - private declarations for forward & reverse DCT subsystems +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; [TAB8] + +; Each IDCT routine is responsible for range-limiting its results and +; converting them to unsigned form (0..MAXJSAMPLE). The raw outputs could +; be quite far out of range if the input data is corrupt, so a bulletproof +; range-limiting step is required. We use a mask-and-table-lookup method +; to do the combined operations quickly. +; +%define RANGE_MASK (MAXJSAMPLE * 4 + 3) ; 2 bits wider than legal samples + +%define ROW(n,b,s) ((b)+(n)*(s)) +%define COL(n,b,s) ((b)+(n)*(s)*DCTSIZE) + +%define DWBLOCK(m,n,b,s) ((b)+(m)*DCTSIZE*(s)+(n)*SIZEOF_DWORD) +%define MMBLOCK(m,n,b,s) ((b)+(m)*DCTSIZE*(s)+(n)*SIZEOF_MMWORD) +%define XMMBLOCK(m,n,b,s) ((b)+(m)*DCTSIZE*(s)+(n)*SIZEOF_XMMWORD) + +; -------------------------------------------------------------------------- diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdmerge-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jdmerge-mmx.asm new file mode 100644 index 000000000..2daa7fafb --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdmerge-mmx.asm @@ -0,0 +1,126 @@ +; +; jdmerge.asm - merged upsampling/color conversion (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + +%define SCALEBITS 16 + +F_0_344 equ 22554 ; FIX(0.34414) +F_0_714 equ 46802 ; FIX(0.71414) +F_1_402 equ 91881 ; FIX(1.40200) +F_1_772 equ 116130 ; FIX(1.77200) +F_0_402 equ (F_1_402 - 65536) ; FIX(1.40200) - FIX(1) +F_0_285 equ ( 65536 - F_0_714) ; FIX(1) - FIX(0.71414) +F_0_228 equ (131072 - F_1_772) ; FIX(2) - FIX(1.77200) + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_merged_upsample_mmx) + +EXTN(jconst_merged_upsample_mmx): + +PW_F0402 times 4 dw F_0_402 +PW_MF0228 times 4 dw -F_0_228 +PW_MF0344_F0285 times 2 dw -F_0_344, F_0_285 +PW_ONE times 4 dw 1 +PD_ONEHALF times 2 dd 1 << (SCALEBITS-1) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 + +%include "jdmrgext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGB_RED +%define RGB_GREEN EXT_RGB_GREEN +%define RGB_BLUE EXT_RGB_BLUE +%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE +%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extrgb_merged_upsample_mmx +%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extrgb_merged_upsample_mmx +%include "jdmrgext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGBX_RED +%define RGB_GREEN EXT_RGBX_GREEN +%define RGB_BLUE EXT_RGBX_BLUE +%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE +%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extrgbx_merged_upsample_mmx +%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extrgbx_merged_upsample_mmx +%include "jdmrgext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGR_RED +%define RGB_GREEN EXT_BGR_GREEN +%define RGB_BLUE EXT_BGR_BLUE +%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE +%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extbgr_merged_upsample_mmx +%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extbgr_merged_upsample_mmx +%include "jdmrgext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGRX_RED +%define RGB_GREEN EXT_BGRX_GREEN +%define RGB_BLUE EXT_BGRX_BLUE +%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE +%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extbgrx_merged_upsample_mmx +%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extbgrx_merged_upsample_mmx +%include "jdmrgext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XBGR_RED +%define RGB_GREEN EXT_XBGR_GREEN +%define RGB_BLUE EXT_XBGR_BLUE +%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE +%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extxbgr_merged_upsample_mmx +%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extxbgr_merged_upsample_mmx +%include "jdmrgext-mmx.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XRGB_RED +%define RGB_GREEN EXT_XRGB_GREEN +%define RGB_BLUE EXT_XRGB_BLUE +%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE +%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extxrgb_merged_upsample_mmx +%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extxrgb_merged_upsample_mmx +%include "jdmrgext-mmx.asm" diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdmerge-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jdmerge-sse2-64.asm new file mode 100644 index 000000000..8f953c71e --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdmerge-sse2-64.asm @@ -0,0 +1,126 @@ +; +; jdmerge.asm - merged upsampling/color conversion (64-bit SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + +%define SCALEBITS 16 + +F_0_344 equ 22554 ; FIX(0.34414) +F_0_714 equ 46802 ; FIX(0.71414) +F_1_402 equ 91881 ; FIX(1.40200) +F_1_772 equ 116130 ; FIX(1.77200) +F_0_402 equ (F_1_402 - 65536) ; FIX(1.40200) - FIX(1) +F_0_285 equ ( 65536 - F_0_714) ; FIX(1) - FIX(0.71414) +F_0_228 equ (131072 - F_1_772) ; FIX(2) - FIX(1.77200) + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_merged_upsample_sse2) + +EXTN(jconst_merged_upsample_sse2): + +PW_F0402 times 8 dw F_0_402 +PW_MF0228 times 8 dw -F_0_228 +PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285 +PW_ONE times 8 dw 1 +PD_ONEHALF times 4 dd 1 << (SCALEBITS-1) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 + +%include "jdmrgext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGB_RED +%define RGB_GREEN EXT_RGB_GREEN +%define RGB_BLUE EXT_RGB_BLUE +%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE +%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgb_merged_upsample_sse2 +%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgb_merged_upsample_sse2 +%include "jdmrgext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGBX_RED +%define RGB_GREEN EXT_RGBX_GREEN +%define RGB_BLUE EXT_RGBX_BLUE +%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE +%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgbx_merged_upsample_sse2 +%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgbx_merged_upsample_sse2 +%include "jdmrgext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGR_RED +%define RGB_GREEN EXT_BGR_GREEN +%define RGB_BLUE EXT_BGR_BLUE +%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE +%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgr_merged_upsample_sse2 +%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgr_merged_upsample_sse2 +%include "jdmrgext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGRX_RED +%define RGB_GREEN EXT_BGRX_GREEN +%define RGB_BLUE EXT_BGRX_BLUE +%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE +%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgrx_merged_upsample_sse2 +%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgrx_merged_upsample_sse2 +%include "jdmrgext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XBGR_RED +%define RGB_GREEN EXT_XBGR_GREEN +%define RGB_BLUE EXT_XBGR_BLUE +%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE +%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxbgr_merged_upsample_sse2 +%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxbgr_merged_upsample_sse2 +%include "jdmrgext-sse2-64.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XRGB_RED +%define RGB_GREEN EXT_XRGB_GREEN +%define RGB_BLUE EXT_XRGB_BLUE +%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE +%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxrgb_merged_upsample_sse2 +%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxrgb_merged_upsample_sse2 +%include "jdmrgext-sse2-64.asm" diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdmerge-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jdmerge-sse2.asm new file mode 100644 index 000000000..d22e82815 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdmerge-sse2.asm @@ -0,0 +1,126 @@ +; +; jdmerge.asm - merged upsampling/color conversion (SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + +%define SCALEBITS 16 + +F_0_344 equ 22554 ; FIX(0.34414) +F_0_714 equ 46802 ; FIX(0.71414) +F_1_402 equ 91881 ; FIX(1.40200) +F_1_772 equ 116130 ; FIX(1.77200) +F_0_402 equ (F_1_402 - 65536) ; FIX(1.40200) - FIX(1) +F_0_285 equ ( 65536 - F_0_714) ; FIX(1) - FIX(0.71414) +F_0_228 equ (131072 - F_1_772) ; FIX(2) - FIX(1.77200) + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_merged_upsample_sse2) + +EXTN(jconst_merged_upsample_sse2): + +PW_F0402 times 8 dw F_0_402 +PW_MF0228 times 8 dw -F_0_228 +PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285 +PW_ONE times 8 dw 1 +PD_ONEHALF times 4 dd 1 << (SCALEBITS-1) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 + +%include "jdmrgext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGB_RED +%define RGB_GREEN EXT_RGB_GREEN +%define RGB_BLUE EXT_RGB_BLUE +%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE +%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgb_merged_upsample_sse2 +%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgb_merged_upsample_sse2 +%include "jdmrgext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_RGBX_RED +%define RGB_GREEN EXT_RGBX_GREEN +%define RGB_BLUE EXT_RGBX_BLUE +%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE +%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgbx_merged_upsample_sse2 +%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgbx_merged_upsample_sse2 +%include "jdmrgext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGR_RED +%define RGB_GREEN EXT_BGR_GREEN +%define RGB_BLUE EXT_BGR_BLUE +%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE +%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgr_merged_upsample_sse2 +%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgr_merged_upsample_sse2 +%include "jdmrgext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_BGRX_RED +%define RGB_GREEN EXT_BGRX_GREEN +%define RGB_BLUE EXT_BGRX_BLUE +%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE +%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgrx_merged_upsample_sse2 +%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgrx_merged_upsample_sse2 +%include "jdmrgext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XBGR_RED +%define RGB_GREEN EXT_XBGR_GREEN +%define RGB_BLUE EXT_XBGR_BLUE +%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE +%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxbgr_merged_upsample_sse2 +%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxbgr_merged_upsample_sse2 +%include "jdmrgext-sse2.asm" + +%undef RGB_RED +%undef RGB_GREEN +%undef RGB_BLUE +%undef RGB_PIXELSIZE +%define RGB_RED EXT_XRGB_RED +%define RGB_GREEN EXT_XRGB_GREEN +%define RGB_BLUE EXT_XRGB_BLUE +%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE +%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxrgb_merged_upsample_sse2 +%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxrgb_merged_upsample_sse2 +%include "jdmrgext-sse2.asm" diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdmrgext-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jdmrgext-mmx.asm new file mode 100644 index 000000000..a92e9341c --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdmrgext-mmx.asm @@ -0,0 +1,464 @@ +; +; jdmrgext.asm - merged upsampling/color conversion (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jcolsamp.inc" + +; -------------------------------------------------------------------------- +; +; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical. +; +; GLOBAL(void) +; jsimd_h2v1_merged_upsample_mmx (JDIMENSION output_width, +; JSAMPIMAGE input_buf, +; JDIMENSION in_row_group_ctr, +; JSAMPARRAY output_buf); +; + +%define output_width(b) (b)+8 ; JDIMENSION output_width +%define input_buf(b) (b)+12 ; JSAMPIMAGE input_buf +%define in_row_group_ctr(b) (b)+16 ; JDIMENSION in_row_group_ctr +%define output_buf(b) (b)+20 ; JSAMPARRAY output_buf + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_MMWORD ; mmword wk[WK_NUM] +%define WK_NUM 3 +%define gotptr wk(0)-SIZEOF_POINTER ; void * gotptr + + align 16 + global EXTN(jsimd_h2v1_merged_upsample_mmx) + +EXTN(jsimd_h2v1_merged_upsample_mmx): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_MMWORD) ; align to 64 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic eax ; make a room for GOT address + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + movpic POINTER [gotptr], ebx ; save GOT address + + mov ecx, JDIMENSION [output_width(eax)] ; col + test ecx,ecx + jz near .return + + push ecx + + mov edi, JSAMPIMAGE [input_buf(eax)] + mov ecx, JDIMENSION [in_row_group_ctr(eax)] + mov esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY] + mov ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY] + mov edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY] + mov edi, JSAMPARRAY [output_buf(eax)] + mov esi, JSAMPROW [esi+ecx*SIZEOF_JSAMPROW] ; inptr0 + mov ebx, JSAMPROW [ebx+ecx*SIZEOF_JSAMPROW] ; inptr1 + mov edx, JSAMPROW [edx+ecx*SIZEOF_JSAMPROW] ; inptr2 + mov edi, JSAMPROW [edi] ; outptr + + pop ecx ; col + + alignx 16,7 +.columnloop: + movpic eax, POINTER [gotptr] ; load GOT address (eax) + + movq mm6, MMWORD [ebx] ; mm6=Cb(01234567) + movq mm7, MMWORD [edx] ; mm7=Cr(01234567) + + pxor mm1,mm1 ; mm1=(all 0's) + pcmpeqw mm3,mm3 + psllw mm3,7 ; mm3={0xFF80 0xFF80 0xFF80 0xFF80} + + movq mm4,mm6 + punpckhbw mm6,mm1 ; mm6=Cb(4567)=CbH + punpcklbw mm4,mm1 ; mm4=Cb(0123)=CbL + movq mm0,mm7 + punpckhbw mm7,mm1 ; mm7=Cr(4567)=CrH + punpcklbw mm0,mm1 ; mm0=Cr(0123)=CrL + + paddw mm6,mm3 + paddw mm4,mm3 + paddw mm7,mm3 + paddw mm0,mm3 + + ; (Original) + ; R = Y + 1.40200 * Cr + ; G = Y - 0.34414 * Cb - 0.71414 * Cr + ; B = Y + 1.77200 * Cb + ; + ; (This implementation) + ; R = Y + 0.40200 * Cr + Cr + ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr + ; B = Y - 0.22800 * Cb + Cb + Cb + + movq mm5,mm6 ; mm5=CbH + movq mm2,mm4 ; mm2=CbL + paddw mm6,mm6 ; mm6=2*CbH + paddw mm4,mm4 ; mm4=2*CbL + movq mm1,mm7 ; mm1=CrH + movq mm3,mm0 ; mm3=CrL + paddw mm7,mm7 ; mm7=2*CrH + paddw mm0,mm0 ; mm0=2*CrL + + pmulhw mm6,[GOTOFF(eax,PW_MF0228)] ; mm6=(2*CbH * -FIX(0.22800)) + pmulhw mm4,[GOTOFF(eax,PW_MF0228)] ; mm4=(2*CbL * -FIX(0.22800)) + pmulhw mm7,[GOTOFF(eax,PW_F0402)] ; mm7=(2*CrH * FIX(0.40200)) + pmulhw mm0,[GOTOFF(eax,PW_F0402)] ; mm0=(2*CrL * FIX(0.40200)) + + paddw mm6,[GOTOFF(eax,PW_ONE)] + paddw mm4,[GOTOFF(eax,PW_ONE)] + psraw mm6,1 ; mm6=(CbH * -FIX(0.22800)) + psraw mm4,1 ; mm4=(CbL * -FIX(0.22800)) + paddw mm7,[GOTOFF(eax,PW_ONE)] + paddw mm0,[GOTOFF(eax,PW_ONE)] + psraw mm7,1 ; mm7=(CrH * FIX(0.40200)) + psraw mm0,1 ; mm0=(CrL * FIX(0.40200)) + + paddw mm6,mm5 + paddw mm4,mm2 + paddw mm6,mm5 ; mm6=(CbH * FIX(1.77200))=(B-Y)H + paddw mm4,mm2 ; mm4=(CbL * FIX(1.77200))=(B-Y)L + paddw mm7,mm1 ; mm7=(CrH * FIX(1.40200))=(R-Y)H + paddw mm0,mm3 ; mm0=(CrL * FIX(1.40200))=(R-Y)L + + movq MMWORD [wk(0)], mm6 ; wk(0)=(B-Y)H + movq MMWORD [wk(1)], mm7 ; wk(1)=(R-Y)H + + movq mm6,mm5 + movq mm7,mm2 + punpcklwd mm5,mm1 + punpckhwd mm6,mm1 + pmaddwd mm5,[GOTOFF(eax,PW_MF0344_F0285)] + pmaddwd mm6,[GOTOFF(eax,PW_MF0344_F0285)] + punpcklwd mm2,mm3 + punpckhwd mm7,mm3 + pmaddwd mm2,[GOTOFF(eax,PW_MF0344_F0285)] + pmaddwd mm7,[GOTOFF(eax,PW_MF0344_F0285)] + + paddd mm5,[GOTOFF(eax,PD_ONEHALF)] + paddd mm6,[GOTOFF(eax,PD_ONEHALF)] + psrad mm5,SCALEBITS + psrad mm6,SCALEBITS + paddd mm2,[GOTOFF(eax,PD_ONEHALF)] + paddd mm7,[GOTOFF(eax,PD_ONEHALF)] + psrad mm2,SCALEBITS + psrad mm7,SCALEBITS + + packssdw mm5,mm6 ; mm5=CbH*-FIX(0.344)+CrH*FIX(0.285) + packssdw mm2,mm7 ; mm2=CbL*-FIX(0.344)+CrL*FIX(0.285) + psubw mm5,mm1 ; mm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H + psubw mm2,mm3 ; mm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L + + movq MMWORD [wk(2)], mm5 ; wk(2)=(G-Y)H + + mov al,2 ; Yctr + jmp short .Yloop_1st + alignx 16,7 + +.Yloop_2nd: + movq mm0, MMWORD [wk(1)] ; mm0=(R-Y)H + movq mm2, MMWORD [wk(2)] ; mm2=(G-Y)H + movq mm4, MMWORD [wk(0)] ; mm4=(B-Y)H + alignx 16,7 + +.Yloop_1st: + movq mm7, MMWORD [esi] ; mm7=Y(01234567) + + pcmpeqw mm6,mm6 + psrlw mm6,BYTE_BIT ; mm6={0xFF 0x00 0xFF 0x00 ..} + pand mm6,mm7 ; mm6=Y(0246)=YE + psrlw mm7,BYTE_BIT ; mm7=Y(1357)=YO + + movq mm1,mm0 ; mm1=mm0=(R-Y)(L/H) + movq mm3,mm2 ; mm3=mm2=(G-Y)(L/H) + movq mm5,mm4 ; mm5=mm4=(B-Y)(L/H) + + paddw mm0,mm6 ; mm0=((R-Y)+YE)=RE=(R0 R2 R4 R6) + paddw mm1,mm7 ; mm1=((R-Y)+YO)=RO=(R1 R3 R5 R7) + packuswb mm0,mm0 ; mm0=(R0 R2 R4 R6 ** ** ** **) + packuswb mm1,mm1 ; mm1=(R1 R3 R5 R7 ** ** ** **) + + paddw mm2,mm6 ; mm2=((G-Y)+YE)=GE=(G0 G2 G4 G6) + paddw mm3,mm7 ; mm3=((G-Y)+YO)=GO=(G1 G3 G5 G7) + packuswb mm2,mm2 ; mm2=(G0 G2 G4 G6 ** ** ** **) + packuswb mm3,mm3 ; mm3=(G1 G3 G5 G7 ** ** ** **) + + paddw mm4,mm6 ; mm4=((B-Y)+YE)=BE=(B0 B2 B4 B6) + paddw mm5,mm7 ; mm5=((B-Y)+YO)=BO=(B1 B3 B5 B7) + packuswb mm4,mm4 ; mm4=(B0 B2 B4 B6 ** ** ** **) + packuswb mm5,mm5 ; mm5=(B1 B3 B5 B7 ** ** ** **) + +%if RGB_PIXELSIZE == 3 ; --------------- + + ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **) + ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **) + ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **) + ; mmG=(** ** ** ** ** ** ** **), mmH=(** ** ** ** ** ** ** **) + + punpcklbw mmA,mmC ; mmA=(00 10 02 12 04 14 06 16) + punpcklbw mmE,mmB ; mmE=(20 01 22 03 24 05 26 07) + punpcklbw mmD,mmF ; mmD=(11 21 13 23 15 25 17 27) + + movq mmG,mmA + movq mmH,mmA + punpcklwd mmA,mmE ; mmA=(00 10 20 01 02 12 22 03) + punpckhwd mmG,mmE ; mmG=(04 14 24 05 06 16 26 07) + + psrlq mmH,2*BYTE_BIT ; mmH=(02 12 04 14 06 16 -- --) + psrlq mmE,2*BYTE_BIT ; mmE=(22 03 24 05 26 07 -- --) + + movq mmC,mmD + movq mmB,mmD + punpcklwd mmD,mmH ; mmD=(11 21 02 12 13 23 04 14) + punpckhwd mmC,mmH ; mmC=(15 25 06 16 17 27 -- --) + + psrlq mmB,2*BYTE_BIT ; mmB=(13 23 15 25 17 27 -- --) + + movq mmF,mmE + punpcklwd mmE,mmB ; mmE=(22 03 13 23 24 05 15 25) + punpckhwd mmF,mmB ; mmF=(26 07 17 27 -- -- -- --) + + punpckldq mmA,mmD ; mmA=(00 10 20 01 11 21 02 12) + punpckldq mmE,mmG ; mmE=(22 03 13 23 04 14 24 05) + punpckldq mmC,mmF ; mmC=(15 25 06 16 26 07 17 27) + + cmp ecx, byte SIZEOF_MMWORD + jb short .column_st16 + + movq MMWORD [edi+0*SIZEOF_MMWORD], mmA + movq MMWORD [edi+1*SIZEOF_MMWORD], mmE + movq MMWORD [edi+2*SIZEOF_MMWORD], mmC + + sub ecx, byte SIZEOF_MMWORD + jz near .endcolumn + + add edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD ; outptr + add esi, byte SIZEOF_MMWORD ; inptr0 + dec al ; Yctr + jnz near .Yloop_2nd + + add ebx, byte SIZEOF_MMWORD ; inptr1 + add edx, byte SIZEOF_MMWORD ; inptr2 + jmp near .columnloop + alignx 16,7 + +.column_st16: + lea ecx, [ecx+ecx*2] ; imul ecx, RGB_PIXELSIZE + cmp ecx, byte 2*SIZEOF_MMWORD + jb short .column_st8 + movq MMWORD [edi+0*SIZEOF_MMWORD], mmA + movq MMWORD [edi+1*SIZEOF_MMWORD], mmE + movq mmA,mmC + sub ecx, byte 2*SIZEOF_MMWORD + add edi, byte 2*SIZEOF_MMWORD + jmp short .column_st4 +.column_st8: + cmp ecx, byte SIZEOF_MMWORD + jb short .column_st4 + movq MMWORD [edi+0*SIZEOF_MMWORD], mmA + movq mmA,mmE + sub ecx, byte SIZEOF_MMWORD + add edi, byte SIZEOF_MMWORD +.column_st4: + movd eax,mmA + cmp ecx, byte SIZEOF_DWORD + jb short .column_st2 + mov DWORD [edi+0*SIZEOF_DWORD], eax + psrlq mmA,DWORD_BIT + movd eax,mmA + sub ecx, byte SIZEOF_DWORD + add edi, byte SIZEOF_DWORD +.column_st2: + cmp ecx, byte SIZEOF_WORD + jb short .column_st1 + mov WORD [edi+0*SIZEOF_WORD], ax + shr eax,WORD_BIT + sub ecx, byte SIZEOF_WORD + add edi, byte SIZEOF_WORD +.column_st1: + cmp ecx, byte SIZEOF_BYTE + jb short .endcolumn + mov BYTE [edi+0*SIZEOF_BYTE], al + +%else ; RGB_PIXELSIZE == 4 ; ----------- + +%ifdef RGBX_FILLER_0XFF + pcmpeqb mm6,mm6 ; mm6=(X0 X2 X4 X6 ** ** ** **) + pcmpeqb mm7,mm7 ; mm7=(X1 X3 X5 X7 ** ** ** **) +%else + pxor mm6,mm6 ; mm6=(X0 X2 X4 X6 ** ** ** **) + pxor mm7,mm7 ; mm7=(X1 X3 X5 X7 ** ** ** **) +%endif + ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **) + ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **) + ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **) + ; mmG=(30 32 34 36 ** ** ** **), mmH=(31 33 35 37 ** ** ** **) + + punpcklbw mmA,mmC ; mmA=(00 10 02 12 04 14 06 16) + punpcklbw mmE,mmG ; mmE=(20 30 22 32 24 34 26 36) + punpcklbw mmB,mmD ; mmB=(01 11 03 13 05 15 07 17) + punpcklbw mmF,mmH ; mmF=(21 31 23 33 25 35 27 37) + + movq mmC,mmA + punpcklwd mmA,mmE ; mmA=(00 10 20 30 02 12 22 32) + punpckhwd mmC,mmE ; mmC=(04 14 24 34 06 16 26 36) + movq mmG,mmB + punpcklwd mmB,mmF ; mmB=(01 11 21 31 03 13 23 33) + punpckhwd mmG,mmF ; mmG=(05 15 25 35 07 17 27 37) + + movq mmD,mmA + punpckldq mmA,mmB ; mmA=(00 10 20 30 01 11 21 31) + punpckhdq mmD,mmB ; mmD=(02 12 22 32 03 13 23 33) + movq mmH,mmC + punpckldq mmC,mmG ; mmC=(04 14 24 34 05 15 25 35) + punpckhdq mmH,mmG ; mmH=(06 16 26 36 07 17 27 37) + + cmp ecx, byte SIZEOF_MMWORD + jb short .column_st16 + + movq MMWORD [edi+0*SIZEOF_MMWORD], mmA + movq MMWORD [edi+1*SIZEOF_MMWORD], mmD + movq MMWORD [edi+2*SIZEOF_MMWORD], mmC + movq MMWORD [edi+3*SIZEOF_MMWORD], mmH + + sub ecx, byte SIZEOF_MMWORD + jz short .endcolumn + + add edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD ; outptr + add esi, byte SIZEOF_MMWORD ; inptr0 + dec al ; Yctr + jnz near .Yloop_2nd + + add ebx, byte SIZEOF_MMWORD ; inptr1 + add edx, byte SIZEOF_MMWORD ; inptr2 + jmp near .columnloop + alignx 16,7 + +.column_st16: + cmp ecx, byte SIZEOF_MMWORD/2 + jb short .column_st8 + movq MMWORD [edi+0*SIZEOF_MMWORD], mmA + movq MMWORD [edi+1*SIZEOF_MMWORD], mmD + movq mmA,mmC + movq mmD,mmH + sub ecx, byte SIZEOF_MMWORD/2 + add edi, byte 2*SIZEOF_MMWORD +.column_st8: + cmp ecx, byte SIZEOF_MMWORD/4 + jb short .column_st4 + movq MMWORD [edi+0*SIZEOF_MMWORD], mmA + movq mmA,mmD + sub ecx, byte SIZEOF_MMWORD/4 + add edi, byte 1*SIZEOF_MMWORD +.column_st4: + cmp ecx, byte SIZEOF_MMWORD/8 + jb short .endcolumn + movd DWORD [edi+0*SIZEOF_DWORD], mmA + +%endif ; RGB_PIXELSIZE ; --------------- + +.endcolumn: + emms ; empty MMX state + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; -------------------------------------------------------------------------- +; +; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical. +; +; GLOBAL(void) +; jsimd_h2v2_merged_upsample_mmx (JDIMENSION output_width, +; JSAMPIMAGE input_buf, +; JDIMENSION in_row_group_ctr, +; JSAMPARRAY output_buf); +; + +%define output_width(b) (b)+8 ; JDIMENSION output_width +%define input_buf(b) (b)+12 ; JSAMPIMAGE input_buf +%define in_row_group_ctr(b) (b)+16 ; JDIMENSION in_row_group_ctr +%define output_buf(b) (b)+20 ; JSAMPARRAY output_buf + + align 16 + global EXTN(jsimd_h2v2_merged_upsample_mmx) + +EXTN(jsimd_h2v2_merged_upsample_mmx): + push ebp + mov ebp,esp + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + mov eax, JDIMENSION [output_width(ebp)] + + mov edi, JSAMPIMAGE [input_buf(ebp)] + mov ecx, JDIMENSION [in_row_group_ctr(ebp)] + mov esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY] + mov ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY] + mov edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY] + mov edi, JSAMPARRAY [output_buf(ebp)] + lea esi, [esi+ecx*SIZEOF_JSAMPROW] + + push edx ; inptr2 + push ebx ; inptr1 + push esi ; inptr00 + mov ebx,esp + + push edi ; output_buf (outptr0) + push ecx ; in_row_group_ctr + push ebx ; input_buf + push eax ; output_width + + call near EXTN(jsimd_h2v1_merged_upsample_mmx) + + add esi, byte SIZEOF_JSAMPROW ; inptr01 + add edi, byte SIZEOF_JSAMPROW ; outptr1 + mov POINTER [ebx+0*SIZEOF_POINTER], esi + mov POINTER [ebx-1*SIZEOF_POINTER], edi + + call near EXTN(jsimd_h2v1_merged_upsample_mmx) + + add esp, byte 7*SIZEOF_DWORD + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdmrgext-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jdmrgext-sse2-64.asm new file mode 100644 index 000000000..ff127b573 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdmrgext-sse2-64.asm @@ -0,0 +1,538 @@ +; +; jdmrgext.asm - merged upsampling/color conversion (64-bit SSE2) +; +; Copyright 2009, 2012 Pierre Ossman for Cendio AB +; Copyright 2009, 2012 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jcolsamp.inc" + +; -------------------------------------------------------------------------- +; +; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical. +; +; GLOBAL(void) +; jsimd_h2v1_merged_upsample_sse2 (JDIMENSION output_width, +; JSAMPIMAGE input_buf, +; JDIMENSION in_row_group_ctr, +; JSAMPARRAY output_buf); +; + +; r10 = JDIMENSION output_width +; r11 = JSAMPIMAGE input_buf +; r12 = JDIMENSION in_row_group_ctr +; r13 = JSAMPARRAY output_buf + +%define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 3 + + align 16 + global EXTN(jsimd_h2v1_merged_upsample_sse2) + +EXTN(jsimd_h2v1_merged_upsample_sse2): + push rbp + mov rax,rsp ; rax = original rbp + sub rsp, byte 4 + and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [rsp],rax + mov rbp,rsp ; rbp = aligned rbp + lea rsp, [wk(0)] + collect_args + push rbx + + mov rcx, r10 ; col + test rcx,rcx + jz near .return + + push rcx + + mov rdi, r11 + mov rcx, r12 + mov rsi, JSAMPARRAY [rdi+0*SIZEOF_JSAMPARRAY] + mov rbx, JSAMPARRAY [rdi+1*SIZEOF_JSAMPARRAY] + mov rdx, JSAMPARRAY [rdi+2*SIZEOF_JSAMPARRAY] + mov rdi, r13 + mov rsi, JSAMPROW [rsi+rcx*SIZEOF_JSAMPROW] ; inptr0 + mov rbx, JSAMPROW [rbx+rcx*SIZEOF_JSAMPROW] ; inptr1 + mov rdx, JSAMPROW [rdx+rcx*SIZEOF_JSAMPROW] ; inptr2 + mov rdi, JSAMPROW [rdi] ; outptr + + pop rcx ; col + +.columnloop: + + movdqa xmm6, XMMWORD [rbx] ; xmm6=Cb(0123456789ABCDEF) + movdqa xmm7, XMMWORD [rdx] ; xmm7=Cr(0123456789ABCDEF) + + pxor xmm1,xmm1 ; xmm1=(all 0's) + pcmpeqw xmm3,xmm3 + psllw xmm3,7 ; xmm3={0xFF80 0xFF80 0xFF80 0xFF80 ..} + + movdqa xmm4,xmm6 + punpckhbw xmm6,xmm1 ; xmm6=Cb(89ABCDEF)=CbH + punpcklbw xmm4,xmm1 ; xmm4=Cb(01234567)=CbL + movdqa xmm0,xmm7 + punpckhbw xmm7,xmm1 ; xmm7=Cr(89ABCDEF)=CrH + punpcklbw xmm0,xmm1 ; xmm0=Cr(01234567)=CrL + + paddw xmm6,xmm3 + paddw xmm4,xmm3 + paddw xmm7,xmm3 + paddw xmm0,xmm3 + + ; (Original) + ; R = Y + 1.40200 * Cr + ; G = Y - 0.34414 * Cb - 0.71414 * Cr + ; B = Y + 1.77200 * Cb + ; + ; (This implementation) + ; R = Y + 0.40200 * Cr + Cr + ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr + ; B = Y - 0.22800 * Cb + Cb + Cb + + movdqa xmm5,xmm6 ; xmm5=CbH + movdqa xmm2,xmm4 ; xmm2=CbL + paddw xmm6,xmm6 ; xmm6=2*CbH + paddw xmm4,xmm4 ; xmm4=2*CbL + movdqa xmm1,xmm7 ; xmm1=CrH + movdqa xmm3,xmm0 ; xmm3=CrL + paddw xmm7,xmm7 ; xmm7=2*CrH + paddw xmm0,xmm0 ; xmm0=2*CrL + + pmulhw xmm6,[rel PW_MF0228] ; xmm6=(2*CbH * -FIX(0.22800)) + pmulhw xmm4,[rel PW_MF0228] ; xmm4=(2*CbL * -FIX(0.22800)) + pmulhw xmm7,[rel PW_F0402] ; xmm7=(2*CrH * FIX(0.40200)) + pmulhw xmm0,[rel PW_F0402] ; xmm0=(2*CrL * FIX(0.40200)) + + paddw xmm6,[rel PW_ONE] + paddw xmm4,[rel PW_ONE] + psraw xmm6,1 ; xmm6=(CbH * -FIX(0.22800)) + psraw xmm4,1 ; xmm4=(CbL * -FIX(0.22800)) + paddw xmm7,[rel PW_ONE] + paddw xmm0,[rel PW_ONE] + psraw xmm7,1 ; xmm7=(CrH * FIX(0.40200)) + psraw xmm0,1 ; xmm0=(CrL * FIX(0.40200)) + + paddw xmm6,xmm5 + paddw xmm4,xmm2 + paddw xmm6,xmm5 ; xmm6=(CbH * FIX(1.77200))=(B-Y)H + paddw xmm4,xmm2 ; xmm4=(CbL * FIX(1.77200))=(B-Y)L + paddw xmm7,xmm1 ; xmm7=(CrH * FIX(1.40200))=(R-Y)H + paddw xmm0,xmm3 ; xmm0=(CrL * FIX(1.40200))=(R-Y)L + + movdqa XMMWORD [wk(0)], xmm6 ; wk(0)=(B-Y)H + movdqa XMMWORD [wk(1)], xmm7 ; wk(1)=(R-Y)H + + movdqa xmm6,xmm5 + movdqa xmm7,xmm2 + punpcklwd xmm5,xmm1 + punpckhwd xmm6,xmm1 + pmaddwd xmm5,[rel PW_MF0344_F0285] + pmaddwd xmm6,[rel PW_MF0344_F0285] + punpcklwd xmm2,xmm3 + punpckhwd xmm7,xmm3 + pmaddwd xmm2,[rel PW_MF0344_F0285] + pmaddwd xmm7,[rel PW_MF0344_F0285] + + paddd xmm5,[rel PD_ONEHALF] + paddd xmm6,[rel PD_ONEHALF] + psrad xmm5,SCALEBITS + psrad xmm6,SCALEBITS + paddd xmm2,[rel PD_ONEHALF] + paddd xmm7,[rel PD_ONEHALF] + psrad xmm2,SCALEBITS + psrad xmm7,SCALEBITS + + packssdw xmm5,xmm6 ; xmm5=CbH*-FIX(0.344)+CrH*FIX(0.285) + packssdw xmm2,xmm7 ; xmm2=CbL*-FIX(0.344)+CrL*FIX(0.285) + psubw xmm5,xmm1 ; xmm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H + psubw xmm2,xmm3 ; xmm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L + + movdqa XMMWORD [wk(2)], xmm5 ; wk(2)=(G-Y)H + + mov al,2 ; Yctr + jmp short .Yloop_1st + +.Yloop_2nd: + movdqa xmm0, XMMWORD [wk(1)] ; xmm0=(R-Y)H + movdqa xmm2, XMMWORD [wk(2)] ; xmm2=(G-Y)H + movdqa xmm4, XMMWORD [wk(0)] ; xmm4=(B-Y)H + +.Yloop_1st: + movdqa xmm7, XMMWORD [rsi] ; xmm7=Y(0123456789ABCDEF) + + pcmpeqw xmm6,xmm6 + psrlw xmm6,BYTE_BIT ; xmm6={0xFF 0x00 0xFF 0x00 ..} + pand xmm6,xmm7 ; xmm6=Y(02468ACE)=YE + psrlw xmm7,BYTE_BIT ; xmm7=Y(13579BDF)=YO + + movdqa xmm1,xmm0 ; xmm1=xmm0=(R-Y)(L/H) + movdqa xmm3,xmm2 ; xmm3=xmm2=(G-Y)(L/H) + movdqa xmm5,xmm4 ; xmm5=xmm4=(B-Y)(L/H) + + paddw xmm0,xmm6 ; xmm0=((R-Y)+YE)=RE=R(02468ACE) + paddw xmm1,xmm7 ; xmm1=((R-Y)+YO)=RO=R(13579BDF) + packuswb xmm0,xmm0 ; xmm0=R(02468ACE********) + packuswb xmm1,xmm1 ; xmm1=R(13579BDF********) + + paddw xmm2,xmm6 ; xmm2=((G-Y)+YE)=GE=G(02468ACE) + paddw xmm3,xmm7 ; xmm3=((G-Y)+YO)=GO=G(13579BDF) + packuswb xmm2,xmm2 ; xmm2=G(02468ACE********) + packuswb xmm3,xmm3 ; xmm3=G(13579BDF********) + + paddw xmm4,xmm6 ; xmm4=((B-Y)+YE)=BE=B(02468ACE) + paddw xmm5,xmm7 ; xmm5=((B-Y)+YO)=BO=B(13579BDF) + packuswb xmm4,xmm4 ; xmm4=B(02468ACE********) + packuswb xmm5,xmm5 ; xmm5=B(13579BDF********) + +%if RGB_PIXELSIZE == 3 ; --------------- + + ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **) + ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **) + ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **) + ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **) + + punpcklbw xmmA,xmmC ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E) + punpcklbw xmmE,xmmB ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F) + punpcklbw xmmD,xmmF ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F) + + movdqa xmmG,xmmA + movdqa xmmH,xmmA + punpcklwd xmmA,xmmE ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07) + punpckhwd xmmG,xmmE ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F) + + psrldq xmmH,2 ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --) + psrldq xmmE,2 ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --) + + movdqa xmmC,xmmD + movdqa xmmB,xmmD + punpcklwd xmmD,xmmH ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18) + punpckhwd xmmC,xmmH ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --) + + psrldq xmmB,2 ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --) + + movdqa xmmF,xmmE + punpcklwd xmmE,xmmB ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29) + punpckhwd xmmF,xmmB ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --) + + pshufd xmmH,xmmA,0x4E; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03) + movdqa xmmB,xmmE + punpckldq xmmA,xmmD ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14) + punpckldq xmmE,xmmH ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07) + punpckhdq xmmD,xmmB ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29) + + pshufd xmmH,xmmG,0x4E; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B) + movdqa xmmB,xmmF + punpckldq xmmG,xmmC ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C) + punpckldq xmmF,xmmH ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F) + punpckhdq xmmC,xmmB ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --) + + punpcklqdq xmmA,xmmE ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05) + punpcklqdq xmmD,xmmG ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A) + punpcklqdq xmmF,xmmC ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F) + + cmp rcx, byte SIZEOF_XMMWORD + jb short .column_st32 + + test rdi, SIZEOF_XMMWORD-1 + jnz short .out1 + ; --(aligned)------------------- + movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD + movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmF + jmp short .out0 +.out1: ; --(unaligned)----------------- + movdqu XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD + movdqu XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmF +.out0: + add rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; outptr + sub rcx, byte SIZEOF_XMMWORD + jz near .endcolumn + + add rsi, byte SIZEOF_XMMWORD ; inptr0 + dec al ; Yctr + jnz near .Yloop_2nd + + add rbx, byte SIZEOF_XMMWORD ; inptr1 + add rdx, byte SIZEOF_XMMWORD ; inptr2 + jmp near .columnloop + +.column_st32: + lea rcx, [rcx+rcx*2] ; imul ecx, RGB_PIXELSIZE + cmp rcx, byte 2*SIZEOF_XMMWORD + jb short .column_st16 + movdqu XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD + add rdi, byte 2*SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmF + sub rcx, byte 2*SIZEOF_XMMWORD + jmp short .column_st15 +.column_st16: + cmp rcx, byte SIZEOF_XMMWORD + jb short .column_st15 + movdqu XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + add rdi, byte SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmD + sub rcx, byte SIZEOF_XMMWORD +.column_st15: + ; Store the lower 8 bytes of xmmA to the output when it has enough + ; space. + cmp rcx, byte SIZEOF_MMWORD + jb short .column_st7 + movq XMM_MMWORD [rdi], xmmA + add rdi, byte SIZEOF_MMWORD + sub rcx, byte SIZEOF_MMWORD + psrldq xmmA, SIZEOF_MMWORD +.column_st7: + ; Store the lower 4 bytes of xmmA to the output when it has enough + ; space. + cmp rcx, byte SIZEOF_DWORD + jb short .column_st3 + movd XMM_DWORD [rdi], xmmA + add rdi, byte SIZEOF_DWORD + sub rcx, byte SIZEOF_DWORD + psrldq xmmA, SIZEOF_DWORD +.column_st3: + ; Store the lower 2 bytes of rax to the output when it has enough + ; space. + movd eax, xmmA + cmp rcx, byte SIZEOF_WORD + jb short .column_st1 + mov WORD [rdi], ax + add rdi, byte SIZEOF_WORD + sub rcx, byte SIZEOF_WORD + shr rax, 16 +.column_st1: + ; Store the lower 1 byte of rax to the output when it has enough + ; space. + test rcx, rcx + jz short .endcolumn + mov BYTE [rdi], al + +%else ; RGB_PIXELSIZE == 4 ; ----------- + +%ifdef RGBX_FILLER_0XFF + pcmpeqb xmm6,xmm6 ; xmm6=XE=X(02468ACE********) + pcmpeqb xmm7,xmm7 ; xmm7=XO=X(13579BDF********) +%else + pxor xmm6,xmm6 ; xmm6=XE=X(02468ACE********) + pxor xmm7,xmm7 ; xmm7=XO=X(13579BDF********) +%endif + ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **) + ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **) + ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **) + ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **) + + punpcklbw xmmA,xmmC ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E) + punpcklbw xmmE,xmmG ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E) + punpcklbw xmmB,xmmD ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F) + punpcklbw xmmF,xmmH ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F) + + movdqa xmmC,xmmA + punpcklwd xmmA,xmmE ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36) + punpckhwd xmmC,xmmE ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E) + movdqa xmmG,xmmB + punpcklwd xmmB,xmmF ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37) + punpckhwd xmmG,xmmF ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F) + + movdqa xmmD,xmmA + punpckldq xmmA,xmmB ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33) + punpckhdq xmmD,xmmB ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37) + movdqa xmmH,xmmC + punpckldq xmmC,xmmG ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B) + punpckhdq xmmH,xmmG ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F) + + cmp rcx, byte SIZEOF_XMMWORD + jb short .column_st32 + + test rdi, SIZEOF_XMMWORD-1 + jnz short .out1 + ; --(aligned)------------------- + movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD + movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmC + movntdq XMMWORD [rdi+3*SIZEOF_XMMWORD], xmmH + jmp short .out0 +.out1: ; --(unaligned)----------------- + movdqu XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD + movdqu XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmC + movdqu XMMWORD [rdi+3*SIZEOF_XMMWORD], xmmH +.out0: + add rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; outptr + sub rcx, byte SIZEOF_XMMWORD + jz near .endcolumn + + add rsi, byte SIZEOF_XMMWORD ; inptr0 + dec al ; Yctr + jnz near .Yloop_2nd + + add rbx, byte SIZEOF_XMMWORD ; inptr1 + add rdx, byte SIZEOF_XMMWORD ; inptr2 + jmp near .columnloop + +.column_st32: + cmp rcx, byte SIZEOF_XMMWORD/2 + jb short .column_st16 + movdqu XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD + add rdi, byte 2*SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmC + movdqa xmmD,xmmH + sub rcx, byte SIZEOF_XMMWORD/2 +.column_st16: + cmp rcx, byte SIZEOF_XMMWORD/4 + jb short .column_st15 + movdqu XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA + add rdi, byte SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmD + sub rcx, byte SIZEOF_XMMWORD/4 +.column_st15: + ; Store two pixels (8 bytes) of xmmA to the output when it has enough + ; space. + cmp rcx, byte SIZEOF_XMMWORD/8 + jb short .column_st7 + movq XMM_MMWORD [rdi], xmmA + add rdi, byte SIZEOF_XMMWORD/8*4 + sub rcx, byte SIZEOF_XMMWORD/8 + psrldq xmmA, SIZEOF_XMMWORD/8*4 +.column_st7: + ; Store one pixel (4 bytes) of xmmA to the output when it has enough + ; space. + test rcx, rcx + jz short .endcolumn + movd XMM_DWORD [rdi], xmmA + +%endif ; RGB_PIXELSIZE ; --------------- + +.endcolumn: + sfence ; flush the write buffer + +.return: + pop rbx + uncollect_args + mov rsp,rbp ; rsp <- aligned rbp + pop rsp ; rsp <- original rbp + pop rbp + ret + +; -------------------------------------------------------------------------- +; +; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical. +; +; GLOBAL(void) +; jsimd_h2v2_merged_upsample_sse2 (JDIMENSION output_width, +; JSAMPIMAGE input_buf, +; JDIMENSION in_row_group_ctr, +; JSAMPARRAY output_buf); +; + +; r10 = JDIMENSION output_width +; r11 = JSAMPIMAGE input_buf +; r12 = JDIMENSION in_row_group_ctr +; r13 = JSAMPARRAY output_buf + + align 16 + global EXTN(jsimd_h2v2_merged_upsample_sse2) + +EXTN(jsimd_h2v2_merged_upsample_sse2): + push rbp + mov rax,rsp + mov rbp,rsp + collect_args + push rbx + + mov rax, r10 + + mov rdi, r11 + mov rcx, r12 + mov rsi, JSAMPARRAY [rdi+0*SIZEOF_JSAMPARRAY] + mov rbx, JSAMPARRAY [rdi+1*SIZEOF_JSAMPARRAY] + mov rdx, JSAMPARRAY [rdi+2*SIZEOF_JSAMPARRAY] + mov rdi, r13 + lea rsi, [rsi+rcx*SIZEOF_JSAMPROW] + + push rdx ; inptr2 + push rbx ; inptr1 + push rsi ; inptr00 + mov rbx,rsp + + push rdi + push rcx + push rax + + %ifdef WIN64 + mov r8, rcx + mov r9, rdi + mov rcx, rax + mov rdx, rbx + %else + mov rdx, rcx + mov rcx, rdi + mov rdi, rax + mov rsi, rbx + %endif + + call EXTN(jsimd_h2v1_merged_upsample_sse2) + + pop rax + pop rcx + pop rdi + pop rsi + pop rbx + pop rdx + + add rdi, byte SIZEOF_JSAMPROW ; outptr1 + add rsi, byte SIZEOF_JSAMPROW ; inptr01 + + push rdx ; inptr2 + push rbx ; inptr1 + push rsi ; inptr00 + mov rbx,rsp + + push rdi + push rcx + push rax + + %ifdef WIN64 + mov r8, rcx + mov r9, rdi + mov rcx, rax + mov rdx, rbx + %else + mov rdx, rcx + mov rcx, rdi + mov rdi, rax + mov rsi, rbx + %endif + + call EXTN(jsimd_h2v1_merged_upsample_sse2) + + pop rax + pop rcx + pop rdi + pop rsi + pop rbx + pop rdx + + pop rbx + uncollect_args + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdmrgext-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jdmrgext-sse2.asm new file mode 100644 index 000000000..c47916f64 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdmrgext-sse2.asm @@ -0,0 +1,519 @@ +; +; jdmrgext.asm - merged upsampling/color conversion (SSE2) +; +; Copyright 2009, 2012 Pierre Ossman for Cendio AB +; Copyright 2012 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jcolsamp.inc" + +; -------------------------------------------------------------------------- +; +; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical. +; +; GLOBAL(void) +; jsimd_h2v1_merged_upsample_sse2 (JDIMENSION output_width, +; JSAMPIMAGE input_buf, +; JDIMENSION in_row_group_ctr, +; JSAMPARRAY output_buf); +; + +%define output_width(b) (b)+8 ; JDIMENSION output_width +%define input_buf(b) (b)+12 ; JSAMPIMAGE input_buf +%define in_row_group_ctr(b) (b)+16 ; JDIMENSION in_row_group_ctr +%define output_buf(b) (b)+20 ; JSAMPARRAY output_buf + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 3 +%define gotptr wk(0)-SIZEOF_POINTER ; void * gotptr + + align 16 + global EXTN(jsimd_h2v1_merged_upsample_sse2) + +EXTN(jsimd_h2v1_merged_upsample_sse2): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic eax ; make a room for GOT address + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + movpic POINTER [gotptr], ebx ; save GOT address + + mov ecx, JDIMENSION [output_width(eax)] ; col + test ecx,ecx + jz near .return + + push ecx + + mov edi, JSAMPIMAGE [input_buf(eax)] + mov ecx, JDIMENSION [in_row_group_ctr(eax)] + mov esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY] + mov ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY] + mov edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY] + mov edi, JSAMPARRAY [output_buf(eax)] + mov esi, JSAMPROW [esi+ecx*SIZEOF_JSAMPROW] ; inptr0 + mov ebx, JSAMPROW [ebx+ecx*SIZEOF_JSAMPROW] ; inptr1 + mov edx, JSAMPROW [edx+ecx*SIZEOF_JSAMPROW] ; inptr2 + mov edi, JSAMPROW [edi] ; outptr + + pop ecx ; col + + alignx 16,7 +.columnloop: + movpic eax, POINTER [gotptr] ; load GOT address (eax) + + movdqa xmm6, XMMWORD [ebx] ; xmm6=Cb(0123456789ABCDEF) + movdqa xmm7, XMMWORD [edx] ; xmm7=Cr(0123456789ABCDEF) + + pxor xmm1,xmm1 ; xmm1=(all 0's) + pcmpeqw xmm3,xmm3 + psllw xmm3,7 ; xmm3={0xFF80 0xFF80 0xFF80 0xFF80 ..} + + movdqa xmm4,xmm6 + punpckhbw xmm6,xmm1 ; xmm6=Cb(89ABCDEF)=CbH + punpcklbw xmm4,xmm1 ; xmm4=Cb(01234567)=CbL + movdqa xmm0,xmm7 + punpckhbw xmm7,xmm1 ; xmm7=Cr(89ABCDEF)=CrH + punpcklbw xmm0,xmm1 ; xmm0=Cr(01234567)=CrL + + paddw xmm6,xmm3 + paddw xmm4,xmm3 + paddw xmm7,xmm3 + paddw xmm0,xmm3 + + ; (Original) + ; R = Y + 1.40200 * Cr + ; G = Y - 0.34414 * Cb - 0.71414 * Cr + ; B = Y + 1.77200 * Cb + ; + ; (This implementation) + ; R = Y + 0.40200 * Cr + Cr + ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr + ; B = Y - 0.22800 * Cb + Cb + Cb + + movdqa xmm5,xmm6 ; xmm5=CbH + movdqa xmm2,xmm4 ; xmm2=CbL + paddw xmm6,xmm6 ; xmm6=2*CbH + paddw xmm4,xmm4 ; xmm4=2*CbL + movdqa xmm1,xmm7 ; xmm1=CrH + movdqa xmm3,xmm0 ; xmm3=CrL + paddw xmm7,xmm7 ; xmm7=2*CrH + paddw xmm0,xmm0 ; xmm0=2*CrL + + pmulhw xmm6,[GOTOFF(eax,PW_MF0228)] ; xmm6=(2*CbH * -FIX(0.22800)) + pmulhw xmm4,[GOTOFF(eax,PW_MF0228)] ; xmm4=(2*CbL * -FIX(0.22800)) + pmulhw xmm7,[GOTOFF(eax,PW_F0402)] ; xmm7=(2*CrH * FIX(0.40200)) + pmulhw xmm0,[GOTOFF(eax,PW_F0402)] ; xmm0=(2*CrL * FIX(0.40200)) + + paddw xmm6,[GOTOFF(eax,PW_ONE)] + paddw xmm4,[GOTOFF(eax,PW_ONE)] + psraw xmm6,1 ; xmm6=(CbH * -FIX(0.22800)) + psraw xmm4,1 ; xmm4=(CbL * -FIX(0.22800)) + paddw xmm7,[GOTOFF(eax,PW_ONE)] + paddw xmm0,[GOTOFF(eax,PW_ONE)] + psraw xmm7,1 ; xmm7=(CrH * FIX(0.40200)) + psraw xmm0,1 ; xmm0=(CrL * FIX(0.40200)) + + paddw xmm6,xmm5 + paddw xmm4,xmm2 + paddw xmm6,xmm5 ; xmm6=(CbH * FIX(1.77200))=(B-Y)H + paddw xmm4,xmm2 ; xmm4=(CbL * FIX(1.77200))=(B-Y)L + paddw xmm7,xmm1 ; xmm7=(CrH * FIX(1.40200))=(R-Y)H + paddw xmm0,xmm3 ; xmm0=(CrL * FIX(1.40200))=(R-Y)L + + movdqa XMMWORD [wk(0)], xmm6 ; wk(0)=(B-Y)H + movdqa XMMWORD [wk(1)], xmm7 ; wk(1)=(R-Y)H + + movdqa xmm6,xmm5 + movdqa xmm7,xmm2 + punpcklwd xmm5,xmm1 + punpckhwd xmm6,xmm1 + pmaddwd xmm5,[GOTOFF(eax,PW_MF0344_F0285)] + pmaddwd xmm6,[GOTOFF(eax,PW_MF0344_F0285)] + punpcklwd xmm2,xmm3 + punpckhwd xmm7,xmm3 + pmaddwd xmm2,[GOTOFF(eax,PW_MF0344_F0285)] + pmaddwd xmm7,[GOTOFF(eax,PW_MF0344_F0285)] + + paddd xmm5,[GOTOFF(eax,PD_ONEHALF)] + paddd xmm6,[GOTOFF(eax,PD_ONEHALF)] + psrad xmm5,SCALEBITS + psrad xmm6,SCALEBITS + paddd xmm2,[GOTOFF(eax,PD_ONEHALF)] + paddd xmm7,[GOTOFF(eax,PD_ONEHALF)] + psrad xmm2,SCALEBITS + psrad xmm7,SCALEBITS + + packssdw xmm5,xmm6 ; xmm5=CbH*-FIX(0.344)+CrH*FIX(0.285) + packssdw xmm2,xmm7 ; xmm2=CbL*-FIX(0.344)+CrL*FIX(0.285) + psubw xmm5,xmm1 ; xmm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H + psubw xmm2,xmm3 ; xmm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L + + movdqa XMMWORD [wk(2)], xmm5 ; wk(2)=(G-Y)H + + mov al,2 ; Yctr + jmp short .Yloop_1st + alignx 16,7 + +.Yloop_2nd: + movdqa xmm0, XMMWORD [wk(1)] ; xmm0=(R-Y)H + movdqa xmm2, XMMWORD [wk(2)] ; xmm2=(G-Y)H + movdqa xmm4, XMMWORD [wk(0)] ; xmm4=(B-Y)H + alignx 16,7 + +.Yloop_1st: + movdqa xmm7, XMMWORD [esi] ; xmm7=Y(0123456789ABCDEF) + + pcmpeqw xmm6,xmm6 + psrlw xmm6,BYTE_BIT ; xmm6={0xFF 0x00 0xFF 0x00 ..} + pand xmm6,xmm7 ; xmm6=Y(02468ACE)=YE + psrlw xmm7,BYTE_BIT ; xmm7=Y(13579BDF)=YO + + movdqa xmm1,xmm0 ; xmm1=xmm0=(R-Y)(L/H) + movdqa xmm3,xmm2 ; xmm3=xmm2=(G-Y)(L/H) + movdqa xmm5,xmm4 ; xmm5=xmm4=(B-Y)(L/H) + + paddw xmm0,xmm6 ; xmm0=((R-Y)+YE)=RE=R(02468ACE) + paddw xmm1,xmm7 ; xmm1=((R-Y)+YO)=RO=R(13579BDF) + packuswb xmm0,xmm0 ; xmm0=R(02468ACE********) + packuswb xmm1,xmm1 ; xmm1=R(13579BDF********) + + paddw xmm2,xmm6 ; xmm2=((G-Y)+YE)=GE=G(02468ACE) + paddw xmm3,xmm7 ; xmm3=((G-Y)+YO)=GO=G(13579BDF) + packuswb xmm2,xmm2 ; xmm2=G(02468ACE********) + packuswb xmm3,xmm3 ; xmm3=G(13579BDF********) + + paddw xmm4,xmm6 ; xmm4=((B-Y)+YE)=BE=B(02468ACE) + paddw xmm5,xmm7 ; xmm5=((B-Y)+YO)=BO=B(13579BDF) + packuswb xmm4,xmm4 ; xmm4=B(02468ACE********) + packuswb xmm5,xmm5 ; xmm5=B(13579BDF********) + +%if RGB_PIXELSIZE == 3 ; --------------- + + ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **) + ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **) + ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **) + ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **) + + punpcklbw xmmA,xmmC ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E) + punpcklbw xmmE,xmmB ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F) + punpcklbw xmmD,xmmF ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F) + + movdqa xmmG,xmmA + movdqa xmmH,xmmA + punpcklwd xmmA,xmmE ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07) + punpckhwd xmmG,xmmE ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F) + + psrldq xmmH,2 ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --) + psrldq xmmE,2 ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --) + + movdqa xmmC,xmmD + movdqa xmmB,xmmD + punpcklwd xmmD,xmmH ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18) + punpckhwd xmmC,xmmH ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --) + + psrldq xmmB,2 ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --) + + movdqa xmmF,xmmE + punpcklwd xmmE,xmmB ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29) + punpckhwd xmmF,xmmB ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --) + + pshufd xmmH,xmmA,0x4E; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03) + movdqa xmmB,xmmE + punpckldq xmmA,xmmD ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14) + punpckldq xmmE,xmmH ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07) + punpckhdq xmmD,xmmB ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29) + + pshufd xmmH,xmmG,0x4E; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B) + movdqa xmmB,xmmF + punpckldq xmmG,xmmC ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C) + punpckldq xmmF,xmmH ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F) + punpckhdq xmmC,xmmB ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --) + + punpcklqdq xmmA,xmmE ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05) + punpcklqdq xmmD,xmmG ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A) + punpcklqdq xmmF,xmmC ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F) + + cmp ecx, byte SIZEOF_XMMWORD + jb short .column_st32 + + test edi, SIZEOF_XMMWORD-1 + jnz short .out1 + ; --(aligned)------------------- + movntdq XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + movntdq XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD + movntdq XMMWORD [edi+2*SIZEOF_XMMWORD], xmmF + jmp short .out0 +.out1: ; --(unaligned)----------------- + movdqu XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD + movdqu XMMWORD [edi+2*SIZEOF_XMMWORD], xmmF +.out0: + add edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; outptr + sub ecx, byte SIZEOF_XMMWORD + jz near .endcolumn + + add esi, byte SIZEOF_XMMWORD ; inptr0 + dec al ; Yctr + jnz near .Yloop_2nd + + add ebx, byte SIZEOF_XMMWORD ; inptr1 + add edx, byte SIZEOF_XMMWORD ; inptr2 + jmp near .columnloop + alignx 16,7 + +.column_st32: + lea ecx, [ecx+ecx*2] ; imul ecx, RGB_PIXELSIZE + cmp ecx, byte 2*SIZEOF_XMMWORD + jb short .column_st16 + movdqu XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD + add edi, byte 2*SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmF + sub ecx, byte 2*SIZEOF_XMMWORD + jmp short .column_st15 +.column_st16: + cmp ecx, byte SIZEOF_XMMWORD + jb short .column_st15 + movdqu XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + add edi, byte SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmD + sub ecx, byte SIZEOF_XMMWORD +.column_st15: + ; Store the lower 8 bytes of xmmA to the output when it has enough + ; space. + cmp ecx, byte SIZEOF_MMWORD + jb short .column_st7 + movq XMM_MMWORD [edi], xmmA + add edi, byte SIZEOF_MMWORD + sub ecx, byte SIZEOF_MMWORD + psrldq xmmA, SIZEOF_MMWORD +.column_st7: + ; Store the lower 4 bytes of xmmA to the output when it has enough + ; space. + cmp ecx, byte SIZEOF_DWORD + jb short .column_st3 + movd XMM_DWORD [edi], xmmA + add edi, byte SIZEOF_DWORD + sub ecx, byte SIZEOF_DWORD + psrldq xmmA, SIZEOF_DWORD +.column_st3: + ; Store the lower 2 bytes of eax to the output when it has enough + ; space. + movd eax, xmmA + cmp ecx, byte SIZEOF_WORD + jb short .column_st1 + mov WORD [edi], ax + add edi, byte SIZEOF_WORD + sub ecx, byte SIZEOF_WORD + shr eax, 16 +.column_st1: + ; Store the lower 1 byte of eax to the output when it has enough + ; space. + test ecx, ecx + jz short .endcolumn + mov BYTE [edi], al + +%else ; RGB_PIXELSIZE == 4 ; ----------- + +%ifdef RGBX_FILLER_0XFF + pcmpeqb xmm6,xmm6 ; xmm6=XE=X(02468ACE********) + pcmpeqb xmm7,xmm7 ; xmm7=XO=X(13579BDF********) +%else + pxor xmm6,xmm6 ; xmm6=XE=X(02468ACE********) + pxor xmm7,xmm7 ; xmm7=XO=X(13579BDF********) +%endif + ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **) + ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **) + ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **) + ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **) + + punpcklbw xmmA,xmmC ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E) + punpcklbw xmmE,xmmG ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E) + punpcklbw xmmB,xmmD ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F) + punpcklbw xmmF,xmmH ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F) + + movdqa xmmC,xmmA + punpcklwd xmmA,xmmE ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36) + punpckhwd xmmC,xmmE ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E) + movdqa xmmG,xmmB + punpcklwd xmmB,xmmF ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37) + punpckhwd xmmG,xmmF ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F) + + movdqa xmmD,xmmA + punpckldq xmmA,xmmB ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33) + punpckhdq xmmD,xmmB ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37) + movdqa xmmH,xmmC + punpckldq xmmC,xmmG ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B) + punpckhdq xmmH,xmmG ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F) + + cmp ecx, byte SIZEOF_XMMWORD + jb short .column_st32 + + test edi, SIZEOF_XMMWORD-1 + jnz short .out1 + ; --(aligned)------------------- + movntdq XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + movntdq XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD + movntdq XMMWORD [edi+2*SIZEOF_XMMWORD], xmmC + movntdq XMMWORD [edi+3*SIZEOF_XMMWORD], xmmH + jmp short .out0 +.out1: ; --(unaligned)----------------- + movdqu XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD + movdqu XMMWORD [edi+2*SIZEOF_XMMWORD], xmmC + movdqu XMMWORD [edi+3*SIZEOF_XMMWORD], xmmH +.out0: + add edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD ; outptr + sub ecx, byte SIZEOF_XMMWORD + jz near .endcolumn + + add esi, byte SIZEOF_XMMWORD ; inptr0 + dec al ; Yctr + jnz near .Yloop_2nd + + add ebx, byte SIZEOF_XMMWORD ; inptr1 + add edx, byte SIZEOF_XMMWORD ; inptr2 + jmp near .columnloop + alignx 16,7 + +.column_st32: + cmp ecx, byte SIZEOF_XMMWORD/2 + jb short .column_st16 + movdqu XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + movdqu XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD + add edi, byte 2*SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmC + movdqa xmmD,xmmH + sub ecx, byte SIZEOF_XMMWORD/2 +.column_st16: + cmp ecx, byte SIZEOF_XMMWORD/4 + jb short .column_st15 + movdqu XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA + add edi, byte SIZEOF_XMMWORD ; outptr + movdqa xmmA,xmmD + sub ecx, byte SIZEOF_XMMWORD/4 +.column_st15: + ; Store two pixels (8 bytes) of xmmA to the output when it has enough + ; space. + cmp ecx, byte SIZEOF_XMMWORD/8 + jb short .column_st7 + movq XMM_MMWORD [edi], xmmA + add edi, byte SIZEOF_XMMWORD/8*4 + sub ecx, byte SIZEOF_XMMWORD/8 + psrldq xmmA, SIZEOF_XMMWORD/8*4 +.column_st7: + ; Store one pixel (4 bytes) of xmmA to the output when it has enough + ; space. + test ecx, ecx + jz short .endcolumn + movd XMM_DWORD [edi], xmmA + +%endif ; RGB_PIXELSIZE ; --------------- + +.endcolumn: + sfence ; flush the write buffer + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; -------------------------------------------------------------------------- +; +; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical. +; +; GLOBAL(void) +; jsimd_h2v2_merged_upsample_sse2 (JDIMENSION output_width, +; JSAMPIMAGE input_buf, +; JDIMENSION in_row_group_ctr, +; JSAMPARRAY output_buf); +; + +%define output_width(b) (b)+8 ; JDIMENSION output_width +%define input_buf(b) (b)+12 ; JSAMPIMAGE input_buf +%define in_row_group_ctr(b) (b)+16 ; JDIMENSION in_row_group_ctr +%define output_buf(b) (b)+20 ; JSAMPARRAY output_buf + + align 16 + global EXTN(jsimd_h2v2_merged_upsample_sse2) + +EXTN(jsimd_h2v2_merged_upsample_sse2): + push ebp + mov ebp,esp + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + mov eax, POINTER [output_width(ebp)] + + mov edi, JSAMPIMAGE [input_buf(ebp)] + mov ecx, JDIMENSION [in_row_group_ctr(ebp)] + mov esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY] + mov ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY] + mov edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY] + mov edi, JSAMPARRAY [output_buf(ebp)] + lea esi, [esi+ecx*SIZEOF_JSAMPROW] + + push edx ; inptr2 + push ebx ; inptr1 + push esi ; inptr00 + mov ebx,esp + + push edi ; output_buf (outptr0) + push ecx ; in_row_group_ctr + push ebx ; input_buf + push eax ; output_width + + call near EXTN(jsimd_h2v1_merged_upsample_sse2) + + add esi, byte SIZEOF_JSAMPROW ; inptr01 + add edi, byte SIZEOF_JSAMPROW ; outptr1 + mov POINTER [ebx+0*SIZEOF_POINTER], esi + mov POINTER [ebx-1*SIZEOF_POINTER], edi + + call near EXTN(jsimd_h2v1_merged_upsample_sse2) + + add esp, byte 7*SIZEOF_DWORD + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdsample-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jdsample-mmx.asm new file mode 100644 index 000000000..88564e439 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdsample-mmx.asm @@ -0,0 +1,737 @@ +; +; jdsample.asm - upsampling (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_fancy_upsample_mmx) + +EXTN(jconst_fancy_upsample_mmx): + +PW_ONE times 4 dw 1 +PW_TWO times 4 dw 2 +PW_THREE times 4 dw 3 +PW_SEVEN times 4 dw 7 +PW_EIGHT times 4 dw 8 + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical. +; +; The upsampling algorithm is linear interpolation between pixel centers, +; also known as a "triangle filter". This is a good compromise between +; speed and visual quality. The centers of the output pixels are 1/4 and 3/4 +; of the way between input pixel centers. +; +; GLOBAL(void) +; jsimd_h2v1_fancy_upsample_mmx (int max_v_samp_factor, +; JDIMENSION downsampled_width, +; JSAMPARRAY input_data, +; JSAMPARRAY * output_data_ptr); +; + +%define max_v_samp(b) (b)+8 ; int max_v_samp_factor +%define downsamp_width(b) (b)+12 ; JDIMENSION downsampled_width +%define input_data(b) (b)+16 ; JSAMPARRAY input_data +%define output_data_ptr(b) (b)+20 ; JSAMPARRAY * output_data_ptr + + align 16 + global EXTN(jsimd_h2v1_fancy_upsample_mmx) + +EXTN(jsimd_h2v1_fancy_upsample_mmx): + push ebp + mov ebp,esp + pushpic ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + + mov eax, JDIMENSION [downsamp_width(ebp)] ; colctr + test eax,eax + jz near .return + + mov ecx, INT [max_v_samp(ebp)] ; rowctr + test ecx,ecx + jz near .return + + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + mov edi, POINTER [output_data_ptr(ebp)] + mov edi, JSAMPARRAY [edi] ; output_data + alignx 16,7 +.rowloop: + push eax ; colctr + push edi + push esi + + mov esi, JSAMPROW [esi] ; inptr + mov edi, JSAMPROW [edi] ; outptr + + test eax, SIZEOF_MMWORD-1 + jz short .skip + mov dl, JSAMPLE [esi+(eax-1)*SIZEOF_JSAMPLE] + mov JSAMPLE [esi+eax*SIZEOF_JSAMPLE], dl ; insert a dummy sample +.skip: + pxor mm0,mm0 ; mm0=(all 0's) + pcmpeqb mm7,mm7 + psrlq mm7,(SIZEOF_MMWORD-1)*BYTE_BIT + pand mm7, MMWORD [esi+0*SIZEOF_MMWORD] + + add eax, byte SIZEOF_MMWORD-1 + and eax, byte -SIZEOF_MMWORD + cmp eax, byte SIZEOF_MMWORD + ja short .columnloop + alignx 16,7 + +.columnloop_last: + pcmpeqb mm6,mm6 + psllq mm6,(SIZEOF_MMWORD-1)*BYTE_BIT + pand mm6, MMWORD [esi+0*SIZEOF_MMWORD] + jmp short .upsample + alignx 16,7 + +.columnloop: + movq mm6, MMWORD [esi+1*SIZEOF_MMWORD] + psllq mm6,(SIZEOF_MMWORD-1)*BYTE_BIT + +.upsample: + movq mm1, MMWORD [esi+0*SIZEOF_MMWORD] + movq mm2,mm1 + movq mm3,mm1 ; mm1=( 0 1 2 3 4 5 6 7) + psllq mm2,BYTE_BIT ; mm2=( - 0 1 2 3 4 5 6) + psrlq mm3,BYTE_BIT ; mm3=( 1 2 3 4 5 6 7 -) + + por mm2,mm7 ; mm2=(-1 0 1 2 3 4 5 6) + por mm3,mm6 ; mm3=( 1 2 3 4 5 6 7 8) + + movq mm7,mm1 + psrlq mm7,(SIZEOF_MMWORD-1)*BYTE_BIT ; mm7=( 7 - - - - - - -) + + movq mm4,mm1 + punpcklbw mm1,mm0 ; mm1=( 0 1 2 3) + punpckhbw mm4,mm0 ; mm4=( 4 5 6 7) + movq mm5,mm2 + punpcklbw mm2,mm0 ; mm2=(-1 0 1 2) + punpckhbw mm5,mm0 ; mm5=( 3 4 5 6) + movq mm6,mm3 + punpcklbw mm3,mm0 ; mm3=( 1 2 3 4) + punpckhbw mm6,mm0 ; mm6=( 5 6 7 8) + + pmullw mm1,[GOTOFF(ebx,PW_THREE)] + pmullw mm4,[GOTOFF(ebx,PW_THREE)] + paddw mm2,[GOTOFF(ebx,PW_ONE)] + paddw mm5,[GOTOFF(ebx,PW_ONE)] + paddw mm3,[GOTOFF(ebx,PW_TWO)] + paddw mm6,[GOTOFF(ebx,PW_TWO)] + + paddw mm2,mm1 + paddw mm5,mm4 + psrlw mm2,2 ; mm2=OutLE=( 0 2 4 6) + psrlw mm5,2 ; mm5=OutHE=( 8 10 12 14) + paddw mm3,mm1 + paddw mm6,mm4 + psrlw mm3,2 ; mm3=OutLO=( 1 3 5 7) + psrlw mm6,2 ; mm6=OutHO=( 9 11 13 15) + + psllw mm3,BYTE_BIT + psllw mm6,BYTE_BIT + por mm2,mm3 ; mm2=OutL=( 0 1 2 3 4 5 6 7) + por mm5,mm6 ; mm5=OutH=( 8 9 10 11 12 13 14 15) + + movq MMWORD [edi+0*SIZEOF_MMWORD], mm2 + movq MMWORD [edi+1*SIZEOF_MMWORD], mm5 + + sub eax, byte SIZEOF_MMWORD + add esi, byte 1*SIZEOF_MMWORD ; inptr + add edi, byte 2*SIZEOF_MMWORD ; outptr + cmp eax, byte SIZEOF_MMWORD + ja near .columnloop + test eax,eax + jnz near .columnloop_last + + pop esi + pop edi + pop eax + + add esi, byte SIZEOF_JSAMPROW ; input_data + add edi, byte SIZEOF_JSAMPROW ; output_data + dec ecx ; rowctr + jg near .rowloop + + emms ; empty MMX state + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + poppic ebx + pop ebp + ret + +; -------------------------------------------------------------------------- +; +; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical. +; Again a triangle filter; see comments for h2v1 case, above. +; +; GLOBAL(void) +; jsimd_h2v2_fancy_upsample_mmx (int max_v_samp_factor, +; JDIMENSION downsampled_width, +; JSAMPARRAY input_data, +; JSAMPARRAY * output_data_ptr); +; + +%define max_v_samp(b) (b)+8 ; int max_v_samp_factor +%define downsamp_width(b) (b)+12 ; JDIMENSION downsampled_width +%define input_data(b) (b)+16 ; JSAMPARRAY input_data +%define output_data_ptr(b) (b)+20 ; JSAMPARRAY * output_data_ptr + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_MMWORD ; mmword wk[WK_NUM] +%define WK_NUM 4 +%define gotptr wk(0)-SIZEOF_POINTER ; void * gotptr + + align 16 + global EXTN(jsimd_h2v2_fancy_upsample_mmx) + +EXTN(jsimd_h2v2_fancy_upsample_mmx): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_MMWORD) ; align to 64 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic eax ; make a room for GOT address + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + movpic POINTER [gotptr], ebx ; save GOT address + + mov edx,eax ; edx = original ebp + mov eax, JDIMENSION [downsamp_width(edx)] ; colctr + test eax,eax + jz near .return + + mov ecx, INT [max_v_samp(edx)] ; rowctr + test ecx,ecx + jz near .return + + mov esi, JSAMPARRAY [input_data(edx)] ; input_data + mov edi, POINTER [output_data_ptr(edx)] + mov edi, JSAMPARRAY [edi] ; output_data + alignx 16,7 +.rowloop: + push eax ; colctr + push ecx + push edi + push esi + + mov ecx, JSAMPROW [esi-1*SIZEOF_JSAMPROW] ; inptr1(above) + mov ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW] ; inptr0 + mov esi, JSAMPROW [esi+1*SIZEOF_JSAMPROW] ; inptr1(below) + mov edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] ; outptr0 + mov edi, JSAMPROW [edi+1*SIZEOF_JSAMPROW] ; outptr1 + + test eax, SIZEOF_MMWORD-1 + jz short .skip + push edx + mov dl, JSAMPLE [ecx+(eax-1)*SIZEOF_JSAMPLE] + mov JSAMPLE [ecx+eax*SIZEOF_JSAMPLE], dl + mov dl, JSAMPLE [ebx+(eax-1)*SIZEOF_JSAMPLE] + mov JSAMPLE [ebx+eax*SIZEOF_JSAMPLE], dl + mov dl, JSAMPLE [esi+(eax-1)*SIZEOF_JSAMPLE] + mov JSAMPLE [esi+eax*SIZEOF_JSAMPLE], dl ; insert a dummy sample + pop edx +.skip: + ; -- process the first column block + + movq mm0, MMWORD [ebx+0*SIZEOF_MMWORD] ; mm0=row[ 0][0] + movq mm1, MMWORD [ecx+0*SIZEOF_MMWORD] ; mm1=row[-1][0] + movq mm2, MMWORD [esi+0*SIZEOF_MMWORD] ; mm2=row[+1][0] + + pushpic ebx + movpic ebx, POINTER [gotptr] ; load GOT address + + pxor mm3,mm3 ; mm3=(all 0's) + movq mm4,mm0 + punpcklbw mm0,mm3 ; mm0=row[ 0][0]( 0 1 2 3) + punpckhbw mm4,mm3 ; mm4=row[ 0][0]( 4 5 6 7) + movq mm5,mm1 + punpcklbw mm1,mm3 ; mm1=row[-1][0]( 0 1 2 3) + punpckhbw mm5,mm3 ; mm5=row[-1][0]( 4 5 6 7) + movq mm6,mm2 + punpcklbw mm2,mm3 ; mm2=row[+1][0]( 0 1 2 3) + punpckhbw mm6,mm3 ; mm6=row[+1][0]( 4 5 6 7) + + pmullw mm0,[GOTOFF(ebx,PW_THREE)] + pmullw mm4,[GOTOFF(ebx,PW_THREE)] + + pcmpeqb mm7,mm7 + psrlq mm7,(SIZEOF_MMWORD-2)*BYTE_BIT + + paddw mm1,mm0 ; mm1=Int0L=( 0 1 2 3) + paddw mm5,mm4 ; mm5=Int0H=( 4 5 6 7) + paddw mm2,mm0 ; mm2=Int1L=( 0 1 2 3) + paddw mm6,mm4 ; mm6=Int1H=( 4 5 6 7) + + movq MMWORD [edx+0*SIZEOF_MMWORD], mm1 ; temporarily save + movq MMWORD [edx+1*SIZEOF_MMWORD], mm5 ; the intermediate data + movq MMWORD [edi+0*SIZEOF_MMWORD], mm2 + movq MMWORD [edi+1*SIZEOF_MMWORD], mm6 + + pand mm1,mm7 ; mm1=( 0 - - -) + pand mm2,mm7 ; mm2=( 0 - - -) + + movq MMWORD [wk(0)], mm1 + movq MMWORD [wk(1)], mm2 + + poppic ebx + + add eax, byte SIZEOF_MMWORD-1 + and eax, byte -SIZEOF_MMWORD + cmp eax, byte SIZEOF_MMWORD + ja short .columnloop + alignx 16,7 + +.columnloop_last: + ; -- process the last column block + + pushpic ebx + movpic ebx, POINTER [gotptr] ; load GOT address + + pcmpeqb mm1,mm1 + psllq mm1,(SIZEOF_MMWORD-2)*BYTE_BIT + movq mm2,mm1 + + pand mm1, MMWORD [edx+1*SIZEOF_MMWORD] ; mm1=( - - - 7) + pand mm2, MMWORD [edi+1*SIZEOF_MMWORD] ; mm2=( - - - 7) + + movq MMWORD [wk(2)], mm1 + movq MMWORD [wk(3)], mm2 + + jmp short .upsample + alignx 16,7 + +.columnloop: + ; -- process the next column block + + movq mm0, MMWORD [ebx+1*SIZEOF_MMWORD] ; mm0=row[ 0][1] + movq mm1, MMWORD [ecx+1*SIZEOF_MMWORD] ; mm1=row[-1][1] + movq mm2, MMWORD [esi+1*SIZEOF_MMWORD] ; mm2=row[+1][1] + + pushpic ebx + movpic ebx, POINTER [gotptr] ; load GOT address + + pxor mm3,mm3 ; mm3=(all 0's) + movq mm4,mm0 + punpcklbw mm0,mm3 ; mm0=row[ 0][1]( 0 1 2 3) + punpckhbw mm4,mm3 ; mm4=row[ 0][1]( 4 5 6 7) + movq mm5,mm1 + punpcklbw mm1,mm3 ; mm1=row[-1][1]( 0 1 2 3) + punpckhbw mm5,mm3 ; mm5=row[-1][1]( 4 5 6 7) + movq mm6,mm2 + punpcklbw mm2,mm3 ; mm2=row[+1][1]( 0 1 2 3) + punpckhbw mm6,mm3 ; mm6=row[+1][1]( 4 5 6 7) + + pmullw mm0,[GOTOFF(ebx,PW_THREE)] + pmullw mm4,[GOTOFF(ebx,PW_THREE)] + + paddw mm1,mm0 ; mm1=Int0L=( 0 1 2 3) + paddw mm5,mm4 ; mm5=Int0H=( 4 5 6 7) + paddw mm2,mm0 ; mm2=Int1L=( 0 1 2 3) + paddw mm6,mm4 ; mm6=Int1H=( 4 5 6 7) + + movq MMWORD [edx+2*SIZEOF_MMWORD], mm1 ; temporarily save + movq MMWORD [edx+3*SIZEOF_MMWORD], mm5 ; the intermediate data + movq MMWORD [edi+2*SIZEOF_MMWORD], mm2 + movq MMWORD [edi+3*SIZEOF_MMWORD], mm6 + + psllq mm1,(SIZEOF_MMWORD-2)*BYTE_BIT ; mm1=( - - - 0) + psllq mm2,(SIZEOF_MMWORD-2)*BYTE_BIT ; mm2=( - - - 0) + + movq MMWORD [wk(2)], mm1 + movq MMWORD [wk(3)], mm2 + +.upsample: + ; -- process the upper row + + movq mm7, MMWORD [edx+0*SIZEOF_MMWORD] ; mm7=Int0L=( 0 1 2 3) + movq mm3, MMWORD [edx+1*SIZEOF_MMWORD] ; mm3=Int0H=( 4 5 6 7) + + movq mm0,mm7 + movq mm4,mm3 + psrlq mm0,2*BYTE_BIT ; mm0=( 1 2 3 -) + psllq mm4,(SIZEOF_MMWORD-2)*BYTE_BIT ; mm4=( - - - 4) + movq mm5,mm7 + movq mm6,mm3 + psrlq mm5,(SIZEOF_MMWORD-2)*BYTE_BIT ; mm5=( 3 - - -) + psllq mm6,2*BYTE_BIT ; mm6=( - 4 5 6) + + por mm0,mm4 ; mm0=( 1 2 3 4) + por mm5,mm6 ; mm5=( 3 4 5 6) + + movq mm1,mm7 + movq mm2,mm3 + psllq mm1,2*BYTE_BIT ; mm1=( - 0 1 2) + psrlq mm2,2*BYTE_BIT ; mm2=( 5 6 7 -) + movq mm4,mm3 + psrlq mm4,(SIZEOF_MMWORD-2)*BYTE_BIT ; mm4=( 7 - - -) + + por mm1, MMWORD [wk(0)] ; mm1=(-1 0 1 2) + por mm2, MMWORD [wk(2)] ; mm2=( 5 6 7 8) + + movq MMWORD [wk(0)], mm4 + + pmullw mm7,[GOTOFF(ebx,PW_THREE)] + pmullw mm3,[GOTOFF(ebx,PW_THREE)] + paddw mm1,[GOTOFF(ebx,PW_EIGHT)] + paddw mm5,[GOTOFF(ebx,PW_EIGHT)] + paddw mm0,[GOTOFF(ebx,PW_SEVEN)] + paddw mm2,[GOTOFF(ebx,PW_SEVEN)] + + paddw mm1,mm7 + paddw mm5,mm3 + psrlw mm1,4 ; mm1=Out0LE=( 0 2 4 6) + psrlw mm5,4 ; mm5=Out0HE=( 8 10 12 14) + paddw mm0,mm7 + paddw mm2,mm3 + psrlw mm0,4 ; mm0=Out0LO=( 1 3 5 7) + psrlw mm2,4 ; mm2=Out0HO=( 9 11 13 15) + + psllw mm0,BYTE_BIT + psllw mm2,BYTE_BIT + por mm1,mm0 ; mm1=Out0L=( 0 1 2 3 4 5 6 7) + por mm5,mm2 ; mm5=Out0H=( 8 9 10 11 12 13 14 15) + + movq MMWORD [edx+0*SIZEOF_MMWORD], mm1 + movq MMWORD [edx+1*SIZEOF_MMWORD], mm5 + + ; -- process the lower row + + movq mm6, MMWORD [edi+0*SIZEOF_MMWORD] ; mm6=Int1L=( 0 1 2 3) + movq mm4, MMWORD [edi+1*SIZEOF_MMWORD] ; mm4=Int1H=( 4 5 6 7) + + movq mm7,mm6 + movq mm3,mm4 + psrlq mm7,2*BYTE_BIT ; mm7=( 1 2 3 -) + psllq mm3,(SIZEOF_MMWORD-2)*BYTE_BIT ; mm3=( - - - 4) + movq mm0,mm6 + movq mm2,mm4 + psrlq mm0,(SIZEOF_MMWORD-2)*BYTE_BIT ; mm0=( 3 - - -) + psllq mm2,2*BYTE_BIT ; mm2=( - 4 5 6) + + por mm7,mm3 ; mm7=( 1 2 3 4) + por mm0,mm2 ; mm0=( 3 4 5 6) + + movq mm1,mm6 + movq mm5,mm4 + psllq mm1,2*BYTE_BIT ; mm1=( - 0 1 2) + psrlq mm5,2*BYTE_BIT ; mm5=( 5 6 7 -) + movq mm3,mm4 + psrlq mm3,(SIZEOF_MMWORD-2)*BYTE_BIT ; mm3=( 7 - - -) + + por mm1, MMWORD [wk(1)] ; mm1=(-1 0 1 2) + por mm5, MMWORD [wk(3)] ; mm5=( 5 6 7 8) + + movq MMWORD [wk(1)], mm3 + + pmullw mm6,[GOTOFF(ebx,PW_THREE)] + pmullw mm4,[GOTOFF(ebx,PW_THREE)] + paddw mm1,[GOTOFF(ebx,PW_EIGHT)] + paddw mm0,[GOTOFF(ebx,PW_EIGHT)] + paddw mm7,[GOTOFF(ebx,PW_SEVEN)] + paddw mm5,[GOTOFF(ebx,PW_SEVEN)] + + paddw mm1,mm6 + paddw mm0,mm4 + psrlw mm1,4 ; mm1=Out1LE=( 0 2 4 6) + psrlw mm0,4 ; mm0=Out1HE=( 8 10 12 14) + paddw mm7,mm6 + paddw mm5,mm4 + psrlw mm7,4 ; mm7=Out1LO=( 1 3 5 7) + psrlw mm5,4 ; mm5=Out1HO=( 9 11 13 15) + + psllw mm7,BYTE_BIT + psllw mm5,BYTE_BIT + por mm1,mm7 ; mm1=Out1L=( 0 1 2 3 4 5 6 7) + por mm0,mm5 ; mm0=Out1H=( 8 9 10 11 12 13 14 15) + + movq MMWORD [edi+0*SIZEOF_MMWORD], mm1 + movq MMWORD [edi+1*SIZEOF_MMWORD], mm0 + + poppic ebx + + sub eax, byte SIZEOF_MMWORD + add ecx, byte 1*SIZEOF_MMWORD ; inptr1(above) + add ebx, byte 1*SIZEOF_MMWORD ; inptr0 + add esi, byte 1*SIZEOF_MMWORD ; inptr1(below) + add edx, byte 2*SIZEOF_MMWORD ; outptr0 + add edi, byte 2*SIZEOF_MMWORD ; outptr1 + cmp eax, byte SIZEOF_MMWORD + ja near .columnloop + test eax,eax + jnz near .columnloop_last + + pop esi + pop edi + pop ecx + pop eax + + add esi, byte 1*SIZEOF_JSAMPROW ; input_data + add edi, byte 2*SIZEOF_JSAMPROW ; output_data + sub ecx, byte 2 ; rowctr + jg near .rowloop + + emms ; empty MMX state + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; -------------------------------------------------------------------------- +; +; Fast processing for the common case of 2:1 horizontal and 1:1 vertical. +; It's still a box filter. +; +; GLOBAL(void) +; jsimd_h2v1_upsample_mmx (int max_v_samp_factor, +; JDIMENSION output_width, +; JSAMPARRAY input_data, +; JSAMPARRAY * output_data_ptr); +; + +%define max_v_samp(b) (b)+8 ; int max_v_samp_factor +%define output_width(b) (b)+12 ; JDIMENSION output_width +%define input_data(b) (b)+16 ; JSAMPARRAY input_data +%define output_data_ptr(b) (b)+20 ; JSAMPARRAY * output_data_ptr + + align 16 + global EXTN(jsimd_h2v1_upsample_mmx) + +EXTN(jsimd_h2v1_upsample_mmx): + push ebp + mov ebp,esp +; push ebx ; unused +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + mov edx, JDIMENSION [output_width(ebp)] + add edx, byte (2*SIZEOF_MMWORD)-1 + and edx, byte -(2*SIZEOF_MMWORD) + jz short .return + + mov ecx, INT [max_v_samp(ebp)] ; rowctr + test ecx,ecx + jz short .return + + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + mov edi, POINTER [output_data_ptr(ebp)] + mov edi, JSAMPARRAY [edi] ; output_data + alignx 16,7 +.rowloop: + push edi + push esi + + mov esi, JSAMPROW [esi] ; inptr + mov edi, JSAMPROW [edi] ; outptr + mov eax,edx ; colctr + alignx 16,7 +.columnloop: + + movq mm0, MMWORD [esi+0*SIZEOF_MMWORD] + + movq mm1,mm0 + punpcklbw mm0,mm0 + punpckhbw mm1,mm1 + + movq MMWORD [edi+0*SIZEOF_MMWORD], mm0 + movq MMWORD [edi+1*SIZEOF_MMWORD], mm1 + + sub eax, byte 2*SIZEOF_MMWORD + jz short .nextrow + + movq mm2, MMWORD [esi+1*SIZEOF_MMWORD] + + movq mm3,mm2 + punpcklbw mm2,mm2 + punpckhbw mm3,mm3 + + movq MMWORD [edi+2*SIZEOF_MMWORD], mm2 + movq MMWORD [edi+3*SIZEOF_MMWORD], mm3 + + sub eax, byte 2*SIZEOF_MMWORD + jz short .nextrow + + add esi, byte 2*SIZEOF_MMWORD ; inptr + add edi, byte 4*SIZEOF_MMWORD ; outptr + jmp short .columnloop + alignx 16,7 + +.nextrow: + pop esi + pop edi + + add esi, byte SIZEOF_JSAMPROW ; input_data + add edi, byte SIZEOF_JSAMPROW ; output_data + dec ecx ; rowctr + jg short .rowloop + + emms ; empty MMX state + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved +; pop ebx ; unused + pop ebp + ret + +; -------------------------------------------------------------------------- +; +; Fast processing for the common case of 2:1 horizontal and 2:1 vertical. +; It's still a box filter. +; +; GLOBAL(void) +; jsimd_h2v2_upsample_mmx (int max_v_samp_factor, +; JDIMENSION output_width, +; JSAMPARRAY input_data, +; JSAMPARRAY * output_data_ptr); +; + +%define max_v_samp(b) (b)+8 ; int max_v_samp_factor +%define output_width(b) (b)+12 ; JDIMENSION output_width +%define input_data(b) (b)+16 ; JSAMPARRAY input_data +%define output_data_ptr(b) (b)+20 ; JSAMPARRAY * output_data_ptr + + align 16 + global EXTN(jsimd_h2v2_upsample_mmx) + +EXTN(jsimd_h2v2_upsample_mmx): + push ebp + mov ebp,esp + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + mov edx, JDIMENSION [output_width(ebp)] + add edx, byte (2*SIZEOF_MMWORD)-1 + and edx, byte -(2*SIZEOF_MMWORD) + jz near .return + + mov ecx, INT [max_v_samp(ebp)] ; rowctr + test ecx,ecx + jz short .return + + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + mov edi, POINTER [output_data_ptr(ebp)] + mov edi, JSAMPARRAY [edi] ; output_data + alignx 16,7 +.rowloop: + push edi + push esi + + mov esi, JSAMPROW [esi] ; inptr + mov ebx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] ; outptr0 + mov edi, JSAMPROW [edi+1*SIZEOF_JSAMPROW] ; outptr1 + mov eax,edx ; colctr + alignx 16,7 +.columnloop: + + movq mm0, MMWORD [esi+0*SIZEOF_MMWORD] + + movq mm1,mm0 + punpcklbw mm0,mm0 + punpckhbw mm1,mm1 + + movq MMWORD [ebx+0*SIZEOF_MMWORD], mm0 + movq MMWORD [ebx+1*SIZEOF_MMWORD], mm1 + movq MMWORD [edi+0*SIZEOF_MMWORD], mm0 + movq MMWORD [edi+1*SIZEOF_MMWORD], mm1 + + sub eax, byte 2*SIZEOF_MMWORD + jz short .nextrow + + movq mm2, MMWORD [esi+1*SIZEOF_MMWORD] + + movq mm3,mm2 + punpcklbw mm2,mm2 + punpckhbw mm3,mm3 + + movq MMWORD [ebx+2*SIZEOF_MMWORD], mm2 + movq MMWORD [ebx+3*SIZEOF_MMWORD], mm3 + movq MMWORD [edi+2*SIZEOF_MMWORD], mm2 + movq MMWORD [edi+3*SIZEOF_MMWORD], mm3 + + sub eax, byte 2*SIZEOF_MMWORD + jz short .nextrow + + add esi, byte 2*SIZEOF_MMWORD ; inptr + add ebx, byte 4*SIZEOF_MMWORD ; outptr0 + add edi, byte 4*SIZEOF_MMWORD ; outptr1 + jmp short .columnloop + alignx 16,7 + +.nextrow: + pop esi + pop edi + + add esi, byte 1*SIZEOF_JSAMPROW ; input_data + add edi, byte 2*SIZEOF_JSAMPROW ; output_data + sub ecx, byte 2 ; rowctr + jg short .rowloop + + emms ; empty MMX state + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdsample-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jdsample-sse2-64.asm new file mode 100644 index 000000000..335ce2aab --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdsample-sse2-64.asm @@ -0,0 +1,671 @@ +; +; jdsample.asm - upsampling (64-bit SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_fancy_upsample_sse2) + +EXTN(jconst_fancy_upsample_sse2): + +PW_ONE times 8 dw 1 +PW_TWO times 8 dw 2 +PW_THREE times 8 dw 3 +PW_SEVEN times 8 dw 7 +PW_EIGHT times 8 dw 8 + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 +; +; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical. +; +; The upsampling algorithm is linear interpolation between pixel centers, +; also known as a "triangle filter". This is a good compromise between +; speed and visual quality. The centers of the output pixels are 1/4 and 3/4 +; of the way between input pixel centers. +; +; GLOBAL(void) +; jsimd_h2v1_fancy_upsample_sse2 (int max_v_samp_factor, +; JDIMENSION downsampled_width, +; JSAMPARRAY input_data, +; JSAMPARRAY * output_data_ptr); +; + +; r10 = int max_v_samp_factor +; r11 = JDIMENSION downsampled_width +; r12 = JSAMPARRAY input_data +; r13 = JSAMPARRAY * output_data_ptr + + align 16 + global EXTN(jsimd_h2v1_fancy_upsample_sse2) + +EXTN(jsimd_h2v1_fancy_upsample_sse2): + push rbp + mov rax,rsp + mov rbp,rsp + collect_args + + mov rax, r11 ; colctr + test rax,rax + jz near .return + + mov rcx, r10 ; rowctr + test rcx,rcx + jz near .return + + mov rsi, r12 ; input_data + mov rdi, r13 + mov rdi, JSAMPARRAY [rdi] ; output_data +.rowloop: + push rax ; colctr + push rdi + push rsi + + mov rsi, JSAMPROW [rsi] ; inptr + mov rdi, JSAMPROW [rdi] ; outptr + + test rax, SIZEOF_XMMWORD-1 + jz short .skip + mov dl, JSAMPLE [rsi+(rax-1)*SIZEOF_JSAMPLE] + mov JSAMPLE [rsi+rax*SIZEOF_JSAMPLE], dl ; insert a dummy sample +.skip: + pxor xmm0,xmm0 ; xmm0=(all 0's) + pcmpeqb xmm7,xmm7 + psrldq xmm7,(SIZEOF_XMMWORD-1) + pand xmm7, XMMWORD [rsi+0*SIZEOF_XMMWORD] + + add rax, byte SIZEOF_XMMWORD-1 + and rax, byte -SIZEOF_XMMWORD + cmp rax, byte SIZEOF_XMMWORD + ja short .columnloop + +.columnloop_last: + pcmpeqb xmm6,xmm6 + pslldq xmm6,(SIZEOF_XMMWORD-1) + pand xmm6, XMMWORD [rsi+0*SIZEOF_XMMWORD] + jmp short .upsample + +.columnloop: + movdqa xmm6, XMMWORD [rsi+1*SIZEOF_XMMWORD] + pslldq xmm6,(SIZEOF_XMMWORD-1) + +.upsample: + movdqa xmm1, XMMWORD [rsi+0*SIZEOF_XMMWORD] + movdqa xmm2,xmm1 + movdqa xmm3,xmm1 ; xmm1=( 0 1 2 ... 13 14 15) + pslldq xmm2,1 ; xmm2=(-- 0 1 ... 12 13 14) + psrldq xmm3,1 ; xmm3=( 1 2 3 ... 14 15 --) + + por xmm2,xmm7 ; xmm2=(-1 0 1 ... 12 13 14) + por xmm3,xmm6 ; xmm3=( 1 2 3 ... 14 15 16) + + movdqa xmm7,xmm1 + psrldq xmm7,(SIZEOF_XMMWORD-1) ; xmm7=(15 -- -- ... -- -- --) + + movdqa xmm4,xmm1 + punpcklbw xmm1,xmm0 ; xmm1=( 0 1 2 3 4 5 6 7) + punpckhbw xmm4,xmm0 ; xmm4=( 8 9 10 11 12 13 14 15) + movdqa xmm5,xmm2 + punpcklbw xmm2,xmm0 ; xmm2=(-1 0 1 2 3 4 5 6) + punpckhbw xmm5,xmm0 ; xmm5=( 7 8 9 10 11 12 13 14) + movdqa xmm6,xmm3 + punpcklbw xmm3,xmm0 ; xmm3=( 1 2 3 4 5 6 7 8) + punpckhbw xmm6,xmm0 ; xmm6=( 9 10 11 12 13 14 15 16) + + pmullw xmm1,[rel PW_THREE] + pmullw xmm4,[rel PW_THREE] + paddw xmm2,[rel PW_ONE] + paddw xmm5,[rel PW_ONE] + paddw xmm3,[rel PW_TWO] + paddw xmm6,[rel PW_TWO] + + paddw xmm2,xmm1 + paddw xmm5,xmm4 + psrlw xmm2,2 ; xmm2=OutLE=( 0 2 4 6 8 10 12 14) + psrlw xmm5,2 ; xmm5=OutHE=(16 18 20 22 24 26 28 30) + paddw xmm3,xmm1 + paddw xmm6,xmm4 + psrlw xmm3,2 ; xmm3=OutLO=( 1 3 5 7 9 11 13 15) + psrlw xmm6,2 ; xmm6=OutHO=(17 19 21 23 25 27 29 31) + + psllw xmm3,BYTE_BIT + psllw xmm6,BYTE_BIT + por xmm2,xmm3 ; xmm2=OutL=( 0 1 2 ... 13 14 15) + por xmm5,xmm6 ; xmm5=OutH=(16 17 18 ... 29 30 31) + + movdqa XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm2 + movdqa XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm5 + + sub rax, byte SIZEOF_XMMWORD + add rsi, byte 1*SIZEOF_XMMWORD ; inptr + add rdi, byte 2*SIZEOF_XMMWORD ; outptr + cmp rax, byte SIZEOF_XMMWORD + ja near .columnloop + test eax,eax + jnz near .columnloop_last + + pop rsi + pop rdi + pop rax + + add rsi, byte SIZEOF_JSAMPROW ; input_data + add rdi, byte SIZEOF_JSAMPROW ; output_data + dec rcx ; rowctr + jg near .rowloop + +.return: + uncollect_args + pop rbp + ret + +; -------------------------------------------------------------------------- +; +; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical. +; Again a triangle filter; see comments for h2v1 case, above. +; +; GLOBAL(void) +; jsimd_h2v2_fancy_upsample_sse2 (int max_v_samp_factor, +; JDIMENSION downsampled_width, +; JSAMPARRAY input_data, +; JSAMPARRAY * output_data_ptr); +; + +; r10 = int max_v_samp_factor +; r11 = JDIMENSION downsampled_width +; r12 = JSAMPARRAY input_data +; r13 = JSAMPARRAY * output_data_ptr + +%define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 4 + + align 16 + global EXTN(jsimd_h2v2_fancy_upsample_sse2) + +EXTN(jsimd_h2v2_fancy_upsample_sse2): + push rbp + mov rax,rsp ; rax = original rbp + sub rsp, byte 4 + and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [rsp],rax + mov rbp,rsp ; rbp = aligned rbp + lea rsp, [wk(0)] + collect_args + push rbx + + mov rax, r11 ; colctr + test rax,rax + jz near .return + + mov rcx, r10 ; rowctr + test rcx,rcx + jz near .return + + mov rsi, r12 ; input_data + mov rdi, r13 + mov rdi, JSAMPARRAY [rdi] ; output_data +.rowloop: + push rax ; colctr + push rcx + push rdi + push rsi + + mov rcx, JSAMPROW [rsi-1*SIZEOF_JSAMPROW] ; inptr1(above) + mov rbx, JSAMPROW [rsi+0*SIZEOF_JSAMPROW] ; inptr0 + mov rsi, JSAMPROW [rsi+1*SIZEOF_JSAMPROW] ; inptr1(below) + mov rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW] ; outptr0 + mov rdi, JSAMPROW [rdi+1*SIZEOF_JSAMPROW] ; outptr1 + + test rax, SIZEOF_XMMWORD-1 + jz short .skip + push rdx + mov dl, JSAMPLE [rcx+(rax-1)*SIZEOF_JSAMPLE] + mov JSAMPLE [rcx+rax*SIZEOF_JSAMPLE], dl + mov dl, JSAMPLE [rbx+(rax-1)*SIZEOF_JSAMPLE] + mov JSAMPLE [rbx+rax*SIZEOF_JSAMPLE], dl + mov dl, JSAMPLE [rsi+(rax-1)*SIZEOF_JSAMPLE] + mov JSAMPLE [rsi+rax*SIZEOF_JSAMPLE], dl ; insert a dummy sample + pop rdx +.skip: + ; -- process the first column block + + movdqa xmm0, XMMWORD [rbx+0*SIZEOF_XMMWORD] ; xmm0=row[ 0][0] + movdqa xmm1, XMMWORD [rcx+0*SIZEOF_XMMWORD] ; xmm1=row[-1][0] + movdqa xmm2, XMMWORD [rsi+0*SIZEOF_XMMWORD] ; xmm2=row[+1][0] + + pxor xmm3,xmm3 ; xmm3=(all 0's) + movdqa xmm4,xmm0 + punpcklbw xmm0,xmm3 ; xmm0=row[ 0]( 0 1 2 3 4 5 6 7) + punpckhbw xmm4,xmm3 ; xmm4=row[ 0]( 8 9 10 11 12 13 14 15) + movdqa xmm5,xmm1 + punpcklbw xmm1,xmm3 ; xmm1=row[-1]( 0 1 2 3 4 5 6 7) + punpckhbw xmm5,xmm3 ; xmm5=row[-1]( 8 9 10 11 12 13 14 15) + movdqa xmm6,xmm2 + punpcklbw xmm2,xmm3 ; xmm2=row[+1]( 0 1 2 3 4 5 6 7) + punpckhbw xmm6,xmm3 ; xmm6=row[+1]( 8 9 10 11 12 13 14 15) + + pmullw xmm0,[rel PW_THREE] + pmullw xmm4,[rel PW_THREE] + + pcmpeqb xmm7,xmm7 + psrldq xmm7,(SIZEOF_XMMWORD-2) + + paddw xmm1,xmm0 ; xmm1=Int0L=( 0 1 2 3 4 5 6 7) + paddw xmm5,xmm4 ; xmm5=Int0H=( 8 9 10 11 12 13 14 15) + paddw xmm2,xmm0 ; xmm2=Int1L=( 0 1 2 3 4 5 6 7) + paddw xmm6,xmm4 ; xmm6=Int1H=( 8 9 10 11 12 13 14 15) + + movdqa XMMWORD [rdx+0*SIZEOF_XMMWORD], xmm1 ; temporarily save + movdqa XMMWORD [rdx+1*SIZEOF_XMMWORD], xmm5 ; the intermediate data + movdqa XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm2 + movdqa XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm6 + + pand xmm1,xmm7 ; xmm1=( 0 -- -- -- -- -- -- --) + pand xmm2,xmm7 ; xmm2=( 0 -- -- -- -- -- -- --) + + movdqa XMMWORD [wk(0)], xmm1 + movdqa XMMWORD [wk(1)], xmm2 + + add rax, byte SIZEOF_XMMWORD-1 + and rax, byte -SIZEOF_XMMWORD + cmp rax, byte SIZEOF_XMMWORD + ja short .columnloop + +.columnloop_last: + ; -- process the last column block + + pcmpeqb xmm1,xmm1 + pslldq xmm1,(SIZEOF_XMMWORD-2) + movdqa xmm2,xmm1 + + pand xmm1, XMMWORD [rdx+1*SIZEOF_XMMWORD] + pand xmm2, XMMWORD [rdi+1*SIZEOF_XMMWORD] + + movdqa XMMWORD [wk(2)], xmm1 ; xmm1=(-- -- -- -- -- -- -- 15) + movdqa XMMWORD [wk(3)], xmm2 ; xmm2=(-- -- -- -- -- -- -- 15) + + jmp near .upsample + +.columnloop: + ; -- process the next column block + + movdqa xmm0, XMMWORD [rbx+1*SIZEOF_XMMWORD] ; xmm0=row[ 0][1] + movdqa xmm1, XMMWORD [rcx+1*SIZEOF_XMMWORD] ; xmm1=row[-1][1] + movdqa xmm2, XMMWORD [rsi+1*SIZEOF_XMMWORD] ; xmm2=row[+1][1] + + pxor xmm3,xmm3 ; xmm3=(all 0's) + movdqa xmm4,xmm0 + punpcklbw xmm0,xmm3 ; xmm0=row[ 0]( 0 1 2 3 4 5 6 7) + punpckhbw xmm4,xmm3 ; xmm4=row[ 0]( 8 9 10 11 12 13 14 15) + movdqa xmm5,xmm1 + punpcklbw xmm1,xmm3 ; xmm1=row[-1]( 0 1 2 3 4 5 6 7) + punpckhbw xmm5,xmm3 ; xmm5=row[-1]( 8 9 10 11 12 13 14 15) + movdqa xmm6,xmm2 + punpcklbw xmm2,xmm3 ; xmm2=row[+1]( 0 1 2 3 4 5 6 7) + punpckhbw xmm6,xmm3 ; xmm6=row[+1]( 8 9 10 11 12 13 14 15) + + pmullw xmm0,[rel PW_THREE] + pmullw xmm4,[rel PW_THREE] + + paddw xmm1,xmm0 ; xmm1=Int0L=( 0 1 2 3 4 5 6 7) + paddw xmm5,xmm4 ; xmm5=Int0H=( 8 9 10 11 12 13 14 15) + paddw xmm2,xmm0 ; xmm2=Int1L=( 0 1 2 3 4 5 6 7) + paddw xmm6,xmm4 ; xmm6=Int1H=( 8 9 10 11 12 13 14 15) + + movdqa XMMWORD [rdx+2*SIZEOF_XMMWORD], xmm1 ; temporarily save + movdqa XMMWORD [rdx+3*SIZEOF_XMMWORD], xmm5 ; the intermediate data + movdqa XMMWORD [rdi+2*SIZEOF_XMMWORD], xmm2 + movdqa XMMWORD [rdi+3*SIZEOF_XMMWORD], xmm6 + + pslldq xmm1,(SIZEOF_XMMWORD-2) ; xmm1=(-- -- -- -- -- -- -- 0) + pslldq xmm2,(SIZEOF_XMMWORD-2) ; xmm2=(-- -- -- -- -- -- -- 0) + + movdqa XMMWORD [wk(2)], xmm1 + movdqa XMMWORD [wk(3)], xmm2 + +.upsample: + ; -- process the upper row + + movdqa xmm7, XMMWORD [rdx+0*SIZEOF_XMMWORD] + movdqa xmm3, XMMWORD [rdx+1*SIZEOF_XMMWORD] + + movdqa xmm0,xmm7 ; xmm7=Int0L=( 0 1 2 3 4 5 6 7) + movdqa xmm4,xmm3 ; xmm3=Int0H=( 8 9 10 11 12 13 14 15) + psrldq xmm0,2 ; xmm0=( 1 2 3 4 5 6 7 --) + pslldq xmm4,(SIZEOF_XMMWORD-2) ; xmm4=(-- -- -- -- -- -- -- 8) + movdqa xmm5,xmm7 + movdqa xmm6,xmm3 + psrldq xmm5,(SIZEOF_XMMWORD-2) ; xmm5=( 7 -- -- -- -- -- -- --) + pslldq xmm6,2 ; xmm6=(-- 8 9 10 11 12 13 14) + + por xmm0,xmm4 ; xmm0=( 1 2 3 4 5 6 7 8) + por xmm5,xmm6 ; xmm5=( 7 8 9 10 11 12 13 14) + + movdqa xmm1,xmm7 + movdqa xmm2,xmm3 + pslldq xmm1,2 ; xmm1=(-- 0 1 2 3 4 5 6) + psrldq xmm2,2 ; xmm2=( 9 10 11 12 13 14 15 --) + movdqa xmm4,xmm3 + psrldq xmm4,(SIZEOF_XMMWORD-2) ; xmm4=(15 -- -- -- -- -- -- --) + + por xmm1, XMMWORD [wk(0)] ; xmm1=(-1 0 1 2 3 4 5 6) + por xmm2, XMMWORD [wk(2)] ; xmm2=( 9 10 11 12 13 14 15 16) + + movdqa XMMWORD [wk(0)], xmm4 + + pmullw xmm7,[rel PW_THREE] + pmullw xmm3,[rel PW_THREE] + paddw xmm1,[rel PW_EIGHT] + paddw xmm5,[rel PW_EIGHT] + paddw xmm0,[rel PW_SEVEN] + paddw xmm2,[rel PW_SEVEN] + + paddw xmm1,xmm7 + paddw xmm5,xmm3 + psrlw xmm1,4 ; xmm1=Out0LE=( 0 2 4 6 8 10 12 14) + psrlw xmm5,4 ; xmm5=Out0HE=(16 18 20 22 24 26 28 30) + paddw xmm0,xmm7 + paddw xmm2,xmm3 + psrlw xmm0,4 ; xmm0=Out0LO=( 1 3 5 7 9 11 13 15) + psrlw xmm2,4 ; xmm2=Out0HO=(17 19 21 23 25 27 29 31) + + psllw xmm0,BYTE_BIT + psllw xmm2,BYTE_BIT + por xmm1,xmm0 ; xmm1=Out0L=( 0 1 2 ... 13 14 15) + por xmm5,xmm2 ; xmm5=Out0H=(16 17 18 ... 29 30 31) + + movdqa XMMWORD [rdx+0*SIZEOF_XMMWORD], xmm1 + movdqa XMMWORD [rdx+1*SIZEOF_XMMWORD], xmm5 + + ; -- process the lower row + + movdqa xmm6, XMMWORD [rdi+0*SIZEOF_XMMWORD] + movdqa xmm4, XMMWORD [rdi+1*SIZEOF_XMMWORD] + + movdqa xmm7,xmm6 ; xmm6=Int1L=( 0 1 2 3 4 5 6 7) + movdqa xmm3,xmm4 ; xmm4=Int1H=( 8 9 10 11 12 13 14 15) + psrldq xmm7,2 ; xmm7=( 1 2 3 4 5 6 7 --) + pslldq xmm3,(SIZEOF_XMMWORD-2) ; xmm3=(-- -- -- -- -- -- -- 8) + movdqa xmm0,xmm6 + movdqa xmm2,xmm4 + psrldq xmm0,(SIZEOF_XMMWORD-2) ; xmm0=( 7 -- -- -- -- -- -- --) + pslldq xmm2,2 ; xmm2=(-- 8 9 10 11 12 13 14) + + por xmm7,xmm3 ; xmm7=( 1 2 3 4 5 6 7 8) + por xmm0,xmm2 ; xmm0=( 7 8 9 10 11 12 13 14) + + movdqa xmm1,xmm6 + movdqa xmm5,xmm4 + pslldq xmm1,2 ; xmm1=(-- 0 1 2 3 4 5 6) + psrldq xmm5,2 ; xmm5=( 9 10 11 12 13 14 15 --) + movdqa xmm3,xmm4 + psrldq xmm3,(SIZEOF_XMMWORD-2) ; xmm3=(15 -- -- -- -- -- -- --) + + por xmm1, XMMWORD [wk(1)] ; xmm1=(-1 0 1 2 3 4 5 6) + por xmm5, XMMWORD [wk(3)] ; xmm5=( 9 10 11 12 13 14 15 16) + + movdqa XMMWORD [wk(1)], xmm3 + + pmullw xmm6,[rel PW_THREE] + pmullw xmm4,[rel PW_THREE] + paddw xmm1,[rel PW_EIGHT] + paddw xmm0,[rel PW_EIGHT] + paddw xmm7,[rel PW_SEVEN] + paddw xmm5,[rel PW_SEVEN] + + paddw xmm1,xmm6 + paddw xmm0,xmm4 + psrlw xmm1,4 ; xmm1=Out1LE=( 0 2 4 6 8 10 12 14) + psrlw xmm0,4 ; xmm0=Out1HE=(16 18 20 22 24 26 28 30) + paddw xmm7,xmm6 + paddw xmm5,xmm4 + psrlw xmm7,4 ; xmm7=Out1LO=( 1 3 5 7 9 11 13 15) + psrlw xmm5,4 ; xmm5=Out1HO=(17 19 21 23 25 27 29 31) + + psllw xmm7,BYTE_BIT + psllw xmm5,BYTE_BIT + por xmm1,xmm7 ; xmm1=Out1L=( 0 1 2 ... 13 14 15) + por xmm0,xmm5 ; xmm0=Out1H=(16 17 18 ... 29 30 31) + + movdqa XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm1 + movdqa XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm0 + + sub rax, byte SIZEOF_XMMWORD + add rcx, byte 1*SIZEOF_XMMWORD ; inptr1(above) + add rbx, byte 1*SIZEOF_XMMWORD ; inptr0 + add rsi, byte 1*SIZEOF_XMMWORD ; inptr1(below) + add rdx, byte 2*SIZEOF_XMMWORD ; outptr0 + add rdi, byte 2*SIZEOF_XMMWORD ; outptr1 + cmp rax, byte SIZEOF_XMMWORD + ja near .columnloop + test rax,rax + jnz near .columnloop_last + + pop rsi + pop rdi + pop rcx + pop rax + + add rsi, byte 1*SIZEOF_JSAMPROW ; input_data + add rdi, byte 2*SIZEOF_JSAMPROW ; output_data + sub rcx, byte 2 ; rowctr + jg near .rowloop + +.return: + pop rbx + uncollect_args + mov rsp,rbp ; rsp <- aligned rbp + pop rsp ; rsp <- original rbp + pop rbp + ret + +; -------------------------------------------------------------------------- +; +; Fast processing for the common case of 2:1 horizontal and 1:1 vertical. +; It's still a box filter. +; +; GLOBAL(void) +; jsimd_h2v1_upsample_sse2 (int max_v_samp_factor, +; JDIMENSION output_width, +; JSAMPARRAY input_data, +; JSAMPARRAY * output_data_ptr); +; + +; r10 = int max_v_samp_factor +; r11 = JDIMENSION output_width +; r12 = JSAMPARRAY input_data +; r13 = JSAMPARRAY * output_data_ptr + + align 16 + global EXTN(jsimd_h2v1_upsample_sse2) + +EXTN(jsimd_h2v1_upsample_sse2): + push rbp + mov rax,rsp + mov rbp,rsp + collect_args + + mov rdx, r11 + add rdx, byte (2*SIZEOF_XMMWORD)-1 + and rdx, byte -(2*SIZEOF_XMMWORD) + jz near .return + + mov rcx, r10 ; rowctr + test rcx,rcx + jz short .return + + mov rsi, r12 ; input_data + mov rdi, r13 + mov rdi, JSAMPARRAY [rdi] ; output_data +.rowloop: + push rdi + push rsi + + mov rsi, JSAMPROW [rsi] ; inptr + mov rdi, JSAMPROW [rdi] ; outptr + mov rax,rdx ; colctr +.columnloop: + + movdqa xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD] + + movdqa xmm1,xmm0 + punpcklbw xmm0,xmm0 + punpckhbw xmm1,xmm1 + + movdqa XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0 + movdqa XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm1 + + sub rax, byte 2*SIZEOF_XMMWORD + jz short .nextrow + + movdqa xmm2, XMMWORD [rsi+1*SIZEOF_XMMWORD] + + movdqa xmm3,xmm2 + punpcklbw xmm2,xmm2 + punpckhbw xmm3,xmm3 + + movdqa XMMWORD [rdi+2*SIZEOF_XMMWORD], xmm2 + movdqa XMMWORD [rdi+3*SIZEOF_XMMWORD], xmm3 + + sub rax, byte 2*SIZEOF_XMMWORD + jz short .nextrow + + add rsi, byte 2*SIZEOF_XMMWORD ; inptr + add rdi, byte 4*SIZEOF_XMMWORD ; outptr + jmp short .columnloop + +.nextrow: + pop rsi + pop rdi + + add rsi, byte SIZEOF_JSAMPROW ; input_data + add rdi, byte SIZEOF_JSAMPROW ; output_data + dec rcx ; rowctr + jg short .rowloop + +.return: + uncollect_args + pop rbp + ret + +; -------------------------------------------------------------------------- +; +; Fast processing for the common case of 2:1 horizontal and 2:1 vertical. +; It's still a box filter. +; +; GLOBAL(void) +; jsimd_h2v2_upsample_sse2 (nt max_v_samp_factor, +; JDIMENSION output_width, +; JSAMPARRAY input_data, +; JSAMPARRAY * output_data_ptr); +; + +; r10 = int max_v_samp_factor +; r11 = JDIMENSION output_width +; r12 = JSAMPARRAY input_data +; r13 = JSAMPARRAY * output_data_ptr + + align 16 + global EXTN(jsimd_h2v2_upsample_sse2) + +EXTN(jsimd_h2v2_upsample_sse2): + push rbp + mov rax,rsp + mov rbp,rsp + collect_args + push rbx + + mov rdx, r11 + add rdx, byte (2*SIZEOF_XMMWORD)-1 + and rdx, byte -(2*SIZEOF_XMMWORD) + jz near .return + + mov rcx, r10 ; rowctr + test rcx,rcx + jz near .return + + mov rsi, r12 ; input_data + mov rdi, r13 + mov rdi, JSAMPARRAY [rdi] ; output_data +.rowloop: + push rdi + push rsi + + mov rsi, JSAMPROW [rsi] ; inptr + mov rbx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW] ; outptr0 + mov rdi, JSAMPROW [rdi+1*SIZEOF_JSAMPROW] ; outptr1 + mov rax,rdx ; colctr +.columnloop: + + movdqa xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD] + + movdqa xmm1,xmm0 + punpcklbw xmm0,xmm0 + punpckhbw xmm1,xmm1 + + movdqa XMMWORD [rbx+0*SIZEOF_XMMWORD], xmm0 + movdqa XMMWORD [rbx+1*SIZEOF_XMMWORD], xmm1 + movdqa XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0 + movdqa XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm1 + + sub rax, byte 2*SIZEOF_XMMWORD + jz short .nextrow + + movdqa xmm2, XMMWORD [rsi+1*SIZEOF_XMMWORD] + + movdqa xmm3,xmm2 + punpcklbw xmm2,xmm2 + punpckhbw xmm3,xmm3 + + movdqa XMMWORD [rbx+2*SIZEOF_XMMWORD], xmm2 + movdqa XMMWORD [rbx+3*SIZEOF_XMMWORD], xmm3 + movdqa XMMWORD [rdi+2*SIZEOF_XMMWORD], xmm2 + movdqa XMMWORD [rdi+3*SIZEOF_XMMWORD], xmm3 + + sub rax, byte 2*SIZEOF_XMMWORD + jz short .nextrow + + add rsi, byte 2*SIZEOF_XMMWORD ; inptr + add rbx, byte 4*SIZEOF_XMMWORD ; outptr0 + add rdi, byte 4*SIZEOF_XMMWORD ; outptr1 + jmp short .columnloop + +.nextrow: + pop rsi + pop rdi + + add rsi, byte 1*SIZEOF_JSAMPROW ; input_data + add rdi, byte 2*SIZEOF_JSAMPROW ; output_data + sub rcx, byte 2 ; rowctr + jg near .rowloop + +.return: + pop rbx + uncollect_args + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jdsample-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jdsample-sse2.asm new file mode 100644 index 000000000..51176d459 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jdsample-sse2.asm @@ -0,0 +1,729 @@ +; +; jdsample.asm - upsampling (SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_fancy_upsample_sse2) + +EXTN(jconst_fancy_upsample_sse2): + +PW_ONE times 8 dw 1 +PW_TWO times 8 dw 2 +PW_THREE times 8 dw 3 +PW_SEVEN times 8 dw 7 +PW_EIGHT times 8 dw 8 + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical. +; +; The upsampling algorithm is linear interpolation between pixel centers, +; also known as a "triangle filter". This is a good compromise between +; speed and visual quality. The centers of the output pixels are 1/4 and 3/4 +; of the way between input pixel centers. +; +; GLOBAL(void) +; jsimd_h2v1_fancy_upsample_sse2 (int max_v_samp_factor, +; JDIMENSION downsampled_width, +; JSAMPARRAY input_data, +; JSAMPARRAY * output_data_ptr); +; + +%define max_v_samp(b) (b)+8 ; int max_v_samp_factor +%define downsamp_width(b) (b)+12 ; JDIMENSION downsampled_width +%define input_data(b) (b)+16 ; JSAMPARRAY input_data +%define output_data_ptr(b) (b)+20 ; JSAMPARRAY * output_data_ptr + + align 16 + global EXTN(jsimd_h2v1_fancy_upsample_sse2) + +EXTN(jsimd_h2v1_fancy_upsample_sse2): + push ebp + mov ebp,esp + pushpic ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + + mov eax, JDIMENSION [downsamp_width(ebp)] ; colctr + test eax,eax + jz near .return + + mov ecx, INT [max_v_samp(ebp)] ; rowctr + test ecx,ecx + jz near .return + + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + mov edi, POINTER [output_data_ptr(ebp)] + mov edi, JSAMPARRAY [edi] ; output_data + alignx 16,7 +.rowloop: + push eax ; colctr + push edi + push esi + + mov esi, JSAMPROW [esi] ; inptr + mov edi, JSAMPROW [edi] ; outptr + + test eax, SIZEOF_XMMWORD-1 + jz short .skip + mov dl, JSAMPLE [esi+(eax-1)*SIZEOF_JSAMPLE] + mov JSAMPLE [esi+eax*SIZEOF_JSAMPLE], dl ; insert a dummy sample +.skip: + pxor xmm0,xmm0 ; xmm0=(all 0's) + pcmpeqb xmm7,xmm7 + psrldq xmm7,(SIZEOF_XMMWORD-1) + pand xmm7, XMMWORD [esi+0*SIZEOF_XMMWORD] + + add eax, byte SIZEOF_XMMWORD-1 + and eax, byte -SIZEOF_XMMWORD + cmp eax, byte SIZEOF_XMMWORD + ja short .columnloop + alignx 16,7 + +.columnloop_last: + pcmpeqb xmm6,xmm6 + pslldq xmm6,(SIZEOF_XMMWORD-1) + pand xmm6, XMMWORD [esi+0*SIZEOF_XMMWORD] + jmp short .upsample + alignx 16,7 + +.columnloop: + movdqa xmm6, XMMWORD [esi+1*SIZEOF_XMMWORD] + pslldq xmm6,(SIZEOF_XMMWORD-1) + +.upsample: + movdqa xmm1, XMMWORD [esi+0*SIZEOF_XMMWORD] + movdqa xmm2,xmm1 + movdqa xmm3,xmm1 ; xmm1=( 0 1 2 ... 13 14 15) + pslldq xmm2,1 ; xmm2=(-- 0 1 ... 12 13 14) + psrldq xmm3,1 ; xmm3=( 1 2 3 ... 14 15 --) + + por xmm2,xmm7 ; xmm2=(-1 0 1 ... 12 13 14) + por xmm3,xmm6 ; xmm3=( 1 2 3 ... 14 15 16) + + movdqa xmm7,xmm1 + psrldq xmm7,(SIZEOF_XMMWORD-1) ; xmm7=(15 -- -- ... -- -- --) + + movdqa xmm4,xmm1 + punpcklbw xmm1,xmm0 ; xmm1=( 0 1 2 3 4 5 6 7) + punpckhbw xmm4,xmm0 ; xmm4=( 8 9 10 11 12 13 14 15) + movdqa xmm5,xmm2 + punpcklbw xmm2,xmm0 ; xmm2=(-1 0 1 2 3 4 5 6) + punpckhbw xmm5,xmm0 ; xmm5=( 7 8 9 10 11 12 13 14) + movdqa xmm6,xmm3 + punpcklbw xmm3,xmm0 ; xmm3=( 1 2 3 4 5 6 7 8) + punpckhbw xmm6,xmm0 ; xmm6=( 9 10 11 12 13 14 15 16) + + pmullw xmm1,[GOTOFF(ebx,PW_THREE)] + pmullw xmm4,[GOTOFF(ebx,PW_THREE)] + paddw xmm2,[GOTOFF(ebx,PW_ONE)] + paddw xmm5,[GOTOFF(ebx,PW_ONE)] + paddw xmm3,[GOTOFF(ebx,PW_TWO)] + paddw xmm6,[GOTOFF(ebx,PW_TWO)] + + paddw xmm2,xmm1 + paddw xmm5,xmm4 + psrlw xmm2,2 ; xmm2=OutLE=( 0 2 4 6 8 10 12 14) + psrlw xmm5,2 ; xmm5=OutHE=(16 18 20 22 24 26 28 30) + paddw xmm3,xmm1 + paddw xmm6,xmm4 + psrlw xmm3,2 ; xmm3=OutLO=( 1 3 5 7 9 11 13 15) + psrlw xmm6,2 ; xmm6=OutHO=(17 19 21 23 25 27 29 31) + + psllw xmm3,BYTE_BIT + psllw xmm6,BYTE_BIT + por xmm2,xmm3 ; xmm2=OutL=( 0 1 2 ... 13 14 15) + por xmm5,xmm6 ; xmm5=OutH=(16 17 18 ... 29 30 31) + + movdqa XMMWORD [edi+0*SIZEOF_XMMWORD], xmm2 + movdqa XMMWORD [edi+1*SIZEOF_XMMWORD], xmm5 + + sub eax, byte SIZEOF_XMMWORD + add esi, byte 1*SIZEOF_XMMWORD ; inptr + add edi, byte 2*SIZEOF_XMMWORD ; outptr + cmp eax, byte SIZEOF_XMMWORD + ja near .columnloop + test eax,eax + jnz near .columnloop_last + + pop esi + pop edi + pop eax + + add esi, byte SIZEOF_JSAMPROW ; input_data + add edi, byte SIZEOF_JSAMPROW ; output_data + dec ecx ; rowctr + jg near .rowloop + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + poppic ebx + pop ebp + ret + +; -------------------------------------------------------------------------- +; +; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical. +; Again a triangle filter; see comments for h2v1 case, above. +; +; GLOBAL(void) +; jsimd_h2v2_fancy_upsample_sse2 (int max_v_samp_factor, +; JDIMENSION downsampled_width, +; JSAMPARRAY input_data, +; JSAMPARRAY * output_data_ptr); +; + +%define max_v_samp(b) (b)+8 ; int max_v_samp_factor +%define downsamp_width(b) (b)+12 ; JDIMENSION downsampled_width +%define input_data(b) (b)+16 ; JSAMPARRAY input_data +%define output_data_ptr(b) (b)+20 ; JSAMPARRAY * output_data_ptr + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 4 +%define gotptr wk(0)-SIZEOF_POINTER ; void * gotptr + + align 16 + global EXTN(jsimd_h2v2_fancy_upsample_sse2) + +EXTN(jsimd_h2v2_fancy_upsample_sse2): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic eax ; make a room for GOT address + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + movpic POINTER [gotptr], ebx ; save GOT address + + mov edx,eax ; edx = original ebp + mov eax, JDIMENSION [downsamp_width(edx)] ; colctr + test eax,eax + jz near .return + + mov ecx, INT [max_v_samp(edx)] ; rowctr + test ecx,ecx + jz near .return + + mov esi, JSAMPARRAY [input_data(edx)] ; input_data + mov edi, POINTER [output_data_ptr(edx)] + mov edi, JSAMPARRAY [edi] ; output_data + alignx 16,7 +.rowloop: + push eax ; colctr + push ecx + push edi + push esi + + mov ecx, JSAMPROW [esi-1*SIZEOF_JSAMPROW] ; inptr1(above) + mov ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW] ; inptr0 + mov esi, JSAMPROW [esi+1*SIZEOF_JSAMPROW] ; inptr1(below) + mov edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] ; outptr0 + mov edi, JSAMPROW [edi+1*SIZEOF_JSAMPROW] ; outptr1 + + test eax, SIZEOF_XMMWORD-1 + jz short .skip + push edx + mov dl, JSAMPLE [ecx+(eax-1)*SIZEOF_JSAMPLE] + mov JSAMPLE [ecx+eax*SIZEOF_JSAMPLE], dl + mov dl, JSAMPLE [ebx+(eax-1)*SIZEOF_JSAMPLE] + mov JSAMPLE [ebx+eax*SIZEOF_JSAMPLE], dl + mov dl, JSAMPLE [esi+(eax-1)*SIZEOF_JSAMPLE] + mov JSAMPLE [esi+eax*SIZEOF_JSAMPLE], dl ; insert a dummy sample + pop edx +.skip: + ; -- process the first column block + + movdqa xmm0, XMMWORD [ebx+0*SIZEOF_XMMWORD] ; xmm0=row[ 0][0] + movdqa xmm1, XMMWORD [ecx+0*SIZEOF_XMMWORD] ; xmm1=row[-1][0] + movdqa xmm2, XMMWORD [esi+0*SIZEOF_XMMWORD] ; xmm2=row[+1][0] + + pushpic ebx + movpic ebx, POINTER [gotptr] ; load GOT address + + pxor xmm3,xmm3 ; xmm3=(all 0's) + movdqa xmm4,xmm0 + punpcklbw xmm0,xmm3 ; xmm0=row[ 0]( 0 1 2 3 4 5 6 7) + punpckhbw xmm4,xmm3 ; xmm4=row[ 0]( 8 9 10 11 12 13 14 15) + movdqa xmm5,xmm1 + punpcklbw xmm1,xmm3 ; xmm1=row[-1]( 0 1 2 3 4 5 6 7) + punpckhbw xmm5,xmm3 ; xmm5=row[-1]( 8 9 10 11 12 13 14 15) + movdqa xmm6,xmm2 + punpcklbw xmm2,xmm3 ; xmm2=row[+1]( 0 1 2 3 4 5 6 7) + punpckhbw xmm6,xmm3 ; xmm6=row[+1]( 8 9 10 11 12 13 14 15) + + pmullw xmm0,[GOTOFF(ebx,PW_THREE)] + pmullw xmm4,[GOTOFF(ebx,PW_THREE)] + + pcmpeqb xmm7,xmm7 + psrldq xmm7,(SIZEOF_XMMWORD-2) + + paddw xmm1,xmm0 ; xmm1=Int0L=( 0 1 2 3 4 5 6 7) + paddw xmm5,xmm4 ; xmm5=Int0H=( 8 9 10 11 12 13 14 15) + paddw xmm2,xmm0 ; xmm2=Int1L=( 0 1 2 3 4 5 6 7) + paddw xmm6,xmm4 ; xmm6=Int1H=( 8 9 10 11 12 13 14 15) + + movdqa XMMWORD [edx+0*SIZEOF_XMMWORD], xmm1 ; temporarily save + movdqa XMMWORD [edx+1*SIZEOF_XMMWORD], xmm5 ; the intermediate data + movdqa XMMWORD [edi+0*SIZEOF_XMMWORD], xmm2 + movdqa XMMWORD [edi+1*SIZEOF_XMMWORD], xmm6 + + pand xmm1,xmm7 ; xmm1=( 0 -- -- -- -- -- -- --) + pand xmm2,xmm7 ; xmm2=( 0 -- -- -- -- -- -- --) + + movdqa XMMWORD [wk(0)], xmm1 + movdqa XMMWORD [wk(1)], xmm2 + + poppic ebx + + add eax, byte SIZEOF_XMMWORD-1 + and eax, byte -SIZEOF_XMMWORD + cmp eax, byte SIZEOF_XMMWORD + ja short .columnloop + alignx 16,7 + +.columnloop_last: + ; -- process the last column block + + pushpic ebx + movpic ebx, POINTER [gotptr] ; load GOT address + + pcmpeqb xmm1,xmm1 + pslldq xmm1,(SIZEOF_XMMWORD-2) + movdqa xmm2,xmm1 + + pand xmm1, XMMWORD [edx+1*SIZEOF_XMMWORD] + pand xmm2, XMMWORD [edi+1*SIZEOF_XMMWORD] + + movdqa XMMWORD [wk(2)], xmm1 ; xmm1=(-- -- -- -- -- -- -- 15) + movdqa XMMWORD [wk(3)], xmm2 ; xmm2=(-- -- -- -- -- -- -- 15) + + jmp near .upsample + alignx 16,7 + +.columnloop: + ; -- process the next column block + + movdqa xmm0, XMMWORD [ebx+1*SIZEOF_XMMWORD] ; xmm0=row[ 0][1] + movdqa xmm1, XMMWORD [ecx+1*SIZEOF_XMMWORD] ; xmm1=row[-1][1] + movdqa xmm2, XMMWORD [esi+1*SIZEOF_XMMWORD] ; xmm2=row[+1][1] + + pushpic ebx + movpic ebx, POINTER [gotptr] ; load GOT address + + pxor xmm3,xmm3 ; xmm3=(all 0's) + movdqa xmm4,xmm0 + punpcklbw xmm0,xmm3 ; xmm0=row[ 0]( 0 1 2 3 4 5 6 7) + punpckhbw xmm4,xmm3 ; xmm4=row[ 0]( 8 9 10 11 12 13 14 15) + movdqa xmm5,xmm1 + punpcklbw xmm1,xmm3 ; xmm1=row[-1]( 0 1 2 3 4 5 6 7) + punpckhbw xmm5,xmm3 ; xmm5=row[-1]( 8 9 10 11 12 13 14 15) + movdqa xmm6,xmm2 + punpcklbw xmm2,xmm3 ; xmm2=row[+1]( 0 1 2 3 4 5 6 7) + punpckhbw xmm6,xmm3 ; xmm6=row[+1]( 8 9 10 11 12 13 14 15) + + pmullw xmm0,[GOTOFF(ebx,PW_THREE)] + pmullw xmm4,[GOTOFF(ebx,PW_THREE)] + + paddw xmm1,xmm0 ; xmm1=Int0L=( 0 1 2 3 4 5 6 7) + paddw xmm5,xmm4 ; xmm5=Int0H=( 8 9 10 11 12 13 14 15) + paddw xmm2,xmm0 ; xmm2=Int1L=( 0 1 2 3 4 5 6 7) + paddw xmm6,xmm4 ; xmm6=Int1H=( 8 9 10 11 12 13 14 15) + + movdqa XMMWORD [edx+2*SIZEOF_XMMWORD], xmm1 ; temporarily save + movdqa XMMWORD [edx+3*SIZEOF_XMMWORD], xmm5 ; the intermediate data + movdqa XMMWORD [edi+2*SIZEOF_XMMWORD], xmm2 + movdqa XMMWORD [edi+3*SIZEOF_XMMWORD], xmm6 + + pslldq xmm1,(SIZEOF_XMMWORD-2) ; xmm1=(-- -- -- -- -- -- -- 0) + pslldq xmm2,(SIZEOF_XMMWORD-2) ; xmm2=(-- -- -- -- -- -- -- 0) + + movdqa XMMWORD [wk(2)], xmm1 + movdqa XMMWORD [wk(3)], xmm2 + +.upsample: + ; -- process the upper row + + movdqa xmm7, XMMWORD [edx+0*SIZEOF_XMMWORD] + movdqa xmm3, XMMWORD [edx+1*SIZEOF_XMMWORD] + + movdqa xmm0,xmm7 ; xmm7=Int0L=( 0 1 2 3 4 5 6 7) + movdqa xmm4,xmm3 ; xmm3=Int0H=( 8 9 10 11 12 13 14 15) + psrldq xmm0,2 ; xmm0=( 1 2 3 4 5 6 7 --) + pslldq xmm4,(SIZEOF_XMMWORD-2) ; xmm4=(-- -- -- -- -- -- -- 8) + movdqa xmm5,xmm7 + movdqa xmm6,xmm3 + psrldq xmm5,(SIZEOF_XMMWORD-2) ; xmm5=( 7 -- -- -- -- -- -- --) + pslldq xmm6,2 ; xmm6=(-- 8 9 10 11 12 13 14) + + por xmm0,xmm4 ; xmm0=( 1 2 3 4 5 6 7 8) + por xmm5,xmm6 ; xmm5=( 7 8 9 10 11 12 13 14) + + movdqa xmm1,xmm7 + movdqa xmm2,xmm3 + pslldq xmm1,2 ; xmm1=(-- 0 1 2 3 4 5 6) + psrldq xmm2,2 ; xmm2=( 9 10 11 12 13 14 15 --) + movdqa xmm4,xmm3 + psrldq xmm4,(SIZEOF_XMMWORD-2) ; xmm4=(15 -- -- -- -- -- -- --) + + por xmm1, XMMWORD [wk(0)] ; xmm1=(-1 0 1 2 3 4 5 6) + por xmm2, XMMWORD [wk(2)] ; xmm2=( 9 10 11 12 13 14 15 16) + + movdqa XMMWORD [wk(0)], xmm4 + + pmullw xmm7,[GOTOFF(ebx,PW_THREE)] + pmullw xmm3,[GOTOFF(ebx,PW_THREE)] + paddw xmm1,[GOTOFF(ebx,PW_EIGHT)] + paddw xmm5,[GOTOFF(ebx,PW_EIGHT)] + paddw xmm0,[GOTOFF(ebx,PW_SEVEN)] + paddw xmm2,[GOTOFF(ebx,PW_SEVEN)] + + paddw xmm1,xmm7 + paddw xmm5,xmm3 + psrlw xmm1,4 ; xmm1=Out0LE=( 0 2 4 6 8 10 12 14) + psrlw xmm5,4 ; xmm5=Out0HE=(16 18 20 22 24 26 28 30) + paddw xmm0,xmm7 + paddw xmm2,xmm3 + psrlw xmm0,4 ; xmm0=Out0LO=( 1 3 5 7 9 11 13 15) + psrlw xmm2,4 ; xmm2=Out0HO=(17 19 21 23 25 27 29 31) + + psllw xmm0,BYTE_BIT + psllw xmm2,BYTE_BIT + por xmm1,xmm0 ; xmm1=Out0L=( 0 1 2 ... 13 14 15) + por xmm5,xmm2 ; xmm5=Out0H=(16 17 18 ... 29 30 31) + + movdqa XMMWORD [edx+0*SIZEOF_XMMWORD], xmm1 + movdqa XMMWORD [edx+1*SIZEOF_XMMWORD], xmm5 + + ; -- process the lower row + + movdqa xmm6, XMMWORD [edi+0*SIZEOF_XMMWORD] + movdqa xmm4, XMMWORD [edi+1*SIZEOF_XMMWORD] + + movdqa xmm7,xmm6 ; xmm6=Int1L=( 0 1 2 3 4 5 6 7) + movdqa xmm3,xmm4 ; xmm4=Int1H=( 8 9 10 11 12 13 14 15) + psrldq xmm7,2 ; xmm7=( 1 2 3 4 5 6 7 --) + pslldq xmm3,(SIZEOF_XMMWORD-2) ; xmm3=(-- -- -- -- -- -- -- 8) + movdqa xmm0,xmm6 + movdqa xmm2,xmm4 + psrldq xmm0,(SIZEOF_XMMWORD-2) ; xmm0=( 7 -- -- -- -- -- -- --) + pslldq xmm2,2 ; xmm2=(-- 8 9 10 11 12 13 14) + + por xmm7,xmm3 ; xmm7=( 1 2 3 4 5 6 7 8) + por xmm0,xmm2 ; xmm0=( 7 8 9 10 11 12 13 14) + + movdqa xmm1,xmm6 + movdqa xmm5,xmm4 + pslldq xmm1,2 ; xmm1=(-- 0 1 2 3 4 5 6) + psrldq xmm5,2 ; xmm5=( 9 10 11 12 13 14 15 --) + movdqa xmm3,xmm4 + psrldq xmm3,(SIZEOF_XMMWORD-2) ; xmm3=(15 -- -- -- -- -- -- --) + + por xmm1, XMMWORD [wk(1)] ; xmm1=(-1 0 1 2 3 4 5 6) + por xmm5, XMMWORD [wk(3)] ; xmm5=( 9 10 11 12 13 14 15 16) + + movdqa XMMWORD [wk(1)], xmm3 + + pmullw xmm6,[GOTOFF(ebx,PW_THREE)] + pmullw xmm4,[GOTOFF(ebx,PW_THREE)] + paddw xmm1,[GOTOFF(ebx,PW_EIGHT)] + paddw xmm0,[GOTOFF(ebx,PW_EIGHT)] + paddw xmm7,[GOTOFF(ebx,PW_SEVEN)] + paddw xmm5,[GOTOFF(ebx,PW_SEVEN)] + + paddw xmm1,xmm6 + paddw xmm0,xmm4 + psrlw xmm1,4 ; xmm1=Out1LE=( 0 2 4 6 8 10 12 14) + psrlw xmm0,4 ; xmm0=Out1HE=(16 18 20 22 24 26 28 30) + paddw xmm7,xmm6 + paddw xmm5,xmm4 + psrlw xmm7,4 ; xmm7=Out1LO=( 1 3 5 7 9 11 13 15) + psrlw xmm5,4 ; xmm5=Out1HO=(17 19 21 23 25 27 29 31) + + psllw xmm7,BYTE_BIT + psllw xmm5,BYTE_BIT + por xmm1,xmm7 ; xmm1=Out1L=( 0 1 2 ... 13 14 15) + por xmm0,xmm5 ; xmm0=Out1H=(16 17 18 ... 29 30 31) + + movdqa XMMWORD [edi+0*SIZEOF_XMMWORD], xmm1 + movdqa XMMWORD [edi+1*SIZEOF_XMMWORD], xmm0 + + poppic ebx + + sub eax, byte SIZEOF_XMMWORD + add ecx, byte 1*SIZEOF_XMMWORD ; inptr1(above) + add ebx, byte 1*SIZEOF_XMMWORD ; inptr0 + add esi, byte 1*SIZEOF_XMMWORD ; inptr1(below) + add edx, byte 2*SIZEOF_XMMWORD ; outptr0 + add edi, byte 2*SIZEOF_XMMWORD ; outptr1 + cmp eax, byte SIZEOF_XMMWORD + ja near .columnloop + test eax,eax + jnz near .columnloop_last + + pop esi + pop edi + pop ecx + pop eax + + add esi, byte 1*SIZEOF_JSAMPROW ; input_data + add edi, byte 2*SIZEOF_JSAMPROW ; output_data + sub ecx, byte 2 ; rowctr + jg near .rowloop + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; -------------------------------------------------------------------------- +; +; Fast processing for the common case of 2:1 horizontal and 1:1 vertical. +; It's still a box filter. +; +; GLOBAL(void) +; jsimd_h2v1_upsample_sse2 (int max_v_samp_factor, +; JDIMENSION output_width, +; JSAMPARRAY input_data, +; JSAMPARRAY * output_data_ptr); +; + +%define max_v_samp(b) (b)+8 ; int max_v_samp_factor +%define output_width(b) (b)+12 ; JDIMENSION output_width +%define input_data(b) (b)+16 ; JSAMPARRAY input_data +%define output_data_ptr(b) (b)+20 ; JSAMPARRAY * output_data_ptr + + align 16 + global EXTN(jsimd_h2v1_upsample_sse2) + +EXTN(jsimd_h2v1_upsample_sse2): + push ebp + mov ebp,esp +; push ebx ; unused +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + mov edx, JDIMENSION [output_width(ebp)] + add edx, byte (2*SIZEOF_XMMWORD)-1 + and edx, byte -(2*SIZEOF_XMMWORD) + jz short .return + + mov ecx, INT [max_v_samp(ebp)] ; rowctr + test ecx,ecx + jz short .return + + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + mov edi, POINTER [output_data_ptr(ebp)] + mov edi, JSAMPARRAY [edi] ; output_data + alignx 16,7 +.rowloop: + push edi + push esi + + mov esi, JSAMPROW [esi] ; inptr + mov edi, JSAMPROW [edi] ; outptr + mov eax,edx ; colctr + alignx 16,7 +.columnloop: + + movdqa xmm0, XMMWORD [esi+0*SIZEOF_XMMWORD] + + movdqa xmm1,xmm0 + punpcklbw xmm0,xmm0 + punpckhbw xmm1,xmm1 + + movdqa XMMWORD [edi+0*SIZEOF_XMMWORD], xmm0 + movdqa XMMWORD [edi+1*SIZEOF_XMMWORD], xmm1 + + sub eax, byte 2*SIZEOF_XMMWORD + jz short .nextrow + + movdqa xmm2, XMMWORD [esi+1*SIZEOF_XMMWORD] + + movdqa xmm3,xmm2 + punpcklbw xmm2,xmm2 + punpckhbw xmm3,xmm3 + + movdqa XMMWORD [edi+2*SIZEOF_XMMWORD], xmm2 + movdqa XMMWORD [edi+3*SIZEOF_XMMWORD], xmm3 + + sub eax, byte 2*SIZEOF_XMMWORD + jz short .nextrow + + add esi, byte 2*SIZEOF_XMMWORD ; inptr + add edi, byte 4*SIZEOF_XMMWORD ; outptr + jmp short .columnloop + alignx 16,7 + +.nextrow: + pop esi + pop edi + + add esi, byte SIZEOF_JSAMPROW ; input_data + add edi, byte SIZEOF_JSAMPROW ; output_data + dec ecx ; rowctr + jg short .rowloop + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved +; pop ebx ; unused + pop ebp + ret + +; -------------------------------------------------------------------------- +; +; Fast processing for the common case of 2:1 horizontal and 2:1 vertical. +; It's still a box filter. +; +; GLOBAL(void) +; jsimd_h2v2_upsample_sse2 (nt max_v_samp_factor, +; JDIMENSION output_width, +; JSAMPARRAY input_data, +; JSAMPARRAY * output_data_ptr); +; + +%define max_v_samp(b) (b)+8 ; int max_v_samp_factor +%define output_width(b) (b)+12 ; JDIMENSION output_width +%define input_data(b) (b)+16 ; JSAMPARRAY input_data +%define output_data_ptr(b) (b)+20 ; JSAMPARRAY * output_data_ptr + + align 16 + global EXTN(jsimd_h2v2_upsample_sse2) + +EXTN(jsimd_h2v2_upsample_sse2): + push ebp + mov ebp,esp + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + mov edx, JDIMENSION [output_width(ebp)] + add edx, byte (2*SIZEOF_XMMWORD)-1 + and edx, byte -(2*SIZEOF_XMMWORD) + jz near .return + + mov ecx, INT [max_v_samp(ebp)] ; rowctr + test ecx,ecx + jz near .return + + mov esi, JSAMPARRAY [input_data(ebp)] ; input_data + mov edi, POINTER [output_data_ptr(ebp)] + mov edi, JSAMPARRAY [edi] ; output_data + alignx 16,7 +.rowloop: + push edi + push esi + + mov esi, JSAMPROW [esi] ; inptr + mov ebx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] ; outptr0 + mov edi, JSAMPROW [edi+1*SIZEOF_JSAMPROW] ; outptr1 + mov eax,edx ; colctr + alignx 16,7 +.columnloop: + + movdqa xmm0, XMMWORD [esi+0*SIZEOF_XMMWORD] + + movdqa xmm1,xmm0 + punpcklbw xmm0,xmm0 + punpckhbw xmm1,xmm1 + + movdqa XMMWORD [ebx+0*SIZEOF_XMMWORD], xmm0 + movdqa XMMWORD [ebx+1*SIZEOF_XMMWORD], xmm1 + movdqa XMMWORD [edi+0*SIZEOF_XMMWORD], xmm0 + movdqa XMMWORD [edi+1*SIZEOF_XMMWORD], xmm1 + + sub eax, byte 2*SIZEOF_XMMWORD + jz short .nextrow + + movdqa xmm2, XMMWORD [esi+1*SIZEOF_XMMWORD] + + movdqa xmm3,xmm2 + punpcklbw xmm2,xmm2 + punpckhbw xmm3,xmm3 + + movdqa XMMWORD [ebx+2*SIZEOF_XMMWORD], xmm2 + movdqa XMMWORD [ebx+3*SIZEOF_XMMWORD], xmm3 + movdqa XMMWORD [edi+2*SIZEOF_XMMWORD], xmm2 + movdqa XMMWORD [edi+3*SIZEOF_XMMWORD], xmm3 + + sub eax, byte 2*SIZEOF_XMMWORD + jz short .nextrow + + add esi, byte 2*SIZEOF_XMMWORD ; inptr + add ebx, byte 4*SIZEOF_XMMWORD ; outptr0 + add edi, byte 4*SIZEOF_XMMWORD ; outptr1 + jmp short .columnloop + alignx 16,7 + +.nextrow: + pop esi + pop edi + + add esi, byte 1*SIZEOF_JSAMPROW ; input_data + add edi, byte 2*SIZEOF_JSAMPROW ; output_data + sub ecx, byte 2 ; rowctr + jg short .rowloop + +.return: + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jfdctflt-3dn.asm b/app/src/main/jni/libjpeg-turbo/simd/jfdctflt-3dn.asm new file mode 100644 index 000000000..e23c5218b --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jfdctflt-3dn.asm @@ -0,0 +1,320 @@ +; +; jfdctflt.asm - floating-point FDCT (3DNow!) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a floating-point implementation of the forward DCT +; (Discrete Cosine Transform). The following code is based directly on +; the IJG's original jfdctflt.c; see the jfdctflt.c for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_fdct_float_3dnow) + +EXTN(jconst_fdct_float_3dnow): + +PD_0_382 times 2 dd 0.382683432365089771728460 +PD_0_707 times 2 dd 0.707106781186547524400844 +PD_0_541 times 2 dd 0.541196100146196984399723 +PD_1_306 times 2 dd 1.306562964876376527856643 + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform the forward DCT on one block of samples. +; +; GLOBAL(void) +; jsimd_fdct_float_3dnow (FAST_FLOAT * data) +; + +%define data(b) (b)+8 ; FAST_FLOAT * data + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_MMWORD ; mmword wk[WK_NUM] +%define WK_NUM 2 + + align 16 + global EXTN(jsimd_fdct_float_3dnow) + +EXTN(jsimd_fdct_float_3dnow): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_MMWORD) ; align to 64 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved +; push esi ; unused +; push edi ; unused + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process rows. + + mov edx, POINTER [data(eax)] ; (FAST_FLOAT *) + mov ecx, DCTSIZE/2 + alignx 16,7 +.rowloop: + + movq mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)] + movq mm1, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)] + movq mm2, MMWORD [MMBLOCK(0,3,edx,SIZEOF_FAST_FLOAT)] + movq mm3, MMWORD [MMBLOCK(1,3,edx,SIZEOF_FAST_FLOAT)] + + ; mm0=(00 01), mm1=(10 11), mm2=(06 07), mm3=(16 17) + + movq mm4,mm0 ; transpose coefficients + punpckldq mm0,mm1 ; mm0=(00 10)=data0 + punpckhdq mm4,mm1 ; mm4=(01 11)=data1 + movq mm5,mm2 ; transpose coefficients + punpckldq mm2,mm3 ; mm2=(06 16)=data6 + punpckhdq mm5,mm3 ; mm5=(07 17)=data7 + + movq mm6,mm4 + movq mm7,mm0 + pfsub mm4,mm2 ; mm4=data1-data6=tmp6 + pfsub mm0,mm5 ; mm0=data0-data7=tmp7 + pfadd mm6,mm2 ; mm6=data1+data6=tmp1 + pfadd mm7,mm5 ; mm7=data0+data7=tmp0 + + movq mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)] + movq mm3, MMWORD [MMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)] + movq mm2, MMWORD [MMBLOCK(0,2,edx,SIZEOF_FAST_FLOAT)] + movq mm5, MMWORD [MMBLOCK(1,2,edx,SIZEOF_FAST_FLOAT)] + + ; mm1=(02 03), mm3=(12 13), mm2=(04 05), mm5=(14 15) + + movq MMWORD [wk(0)], mm4 ; wk(0)=tmp6 + movq MMWORD [wk(1)], mm0 ; wk(1)=tmp7 + + movq mm4,mm1 ; transpose coefficients + punpckldq mm1,mm3 ; mm1=(02 12)=data2 + punpckhdq mm4,mm3 ; mm4=(03 13)=data3 + movq mm0,mm2 ; transpose coefficients + punpckldq mm2,mm5 ; mm2=(04 14)=data4 + punpckhdq mm0,mm5 ; mm0=(05 15)=data5 + + movq mm3,mm4 + movq mm5,mm1 + pfadd mm4,mm2 ; mm4=data3+data4=tmp3 + pfadd mm1,mm0 ; mm1=data2+data5=tmp2 + pfsub mm3,mm2 ; mm3=data3-data4=tmp4 + pfsub mm5,mm0 ; mm5=data2-data5=tmp5 + + ; -- Even part + + movq mm2,mm7 + movq mm0,mm6 + pfsub mm7,mm4 ; mm7=tmp13 + pfsub mm6,mm1 ; mm6=tmp12 + pfadd mm2,mm4 ; mm2=tmp10 + pfadd mm0,mm1 ; mm0=tmp11 + + pfadd mm6,mm7 + pfmul mm6,[GOTOFF(ebx,PD_0_707)] ; mm6=z1 + + movq mm4,mm2 + movq mm1,mm7 + pfsub mm2,mm0 ; mm2=data4 + pfsub mm7,mm6 ; mm7=data6 + pfadd mm4,mm0 ; mm4=data0 + pfadd mm1,mm6 ; mm1=data2 + + movq MMWORD [MMBLOCK(0,2,edx,SIZEOF_FAST_FLOAT)], mm2 + movq MMWORD [MMBLOCK(0,3,edx,SIZEOF_FAST_FLOAT)], mm7 + movq MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)], mm4 + movq MMWORD [MMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)], mm1 + + ; -- Odd part + + movq mm0, MMWORD [wk(0)] ; mm0=tmp6 + movq mm6, MMWORD [wk(1)] ; mm6=tmp7 + + pfadd mm3,mm5 ; mm3=tmp10 + pfadd mm5,mm0 ; mm5=tmp11 + pfadd mm0,mm6 ; mm0=tmp12, mm6=tmp7 + + pfmul mm5,[GOTOFF(ebx,PD_0_707)] ; mm5=z3 + + movq mm2,mm3 ; mm2=tmp10 + pfsub mm3,mm0 + pfmul mm3,[GOTOFF(ebx,PD_0_382)] ; mm3=z5 + pfmul mm2,[GOTOFF(ebx,PD_0_541)] ; mm2=MULTIPLY(tmp10,FIX_0_54119610) + pfmul mm0,[GOTOFF(ebx,PD_1_306)] ; mm0=MULTIPLY(tmp12,FIX_1_30656296) + pfadd mm2,mm3 ; mm2=z2 + pfadd mm0,mm3 ; mm0=z4 + + movq mm7,mm6 + pfsub mm6,mm5 ; mm6=z13 + pfadd mm7,mm5 ; mm7=z11 + + movq mm4,mm6 + movq mm1,mm7 + pfsub mm6,mm2 ; mm6=data3 + pfsub mm7,mm0 ; mm7=data7 + pfadd mm4,mm2 ; mm4=data5 + pfadd mm1,mm0 ; mm1=data1 + + movq MMWORD [MMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)], mm6 + movq MMWORD [MMBLOCK(1,3,edx,SIZEOF_FAST_FLOAT)], mm7 + movq MMWORD [MMBLOCK(1,2,edx,SIZEOF_FAST_FLOAT)], mm4 + movq MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)], mm1 + + add edx, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT + dec ecx + jnz near .rowloop + + ; ---- Pass 2: process columns. + + mov edx, POINTER [data(eax)] ; (FAST_FLOAT *) + mov ecx, DCTSIZE/2 + alignx 16,7 +.columnloop: + + movq mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)] + movq mm1, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)] + movq mm2, MMWORD [MMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)] + movq mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)] + + ; mm0=(00 10), mm1=(01 11), mm2=(60 70), mm3=(61 71) + + movq mm4,mm0 ; transpose coefficients + punpckldq mm0,mm1 ; mm0=(00 01)=data0 + punpckhdq mm4,mm1 ; mm4=(10 11)=data1 + movq mm5,mm2 ; transpose coefficients + punpckldq mm2,mm3 ; mm2=(60 61)=data6 + punpckhdq mm5,mm3 ; mm5=(70 71)=data7 + + movq mm6,mm4 + movq mm7,mm0 + pfsub mm4,mm2 ; mm4=data1-data6=tmp6 + pfsub mm0,mm5 ; mm0=data0-data7=tmp7 + pfadd mm6,mm2 ; mm6=data1+data6=tmp1 + pfadd mm7,mm5 ; mm7=data0+data7=tmp0 + + movq mm1, MMWORD [MMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)] + movq mm3, MMWORD [MMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)] + movq mm2, MMWORD [MMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)] + movq mm5, MMWORD [MMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)] + + ; mm1=(20 30), mm3=(21 31), mm2=(40 50), mm5=(41 51) + + movq MMWORD [wk(0)], mm4 ; wk(0)=tmp6 + movq MMWORD [wk(1)], mm0 ; wk(1)=tmp7 + + movq mm4,mm1 ; transpose coefficients + punpckldq mm1,mm3 ; mm1=(20 21)=data2 + punpckhdq mm4,mm3 ; mm4=(30 31)=data3 + movq mm0,mm2 ; transpose coefficients + punpckldq mm2,mm5 ; mm2=(40 41)=data4 + punpckhdq mm0,mm5 ; mm0=(50 51)=data5 + + movq mm3,mm4 + movq mm5,mm1 + pfadd mm4,mm2 ; mm4=data3+data4=tmp3 + pfadd mm1,mm0 ; mm1=data2+data5=tmp2 + pfsub mm3,mm2 ; mm3=data3-data4=tmp4 + pfsub mm5,mm0 ; mm5=data2-data5=tmp5 + + ; -- Even part + + movq mm2,mm7 + movq mm0,mm6 + pfsub mm7,mm4 ; mm7=tmp13 + pfsub mm6,mm1 ; mm6=tmp12 + pfadd mm2,mm4 ; mm2=tmp10 + pfadd mm0,mm1 ; mm0=tmp11 + + pfadd mm6,mm7 + pfmul mm6,[GOTOFF(ebx,PD_0_707)] ; mm6=z1 + + movq mm4,mm2 + movq mm1,mm7 + pfsub mm2,mm0 ; mm2=data4 + pfsub mm7,mm6 ; mm7=data6 + pfadd mm4,mm0 ; mm4=data0 + pfadd mm1,mm6 ; mm1=data2 + + movq MMWORD [MMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)], mm2 + movq MMWORD [MMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)], mm7 + movq MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)], mm4 + movq MMWORD [MMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)], mm1 + + ; -- Odd part + + movq mm0, MMWORD [wk(0)] ; mm0=tmp6 + movq mm6, MMWORD [wk(1)] ; mm6=tmp7 + + pfadd mm3,mm5 ; mm3=tmp10 + pfadd mm5,mm0 ; mm5=tmp11 + pfadd mm0,mm6 ; mm0=tmp12, mm6=tmp7 + + pfmul mm5,[GOTOFF(ebx,PD_0_707)] ; mm5=z3 + + movq mm2,mm3 ; mm2=tmp10 + pfsub mm3,mm0 + pfmul mm3,[GOTOFF(ebx,PD_0_382)] ; mm3=z5 + pfmul mm2,[GOTOFF(ebx,PD_0_541)] ; mm2=MULTIPLY(tmp10,FIX_0_54119610) + pfmul mm0,[GOTOFF(ebx,PD_1_306)] ; mm0=MULTIPLY(tmp12,FIX_1_30656296) + pfadd mm2,mm3 ; mm2=z2 + pfadd mm0,mm3 ; mm0=z4 + + movq mm7,mm6 + pfsub mm6,mm5 ; mm6=z13 + pfadd mm7,mm5 ; mm7=z11 + + movq mm4,mm6 + movq mm1,mm7 + pfsub mm6,mm2 ; mm6=data3 + pfsub mm7,mm0 ; mm7=data7 + pfadd mm4,mm2 ; mm4=data5 + pfadd mm1,mm0 ; mm1=data1 + + movq MMWORD [MMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)], mm6 + movq MMWORD [MMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)], mm7 + movq MMWORD [MMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)], mm4 + movq MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)], mm1 + + add edx, byte 2*SIZEOF_FAST_FLOAT + dec ecx + jnz near .columnloop + + femms ; empty MMX/3DNow! state + +; pop edi ; unused +; pop esi ; unused +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + poppic ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jfdctflt-sse-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jfdctflt-sse-64.asm new file mode 100644 index 000000000..4184e95f6 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jfdctflt-sse-64.asm @@ -0,0 +1,358 @@ +; +; jfdctflt.asm - floating-point FDCT (64-bit SSE) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a floating-point implementation of the forward DCT +; (Discrete Cosine Transform). The following code is based directly on +; the IJG's original jfdctflt.c; see the jfdctflt.c for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%macro unpcklps2 2 ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5) + shufps %1,%2,0x44 +%endmacro + +%macro unpckhps2 2 ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7) + shufps %1,%2,0xEE +%endmacro + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_fdct_float_sse) + +EXTN(jconst_fdct_float_sse): + +PD_0_382 times 4 dd 0.382683432365089771728460 +PD_0_707 times 4 dd 0.707106781186547524400844 +PD_0_541 times 4 dd 0.541196100146196984399723 +PD_1_306 times 4 dd 1.306562964876376527856643 + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 +; +; Perform the forward DCT on one block of samples. +; +; GLOBAL(void) +; jsimd_fdct_float_sse (FAST_FLOAT * data) +; + +; r10 = FAST_FLOAT * data + +%define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 + + align 16 + global EXTN(jsimd_fdct_float_sse) + +EXTN(jsimd_fdct_float_sse): + push rbp + mov rax,rsp ; rax = original rbp + sub rsp, byte 4 + and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [rsp],rax + mov rbp,rsp ; rbp = aligned rbp + lea rsp, [wk(0)] + collect_args + + ; ---- Pass 1: process rows. + + mov rdx, r10 ; (FAST_FLOAT *) + mov rcx, DCTSIZE/4 +.rowloop: + + movaps xmm0, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)] + movaps xmm2, XMMWORD [XMMBLOCK(2,1,rdx,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(3,1,rdx,SIZEOF_FAST_FLOAT)] + + ; xmm0=(20 21 22 23), xmm2=(24 25 26 27) + ; xmm1=(30 31 32 33), xmm3=(34 35 36 37) + + movaps xmm4,xmm0 ; transpose coefficients(phase 1) + unpcklps xmm0,xmm1 ; xmm0=(20 30 21 31) + unpckhps xmm4,xmm1 ; xmm4=(22 32 23 33) + movaps xmm5,xmm2 ; transpose coefficients(phase 1) + unpcklps xmm2,xmm3 ; xmm2=(24 34 25 35) + unpckhps xmm5,xmm3 ; xmm5=(26 36 27 37) + + movaps xmm6, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)] + movaps xmm7, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(0,1,rdx,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(1,1,rdx,SIZEOF_FAST_FLOAT)] + + ; xmm6=(00 01 02 03), xmm1=(04 05 06 07) + ; xmm7=(10 11 12 13), xmm3=(14 15 16 17) + + movaps XMMWORD [wk(0)], xmm4 ; wk(0)=(22 32 23 33) + movaps XMMWORD [wk(1)], xmm2 ; wk(1)=(24 34 25 35) + + movaps xmm4,xmm6 ; transpose coefficients(phase 1) + unpcklps xmm6,xmm7 ; xmm6=(00 10 01 11) + unpckhps xmm4,xmm7 ; xmm4=(02 12 03 13) + movaps xmm2,xmm1 ; transpose coefficients(phase 1) + unpcklps xmm1,xmm3 ; xmm1=(04 14 05 15) + unpckhps xmm2,xmm3 ; xmm2=(06 16 07 17) + + movaps xmm7,xmm6 ; transpose coefficients(phase 2) + unpcklps2 xmm6,xmm0 ; xmm6=(00 10 20 30)=data0 + unpckhps2 xmm7,xmm0 ; xmm7=(01 11 21 31)=data1 + movaps xmm3,xmm2 ; transpose coefficients(phase 2) + unpcklps2 xmm2,xmm5 ; xmm2=(06 16 26 36)=data6 + unpckhps2 xmm3,xmm5 ; xmm3=(07 17 27 37)=data7 + + movaps xmm0,xmm7 + movaps xmm5,xmm6 + subps xmm7,xmm2 ; xmm7=data1-data6=tmp6 + subps xmm6,xmm3 ; xmm6=data0-data7=tmp7 + addps xmm0,xmm2 ; xmm0=data1+data6=tmp1 + addps xmm5,xmm3 ; xmm5=data0+data7=tmp0 + + movaps xmm2, XMMWORD [wk(0)] ; xmm2=(22 32 23 33) + movaps xmm3, XMMWORD [wk(1)] ; xmm3=(24 34 25 35) + movaps XMMWORD [wk(0)], xmm7 ; wk(0)=tmp6 + movaps XMMWORD [wk(1)], xmm6 ; wk(1)=tmp7 + + movaps xmm7,xmm4 ; transpose coefficients(phase 2) + unpcklps2 xmm4,xmm2 ; xmm4=(02 12 22 32)=data2 + unpckhps2 xmm7,xmm2 ; xmm7=(03 13 23 33)=data3 + movaps xmm6,xmm1 ; transpose coefficients(phase 2) + unpcklps2 xmm1,xmm3 ; xmm1=(04 14 24 34)=data4 + unpckhps2 xmm6,xmm3 ; xmm6=(05 15 25 35)=data5 + + movaps xmm2,xmm7 + movaps xmm3,xmm4 + addps xmm7,xmm1 ; xmm7=data3+data4=tmp3 + addps xmm4,xmm6 ; xmm4=data2+data5=tmp2 + subps xmm2,xmm1 ; xmm2=data3-data4=tmp4 + subps xmm3,xmm6 ; xmm3=data2-data5=tmp5 + + ; -- Even part + + movaps xmm1,xmm5 + movaps xmm6,xmm0 + subps xmm5,xmm7 ; xmm5=tmp13 + subps xmm0,xmm4 ; xmm0=tmp12 + addps xmm1,xmm7 ; xmm1=tmp10 + addps xmm6,xmm4 ; xmm6=tmp11 + + addps xmm0,xmm5 + mulps xmm0,[rel PD_0_707] ; xmm0=z1 + + movaps xmm7,xmm1 + movaps xmm4,xmm5 + subps xmm1,xmm6 ; xmm1=data4 + subps xmm5,xmm0 ; xmm5=data6 + addps xmm7,xmm6 ; xmm7=data0 + addps xmm4,xmm0 ; xmm4=data2 + + movaps XMMWORD [XMMBLOCK(0,1,rdx,SIZEOF_FAST_FLOAT)], xmm1 + movaps XMMWORD [XMMBLOCK(2,1,rdx,SIZEOF_FAST_FLOAT)], xmm5 + movaps XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)], xmm7 + movaps XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)], xmm4 + + ; -- Odd part + + movaps xmm6, XMMWORD [wk(0)] ; xmm6=tmp6 + movaps xmm0, XMMWORD [wk(1)] ; xmm0=tmp7 + + addps xmm2,xmm3 ; xmm2=tmp10 + addps xmm3,xmm6 ; xmm3=tmp11 + addps xmm6,xmm0 ; xmm6=tmp12, xmm0=tmp7 + + mulps xmm3,[rel PD_0_707] ; xmm3=z3 + + movaps xmm1,xmm2 ; xmm1=tmp10 + subps xmm2,xmm6 + mulps xmm2,[rel PD_0_382] ; xmm2=z5 + mulps xmm1,[rel PD_0_541] ; xmm1=MULTIPLY(tmp10,FIX_0_541196) + mulps xmm6,[rel PD_1_306] ; xmm6=MULTIPLY(tmp12,FIX_1_306562) + addps xmm1,xmm2 ; xmm1=z2 + addps xmm6,xmm2 ; xmm6=z4 + + movaps xmm5,xmm0 + subps xmm0,xmm3 ; xmm0=z13 + addps xmm5,xmm3 ; xmm5=z11 + + movaps xmm7,xmm0 + movaps xmm4,xmm5 + subps xmm0,xmm1 ; xmm0=data3 + subps xmm5,xmm6 ; xmm5=data7 + addps xmm7,xmm1 ; xmm7=data5 + addps xmm4,xmm6 ; xmm4=data1 + + movaps XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)], xmm0 + movaps XMMWORD [XMMBLOCK(3,1,rdx,SIZEOF_FAST_FLOAT)], xmm5 + movaps XMMWORD [XMMBLOCK(1,1,rdx,SIZEOF_FAST_FLOAT)], xmm7 + movaps XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)], xmm4 + + add rdx, 4*DCTSIZE*SIZEOF_FAST_FLOAT + dec rcx + jnz near .rowloop + + ; ---- Pass 2: process columns. + + mov rdx, r10 ; (FAST_FLOAT *) + mov rcx, DCTSIZE/4 +.columnloop: + + movaps xmm0, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)] + movaps xmm2, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_FAST_FLOAT)] + + ; xmm0=(02 12 22 32), xmm2=(42 52 62 72) + ; xmm1=(03 13 23 33), xmm3=(43 53 63 73) + + movaps xmm4,xmm0 ; transpose coefficients(phase 1) + unpcklps xmm0,xmm1 ; xmm0=(02 03 12 13) + unpckhps xmm4,xmm1 ; xmm4=(22 23 32 33) + movaps xmm5,xmm2 ; transpose coefficients(phase 1) + unpcklps xmm2,xmm3 ; xmm2=(42 43 52 53) + unpckhps xmm5,xmm3 ; xmm5=(62 63 72 73) + + movaps xmm6, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)] + movaps xmm7, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_FAST_FLOAT)] + + ; xmm6=(00 10 20 30), xmm1=(40 50 60 70) + ; xmm7=(01 11 21 31), xmm3=(41 51 61 71) + + movaps XMMWORD [wk(0)], xmm4 ; wk(0)=(22 23 32 33) + movaps XMMWORD [wk(1)], xmm2 ; wk(1)=(42 43 52 53) + + movaps xmm4,xmm6 ; transpose coefficients(phase 1) + unpcklps xmm6,xmm7 ; xmm6=(00 01 10 11) + unpckhps xmm4,xmm7 ; xmm4=(20 21 30 31) + movaps xmm2,xmm1 ; transpose coefficients(phase 1) + unpcklps xmm1,xmm3 ; xmm1=(40 41 50 51) + unpckhps xmm2,xmm3 ; xmm2=(60 61 70 71) + + movaps xmm7,xmm6 ; transpose coefficients(phase 2) + unpcklps2 xmm6,xmm0 ; xmm6=(00 01 02 03)=data0 + unpckhps2 xmm7,xmm0 ; xmm7=(10 11 12 13)=data1 + movaps xmm3,xmm2 ; transpose coefficients(phase 2) + unpcklps2 xmm2,xmm5 ; xmm2=(60 61 62 63)=data6 + unpckhps2 xmm3,xmm5 ; xmm3=(70 71 72 73)=data7 + + movaps xmm0,xmm7 + movaps xmm5,xmm6 + subps xmm7,xmm2 ; xmm7=data1-data6=tmp6 + subps xmm6,xmm3 ; xmm6=data0-data7=tmp7 + addps xmm0,xmm2 ; xmm0=data1+data6=tmp1 + addps xmm5,xmm3 ; xmm5=data0+data7=tmp0 + + movaps xmm2, XMMWORD [wk(0)] ; xmm2=(22 23 32 33) + movaps xmm3, XMMWORD [wk(1)] ; xmm3=(42 43 52 53) + movaps XMMWORD [wk(0)], xmm7 ; wk(0)=tmp6 + movaps XMMWORD [wk(1)], xmm6 ; wk(1)=tmp7 + + movaps xmm7,xmm4 ; transpose coefficients(phase 2) + unpcklps2 xmm4,xmm2 ; xmm4=(20 21 22 23)=data2 + unpckhps2 xmm7,xmm2 ; xmm7=(30 31 32 33)=data3 + movaps xmm6,xmm1 ; transpose coefficients(phase 2) + unpcklps2 xmm1,xmm3 ; xmm1=(40 41 42 43)=data4 + unpckhps2 xmm6,xmm3 ; xmm6=(50 51 52 53)=data5 + + movaps xmm2,xmm7 + movaps xmm3,xmm4 + addps xmm7,xmm1 ; xmm7=data3+data4=tmp3 + addps xmm4,xmm6 ; xmm4=data2+data5=tmp2 + subps xmm2,xmm1 ; xmm2=data3-data4=tmp4 + subps xmm3,xmm6 ; xmm3=data2-data5=tmp5 + + ; -- Even part + + movaps xmm1,xmm5 + movaps xmm6,xmm0 + subps xmm5,xmm7 ; xmm5=tmp13 + subps xmm0,xmm4 ; xmm0=tmp12 + addps xmm1,xmm7 ; xmm1=tmp10 + addps xmm6,xmm4 ; xmm6=tmp11 + + addps xmm0,xmm5 + mulps xmm0,[rel PD_0_707] ; xmm0=z1 + + movaps xmm7,xmm1 + movaps xmm4,xmm5 + subps xmm1,xmm6 ; xmm1=data4 + subps xmm5,xmm0 ; xmm5=data6 + addps xmm7,xmm6 ; xmm7=data0 + addps xmm4,xmm0 ; xmm4=data2 + + movaps XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_FAST_FLOAT)], xmm1 + movaps XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_FAST_FLOAT)], xmm5 + movaps XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)], xmm7 + movaps XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)], xmm4 + + ; -- Odd part + + movaps xmm6, XMMWORD [wk(0)] ; xmm6=tmp6 + movaps xmm0, XMMWORD [wk(1)] ; xmm0=tmp7 + + addps xmm2,xmm3 ; xmm2=tmp10 + addps xmm3,xmm6 ; xmm3=tmp11 + addps xmm6,xmm0 ; xmm6=tmp12, xmm0=tmp7 + + mulps xmm3,[rel PD_0_707] ; xmm3=z3 + + movaps xmm1,xmm2 ; xmm1=tmp10 + subps xmm2,xmm6 + mulps xmm2,[rel PD_0_382] ; xmm2=z5 + mulps xmm1,[rel PD_0_541] ; xmm1=MULTIPLY(tmp10,FIX_0_541196) + mulps xmm6,[rel PD_1_306] ; xmm6=MULTIPLY(tmp12,FIX_1_306562) + addps xmm1,xmm2 ; xmm1=z2 + addps xmm6,xmm2 ; xmm6=z4 + + movaps xmm5,xmm0 + subps xmm0,xmm3 ; xmm0=z13 + addps xmm5,xmm3 ; xmm5=z11 + + movaps xmm7,xmm0 + movaps xmm4,xmm5 + subps xmm0,xmm1 ; xmm0=data3 + subps xmm5,xmm6 ; xmm5=data7 + addps xmm7,xmm1 ; xmm7=data5 + addps xmm4,xmm6 ; xmm4=data1 + + movaps XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)], xmm0 + movaps XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_FAST_FLOAT)], xmm5 + movaps XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_FAST_FLOAT)], xmm7 + movaps XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)], xmm4 + + add rdx, byte 4*SIZEOF_FAST_FLOAT + dec rcx + jnz near .columnloop + + uncollect_args + mov rsp,rbp ; rsp <- aligned rbp + pop rsp ; rsp <- original rbp + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jfdctflt-sse.asm b/app/src/main/jni/libjpeg-turbo/simd/jfdctflt-sse.asm new file mode 100644 index 000000000..477f5c837 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jfdctflt-sse.asm @@ -0,0 +1,370 @@ +; +; jfdctflt.asm - floating-point FDCT (SSE) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a floating-point implementation of the forward DCT +; (Discrete Cosine Transform). The following code is based directly on +; the IJG's original jfdctflt.c; see the jfdctflt.c for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%macro unpcklps2 2 ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5) + shufps %1,%2,0x44 +%endmacro + +%macro unpckhps2 2 ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7) + shufps %1,%2,0xEE +%endmacro + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_fdct_float_sse) + +EXTN(jconst_fdct_float_sse): + +PD_0_382 times 4 dd 0.382683432365089771728460 +PD_0_707 times 4 dd 0.707106781186547524400844 +PD_0_541 times 4 dd 0.541196100146196984399723 +PD_1_306 times 4 dd 1.306562964876376527856643 + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform the forward DCT on one block of samples. +; +; GLOBAL(void) +; jsimd_fdct_float_sse (FAST_FLOAT * data) +; + +%define data(b) (b)+8 ; FAST_FLOAT * data + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 + + align 16 + global EXTN(jsimd_fdct_float_sse) + +EXTN(jsimd_fdct_float_sse): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved +; push esi ; unused +; push edi ; unused + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process rows. + + mov edx, POINTER [data(eax)] ; (FAST_FLOAT *) + mov ecx, DCTSIZE/4 + alignx 16,7 +.rowloop: + + movaps xmm0, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)] + movaps xmm2, XMMWORD [XMMBLOCK(2,1,edx,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(3,1,edx,SIZEOF_FAST_FLOAT)] + + ; xmm0=(20 21 22 23), xmm2=(24 25 26 27) + ; xmm1=(30 31 32 33), xmm3=(34 35 36 37) + + movaps xmm4,xmm0 ; transpose coefficients(phase 1) + unpcklps xmm0,xmm1 ; xmm0=(20 30 21 31) + unpckhps xmm4,xmm1 ; xmm4=(22 32 23 33) + movaps xmm5,xmm2 ; transpose coefficients(phase 1) + unpcklps xmm2,xmm3 ; xmm2=(24 34 25 35) + unpckhps xmm5,xmm3 ; xmm5=(26 36 27 37) + + movaps xmm6, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)] + movaps xmm7, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)] + + ; xmm6=(00 01 02 03), xmm1=(04 05 06 07) + ; xmm7=(10 11 12 13), xmm3=(14 15 16 17) + + movaps XMMWORD [wk(0)], xmm4 ; wk(0)=(22 32 23 33) + movaps XMMWORD [wk(1)], xmm2 ; wk(1)=(24 34 25 35) + + movaps xmm4,xmm6 ; transpose coefficients(phase 1) + unpcklps xmm6,xmm7 ; xmm6=(00 10 01 11) + unpckhps xmm4,xmm7 ; xmm4=(02 12 03 13) + movaps xmm2,xmm1 ; transpose coefficients(phase 1) + unpcklps xmm1,xmm3 ; xmm1=(04 14 05 15) + unpckhps xmm2,xmm3 ; xmm2=(06 16 07 17) + + movaps xmm7,xmm6 ; transpose coefficients(phase 2) + unpcklps2 xmm6,xmm0 ; xmm6=(00 10 20 30)=data0 + unpckhps2 xmm7,xmm0 ; xmm7=(01 11 21 31)=data1 + movaps xmm3,xmm2 ; transpose coefficients(phase 2) + unpcklps2 xmm2,xmm5 ; xmm2=(06 16 26 36)=data6 + unpckhps2 xmm3,xmm5 ; xmm3=(07 17 27 37)=data7 + + movaps xmm0,xmm7 + movaps xmm5,xmm6 + subps xmm7,xmm2 ; xmm7=data1-data6=tmp6 + subps xmm6,xmm3 ; xmm6=data0-data7=tmp7 + addps xmm0,xmm2 ; xmm0=data1+data6=tmp1 + addps xmm5,xmm3 ; xmm5=data0+data7=tmp0 + + movaps xmm2, XMMWORD [wk(0)] ; xmm2=(22 32 23 33) + movaps xmm3, XMMWORD [wk(1)] ; xmm3=(24 34 25 35) + movaps XMMWORD [wk(0)], xmm7 ; wk(0)=tmp6 + movaps XMMWORD [wk(1)], xmm6 ; wk(1)=tmp7 + + movaps xmm7,xmm4 ; transpose coefficients(phase 2) + unpcklps2 xmm4,xmm2 ; xmm4=(02 12 22 32)=data2 + unpckhps2 xmm7,xmm2 ; xmm7=(03 13 23 33)=data3 + movaps xmm6,xmm1 ; transpose coefficients(phase 2) + unpcklps2 xmm1,xmm3 ; xmm1=(04 14 24 34)=data4 + unpckhps2 xmm6,xmm3 ; xmm6=(05 15 25 35)=data5 + + movaps xmm2,xmm7 + movaps xmm3,xmm4 + addps xmm7,xmm1 ; xmm7=data3+data4=tmp3 + addps xmm4,xmm6 ; xmm4=data2+data5=tmp2 + subps xmm2,xmm1 ; xmm2=data3-data4=tmp4 + subps xmm3,xmm6 ; xmm3=data2-data5=tmp5 + + ; -- Even part + + movaps xmm1,xmm5 + movaps xmm6,xmm0 + subps xmm5,xmm7 ; xmm5=tmp13 + subps xmm0,xmm4 ; xmm0=tmp12 + addps xmm1,xmm7 ; xmm1=tmp10 + addps xmm6,xmm4 ; xmm6=tmp11 + + addps xmm0,xmm5 + mulps xmm0,[GOTOFF(ebx,PD_0_707)] ; xmm0=z1 + + movaps xmm7,xmm1 + movaps xmm4,xmm5 + subps xmm1,xmm6 ; xmm1=data4 + subps xmm5,xmm0 ; xmm5=data6 + addps xmm7,xmm6 ; xmm7=data0 + addps xmm4,xmm0 ; xmm4=data2 + + movaps XMMWORD [XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)], xmm1 + movaps XMMWORD [XMMBLOCK(2,1,edx,SIZEOF_FAST_FLOAT)], xmm5 + movaps XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)], xmm7 + movaps XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)], xmm4 + + ; -- Odd part + + movaps xmm6, XMMWORD [wk(0)] ; xmm6=tmp6 + movaps xmm0, XMMWORD [wk(1)] ; xmm0=tmp7 + + addps xmm2,xmm3 ; xmm2=tmp10 + addps xmm3,xmm6 ; xmm3=tmp11 + addps xmm6,xmm0 ; xmm6=tmp12, xmm0=tmp7 + + mulps xmm3,[GOTOFF(ebx,PD_0_707)] ; xmm3=z3 + + movaps xmm1,xmm2 ; xmm1=tmp10 + subps xmm2,xmm6 + mulps xmm2,[GOTOFF(ebx,PD_0_382)] ; xmm2=z5 + mulps xmm1,[GOTOFF(ebx,PD_0_541)] ; xmm1=MULTIPLY(tmp10,FIX_0_541196) + mulps xmm6,[GOTOFF(ebx,PD_1_306)] ; xmm6=MULTIPLY(tmp12,FIX_1_306562) + addps xmm1,xmm2 ; xmm1=z2 + addps xmm6,xmm2 ; xmm6=z4 + + movaps xmm5,xmm0 + subps xmm0,xmm3 ; xmm0=z13 + addps xmm5,xmm3 ; xmm5=z11 + + movaps xmm7,xmm0 + movaps xmm4,xmm5 + subps xmm0,xmm1 ; xmm0=data3 + subps xmm5,xmm6 ; xmm5=data7 + addps xmm7,xmm1 ; xmm7=data5 + addps xmm4,xmm6 ; xmm4=data1 + + movaps XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)], xmm0 + movaps XMMWORD [XMMBLOCK(3,1,edx,SIZEOF_FAST_FLOAT)], xmm5 + movaps XMMWORD [XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)], xmm7 + movaps XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)], xmm4 + + add edx, 4*DCTSIZE*SIZEOF_FAST_FLOAT + dec ecx + jnz near .rowloop + + ; ---- Pass 2: process columns. + + mov edx, POINTER [data(eax)] ; (FAST_FLOAT *) + mov ecx, DCTSIZE/4 + alignx 16,7 +.columnloop: + + movaps xmm0, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)] + movaps xmm2, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)] + + ; xmm0=(02 12 22 32), xmm2=(42 52 62 72) + ; xmm1=(03 13 23 33), xmm3=(43 53 63 73) + + movaps xmm4,xmm0 ; transpose coefficients(phase 1) + unpcklps xmm0,xmm1 ; xmm0=(02 03 12 13) + unpckhps xmm4,xmm1 ; xmm4=(22 23 32 33) + movaps xmm5,xmm2 ; transpose coefficients(phase 1) + unpcklps xmm2,xmm3 ; xmm2=(42 43 52 53) + unpckhps xmm5,xmm3 ; xmm5=(62 63 72 73) + + movaps xmm6, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)] + movaps xmm7, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)] + + ; xmm6=(00 10 20 30), xmm1=(40 50 60 70) + ; xmm7=(01 11 21 31), xmm3=(41 51 61 71) + + movaps XMMWORD [wk(0)], xmm4 ; wk(0)=(22 23 32 33) + movaps XMMWORD [wk(1)], xmm2 ; wk(1)=(42 43 52 53) + + movaps xmm4,xmm6 ; transpose coefficients(phase 1) + unpcklps xmm6,xmm7 ; xmm6=(00 01 10 11) + unpckhps xmm4,xmm7 ; xmm4=(20 21 30 31) + movaps xmm2,xmm1 ; transpose coefficients(phase 1) + unpcklps xmm1,xmm3 ; xmm1=(40 41 50 51) + unpckhps xmm2,xmm3 ; xmm2=(60 61 70 71) + + movaps xmm7,xmm6 ; transpose coefficients(phase 2) + unpcklps2 xmm6,xmm0 ; xmm6=(00 01 02 03)=data0 + unpckhps2 xmm7,xmm0 ; xmm7=(10 11 12 13)=data1 + movaps xmm3,xmm2 ; transpose coefficients(phase 2) + unpcklps2 xmm2,xmm5 ; xmm2=(60 61 62 63)=data6 + unpckhps2 xmm3,xmm5 ; xmm3=(70 71 72 73)=data7 + + movaps xmm0,xmm7 + movaps xmm5,xmm6 + subps xmm7,xmm2 ; xmm7=data1-data6=tmp6 + subps xmm6,xmm3 ; xmm6=data0-data7=tmp7 + addps xmm0,xmm2 ; xmm0=data1+data6=tmp1 + addps xmm5,xmm3 ; xmm5=data0+data7=tmp0 + + movaps xmm2, XMMWORD [wk(0)] ; xmm2=(22 23 32 33) + movaps xmm3, XMMWORD [wk(1)] ; xmm3=(42 43 52 53) + movaps XMMWORD [wk(0)], xmm7 ; wk(0)=tmp6 + movaps XMMWORD [wk(1)], xmm6 ; wk(1)=tmp7 + + movaps xmm7,xmm4 ; transpose coefficients(phase 2) + unpcklps2 xmm4,xmm2 ; xmm4=(20 21 22 23)=data2 + unpckhps2 xmm7,xmm2 ; xmm7=(30 31 32 33)=data3 + movaps xmm6,xmm1 ; transpose coefficients(phase 2) + unpcklps2 xmm1,xmm3 ; xmm1=(40 41 42 43)=data4 + unpckhps2 xmm6,xmm3 ; xmm6=(50 51 52 53)=data5 + + movaps xmm2,xmm7 + movaps xmm3,xmm4 + addps xmm7,xmm1 ; xmm7=data3+data4=tmp3 + addps xmm4,xmm6 ; xmm4=data2+data5=tmp2 + subps xmm2,xmm1 ; xmm2=data3-data4=tmp4 + subps xmm3,xmm6 ; xmm3=data2-data5=tmp5 + + ; -- Even part + + movaps xmm1,xmm5 + movaps xmm6,xmm0 + subps xmm5,xmm7 ; xmm5=tmp13 + subps xmm0,xmm4 ; xmm0=tmp12 + addps xmm1,xmm7 ; xmm1=tmp10 + addps xmm6,xmm4 ; xmm6=tmp11 + + addps xmm0,xmm5 + mulps xmm0,[GOTOFF(ebx,PD_0_707)] ; xmm0=z1 + + movaps xmm7,xmm1 + movaps xmm4,xmm5 + subps xmm1,xmm6 ; xmm1=data4 + subps xmm5,xmm0 ; xmm5=data6 + addps xmm7,xmm6 ; xmm7=data0 + addps xmm4,xmm0 ; xmm4=data2 + + movaps XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)], xmm1 + movaps XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)], xmm5 + movaps XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)], xmm7 + movaps XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)], xmm4 + + ; -- Odd part + + movaps xmm6, XMMWORD [wk(0)] ; xmm6=tmp6 + movaps xmm0, XMMWORD [wk(1)] ; xmm0=tmp7 + + addps xmm2,xmm3 ; xmm2=tmp10 + addps xmm3,xmm6 ; xmm3=tmp11 + addps xmm6,xmm0 ; xmm6=tmp12, xmm0=tmp7 + + mulps xmm3,[GOTOFF(ebx,PD_0_707)] ; xmm3=z3 + + movaps xmm1,xmm2 ; xmm1=tmp10 + subps xmm2,xmm6 + mulps xmm2,[GOTOFF(ebx,PD_0_382)] ; xmm2=z5 + mulps xmm1,[GOTOFF(ebx,PD_0_541)] ; xmm1=MULTIPLY(tmp10,FIX_0_541196) + mulps xmm6,[GOTOFF(ebx,PD_1_306)] ; xmm6=MULTIPLY(tmp12,FIX_1_306562) + addps xmm1,xmm2 ; xmm1=z2 + addps xmm6,xmm2 ; xmm6=z4 + + movaps xmm5,xmm0 + subps xmm0,xmm3 ; xmm0=z13 + addps xmm5,xmm3 ; xmm5=z11 + + movaps xmm7,xmm0 + movaps xmm4,xmm5 + subps xmm0,xmm1 ; xmm0=data3 + subps xmm5,xmm6 ; xmm5=data7 + addps xmm7,xmm1 ; xmm7=data5 + addps xmm4,xmm6 ; xmm4=data1 + + movaps XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)], xmm0 + movaps XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)], xmm5 + movaps XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)], xmm7 + movaps XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)], xmm4 + + add edx, byte 4*SIZEOF_FAST_FLOAT + dec ecx + jnz near .columnloop + +; pop edi ; unused +; pop esi ; unused +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + poppic ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jfdctfst-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jfdctfst-mmx.asm new file mode 100644 index 000000000..2c2d20c2e --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jfdctfst-mmx.asm @@ -0,0 +1,397 @@ +; +; jfdctfst.asm - fast integer FDCT (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a fast, not so accurate integer implementation of +; the forward DCT (Discrete Cosine Transform). The following code is +; based directly on the IJG's original jfdctfst.c; see the jfdctfst.c +; for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 8 ; 14 is also OK. + +%if CONST_BITS == 8 +F_0_382 equ 98 ; FIX(0.382683433) +F_0_541 equ 139 ; FIX(0.541196100) +F_0_707 equ 181 ; FIX(0.707106781) +F_1_306 equ 334 ; FIX(1.306562965) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_0_382 equ DESCALE( 410903207,30-CONST_BITS) ; FIX(0.382683433) +F_0_541 equ DESCALE( 581104887,30-CONST_BITS) ; FIX(0.541196100) +F_0_707 equ DESCALE( 759250124,30-CONST_BITS) ; FIX(0.707106781) +F_1_306 equ DESCALE(1402911301,30-CONST_BITS) ; FIX(1.306562965) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + +; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow) +; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw) + +%define PRE_MULTIPLY_SCALE_BITS 2 +%define CONST_SHIFT (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS) + + alignz 16 + global EXTN(jconst_fdct_ifast_mmx) + +EXTN(jconst_fdct_ifast_mmx): + +PW_F0707 times 4 dw F_0_707 << CONST_SHIFT +PW_F0382 times 4 dw F_0_382 << CONST_SHIFT +PW_F0541 times 4 dw F_0_541 << CONST_SHIFT +PW_F1306 times 4 dw F_1_306 << CONST_SHIFT + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform the forward DCT on one block of samples. +; +; GLOBAL(void) +; jsimd_fdct_ifast_mmx (DCTELEM * data) +; + +%define data(b) (b)+8 ; DCTELEM * data + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_MMWORD ; mmword wk[WK_NUM] +%define WK_NUM 2 + + align 16 + global EXTN(jsimd_fdct_ifast_mmx) + +EXTN(jsimd_fdct_ifast_mmx): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_MMWORD) ; align to 64 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved +; push esi ; unused +; push edi ; unused + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process rows. + + mov edx, POINTER [data(eax)] ; (DCTELEM *) + mov ecx, DCTSIZE/4 + alignx 16,7 +.rowloop: + + movq mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)] + movq mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)] + movq mm2, MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)] + movq mm3, MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)] + + ; mm0=(20 21 22 23), mm2=(24 25 26 27) + ; mm1=(30 31 32 33), mm3=(34 35 36 37) + + movq mm4,mm0 ; transpose coefficients(phase 1) + punpcklwd mm0,mm1 ; mm0=(20 30 21 31) + punpckhwd mm4,mm1 ; mm4=(22 32 23 33) + movq mm5,mm2 ; transpose coefficients(phase 1) + punpcklwd mm2,mm3 ; mm2=(24 34 25 35) + punpckhwd mm5,mm3 ; mm5=(26 36 27 37) + + movq mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)] + movq mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)] + movq mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)] + movq mm3, MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)] + + ; mm6=(00 01 02 03), mm1=(04 05 06 07) + ; mm7=(10 11 12 13), mm3=(14 15 16 17) + + movq MMWORD [wk(0)], mm4 ; wk(0)=(22 32 23 33) + movq MMWORD [wk(1)], mm2 ; wk(1)=(24 34 25 35) + + movq mm4,mm6 ; transpose coefficients(phase 1) + punpcklwd mm6,mm7 ; mm6=(00 10 01 11) + punpckhwd mm4,mm7 ; mm4=(02 12 03 13) + movq mm2,mm1 ; transpose coefficients(phase 1) + punpcklwd mm1,mm3 ; mm1=(04 14 05 15) + punpckhwd mm2,mm3 ; mm2=(06 16 07 17) + + movq mm7,mm6 ; transpose coefficients(phase 2) + punpckldq mm6,mm0 ; mm6=(00 10 20 30)=data0 + punpckhdq mm7,mm0 ; mm7=(01 11 21 31)=data1 + movq mm3,mm2 ; transpose coefficients(phase 2) + punpckldq mm2,mm5 ; mm2=(06 16 26 36)=data6 + punpckhdq mm3,mm5 ; mm3=(07 17 27 37)=data7 + + movq mm0,mm7 + movq mm5,mm6 + psubw mm7,mm2 ; mm7=data1-data6=tmp6 + psubw mm6,mm3 ; mm6=data0-data7=tmp7 + paddw mm0,mm2 ; mm0=data1+data6=tmp1 + paddw mm5,mm3 ; mm5=data0+data7=tmp0 + + movq mm2, MMWORD [wk(0)] ; mm2=(22 32 23 33) + movq mm3, MMWORD [wk(1)] ; mm3=(24 34 25 35) + movq MMWORD [wk(0)], mm7 ; wk(0)=tmp6 + movq MMWORD [wk(1)], mm6 ; wk(1)=tmp7 + + movq mm7,mm4 ; transpose coefficients(phase 2) + punpckldq mm4,mm2 ; mm4=(02 12 22 32)=data2 + punpckhdq mm7,mm2 ; mm7=(03 13 23 33)=data3 + movq mm6,mm1 ; transpose coefficients(phase 2) + punpckldq mm1,mm3 ; mm1=(04 14 24 34)=data4 + punpckhdq mm6,mm3 ; mm6=(05 15 25 35)=data5 + + movq mm2,mm7 + movq mm3,mm4 + paddw mm7,mm1 ; mm7=data3+data4=tmp3 + paddw mm4,mm6 ; mm4=data2+data5=tmp2 + psubw mm2,mm1 ; mm2=data3-data4=tmp4 + psubw mm3,mm6 ; mm3=data2-data5=tmp5 + + ; -- Even part + + movq mm1,mm5 + movq mm6,mm0 + psubw mm5,mm7 ; mm5=tmp13 + psubw mm0,mm4 ; mm0=tmp12 + paddw mm1,mm7 ; mm1=tmp10 + paddw mm6,mm4 ; mm6=tmp11 + + paddw mm0,mm5 + psllw mm0,PRE_MULTIPLY_SCALE_BITS + pmulhw mm0,[GOTOFF(ebx,PW_F0707)] ; mm0=z1 + + movq mm7,mm1 + movq mm4,mm5 + psubw mm1,mm6 ; mm1=data4 + psubw mm5,mm0 ; mm5=data6 + paddw mm7,mm6 ; mm7=data0 + paddw mm4,mm0 ; mm4=data2 + + movq MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)], mm1 + movq MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)], mm5 + movq MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm7 + movq MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4 + + ; -- Odd part + + movq mm6, MMWORD [wk(0)] ; mm6=tmp6 + movq mm0, MMWORD [wk(1)] ; mm0=tmp7 + + paddw mm2,mm3 ; mm2=tmp10 + paddw mm3,mm6 ; mm3=tmp11 + paddw mm6,mm0 ; mm6=tmp12, mm0=tmp7 + + psllw mm2,PRE_MULTIPLY_SCALE_BITS + psllw mm6,PRE_MULTIPLY_SCALE_BITS + + psllw mm3,PRE_MULTIPLY_SCALE_BITS + pmulhw mm3,[GOTOFF(ebx,PW_F0707)] ; mm3=z3 + + movq mm1,mm2 ; mm1=tmp10 + psubw mm2,mm6 + pmulhw mm2,[GOTOFF(ebx,PW_F0382)] ; mm2=z5 + pmulhw mm1,[GOTOFF(ebx,PW_F0541)] ; mm1=MULTIPLY(tmp10,FIX_0_54119610) + pmulhw mm6,[GOTOFF(ebx,PW_F1306)] ; mm6=MULTIPLY(tmp12,FIX_1_30656296) + paddw mm1,mm2 ; mm1=z2 + paddw mm6,mm2 ; mm6=z4 + + movq mm5,mm0 + psubw mm0,mm3 ; mm0=z13 + paddw mm5,mm3 ; mm5=z11 + + movq mm7,mm0 + movq mm4,mm5 + psubw mm0,mm1 ; mm0=data3 + psubw mm5,mm6 ; mm5=data7 + paddw mm7,mm1 ; mm7=data5 + paddw mm4,mm6 ; mm4=data1 + + movq MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm0 + movq MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)], mm5 + movq MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)], mm7 + movq MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm4 + + add edx, byte 4*DCTSIZE*SIZEOF_DCTELEM + dec ecx + jnz near .rowloop + + ; ---- Pass 2: process columns. + + mov edx, POINTER [data(eax)] ; (DCTELEM *) + mov ecx, DCTSIZE/4 + alignx 16,7 +.columnloop: + + movq mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)] + movq mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)] + movq mm2, MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)] + movq mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)] + + ; mm0=(02 12 22 32), mm2=(42 52 62 72) + ; mm1=(03 13 23 33), mm3=(43 53 63 73) + + movq mm4,mm0 ; transpose coefficients(phase 1) + punpcklwd mm0,mm1 ; mm0=(02 03 12 13) + punpckhwd mm4,mm1 ; mm4=(22 23 32 33) + movq mm5,mm2 ; transpose coefficients(phase 1) + punpcklwd mm2,mm3 ; mm2=(42 43 52 53) + punpckhwd mm5,mm3 ; mm5=(62 63 72 73) + + movq mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)] + movq mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)] + movq mm1, MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)] + movq mm3, MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)] + + ; mm6=(00 10 20 30), mm1=(40 50 60 70) + ; mm7=(01 11 21 31), mm3=(41 51 61 71) + + movq MMWORD [wk(0)], mm4 ; wk(0)=(22 23 32 33) + movq MMWORD [wk(1)], mm2 ; wk(1)=(42 43 52 53) + + movq mm4,mm6 ; transpose coefficients(phase 1) + punpcklwd mm6,mm7 ; mm6=(00 01 10 11) + punpckhwd mm4,mm7 ; mm4=(20 21 30 31) + movq mm2,mm1 ; transpose coefficients(phase 1) + punpcklwd mm1,mm3 ; mm1=(40 41 50 51) + punpckhwd mm2,mm3 ; mm2=(60 61 70 71) + + movq mm7,mm6 ; transpose coefficients(phase 2) + punpckldq mm6,mm0 ; mm6=(00 01 02 03)=data0 + punpckhdq mm7,mm0 ; mm7=(10 11 12 13)=data1 + movq mm3,mm2 ; transpose coefficients(phase 2) + punpckldq mm2,mm5 ; mm2=(60 61 62 63)=data6 + punpckhdq mm3,mm5 ; mm3=(70 71 72 73)=data7 + + movq mm0,mm7 + movq mm5,mm6 + psubw mm7,mm2 ; mm7=data1-data6=tmp6 + psubw mm6,mm3 ; mm6=data0-data7=tmp7 + paddw mm0,mm2 ; mm0=data1+data6=tmp1 + paddw mm5,mm3 ; mm5=data0+data7=tmp0 + + movq mm2, MMWORD [wk(0)] ; mm2=(22 23 32 33) + movq mm3, MMWORD [wk(1)] ; mm3=(42 43 52 53) + movq MMWORD [wk(0)], mm7 ; wk(0)=tmp6 + movq MMWORD [wk(1)], mm6 ; wk(1)=tmp7 + + movq mm7,mm4 ; transpose coefficients(phase 2) + punpckldq mm4,mm2 ; mm4=(20 21 22 23)=data2 + punpckhdq mm7,mm2 ; mm7=(30 31 32 33)=data3 + movq mm6,mm1 ; transpose coefficients(phase 2) + punpckldq mm1,mm3 ; mm1=(40 41 42 43)=data4 + punpckhdq mm6,mm3 ; mm6=(50 51 52 53)=data5 + + movq mm2,mm7 + movq mm3,mm4 + paddw mm7,mm1 ; mm7=data3+data4=tmp3 + paddw mm4,mm6 ; mm4=data2+data5=tmp2 + psubw mm2,mm1 ; mm2=data3-data4=tmp4 + psubw mm3,mm6 ; mm3=data2-data5=tmp5 + + ; -- Even part + + movq mm1,mm5 + movq mm6,mm0 + psubw mm5,mm7 ; mm5=tmp13 + psubw mm0,mm4 ; mm0=tmp12 + paddw mm1,mm7 ; mm1=tmp10 + paddw mm6,mm4 ; mm6=tmp11 + + paddw mm0,mm5 + psllw mm0,PRE_MULTIPLY_SCALE_BITS + pmulhw mm0,[GOTOFF(ebx,PW_F0707)] ; mm0=z1 + + movq mm7,mm1 + movq mm4,mm5 + psubw mm1,mm6 ; mm1=data4 + psubw mm5,mm0 ; mm5=data6 + paddw mm7,mm6 ; mm7=data0 + paddw mm4,mm0 ; mm4=data2 + + movq MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)], mm1 + movq MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)], mm5 + movq MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm7 + movq MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4 + + ; -- Odd part + + movq mm6, MMWORD [wk(0)] ; mm6=tmp6 + movq mm0, MMWORD [wk(1)] ; mm0=tmp7 + + paddw mm2,mm3 ; mm2=tmp10 + paddw mm3,mm6 ; mm3=tmp11 + paddw mm6,mm0 ; mm6=tmp12, mm0=tmp7 + + psllw mm2,PRE_MULTIPLY_SCALE_BITS + psllw mm6,PRE_MULTIPLY_SCALE_BITS + + psllw mm3,PRE_MULTIPLY_SCALE_BITS + pmulhw mm3,[GOTOFF(ebx,PW_F0707)] ; mm3=z3 + + movq mm1,mm2 ; mm1=tmp10 + psubw mm2,mm6 + pmulhw mm2,[GOTOFF(ebx,PW_F0382)] ; mm2=z5 + pmulhw mm1,[GOTOFF(ebx,PW_F0541)] ; mm1=MULTIPLY(tmp10,FIX_0_54119610) + pmulhw mm6,[GOTOFF(ebx,PW_F1306)] ; mm6=MULTIPLY(tmp12,FIX_1_30656296) + paddw mm1,mm2 ; mm1=z2 + paddw mm6,mm2 ; mm6=z4 + + movq mm5,mm0 + psubw mm0,mm3 ; mm0=z13 + paddw mm5,mm3 ; mm5=z11 + + movq mm7,mm0 + movq mm4,mm5 + psubw mm0,mm1 ; mm0=data3 + psubw mm5,mm6 ; mm5=data7 + paddw mm7,mm1 ; mm7=data5 + paddw mm4,mm6 ; mm4=data1 + + movq MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm0 + movq MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)], mm5 + movq MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)], mm7 + movq MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm4 + + add edx, byte 4*SIZEOF_DCTELEM + dec ecx + jnz near .columnloop + + emms ; empty MMX state + +; pop edi ; unused +; pop esi ; unused +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + poppic ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jfdctfst-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jfdctfst-sse2-64.asm new file mode 100644 index 000000000..13c4efc0b --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jfdctfst-sse2-64.asm @@ -0,0 +1,392 @@ +; +; jfdctfst.asm - fast integer FDCT (64-bit SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a fast, not so accurate integer implementation of +; the forward DCT (Discrete Cosine Transform). The following code is +; based directly on the IJG's original jfdctfst.c; see the jfdctfst.c +; for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 8 ; 14 is also OK. + +%if CONST_BITS == 8 +F_0_382 equ 98 ; FIX(0.382683433) +F_0_541 equ 139 ; FIX(0.541196100) +F_0_707 equ 181 ; FIX(0.707106781) +F_1_306 equ 334 ; FIX(1.306562965) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_0_382 equ DESCALE( 410903207,30-CONST_BITS) ; FIX(0.382683433) +F_0_541 equ DESCALE( 581104887,30-CONST_BITS) ; FIX(0.541196100) +F_0_707 equ DESCALE( 759250124,30-CONST_BITS) ; FIX(0.707106781) +F_1_306 equ DESCALE(1402911301,30-CONST_BITS) ; FIX(1.306562965) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + +; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow) +; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw) + +%define PRE_MULTIPLY_SCALE_BITS 2 +%define CONST_SHIFT (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS) + + alignz 16 + global EXTN(jconst_fdct_ifast_sse2) + +EXTN(jconst_fdct_ifast_sse2): + +PW_F0707 times 8 dw F_0_707 << CONST_SHIFT +PW_F0382 times 8 dw F_0_382 << CONST_SHIFT +PW_F0541 times 8 dw F_0_541 << CONST_SHIFT +PW_F1306 times 8 dw F_1_306 << CONST_SHIFT + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 +; +; Perform the forward DCT on one block of samples. +; +; GLOBAL(void) +; jsimd_fdct_ifast_sse2 (DCTELEM * data) +; + +; r10 = DCTELEM * data + +%define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 + + align 16 + global EXTN(jsimd_fdct_ifast_sse2) + +EXTN(jsimd_fdct_ifast_sse2): + push rbp + mov rax,rsp ; rax = original rbp + sub rsp, byte 4 + and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [rsp],rax + mov rbp,rsp ; rbp = aligned rbp + lea rsp, [wk(0)] + collect_args + + ; ---- Pass 1: process rows. + + mov rdx, r10 ; (DCTELEM *) + + movdqa xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)] + movdqa xmm1, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)] + movdqa xmm2, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)] + movdqa xmm3, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)] + + ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27) + ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37) + + movdqa xmm4,xmm0 ; transpose coefficients(phase 1) + punpcklwd xmm0,xmm1 ; xmm0=(00 10 01 11 02 12 03 13) + punpckhwd xmm4,xmm1 ; xmm4=(04 14 05 15 06 16 07 17) + movdqa xmm5,xmm2 ; transpose coefficients(phase 1) + punpcklwd xmm2,xmm3 ; xmm2=(20 30 21 31 22 32 23 33) + punpckhwd xmm5,xmm3 ; xmm5=(24 34 25 35 26 36 27 37) + + movdqa xmm6, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)] + movdqa xmm7, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)] + movdqa xmm1, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)] + movdqa xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)] + + ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62) + ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63) + + movdqa XMMWORD [wk(0)], xmm2 ; wk(0)=(20 30 21 31 22 32 23 33) + movdqa XMMWORD [wk(1)], xmm5 ; wk(1)=(24 34 25 35 26 36 27 37) + + movdqa xmm2,xmm6 ; transpose coefficients(phase 1) + punpcklwd xmm6,xmm7 ; xmm6=(40 50 41 51 42 52 43 53) + punpckhwd xmm2,xmm7 ; xmm2=(44 54 45 55 46 56 47 57) + movdqa xmm5,xmm1 ; transpose coefficients(phase 1) + punpcklwd xmm1,xmm3 ; xmm1=(60 70 61 71 62 72 63 73) + punpckhwd xmm5,xmm3 ; xmm5=(64 74 65 75 66 76 67 77) + + movdqa xmm7,xmm6 ; transpose coefficients(phase 2) + punpckldq xmm6,xmm1 ; xmm6=(40 50 60 70 41 51 61 71) + punpckhdq xmm7,xmm1 ; xmm7=(42 52 62 72 43 53 63 73) + movdqa xmm3,xmm2 ; transpose coefficients(phase 2) + punpckldq xmm2,xmm5 ; xmm2=(44 54 64 74 45 55 65 75) + punpckhdq xmm3,xmm5 ; xmm3=(46 56 66 76 47 57 67 77) + + movdqa xmm1, XMMWORD [wk(0)] ; xmm1=(20 30 21 31 22 32 23 33) + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=(24 34 25 35 26 36 27 37) + movdqa XMMWORD [wk(0)], xmm7 ; wk(0)=(42 52 62 72 43 53 63 73) + movdqa XMMWORD [wk(1)], xmm2 ; wk(1)=(44 54 64 74 45 55 65 75) + + movdqa xmm7,xmm0 ; transpose coefficients(phase 2) + punpckldq xmm0,xmm1 ; xmm0=(00 10 20 30 01 11 21 31) + punpckhdq xmm7,xmm1 ; xmm7=(02 12 22 32 03 13 23 33) + movdqa xmm2,xmm4 ; transpose coefficients(phase 2) + punpckldq xmm4,xmm5 ; xmm4=(04 14 24 34 05 15 25 35) + punpckhdq xmm2,xmm5 ; xmm2=(06 16 26 36 07 17 27 37) + + movdqa xmm1,xmm0 ; transpose coefficients(phase 3) + punpcklqdq xmm0,xmm6 ; xmm0=(00 10 20 30 40 50 60 70)=data0 + punpckhqdq xmm1,xmm6 ; xmm1=(01 11 21 31 41 51 61 71)=data1 + movdqa xmm5,xmm2 ; transpose coefficients(phase 3) + punpcklqdq xmm2,xmm3 ; xmm2=(06 16 26 36 46 56 66 76)=data6 + punpckhqdq xmm5,xmm3 ; xmm5=(07 17 27 37 47 57 67 77)=data7 + + movdqa xmm6,xmm1 + movdqa xmm3,xmm0 + psubw xmm1,xmm2 ; xmm1=data1-data6=tmp6 + psubw xmm0,xmm5 ; xmm0=data0-data7=tmp7 + paddw xmm6,xmm2 ; xmm6=data1+data6=tmp1 + paddw xmm3,xmm5 ; xmm3=data0+data7=tmp0 + + movdqa xmm2, XMMWORD [wk(0)] ; xmm2=(42 52 62 72 43 53 63 73) + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=(44 54 64 74 45 55 65 75) + movdqa XMMWORD [wk(0)], xmm1 ; wk(0)=tmp6 + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=tmp7 + + movdqa xmm1,xmm7 ; transpose coefficients(phase 3) + punpcklqdq xmm7,xmm2 ; xmm7=(02 12 22 32 42 52 62 72)=data2 + punpckhqdq xmm1,xmm2 ; xmm1=(03 13 23 33 43 53 63 73)=data3 + movdqa xmm0,xmm4 ; transpose coefficients(phase 3) + punpcklqdq xmm4,xmm5 ; xmm4=(04 14 24 34 44 54 64 74)=data4 + punpckhqdq xmm0,xmm5 ; xmm0=(05 15 25 35 45 55 65 75)=data5 + + movdqa xmm2,xmm1 + movdqa xmm5,xmm7 + paddw xmm1,xmm4 ; xmm1=data3+data4=tmp3 + paddw xmm7,xmm0 ; xmm7=data2+data5=tmp2 + psubw xmm2,xmm4 ; xmm2=data3-data4=tmp4 + psubw xmm5,xmm0 ; xmm5=data2-data5=tmp5 + + ; -- Even part + + movdqa xmm4,xmm3 + movdqa xmm0,xmm6 + psubw xmm3,xmm1 ; xmm3=tmp13 + psubw xmm6,xmm7 ; xmm6=tmp12 + paddw xmm4,xmm1 ; xmm4=tmp10 + paddw xmm0,xmm7 ; xmm0=tmp11 + + paddw xmm6,xmm3 + psllw xmm6,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm6,[rel PW_F0707] ; xmm6=z1 + + movdqa xmm1,xmm4 + movdqa xmm7,xmm3 + psubw xmm4,xmm0 ; xmm4=data4 + psubw xmm3,xmm6 ; xmm3=data6 + paddw xmm1,xmm0 ; xmm1=data0 + paddw xmm7,xmm6 ; xmm7=data2 + + movdqa xmm0, XMMWORD [wk(0)] ; xmm0=tmp6 + movdqa xmm6, XMMWORD [wk(1)] ; xmm6=tmp7 + movdqa XMMWORD [wk(0)], xmm4 ; wk(0)=data4 + movdqa XMMWORD [wk(1)], xmm3 ; wk(1)=data6 + + ; -- Odd part + + paddw xmm2,xmm5 ; xmm2=tmp10 + paddw xmm5,xmm0 ; xmm5=tmp11 + paddw xmm0,xmm6 ; xmm0=tmp12, xmm6=tmp7 + + psllw xmm2,PRE_MULTIPLY_SCALE_BITS + psllw xmm0,PRE_MULTIPLY_SCALE_BITS + + psllw xmm5,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm5,[rel PW_F0707] ; xmm5=z3 + + movdqa xmm4,xmm2 ; xmm4=tmp10 + psubw xmm2,xmm0 + pmulhw xmm2,[rel PW_F0382] ; xmm2=z5 + pmulhw xmm4,[rel PW_F0541] ; xmm4=MULTIPLY(tmp10,FIX_0_541196) + pmulhw xmm0,[rel PW_F1306] ; xmm0=MULTIPLY(tmp12,FIX_1_306562) + paddw xmm4,xmm2 ; xmm4=z2 + paddw xmm0,xmm2 ; xmm0=z4 + + movdqa xmm3,xmm6 + psubw xmm6,xmm5 ; xmm6=z13 + paddw xmm3,xmm5 ; xmm3=z11 + + movdqa xmm2,xmm6 + movdqa xmm5,xmm3 + psubw xmm6,xmm4 ; xmm6=data3 + psubw xmm3,xmm0 ; xmm3=data7 + paddw xmm2,xmm4 ; xmm2=data5 + paddw xmm5,xmm0 ; xmm5=data1 + + ; ---- Pass 2: process columns. + + ; xmm1=(00 10 20 30 40 50 60 70), xmm7=(02 12 22 32 42 52 62 72) + ; xmm5=(01 11 21 31 41 51 61 71), xmm6=(03 13 23 33 43 53 63 73) + + movdqa xmm4,xmm1 ; transpose coefficients(phase 1) + punpcklwd xmm1,xmm5 ; xmm1=(00 01 10 11 20 21 30 31) + punpckhwd xmm4,xmm5 ; xmm4=(40 41 50 51 60 61 70 71) + movdqa xmm0,xmm7 ; transpose coefficients(phase 1) + punpcklwd xmm7,xmm6 ; xmm7=(02 03 12 13 22 23 32 33) + punpckhwd xmm0,xmm6 ; xmm0=(42 43 52 53 62 63 72 73) + + movdqa xmm5, XMMWORD [wk(0)] ; xmm5=col4 + movdqa xmm6, XMMWORD [wk(1)] ; xmm6=col6 + + ; xmm5=(04 14 24 34 44 54 64 74), xmm6=(06 16 26 36 46 56 66 76) + ; xmm2=(05 15 25 35 45 55 65 75), xmm3=(07 17 27 37 47 57 67 77) + + movdqa XMMWORD [wk(0)], xmm7 ; wk(0)=(02 03 12 13 22 23 32 33) + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=(42 43 52 53 62 63 72 73) + + movdqa xmm7,xmm5 ; transpose coefficients(phase 1) + punpcklwd xmm5,xmm2 ; xmm5=(04 05 14 15 24 25 34 35) + punpckhwd xmm7,xmm2 ; xmm7=(44 45 54 55 64 65 74 75) + movdqa xmm0,xmm6 ; transpose coefficients(phase 1) + punpcklwd xmm6,xmm3 ; xmm6=(06 07 16 17 26 27 36 37) + punpckhwd xmm0,xmm3 ; xmm0=(46 47 56 57 66 67 76 77) + + movdqa xmm2,xmm5 ; transpose coefficients(phase 2) + punpckldq xmm5,xmm6 ; xmm5=(04 05 06 07 14 15 16 17) + punpckhdq xmm2,xmm6 ; xmm2=(24 25 26 27 34 35 36 37) + movdqa xmm3,xmm7 ; transpose coefficients(phase 2) + punpckldq xmm7,xmm0 ; xmm7=(44 45 46 47 54 55 56 57) + punpckhdq xmm3,xmm0 ; xmm3=(64 65 66 67 74 75 76 77) + + movdqa xmm6, XMMWORD [wk(0)] ; xmm6=(02 03 12 13 22 23 32 33) + movdqa xmm0, XMMWORD [wk(1)] ; xmm0=(42 43 52 53 62 63 72 73) + movdqa XMMWORD [wk(0)], xmm2 ; wk(0)=(24 25 26 27 34 35 36 37) + movdqa XMMWORD [wk(1)], xmm7 ; wk(1)=(44 45 46 47 54 55 56 57) + + movdqa xmm2,xmm1 ; transpose coefficients(phase 2) + punpckldq xmm1,xmm6 ; xmm1=(00 01 02 03 10 11 12 13) + punpckhdq xmm2,xmm6 ; xmm2=(20 21 22 23 30 31 32 33) + movdqa xmm7,xmm4 ; transpose coefficients(phase 2) + punpckldq xmm4,xmm0 ; xmm4=(40 41 42 43 50 51 52 53) + punpckhdq xmm7,xmm0 ; xmm7=(60 61 62 63 70 71 72 73) + + movdqa xmm6,xmm1 ; transpose coefficients(phase 3) + punpcklqdq xmm1,xmm5 ; xmm1=(00 01 02 03 04 05 06 07)=data0 + punpckhqdq xmm6,xmm5 ; xmm6=(10 11 12 13 14 15 16 17)=data1 + movdqa xmm0,xmm7 ; transpose coefficients(phase 3) + punpcklqdq xmm7,xmm3 ; xmm7=(60 61 62 63 64 65 66 67)=data6 + punpckhqdq xmm0,xmm3 ; xmm0=(70 71 72 73 74 75 76 77)=data7 + + movdqa xmm5,xmm6 + movdqa xmm3,xmm1 + psubw xmm6,xmm7 ; xmm6=data1-data6=tmp6 + psubw xmm1,xmm0 ; xmm1=data0-data7=tmp7 + paddw xmm5,xmm7 ; xmm5=data1+data6=tmp1 + paddw xmm3,xmm0 ; xmm3=data0+data7=tmp0 + + movdqa xmm7, XMMWORD [wk(0)] ; xmm7=(24 25 26 27 34 35 36 37) + movdqa xmm0, XMMWORD [wk(1)] ; xmm0=(44 45 46 47 54 55 56 57) + movdqa XMMWORD [wk(0)], xmm6 ; wk(0)=tmp6 + movdqa XMMWORD [wk(1)], xmm1 ; wk(1)=tmp7 + + movdqa xmm6,xmm2 ; transpose coefficients(phase 3) + punpcklqdq xmm2,xmm7 ; xmm2=(20 21 22 23 24 25 26 27)=data2 + punpckhqdq xmm6,xmm7 ; xmm6=(30 31 32 33 34 35 36 37)=data3 + movdqa xmm1,xmm4 ; transpose coefficients(phase 3) + punpcklqdq xmm4,xmm0 ; xmm4=(40 41 42 43 44 45 46 47)=data4 + punpckhqdq xmm1,xmm0 ; xmm1=(50 51 52 53 54 55 56 57)=data5 + + movdqa xmm7,xmm6 + movdqa xmm0,xmm2 + paddw xmm6,xmm4 ; xmm6=data3+data4=tmp3 + paddw xmm2,xmm1 ; xmm2=data2+data5=tmp2 + psubw xmm7,xmm4 ; xmm7=data3-data4=tmp4 + psubw xmm0,xmm1 ; xmm0=data2-data5=tmp5 + + ; -- Even part + + movdqa xmm4,xmm3 + movdqa xmm1,xmm5 + psubw xmm3,xmm6 ; xmm3=tmp13 + psubw xmm5,xmm2 ; xmm5=tmp12 + paddw xmm4,xmm6 ; xmm4=tmp10 + paddw xmm1,xmm2 ; xmm1=tmp11 + + paddw xmm5,xmm3 + psllw xmm5,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm5,[rel PW_F0707] ; xmm5=z1 + + movdqa xmm6,xmm4 + movdqa xmm2,xmm3 + psubw xmm4,xmm1 ; xmm4=data4 + psubw xmm3,xmm5 ; xmm3=data6 + paddw xmm6,xmm1 ; xmm6=data0 + paddw xmm2,xmm5 ; xmm2=data2 + + movdqa XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)], xmm4 + movdqa XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)], xmm3 + movdqa XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)], xmm6 + movdqa XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)], xmm2 + + ; -- Odd part + + movdqa xmm1, XMMWORD [wk(0)] ; xmm1=tmp6 + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=tmp7 + + paddw xmm7,xmm0 ; xmm7=tmp10 + paddw xmm0,xmm1 ; xmm0=tmp11 + paddw xmm1,xmm5 ; xmm1=tmp12, xmm5=tmp7 + + psllw xmm7,PRE_MULTIPLY_SCALE_BITS + psllw xmm1,PRE_MULTIPLY_SCALE_BITS + + psllw xmm0,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm0,[rel PW_F0707] ; xmm0=z3 + + movdqa xmm4,xmm7 ; xmm4=tmp10 + psubw xmm7,xmm1 + pmulhw xmm7,[rel PW_F0382] ; xmm7=z5 + pmulhw xmm4,[rel PW_F0541] ; xmm4=MULTIPLY(tmp10,FIX_0_541196) + pmulhw xmm1,[rel PW_F1306] ; xmm1=MULTIPLY(tmp12,FIX_1_306562) + paddw xmm4,xmm7 ; xmm4=z2 + paddw xmm1,xmm7 ; xmm1=z4 + + movdqa xmm3,xmm5 + psubw xmm5,xmm0 ; xmm5=z13 + paddw xmm3,xmm0 ; xmm3=z11 + + movdqa xmm6,xmm5 + movdqa xmm2,xmm3 + psubw xmm5,xmm4 ; xmm5=data3 + psubw xmm3,xmm1 ; xmm3=data7 + paddw xmm6,xmm4 ; xmm6=data5 + paddw xmm2,xmm1 ; xmm2=data1 + + movdqa XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)], xmm5 + movdqa XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)], xmm3 + movdqa XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)], xmm6 + movdqa XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)], xmm2 + + uncollect_args + mov rsp,rbp ; rsp <- aligned rbp + pop rsp ; rsp <- original rbp + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jfdctfst-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jfdctfst-sse2.asm new file mode 100644 index 000000000..e6e4b72f6 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jfdctfst-sse2.asm @@ -0,0 +1,404 @@ +; +; jfdctfst.asm - fast integer FDCT (SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a fast, not so accurate integer implementation of +; the forward DCT (Discrete Cosine Transform). The following code is +; based directly on the IJG's original jfdctfst.c; see the jfdctfst.c +; for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 8 ; 14 is also OK. + +%if CONST_BITS == 8 +F_0_382 equ 98 ; FIX(0.382683433) +F_0_541 equ 139 ; FIX(0.541196100) +F_0_707 equ 181 ; FIX(0.707106781) +F_1_306 equ 334 ; FIX(1.306562965) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_0_382 equ DESCALE( 410903207,30-CONST_BITS) ; FIX(0.382683433) +F_0_541 equ DESCALE( 581104887,30-CONST_BITS) ; FIX(0.541196100) +F_0_707 equ DESCALE( 759250124,30-CONST_BITS) ; FIX(0.707106781) +F_1_306 equ DESCALE(1402911301,30-CONST_BITS) ; FIX(1.306562965) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + +; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow) +; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw) + +%define PRE_MULTIPLY_SCALE_BITS 2 +%define CONST_SHIFT (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS) + + alignz 16 + global EXTN(jconst_fdct_ifast_sse2) + +EXTN(jconst_fdct_ifast_sse2): + +PW_F0707 times 8 dw F_0_707 << CONST_SHIFT +PW_F0382 times 8 dw F_0_382 << CONST_SHIFT +PW_F0541 times 8 dw F_0_541 << CONST_SHIFT +PW_F1306 times 8 dw F_1_306 << CONST_SHIFT + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform the forward DCT on one block of samples. +; +; GLOBAL(void) +; jsimd_fdct_ifast_sse2 (DCTELEM * data) +; + +%define data(b) (b)+8 ; DCTELEM * data + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 + + align 16 + global EXTN(jsimd_fdct_ifast_sse2) + +EXTN(jsimd_fdct_ifast_sse2): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic ebx +; push ecx ; unused +; push edx ; need not be preserved +; push esi ; unused +; push edi ; unused + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process rows. + + mov edx, POINTER [data(eax)] ; (DCTELEM *) + + movdqa xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)] + movdqa xmm1, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)] + movdqa xmm2, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)] + movdqa xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)] + + ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27) + ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37) + + movdqa xmm4,xmm0 ; transpose coefficients(phase 1) + punpcklwd xmm0,xmm1 ; xmm0=(00 10 01 11 02 12 03 13) + punpckhwd xmm4,xmm1 ; xmm4=(04 14 05 15 06 16 07 17) + movdqa xmm5,xmm2 ; transpose coefficients(phase 1) + punpcklwd xmm2,xmm3 ; xmm2=(20 30 21 31 22 32 23 33) + punpckhwd xmm5,xmm3 ; xmm5=(24 34 25 35 26 36 27 37) + + movdqa xmm6, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)] + movdqa xmm7, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)] + movdqa xmm1, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)] + movdqa xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)] + + ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62) + ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63) + + movdqa XMMWORD [wk(0)], xmm2 ; wk(0)=(20 30 21 31 22 32 23 33) + movdqa XMMWORD [wk(1)], xmm5 ; wk(1)=(24 34 25 35 26 36 27 37) + + movdqa xmm2,xmm6 ; transpose coefficients(phase 1) + punpcklwd xmm6,xmm7 ; xmm6=(40 50 41 51 42 52 43 53) + punpckhwd xmm2,xmm7 ; xmm2=(44 54 45 55 46 56 47 57) + movdqa xmm5,xmm1 ; transpose coefficients(phase 1) + punpcklwd xmm1,xmm3 ; xmm1=(60 70 61 71 62 72 63 73) + punpckhwd xmm5,xmm3 ; xmm5=(64 74 65 75 66 76 67 77) + + movdqa xmm7,xmm6 ; transpose coefficients(phase 2) + punpckldq xmm6,xmm1 ; xmm6=(40 50 60 70 41 51 61 71) + punpckhdq xmm7,xmm1 ; xmm7=(42 52 62 72 43 53 63 73) + movdqa xmm3,xmm2 ; transpose coefficients(phase 2) + punpckldq xmm2,xmm5 ; xmm2=(44 54 64 74 45 55 65 75) + punpckhdq xmm3,xmm5 ; xmm3=(46 56 66 76 47 57 67 77) + + movdqa xmm1, XMMWORD [wk(0)] ; xmm1=(20 30 21 31 22 32 23 33) + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=(24 34 25 35 26 36 27 37) + movdqa XMMWORD [wk(0)], xmm7 ; wk(0)=(42 52 62 72 43 53 63 73) + movdqa XMMWORD [wk(1)], xmm2 ; wk(1)=(44 54 64 74 45 55 65 75) + + movdqa xmm7,xmm0 ; transpose coefficients(phase 2) + punpckldq xmm0,xmm1 ; xmm0=(00 10 20 30 01 11 21 31) + punpckhdq xmm7,xmm1 ; xmm7=(02 12 22 32 03 13 23 33) + movdqa xmm2,xmm4 ; transpose coefficients(phase 2) + punpckldq xmm4,xmm5 ; xmm4=(04 14 24 34 05 15 25 35) + punpckhdq xmm2,xmm5 ; xmm2=(06 16 26 36 07 17 27 37) + + movdqa xmm1,xmm0 ; transpose coefficients(phase 3) + punpcklqdq xmm0,xmm6 ; xmm0=(00 10 20 30 40 50 60 70)=data0 + punpckhqdq xmm1,xmm6 ; xmm1=(01 11 21 31 41 51 61 71)=data1 + movdqa xmm5,xmm2 ; transpose coefficients(phase 3) + punpcklqdq xmm2,xmm3 ; xmm2=(06 16 26 36 46 56 66 76)=data6 + punpckhqdq xmm5,xmm3 ; xmm5=(07 17 27 37 47 57 67 77)=data7 + + movdqa xmm6,xmm1 + movdqa xmm3,xmm0 + psubw xmm1,xmm2 ; xmm1=data1-data6=tmp6 + psubw xmm0,xmm5 ; xmm0=data0-data7=tmp7 + paddw xmm6,xmm2 ; xmm6=data1+data6=tmp1 + paddw xmm3,xmm5 ; xmm3=data0+data7=tmp0 + + movdqa xmm2, XMMWORD [wk(0)] ; xmm2=(42 52 62 72 43 53 63 73) + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=(44 54 64 74 45 55 65 75) + movdqa XMMWORD [wk(0)], xmm1 ; wk(0)=tmp6 + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=tmp7 + + movdqa xmm1,xmm7 ; transpose coefficients(phase 3) + punpcklqdq xmm7,xmm2 ; xmm7=(02 12 22 32 42 52 62 72)=data2 + punpckhqdq xmm1,xmm2 ; xmm1=(03 13 23 33 43 53 63 73)=data3 + movdqa xmm0,xmm4 ; transpose coefficients(phase 3) + punpcklqdq xmm4,xmm5 ; xmm4=(04 14 24 34 44 54 64 74)=data4 + punpckhqdq xmm0,xmm5 ; xmm0=(05 15 25 35 45 55 65 75)=data5 + + movdqa xmm2,xmm1 + movdqa xmm5,xmm7 + paddw xmm1,xmm4 ; xmm1=data3+data4=tmp3 + paddw xmm7,xmm0 ; xmm7=data2+data5=tmp2 + psubw xmm2,xmm4 ; xmm2=data3-data4=tmp4 + psubw xmm5,xmm0 ; xmm5=data2-data5=tmp5 + + ; -- Even part + + movdqa xmm4,xmm3 + movdqa xmm0,xmm6 + psubw xmm3,xmm1 ; xmm3=tmp13 + psubw xmm6,xmm7 ; xmm6=tmp12 + paddw xmm4,xmm1 ; xmm4=tmp10 + paddw xmm0,xmm7 ; xmm0=tmp11 + + paddw xmm6,xmm3 + psllw xmm6,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm6,[GOTOFF(ebx,PW_F0707)] ; xmm6=z1 + + movdqa xmm1,xmm4 + movdqa xmm7,xmm3 + psubw xmm4,xmm0 ; xmm4=data4 + psubw xmm3,xmm6 ; xmm3=data6 + paddw xmm1,xmm0 ; xmm1=data0 + paddw xmm7,xmm6 ; xmm7=data2 + + movdqa xmm0, XMMWORD [wk(0)] ; xmm0=tmp6 + movdqa xmm6, XMMWORD [wk(1)] ; xmm6=tmp7 + movdqa XMMWORD [wk(0)], xmm4 ; wk(0)=data4 + movdqa XMMWORD [wk(1)], xmm3 ; wk(1)=data6 + + ; -- Odd part + + paddw xmm2,xmm5 ; xmm2=tmp10 + paddw xmm5,xmm0 ; xmm5=tmp11 + paddw xmm0,xmm6 ; xmm0=tmp12, xmm6=tmp7 + + psllw xmm2,PRE_MULTIPLY_SCALE_BITS + psllw xmm0,PRE_MULTIPLY_SCALE_BITS + + psllw xmm5,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm5,[GOTOFF(ebx,PW_F0707)] ; xmm5=z3 + + movdqa xmm4,xmm2 ; xmm4=tmp10 + psubw xmm2,xmm0 + pmulhw xmm2,[GOTOFF(ebx,PW_F0382)] ; xmm2=z5 + pmulhw xmm4,[GOTOFF(ebx,PW_F0541)] ; xmm4=MULTIPLY(tmp10,FIX_0_541196) + pmulhw xmm0,[GOTOFF(ebx,PW_F1306)] ; xmm0=MULTIPLY(tmp12,FIX_1_306562) + paddw xmm4,xmm2 ; xmm4=z2 + paddw xmm0,xmm2 ; xmm0=z4 + + movdqa xmm3,xmm6 + psubw xmm6,xmm5 ; xmm6=z13 + paddw xmm3,xmm5 ; xmm3=z11 + + movdqa xmm2,xmm6 + movdqa xmm5,xmm3 + psubw xmm6,xmm4 ; xmm6=data3 + psubw xmm3,xmm0 ; xmm3=data7 + paddw xmm2,xmm4 ; xmm2=data5 + paddw xmm5,xmm0 ; xmm5=data1 + + ; ---- Pass 2: process columns. + +; mov edx, POINTER [data(eax)] ; (DCTELEM *) + + ; xmm1=(00 10 20 30 40 50 60 70), xmm7=(02 12 22 32 42 52 62 72) + ; xmm5=(01 11 21 31 41 51 61 71), xmm6=(03 13 23 33 43 53 63 73) + + movdqa xmm4,xmm1 ; transpose coefficients(phase 1) + punpcklwd xmm1,xmm5 ; xmm1=(00 01 10 11 20 21 30 31) + punpckhwd xmm4,xmm5 ; xmm4=(40 41 50 51 60 61 70 71) + movdqa xmm0,xmm7 ; transpose coefficients(phase 1) + punpcklwd xmm7,xmm6 ; xmm7=(02 03 12 13 22 23 32 33) + punpckhwd xmm0,xmm6 ; xmm0=(42 43 52 53 62 63 72 73) + + movdqa xmm5, XMMWORD [wk(0)] ; xmm5=col4 + movdqa xmm6, XMMWORD [wk(1)] ; xmm6=col6 + + ; xmm5=(04 14 24 34 44 54 64 74), xmm6=(06 16 26 36 46 56 66 76) + ; xmm2=(05 15 25 35 45 55 65 75), xmm3=(07 17 27 37 47 57 67 77) + + movdqa XMMWORD [wk(0)], xmm7 ; wk(0)=(02 03 12 13 22 23 32 33) + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=(42 43 52 53 62 63 72 73) + + movdqa xmm7,xmm5 ; transpose coefficients(phase 1) + punpcklwd xmm5,xmm2 ; xmm5=(04 05 14 15 24 25 34 35) + punpckhwd xmm7,xmm2 ; xmm7=(44 45 54 55 64 65 74 75) + movdqa xmm0,xmm6 ; transpose coefficients(phase 1) + punpcklwd xmm6,xmm3 ; xmm6=(06 07 16 17 26 27 36 37) + punpckhwd xmm0,xmm3 ; xmm0=(46 47 56 57 66 67 76 77) + + movdqa xmm2,xmm5 ; transpose coefficients(phase 2) + punpckldq xmm5,xmm6 ; xmm5=(04 05 06 07 14 15 16 17) + punpckhdq xmm2,xmm6 ; xmm2=(24 25 26 27 34 35 36 37) + movdqa xmm3,xmm7 ; transpose coefficients(phase 2) + punpckldq xmm7,xmm0 ; xmm7=(44 45 46 47 54 55 56 57) + punpckhdq xmm3,xmm0 ; xmm3=(64 65 66 67 74 75 76 77) + + movdqa xmm6, XMMWORD [wk(0)] ; xmm6=(02 03 12 13 22 23 32 33) + movdqa xmm0, XMMWORD [wk(1)] ; xmm0=(42 43 52 53 62 63 72 73) + movdqa XMMWORD [wk(0)], xmm2 ; wk(0)=(24 25 26 27 34 35 36 37) + movdqa XMMWORD [wk(1)], xmm7 ; wk(1)=(44 45 46 47 54 55 56 57) + + movdqa xmm2,xmm1 ; transpose coefficients(phase 2) + punpckldq xmm1,xmm6 ; xmm1=(00 01 02 03 10 11 12 13) + punpckhdq xmm2,xmm6 ; xmm2=(20 21 22 23 30 31 32 33) + movdqa xmm7,xmm4 ; transpose coefficients(phase 2) + punpckldq xmm4,xmm0 ; xmm4=(40 41 42 43 50 51 52 53) + punpckhdq xmm7,xmm0 ; xmm7=(60 61 62 63 70 71 72 73) + + movdqa xmm6,xmm1 ; transpose coefficients(phase 3) + punpcklqdq xmm1,xmm5 ; xmm1=(00 01 02 03 04 05 06 07)=data0 + punpckhqdq xmm6,xmm5 ; xmm6=(10 11 12 13 14 15 16 17)=data1 + movdqa xmm0,xmm7 ; transpose coefficients(phase 3) + punpcklqdq xmm7,xmm3 ; xmm7=(60 61 62 63 64 65 66 67)=data6 + punpckhqdq xmm0,xmm3 ; xmm0=(70 71 72 73 74 75 76 77)=data7 + + movdqa xmm5,xmm6 + movdqa xmm3,xmm1 + psubw xmm6,xmm7 ; xmm6=data1-data6=tmp6 + psubw xmm1,xmm0 ; xmm1=data0-data7=tmp7 + paddw xmm5,xmm7 ; xmm5=data1+data6=tmp1 + paddw xmm3,xmm0 ; xmm3=data0+data7=tmp0 + + movdqa xmm7, XMMWORD [wk(0)] ; xmm7=(24 25 26 27 34 35 36 37) + movdqa xmm0, XMMWORD [wk(1)] ; xmm0=(44 45 46 47 54 55 56 57) + movdqa XMMWORD [wk(0)], xmm6 ; wk(0)=tmp6 + movdqa XMMWORD [wk(1)], xmm1 ; wk(1)=tmp7 + + movdqa xmm6,xmm2 ; transpose coefficients(phase 3) + punpcklqdq xmm2,xmm7 ; xmm2=(20 21 22 23 24 25 26 27)=data2 + punpckhqdq xmm6,xmm7 ; xmm6=(30 31 32 33 34 35 36 37)=data3 + movdqa xmm1,xmm4 ; transpose coefficients(phase 3) + punpcklqdq xmm4,xmm0 ; xmm4=(40 41 42 43 44 45 46 47)=data4 + punpckhqdq xmm1,xmm0 ; xmm1=(50 51 52 53 54 55 56 57)=data5 + + movdqa xmm7,xmm6 + movdqa xmm0,xmm2 + paddw xmm6,xmm4 ; xmm6=data3+data4=tmp3 + paddw xmm2,xmm1 ; xmm2=data2+data5=tmp2 + psubw xmm7,xmm4 ; xmm7=data3-data4=tmp4 + psubw xmm0,xmm1 ; xmm0=data2-data5=tmp5 + + ; -- Even part + + movdqa xmm4,xmm3 + movdqa xmm1,xmm5 + psubw xmm3,xmm6 ; xmm3=tmp13 + psubw xmm5,xmm2 ; xmm5=tmp12 + paddw xmm4,xmm6 ; xmm4=tmp10 + paddw xmm1,xmm2 ; xmm1=tmp11 + + paddw xmm5,xmm3 + psllw xmm5,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm5,[GOTOFF(ebx,PW_F0707)] ; xmm5=z1 + + movdqa xmm6,xmm4 + movdqa xmm2,xmm3 + psubw xmm4,xmm1 ; xmm4=data4 + psubw xmm3,xmm5 ; xmm3=data6 + paddw xmm6,xmm1 ; xmm6=data0 + paddw xmm2,xmm5 ; xmm2=data2 + + movdqa XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)], xmm4 + movdqa XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)], xmm3 + movdqa XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)], xmm6 + movdqa XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)], xmm2 + + ; -- Odd part + + movdqa xmm1, XMMWORD [wk(0)] ; xmm1=tmp6 + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=tmp7 + + paddw xmm7,xmm0 ; xmm7=tmp10 + paddw xmm0,xmm1 ; xmm0=tmp11 + paddw xmm1,xmm5 ; xmm1=tmp12, xmm5=tmp7 + + psllw xmm7,PRE_MULTIPLY_SCALE_BITS + psllw xmm1,PRE_MULTIPLY_SCALE_BITS + + psllw xmm0,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm0,[GOTOFF(ebx,PW_F0707)] ; xmm0=z3 + + movdqa xmm4,xmm7 ; xmm4=tmp10 + psubw xmm7,xmm1 + pmulhw xmm7,[GOTOFF(ebx,PW_F0382)] ; xmm7=z5 + pmulhw xmm4,[GOTOFF(ebx,PW_F0541)] ; xmm4=MULTIPLY(tmp10,FIX_0_541196) + pmulhw xmm1,[GOTOFF(ebx,PW_F1306)] ; xmm1=MULTIPLY(tmp12,FIX_1_306562) + paddw xmm4,xmm7 ; xmm4=z2 + paddw xmm1,xmm7 ; xmm1=z4 + + movdqa xmm3,xmm5 + psubw xmm5,xmm0 ; xmm5=z13 + paddw xmm3,xmm0 ; xmm3=z11 + + movdqa xmm6,xmm5 + movdqa xmm2,xmm3 + psubw xmm5,xmm4 ; xmm5=data3 + psubw xmm3,xmm1 ; xmm3=data7 + paddw xmm6,xmm4 ; xmm6=data5 + paddw xmm2,xmm1 ; xmm2=data1 + + movdqa XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)], xmm5 + movdqa XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)], xmm3 + movdqa XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)], xmm6 + movdqa XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)], xmm2 + +; pop edi ; unused +; pop esi ; unused +; pop edx ; need not be preserved +; pop ecx ; unused + poppic ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jfdctint-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jfdctint-mmx.asm new file mode 100644 index 000000000..8536ae2cf --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jfdctint-mmx.asm @@ -0,0 +1,622 @@ +; +; jfdctint.asm - accurate integer FDCT (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a slow-but-accurate integer implementation of the +; forward DCT (Discrete Cosine Transform). The following code is based +; directly on the IJG's original jfdctint.c; see the jfdctint.c for +; more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 13 +%define PASS1_BITS 2 + +%define DESCALE_P1 (CONST_BITS-PASS1_BITS) +%define DESCALE_P2 (CONST_BITS+PASS1_BITS) + +%if CONST_BITS == 13 +F_0_298 equ 2446 ; FIX(0.298631336) +F_0_390 equ 3196 ; FIX(0.390180644) +F_0_541 equ 4433 ; FIX(0.541196100) +F_0_765 equ 6270 ; FIX(0.765366865) +F_0_899 equ 7373 ; FIX(0.899976223) +F_1_175 equ 9633 ; FIX(1.175875602) +F_1_501 equ 12299 ; FIX(1.501321110) +F_1_847 equ 15137 ; FIX(1.847759065) +F_1_961 equ 16069 ; FIX(1.961570560) +F_2_053 equ 16819 ; FIX(2.053119869) +F_2_562 equ 20995 ; FIX(2.562915447) +F_3_072 equ 25172 ; FIX(3.072711026) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_0_298 equ DESCALE( 320652955,30-CONST_BITS) ; FIX(0.298631336) +F_0_390 equ DESCALE( 418953276,30-CONST_BITS) ; FIX(0.390180644) +F_0_541 equ DESCALE( 581104887,30-CONST_BITS) ; FIX(0.541196100) +F_0_765 equ DESCALE( 821806413,30-CONST_BITS) ; FIX(0.765366865) +F_0_899 equ DESCALE( 966342111,30-CONST_BITS) ; FIX(0.899976223) +F_1_175 equ DESCALE(1262586813,30-CONST_BITS) ; FIX(1.175875602) +F_1_501 equ DESCALE(1612031267,30-CONST_BITS) ; FIX(1.501321110) +F_1_847 equ DESCALE(1984016188,30-CONST_BITS) ; FIX(1.847759065) +F_1_961 equ DESCALE(2106220350,30-CONST_BITS) ; FIX(1.961570560) +F_2_053 equ DESCALE(2204520673,30-CONST_BITS) ; FIX(2.053119869) +F_2_562 equ DESCALE(2751909506,30-CONST_BITS) ; FIX(2.562915447) +F_3_072 equ DESCALE(3299298341,30-CONST_BITS) ; FIX(3.072711026) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_fdct_islow_mmx) + +EXTN(jconst_fdct_islow_mmx): + +PW_F130_F054 times 2 dw (F_0_541+F_0_765), F_0_541 +PW_F054_MF130 times 2 dw F_0_541, (F_0_541-F_1_847) +PW_MF078_F117 times 2 dw (F_1_175-F_1_961), F_1_175 +PW_F117_F078 times 2 dw F_1_175, (F_1_175-F_0_390) +PW_MF060_MF089 times 2 dw (F_0_298-F_0_899),-F_0_899 +PW_MF089_F060 times 2 dw -F_0_899, (F_1_501-F_0_899) +PW_MF050_MF256 times 2 dw (F_2_053-F_2_562),-F_2_562 +PW_MF256_F050 times 2 dw -F_2_562, (F_3_072-F_2_562) +PD_DESCALE_P1 times 2 dd 1 << (DESCALE_P1-1) +PD_DESCALE_P2 times 2 dd 1 << (DESCALE_P2-1) +PW_DESCALE_P2X times 4 dw 1 << (PASS1_BITS-1) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform the forward DCT on one block of samples. +; +; GLOBAL(void) +; jsimd_fdct_islow_mmx (DCTELEM * data) +; + +%define data(b) (b)+8 ; DCTELEM * data + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_MMWORD ; mmword wk[WK_NUM] +%define WK_NUM 2 + + align 16 + global EXTN(jsimd_fdct_islow_mmx) + +EXTN(jsimd_fdct_islow_mmx): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_MMWORD) ; align to 64 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved +; push esi ; unused +; push edi ; unused + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process rows. + + mov edx, POINTER [data(eax)] ; (DCTELEM *) + mov ecx, DCTSIZE/4 + alignx 16,7 +.rowloop: + + movq mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)] + movq mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)] + movq mm2, MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)] + movq mm3, MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)] + + ; mm0=(20 21 22 23), mm2=(24 25 26 27) + ; mm1=(30 31 32 33), mm3=(34 35 36 37) + + movq mm4,mm0 ; transpose coefficients(phase 1) + punpcklwd mm0,mm1 ; mm0=(20 30 21 31) + punpckhwd mm4,mm1 ; mm4=(22 32 23 33) + movq mm5,mm2 ; transpose coefficients(phase 1) + punpcklwd mm2,mm3 ; mm2=(24 34 25 35) + punpckhwd mm5,mm3 ; mm5=(26 36 27 37) + + movq mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)] + movq mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)] + movq mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)] + movq mm3, MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)] + + ; mm6=(00 01 02 03), mm1=(04 05 06 07) + ; mm7=(10 11 12 13), mm3=(14 15 16 17) + + movq MMWORD [wk(0)], mm4 ; wk(0)=(22 32 23 33) + movq MMWORD [wk(1)], mm2 ; wk(1)=(24 34 25 35) + + movq mm4,mm6 ; transpose coefficients(phase 1) + punpcklwd mm6,mm7 ; mm6=(00 10 01 11) + punpckhwd mm4,mm7 ; mm4=(02 12 03 13) + movq mm2,mm1 ; transpose coefficients(phase 1) + punpcklwd mm1,mm3 ; mm1=(04 14 05 15) + punpckhwd mm2,mm3 ; mm2=(06 16 07 17) + + movq mm7,mm6 ; transpose coefficients(phase 2) + punpckldq mm6,mm0 ; mm6=(00 10 20 30)=data0 + punpckhdq mm7,mm0 ; mm7=(01 11 21 31)=data1 + movq mm3,mm2 ; transpose coefficients(phase 2) + punpckldq mm2,mm5 ; mm2=(06 16 26 36)=data6 + punpckhdq mm3,mm5 ; mm3=(07 17 27 37)=data7 + + movq mm0,mm7 + movq mm5,mm6 + psubw mm7,mm2 ; mm7=data1-data6=tmp6 + psubw mm6,mm3 ; mm6=data0-data7=tmp7 + paddw mm0,mm2 ; mm0=data1+data6=tmp1 + paddw mm5,mm3 ; mm5=data0+data7=tmp0 + + movq mm2, MMWORD [wk(0)] ; mm2=(22 32 23 33) + movq mm3, MMWORD [wk(1)] ; mm3=(24 34 25 35) + movq MMWORD [wk(0)], mm7 ; wk(0)=tmp6 + movq MMWORD [wk(1)], mm6 ; wk(1)=tmp7 + + movq mm7,mm4 ; transpose coefficients(phase 2) + punpckldq mm4,mm2 ; mm4=(02 12 22 32)=data2 + punpckhdq mm7,mm2 ; mm7=(03 13 23 33)=data3 + movq mm6,mm1 ; transpose coefficients(phase 2) + punpckldq mm1,mm3 ; mm1=(04 14 24 34)=data4 + punpckhdq mm6,mm3 ; mm6=(05 15 25 35)=data5 + + movq mm2,mm7 + movq mm3,mm4 + paddw mm7,mm1 ; mm7=data3+data4=tmp3 + paddw mm4,mm6 ; mm4=data2+data5=tmp2 + psubw mm2,mm1 ; mm2=data3-data4=tmp4 + psubw mm3,mm6 ; mm3=data2-data5=tmp5 + + ; -- Even part + + movq mm1,mm5 + movq mm6,mm0 + paddw mm5,mm7 ; mm5=tmp10 + paddw mm0,mm4 ; mm0=tmp11 + psubw mm1,mm7 ; mm1=tmp13 + psubw mm6,mm4 ; mm6=tmp12 + + movq mm7,mm5 + paddw mm5,mm0 ; mm5=tmp10+tmp11 + psubw mm7,mm0 ; mm7=tmp10-tmp11 + + psllw mm5,PASS1_BITS ; mm5=data0 + psllw mm7,PASS1_BITS ; mm7=data4 + + movq MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm5 + movq MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)], mm7 + + ; (Original) + ; z1 = (tmp12 + tmp13) * 0.541196100; + ; data2 = z1 + tmp13 * 0.765366865; + ; data6 = z1 + tmp12 * -1.847759065; + ; + ; (This implementation) + ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100; + ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065); + + movq mm4,mm1 ; mm1=tmp13 + movq mm0,mm1 + punpcklwd mm4,mm6 ; mm6=tmp12 + punpckhwd mm0,mm6 + movq mm1,mm4 + movq mm6,mm0 + pmaddwd mm4,[GOTOFF(ebx,PW_F130_F054)] ; mm4=data2L + pmaddwd mm0,[GOTOFF(ebx,PW_F130_F054)] ; mm0=data2H + pmaddwd mm1,[GOTOFF(ebx,PW_F054_MF130)] ; mm1=data6L + pmaddwd mm6,[GOTOFF(ebx,PW_F054_MF130)] ; mm6=data6H + + paddd mm4,[GOTOFF(ebx,PD_DESCALE_P1)] + paddd mm0,[GOTOFF(ebx,PD_DESCALE_P1)] + psrad mm4,DESCALE_P1 + psrad mm0,DESCALE_P1 + paddd mm1,[GOTOFF(ebx,PD_DESCALE_P1)] + paddd mm6,[GOTOFF(ebx,PD_DESCALE_P1)] + psrad mm1,DESCALE_P1 + psrad mm6,DESCALE_P1 + + packssdw mm4,mm0 ; mm4=data2 + packssdw mm1,mm6 ; mm1=data6 + + movq MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4 + movq MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)], mm1 + + ; -- Odd part + + movq mm5, MMWORD [wk(0)] ; mm5=tmp6 + movq mm7, MMWORD [wk(1)] ; mm7=tmp7 + + movq mm0,mm2 ; mm2=tmp4 + movq mm6,mm3 ; mm3=tmp5 + paddw mm0,mm5 ; mm0=z3 + paddw mm6,mm7 ; mm6=z4 + + ; (Original) + ; z5 = (z3 + z4) * 1.175875602; + ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; + ; z3 += z5; z4 += z5; + ; + ; (This implementation) + ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; + ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); + + movq mm4,mm0 + movq mm1,mm0 + punpcklwd mm4,mm6 + punpckhwd mm1,mm6 + movq mm0,mm4 + movq mm6,mm1 + pmaddwd mm4,[GOTOFF(ebx,PW_MF078_F117)] ; mm4=z3L + pmaddwd mm1,[GOTOFF(ebx,PW_MF078_F117)] ; mm1=z3H + pmaddwd mm0,[GOTOFF(ebx,PW_F117_F078)] ; mm0=z4L + pmaddwd mm6,[GOTOFF(ebx,PW_F117_F078)] ; mm6=z4H + + movq MMWORD [wk(0)], mm4 ; wk(0)=z3L + movq MMWORD [wk(1)], mm1 ; wk(1)=z3H + + ; (Original) + ; z1 = tmp4 + tmp7; z2 = tmp5 + tmp6; + ; tmp4 = tmp4 * 0.298631336; tmp5 = tmp5 * 2.053119869; + ; tmp6 = tmp6 * 3.072711026; tmp7 = tmp7 * 1.501321110; + ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; + ; data7 = tmp4 + z1 + z3; data5 = tmp5 + z2 + z4; + ; data3 = tmp6 + z2 + z3; data1 = tmp7 + z1 + z4; + ; + ; (This implementation) + ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223; + ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447; + ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447); + ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223); + ; data7 = tmp4 + z3; data5 = tmp5 + z4; + ; data3 = tmp6 + z3; data1 = tmp7 + z4; + + movq mm4,mm2 + movq mm1,mm2 + punpcklwd mm4,mm7 + punpckhwd mm1,mm7 + movq mm2,mm4 + movq mm7,mm1 + pmaddwd mm4,[GOTOFF(ebx,PW_MF060_MF089)] ; mm4=tmp4L + pmaddwd mm1,[GOTOFF(ebx,PW_MF060_MF089)] ; mm1=tmp4H + pmaddwd mm2,[GOTOFF(ebx,PW_MF089_F060)] ; mm2=tmp7L + pmaddwd mm7,[GOTOFF(ebx,PW_MF089_F060)] ; mm7=tmp7H + + paddd mm4, MMWORD [wk(0)] ; mm4=data7L + paddd mm1, MMWORD [wk(1)] ; mm1=data7H + paddd mm2,mm0 ; mm2=data1L + paddd mm7,mm6 ; mm7=data1H + + paddd mm4,[GOTOFF(ebx,PD_DESCALE_P1)] + paddd mm1,[GOTOFF(ebx,PD_DESCALE_P1)] + psrad mm4,DESCALE_P1 + psrad mm1,DESCALE_P1 + paddd mm2,[GOTOFF(ebx,PD_DESCALE_P1)] + paddd mm7,[GOTOFF(ebx,PD_DESCALE_P1)] + psrad mm2,DESCALE_P1 + psrad mm7,DESCALE_P1 + + packssdw mm4,mm1 ; mm4=data7 + packssdw mm2,mm7 ; mm2=data1 + + movq MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)], mm4 + movq MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm2 + + movq mm1,mm3 + movq mm7,mm3 + punpcklwd mm1,mm5 + punpckhwd mm7,mm5 + movq mm3,mm1 + movq mm5,mm7 + pmaddwd mm1,[GOTOFF(ebx,PW_MF050_MF256)] ; mm1=tmp5L + pmaddwd mm7,[GOTOFF(ebx,PW_MF050_MF256)] ; mm7=tmp5H + pmaddwd mm3,[GOTOFF(ebx,PW_MF256_F050)] ; mm3=tmp6L + pmaddwd mm5,[GOTOFF(ebx,PW_MF256_F050)] ; mm5=tmp6H + + paddd mm1,mm0 ; mm1=data5L + paddd mm7,mm6 ; mm7=data5H + paddd mm3, MMWORD [wk(0)] ; mm3=data3L + paddd mm5, MMWORD [wk(1)] ; mm5=data3H + + paddd mm1,[GOTOFF(ebx,PD_DESCALE_P1)] + paddd mm7,[GOTOFF(ebx,PD_DESCALE_P1)] + psrad mm1,DESCALE_P1 + psrad mm7,DESCALE_P1 + paddd mm3,[GOTOFF(ebx,PD_DESCALE_P1)] + paddd mm5,[GOTOFF(ebx,PD_DESCALE_P1)] + psrad mm3,DESCALE_P1 + psrad mm5,DESCALE_P1 + + packssdw mm1,mm7 ; mm1=data5 + packssdw mm3,mm5 ; mm3=data3 + + movq MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)], mm1 + movq MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm3 + + add edx, byte 4*DCTSIZE*SIZEOF_DCTELEM + dec ecx + jnz near .rowloop + + ; ---- Pass 2: process columns. + + mov edx, POINTER [data(eax)] ; (DCTELEM *) + mov ecx, DCTSIZE/4 + alignx 16,7 +.columnloop: + + movq mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)] + movq mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)] + movq mm2, MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)] + movq mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)] + + ; mm0=(02 12 22 32), mm2=(42 52 62 72) + ; mm1=(03 13 23 33), mm3=(43 53 63 73) + + movq mm4,mm0 ; transpose coefficients(phase 1) + punpcklwd mm0,mm1 ; mm0=(02 03 12 13) + punpckhwd mm4,mm1 ; mm4=(22 23 32 33) + movq mm5,mm2 ; transpose coefficients(phase 1) + punpcklwd mm2,mm3 ; mm2=(42 43 52 53) + punpckhwd mm5,mm3 ; mm5=(62 63 72 73) + + movq mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)] + movq mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)] + movq mm1, MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)] + movq mm3, MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)] + + ; mm6=(00 10 20 30), mm1=(40 50 60 70) + ; mm7=(01 11 21 31), mm3=(41 51 61 71) + + movq MMWORD [wk(0)], mm4 ; wk(0)=(22 23 32 33) + movq MMWORD [wk(1)], mm2 ; wk(1)=(42 43 52 53) + + movq mm4,mm6 ; transpose coefficients(phase 1) + punpcklwd mm6,mm7 ; mm6=(00 01 10 11) + punpckhwd mm4,mm7 ; mm4=(20 21 30 31) + movq mm2,mm1 ; transpose coefficients(phase 1) + punpcklwd mm1,mm3 ; mm1=(40 41 50 51) + punpckhwd mm2,mm3 ; mm2=(60 61 70 71) + + movq mm7,mm6 ; transpose coefficients(phase 2) + punpckldq mm6,mm0 ; mm6=(00 01 02 03)=data0 + punpckhdq mm7,mm0 ; mm7=(10 11 12 13)=data1 + movq mm3,mm2 ; transpose coefficients(phase 2) + punpckldq mm2,mm5 ; mm2=(60 61 62 63)=data6 + punpckhdq mm3,mm5 ; mm3=(70 71 72 73)=data7 + + movq mm0,mm7 + movq mm5,mm6 + psubw mm7,mm2 ; mm7=data1-data6=tmp6 + psubw mm6,mm3 ; mm6=data0-data7=tmp7 + paddw mm0,mm2 ; mm0=data1+data6=tmp1 + paddw mm5,mm3 ; mm5=data0+data7=tmp0 + + movq mm2, MMWORD [wk(0)] ; mm2=(22 23 32 33) + movq mm3, MMWORD [wk(1)] ; mm3=(42 43 52 53) + movq MMWORD [wk(0)], mm7 ; wk(0)=tmp6 + movq MMWORD [wk(1)], mm6 ; wk(1)=tmp7 + + movq mm7,mm4 ; transpose coefficients(phase 2) + punpckldq mm4,mm2 ; mm4=(20 21 22 23)=data2 + punpckhdq mm7,mm2 ; mm7=(30 31 32 33)=data3 + movq mm6,mm1 ; transpose coefficients(phase 2) + punpckldq mm1,mm3 ; mm1=(40 41 42 43)=data4 + punpckhdq mm6,mm3 ; mm6=(50 51 52 53)=data5 + + movq mm2,mm7 + movq mm3,mm4 + paddw mm7,mm1 ; mm7=data3+data4=tmp3 + paddw mm4,mm6 ; mm4=data2+data5=tmp2 + psubw mm2,mm1 ; mm2=data3-data4=tmp4 + psubw mm3,mm6 ; mm3=data2-data5=tmp5 + + ; -- Even part + + movq mm1,mm5 + movq mm6,mm0 + paddw mm5,mm7 ; mm5=tmp10 + paddw mm0,mm4 ; mm0=tmp11 + psubw mm1,mm7 ; mm1=tmp13 + psubw mm6,mm4 ; mm6=tmp12 + + movq mm7,mm5 + paddw mm5,mm0 ; mm5=tmp10+tmp11 + psubw mm7,mm0 ; mm7=tmp10-tmp11 + + paddw mm5,[GOTOFF(ebx,PW_DESCALE_P2X)] + paddw mm7,[GOTOFF(ebx,PW_DESCALE_P2X)] + psraw mm5,PASS1_BITS ; mm5=data0 + psraw mm7,PASS1_BITS ; mm7=data4 + + movq MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm5 + movq MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)], mm7 + + ; (Original) + ; z1 = (tmp12 + tmp13) * 0.541196100; + ; data2 = z1 + tmp13 * 0.765366865; + ; data6 = z1 + tmp12 * -1.847759065; + ; + ; (This implementation) + ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100; + ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065); + + movq mm4,mm1 ; mm1=tmp13 + movq mm0,mm1 + punpcklwd mm4,mm6 ; mm6=tmp12 + punpckhwd mm0,mm6 + movq mm1,mm4 + movq mm6,mm0 + pmaddwd mm4,[GOTOFF(ebx,PW_F130_F054)] ; mm4=data2L + pmaddwd mm0,[GOTOFF(ebx,PW_F130_F054)] ; mm0=data2H + pmaddwd mm1,[GOTOFF(ebx,PW_F054_MF130)] ; mm1=data6L + pmaddwd mm6,[GOTOFF(ebx,PW_F054_MF130)] ; mm6=data6H + + paddd mm4,[GOTOFF(ebx,PD_DESCALE_P2)] + paddd mm0,[GOTOFF(ebx,PD_DESCALE_P2)] + psrad mm4,DESCALE_P2 + psrad mm0,DESCALE_P2 + paddd mm1,[GOTOFF(ebx,PD_DESCALE_P2)] + paddd mm6,[GOTOFF(ebx,PD_DESCALE_P2)] + psrad mm1,DESCALE_P2 + psrad mm6,DESCALE_P2 + + packssdw mm4,mm0 ; mm4=data2 + packssdw mm1,mm6 ; mm1=data6 + + movq MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4 + movq MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)], mm1 + + ; -- Odd part + + movq mm5, MMWORD [wk(0)] ; mm5=tmp6 + movq mm7, MMWORD [wk(1)] ; mm7=tmp7 + + movq mm0,mm2 ; mm2=tmp4 + movq mm6,mm3 ; mm3=tmp5 + paddw mm0,mm5 ; mm0=z3 + paddw mm6,mm7 ; mm6=z4 + + ; (Original) + ; z5 = (z3 + z4) * 1.175875602; + ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; + ; z3 += z5; z4 += z5; + ; + ; (This implementation) + ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; + ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); + + movq mm4,mm0 + movq mm1,mm0 + punpcklwd mm4,mm6 + punpckhwd mm1,mm6 + movq mm0,mm4 + movq mm6,mm1 + pmaddwd mm4,[GOTOFF(ebx,PW_MF078_F117)] ; mm4=z3L + pmaddwd mm1,[GOTOFF(ebx,PW_MF078_F117)] ; mm1=z3H + pmaddwd mm0,[GOTOFF(ebx,PW_F117_F078)] ; mm0=z4L + pmaddwd mm6,[GOTOFF(ebx,PW_F117_F078)] ; mm6=z4H + + movq MMWORD [wk(0)], mm4 ; wk(0)=z3L + movq MMWORD [wk(1)], mm1 ; wk(1)=z3H + + ; (Original) + ; z1 = tmp4 + tmp7; z2 = tmp5 + tmp6; + ; tmp4 = tmp4 * 0.298631336; tmp5 = tmp5 * 2.053119869; + ; tmp6 = tmp6 * 3.072711026; tmp7 = tmp7 * 1.501321110; + ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; + ; data7 = tmp4 + z1 + z3; data5 = tmp5 + z2 + z4; + ; data3 = tmp6 + z2 + z3; data1 = tmp7 + z1 + z4; + ; + ; (This implementation) + ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223; + ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447; + ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447); + ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223); + ; data7 = tmp4 + z3; data5 = tmp5 + z4; + ; data3 = tmp6 + z3; data1 = tmp7 + z4; + + movq mm4,mm2 + movq mm1,mm2 + punpcklwd mm4,mm7 + punpckhwd mm1,mm7 + movq mm2,mm4 + movq mm7,mm1 + pmaddwd mm4,[GOTOFF(ebx,PW_MF060_MF089)] ; mm4=tmp4L + pmaddwd mm1,[GOTOFF(ebx,PW_MF060_MF089)] ; mm1=tmp4H + pmaddwd mm2,[GOTOFF(ebx,PW_MF089_F060)] ; mm2=tmp7L + pmaddwd mm7,[GOTOFF(ebx,PW_MF089_F060)] ; mm7=tmp7H + + paddd mm4, MMWORD [wk(0)] ; mm4=data7L + paddd mm1, MMWORD [wk(1)] ; mm1=data7H + paddd mm2,mm0 ; mm2=data1L + paddd mm7,mm6 ; mm7=data1H + + paddd mm4,[GOTOFF(ebx,PD_DESCALE_P2)] + paddd mm1,[GOTOFF(ebx,PD_DESCALE_P2)] + psrad mm4,DESCALE_P2 + psrad mm1,DESCALE_P2 + paddd mm2,[GOTOFF(ebx,PD_DESCALE_P2)] + paddd mm7,[GOTOFF(ebx,PD_DESCALE_P2)] + psrad mm2,DESCALE_P2 + psrad mm7,DESCALE_P2 + + packssdw mm4,mm1 ; mm4=data7 + packssdw mm2,mm7 ; mm2=data1 + + movq MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)], mm4 + movq MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm2 + + movq mm1,mm3 + movq mm7,mm3 + punpcklwd mm1,mm5 + punpckhwd mm7,mm5 + movq mm3,mm1 + movq mm5,mm7 + pmaddwd mm1,[GOTOFF(ebx,PW_MF050_MF256)] ; mm1=tmp5L + pmaddwd mm7,[GOTOFF(ebx,PW_MF050_MF256)] ; mm7=tmp5H + pmaddwd mm3,[GOTOFF(ebx,PW_MF256_F050)] ; mm3=tmp6L + pmaddwd mm5,[GOTOFF(ebx,PW_MF256_F050)] ; mm5=tmp6H + + paddd mm1,mm0 ; mm1=data5L + paddd mm7,mm6 ; mm7=data5H + paddd mm3, MMWORD [wk(0)] ; mm3=data3L + paddd mm5, MMWORD [wk(1)] ; mm5=data3H + + paddd mm1,[GOTOFF(ebx,PD_DESCALE_P2)] + paddd mm7,[GOTOFF(ebx,PD_DESCALE_P2)] + psrad mm1,DESCALE_P2 + psrad mm7,DESCALE_P2 + paddd mm3,[GOTOFF(ebx,PD_DESCALE_P2)] + paddd mm5,[GOTOFF(ebx,PD_DESCALE_P2)] + psrad mm3,DESCALE_P2 + psrad mm5,DESCALE_P2 + + packssdw mm1,mm7 ; mm1=data5 + packssdw mm3,mm5 ; mm3=data3 + + movq MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)], mm1 + movq MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm3 + + add edx, byte 4*SIZEOF_DCTELEM + dec ecx + jnz near .columnloop + + emms ; empty MMX state + +; pop edi ; unused +; pop esi ; unused +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + poppic ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jfdctint-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jfdctint-sse2-64.asm new file mode 100644 index 000000000..53ec52769 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jfdctint-sse2-64.asm @@ -0,0 +1,622 @@ +; +; jfdctint.asm - accurate integer FDCT (64-bit SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a slow-but-accurate integer implementation of the +; forward DCT (Discrete Cosine Transform). The following code is based +; directly on the IJG's original jfdctint.c; see the jfdctint.c for +; more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 13 +%define PASS1_BITS 2 + +%define DESCALE_P1 (CONST_BITS-PASS1_BITS) +%define DESCALE_P2 (CONST_BITS+PASS1_BITS) + +%if CONST_BITS == 13 +F_0_298 equ 2446 ; FIX(0.298631336) +F_0_390 equ 3196 ; FIX(0.390180644) +F_0_541 equ 4433 ; FIX(0.541196100) +F_0_765 equ 6270 ; FIX(0.765366865) +F_0_899 equ 7373 ; FIX(0.899976223) +F_1_175 equ 9633 ; FIX(1.175875602) +F_1_501 equ 12299 ; FIX(1.501321110) +F_1_847 equ 15137 ; FIX(1.847759065) +F_1_961 equ 16069 ; FIX(1.961570560) +F_2_053 equ 16819 ; FIX(2.053119869) +F_2_562 equ 20995 ; FIX(2.562915447) +F_3_072 equ 25172 ; FIX(3.072711026) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_0_298 equ DESCALE( 320652955,30-CONST_BITS) ; FIX(0.298631336) +F_0_390 equ DESCALE( 418953276,30-CONST_BITS) ; FIX(0.390180644) +F_0_541 equ DESCALE( 581104887,30-CONST_BITS) ; FIX(0.541196100) +F_0_765 equ DESCALE( 821806413,30-CONST_BITS) ; FIX(0.765366865) +F_0_899 equ DESCALE( 966342111,30-CONST_BITS) ; FIX(0.899976223) +F_1_175 equ DESCALE(1262586813,30-CONST_BITS) ; FIX(1.175875602) +F_1_501 equ DESCALE(1612031267,30-CONST_BITS) ; FIX(1.501321110) +F_1_847 equ DESCALE(1984016188,30-CONST_BITS) ; FIX(1.847759065) +F_1_961 equ DESCALE(2106220350,30-CONST_BITS) ; FIX(1.961570560) +F_2_053 equ DESCALE(2204520673,30-CONST_BITS) ; FIX(2.053119869) +F_2_562 equ DESCALE(2751909506,30-CONST_BITS) ; FIX(2.562915447) +F_3_072 equ DESCALE(3299298341,30-CONST_BITS) ; FIX(3.072711026) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_fdct_islow_sse2) + +EXTN(jconst_fdct_islow_sse2): + +PW_F130_F054 times 4 dw (F_0_541+F_0_765), F_0_541 +PW_F054_MF130 times 4 dw F_0_541, (F_0_541-F_1_847) +PW_MF078_F117 times 4 dw (F_1_175-F_1_961), F_1_175 +PW_F117_F078 times 4 dw F_1_175, (F_1_175-F_0_390) +PW_MF060_MF089 times 4 dw (F_0_298-F_0_899),-F_0_899 +PW_MF089_F060 times 4 dw -F_0_899, (F_1_501-F_0_899) +PW_MF050_MF256 times 4 dw (F_2_053-F_2_562),-F_2_562 +PW_MF256_F050 times 4 dw -F_2_562, (F_3_072-F_2_562) +PD_DESCALE_P1 times 4 dd 1 << (DESCALE_P1-1) +PD_DESCALE_P2 times 4 dd 1 << (DESCALE_P2-1) +PW_DESCALE_P2X times 8 dw 1 << (PASS1_BITS-1) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 +; +; Perform the forward DCT on one block of samples. +; +; GLOBAL(void) +; jsimd_fdct_islow_sse2 (DCTELEM * data) +; + +; r10 = DCTELEM * data + +%define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 6 + + align 16 + global EXTN(jsimd_fdct_islow_sse2) + +EXTN(jsimd_fdct_islow_sse2): + push rbp + mov rax,rsp ; rax = original rbp + sub rsp, byte 4 + and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [rsp],rax + mov rbp,rsp ; rbp = aligned rbp + lea rsp, [wk(0)] + collect_args + + ; ---- Pass 1: process rows. + + mov rdx, r10 ; (DCTELEM *) + + movdqa xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)] + movdqa xmm1, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)] + movdqa xmm2, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)] + movdqa xmm3, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)] + + ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27) + ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37) + + movdqa xmm4,xmm0 ; transpose coefficients(phase 1) + punpcklwd xmm0,xmm1 ; xmm0=(00 10 01 11 02 12 03 13) + punpckhwd xmm4,xmm1 ; xmm4=(04 14 05 15 06 16 07 17) + movdqa xmm5,xmm2 ; transpose coefficients(phase 1) + punpcklwd xmm2,xmm3 ; xmm2=(20 30 21 31 22 32 23 33) + punpckhwd xmm5,xmm3 ; xmm5=(24 34 25 35 26 36 27 37) + + movdqa xmm6, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)] + movdqa xmm7, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)] + movdqa xmm1, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)] + movdqa xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)] + + ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62) + ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63) + + movdqa XMMWORD [wk(0)], xmm2 ; wk(0)=(20 30 21 31 22 32 23 33) + movdqa XMMWORD [wk(1)], xmm5 ; wk(1)=(24 34 25 35 26 36 27 37) + + movdqa xmm2,xmm6 ; transpose coefficients(phase 1) + punpcklwd xmm6,xmm7 ; xmm6=(40 50 41 51 42 52 43 53) + punpckhwd xmm2,xmm7 ; xmm2=(44 54 45 55 46 56 47 57) + movdqa xmm5,xmm1 ; transpose coefficients(phase 1) + punpcklwd xmm1,xmm3 ; xmm1=(60 70 61 71 62 72 63 73) + punpckhwd xmm5,xmm3 ; xmm5=(64 74 65 75 66 76 67 77) + + movdqa xmm7,xmm6 ; transpose coefficients(phase 2) + punpckldq xmm6,xmm1 ; xmm6=(40 50 60 70 41 51 61 71) + punpckhdq xmm7,xmm1 ; xmm7=(42 52 62 72 43 53 63 73) + movdqa xmm3,xmm2 ; transpose coefficients(phase 2) + punpckldq xmm2,xmm5 ; xmm2=(44 54 64 74 45 55 65 75) + punpckhdq xmm3,xmm5 ; xmm3=(46 56 66 76 47 57 67 77) + + movdqa xmm1, XMMWORD [wk(0)] ; xmm1=(20 30 21 31 22 32 23 33) + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=(24 34 25 35 26 36 27 37) + movdqa XMMWORD [wk(2)], xmm7 ; wk(2)=(42 52 62 72 43 53 63 73) + movdqa XMMWORD [wk(3)], xmm2 ; wk(3)=(44 54 64 74 45 55 65 75) + + movdqa xmm7,xmm0 ; transpose coefficients(phase 2) + punpckldq xmm0,xmm1 ; xmm0=(00 10 20 30 01 11 21 31) + punpckhdq xmm7,xmm1 ; xmm7=(02 12 22 32 03 13 23 33) + movdqa xmm2,xmm4 ; transpose coefficients(phase 2) + punpckldq xmm4,xmm5 ; xmm4=(04 14 24 34 05 15 25 35) + punpckhdq xmm2,xmm5 ; xmm2=(06 16 26 36 07 17 27 37) + + movdqa xmm1,xmm0 ; transpose coefficients(phase 3) + punpcklqdq xmm0,xmm6 ; xmm0=(00 10 20 30 40 50 60 70)=data0 + punpckhqdq xmm1,xmm6 ; xmm1=(01 11 21 31 41 51 61 71)=data1 + movdqa xmm5,xmm2 ; transpose coefficients(phase 3) + punpcklqdq xmm2,xmm3 ; xmm2=(06 16 26 36 46 56 66 76)=data6 + punpckhqdq xmm5,xmm3 ; xmm5=(07 17 27 37 47 57 67 77)=data7 + + movdqa xmm6,xmm1 + movdqa xmm3,xmm0 + psubw xmm1,xmm2 ; xmm1=data1-data6=tmp6 + psubw xmm0,xmm5 ; xmm0=data0-data7=tmp7 + paddw xmm6,xmm2 ; xmm6=data1+data6=tmp1 + paddw xmm3,xmm5 ; xmm3=data0+data7=tmp0 + + movdqa xmm2, XMMWORD [wk(2)] ; xmm2=(42 52 62 72 43 53 63 73) + movdqa xmm5, XMMWORD [wk(3)] ; xmm5=(44 54 64 74 45 55 65 75) + movdqa XMMWORD [wk(0)], xmm1 ; wk(0)=tmp6 + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=tmp7 + + movdqa xmm1,xmm7 ; transpose coefficients(phase 3) + punpcklqdq xmm7,xmm2 ; xmm7=(02 12 22 32 42 52 62 72)=data2 + punpckhqdq xmm1,xmm2 ; xmm1=(03 13 23 33 43 53 63 73)=data3 + movdqa xmm0,xmm4 ; transpose coefficients(phase 3) + punpcklqdq xmm4,xmm5 ; xmm4=(04 14 24 34 44 54 64 74)=data4 + punpckhqdq xmm0,xmm5 ; xmm0=(05 15 25 35 45 55 65 75)=data5 + + movdqa xmm2,xmm1 + movdqa xmm5,xmm7 + paddw xmm1,xmm4 ; xmm1=data3+data4=tmp3 + paddw xmm7,xmm0 ; xmm7=data2+data5=tmp2 + psubw xmm2,xmm4 ; xmm2=data3-data4=tmp4 + psubw xmm5,xmm0 ; xmm5=data2-data5=tmp5 + + ; -- Even part + + movdqa xmm4,xmm3 + movdqa xmm0,xmm6 + paddw xmm3,xmm1 ; xmm3=tmp10 + paddw xmm6,xmm7 ; xmm6=tmp11 + psubw xmm4,xmm1 ; xmm4=tmp13 + psubw xmm0,xmm7 ; xmm0=tmp12 + + movdqa xmm1,xmm3 + paddw xmm3,xmm6 ; xmm3=tmp10+tmp11 + psubw xmm1,xmm6 ; xmm1=tmp10-tmp11 + + psllw xmm3,PASS1_BITS ; xmm3=data0 + psllw xmm1,PASS1_BITS ; xmm1=data4 + + movdqa XMMWORD [wk(2)], xmm3 ; wk(2)=data0 + movdqa XMMWORD [wk(3)], xmm1 ; wk(3)=data4 + + ; (Original) + ; z1 = (tmp12 + tmp13) * 0.541196100; + ; data2 = z1 + tmp13 * 0.765366865; + ; data6 = z1 + tmp12 * -1.847759065; + ; + ; (This implementation) + ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100; + ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065); + + movdqa xmm7,xmm4 ; xmm4=tmp13 + movdqa xmm6,xmm4 + punpcklwd xmm7,xmm0 ; xmm0=tmp12 + punpckhwd xmm6,xmm0 + movdqa xmm4,xmm7 + movdqa xmm0,xmm6 + pmaddwd xmm7,[rel PW_F130_F054] ; xmm7=data2L + pmaddwd xmm6,[rel PW_F130_F054] ; xmm6=data2H + pmaddwd xmm4,[rel PW_F054_MF130] ; xmm4=data6L + pmaddwd xmm0,[rel PW_F054_MF130] ; xmm0=data6H + + paddd xmm7,[rel PD_DESCALE_P1] + paddd xmm6,[rel PD_DESCALE_P1] + psrad xmm7,DESCALE_P1 + psrad xmm6,DESCALE_P1 + paddd xmm4,[rel PD_DESCALE_P1] + paddd xmm0,[rel PD_DESCALE_P1] + psrad xmm4,DESCALE_P1 + psrad xmm0,DESCALE_P1 + + packssdw xmm7,xmm6 ; xmm7=data2 + packssdw xmm4,xmm0 ; xmm4=data6 + + movdqa XMMWORD [wk(4)], xmm7 ; wk(4)=data2 + movdqa XMMWORD [wk(5)], xmm4 ; wk(5)=data6 + + ; -- Odd part + + movdqa xmm3, XMMWORD [wk(0)] ; xmm3=tmp6 + movdqa xmm1, XMMWORD [wk(1)] ; xmm1=tmp7 + + movdqa xmm6,xmm2 ; xmm2=tmp4 + movdqa xmm0,xmm5 ; xmm5=tmp5 + paddw xmm6,xmm3 ; xmm6=z3 + paddw xmm0,xmm1 ; xmm0=z4 + + ; (Original) + ; z5 = (z3 + z4) * 1.175875602; + ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; + ; z3 += z5; z4 += z5; + ; + ; (This implementation) + ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; + ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); + + movdqa xmm7,xmm6 + movdqa xmm4,xmm6 + punpcklwd xmm7,xmm0 + punpckhwd xmm4,xmm0 + movdqa xmm6,xmm7 + movdqa xmm0,xmm4 + pmaddwd xmm7,[rel PW_MF078_F117] ; xmm7=z3L + pmaddwd xmm4,[rel PW_MF078_F117] ; xmm4=z3H + pmaddwd xmm6,[rel PW_F117_F078] ; xmm6=z4L + pmaddwd xmm0,[rel PW_F117_F078] ; xmm0=z4H + + movdqa XMMWORD [wk(0)], xmm7 ; wk(0)=z3L + movdqa XMMWORD [wk(1)], xmm4 ; wk(1)=z3H + + ; (Original) + ; z1 = tmp4 + tmp7; z2 = tmp5 + tmp6; + ; tmp4 = tmp4 * 0.298631336; tmp5 = tmp5 * 2.053119869; + ; tmp6 = tmp6 * 3.072711026; tmp7 = tmp7 * 1.501321110; + ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; + ; data7 = tmp4 + z1 + z3; data5 = tmp5 + z2 + z4; + ; data3 = tmp6 + z2 + z3; data1 = tmp7 + z1 + z4; + ; + ; (This implementation) + ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223; + ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447; + ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447); + ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223); + ; data7 = tmp4 + z3; data5 = tmp5 + z4; + ; data3 = tmp6 + z3; data1 = tmp7 + z4; + + movdqa xmm7,xmm2 + movdqa xmm4,xmm2 + punpcklwd xmm7,xmm1 + punpckhwd xmm4,xmm1 + movdqa xmm2,xmm7 + movdqa xmm1,xmm4 + pmaddwd xmm7,[rel PW_MF060_MF089] ; xmm7=tmp4L + pmaddwd xmm4,[rel PW_MF060_MF089] ; xmm4=tmp4H + pmaddwd xmm2,[rel PW_MF089_F060] ; xmm2=tmp7L + pmaddwd xmm1,[rel PW_MF089_F060] ; xmm1=tmp7H + + paddd xmm7, XMMWORD [wk(0)] ; xmm7=data7L + paddd xmm4, XMMWORD [wk(1)] ; xmm4=data7H + paddd xmm2,xmm6 ; xmm2=data1L + paddd xmm1,xmm0 ; xmm1=data1H + + paddd xmm7,[rel PD_DESCALE_P1] + paddd xmm4,[rel PD_DESCALE_P1] + psrad xmm7,DESCALE_P1 + psrad xmm4,DESCALE_P1 + paddd xmm2,[rel PD_DESCALE_P1] + paddd xmm1,[rel PD_DESCALE_P1] + psrad xmm2,DESCALE_P1 + psrad xmm1,DESCALE_P1 + + packssdw xmm7,xmm4 ; xmm7=data7 + packssdw xmm2,xmm1 ; xmm2=data1 + + movdqa xmm4,xmm5 + movdqa xmm1,xmm5 + punpcklwd xmm4,xmm3 + punpckhwd xmm1,xmm3 + movdqa xmm5,xmm4 + movdqa xmm3,xmm1 + pmaddwd xmm4,[rel PW_MF050_MF256] ; xmm4=tmp5L + pmaddwd xmm1,[rel PW_MF050_MF256] ; xmm1=tmp5H + pmaddwd xmm5,[rel PW_MF256_F050] ; xmm5=tmp6L + pmaddwd xmm3,[rel PW_MF256_F050] ; xmm3=tmp6H + + paddd xmm4,xmm6 ; xmm4=data5L + paddd xmm1,xmm0 ; xmm1=data5H + paddd xmm5, XMMWORD [wk(0)] ; xmm5=data3L + paddd xmm3, XMMWORD [wk(1)] ; xmm3=data3H + + paddd xmm4,[rel PD_DESCALE_P1] + paddd xmm1,[rel PD_DESCALE_P1] + psrad xmm4,DESCALE_P1 + psrad xmm1,DESCALE_P1 + paddd xmm5,[rel PD_DESCALE_P1] + paddd xmm3,[rel PD_DESCALE_P1] + psrad xmm5,DESCALE_P1 + psrad xmm3,DESCALE_P1 + + packssdw xmm4,xmm1 ; xmm4=data5 + packssdw xmm5,xmm3 ; xmm5=data3 + + ; ---- Pass 2: process columns. + + movdqa xmm6, XMMWORD [wk(2)] ; xmm6=col0 + movdqa xmm0, XMMWORD [wk(4)] ; xmm0=col2 + + ; xmm6=(00 10 20 30 40 50 60 70), xmm0=(02 12 22 32 42 52 62 72) + ; xmm2=(01 11 21 31 41 51 61 71), xmm5=(03 13 23 33 43 53 63 73) + + movdqa xmm1,xmm6 ; transpose coefficients(phase 1) + punpcklwd xmm6,xmm2 ; xmm6=(00 01 10 11 20 21 30 31) + punpckhwd xmm1,xmm2 ; xmm1=(40 41 50 51 60 61 70 71) + movdqa xmm3,xmm0 ; transpose coefficients(phase 1) + punpcklwd xmm0,xmm5 ; xmm0=(02 03 12 13 22 23 32 33) + punpckhwd xmm3,xmm5 ; xmm3=(42 43 52 53 62 63 72 73) + + movdqa xmm2, XMMWORD [wk(3)] ; xmm2=col4 + movdqa xmm5, XMMWORD [wk(5)] ; xmm5=col6 + + ; xmm2=(04 14 24 34 44 54 64 74), xmm5=(06 16 26 36 46 56 66 76) + ; xmm4=(05 15 25 35 45 55 65 75), xmm7=(07 17 27 37 47 57 67 77) + + movdqa XMMWORD [wk(0)], xmm0 ; wk(0)=(02 03 12 13 22 23 32 33) + movdqa XMMWORD [wk(1)], xmm3 ; wk(1)=(42 43 52 53 62 63 72 73) + + movdqa xmm0,xmm2 ; transpose coefficients(phase 1) + punpcklwd xmm2,xmm4 ; xmm2=(04 05 14 15 24 25 34 35) + punpckhwd xmm0,xmm4 ; xmm0=(44 45 54 55 64 65 74 75) + movdqa xmm3,xmm5 ; transpose coefficients(phase 1) + punpcklwd xmm5,xmm7 ; xmm5=(06 07 16 17 26 27 36 37) + punpckhwd xmm3,xmm7 ; xmm3=(46 47 56 57 66 67 76 77) + + movdqa xmm4,xmm2 ; transpose coefficients(phase 2) + punpckldq xmm2,xmm5 ; xmm2=(04 05 06 07 14 15 16 17) + punpckhdq xmm4,xmm5 ; xmm4=(24 25 26 27 34 35 36 37) + movdqa xmm7,xmm0 ; transpose coefficients(phase 2) + punpckldq xmm0,xmm3 ; xmm0=(44 45 46 47 54 55 56 57) + punpckhdq xmm7,xmm3 ; xmm7=(64 65 66 67 74 75 76 77) + + movdqa xmm5, XMMWORD [wk(0)] ; xmm5=(02 03 12 13 22 23 32 33) + movdqa xmm3, XMMWORD [wk(1)] ; xmm3=(42 43 52 53 62 63 72 73) + movdqa XMMWORD [wk(2)], xmm4 ; wk(2)=(24 25 26 27 34 35 36 37) + movdqa XMMWORD [wk(3)], xmm0 ; wk(3)=(44 45 46 47 54 55 56 57) + + movdqa xmm4,xmm6 ; transpose coefficients(phase 2) + punpckldq xmm6,xmm5 ; xmm6=(00 01 02 03 10 11 12 13) + punpckhdq xmm4,xmm5 ; xmm4=(20 21 22 23 30 31 32 33) + movdqa xmm0,xmm1 ; transpose coefficients(phase 2) + punpckldq xmm1,xmm3 ; xmm1=(40 41 42 43 50 51 52 53) + punpckhdq xmm0,xmm3 ; xmm0=(60 61 62 63 70 71 72 73) + + movdqa xmm5,xmm6 ; transpose coefficients(phase 3) + punpcklqdq xmm6,xmm2 ; xmm6=(00 01 02 03 04 05 06 07)=data0 + punpckhqdq xmm5,xmm2 ; xmm5=(10 11 12 13 14 15 16 17)=data1 + movdqa xmm3,xmm0 ; transpose coefficients(phase 3) + punpcklqdq xmm0,xmm7 ; xmm0=(60 61 62 63 64 65 66 67)=data6 + punpckhqdq xmm3,xmm7 ; xmm3=(70 71 72 73 74 75 76 77)=data7 + + movdqa xmm2,xmm5 + movdqa xmm7,xmm6 + psubw xmm5,xmm0 ; xmm5=data1-data6=tmp6 + psubw xmm6,xmm3 ; xmm6=data0-data7=tmp7 + paddw xmm2,xmm0 ; xmm2=data1+data6=tmp1 + paddw xmm7,xmm3 ; xmm7=data0+data7=tmp0 + + movdqa xmm0, XMMWORD [wk(2)] ; xmm0=(24 25 26 27 34 35 36 37) + movdqa xmm3, XMMWORD [wk(3)] ; xmm3=(44 45 46 47 54 55 56 57) + movdqa XMMWORD [wk(0)], xmm5 ; wk(0)=tmp6 + movdqa XMMWORD [wk(1)], xmm6 ; wk(1)=tmp7 + + movdqa xmm5,xmm4 ; transpose coefficients(phase 3) + punpcklqdq xmm4,xmm0 ; xmm4=(20 21 22 23 24 25 26 27)=data2 + punpckhqdq xmm5,xmm0 ; xmm5=(30 31 32 33 34 35 36 37)=data3 + movdqa xmm6,xmm1 ; transpose coefficients(phase 3) + punpcklqdq xmm1,xmm3 ; xmm1=(40 41 42 43 44 45 46 47)=data4 + punpckhqdq xmm6,xmm3 ; xmm6=(50 51 52 53 54 55 56 57)=data5 + + movdqa xmm0,xmm5 + movdqa xmm3,xmm4 + paddw xmm5,xmm1 ; xmm5=data3+data4=tmp3 + paddw xmm4,xmm6 ; xmm4=data2+data5=tmp2 + psubw xmm0,xmm1 ; xmm0=data3-data4=tmp4 + psubw xmm3,xmm6 ; xmm3=data2-data5=tmp5 + + ; -- Even part + + movdqa xmm1,xmm7 + movdqa xmm6,xmm2 + paddw xmm7,xmm5 ; xmm7=tmp10 + paddw xmm2,xmm4 ; xmm2=tmp11 + psubw xmm1,xmm5 ; xmm1=tmp13 + psubw xmm6,xmm4 ; xmm6=tmp12 + + movdqa xmm5,xmm7 + paddw xmm7,xmm2 ; xmm7=tmp10+tmp11 + psubw xmm5,xmm2 ; xmm5=tmp10-tmp11 + + paddw xmm7,[rel PW_DESCALE_P2X] + paddw xmm5,[rel PW_DESCALE_P2X] + psraw xmm7,PASS1_BITS ; xmm7=data0 + psraw xmm5,PASS1_BITS ; xmm5=data4 + + movdqa XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)], xmm7 + movdqa XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)], xmm5 + + ; (Original) + ; z1 = (tmp12 + tmp13) * 0.541196100; + ; data2 = z1 + tmp13 * 0.765366865; + ; data6 = z1 + tmp12 * -1.847759065; + ; + ; (This implementation) + ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100; + ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065); + + movdqa xmm4,xmm1 ; xmm1=tmp13 + movdqa xmm2,xmm1 + punpcklwd xmm4,xmm6 ; xmm6=tmp12 + punpckhwd xmm2,xmm6 + movdqa xmm1,xmm4 + movdqa xmm6,xmm2 + pmaddwd xmm4,[rel PW_F130_F054] ; xmm4=data2L + pmaddwd xmm2,[rel PW_F130_F054] ; xmm2=data2H + pmaddwd xmm1,[rel PW_F054_MF130] ; xmm1=data6L + pmaddwd xmm6,[rel PW_F054_MF130] ; xmm6=data6H + + paddd xmm4,[rel PD_DESCALE_P2] + paddd xmm2,[rel PD_DESCALE_P2] + psrad xmm4,DESCALE_P2 + psrad xmm2,DESCALE_P2 + paddd xmm1,[rel PD_DESCALE_P2] + paddd xmm6,[rel PD_DESCALE_P2] + psrad xmm1,DESCALE_P2 + psrad xmm6,DESCALE_P2 + + packssdw xmm4,xmm2 ; xmm4=data2 + packssdw xmm1,xmm6 ; xmm1=data6 + + movdqa XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)], xmm4 + movdqa XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)], xmm1 + + ; -- Odd part + + movdqa xmm7, XMMWORD [wk(0)] ; xmm7=tmp6 + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=tmp7 + + movdqa xmm2,xmm0 ; xmm0=tmp4 + movdqa xmm6,xmm3 ; xmm3=tmp5 + paddw xmm2,xmm7 ; xmm2=z3 + paddw xmm6,xmm5 ; xmm6=z4 + + ; (Original) + ; z5 = (z3 + z4) * 1.175875602; + ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; + ; z3 += z5; z4 += z5; + ; + ; (This implementation) + ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; + ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); + + movdqa xmm4,xmm2 + movdqa xmm1,xmm2 + punpcklwd xmm4,xmm6 + punpckhwd xmm1,xmm6 + movdqa xmm2,xmm4 + movdqa xmm6,xmm1 + pmaddwd xmm4,[rel PW_MF078_F117] ; xmm4=z3L + pmaddwd xmm1,[rel PW_MF078_F117] ; xmm1=z3H + pmaddwd xmm2,[rel PW_F117_F078] ; xmm2=z4L + pmaddwd xmm6,[rel PW_F117_F078] ; xmm6=z4H + + movdqa XMMWORD [wk(0)], xmm4 ; wk(0)=z3L + movdqa XMMWORD [wk(1)], xmm1 ; wk(1)=z3H + + ; (Original) + ; z1 = tmp4 + tmp7; z2 = tmp5 + tmp6; + ; tmp4 = tmp4 * 0.298631336; tmp5 = tmp5 * 2.053119869; + ; tmp6 = tmp6 * 3.072711026; tmp7 = tmp7 * 1.501321110; + ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; + ; data7 = tmp4 + z1 + z3; data5 = tmp5 + z2 + z4; + ; data3 = tmp6 + z2 + z3; data1 = tmp7 + z1 + z4; + ; + ; (This implementation) + ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223; + ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447; + ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447); + ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223); + ; data7 = tmp4 + z3; data5 = tmp5 + z4; + ; data3 = tmp6 + z3; data1 = tmp7 + z4; + + movdqa xmm4,xmm0 + movdqa xmm1,xmm0 + punpcklwd xmm4,xmm5 + punpckhwd xmm1,xmm5 + movdqa xmm0,xmm4 + movdqa xmm5,xmm1 + pmaddwd xmm4,[rel PW_MF060_MF089] ; xmm4=tmp4L + pmaddwd xmm1,[rel PW_MF060_MF089] ; xmm1=tmp4H + pmaddwd xmm0,[rel PW_MF089_F060] ; xmm0=tmp7L + pmaddwd xmm5,[rel PW_MF089_F060] ; xmm5=tmp7H + + paddd xmm4, XMMWORD [wk(0)] ; xmm4=data7L + paddd xmm1, XMMWORD [wk(1)] ; xmm1=data7H + paddd xmm0,xmm2 ; xmm0=data1L + paddd xmm5,xmm6 ; xmm5=data1H + + paddd xmm4,[rel PD_DESCALE_P2] + paddd xmm1,[rel PD_DESCALE_P2] + psrad xmm4,DESCALE_P2 + psrad xmm1,DESCALE_P2 + paddd xmm0,[rel PD_DESCALE_P2] + paddd xmm5,[rel PD_DESCALE_P2] + psrad xmm0,DESCALE_P2 + psrad xmm5,DESCALE_P2 + + packssdw xmm4,xmm1 ; xmm4=data7 + packssdw xmm0,xmm5 ; xmm0=data1 + + movdqa XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)], xmm4 + movdqa XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)], xmm0 + + movdqa xmm1,xmm3 + movdqa xmm5,xmm3 + punpcklwd xmm1,xmm7 + punpckhwd xmm5,xmm7 + movdqa xmm3,xmm1 + movdqa xmm7,xmm5 + pmaddwd xmm1,[rel PW_MF050_MF256] ; xmm1=tmp5L + pmaddwd xmm5,[rel PW_MF050_MF256] ; xmm5=tmp5H + pmaddwd xmm3,[rel PW_MF256_F050] ; xmm3=tmp6L + pmaddwd xmm7,[rel PW_MF256_F050] ; xmm7=tmp6H + + paddd xmm1,xmm2 ; xmm1=data5L + paddd xmm5,xmm6 ; xmm5=data5H + paddd xmm3, XMMWORD [wk(0)] ; xmm3=data3L + paddd xmm7, XMMWORD [wk(1)] ; xmm7=data3H + + paddd xmm1,[rel PD_DESCALE_P2] + paddd xmm5,[rel PD_DESCALE_P2] + psrad xmm1,DESCALE_P2 + psrad xmm5,DESCALE_P2 + paddd xmm3,[rel PD_DESCALE_P2] + paddd xmm7,[rel PD_DESCALE_P2] + psrad xmm3,DESCALE_P2 + psrad xmm7,DESCALE_P2 + + packssdw xmm1,xmm5 ; xmm1=data5 + packssdw xmm3,xmm7 ; xmm3=data3 + + movdqa XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)], xmm1 + movdqa XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)], xmm3 + + uncollect_args + mov rsp,rbp ; rsp <- aligned rbp + pop rsp ; rsp <- original rbp + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jfdctint-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jfdctint-sse2.asm new file mode 100644 index 000000000..85133f80e --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jfdctint-sse2.asm @@ -0,0 +1,634 @@ +; +; jfdctint.asm - accurate integer FDCT (SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a slow-but-accurate integer implementation of the +; forward DCT (Discrete Cosine Transform). The following code is based +; directly on the IJG's original jfdctint.c; see the jfdctint.c for +; more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 13 +%define PASS1_BITS 2 + +%define DESCALE_P1 (CONST_BITS-PASS1_BITS) +%define DESCALE_P2 (CONST_BITS+PASS1_BITS) + +%if CONST_BITS == 13 +F_0_298 equ 2446 ; FIX(0.298631336) +F_0_390 equ 3196 ; FIX(0.390180644) +F_0_541 equ 4433 ; FIX(0.541196100) +F_0_765 equ 6270 ; FIX(0.765366865) +F_0_899 equ 7373 ; FIX(0.899976223) +F_1_175 equ 9633 ; FIX(1.175875602) +F_1_501 equ 12299 ; FIX(1.501321110) +F_1_847 equ 15137 ; FIX(1.847759065) +F_1_961 equ 16069 ; FIX(1.961570560) +F_2_053 equ 16819 ; FIX(2.053119869) +F_2_562 equ 20995 ; FIX(2.562915447) +F_3_072 equ 25172 ; FIX(3.072711026) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_0_298 equ DESCALE( 320652955,30-CONST_BITS) ; FIX(0.298631336) +F_0_390 equ DESCALE( 418953276,30-CONST_BITS) ; FIX(0.390180644) +F_0_541 equ DESCALE( 581104887,30-CONST_BITS) ; FIX(0.541196100) +F_0_765 equ DESCALE( 821806413,30-CONST_BITS) ; FIX(0.765366865) +F_0_899 equ DESCALE( 966342111,30-CONST_BITS) ; FIX(0.899976223) +F_1_175 equ DESCALE(1262586813,30-CONST_BITS) ; FIX(1.175875602) +F_1_501 equ DESCALE(1612031267,30-CONST_BITS) ; FIX(1.501321110) +F_1_847 equ DESCALE(1984016188,30-CONST_BITS) ; FIX(1.847759065) +F_1_961 equ DESCALE(2106220350,30-CONST_BITS) ; FIX(1.961570560) +F_2_053 equ DESCALE(2204520673,30-CONST_BITS) ; FIX(2.053119869) +F_2_562 equ DESCALE(2751909506,30-CONST_BITS) ; FIX(2.562915447) +F_3_072 equ DESCALE(3299298341,30-CONST_BITS) ; FIX(3.072711026) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_fdct_islow_sse2) + +EXTN(jconst_fdct_islow_sse2): + +PW_F130_F054 times 4 dw (F_0_541+F_0_765), F_0_541 +PW_F054_MF130 times 4 dw F_0_541, (F_0_541-F_1_847) +PW_MF078_F117 times 4 dw (F_1_175-F_1_961), F_1_175 +PW_F117_F078 times 4 dw F_1_175, (F_1_175-F_0_390) +PW_MF060_MF089 times 4 dw (F_0_298-F_0_899),-F_0_899 +PW_MF089_F060 times 4 dw -F_0_899, (F_1_501-F_0_899) +PW_MF050_MF256 times 4 dw (F_2_053-F_2_562),-F_2_562 +PW_MF256_F050 times 4 dw -F_2_562, (F_3_072-F_2_562) +PD_DESCALE_P1 times 4 dd 1 << (DESCALE_P1-1) +PD_DESCALE_P2 times 4 dd 1 << (DESCALE_P2-1) +PW_DESCALE_P2X times 8 dw 1 << (PASS1_BITS-1) + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform the forward DCT on one block of samples. +; +; GLOBAL(void) +; jsimd_fdct_islow_sse2 (DCTELEM * data) +; + +%define data(b) (b)+8 ; DCTELEM * data + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 6 + + align 16 + global EXTN(jsimd_fdct_islow_sse2) + +EXTN(jsimd_fdct_islow_sse2): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic ebx +; push ecx ; unused +; push edx ; need not be preserved +; push esi ; unused +; push edi ; unused + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process rows. + + mov edx, POINTER [data(eax)] ; (DCTELEM *) + + movdqa xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)] + movdqa xmm1, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)] + movdqa xmm2, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)] + movdqa xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)] + + ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27) + ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37) + + movdqa xmm4,xmm0 ; transpose coefficients(phase 1) + punpcklwd xmm0,xmm1 ; xmm0=(00 10 01 11 02 12 03 13) + punpckhwd xmm4,xmm1 ; xmm4=(04 14 05 15 06 16 07 17) + movdqa xmm5,xmm2 ; transpose coefficients(phase 1) + punpcklwd xmm2,xmm3 ; xmm2=(20 30 21 31 22 32 23 33) + punpckhwd xmm5,xmm3 ; xmm5=(24 34 25 35 26 36 27 37) + + movdqa xmm6, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)] + movdqa xmm7, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)] + movdqa xmm1, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)] + movdqa xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)] + + ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62) + ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63) + + movdqa XMMWORD [wk(0)], xmm2 ; wk(0)=(20 30 21 31 22 32 23 33) + movdqa XMMWORD [wk(1)], xmm5 ; wk(1)=(24 34 25 35 26 36 27 37) + + movdqa xmm2,xmm6 ; transpose coefficients(phase 1) + punpcklwd xmm6,xmm7 ; xmm6=(40 50 41 51 42 52 43 53) + punpckhwd xmm2,xmm7 ; xmm2=(44 54 45 55 46 56 47 57) + movdqa xmm5,xmm1 ; transpose coefficients(phase 1) + punpcklwd xmm1,xmm3 ; xmm1=(60 70 61 71 62 72 63 73) + punpckhwd xmm5,xmm3 ; xmm5=(64 74 65 75 66 76 67 77) + + movdqa xmm7,xmm6 ; transpose coefficients(phase 2) + punpckldq xmm6,xmm1 ; xmm6=(40 50 60 70 41 51 61 71) + punpckhdq xmm7,xmm1 ; xmm7=(42 52 62 72 43 53 63 73) + movdqa xmm3,xmm2 ; transpose coefficients(phase 2) + punpckldq xmm2,xmm5 ; xmm2=(44 54 64 74 45 55 65 75) + punpckhdq xmm3,xmm5 ; xmm3=(46 56 66 76 47 57 67 77) + + movdqa xmm1, XMMWORD [wk(0)] ; xmm1=(20 30 21 31 22 32 23 33) + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=(24 34 25 35 26 36 27 37) + movdqa XMMWORD [wk(2)], xmm7 ; wk(2)=(42 52 62 72 43 53 63 73) + movdqa XMMWORD [wk(3)], xmm2 ; wk(3)=(44 54 64 74 45 55 65 75) + + movdqa xmm7,xmm0 ; transpose coefficients(phase 2) + punpckldq xmm0,xmm1 ; xmm0=(00 10 20 30 01 11 21 31) + punpckhdq xmm7,xmm1 ; xmm7=(02 12 22 32 03 13 23 33) + movdqa xmm2,xmm4 ; transpose coefficients(phase 2) + punpckldq xmm4,xmm5 ; xmm4=(04 14 24 34 05 15 25 35) + punpckhdq xmm2,xmm5 ; xmm2=(06 16 26 36 07 17 27 37) + + movdqa xmm1,xmm0 ; transpose coefficients(phase 3) + punpcklqdq xmm0,xmm6 ; xmm0=(00 10 20 30 40 50 60 70)=data0 + punpckhqdq xmm1,xmm6 ; xmm1=(01 11 21 31 41 51 61 71)=data1 + movdqa xmm5,xmm2 ; transpose coefficients(phase 3) + punpcklqdq xmm2,xmm3 ; xmm2=(06 16 26 36 46 56 66 76)=data6 + punpckhqdq xmm5,xmm3 ; xmm5=(07 17 27 37 47 57 67 77)=data7 + + movdqa xmm6,xmm1 + movdqa xmm3,xmm0 + psubw xmm1,xmm2 ; xmm1=data1-data6=tmp6 + psubw xmm0,xmm5 ; xmm0=data0-data7=tmp7 + paddw xmm6,xmm2 ; xmm6=data1+data6=tmp1 + paddw xmm3,xmm5 ; xmm3=data0+data7=tmp0 + + movdqa xmm2, XMMWORD [wk(2)] ; xmm2=(42 52 62 72 43 53 63 73) + movdqa xmm5, XMMWORD [wk(3)] ; xmm5=(44 54 64 74 45 55 65 75) + movdqa XMMWORD [wk(0)], xmm1 ; wk(0)=tmp6 + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=tmp7 + + movdqa xmm1,xmm7 ; transpose coefficients(phase 3) + punpcklqdq xmm7,xmm2 ; xmm7=(02 12 22 32 42 52 62 72)=data2 + punpckhqdq xmm1,xmm2 ; xmm1=(03 13 23 33 43 53 63 73)=data3 + movdqa xmm0,xmm4 ; transpose coefficients(phase 3) + punpcklqdq xmm4,xmm5 ; xmm4=(04 14 24 34 44 54 64 74)=data4 + punpckhqdq xmm0,xmm5 ; xmm0=(05 15 25 35 45 55 65 75)=data5 + + movdqa xmm2,xmm1 + movdqa xmm5,xmm7 + paddw xmm1,xmm4 ; xmm1=data3+data4=tmp3 + paddw xmm7,xmm0 ; xmm7=data2+data5=tmp2 + psubw xmm2,xmm4 ; xmm2=data3-data4=tmp4 + psubw xmm5,xmm0 ; xmm5=data2-data5=tmp5 + + ; -- Even part + + movdqa xmm4,xmm3 + movdqa xmm0,xmm6 + paddw xmm3,xmm1 ; xmm3=tmp10 + paddw xmm6,xmm7 ; xmm6=tmp11 + psubw xmm4,xmm1 ; xmm4=tmp13 + psubw xmm0,xmm7 ; xmm0=tmp12 + + movdqa xmm1,xmm3 + paddw xmm3,xmm6 ; xmm3=tmp10+tmp11 + psubw xmm1,xmm6 ; xmm1=tmp10-tmp11 + + psllw xmm3,PASS1_BITS ; xmm3=data0 + psllw xmm1,PASS1_BITS ; xmm1=data4 + + movdqa XMMWORD [wk(2)], xmm3 ; wk(2)=data0 + movdqa XMMWORD [wk(3)], xmm1 ; wk(3)=data4 + + ; (Original) + ; z1 = (tmp12 + tmp13) * 0.541196100; + ; data2 = z1 + tmp13 * 0.765366865; + ; data6 = z1 + tmp12 * -1.847759065; + ; + ; (This implementation) + ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100; + ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065); + + movdqa xmm7,xmm4 ; xmm4=tmp13 + movdqa xmm6,xmm4 + punpcklwd xmm7,xmm0 ; xmm0=tmp12 + punpckhwd xmm6,xmm0 + movdqa xmm4,xmm7 + movdqa xmm0,xmm6 + pmaddwd xmm7,[GOTOFF(ebx,PW_F130_F054)] ; xmm7=data2L + pmaddwd xmm6,[GOTOFF(ebx,PW_F130_F054)] ; xmm6=data2H + pmaddwd xmm4,[GOTOFF(ebx,PW_F054_MF130)] ; xmm4=data6L + pmaddwd xmm0,[GOTOFF(ebx,PW_F054_MF130)] ; xmm0=data6H + + paddd xmm7,[GOTOFF(ebx,PD_DESCALE_P1)] + paddd xmm6,[GOTOFF(ebx,PD_DESCALE_P1)] + psrad xmm7,DESCALE_P1 + psrad xmm6,DESCALE_P1 + paddd xmm4,[GOTOFF(ebx,PD_DESCALE_P1)] + paddd xmm0,[GOTOFF(ebx,PD_DESCALE_P1)] + psrad xmm4,DESCALE_P1 + psrad xmm0,DESCALE_P1 + + packssdw xmm7,xmm6 ; xmm7=data2 + packssdw xmm4,xmm0 ; xmm4=data6 + + movdqa XMMWORD [wk(4)], xmm7 ; wk(4)=data2 + movdqa XMMWORD [wk(5)], xmm4 ; wk(5)=data6 + + ; -- Odd part + + movdqa xmm3, XMMWORD [wk(0)] ; xmm3=tmp6 + movdqa xmm1, XMMWORD [wk(1)] ; xmm1=tmp7 + + movdqa xmm6,xmm2 ; xmm2=tmp4 + movdqa xmm0,xmm5 ; xmm5=tmp5 + paddw xmm6,xmm3 ; xmm6=z3 + paddw xmm0,xmm1 ; xmm0=z4 + + ; (Original) + ; z5 = (z3 + z4) * 1.175875602; + ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; + ; z3 += z5; z4 += z5; + ; + ; (This implementation) + ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; + ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); + + movdqa xmm7,xmm6 + movdqa xmm4,xmm6 + punpcklwd xmm7,xmm0 + punpckhwd xmm4,xmm0 + movdqa xmm6,xmm7 + movdqa xmm0,xmm4 + pmaddwd xmm7,[GOTOFF(ebx,PW_MF078_F117)] ; xmm7=z3L + pmaddwd xmm4,[GOTOFF(ebx,PW_MF078_F117)] ; xmm4=z3H + pmaddwd xmm6,[GOTOFF(ebx,PW_F117_F078)] ; xmm6=z4L + pmaddwd xmm0,[GOTOFF(ebx,PW_F117_F078)] ; xmm0=z4H + + movdqa XMMWORD [wk(0)], xmm7 ; wk(0)=z3L + movdqa XMMWORD [wk(1)], xmm4 ; wk(1)=z3H + + ; (Original) + ; z1 = tmp4 + tmp7; z2 = tmp5 + tmp6; + ; tmp4 = tmp4 * 0.298631336; tmp5 = tmp5 * 2.053119869; + ; tmp6 = tmp6 * 3.072711026; tmp7 = tmp7 * 1.501321110; + ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; + ; data7 = tmp4 + z1 + z3; data5 = tmp5 + z2 + z4; + ; data3 = tmp6 + z2 + z3; data1 = tmp7 + z1 + z4; + ; + ; (This implementation) + ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223; + ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447; + ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447); + ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223); + ; data7 = tmp4 + z3; data5 = tmp5 + z4; + ; data3 = tmp6 + z3; data1 = tmp7 + z4; + + movdqa xmm7,xmm2 + movdqa xmm4,xmm2 + punpcklwd xmm7,xmm1 + punpckhwd xmm4,xmm1 + movdqa xmm2,xmm7 + movdqa xmm1,xmm4 + pmaddwd xmm7,[GOTOFF(ebx,PW_MF060_MF089)] ; xmm7=tmp4L + pmaddwd xmm4,[GOTOFF(ebx,PW_MF060_MF089)] ; xmm4=tmp4H + pmaddwd xmm2,[GOTOFF(ebx,PW_MF089_F060)] ; xmm2=tmp7L + pmaddwd xmm1,[GOTOFF(ebx,PW_MF089_F060)] ; xmm1=tmp7H + + paddd xmm7, XMMWORD [wk(0)] ; xmm7=data7L + paddd xmm4, XMMWORD [wk(1)] ; xmm4=data7H + paddd xmm2,xmm6 ; xmm2=data1L + paddd xmm1,xmm0 ; xmm1=data1H + + paddd xmm7,[GOTOFF(ebx,PD_DESCALE_P1)] + paddd xmm4,[GOTOFF(ebx,PD_DESCALE_P1)] + psrad xmm7,DESCALE_P1 + psrad xmm4,DESCALE_P1 + paddd xmm2,[GOTOFF(ebx,PD_DESCALE_P1)] + paddd xmm1,[GOTOFF(ebx,PD_DESCALE_P1)] + psrad xmm2,DESCALE_P1 + psrad xmm1,DESCALE_P1 + + packssdw xmm7,xmm4 ; xmm7=data7 + packssdw xmm2,xmm1 ; xmm2=data1 + + movdqa xmm4,xmm5 + movdqa xmm1,xmm5 + punpcklwd xmm4,xmm3 + punpckhwd xmm1,xmm3 + movdqa xmm5,xmm4 + movdqa xmm3,xmm1 + pmaddwd xmm4,[GOTOFF(ebx,PW_MF050_MF256)] ; xmm4=tmp5L + pmaddwd xmm1,[GOTOFF(ebx,PW_MF050_MF256)] ; xmm1=tmp5H + pmaddwd xmm5,[GOTOFF(ebx,PW_MF256_F050)] ; xmm5=tmp6L + pmaddwd xmm3,[GOTOFF(ebx,PW_MF256_F050)] ; xmm3=tmp6H + + paddd xmm4,xmm6 ; xmm4=data5L + paddd xmm1,xmm0 ; xmm1=data5H + paddd xmm5, XMMWORD [wk(0)] ; xmm5=data3L + paddd xmm3, XMMWORD [wk(1)] ; xmm3=data3H + + paddd xmm4,[GOTOFF(ebx,PD_DESCALE_P1)] + paddd xmm1,[GOTOFF(ebx,PD_DESCALE_P1)] + psrad xmm4,DESCALE_P1 + psrad xmm1,DESCALE_P1 + paddd xmm5,[GOTOFF(ebx,PD_DESCALE_P1)] + paddd xmm3,[GOTOFF(ebx,PD_DESCALE_P1)] + psrad xmm5,DESCALE_P1 + psrad xmm3,DESCALE_P1 + + packssdw xmm4,xmm1 ; xmm4=data5 + packssdw xmm5,xmm3 ; xmm5=data3 + + ; ---- Pass 2: process columns. + +; mov edx, POINTER [data(eax)] ; (DCTELEM *) + + movdqa xmm6, XMMWORD [wk(2)] ; xmm6=col0 + movdqa xmm0, XMMWORD [wk(4)] ; xmm0=col2 + + ; xmm6=(00 10 20 30 40 50 60 70), xmm0=(02 12 22 32 42 52 62 72) + ; xmm2=(01 11 21 31 41 51 61 71), xmm5=(03 13 23 33 43 53 63 73) + + movdqa xmm1,xmm6 ; transpose coefficients(phase 1) + punpcklwd xmm6,xmm2 ; xmm6=(00 01 10 11 20 21 30 31) + punpckhwd xmm1,xmm2 ; xmm1=(40 41 50 51 60 61 70 71) + movdqa xmm3,xmm0 ; transpose coefficients(phase 1) + punpcklwd xmm0,xmm5 ; xmm0=(02 03 12 13 22 23 32 33) + punpckhwd xmm3,xmm5 ; xmm3=(42 43 52 53 62 63 72 73) + + movdqa xmm2, XMMWORD [wk(3)] ; xmm2=col4 + movdqa xmm5, XMMWORD [wk(5)] ; xmm5=col6 + + ; xmm2=(04 14 24 34 44 54 64 74), xmm5=(06 16 26 36 46 56 66 76) + ; xmm4=(05 15 25 35 45 55 65 75), xmm7=(07 17 27 37 47 57 67 77) + + movdqa XMMWORD [wk(0)], xmm0 ; wk(0)=(02 03 12 13 22 23 32 33) + movdqa XMMWORD [wk(1)], xmm3 ; wk(1)=(42 43 52 53 62 63 72 73) + + movdqa xmm0,xmm2 ; transpose coefficients(phase 1) + punpcklwd xmm2,xmm4 ; xmm2=(04 05 14 15 24 25 34 35) + punpckhwd xmm0,xmm4 ; xmm0=(44 45 54 55 64 65 74 75) + movdqa xmm3,xmm5 ; transpose coefficients(phase 1) + punpcklwd xmm5,xmm7 ; xmm5=(06 07 16 17 26 27 36 37) + punpckhwd xmm3,xmm7 ; xmm3=(46 47 56 57 66 67 76 77) + + movdqa xmm4,xmm2 ; transpose coefficients(phase 2) + punpckldq xmm2,xmm5 ; xmm2=(04 05 06 07 14 15 16 17) + punpckhdq xmm4,xmm5 ; xmm4=(24 25 26 27 34 35 36 37) + movdqa xmm7,xmm0 ; transpose coefficients(phase 2) + punpckldq xmm0,xmm3 ; xmm0=(44 45 46 47 54 55 56 57) + punpckhdq xmm7,xmm3 ; xmm7=(64 65 66 67 74 75 76 77) + + movdqa xmm5, XMMWORD [wk(0)] ; xmm5=(02 03 12 13 22 23 32 33) + movdqa xmm3, XMMWORD [wk(1)] ; xmm3=(42 43 52 53 62 63 72 73) + movdqa XMMWORD [wk(2)], xmm4 ; wk(2)=(24 25 26 27 34 35 36 37) + movdqa XMMWORD [wk(3)], xmm0 ; wk(3)=(44 45 46 47 54 55 56 57) + + movdqa xmm4,xmm6 ; transpose coefficients(phase 2) + punpckldq xmm6,xmm5 ; xmm6=(00 01 02 03 10 11 12 13) + punpckhdq xmm4,xmm5 ; xmm4=(20 21 22 23 30 31 32 33) + movdqa xmm0,xmm1 ; transpose coefficients(phase 2) + punpckldq xmm1,xmm3 ; xmm1=(40 41 42 43 50 51 52 53) + punpckhdq xmm0,xmm3 ; xmm0=(60 61 62 63 70 71 72 73) + + movdqa xmm5,xmm6 ; transpose coefficients(phase 3) + punpcklqdq xmm6,xmm2 ; xmm6=(00 01 02 03 04 05 06 07)=data0 + punpckhqdq xmm5,xmm2 ; xmm5=(10 11 12 13 14 15 16 17)=data1 + movdqa xmm3,xmm0 ; transpose coefficients(phase 3) + punpcklqdq xmm0,xmm7 ; xmm0=(60 61 62 63 64 65 66 67)=data6 + punpckhqdq xmm3,xmm7 ; xmm3=(70 71 72 73 74 75 76 77)=data7 + + movdqa xmm2,xmm5 + movdqa xmm7,xmm6 + psubw xmm5,xmm0 ; xmm5=data1-data6=tmp6 + psubw xmm6,xmm3 ; xmm6=data0-data7=tmp7 + paddw xmm2,xmm0 ; xmm2=data1+data6=tmp1 + paddw xmm7,xmm3 ; xmm7=data0+data7=tmp0 + + movdqa xmm0, XMMWORD [wk(2)] ; xmm0=(24 25 26 27 34 35 36 37) + movdqa xmm3, XMMWORD [wk(3)] ; xmm3=(44 45 46 47 54 55 56 57) + movdqa XMMWORD [wk(0)], xmm5 ; wk(0)=tmp6 + movdqa XMMWORD [wk(1)], xmm6 ; wk(1)=tmp7 + + movdqa xmm5,xmm4 ; transpose coefficients(phase 3) + punpcklqdq xmm4,xmm0 ; xmm4=(20 21 22 23 24 25 26 27)=data2 + punpckhqdq xmm5,xmm0 ; xmm5=(30 31 32 33 34 35 36 37)=data3 + movdqa xmm6,xmm1 ; transpose coefficients(phase 3) + punpcklqdq xmm1,xmm3 ; xmm1=(40 41 42 43 44 45 46 47)=data4 + punpckhqdq xmm6,xmm3 ; xmm6=(50 51 52 53 54 55 56 57)=data5 + + movdqa xmm0,xmm5 + movdqa xmm3,xmm4 + paddw xmm5,xmm1 ; xmm5=data3+data4=tmp3 + paddw xmm4,xmm6 ; xmm4=data2+data5=tmp2 + psubw xmm0,xmm1 ; xmm0=data3-data4=tmp4 + psubw xmm3,xmm6 ; xmm3=data2-data5=tmp5 + + ; -- Even part + + movdqa xmm1,xmm7 + movdqa xmm6,xmm2 + paddw xmm7,xmm5 ; xmm7=tmp10 + paddw xmm2,xmm4 ; xmm2=tmp11 + psubw xmm1,xmm5 ; xmm1=tmp13 + psubw xmm6,xmm4 ; xmm6=tmp12 + + movdqa xmm5,xmm7 + paddw xmm7,xmm2 ; xmm7=tmp10+tmp11 + psubw xmm5,xmm2 ; xmm5=tmp10-tmp11 + + paddw xmm7,[GOTOFF(ebx,PW_DESCALE_P2X)] + paddw xmm5,[GOTOFF(ebx,PW_DESCALE_P2X)] + psraw xmm7,PASS1_BITS ; xmm7=data0 + psraw xmm5,PASS1_BITS ; xmm5=data4 + + movdqa XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)], xmm7 + movdqa XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)], xmm5 + + ; (Original) + ; z1 = (tmp12 + tmp13) * 0.541196100; + ; data2 = z1 + tmp13 * 0.765366865; + ; data6 = z1 + tmp12 * -1.847759065; + ; + ; (This implementation) + ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100; + ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065); + + movdqa xmm4,xmm1 ; xmm1=tmp13 + movdqa xmm2,xmm1 + punpcklwd xmm4,xmm6 ; xmm6=tmp12 + punpckhwd xmm2,xmm6 + movdqa xmm1,xmm4 + movdqa xmm6,xmm2 + pmaddwd xmm4,[GOTOFF(ebx,PW_F130_F054)] ; xmm4=data2L + pmaddwd xmm2,[GOTOFF(ebx,PW_F130_F054)] ; xmm2=data2H + pmaddwd xmm1,[GOTOFF(ebx,PW_F054_MF130)] ; xmm1=data6L + pmaddwd xmm6,[GOTOFF(ebx,PW_F054_MF130)] ; xmm6=data6H + + paddd xmm4,[GOTOFF(ebx,PD_DESCALE_P2)] + paddd xmm2,[GOTOFF(ebx,PD_DESCALE_P2)] + psrad xmm4,DESCALE_P2 + psrad xmm2,DESCALE_P2 + paddd xmm1,[GOTOFF(ebx,PD_DESCALE_P2)] + paddd xmm6,[GOTOFF(ebx,PD_DESCALE_P2)] + psrad xmm1,DESCALE_P2 + psrad xmm6,DESCALE_P2 + + packssdw xmm4,xmm2 ; xmm4=data2 + packssdw xmm1,xmm6 ; xmm1=data6 + + movdqa XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)], xmm4 + movdqa XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)], xmm1 + + ; -- Odd part + + movdqa xmm7, XMMWORD [wk(0)] ; xmm7=tmp6 + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=tmp7 + + movdqa xmm2,xmm0 ; xmm0=tmp4 + movdqa xmm6,xmm3 ; xmm3=tmp5 + paddw xmm2,xmm7 ; xmm2=z3 + paddw xmm6,xmm5 ; xmm6=z4 + + ; (Original) + ; z5 = (z3 + z4) * 1.175875602; + ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; + ; z3 += z5; z4 += z5; + ; + ; (This implementation) + ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; + ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); + + movdqa xmm4,xmm2 + movdqa xmm1,xmm2 + punpcklwd xmm4,xmm6 + punpckhwd xmm1,xmm6 + movdqa xmm2,xmm4 + movdqa xmm6,xmm1 + pmaddwd xmm4,[GOTOFF(ebx,PW_MF078_F117)] ; xmm4=z3L + pmaddwd xmm1,[GOTOFF(ebx,PW_MF078_F117)] ; xmm1=z3H + pmaddwd xmm2,[GOTOFF(ebx,PW_F117_F078)] ; xmm2=z4L + pmaddwd xmm6,[GOTOFF(ebx,PW_F117_F078)] ; xmm6=z4H + + movdqa XMMWORD [wk(0)], xmm4 ; wk(0)=z3L + movdqa XMMWORD [wk(1)], xmm1 ; wk(1)=z3H + + ; (Original) + ; z1 = tmp4 + tmp7; z2 = tmp5 + tmp6; + ; tmp4 = tmp4 * 0.298631336; tmp5 = tmp5 * 2.053119869; + ; tmp6 = tmp6 * 3.072711026; tmp7 = tmp7 * 1.501321110; + ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; + ; data7 = tmp4 + z1 + z3; data5 = tmp5 + z2 + z4; + ; data3 = tmp6 + z2 + z3; data1 = tmp7 + z1 + z4; + ; + ; (This implementation) + ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223; + ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447; + ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447); + ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223); + ; data7 = tmp4 + z3; data5 = tmp5 + z4; + ; data3 = tmp6 + z3; data1 = tmp7 + z4; + + movdqa xmm4,xmm0 + movdqa xmm1,xmm0 + punpcklwd xmm4,xmm5 + punpckhwd xmm1,xmm5 + movdqa xmm0,xmm4 + movdqa xmm5,xmm1 + pmaddwd xmm4,[GOTOFF(ebx,PW_MF060_MF089)] ; xmm4=tmp4L + pmaddwd xmm1,[GOTOFF(ebx,PW_MF060_MF089)] ; xmm1=tmp4H + pmaddwd xmm0,[GOTOFF(ebx,PW_MF089_F060)] ; xmm0=tmp7L + pmaddwd xmm5,[GOTOFF(ebx,PW_MF089_F060)] ; xmm5=tmp7H + + paddd xmm4, XMMWORD [wk(0)] ; xmm4=data7L + paddd xmm1, XMMWORD [wk(1)] ; xmm1=data7H + paddd xmm0,xmm2 ; xmm0=data1L + paddd xmm5,xmm6 ; xmm5=data1H + + paddd xmm4,[GOTOFF(ebx,PD_DESCALE_P2)] + paddd xmm1,[GOTOFF(ebx,PD_DESCALE_P2)] + psrad xmm4,DESCALE_P2 + psrad xmm1,DESCALE_P2 + paddd xmm0,[GOTOFF(ebx,PD_DESCALE_P2)] + paddd xmm5,[GOTOFF(ebx,PD_DESCALE_P2)] + psrad xmm0,DESCALE_P2 + psrad xmm5,DESCALE_P2 + + packssdw xmm4,xmm1 ; xmm4=data7 + packssdw xmm0,xmm5 ; xmm0=data1 + + movdqa XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)], xmm4 + movdqa XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)], xmm0 + + movdqa xmm1,xmm3 + movdqa xmm5,xmm3 + punpcklwd xmm1,xmm7 + punpckhwd xmm5,xmm7 + movdqa xmm3,xmm1 + movdqa xmm7,xmm5 + pmaddwd xmm1,[GOTOFF(ebx,PW_MF050_MF256)] ; xmm1=tmp5L + pmaddwd xmm5,[GOTOFF(ebx,PW_MF050_MF256)] ; xmm5=tmp5H + pmaddwd xmm3,[GOTOFF(ebx,PW_MF256_F050)] ; xmm3=tmp6L + pmaddwd xmm7,[GOTOFF(ebx,PW_MF256_F050)] ; xmm7=tmp6H + + paddd xmm1,xmm2 ; xmm1=data5L + paddd xmm5,xmm6 ; xmm5=data5H + paddd xmm3, XMMWORD [wk(0)] ; xmm3=data3L + paddd xmm7, XMMWORD [wk(1)] ; xmm7=data3H + + paddd xmm1,[GOTOFF(ebx,PD_DESCALE_P2)] + paddd xmm5,[GOTOFF(ebx,PD_DESCALE_P2)] + psrad xmm1,DESCALE_P2 + psrad xmm5,DESCALE_P2 + paddd xmm3,[GOTOFF(ebx,PD_DESCALE_P2)] + paddd xmm7,[GOTOFF(ebx,PD_DESCALE_P2)] + psrad xmm3,DESCALE_P2 + psrad xmm7,DESCALE_P2 + + packssdw xmm1,xmm5 ; xmm1=data5 + packssdw xmm3,xmm7 ; xmm3=data3 + + movdqa XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)], xmm1 + movdqa XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)], xmm3 + +; pop edi ; unused +; pop esi ; unused +; pop edx ; need not be preserved +; pop ecx ; unused + poppic ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jidctflt-3dn.asm b/app/src/main/jni/libjpeg-turbo/simd/jidctflt-3dn.asm new file mode 100644 index 000000000..ea2e27011 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jidctflt-3dn.asm @@ -0,0 +1,452 @@ +; +; jidctflt.asm - floating-point IDCT (3DNow! & MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a floating-point implementation of the inverse DCT +; (Discrete Cosine Transform). The following code is based directly on +; the IJG's original jidctflt.c; see the jidctflt.c for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_idct_float_3dnow) + +EXTN(jconst_idct_float_3dnow): + +PD_1_414 times 2 dd 1.414213562373095048801689 +PD_1_847 times 2 dd 1.847759065022573512256366 +PD_1_082 times 2 dd 1.082392200292393968799446 +PD_2_613 times 2 dd 2.613125929752753055713286 +PD_RNDINT_MAGIC times 2 dd 100663296.0 ; (float)(0x00C00000 << 3) +PB_CENTERJSAMP times 8 db CENTERJSAMPLE + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform dequantization and inverse DCT on one block of coefficients. +; +; GLOBAL(void) +; jsimd_idct_float_3dnow (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +%define dct_table(b) (b)+8 ; void * dct_table +%define coef_block(b) (b)+12 ; JCOEFPTR coef_block +%define output_buf(b) (b)+16 ; JSAMPARRAY output_buf +%define output_col(b) (b)+20 ; JDIMENSION output_col + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_MMWORD ; mmword wk[WK_NUM] +%define WK_NUM 2 +%define workspace wk(0)-DCTSIZE2*SIZEOF_FAST_FLOAT + ; FAST_FLOAT workspace[DCTSIZE2] + + align 16 + global EXTN(jsimd_idct_float_3dnow) + +EXTN(jsimd_idct_float_3dnow): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_MMWORD) ; align to 64 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [workspace] + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process columns from input, store into work array. + +; mov eax, [original_ebp] + mov edx, POINTER [dct_table(eax)] ; quantptr + mov esi, JCOEFPTR [coef_block(eax)] ; inptr + lea edi, [workspace] ; FAST_FLOAT * wsptr + mov ecx, DCTSIZE/2 ; ctr + alignx 16,7 +.columnloop: +%ifndef NO_ZERO_COLUMN_TEST_FLOAT_3DNOW + mov eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)] + jnz short .columnDCT + + pushpic ebx ; save GOT address + mov ebx, DWORD [DWBLOCK(3,0,esi,SIZEOF_JCOEF)] + mov eax, DWORD [DWBLOCK(4,0,esi,SIZEOF_JCOEF)] + or ebx, DWORD [DWBLOCK(5,0,esi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(6,0,esi,SIZEOF_JCOEF)] + or ebx, DWORD [DWBLOCK(7,0,esi,SIZEOF_JCOEF)] + or eax,ebx + poppic ebx ; restore GOT address + jnz short .columnDCT + + ; -- AC terms all zero + + movd mm0, DWORD [DWBLOCK(0,0,esi,SIZEOF_JCOEF)] + + punpcklwd mm0,mm0 + psrad mm0,(DWORD_BIT-WORD_BIT) + pi2fd mm0,mm0 + + pfmul mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + + movq mm1,mm0 + punpckldq mm0,mm0 + punpckhdq mm1,mm1 + + movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], mm0 + movq MMWORD [MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], mm0 + movq MMWORD [MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT)], mm0 + movq MMWORD [MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT)], mm0 + movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], mm1 + movq MMWORD [MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], mm1 + movq MMWORD [MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT)], mm1 + movq MMWORD [MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT)], mm1 + jmp near .nextcolumn + alignx 16,7 +%endif +.columnDCT: + + ; -- Even part + + movd mm0, DWORD [DWBLOCK(0,0,esi,SIZEOF_JCOEF)] + movd mm1, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)] + movd mm2, DWORD [DWBLOCK(4,0,esi,SIZEOF_JCOEF)] + movd mm3, DWORD [DWBLOCK(6,0,esi,SIZEOF_JCOEF)] + + punpcklwd mm0,mm0 + punpcklwd mm1,mm1 + psrad mm0,(DWORD_BIT-WORD_BIT) + psrad mm1,(DWORD_BIT-WORD_BIT) + pi2fd mm0,mm0 + pi2fd mm1,mm1 + + pfmul mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + pfmul mm1, MMWORD [MMBLOCK(2,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + + punpcklwd mm2,mm2 + punpcklwd mm3,mm3 + psrad mm2,(DWORD_BIT-WORD_BIT) + psrad mm3,(DWORD_BIT-WORD_BIT) + pi2fd mm2,mm2 + pi2fd mm3,mm3 + + pfmul mm2, MMWORD [MMBLOCK(4,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + pfmul mm3, MMWORD [MMBLOCK(6,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + + movq mm4,mm0 + movq mm5,mm1 + pfsub mm0,mm2 ; mm0=tmp11 + pfsub mm1,mm3 + pfadd mm4,mm2 ; mm4=tmp10 + pfadd mm5,mm3 ; mm5=tmp13 + + pfmul mm1,[GOTOFF(ebx,PD_1_414)] + pfsub mm1,mm5 ; mm1=tmp12 + + movq mm6,mm4 + movq mm7,mm0 + pfsub mm4,mm5 ; mm4=tmp3 + pfsub mm0,mm1 ; mm0=tmp2 + pfadd mm6,mm5 ; mm6=tmp0 + pfadd mm7,mm1 ; mm7=tmp1 + + movq MMWORD [wk(1)], mm4 ; tmp3 + movq MMWORD [wk(0)], mm0 ; tmp2 + + ; -- Odd part + + movd mm2, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)] + movd mm3, DWORD [DWBLOCK(3,0,esi,SIZEOF_JCOEF)] + movd mm5, DWORD [DWBLOCK(5,0,esi,SIZEOF_JCOEF)] + movd mm1, DWORD [DWBLOCK(7,0,esi,SIZEOF_JCOEF)] + + punpcklwd mm2,mm2 + punpcklwd mm3,mm3 + psrad mm2,(DWORD_BIT-WORD_BIT) + psrad mm3,(DWORD_BIT-WORD_BIT) + pi2fd mm2,mm2 + pi2fd mm3,mm3 + + pfmul mm2, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + pfmul mm3, MMWORD [MMBLOCK(3,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + + punpcklwd mm5,mm5 + punpcklwd mm1,mm1 + psrad mm5,(DWORD_BIT-WORD_BIT) + psrad mm1,(DWORD_BIT-WORD_BIT) + pi2fd mm5,mm5 + pi2fd mm1,mm1 + + pfmul mm5, MMWORD [MMBLOCK(5,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + pfmul mm1, MMWORD [MMBLOCK(7,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + + movq mm4,mm2 + movq mm0,mm5 + pfadd mm2,mm1 ; mm2=z11 + pfadd mm5,mm3 ; mm5=z13 + pfsub mm4,mm1 ; mm4=z12 + pfsub mm0,mm3 ; mm0=z10 + + movq mm1,mm2 + pfsub mm2,mm5 + pfadd mm1,mm5 ; mm1=tmp7 + + pfmul mm2,[GOTOFF(ebx,PD_1_414)] ; mm2=tmp11 + + movq mm3,mm0 + pfadd mm0,mm4 + pfmul mm0,[GOTOFF(ebx,PD_1_847)] ; mm0=z5 + pfmul mm3,[GOTOFF(ebx,PD_2_613)] ; mm3=(z10 * 2.613125930) + pfmul mm4,[GOTOFF(ebx,PD_1_082)] ; mm4=(z12 * 1.082392200) + pfsubr mm3,mm0 ; mm3=tmp12 + pfsub mm4,mm0 ; mm4=tmp10 + + ; -- Final output stage + + pfsub mm3,mm1 ; mm3=tmp6 + movq mm5,mm6 + movq mm0,mm7 + pfadd mm6,mm1 ; mm6=data0=(00 01) + pfadd mm7,mm3 ; mm7=data1=(10 11) + pfsub mm5,mm1 ; mm5=data7=(70 71) + pfsub mm0,mm3 ; mm0=data6=(60 61) + pfsub mm2,mm3 ; mm2=tmp5 + + movq mm1,mm6 ; transpose coefficients + punpckldq mm6,mm7 ; mm6=(00 10) + punpckhdq mm1,mm7 ; mm1=(01 11) + movq mm3,mm0 ; transpose coefficients + punpckldq mm0,mm5 ; mm0=(60 70) + punpckhdq mm3,mm5 ; mm3=(61 71) + + movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], mm6 + movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], mm1 + movq MMWORD [MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT)], mm0 + movq MMWORD [MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT)], mm3 + + movq mm7, MMWORD [wk(0)] ; mm7=tmp2 + movq mm5, MMWORD [wk(1)] ; mm5=tmp3 + + pfadd mm4,mm2 ; mm4=tmp4 + movq mm6,mm7 + movq mm1,mm5 + pfadd mm7,mm2 ; mm7=data2=(20 21) + pfadd mm5,mm4 ; mm5=data4=(40 41) + pfsub mm6,mm2 ; mm6=data5=(50 51) + pfsub mm1,mm4 ; mm1=data3=(30 31) + + movq mm0,mm7 ; transpose coefficients + punpckldq mm7,mm1 ; mm7=(20 30) + punpckhdq mm0,mm1 ; mm0=(21 31) + movq mm3,mm5 ; transpose coefficients + punpckldq mm5,mm6 ; mm5=(40 50) + punpckhdq mm3,mm6 ; mm3=(41 51) + + movq MMWORD [MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], mm7 + movq MMWORD [MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], mm0 + movq MMWORD [MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT)], mm5 + movq MMWORD [MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT)], mm3 + +.nextcolumn: + add esi, byte 2*SIZEOF_JCOEF ; coef_block + add edx, byte 2*SIZEOF_FLOAT_MULT_TYPE ; quantptr + add edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT ; wsptr + dec ecx ; ctr + jnz near .columnloop + + ; -- Prefetch the next coefficient block + + prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32] + prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32] + prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32] + prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32] + + ; ---- Pass 2: process rows from work array, store into output array. + + mov eax, [original_ebp] + lea esi, [workspace] ; FAST_FLOAT * wsptr + mov edi, JSAMPARRAY [output_buf(eax)] ; (JSAMPROW *) + mov eax, JDIMENSION [output_col(eax)] + mov ecx, DCTSIZE/2 ; ctr + alignx 16,7 +.rowloop: + + ; -- Even part + + movq mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)] + movq mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_FAST_FLOAT)] + movq mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_FAST_FLOAT)] + movq mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_FAST_FLOAT)] + + movq mm4,mm0 + movq mm5,mm1 + pfsub mm0,mm2 ; mm0=tmp11 + pfsub mm1,mm3 + pfadd mm4,mm2 ; mm4=tmp10 + pfadd mm5,mm3 ; mm5=tmp13 + + pfmul mm1,[GOTOFF(ebx,PD_1_414)] + pfsub mm1,mm5 ; mm1=tmp12 + + movq mm6,mm4 + movq mm7,mm0 + pfsub mm4,mm5 ; mm4=tmp3 + pfsub mm0,mm1 ; mm0=tmp2 + pfadd mm6,mm5 ; mm6=tmp0 + pfadd mm7,mm1 ; mm7=tmp1 + + movq MMWORD [wk(1)], mm4 ; tmp3 + movq MMWORD [wk(0)], mm0 ; tmp2 + + ; -- Odd part + + movq mm2, MMWORD [MMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)] + movq mm3, MMWORD [MMBLOCK(3,0,esi,SIZEOF_FAST_FLOAT)] + movq mm5, MMWORD [MMBLOCK(5,0,esi,SIZEOF_FAST_FLOAT)] + movq mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_FAST_FLOAT)] + + movq mm4,mm2 + movq mm0,mm5 + pfadd mm2,mm1 ; mm2=z11 + pfadd mm5,mm3 ; mm5=z13 + pfsub mm4,mm1 ; mm4=z12 + pfsub mm0,mm3 ; mm0=z10 + + movq mm1,mm2 + pfsub mm2,mm5 + pfadd mm1,mm5 ; mm1=tmp7 + + pfmul mm2,[GOTOFF(ebx,PD_1_414)] ; mm2=tmp11 + + movq mm3,mm0 + pfadd mm0,mm4 + pfmul mm0,[GOTOFF(ebx,PD_1_847)] ; mm0=z5 + pfmul mm3,[GOTOFF(ebx,PD_2_613)] ; mm3=(z10 * 2.613125930) + pfmul mm4,[GOTOFF(ebx,PD_1_082)] ; mm4=(z12 * 1.082392200) + pfsubr mm3,mm0 ; mm3=tmp12 + pfsub mm4,mm0 ; mm4=tmp10 + + ; -- Final output stage + + pfsub mm3,mm1 ; mm3=tmp6 + movq mm5,mm6 + movq mm0,mm7 + pfadd mm6,mm1 ; mm6=data0=(00 10) + pfadd mm7,mm3 ; mm7=data1=(01 11) + pfsub mm5,mm1 ; mm5=data7=(07 17) + pfsub mm0,mm3 ; mm0=data6=(06 16) + pfsub mm2,mm3 ; mm2=tmp5 + + movq mm1,[GOTOFF(ebx,PD_RNDINT_MAGIC)] ; mm1=[PD_RNDINT_MAGIC] + pcmpeqd mm3,mm3 + psrld mm3,WORD_BIT ; mm3={0xFFFF 0x0000 0xFFFF 0x0000} + + pfadd mm6,mm1 ; mm6=roundint(data0/8)=(00 ** 10 **) + pfadd mm7,mm1 ; mm7=roundint(data1/8)=(01 ** 11 **) + pfadd mm0,mm1 ; mm0=roundint(data6/8)=(06 ** 16 **) + pfadd mm5,mm1 ; mm5=roundint(data7/8)=(07 ** 17 **) + + pand mm6,mm3 ; mm6=(00 -- 10 --) + pslld mm7,WORD_BIT ; mm7=(-- 01 -- 11) + pand mm0,mm3 ; mm0=(06 -- 16 --) + pslld mm5,WORD_BIT ; mm5=(-- 07 -- 17) + por mm6,mm7 ; mm6=(00 01 10 11) + por mm0,mm5 ; mm0=(06 07 16 17) + + movq mm1, MMWORD [wk(0)] ; mm1=tmp2 + movq mm3, MMWORD [wk(1)] ; mm3=tmp3 + + pfadd mm4,mm2 ; mm4=tmp4 + movq mm7,mm1 + movq mm5,mm3 + pfadd mm1,mm2 ; mm1=data2=(02 12) + pfadd mm3,mm4 ; mm3=data4=(04 14) + pfsub mm7,mm2 ; mm7=data5=(05 15) + pfsub mm5,mm4 ; mm5=data3=(03 13) + + movq mm2,[GOTOFF(ebx,PD_RNDINT_MAGIC)] ; mm2=[PD_RNDINT_MAGIC] + pcmpeqd mm4,mm4 + psrld mm4,WORD_BIT ; mm4={0xFFFF 0x0000 0xFFFF 0x0000} + + pfadd mm3,mm2 ; mm3=roundint(data4/8)=(04 ** 14 **) + pfadd mm7,mm2 ; mm7=roundint(data5/8)=(05 ** 15 **) + pfadd mm1,mm2 ; mm1=roundint(data2/8)=(02 ** 12 **) + pfadd mm5,mm2 ; mm5=roundint(data3/8)=(03 ** 13 **) + + pand mm3,mm4 ; mm3=(04 -- 14 --) + pslld mm7,WORD_BIT ; mm7=(-- 05 -- 15) + pand mm1,mm4 ; mm1=(02 -- 12 --) + pslld mm5,WORD_BIT ; mm5=(-- 03 -- 13) + por mm3,mm7 ; mm3=(04 05 14 15) + por mm1,mm5 ; mm1=(02 03 12 13) + + movq mm2,[GOTOFF(ebx,PB_CENTERJSAMP)] ; mm2=[PB_CENTERJSAMP] + + packsswb mm6,mm3 ; mm6=(00 01 10 11 04 05 14 15) + packsswb mm1,mm0 ; mm1=(02 03 12 13 06 07 16 17) + paddb mm6,mm2 + paddb mm1,mm2 + + movq mm4,mm6 ; transpose coefficients(phase 2) + punpcklwd mm6,mm1 ; mm6=(00 01 02 03 10 11 12 13) + punpckhwd mm4,mm1 ; mm4=(04 05 06 07 14 15 16 17) + + movq mm7,mm6 ; transpose coefficients(phase 3) + punpckldq mm6,mm4 ; mm6=(00 01 02 03 04 05 06 07) + punpckhdq mm7,mm4 ; mm7=(10 11 12 13 14 15 16 17) + + pushpic ebx ; save GOT address + + mov edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] + mov ebx, JSAMPROW [edi+1*SIZEOF_JSAMPROW] + movq MMWORD [edx+eax*SIZEOF_JSAMPLE], mm6 + movq MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm7 + + poppic ebx ; restore GOT address + + add esi, byte 2*SIZEOF_FAST_FLOAT ; wsptr + add edi, byte 2*SIZEOF_JSAMPROW + dec ecx ; ctr + jnz near .rowloop + + femms ; empty MMX/3DNow! state + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jidctflt-sse.asm b/app/src/main/jni/libjpeg-turbo/simd/jidctflt-sse.asm new file mode 100644 index 000000000..6160e417c --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jidctflt-sse.asm @@ -0,0 +1,572 @@ +; +; jidctflt.asm - floating-point IDCT (SSE & MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a floating-point implementation of the inverse DCT +; (Discrete Cosine Transform). The following code is based directly on +; the IJG's original jidctflt.c; see the jidctflt.c for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%macro unpcklps2 2 ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5) + shufps %1,%2,0x44 +%endmacro + +%macro unpckhps2 2 ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7) + shufps %1,%2,0xEE +%endmacro + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_idct_float_sse) + +EXTN(jconst_idct_float_sse): + +PD_1_414 times 4 dd 1.414213562373095048801689 +PD_1_847 times 4 dd 1.847759065022573512256366 +PD_1_082 times 4 dd 1.082392200292393968799446 +PD_M2_613 times 4 dd -2.613125929752753055713286 +PD_0_125 times 4 dd 0.125 ; 1/8 +PB_CENTERJSAMP times 8 db CENTERJSAMPLE + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform dequantization and inverse DCT on one block of coefficients. +; +; GLOBAL(void) +; jsimd_idct_float_sse (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +%define dct_table(b) (b)+8 ; void * dct_table +%define coef_block(b) (b)+12 ; JCOEFPTR coef_block +%define output_buf(b) (b)+16 ; JSAMPARRAY output_buf +%define output_col(b) (b)+20 ; JDIMENSION output_col + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 +%define workspace wk(0)-DCTSIZE2*SIZEOF_FAST_FLOAT + ; FAST_FLOAT workspace[DCTSIZE2] + + align 16 + global EXTN(jsimd_idct_float_sse) + +EXTN(jsimd_idct_float_sse): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [workspace] + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process columns from input, store into work array. + +; mov eax, [original_ebp] + mov edx, POINTER [dct_table(eax)] ; quantptr + mov esi, JCOEFPTR [coef_block(eax)] ; inptr + lea edi, [workspace] ; FAST_FLOAT * wsptr + mov ecx, DCTSIZE/4 ; ctr + alignx 16,7 +.columnloop: +%ifndef NO_ZERO_COLUMN_TEST_FLOAT_SSE + mov eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)] + jnz near .columnDCT + + movq mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)] + por mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + por mm1, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)] + por mm0, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + por mm1, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)] + por mm0, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + por mm1,mm0 + packsswb mm1,mm1 + movd eax,mm1 + test eax,eax + jnz short .columnDCT + + ; -- AC terms all zero + + movq mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + + punpckhwd mm1,mm0 ; mm1=(** 02 ** 03) + punpcklwd mm0,mm0 ; mm0=(00 00 01 01) + psrad mm1,(DWORD_BIT-WORD_BIT) ; mm1=in0H=(02 03) + psrad mm0,(DWORD_BIT-WORD_BIT) ; mm0=in0L=(00 01) + cvtpi2ps xmm3,mm1 ; xmm3=(02 03 ** **) + cvtpi2ps xmm0,mm0 ; xmm0=(00 01 ** **) + movlhps xmm0,xmm3 ; xmm0=in0=(00 01 02 03) + + mulps xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + + movaps xmm1,xmm0 + movaps xmm2,xmm0 + movaps xmm3,xmm0 + + shufps xmm0,xmm0,0x00 ; xmm0=(00 00 00 00) + shufps xmm1,xmm1,0x55 ; xmm1=(01 01 01 01) + shufps xmm2,xmm2,0xAA ; xmm2=(02 02 02 02) + shufps xmm3,xmm3,0xFF ; xmm3=(03 03 03 03) + + movaps XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm0 + movaps XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm0 + movaps XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm1 + movaps XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm1 + movaps XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT)], xmm2 + movaps XMMWORD [XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT)], xmm2 + movaps XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT)], xmm3 + movaps XMMWORD [XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT)], xmm3 + jmp near .nextcolumn + alignx 16,7 +%endif +.columnDCT: + + ; -- Even part + + movq mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)] + movq mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)] + movq mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)] + + punpckhwd mm4,mm0 ; mm4=(** 02 ** 03) + punpcklwd mm0,mm0 ; mm0=(00 00 01 01) + punpckhwd mm5,mm1 ; mm5=(** 22 ** 23) + punpcklwd mm1,mm1 ; mm1=(20 20 21 21) + + psrad mm4,(DWORD_BIT-WORD_BIT) ; mm4=in0H=(02 03) + psrad mm0,(DWORD_BIT-WORD_BIT) ; mm0=in0L=(00 01) + cvtpi2ps xmm4,mm4 ; xmm4=(02 03 ** **) + cvtpi2ps xmm0,mm0 ; xmm0=(00 01 ** **) + psrad mm5,(DWORD_BIT-WORD_BIT) ; mm5=in2H=(22 23) + psrad mm1,(DWORD_BIT-WORD_BIT) ; mm1=in2L=(20 21) + cvtpi2ps xmm5,mm5 ; xmm5=(22 23 ** **) + cvtpi2ps xmm1,mm1 ; xmm1=(20 21 ** **) + + punpckhwd mm6,mm2 ; mm6=(** 42 ** 43) + punpcklwd mm2,mm2 ; mm2=(40 40 41 41) + punpckhwd mm7,mm3 ; mm7=(** 62 ** 63) + punpcklwd mm3,mm3 ; mm3=(60 60 61 61) + + psrad mm6,(DWORD_BIT-WORD_BIT) ; mm6=in4H=(42 43) + psrad mm2,(DWORD_BIT-WORD_BIT) ; mm2=in4L=(40 41) + cvtpi2ps xmm6,mm6 ; xmm6=(42 43 ** **) + cvtpi2ps xmm2,mm2 ; xmm2=(40 41 ** **) + psrad mm7,(DWORD_BIT-WORD_BIT) ; mm7=in6H=(62 63) + psrad mm3,(DWORD_BIT-WORD_BIT) ; mm3=in6L=(60 61) + cvtpi2ps xmm7,mm7 ; xmm7=(62 63 ** **) + cvtpi2ps xmm3,mm3 ; xmm3=(60 61 ** **) + + movlhps xmm0,xmm4 ; xmm0=in0=(00 01 02 03) + movlhps xmm1,xmm5 ; xmm1=in2=(20 21 22 23) + mulps xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm1, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + + movlhps xmm2,xmm6 ; xmm2=in4=(40 41 42 43) + movlhps xmm3,xmm7 ; xmm3=in6=(60 61 62 63) + mulps xmm2, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm3, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + + movaps xmm4,xmm0 + movaps xmm5,xmm1 + subps xmm0,xmm2 ; xmm0=tmp11 + subps xmm1,xmm3 + addps xmm4,xmm2 ; xmm4=tmp10 + addps xmm5,xmm3 ; xmm5=tmp13 + + mulps xmm1,[GOTOFF(ebx,PD_1_414)] + subps xmm1,xmm5 ; xmm1=tmp12 + + movaps xmm6,xmm4 + movaps xmm7,xmm0 + subps xmm4,xmm5 ; xmm4=tmp3 + subps xmm0,xmm1 ; xmm0=tmp2 + addps xmm6,xmm5 ; xmm6=tmp0 + addps xmm7,xmm1 ; xmm7=tmp1 + + movaps XMMWORD [wk(1)], xmm4 ; tmp3 + movaps XMMWORD [wk(0)], xmm0 ; tmp2 + + ; -- Odd part + + movq mm4, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + movq mm5, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + + punpckhwd mm6,mm4 ; mm6=(** 12 ** 13) + punpcklwd mm4,mm4 ; mm4=(10 10 11 11) + punpckhwd mm2,mm0 ; mm2=(** 32 ** 33) + punpcklwd mm0,mm0 ; mm0=(30 30 31 31) + + psrad mm6,(DWORD_BIT-WORD_BIT) ; mm6=in1H=(12 13) + psrad mm4,(DWORD_BIT-WORD_BIT) ; mm4=in1L=(10 11) + cvtpi2ps xmm4,mm6 ; xmm4=(12 13 ** **) + cvtpi2ps xmm2,mm4 ; xmm2=(10 11 ** **) + psrad mm2,(DWORD_BIT-WORD_BIT) ; mm2=in3H=(32 33) + psrad mm0,(DWORD_BIT-WORD_BIT) ; mm0=in3L=(30 31) + cvtpi2ps xmm0,mm2 ; xmm0=(32 33 ** **) + cvtpi2ps xmm3,mm0 ; xmm3=(30 31 ** **) + + punpckhwd mm7,mm5 ; mm7=(** 52 ** 53) + punpcklwd mm5,mm5 ; mm5=(50 50 51 51) + punpckhwd mm3,mm1 ; mm3=(** 72 ** 73) + punpcklwd mm1,mm1 ; mm1=(70 70 71 71) + + movlhps xmm2,xmm4 ; xmm2=in1=(10 11 12 13) + movlhps xmm3,xmm0 ; xmm3=in3=(30 31 32 33) + + psrad mm7,(DWORD_BIT-WORD_BIT) ; mm7=in5H=(52 53) + psrad mm5,(DWORD_BIT-WORD_BIT) ; mm5=in5L=(50 51) + cvtpi2ps xmm4,mm7 ; xmm4=(52 53 ** **) + cvtpi2ps xmm5,mm5 ; xmm5=(50 51 ** **) + psrad mm3,(DWORD_BIT-WORD_BIT) ; mm3=in7H=(72 73) + psrad mm1,(DWORD_BIT-WORD_BIT) ; mm1=in7L=(70 71) + cvtpi2ps xmm0,mm3 ; xmm0=(72 73 ** **) + cvtpi2ps xmm1,mm1 ; xmm1=(70 71 ** **) + + mulps xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + + movlhps xmm5,xmm4 ; xmm5=in5=(50 51 52 53) + movlhps xmm1,xmm0 ; xmm1=in7=(70 71 72 73) + mulps xmm5, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm1, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + + movaps xmm4,xmm2 + movaps xmm0,xmm5 + addps xmm2,xmm1 ; xmm2=z11 + addps xmm5,xmm3 ; xmm5=z13 + subps xmm4,xmm1 ; xmm4=z12 + subps xmm0,xmm3 ; xmm0=z10 + + movaps xmm1,xmm2 + subps xmm2,xmm5 + addps xmm1,xmm5 ; xmm1=tmp7 + + mulps xmm2,[GOTOFF(ebx,PD_1_414)] ; xmm2=tmp11 + + movaps xmm3,xmm0 + addps xmm0,xmm4 + mulps xmm0,[GOTOFF(ebx,PD_1_847)] ; xmm0=z5 + mulps xmm3,[GOTOFF(ebx,PD_M2_613)] ; xmm3=(z10 * -2.613125930) + mulps xmm4,[GOTOFF(ebx,PD_1_082)] ; xmm4=(z12 * 1.082392200) + addps xmm3,xmm0 ; xmm3=tmp12 + subps xmm4,xmm0 ; xmm4=tmp10 + + ; -- Final output stage + + subps xmm3,xmm1 ; xmm3=tmp6 + movaps xmm5,xmm6 + movaps xmm0,xmm7 + addps xmm6,xmm1 ; xmm6=data0=(00 01 02 03) + addps xmm7,xmm3 ; xmm7=data1=(10 11 12 13) + subps xmm5,xmm1 ; xmm5=data7=(70 71 72 73) + subps xmm0,xmm3 ; xmm0=data6=(60 61 62 63) + subps xmm2,xmm3 ; xmm2=tmp5 + + movaps xmm1,xmm6 ; transpose coefficients(phase 1) + unpcklps xmm6,xmm7 ; xmm6=(00 10 01 11) + unpckhps xmm1,xmm7 ; xmm1=(02 12 03 13) + movaps xmm3,xmm0 ; transpose coefficients(phase 1) + unpcklps xmm0,xmm5 ; xmm0=(60 70 61 71) + unpckhps xmm3,xmm5 ; xmm3=(62 72 63 73) + + movaps xmm7, XMMWORD [wk(0)] ; xmm7=tmp2 + movaps xmm5, XMMWORD [wk(1)] ; xmm5=tmp3 + + movaps XMMWORD [wk(0)], xmm0 ; wk(0)=(60 70 61 71) + movaps XMMWORD [wk(1)], xmm3 ; wk(1)=(62 72 63 73) + + addps xmm4,xmm2 ; xmm4=tmp4 + movaps xmm0,xmm7 + movaps xmm3,xmm5 + addps xmm7,xmm2 ; xmm7=data2=(20 21 22 23) + addps xmm5,xmm4 ; xmm5=data4=(40 41 42 43) + subps xmm0,xmm2 ; xmm0=data5=(50 51 52 53) + subps xmm3,xmm4 ; xmm3=data3=(30 31 32 33) + + movaps xmm2,xmm7 ; transpose coefficients(phase 1) + unpcklps xmm7,xmm3 ; xmm7=(20 30 21 31) + unpckhps xmm2,xmm3 ; xmm2=(22 32 23 33) + movaps xmm4,xmm5 ; transpose coefficients(phase 1) + unpcklps xmm5,xmm0 ; xmm5=(40 50 41 51) + unpckhps xmm4,xmm0 ; xmm4=(42 52 43 53) + + movaps xmm3,xmm6 ; transpose coefficients(phase 2) + unpcklps2 xmm6,xmm7 ; xmm6=(00 10 20 30) + unpckhps2 xmm3,xmm7 ; xmm3=(01 11 21 31) + movaps xmm0,xmm1 ; transpose coefficients(phase 2) + unpcklps2 xmm1,xmm2 ; xmm1=(02 12 22 32) + unpckhps2 xmm0,xmm2 ; xmm0=(03 13 23 33) + + movaps xmm7, XMMWORD [wk(0)] ; xmm7=(60 70 61 71) + movaps xmm2, XMMWORD [wk(1)] ; xmm2=(62 72 63 73) + + movaps XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm6 + movaps XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm3 + movaps XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT)], xmm1 + movaps XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT)], xmm0 + + movaps xmm6,xmm5 ; transpose coefficients(phase 2) + unpcklps2 xmm5,xmm7 ; xmm5=(40 50 60 70) + unpckhps2 xmm6,xmm7 ; xmm6=(41 51 61 71) + movaps xmm3,xmm4 ; transpose coefficients(phase 2) + unpcklps2 xmm4,xmm2 ; xmm4=(42 52 62 72) + unpckhps2 xmm3,xmm2 ; xmm3=(43 53 63 73) + + movaps XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm5 + movaps XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm6 + movaps XMMWORD [XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT)], xmm4 + movaps XMMWORD [XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT)], xmm3 + +.nextcolumn: + add esi, byte 4*SIZEOF_JCOEF ; coef_block + add edx, byte 4*SIZEOF_FLOAT_MULT_TYPE ; quantptr + add edi, 4*DCTSIZE*SIZEOF_FAST_FLOAT ; wsptr + dec ecx ; ctr + jnz near .columnloop + + ; -- Prefetch the next coefficient block + + prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32] + prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32] + prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32] + prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32] + + ; ---- Pass 2: process rows from work array, store into output array. + + mov eax, [original_ebp] + lea esi, [workspace] ; FAST_FLOAT * wsptr + mov edi, JSAMPARRAY [output_buf(eax)] ; (JSAMPROW *) + mov eax, JDIMENSION [output_col(eax)] + mov ecx, DCTSIZE/4 ; ctr + alignx 16,7 +.rowloop: + + ; -- Even part + + movaps xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm2, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_FAST_FLOAT)] + + movaps xmm4,xmm0 + movaps xmm5,xmm1 + subps xmm0,xmm2 ; xmm0=tmp11 + subps xmm1,xmm3 + addps xmm4,xmm2 ; xmm4=tmp10 + addps xmm5,xmm3 ; xmm5=tmp13 + + mulps xmm1,[GOTOFF(ebx,PD_1_414)] + subps xmm1,xmm5 ; xmm1=tmp12 + + movaps xmm6,xmm4 + movaps xmm7,xmm0 + subps xmm4,xmm5 ; xmm4=tmp3 + subps xmm0,xmm1 ; xmm0=tmp2 + addps xmm6,xmm5 ; xmm6=tmp0 + addps xmm7,xmm1 ; xmm7=tmp1 + + movaps XMMWORD [wk(1)], xmm4 ; tmp3 + movaps XMMWORD [wk(0)], xmm0 ; tmp2 + + ; -- Odd part + + movaps xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm5, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_FAST_FLOAT)] + + movaps xmm4,xmm2 + movaps xmm0,xmm5 + addps xmm2,xmm1 ; xmm2=z11 + addps xmm5,xmm3 ; xmm5=z13 + subps xmm4,xmm1 ; xmm4=z12 + subps xmm0,xmm3 ; xmm0=z10 + + movaps xmm1,xmm2 + subps xmm2,xmm5 + addps xmm1,xmm5 ; xmm1=tmp7 + + mulps xmm2,[GOTOFF(ebx,PD_1_414)] ; xmm2=tmp11 + + movaps xmm3,xmm0 + addps xmm0,xmm4 + mulps xmm0,[GOTOFF(ebx,PD_1_847)] ; xmm0=z5 + mulps xmm3,[GOTOFF(ebx,PD_M2_613)] ; xmm3=(z10 * -2.613125930) + mulps xmm4,[GOTOFF(ebx,PD_1_082)] ; xmm4=(z12 * 1.082392200) + addps xmm3,xmm0 ; xmm3=tmp12 + subps xmm4,xmm0 ; xmm4=tmp10 + + ; -- Final output stage + + subps xmm3,xmm1 ; xmm3=tmp6 + movaps xmm5,xmm6 + movaps xmm0,xmm7 + addps xmm6,xmm1 ; xmm6=data0=(00 10 20 30) + addps xmm7,xmm3 ; xmm7=data1=(01 11 21 31) + subps xmm5,xmm1 ; xmm5=data7=(07 17 27 37) + subps xmm0,xmm3 ; xmm0=data6=(06 16 26 36) + subps xmm2,xmm3 ; xmm2=tmp5 + + movaps xmm1,[GOTOFF(ebx,PD_0_125)] ; xmm1=[PD_0_125] + + mulps xmm6,xmm1 ; descale(1/8) + mulps xmm7,xmm1 ; descale(1/8) + mulps xmm5,xmm1 ; descale(1/8) + mulps xmm0,xmm1 ; descale(1/8) + + movhlps xmm3,xmm6 + movhlps xmm1,xmm7 + cvtps2pi mm0,xmm6 ; round to int32, mm0=data0L=(00 10) + cvtps2pi mm1,xmm7 ; round to int32, mm1=data1L=(01 11) + cvtps2pi mm2,xmm3 ; round to int32, mm2=data0H=(20 30) + cvtps2pi mm3,xmm1 ; round to int32, mm3=data1H=(21 31) + packssdw mm0,mm2 ; mm0=data0=(00 10 20 30) + packssdw mm1,mm3 ; mm1=data1=(01 11 21 31) + + movhlps xmm6,xmm5 + movhlps xmm7,xmm0 + cvtps2pi mm4,xmm5 ; round to int32, mm4=data7L=(07 17) + cvtps2pi mm5,xmm0 ; round to int32, mm5=data6L=(06 16) + cvtps2pi mm6,xmm6 ; round to int32, mm6=data7H=(27 37) + cvtps2pi mm7,xmm7 ; round to int32, mm7=data6H=(26 36) + packssdw mm4,mm6 ; mm4=data7=(07 17 27 37) + packssdw mm5,mm7 ; mm5=data6=(06 16 26 36) + + packsswb mm0,mm5 ; mm0=(00 10 20 30 06 16 26 36) + packsswb mm1,mm4 ; mm1=(01 11 21 31 07 17 27 37) + + movaps xmm3, XMMWORD [wk(0)] ; xmm3=tmp2 + movaps xmm1, XMMWORD [wk(1)] ; xmm1=tmp3 + + movaps xmm6,[GOTOFF(ebx,PD_0_125)] ; xmm6=[PD_0_125] + + addps xmm4,xmm2 ; xmm4=tmp4 + movaps xmm5,xmm3 + movaps xmm0,xmm1 + addps xmm3,xmm2 ; xmm3=data2=(02 12 22 32) + addps xmm1,xmm4 ; xmm1=data4=(04 14 24 34) + subps xmm5,xmm2 ; xmm5=data5=(05 15 25 35) + subps xmm0,xmm4 ; xmm0=data3=(03 13 23 33) + + mulps xmm3,xmm6 ; descale(1/8) + mulps xmm1,xmm6 ; descale(1/8) + mulps xmm5,xmm6 ; descale(1/8) + mulps xmm0,xmm6 ; descale(1/8) + + movhlps xmm7,xmm3 + movhlps xmm2,xmm1 + cvtps2pi mm2,xmm3 ; round to int32, mm2=data2L=(02 12) + cvtps2pi mm3,xmm1 ; round to int32, mm3=data4L=(04 14) + cvtps2pi mm6,xmm7 ; round to int32, mm6=data2H=(22 32) + cvtps2pi mm7,xmm2 ; round to int32, mm7=data4H=(24 34) + packssdw mm2,mm6 ; mm2=data2=(02 12 22 32) + packssdw mm3,mm7 ; mm3=data4=(04 14 24 34) + + movhlps xmm4,xmm5 + movhlps xmm6,xmm0 + cvtps2pi mm5,xmm5 ; round to int32, mm5=data5L=(05 15) + cvtps2pi mm4,xmm0 ; round to int32, mm4=data3L=(03 13) + cvtps2pi mm6,xmm4 ; round to int32, mm6=data5H=(25 35) + cvtps2pi mm7,xmm6 ; round to int32, mm7=data3H=(23 33) + packssdw mm5,mm6 ; mm5=data5=(05 15 25 35) + packssdw mm4,mm7 ; mm4=data3=(03 13 23 33) + + movq mm6,[GOTOFF(ebx,PB_CENTERJSAMP)] ; mm6=[PB_CENTERJSAMP] + + packsswb mm2,mm3 ; mm2=(02 12 22 32 04 14 24 34) + packsswb mm4,mm5 ; mm4=(03 13 23 33 05 15 25 35) + + paddb mm0,mm6 + paddb mm1,mm6 + paddb mm2,mm6 + paddb mm4,mm6 + + movq mm7,mm0 ; transpose coefficients(phase 1) + punpcklbw mm0,mm1 ; mm0=(00 01 10 11 20 21 30 31) + punpckhbw mm7,mm1 ; mm7=(06 07 16 17 26 27 36 37) + movq mm3,mm2 ; transpose coefficients(phase 1) + punpcklbw mm2,mm4 ; mm2=(02 03 12 13 22 23 32 33) + punpckhbw mm3,mm4 ; mm3=(04 05 14 15 24 25 34 35) + + movq mm5,mm0 ; transpose coefficients(phase 2) + punpcklwd mm0,mm2 ; mm0=(00 01 02 03 10 11 12 13) + punpckhwd mm5,mm2 ; mm5=(20 21 22 23 30 31 32 33) + movq mm6,mm3 ; transpose coefficients(phase 2) + punpcklwd mm3,mm7 ; mm3=(04 05 06 07 14 15 16 17) + punpckhwd mm6,mm7 ; mm6=(24 25 26 27 34 35 36 37) + + movq mm1,mm0 ; transpose coefficients(phase 3) + punpckldq mm0,mm3 ; mm0=(00 01 02 03 04 05 06 07) + punpckhdq mm1,mm3 ; mm1=(10 11 12 13 14 15 16 17) + movq mm4,mm5 ; transpose coefficients(phase 3) + punpckldq mm5,mm6 ; mm5=(20 21 22 23 24 25 26 27) + punpckhdq mm4,mm6 ; mm4=(30 31 32 33 34 35 36 37) + + pushpic ebx ; save GOT address + + mov edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] + mov ebx, JSAMPROW [edi+1*SIZEOF_JSAMPROW] + movq MMWORD [edx+eax*SIZEOF_JSAMPLE], mm0 + movq MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm1 + mov edx, JSAMPROW [edi+2*SIZEOF_JSAMPROW] + mov ebx, JSAMPROW [edi+3*SIZEOF_JSAMPROW] + movq MMWORD [edx+eax*SIZEOF_JSAMPLE], mm5 + movq MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm4 + + poppic ebx ; restore GOT address + + add esi, byte 4*SIZEOF_FAST_FLOAT ; wsptr + add edi, byte 4*SIZEOF_JSAMPROW + dec ecx ; ctr + jnz near .rowloop + + emms ; empty MMX state + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jidctflt-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jidctflt-sse2-64.asm new file mode 100644 index 000000000..32e4ec222 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jidctflt-sse2-64.asm @@ -0,0 +1,483 @@ +; +; jidctflt.asm - floating-point IDCT (64-bit SSE & SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a floating-point implementation of the inverse DCT +; (Discrete Cosine Transform). The following code is based directly on +; the IJG's original jidctflt.c; see the jidctflt.c for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%macro unpcklps2 2 ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5) + shufps %1,%2,0x44 +%endmacro + +%macro unpckhps2 2 ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7) + shufps %1,%2,0xEE +%endmacro + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_idct_float_sse2) + +EXTN(jconst_idct_float_sse2): + +PD_1_414 times 4 dd 1.414213562373095048801689 +PD_1_847 times 4 dd 1.847759065022573512256366 +PD_1_082 times 4 dd 1.082392200292393968799446 +PD_M2_613 times 4 dd -2.613125929752753055713286 +PD_RNDINT_MAGIC times 4 dd 100663296.0 ; (float)(0x00C00000 << 3) +PB_CENTERJSAMP times 16 db CENTERJSAMPLE + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 +; +; Perform dequantization and inverse DCT on one block of coefficients. +; +; GLOBAL(void) +; jsimd_idct_float_sse2 (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +; r10 = void * dct_table +; r11 = JCOEFPTR coef_block +; r12 = JSAMPARRAY output_buf +; r13 = JDIMENSION output_col + +%define original_rbp rbp+0 +%define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 +%define workspace wk(0)-DCTSIZE2*SIZEOF_FAST_FLOAT + ; FAST_FLOAT workspace[DCTSIZE2] + + align 16 + global EXTN(jsimd_idct_float_sse2) + +EXTN(jsimd_idct_float_sse2): + push rbp + mov rax,rsp ; rax = original rbp + sub rsp, byte 4 + and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [rsp],rax + mov rbp,rsp ; rbp = aligned rbp + lea rsp, [workspace] + collect_args + push rbx + + ; ---- Pass 1: process columns from input, store into work array. + + mov rdx, r10 ; quantptr + mov rsi, r11 ; inptr + lea rdi, [workspace] ; FAST_FLOAT * wsptr + mov rcx, DCTSIZE/4 ; ctr +.columnloop: +%ifndef NO_ZERO_COLUMN_TEST_FLOAT_SSE + mov eax, DWORD [DWBLOCK(1,0,rsi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(2,0,rsi,SIZEOF_JCOEF)] + jnz near .columnDCT + + movq xmm1, XMM_MMWORD [MMBLOCK(1,0,rsi,SIZEOF_JCOEF)] + movq xmm2, XMM_MMWORD [MMBLOCK(2,0,rsi,SIZEOF_JCOEF)] + movq xmm3, XMM_MMWORD [MMBLOCK(3,0,rsi,SIZEOF_JCOEF)] + movq xmm4, XMM_MMWORD [MMBLOCK(4,0,rsi,SIZEOF_JCOEF)] + movq xmm5, XMM_MMWORD [MMBLOCK(5,0,rsi,SIZEOF_JCOEF)] + movq xmm6, XMM_MMWORD [MMBLOCK(6,0,rsi,SIZEOF_JCOEF)] + movq xmm7, XMM_MMWORD [MMBLOCK(7,0,rsi,SIZEOF_JCOEF)] + por xmm1,xmm2 + por xmm3,xmm4 + por xmm5,xmm6 + por xmm1,xmm3 + por xmm5,xmm7 + por xmm1,xmm5 + packsswb xmm1,xmm1 + movd eax,xmm1 + test rax,rax + jnz short .columnDCT + + ; -- AC terms all zero + + movq xmm0, XMM_MMWORD [MMBLOCK(0,0,rsi,SIZEOF_JCOEF)] + + punpcklwd xmm0,xmm0 ; xmm0=(00 00 01 01 02 02 03 03) + psrad xmm0,(DWORD_BIT-WORD_BIT) ; xmm0=in0=(00 01 02 03) + cvtdq2ps xmm0,xmm0 ; xmm0=in0=(00 01 02 03) + + mulps xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FLOAT_MULT_TYPE)] + + movaps xmm1,xmm0 + movaps xmm2,xmm0 + movaps xmm3,xmm0 + + shufps xmm0,xmm0,0x00 ; xmm0=(00 00 00 00) + shufps xmm1,xmm1,0x55 ; xmm1=(01 01 01 01) + shufps xmm2,xmm2,0xAA ; xmm2=(02 02 02 02) + shufps xmm3,xmm3,0xFF ; xmm3=(03 03 03 03) + + movaps XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_FAST_FLOAT)], xmm0 + movaps XMMWORD [XMMBLOCK(0,1,rdi,SIZEOF_FAST_FLOAT)], xmm0 + movaps XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_FAST_FLOAT)], xmm1 + movaps XMMWORD [XMMBLOCK(1,1,rdi,SIZEOF_FAST_FLOAT)], xmm1 + movaps XMMWORD [XMMBLOCK(2,0,rdi,SIZEOF_FAST_FLOAT)], xmm2 + movaps XMMWORD [XMMBLOCK(2,1,rdi,SIZEOF_FAST_FLOAT)], xmm2 + movaps XMMWORD [XMMBLOCK(3,0,rdi,SIZEOF_FAST_FLOAT)], xmm3 + movaps XMMWORD [XMMBLOCK(3,1,rdi,SIZEOF_FAST_FLOAT)], xmm3 + jmp near .nextcolumn +%endif +.columnDCT: + + ; -- Even part + + movq xmm0, XMM_MMWORD [MMBLOCK(0,0,rsi,SIZEOF_JCOEF)] + movq xmm1, XMM_MMWORD [MMBLOCK(2,0,rsi,SIZEOF_JCOEF)] + movq xmm2, XMM_MMWORD [MMBLOCK(4,0,rsi,SIZEOF_JCOEF)] + movq xmm3, XMM_MMWORD [MMBLOCK(6,0,rsi,SIZEOF_JCOEF)] + + punpcklwd xmm0,xmm0 ; xmm0=(00 00 01 01 02 02 03 03) + punpcklwd xmm1,xmm1 ; xmm1=(20 20 21 21 22 22 23 23) + psrad xmm0,(DWORD_BIT-WORD_BIT) ; xmm0=in0=(00 01 02 03) + psrad xmm1,(DWORD_BIT-WORD_BIT) ; xmm1=in2=(20 21 22 23) + cvtdq2ps xmm0,xmm0 ; xmm0=in0=(00 01 02 03) + cvtdq2ps xmm1,xmm1 ; xmm1=in2=(20 21 22 23) + + punpcklwd xmm2,xmm2 ; xmm2=(40 40 41 41 42 42 43 43) + punpcklwd xmm3,xmm3 ; xmm3=(60 60 61 61 62 62 63 63) + psrad xmm2,(DWORD_BIT-WORD_BIT) ; xmm2=in4=(40 41 42 43) + psrad xmm3,(DWORD_BIT-WORD_BIT) ; xmm3=in6=(60 61 62 63) + cvtdq2ps xmm2,xmm2 ; xmm2=in4=(40 41 42 43) + cvtdq2ps xmm3,xmm3 ; xmm3=in6=(60 61 62 63) + + mulps xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm1, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm2, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm3, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_FLOAT_MULT_TYPE)] + + movaps xmm4,xmm0 + movaps xmm5,xmm1 + subps xmm0,xmm2 ; xmm0=tmp11 + subps xmm1,xmm3 + addps xmm4,xmm2 ; xmm4=tmp10 + addps xmm5,xmm3 ; xmm5=tmp13 + + mulps xmm1,[rel PD_1_414] + subps xmm1,xmm5 ; xmm1=tmp12 + + movaps xmm6,xmm4 + movaps xmm7,xmm0 + subps xmm4,xmm5 ; xmm4=tmp3 + subps xmm0,xmm1 ; xmm0=tmp2 + addps xmm6,xmm5 ; xmm6=tmp0 + addps xmm7,xmm1 ; xmm7=tmp1 + + movaps XMMWORD [wk(1)], xmm4 ; tmp3 + movaps XMMWORD [wk(0)], xmm0 ; tmp2 + + ; -- Odd part + + movq xmm2, XMM_MMWORD [MMBLOCK(1,0,rsi,SIZEOF_JCOEF)] + movq xmm3, XMM_MMWORD [MMBLOCK(3,0,rsi,SIZEOF_JCOEF)] + movq xmm5, XMM_MMWORD [MMBLOCK(5,0,rsi,SIZEOF_JCOEF)] + movq xmm1, XMM_MMWORD [MMBLOCK(7,0,rsi,SIZEOF_JCOEF)] + + punpcklwd xmm2,xmm2 ; xmm2=(10 10 11 11 12 12 13 13) + punpcklwd xmm3,xmm3 ; xmm3=(30 30 31 31 32 32 33 33) + psrad xmm2,(DWORD_BIT-WORD_BIT) ; xmm2=in1=(10 11 12 13) + psrad xmm3,(DWORD_BIT-WORD_BIT) ; xmm3=in3=(30 31 32 33) + cvtdq2ps xmm2,xmm2 ; xmm2=in1=(10 11 12 13) + cvtdq2ps xmm3,xmm3 ; xmm3=in3=(30 31 32 33) + + punpcklwd xmm5,xmm5 ; xmm5=(50 50 51 51 52 52 53 53) + punpcklwd xmm1,xmm1 ; xmm1=(70 70 71 71 72 72 73 73) + psrad xmm5,(DWORD_BIT-WORD_BIT) ; xmm5=in5=(50 51 52 53) + psrad xmm1,(DWORD_BIT-WORD_BIT) ; xmm1=in7=(70 71 72 73) + cvtdq2ps xmm5,xmm5 ; xmm5=in5=(50 51 52 53) + cvtdq2ps xmm1,xmm1 ; xmm1=in7=(70 71 72 73) + + mulps xmm2, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm3, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm5, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm1, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_FLOAT_MULT_TYPE)] + + movaps xmm4,xmm2 + movaps xmm0,xmm5 + addps xmm2,xmm1 ; xmm2=z11 + addps xmm5,xmm3 ; xmm5=z13 + subps xmm4,xmm1 ; xmm4=z12 + subps xmm0,xmm3 ; xmm0=z10 + + movaps xmm1,xmm2 + subps xmm2,xmm5 + addps xmm1,xmm5 ; xmm1=tmp7 + + mulps xmm2,[rel PD_1_414] ; xmm2=tmp11 + + movaps xmm3,xmm0 + addps xmm0,xmm4 + mulps xmm0,[rel PD_1_847] ; xmm0=z5 + mulps xmm3,[rel PD_M2_613] ; xmm3=(z10 * -2.613125930) + mulps xmm4,[rel PD_1_082] ; xmm4=(z12 * 1.082392200) + addps xmm3,xmm0 ; xmm3=tmp12 + subps xmm4,xmm0 ; xmm4=tmp10 + + ; -- Final output stage + + subps xmm3,xmm1 ; xmm3=tmp6 + movaps xmm5,xmm6 + movaps xmm0,xmm7 + addps xmm6,xmm1 ; xmm6=data0=(00 01 02 03) + addps xmm7,xmm3 ; xmm7=data1=(10 11 12 13) + subps xmm5,xmm1 ; xmm5=data7=(70 71 72 73) + subps xmm0,xmm3 ; xmm0=data6=(60 61 62 63) + subps xmm2,xmm3 ; xmm2=tmp5 + + movaps xmm1,xmm6 ; transpose coefficients(phase 1) + unpcklps xmm6,xmm7 ; xmm6=(00 10 01 11) + unpckhps xmm1,xmm7 ; xmm1=(02 12 03 13) + movaps xmm3,xmm0 ; transpose coefficients(phase 1) + unpcklps xmm0,xmm5 ; xmm0=(60 70 61 71) + unpckhps xmm3,xmm5 ; xmm3=(62 72 63 73) + + movaps xmm7, XMMWORD [wk(0)] ; xmm7=tmp2 + movaps xmm5, XMMWORD [wk(1)] ; xmm5=tmp3 + + movaps XMMWORD [wk(0)], xmm0 ; wk(0)=(60 70 61 71) + movaps XMMWORD [wk(1)], xmm3 ; wk(1)=(62 72 63 73) + + addps xmm4,xmm2 ; xmm4=tmp4 + movaps xmm0,xmm7 + movaps xmm3,xmm5 + addps xmm7,xmm2 ; xmm7=data2=(20 21 22 23) + addps xmm5,xmm4 ; xmm5=data4=(40 41 42 43) + subps xmm0,xmm2 ; xmm0=data5=(50 51 52 53) + subps xmm3,xmm4 ; xmm3=data3=(30 31 32 33) + + movaps xmm2,xmm7 ; transpose coefficients(phase 1) + unpcklps xmm7,xmm3 ; xmm7=(20 30 21 31) + unpckhps xmm2,xmm3 ; xmm2=(22 32 23 33) + movaps xmm4,xmm5 ; transpose coefficients(phase 1) + unpcklps xmm5,xmm0 ; xmm5=(40 50 41 51) + unpckhps xmm4,xmm0 ; xmm4=(42 52 43 53) + + movaps xmm3,xmm6 ; transpose coefficients(phase 2) + unpcklps2 xmm6,xmm7 ; xmm6=(00 10 20 30) + unpckhps2 xmm3,xmm7 ; xmm3=(01 11 21 31) + movaps xmm0,xmm1 ; transpose coefficients(phase 2) + unpcklps2 xmm1,xmm2 ; xmm1=(02 12 22 32) + unpckhps2 xmm0,xmm2 ; xmm0=(03 13 23 33) + + movaps xmm7, XMMWORD [wk(0)] ; xmm7=(60 70 61 71) + movaps xmm2, XMMWORD [wk(1)] ; xmm2=(62 72 63 73) + + movaps XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_FAST_FLOAT)], xmm6 + movaps XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_FAST_FLOAT)], xmm3 + movaps XMMWORD [XMMBLOCK(2,0,rdi,SIZEOF_FAST_FLOAT)], xmm1 + movaps XMMWORD [XMMBLOCK(3,0,rdi,SIZEOF_FAST_FLOAT)], xmm0 + + movaps xmm6,xmm5 ; transpose coefficients(phase 2) + unpcklps2 xmm5,xmm7 ; xmm5=(40 50 60 70) + unpckhps2 xmm6,xmm7 ; xmm6=(41 51 61 71) + movaps xmm3,xmm4 ; transpose coefficients(phase 2) + unpcklps2 xmm4,xmm2 ; xmm4=(42 52 62 72) + unpckhps2 xmm3,xmm2 ; xmm3=(43 53 63 73) + + movaps XMMWORD [XMMBLOCK(0,1,rdi,SIZEOF_FAST_FLOAT)], xmm5 + movaps XMMWORD [XMMBLOCK(1,1,rdi,SIZEOF_FAST_FLOAT)], xmm6 + movaps XMMWORD [XMMBLOCK(2,1,rdi,SIZEOF_FAST_FLOAT)], xmm4 + movaps XMMWORD [XMMBLOCK(3,1,rdi,SIZEOF_FAST_FLOAT)], xmm3 + +.nextcolumn: + add rsi, byte 4*SIZEOF_JCOEF ; coef_block + add rdx, byte 4*SIZEOF_FLOAT_MULT_TYPE ; quantptr + add rdi, 4*DCTSIZE*SIZEOF_FAST_FLOAT ; wsptr + dec rcx ; ctr + jnz near .columnloop + + ; -- Prefetch the next coefficient block + + prefetchnta [rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32] + prefetchnta [rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32] + prefetchnta [rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32] + prefetchnta [rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32] + + ; ---- Pass 2: process rows from work array, store into output array. + + mov rax, [original_rbp] + lea rsi, [workspace] ; FAST_FLOAT * wsptr + mov rdi, r12 ; (JSAMPROW *) + mov rax, r13 + mov rcx, DCTSIZE/4 ; ctr +.rowloop: + + ; -- Even part + + movaps xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_FAST_FLOAT)] + movaps xmm2, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_FAST_FLOAT)] + + movaps xmm4,xmm0 + movaps xmm5,xmm1 + subps xmm0,xmm2 ; xmm0=tmp11 + subps xmm1,xmm3 + addps xmm4,xmm2 ; xmm4=tmp10 + addps xmm5,xmm3 ; xmm5=tmp13 + + mulps xmm1,[rel PD_1_414] + subps xmm1,xmm5 ; xmm1=tmp12 + + movaps xmm6,xmm4 + movaps xmm7,xmm0 + subps xmm4,xmm5 ; xmm4=tmp3 + subps xmm0,xmm1 ; xmm0=tmp2 + addps xmm6,xmm5 ; xmm6=tmp0 + addps xmm7,xmm1 ; xmm7=tmp1 + + movaps XMMWORD [wk(1)], xmm4 ; tmp3 + movaps XMMWORD [wk(0)], xmm0 ; tmp2 + + ; -- Odd part + + movaps xmm2, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_FAST_FLOAT)] + movaps xmm5, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_FAST_FLOAT)] + + movaps xmm4,xmm2 + movaps xmm0,xmm5 + addps xmm2,xmm1 ; xmm2=z11 + addps xmm5,xmm3 ; xmm5=z13 + subps xmm4,xmm1 ; xmm4=z12 + subps xmm0,xmm3 ; xmm0=z10 + + movaps xmm1,xmm2 + subps xmm2,xmm5 + addps xmm1,xmm5 ; xmm1=tmp7 + + mulps xmm2,[rel PD_1_414] ; xmm2=tmp11 + + movaps xmm3,xmm0 + addps xmm0,xmm4 + mulps xmm0,[rel PD_1_847] ; xmm0=z5 + mulps xmm3,[rel PD_M2_613] ; xmm3=(z10 * -2.613125930) + mulps xmm4,[rel PD_1_082] ; xmm4=(z12 * 1.082392200) + addps xmm3,xmm0 ; xmm3=tmp12 + subps xmm4,xmm0 ; xmm4=tmp10 + + ; -- Final output stage + + subps xmm3,xmm1 ; xmm3=tmp6 + movaps xmm5,xmm6 + movaps xmm0,xmm7 + addps xmm6,xmm1 ; xmm6=data0=(00 10 20 30) + addps xmm7,xmm3 ; xmm7=data1=(01 11 21 31) + subps xmm5,xmm1 ; xmm5=data7=(07 17 27 37) + subps xmm0,xmm3 ; xmm0=data6=(06 16 26 36) + subps xmm2,xmm3 ; xmm2=tmp5 + + movaps xmm1,[rel PD_RNDINT_MAGIC] ; xmm1=[rel PD_RNDINT_MAGIC] + pcmpeqd xmm3,xmm3 + psrld xmm3,WORD_BIT ; xmm3={0xFFFF 0x0000 0xFFFF 0x0000 ..} + + addps xmm6,xmm1 ; xmm6=roundint(data0/8)=(00 ** 10 ** 20 ** 30 **) + addps xmm7,xmm1 ; xmm7=roundint(data1/8)=(01 ** 11 ** 21 ** 31 **) + addps xmm0,xmm1 ; xmm0=roundint(data6/8)=(06 ** 16 ** 26 ** 36 **) + addps xmm5,xmm1 ; xmm5=roundint(data7/8)=(07 ** 17 ** 27 ** 37 **) + + pand xmm6,xmm3 ; xmm6=(00 -- 10 -- 20 -- 30 --) + pslld xmm7,WORD_BIT ; xmm7=(-- 01 -- 11 -- 21 -- 31) + pand xmm0,xmm3 ; xmm0=(06 -- 16 -- 26 -- 36 --) + pslld xmm5,WORD_BIT ; xmm5=(-- 07 -- 17 -- 27 -- 37) + por xmm6,xmm7 ; xmm6=(00 01 10 11 20 21 30 31) + por xmm0,xmm5 ; xmm0=(06 07 16 17 26 27 36 37) + + movaps xmm1, XMMWORD [wk(0)] ; xmm1=tmp2 + movaps xmm3, XMMWORD [wk(1)] ; xmm3=tmp3 + + addps xmm4,xmm2 ; xmm4=tmp4 + movaps xmm7,xmm1 + movaps xmm5,xmm3 + addps xmm1,xmm2 ; xmm1=data2=(02 12 22 32) + addps xmm3,xmm4 ; xmm3=data4=(04 14 24 34) + subps xmm7,xmm2 ; xmm7=data5=(05 15 25 35) + subps xmm5,xmm4 ; xmm5=data3=(03 13 23 33) + + movaps xmm2,[rel PD_RNDINT_MAGIC] ; xmm2=[rel PD_RNDINT_MAGIC] + pcmpeqd xmm4,xmm4 + psrld xmm4,WORD_BIT ; xmm4={0xFFFF 0x0000 0xFFFF 0x0000 ..} + + addps xmm3,xmm2 ; xmm3=roundint(data4/8)=(04 ** 14 ** 24 ** 34 **) + addps xmm7,xmm2 ; xmm7=roundint(data5/8)=(05 ** 15 ** 25 ** 35 **) + addps xmm1,xmm2 ; xmm1=roundint(data2/8)=(02 ** 12 ** 22 ** 32 **) + addps xmm5,xmm2 ; xmm5=roundint(data3/8)=(03 ** 13 ** 23 ** 33 **) + + pand xmm3,xmm4 ; xmm3=(04 -- 14 -- 24 -- 34 --) + pslld xmm7,WORD_BIT ; xmm7=(-- 05 -- 15 -- 25 -- 35) + pand xmm1,xmm4 ; xmm1=(02 -- 12 -- 22 -- 32 --) + pslld xmm5,WORD_BIT ; xmm5=(-- 03 -- 13 -- 23 -- 33) + por xmm3,xmm7 ; xmm3=(04 05 14 15 24 25 34 35) + por xmm1,xmm5 ; xmm1=(02 03 12 13 22 23 32 33) + + movdqa xmm2,[rel PB_CENTERJSAMP] ; xmm2=[rel PB_CENTERJSAMP] + + packsswb xmm6,xmm3 ; xmm6=(00 01 10 11 20 21 30 31 04 05 14 15 24 25 34 35) + packsswb xmm1,xmm0 ; xmm1=(02 03 12 13 22 23 32 33 06 07 16 17 26 27 36 37) + paddb xmm6,xmm2 + paddb xmm1,xmm2 + + movdqa xmm4,xmm6 ; transpose coefficients(phase 2) + punpcklwd xmm6,xmm1 ; xmm6=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33) + punpckhwd xmm4,xmm1 ; xmm4=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37) + + movdqa xmm7,xmm6 ; transpose coefficients(phase 3) + punpckldq xmm6,xmm4 ; xmm6=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17) + punpckhdq xmm7,xmm4 ; xmm7=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37) + + pshufd xmm5,xmm6,0x4E ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07) + pshufd xmm3,xmm7,0x4E ; xmm3=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27) + + mov rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW] + mov rbx, JSAMPROW [rdi+2*SIZEOF_JSAMPROW] + movq XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm6 + movq XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE], xmm7 + mov rdx, JSAMPROW [rdi+1*SIZEOF_JSAMPROW] + mov rbx, JSAMPROW [rdi+3*SIZEOF_JSAMPROW] + movq XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm5 + movq XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE], xmm3 + + add rsi, byte 4*SIZEOF_FAST_FLOAT ; wsptr + add rdi, byte 4*SIZEOF_JSAMPROW + dec rcx ; ctr + jnz near .rowloop + + pop rbx + uncollect_args + mov rsp,rbp ; rsp <- aligned rbp + pop rsp ; rsp <- original rbp + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jidctflt-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jidctflt-sse2.asm new file mode 100644 index 000000000..de2cd1f69 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jidctflt-sse2.asm @@ -0,0 +1,498 @@ +; +; jidctflt.asm - floating-point IDCT (SSE & SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a floating-point implementation of the inverse DCT +; (Discrete Cosine Transform). The following code is based directly on +; the IJG's original jidctflt.c; see the jidctflt.c for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%macro unpcklps2 2 ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5) + shufps %1,%2,0x44 +%endmacro + +%macro unpckhps2 2 ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7) + shufps %1,%2,0xEE +%endmacro + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_idct_float_sse2) + +EXTN(jconst_idct_float_sse2): + +PD_1_414 times 4 dd 1.414213562373095048801689 +PD_1_847 times 4 dd 1.847759065022573512256366 +PD_1_082 times 4 dd 1.082392200292393968799446 +PD_M2_613 times 4 dd -2.613125929752753055713286 +PD_RNDINT_MAGIC times 4 dd 100663296.0 ; (float)(0x00C00000 << 3) +PB_CENTERJSAMP times 16 db CENTERJSAMPLE + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform dequantization and inverse DCT on one block of coefficients. +; +; GLOBAL(void) +; jsimd_idct_float_sse2 (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +%define dct_table(b) (b)+8 ; void * dct_table +%define coef_block(b) (b)+12 ; JCOEFPTR coef_block +%define output_buf(b) (b)+16 ; JSAMPARRAY output_buf +%define output_col(b) (b)+20 ; JDIMENSION output_col + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 +%define workspace wk(0)-DCTSIZE2*SIZEOF_FAST_FLOAT + ; FAST_FLOAT workspace[DCTSIZE2] + + align 16 + global EXTN(jsimd_idct_float_sse2) + +EXTN(jsimd_idct_float_sse2): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [workspace] + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process columns from input, store into work array. + +; mov eax, [original_ebp] + mov edx, POINTER [dct_table(eax)] ; quantptr + mov esi, JCOEFPTR [coef_block(eax)] ; inptr + lea edi, [workspace] ; FAST_FLOAT * wsptr + mov ecx, DCTSIZE/4 ; ctr + alignx 16,7 +.columnloop: +%ifndef NO_ZERO_COLUMN_TEST_FLOAT_SSE + mov eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)] + jnz near .columnDCT + + movq xmm1, XMM_MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq xmm2, XMM_MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)] + movq xmm3, XMM_MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + movq xmm4, XMM_MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)] + movq xmm5, XMM_MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movq xmm6, XMM_MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)] + movq xmm7, XMM_MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + por xmm1,xmm2 + por xmm3,xmm4 + por xmm5,xmm6 + por xmm1,xmm3 + por xmm5,xmm7 + por xmm1,xmm5 + packsswb xmm1,xmm1 + movd eax,xmm1 + test eax,eax + jnz short .columnDCT + + ; -- AC terms all zero + + movq xmm0, XMM_MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + + punpcklwd xmm0,xmm0 ; xmm0=(00 00 01 01 02 02 03 03) + psrad xmm0,(DWORD_BIT-WORD_BIT) ; xmm0=in0=(00 01 02 03) + cvtdq2ps xmm0,xmm0 ; xmm0=in0=(00 01 02 03) + + mulps xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + + movaps xmm1,xmm0 + movaps xmm2,xmm0 + movaps xmm3,xmm0 + + shufps xmm0,xmm0,0x00 ; xmm0=(00 00 00 00) + shufps xmm1,xmm1,0x55 ; xmm1=(01 01 01 01) + shufps xmm2,xmm2,0xAA ; xmm2=(02 02 02 02) + shufps xmm3,xmm3,0xFF ; xmm3=(03 03 03 03) + + movaps XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm0 + movaps XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm0 + movaps XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm1 + movaps XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm1 + movaps XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT)], xmm2 + movaps XMMWORD [XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT)], xmm2 + movaps XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT)], xmm3 + movaps XMMWORD [XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT)], xmm3 + jmp near .nextcolumn + alignx 16,7 +%endif +.columnDCT: + + ; -- Even part + + movq xmm0, XMM_MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + movq xmm1, XMM_MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)] + movq xmm2, XMM_MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)] + movq xmm3, XMM_MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)] + + punpcklwd xmm0,xmm0 ; xmm0=(00 00 01 01 02 02 03 03) + punpcklwd xmm1,xmm1 ; xmm1=(20 20 21 21 22 22 23 23) + psrad xmm0,(DWORD_BIT-WORD_BIT) ; xmm0=in0=(00 01 02 03) + psrad xmm1,(DWORD_BIT-WORD_BIT) ; xmm1=in2=(20 21 22 23) + cvtdq2ps xmm0,xmm0 ; xmm0=in0=(00 01 02 03) + cvtdq2ps xmm1,xmm1 ; xmm1=in2=(20 21 22 23) + + punpcklwd xmm2,xmm2 ; xmm2=(40 40 41 41 42 42 43 43) + punpcklwd xmm3,xmm3 ; xmm3=(60 60 61 61 62 62 63 63) + psrad xmm2,(DWORD_BIT-WORD_BIT) ; xmm2=in4=(40 41 42 43) + psrad xmm3,(DWORD_BIT-WORD_BIT) ; xmm3=in6=(60 61 62 63) + cvtdq2ps xmm2,xmm2 ; xmm2=in4=(40 41 42 43) + cvtdq2ps xmm3,xmm3 ; xmm3=in6=(60 61 62 63) + + mulps xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm1, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm2, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm3, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + + movaps xmm4,xmm0 + movaps xmm5,xmm1 + subps xmm0,xmm2 ; xmm0=tmp11 + subps xmm1,xmm3 + addps xmm4,xmm2 ; xmm4=tmp10 + addps xmm5,xmm3 ; xmm5=tmp13 + + mulps xmm1,[GOTOFF(ebx,PD_1_414)] + subps xmm1,xmm5 ; xmm1=tmp12 + + movaps xmm6,xmm4 + movaps xmm7,xmm0 + subps xmm4,xmm5 ; xmm4=tmp3 + subps xmm0,xmm1 ; xmm0=tmp2 + addps xmm6,xmm5 ; xmm6=tmp0 + addps xmm7,xmm1 ; xmm7=tmp1 + + movaps XMMWORD [wk(1)], xmm4 ; tmp3 + movaps XMMWORD [wk(0)], xmm0 ; tmp2 + + ; -- Odd part + + movq xmm2, XMM_MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq xmm3, XMM_MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + movq xmm5, XMM_MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movq xmm1, XMM_MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + + punpcklwd xmm2,xmm2 ; xmm2=(10 10 11 11 12 12 13 13) + punpcklwd xmm3,xmm3 ; xmm3=(30 30 31 31 32 32 33 33) + psrad xmm2,(DWORD_BIT-WORD_BIT) ; xmm2=in1=(10 11 12 13) + psrad xmm3,(DWORD_BIT-WORD_BIT) ; xmm3=in3=(30 31 32 33) + cvtdq2ps xmm2,xmm2 ; xmm2=in1=(10 11 12 13) + cvtdq2ps xmm3,xmm3 ; xmm3=in3=(30 31 32 33) + + punpcklwd xmm5,xmm5 ; xmm5=(50 50 51 51 52 52 53 53) + punpcklwd xmm1,xmm1 ; xmm1=(70 70 71 71 72 72 73 73) + psrad xmm5,(DWORD_BIT-WORD_BIT) ; xmm5=in5=(50 51 52 53) + psrad xmm1,(DWORD_BIT-WORD_BIT) ; xmm1=in7=(70 71 72 73) + cvtdq2ps xmm5,xmm5 ; xmm5=in5=(50 51 52 53) + cvtdq2ps xmm1,xmm1 ; xmm1=in7=(70 71 72 73) + + mulps xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm5, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + mulps xmm1, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_FLOAT_MULT_TYPE)] + + movaps xmm4,xmm2 + movaps xmm0,xmm5 + addps xmm2,xmm1 ; xmm2=z11 + addps xmm5,xmm3 ; xmm5=z13 + subps xmm4,xmm1 ; xmm4=z12 + subps xmm0,xmm3 ; xmm0=z10 + + movaps xmm1,xmm2 + subps xmm2,xmm5 + addps xmm1,xmm5 ; xmm1=tmp7 + + mulps xmm2,[GOTOFF(ebx,PD_1_414)] ; xmm2=tmp11 + + movaps xmm3,xmm0 + addps xmm0,xmm4 + mulps xmm0,[GOTOFF(ebx,PD_1_847)] ; xmm0=z5 + mulps xmm3,[GOTOFF(ebx,PD_M2_613)] ; xmm3=(z10 * -2.613125930) + mulps xmm4,[GOTOFF(ebx,PD_1_082)] ; xmm4=(z12 * 1.082392200) + addps xmm3,xmm0 ; xmm3=tmp12 + subps xmm4,xmm0 ; xmm4=tmp10 + + ; -- Final output stage + + subps xmm3,xmm1 ; xmm3=tmp6 + movaps xmm5,xmm6 + movaps xmm0,xmm7 + addps xmm6,xmm1 ; xmm6=data0=(00 01 02 03) + addps xmm7,xmm3 ; xmm7=data1=(10 11 12 13) + subps xmm5,xmm1 ; xmm5=data7=(70 71 72 73) + subps xmm0,xmm3 ; xmm0=data6=(60 61 62 63) + subps xmm2,xmm3 ; xmm2=tmp5 + + movaps xmm1,xmm6 ; transpose coefficients(phase 1) + unpcklps xmm6,xmm7 ; xmm6=(00 10 01 11) + unpckhps xmm1,xmm7 ; xmm1=(02 12 03 13) + movaps xmm3,xmm0 ; transpose coefficients(phase 1) + unpcklps xmm0,xmm5 ; xmm0=(60 70 61 71) + unpckhps xmm3,xmm5 ; xmm3=(62 72 63 73) + + movaps xmm7, XMMWORD [wk(0)] ; xmm7=tmp2 + movaps xmm5, XMMWORD [wk(1)] ; xmm5=tmp3 + + movaps XMMWORD [wk(0)], xmm0 ; wk(0)=(60 70 61 71) + movaps XMMWORD [wk(1)], xmm3 ; wk(1)=(62 72 63 73) + + addps xmm4,xmm2 ; xmm4=tmp4 + movaps xmm0,xmm7 + movaps xmm3,xmm5 + addps xmm7,xmm2 ; xmm7=data2=(20 21 22 23) + addps xmm5,xmm4 ; xmm5=data4=(40 41 42 43) + subps xmm0,xmm2 ; xmm0=data5=(50 51 52 53) + subps xmm3,xmm4 ; xmm3=data3=(30 31 32 33) + + movaps xmm2,xmm7 ; transpose coefficients(phase 1) + unpcklps xmm7,xmm3 ; xmm7=(20 30 21 31) + unpckhps xmm2,xmm3 ; xmm2=(22 32 23 33) + movaps xmm4,xmm5 ; transpose coefficients(phase 1) + unpcklps xmm5,xmm0 ; xmm5=(40 50 41 51) + unpckhps xmm4,xmm0 ; xmm4=(42 52 43 53) + + movaps xmm3,xmm6 ; transpose coefficients(phase 2) + unpcklps2 xmm6,xmm7 ; xmm6=(00 10 20 30) + unpckhps2 xmm3,xmm7 ; xmm3=(01 11 21 31) + movaps xmm0,xmm1 ; transpose coefficients(phase 2) + unpcklps2 xmm1,xmm2 ; xmm1=(02 12 22 32) + unpckhps2 xmm0,xmm2 ; xmm0=(03 13 23 33) + + movaps xmm7, XMMWORD [wk(0)] ; xmm7=(60 70 61 71) + movaps xmm2, XMMWORD [wk(1)] ; xmm2=(62 72 63 73) + + movaps XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm6 + movaps XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm3 + movaps XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT)], xmm1 + movaps XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT)], xmm0 + + movaps xmm6,xmm5 ; transpose coefficients(phase 2) + unpcklps2 xmm5,xmm7 ; xmm5=(40 50 60 70) + unpckhps2 xmm6,xmm7 ; xmm6=(41 51 61 71) + movaps xmm3,xmm4 ; transpose coefficients(phase 2) + unpcklps2 xmm4,xmm2 ; xmm4=(42 52 62 72) + unpckhps2 xmm3,xmm2 ; xmm3=(43 53 63 73) + + movaps XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm5 + movaps XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm6 + movaps XMMWORD [XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT)], xmm4 + movaps XMMWORD [XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT)], xmm3 + +.nextcolumn: + add esi, byte 4*SIZEOF_JCOEF ; coef_block + add edx, byte 4*SIZEOF_FLOAT_MULT_TYPE ; quantptr + add edi, 4*DCTSIZE*SIZEOF_FAST_FLOAT ; wsptr + dec ecx ; ctr + jnz near .columnloop + + ; -- Prefetch the next coefficient block + + prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32] + prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32] + prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32] + prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32] + + ; ---- Pass 2: process rows from work array, store into output array. + + mov eax, [original_ebp] + lea esi, [workspace] ; FAST_FLOAT * wsptr + mov edi, JSAMPARRAY [output_buf(eax)] ; (JSAMPROW *) + mov eax, JDIMENSION [output_col(eax)] + mov ecx, DCTSIZE/4 ; ctr + alignx 16,7 +.rowloop: + + ; -- Even part + + movaps xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm2, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_FAST_FLOAT)] + + movaps xmm4,xmm0 + movaps xmm5,xmm1 + subps xmm0,xmm2 ; xmm0=tmp11 + subps xmm1,xmm3 + addps xmm4,xmm2 ; xmm4=tmp10 + addps xmm5,xmm3 ; xmm5=tmp13 + + mulps xmm1,[GOTOFF(ebx,PD_1_414)] + subps xmm1,xmm5 ; xmm1=tmp12 + + movaps xmm6,xmm4 + movaps xmm7,xmm0 + subps xmm4,xmm5 ; xmm4=tmp3 + subps xmm0,xmm1 ; xmm0=tmp2 + addps xmm6,xmm5 ; xmm6=tmp0 + addps xmm7,xmm1 ; xmm7=tmp1 + + movaps XMMWORD [wk(1)], xmm4 ; tmp3 + movaps XMMWORD [wk(0)], xmm0 ; tmp2 + + ; -- Odd part + + movaps xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm5, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_FAST_FLOAT)] + + movaps xmm4,xmm2 + movaps xmm0,xmm5 + addps xmm2,xmm1 ; xmm2=z11 + addps xmm5,xmm3 ; xmm5=z13 + subps xmm4,xmm1 ; xmm4=z12 + subps xmm0,xmm3 ; xmm0=z10 + + movaps xmm1,xmm2 + subps xmm2,xmm5 + addps xmm1,xmm5 ; xmm1=tmp7 + + mulps xmm2,[GOTOFF(ebx,PD_1_414)] ; xmm2=tmp11 + + movaps xmm3,xmm0 + addps xmm0,xmm4 + mulps xmm0,[GOTOFF(ebx,PD_1_847)] ; xmm0=z5 + mulps xmm3,[GOTOFF(ebx,PD_M2_613)] ; xmm3=(z10 * -2.613125930) + mulps xmm4,[GOTOFF(ebx,PD_1_082)] ; xmm4=(z12 * 1.082392200) + addps xmm3,xmm0 ; xmm3=tmp12 + subps xmm4,xmm0 ; xmm4=tmp10 + + ; -- Final output stage + + subps xmm3,xmm1 ; xmm3=tmp6 + movaps xmm5,xmm6 + movaps xmm0,xmm7 + addps xmm6,xmm1 ; xmm6=data0=(00 10 20 30) + addps xmm7,xmm3 ; xmm7=data1=(01 11 21 31) + subps xmm5,xmm1 ; xmm5=data7=(07 17 27 37) + subps xmm0,xmm3 ; xmm0=data6=(06 16 26 36) + subps xmm2,xmm3 ; xmm2=tmp5 + + movaps xmm1,[GOTOFF(ebx,PD_RNDINT_MAGIC)] ; xmm1=[PD_RNDINT_MAGIC] + pcmpeqd xmm3,xmm3 + psrld xmm3,WORD_BIT ; xmm3={0xFFFF 0x0000 0xFFFF 0x0000 ..} + + addps xmm6,xmm1 ; xmm6=roundint(data0/8)=(00 ** 10 ** 20 ** 30 **) + addps xmm7,xmm1 ; xmm7=roundint(data1/8)=(01 ** 11 ** 21 ** 31 **) + addps xmm0,xmm1 ; xmm0=roundint(data6/8)=(06 ** 16 ** 26 ** 36 **) + addps xmm5,xmm1 ; xmm5=roundint(data7/8)=(07 ** 17 ** 27 ** 37 **) + + pand xmm6,xmm3 ; xmm6=(00 -- 10 -- 20 -- 30 --) + pslld xmm7,WORD_BIT ; xmm7=(-- 01 -- 11 -- 21 -- 31) + pand xmm0,xmm3 ; xmm0=(06 -- 16 -- 26 -- 36 --) + pslld xmm5,WORD_BIT ; xmm5=(-- 07 -- 17 -- 27 -- 37) + por xmm6,xmm7 ; xmm6=(00 01 10 11 20 21 30 31) + por xmm0,xmm5 ; xmm0=(06 07 16 17 26 27 36 37) + + movaps xmm1, XMMWORD [wk(0)] ; xmm1=tmp2 + movaps xmm3, XMMWORD [wk(1)] ; xmm3=tmp3 + + addps xmm4,xmm2 ; xmm4=tmp4 + movaps xmm7,xmm1 + movaps xmm5,xmm3 + addps xmm1,xmm2 ; xmm1=data2=(02 12 22 32) + addps xmm3,xmm4 ; xmm3=data4=(04 14 24 34) + subps xmm7,xmm2 ; xmm7=data5=(05 15 25 35) + subps xmm5,xmm4 ; xmm5=data3=(03 13 23 33) + + movaps xmm2,[GOTOFF(ebx,PD_RNDINT_MAGIC)] ; xmm2=[PD_RNDINT_MAGIC] + pcmpeqd xmm4,xmm4 + psrld xmm4,WORD_BIT ; xmm4={0xFFFF 0x0000 0xFFFF 0x0000 ..} + + addps xmm3,xmm2 ; xmm3=roundint(data4/8)=(04 ** 14 ** 24 ** 34 **) + addps xmm7,xmm2 ; xmm7=roundint(data5/8)=(05 ** 15 ** 25 ** 35 **) + addps xmm1,xmm2 ; xmm1=roundint(data2/8)=(02 ** 12 ** 22 ** 32 **) + addps xmm5,xmm2 ; xmm5=roundint(data3/8)=(03 ** 13 ** 23 ** 33 **) + + pand xmm3,xmm4 ; xmm3=(04 -- 14 -- 24 -- 34 --) + pslld xmm7,WORD_BIT ; xmm7=(-- 05 -- 15 -- 25 -- 35) + pand xmm1,xmm4 ; xmm1=(02 -- 12 -- 22 -- 32 --) + pslld xmm5,WORD_BIT ; xmm5=(-- 03 -- 13 -- 23 -- 33) + por xmm3,xmm7 ; xmm3=(04 05 14 15 24 25 34 35) + por xmm1,xmm5 ; xmm1=(02 03 12 13 22 23 32 33) + + movdqa xmm2,[GOTOFF(ebx,PB_CENTERJSAMP)] ; xmm2=[PB_CENTERJSAMP] + + packsswb xmm6,xmm3 ; xmm6=(00 01 10 11 20 21 30 31 04 05 14 15 24 25 34 35) + packsswb xmm1,xmm0 ; xmm1=(02 03 12 13 22 23 32 33 06 07 16 17 26 27 36 37) + paddb xmm6,xmm2 + paddb xmm1,xmm2 + + movdqa xmm4,xmm6 ; transpose coefficients(phase 2) + punpcklwd xmm6,xmm1 ; xmm6=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33) + punpckhwd xmm4,xmm1 ; xmm4=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37) + + movdqa xmm7,xmm6 ; transpose coefficients(phase 3) + punpckldq xmm6,xmm4 ; xmm6=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17) + punpckhdq xmm7,xmm4 ; xmm7=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37) + + pshufd xmm5,xmm6,0x4E ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07) + pshufd xmm3,xmm7,0x4E ; xmm3=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27) + + pushpic ebx ; save GOT address + + mov edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] + mov ebx, JSAMPROW [edi+2*SIZEOF_JSAMPROW] + movq XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm6 + movq XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE], xmm7 + mov edx, JSAMPROW [edi+1*SIZEOF_JSAMPROW] + mov ebx, JSAMPROW [edi+3*SIZEOF_JSAMPROW] + movq XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm5 + movq XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE], xmm3 + + poppic ebx ; restore GOT address + + add esi, byte 4*SIZEOF_FAST_FLOAT ; wsptr + add edi, byte 4*SIZEOF_JSAMPROW + dec ecx ; ctr + jnz near .rowloop + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jidctfst-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jidctfst-mmx.asm new file mode 100644 index 000000000..d97c02a24 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jidctfst-mmx.asm @@ -0,0 +1,500 @@ +; +; jidctfst.asm - fast integer IDCT (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a fast, not so accurate integer implementation of +; the inverse DCT (Discrete Cosine Transform). The following code is +; based directly on the IJG's original jidctfst.c; see the jidctfst.c +; for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 8 ; 14 is also OK. +%define PASS1_BITS 2 + +%if IFAST_SCALE_BITS != PASS1_BITS +%error "'IFAST_SCALE_BITS' must be equal to 'PASS1_BITS'." +%endif + +%if CONST_BITS == 8 +F_1_082 equ 277 ; FIX(1.082392200) +F_1_414 equ 362 ; FIX(1.414213562) +F_1_847 equ 473 ; FIX(1.847759065) +F_2_613 equ 669 ; FIX(2.613125930) +F_1_613 equ (F_2_613 - 256) ; FIX(2.613125930) - FIX(1) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_1_082 equ DESCALE(1162209775,30-CONST_BITS) ; FIX(1.082392200) +F_1_414 equ DESCALE(1518500249,30-CONST_BITS) ; FIX(1.414213562) +F_1_847 equ DESCALE(1984016188,30-CONST_BITS) ; FIX(1.847759065) +F_2_613 equ DESCALE(2805822602,30-CONST_BITS) ; FIX(2.613125930) +F_1_613 equ (F_2_613 - (1 << CONST_BITS)) ; FIX(2.613125930) - FIX(1) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + +; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow) +; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw) + +%define PRE_MULTIPLY_SCALE_BITS 2 +%define CONST_SHIFT (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS) + + alignz 16 + global EXTN(jconst_idct_ifast_mmx) + +EXTN(jconst_idct_ifast_mmx): + +PW_F1414 times 4 dw F_1_414 << CONST_SHIFT +PW_F1847 times 4 dw F_1_847 << CONST_SHIFT +PW_MF1613 times 4 dw -F_1_613 << CONST_SHIFT +PW_F1082 times 4 dw F_1_082 << CONST_SHIFT +PB_CENTERJSAMP times 8 db CENTERJSAMPLE + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform dequantization and inverse DCT on one block of coefficients. +; +; GLOBAL(void) +; jsimd_idct_ifast_mmx (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +%define dct_table(b) (b)+8 ; jpeg_component_info * compptr +%define coef_block(b) (b)+12 ; JCOEFPTR coef_block +%define output_buf(b) (b)+16 ; JSAMPARRAY output_buf +%define output_col(b) (b)+20 ; JDIMENSION output_col + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_MMWORD ; mmword wk[WK_NUM] +%define WK_NUM 2 +%define workspace wk(0)-DCTSIZE2*SIZEOF_JCOEF + ; JCOEF workspace[DCTSIZE2] + + align 16 + global EXTN(jsimd_idct_ifast_mmx) + +EXTN(jsimd_idct_ifast_mmx): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_MMWORD) ; align to 64 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [workspace] + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process columns from input, store into work array. + +; mov eax, [original_ebp] + mov edx, POINTER [dct_table(eax)] ; quantptr + mov esi, JCOEFPTR [coef_block(eax)] ; inptr + lea edi, [workspace] ; JCOEF * wsptr + mov ecx, DCTSIZE/4 ; ctr + alignx 16,7 +.columnloop: +%ifndef NO_ZERO_COLUMN_TEST_IFAST_MMX + mov eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)] + jnz short .columnDCT + + movq mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)] + por mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + por mm1, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)] + por mm0, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + por mm1, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)] + por mm0, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + por mm1,mm0 + packsswb mm1,mm1 + movd eax,mm1 + test eax,eax + jnz short .columnDCT + + ; -- AC terms all zero + + movq mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + pmullw mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_IFAST_MULT_TYPE)] + + movq mm2,mm0 ; mm0=in0=(00 01 02 03) + punpcklwd mm0,mm0 ; mm0=(00 00 01 01) + punpckhwd mm2,mm2 ; mm2=(02 02 03 03) + + movq mm1,mm0 + punpckldq mm0,mm0 ; mm0=(00 00 00 00) + punpckhdq mm1,mm1 ; mm1=(01 01 01 01) + movq mm3,mm2 + punpckldq mm2,mm2 ; mm2=(02 02 02 02) + punpckhdq mm3,mm3 ; mm3=(03 03 03 03) + + movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0 + movq MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm0 + movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm1 + movq MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm1 + movq MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm2 + movq MMWORD [MMBLOCK(2,1,edi,SIZEOF_JCOEF)], mm2 + movq MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm3 + movq MMWORD [MMBLOCK(3,1,edi,SIZEOF_JCOEF)], mm3 + jmp near .nextcolumn + alignx 16,7 +%endif +.columnDCT: + + ; -- Even part + + movq mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)] + pmullw mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_IFAST_MULT_TYPE)] + pmullw mm1, MMWORD [MMBLOCK(2,0,edx,SIZEOF_IFAST_MULT_TYPE)] + movq mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)] + movq mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)] + pmullw mm2, MMWORD [MMBLOCK(4,0,edx,SIZEOF_IFAST_MULT_TYPE)] + pmullw mm3, MMWORD [MMBLOCK(6,0,edx,SIZEOF_IFAST_MULT_TYPE)] + + movq mm4,mm0 + movq mm5,mm1 + psubw mm0,mm2 ; mm0=tmp11 + psubw mm1,mm3 + paddw mm4,mm2 ; mm4=tmp10 + paddw mm5,mm3 ; mm5=tmp13 + + psllw mm1,PRE_MULTIPLY_SCALE_BITS + pmulhw mm1,[GOTOFF(ebx,PW_F1414)] + psubw mm1,mm5 ; mm1=tmp12 + + movq mm6,mm4 + movq mm7,mm0 + psubw mm4,mm5 ; mm4=tmp3 + psubw mm0,mm1 ; mm0=tmp2 + paddw mm6,mm5 ; mm6=tmp0 + paddw mm7,mm1 ; mm7=tmp1 + + movq MMWORD [wk(1)], mm4 ; wk(1)=tmp3 + movq MMWORD [wk(0)], mm0 ; wk(0)=tmp2 + + ; -- Odd part + + movq mm2, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq mm3, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + pmullw mm2, MMWORD [MMBLOCK(1,0,edx,SIZEOF_IFAST_MULT_TYPE)] + pmullw mm3, MMWORD [MMBLOCK(3,0,edx,SIZEOF_IFAST_MULT_TYPE)] + movq mm5, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + pmullw mm5, MMWORD [MMBLOCK(5,0,edx,SIZEOF_IFAST_MULT_TYPE)] + pmullw mm1, MMWORD [MMBLOCK(7,0,edx,SIZEOF_IFAST_MULT_TYPE)] + + movq mm4,mm2 + movq mm0,mm5 + psubw mm2,mm1 ; mm2=z12 + psubw mm5,mm3 ; mm5=z10 + paddw mm4,mm1 ; mm4=z11 + paddw mm0,mm3 ; mm0=z13 + + movq mm1,mm5 ; mm1=z10(unscaled) + psllw mm2,PRE_MULTIPLY_SCALE_BITS + psllw mm5,PRE_MULTIPLY_SCALE_BITS + + movq mm3,mm4 + psubw mm4,mm0 + paddw mm3,mm0 ; mm3=tmp7 + + psllw mm4,PRE_MULTIPLY_SCALE_BITS + pmulhw mm4,[GOTOFF(ebx,PW_F1414)] ; mm4=tmp11 + + ; To avoid overflow... + ; + ; (Original) + ; tmp12 = -2.613125930 * z10 + z5; + ; + ; (This implementation) + ; tmp12 = (-1.613125930 - 1) * z10 + z5; + ; = -1.613125930 * z10 - z10 + z5; + + movq mm0,mm5 + paddw mm5,mm2 + pmulhw mm5,[GOTOFF(ebx,PW_F1847)] ; mm5=z5 + pmulhw mm0,[GOTOFF(ebx,PW_MF1613)] + pmulhw mm2,[GOTOFF(ebx,PW_F1082)] + psubw mm0,mm1 + psubw mm2,mm5 ; mm2=tmp10 + paddw mm0,mm5 ; mm0=tmp12 + + ; -- Final output stage + + psubw mm0,mm3 ; mm0=tmp6 + movq mm1,mm6 + movq mm5,mm7 + paddw mm6,mm3 ; mm6=data0=(00 01 02 03) + paddw mm7,mm0 ; mm7=data1=(10 11 12 13) + psubw mm1,mm3 ; mm1=data7=(70 71 72 73) + psubw mm5,mm0 ; mm5=data6=(60 61 62 63) + psubw mm4,mm0 ; mm4=tmp5 + + movq mm3,mm6 ; transpose coefficients(phase 1) + punpcklwd mm6,mm7 ; mm6=(00 10 01 11) + punpckhwd mm3,mm7 ; mm3=(02 12 03 13) + movq mm0,mm5 ; transpose coefficients(phase 1) + punpcklwd mm5,mm1 ; mm5=(60 70 61 71) + punpckhwd mm0,mm1 ; mm0=(62 72 63 73) + + movq mm7, MMWORD [wk(0)] ; mm7=tmp2 + movq mm1, MMWORD [wk(1)] ; mm1=tmp3 + + movq MMWORD [wk(0)], mm5 ; wk(0)=(60 70 61 71) + movq MMWORD [wk(1)], mm0 ; wk(1)=(62 72 63 73) + + paddw mm2,mm4 ; mm2=tmp4 + movq mm5,mm7 + movq mm0,mm1 + paddw mm7,mm4 ; mm7=data2=(20 21 22 23) + paddw mm1,mm2 ; mm1=data4=(40 41 42 43) + psubw mm5,mm4 ; mm5=data5=(50 51 52 53) + psubw mm0,mm2 ; mm0=data3=(30 31 32 33) + + movq mm4,mm7 ; transpose coefficients(phase 1) + punpcklwd mm7,mm0 ; mm7=(20 30 21 31) + punpckhwd mm4,mm0 ; mm4=(22 32 23 33) + movq mm2,mm1 ; transpose coefficients(phase 1) + punpcklwd mm1,mm5 ; mm1=(40 50 41 51) + punpckhwd mm2,mm5 ; mm2=(42 52 43 53) + + movq mm0,mm6 ; transpose coefficients(phase 2) + punpckldq mm6,mm7 ; mm6=(00 10 20 30) + punpckhdq mm0,mm7 ; mm0=(01 11 21 31) + movq mm5,mm3 ; transpose coefficients(phase 2) + punpckldq mm3,mm4 ; mm3=(02 12 22 32) + punpckhdq mm5,mm4 ; mm5=(03 13 23 33) + + movq mm7, MMWORD [wk(0)] ; mm7=(60 70 61 71) + movq mm4, MMWORD [wk(1)] ; mm4=(62 72 63 73) + + movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm6 + movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm0 + movq MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm3 + movq MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm5 + + movq mm6,mm1 ; transpose coefficients(phase 2) + punpckldq mm1,mm7 ; mm1=(40 50 60 70) + punpckhdq mm6,mm7 ; mm6=(41 51 61 71) + movq mm0,mm2 ; transpose coefficients(phase 2) + punpckldq mm2,mm4 ; mm2=(42 52 62 72) + punpckhdq mm0,mm4 ; mm0=(43 53 63 73) + + movq MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm1 + movq MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm6 + movq MMWORD [MMBLOCK(2,1,edi,SIZEOF_JCOEF)], mm2 + movq MMWORD [MMBLOCK(3,1,edi,SIZEOF_JCOEF)], mm0 + +.nextcolumn: + add esi, byte 4*SIZEOF_JCOEF ; coef_block + add edx, byte 4*SIZEOF_IFAST_MULT_TYPE ; quantptr + add edi, byte 4*DCTSIZE*SIZEOF_JCOEF ; wsptr + dec ecx ; ctr + jnz near .columnloop + + ; ---- Pass 2: process rows from work array, store into output array. + + mov eax, [original_ebp] + lea esi, [workspace] ; JCOEF * wsptr + mov edi, JSAMPARRAY [output_buf(eax)] ; (JSAMPROW *) + mov eax, JDIMENSION [output_col(eax)] + mov ecx, DCTSIZE/4 ; ctr + alignx 16,7 +.rowloop: + + ; -- Even part + + movq mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)] + movq mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)] + movq mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)] + + movq mm4,mm0 + movq mm5,mm1 + psubw mm0,mm2 ; mm0=tmp11 + psubw mm1,mm3 + paddw mm4,mm2 ; mm4=tmp10 + paddw mm5,mm3 ; mm5=tmp13 + + psllw mm1,PRE_MULTIPLY_SCALE_BITS + pmulhw mm1,[GOTOFF(ebx,PW_F1414)] + psubw mm1,mm5 ; mm1=tmp12 + + movq mm6,mm4 + movq mm7,mm0 + psubw mm4,mm5 ; mm4=tmp3 + psubw mm0,mm1 ; mm0=tmp2 + paddw mm6,mm5 ; mm6=tmp0 + paddw mm7,mm1 ; mm7=tmp1 + + movq MMWORD [wk(1)], mm4 ; wk(1)=tmp3 + movq MMWORD [wk(0)], mm0 ; wk(0)=tmp2 + + ; -- Odd part + + movq mm2, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq mm3, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + movq mm5, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + + movq mm4,mm2 + movq mm0,mm5 + psubw mm2,mm1 ; mm2=z12 + psubw mm5,mm3 ; mm5=z10 + paddw mm4,mm1 ; mm4=z11 + paddw mm0,mm3 ; mm0=z13 + + movq mm1,mm5 ; mm1=z10(unscaled) + psllw mm2,PRE_MULTIPLY_SCALE_BITS + psllw mm5,PRE_MULTIPLY_SCALE_BITS + + movq mm3,mm4 + psubw mm4,mm0 + paddw mm3,mm0 ; mm3=tmp7 + + psllw mm4,PRE_MULTIPLY_SCALE_BITS + pmulhw mm4,[GOTOFF(ebx,PW_F1414)] ; mm4=tmp11 + + ; To avoid overflow... + ; + ; (Original) + ; tmp12 = -2.613125930 * z10 + z5; + ; + ; (This implementation) + ; tmp12 = (-1.613125930 - 1) * z10 + z5; + ; = -1.613125930 * z10 - z10 + z5; + + movq mm0,mm5 + paddw mm5,mm2 + pmulhw mm5,[GOTOFF(ebx,PW_F1847)] ; mm5=z5 + pmulhw mm0,[GOTOFF(ebx,PW_MF1613)] + pmulhw mm2,[GOTOFF(ebx,PW_F1082)] + psubw mm0,mm1 + psubw mm2,mm5 ; mm2=tmp10 + paddw mm0,mm5 ; mm0=tmp12 + + ; -- Final output stage + + psubw mm0,mm3 ; mm0=tmp6 + movq mm1,mm6 + movq mm5,mm7 + paddw mm6,mm3 ; mm6=data0=(00 10 20 30) + paddw mm7,mm0 ; mm7=data1=(01 11 21 31) + psraw mm6,(PASS1_BITS+3) ; descale + psraw mm7,(PASS1_BITS+3) ; descale + psubw mm1,mm3 ; mm1=data7=(07 17 27 37) + psubw mm5,mm0 ; mm5=data6=(06 16 26 36) + psraw mm1,(PASS1_BITS+3) ; descale + psraw mm5,(PASS1_BITS+3) ; descale + psubw mm4,mm0 ; mm4=tmp5 + + packsswb mm6,mm5 ; mm6=(00 10 20 30 06 16 26 36) + packsswb mm7,mm1 ; mm7=(01 11 21 31 07 17 27 37) + + movq mm3, MMWORD [wk(0)] ; mm3=tmp2 + movq mm0, MMWORD [wk(1)] ; mm0=tmp3 + + paddw mm2,mm4 ; mm2=tmp4 + movq mm5,mm3 + movq mm1,mm0 + paddw mm3,mm4 ; mm3=data2=(02 12 22 32) + paddw mm0,mm2 ; mm0=data4=(04 14 24 34) + psraw mm3,(PASS1_BITS+3) ; descale + psraw mm0,(PASS1_BITS+3) ; descale + psubw mm5,mm4 ; mm5=data5=(05 15 25 35) + psubw mm1,mm2 ; mm1=data3=(03 13 23 33) + psraw mm5,(PASS1_BITS+3) ; descale + psraw mm1,(PASS1_BITS+3) ; descale + + movq mm4,[GOTOFF(ebx,PB_CENTERJSAMP)] ; mm4=[PB_CENTERJSAMP] + + packsswb mm3,mm0 ; mm3=(02 12 22 32 04 14 24 34) + packsswb mm1,mm5 ; mm1=(03 13 23 33 05 15 25 35) + + paddb mm6,mm4 + paddb mm7,mm4 + paddb mm3,mm4 + paddb mm1,mm4 + + movq mm2,mm6 ; transpose coefficients(phase 1) + punpcklbw mm6,mm7 ; mm6=(00 01 10 11 20 21 30 31) + punpckhbw mm2,mm7 ; mm2=(06 07 16 17 26 27 36 37) + movq mm0,mm3 ; transpose coefficients(phase 1) + punpcklbw mm3,mm1 ; mm3=(02 03 12 13 22 23 32 33) + punpckhbw mm0,mm1 ; mm0=(04 05 14 15 24 25 34 35) + + movq mm5,mm6 ; transpose coefficients(phase 2) + punpcklwd mm6,mm3 ; mm6=(00 01 02 03 10 11 12 13) + punpckhwd mm5,mm3 ; mm5=(20 21 22 23 30 31 32 33) + movq mm4,mm0 ; transpose coefficients(phase 2) + punpcklwd mm0,mm2 ; mm0=(04 05 06 07 14 15 16 17) + punpckhwd mm4,mm2 ; mm4=(24 25 26 27 34 35 36 37) + + movq mm7,mm6 ; transpose coefficients(phase 3) + punpckldq mm6,mm0 ; mm6=(00 01 02 03 04 05 06 07) + punpckhdq mm7,mm0 ; mm7=(10 11 12 13 14 15 16 17) + movq mm1,mm5 ; transpose coefficients(phase 3) + punpckldq mm5,mm4 ; mm5=(20 21 22 23 24 25 26 27) + punpckhdq mm1,mm4 ; mm1=(30 31 32 33 34 35 36 37) + + pushpic ebx ; save GOT address + + mov edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] + mov ebx, JSAMPROW [edi+1*SIZEOF_JSAMPROW] + movq MMWORD [edx+eax*SIZEOF_JSAMPLE], mm6 + movq MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm7 + mov edx, JSAMPROW [edi+2*SIZEOF_JSAMPROW] + mov ebx, JSAMPROW [edi+3*SIZEOF_JSAMPROW] + movq MMWORD [edx+eax*SIZEOF_JSAMPLE], mm5 + movq MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm1 + + poppic ebx ; restore GOT address + + add esi, byte 4*SIZEOF_JCOEF ; wsptr + add edi, byte 4*SIZEOF_JSAMPROW + dec ecx ; ctr + jnz near .rowloop + + emms ; empty MMX state + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jidctfst-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jidctfst-sse2-64.asm new file mode 100644 index 000000000..85669527c --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jidctfst-sse2-64.asm @@ -0,0 +1,492 @@ +; +; jidctfst.asm - fast integer IDCT (64-bit SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/projecpt/showfiles.php?group_id=6208 +; +; This file contains a fast, not so accurate integer implementation of +; the inverse DCT (Discrete Cosine Transform). The following code is +; based directly on the IJG's original jidctfst.c; see the jidctfst.c +; for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 8 ; 14 is also OK. +%define PASS1_BITS 2 + +%if IFAST_SCALE_BITS != PASS1_BITS +%error "'IFAST_SCALE_BITS' must be equal to 'PASS1_BITS'." +%endif + +%if CONST_BITS == 8 +F_1_082 equ 277 ; FIX(1.082392200) +F_1_414 equ 362 ; FIX(1.414213562) +F_1_847 equ 473 ; FIX(1.847759065) +F_2_613 equ 669 ; FIX(2.613125930) +F_1_613 equ (F_2_613 - 256) ; FIX(2.613125930) - FIX(1) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_1_082 equ DESCALE(1162209775,30-CONST_BITS) ; FIX(1.082392200) +F_1_414 equ DESCALE(1518500249,30-CONST_BITS) ; FIX(1.414213562) +F_1_847 equ DESCALE(1984016188,30-CONST_BITS) ; FIX(1.847759065) +F_2_613 equ DESCALE(2805822602,30-CONST_BITS) ; FIX(2.613125930) +F_1_613 equ (F_2_613 - (1 << CONST_BITS)) ; FIX(2.613125930) - FIX(1) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + +; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow) +; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw) + +%define PRE_MULTIPLY_SCALE_BITS 2 +%define CONST_SHIFT (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS) + + alignz 16 + global EXTN(jconst_idct_ifast_sse2) + +EXTN(jconst_idct_ifast_sse2): + +PW_F1414 times 8 dw F_1_414 << CONST_SHIFT +PW_F1847 times 8 dw F_1_847 << CONST_SHIFT +PW_MF1613 times 8 dw -F_1_613 << CONST_SHIFT +PW_F1082 times 8 dw F_1_082 << CONST_SHIFT +PB_CENTERJSAMP times 16 db CENTERJSAMPLE + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 +; +; Perform dequantization and inverse DCT on one block of coefficients. +; +; GLOBAL(void) +; jsimd_idct_ifast_sse2 (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +; r10 = jpeg_component_info * compptr +; r11 = JCOEFPTR coef_block +; r12 = JSAMPARRAY output_buf +; r13 = JDIMENSION output_col + +%define original_rbp rbp+0 +%define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 + + align 16 + global EXTN(jsimd_idct_ifast_sse2) + +EXTN(jsimd_idct_ifast_sse2): + push rbp + mov rax,rsp ; rax = original rbp + sub rsp, byte 4 + and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [rsp],rax + mov rbp,rsp ; rbp = aligned rbp + lea rsp, [wk(0)] + collect_args + + ; ---- Pass 1: process columns from input. + + mov rdx, r10 ; quantptr + mov rsi, r11 ; inptr + +%ifndef NO_ZERO_COLUMN_TEST_IFAST_SSE2 + mov eax, DWORD [DWBLOCK(1,0,rsi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(2,0,rsi,SIZEOF_JCOEF)] + jnz near .columnDCT + + movdqa xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)] + por xmm1, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)] + por xmm1, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)] + por xmm1,xmm0 + packsswb xmm1,xmm1 + packsswb xmm1,xmm1 + movd eax,xmm1 + test rax,rax + jnz short .columnDCT + + ; -- AC terms all zero + + movdqa xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)] + pmullw xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + + movdqa xmm7,xmm0 ; xmm0=in0=(00 01 02 03 04 05 06 07) + punpcklwd xmm0,xmm0 ; xmm0=(00 00 01 01 02 02 03 03) + punpckhwd xmm7,xmm7 ; xmm7=(04 04 05 05 06 06 07 07) + + pshufd xmm6,xmm0,0x00 ; xmm6=col0=(00 00 00 00 00 00 00 00) + pshufd xmm2,xmm0,0x55 ; xmm2=col1=(01 01 01 01 01 01 01 01) + pshufd xmm5,xmm0,0xAA ; xmm5=col2=(02 02 02 02 02 02 02 02) + pshufd xmm0,xmm0,0xFF ; xmm0=col3=(03 03 03 03 03 03 03 03) + pshufd xmm1,xmm7,0x00 ; xmm1=col4=(04 04 04 04 04 04 04 04) + pshufd xmm4,xmm7,0x55 ; xmm4=col5=(05 05 05 05 05 05 05 05) + pshufd xmm3,xmm7,0xAA ; xmm3=col6=(06 06 06 06 06 06 06 06) + pshufd xmm7,xmm7,0xFF ; xmm7=col7=(07 07 07 07 07 07 07 07) + + movdqa XMMWORD [wk(0)], xmm2 ; wk(0)=col1 + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=col3 + jmp near .column_end +%endif +.columnDCT: + + ; -- Even part + + movdqa xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)] + pmullw xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_IFAST_MULT_TYPE)] + pmullw xmm1, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_IFAST_MULT_TYPE)] + movdqa xmm2, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)] + movdqa xmm3, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)] + pmullw xmm2, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_IFAST_MULT_TYPE)] + pmullw xmm3, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_IFAST_MULT_TYPE)] + + movdqa xmm4,xmm0 + movdqa xmm5,xmm1 + psubw xmm0,xmm2 ; xmm0=tmp11 + psubw xmm1,xmm3 + paddw xmm4,xmm2 ; xmm4=tmp10 + paddw xmm5,xmm3 ; xmm5=tmp13 + + psllw xmm1,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm1,[rel PW_F1414] + psubw xmm1,xmm5 ; xmm1=tmp12 + + movdqa xmm6,xmm4 + movdqa xmm7,xmm0 + psubw xmm4,xmm5 ; xmm4=tmp3 + psubw xmm0,xmm1 ; xmm0=tmp2 + paddw xmm6,xmm5 ; xmm6=tmp0 + paddw xmm7,xmm1 ; xmm7=tmp1 + + movdqa XMMWORD [wk(1)], xmm4 ; wk(1)=tmp3 + movdqa XMMWORD [wk(0)], xmm0 ; wk(0)=tmp2 + + ; -- Odd part + + movdqa xmm2, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)] + movdqa xmm3, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)] + pmullw xmm2, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_IFAST_MULT_TYPE)] + pmullw xmm3, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_IFAST_MULT_TYPE)] + movdqa xmm5, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)] + pmullw xmm5, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_IFAST_MULT_TYPE)] + pmullw xmm1, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_IFAST_MULT_TYPE)] + + movdqa xmm4,xmm2 + movdqa xmm0,xmm5 + psubw xmm2,xmm1 ; xmm2=z12 + psubw xmm5,xmm3 ; xmm5=z10 + paddw xmm4,xmm1 ; xmm4=z11 + paddw xmm0,xmm3 ; xmm0=z13 + + movdqa xmm1,xmm5 ; xmm1=z10(unscaled) + psllw xmm2,PRE_MULTIPLY_SCALE_BITS + psllw xmm5,PRE_MULTIPLY_SCALE_BITS + + movdqa xmm3,xmm4 + psubw xmm4,xmm0 + paddw xmm3,xmm0 ; xmm3=tmp7 + + psllw xmm4,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm4,[rel PW_F1414] ; xmm4=tmp11 + + ; To avoid overflow... + ; + ; (Original) + ; tmp12 = -2.613125930 * z10 + z5; + ; + ; (This implementation) + ; tmp12 = (-1.613125930 - 1) * z10 + z5; + ; = -1.613125930 * z10 - z10 + z5; + + movdqa xmm0,xmm5 + paddw xmm5,xmm2 + pmulhw xmm5,[rel PW_F1847] ; xmm5=z5 + pmulhw xmm0,[rel PW_MF1613] + pmulhw xmm2,[rel PW_F1082] + psubw xmm0,xmm1 + psubw xmm2,xmm5 ; xmm2=tmp10 + paddw xmm0,xmm5 ; xmm0=tmp12 + + ; -- Final output stage + + psubw xmm0,xmm3 ; xmm0=tmp6 + movdqa xmm1,xmm6 + movdqa xmm5,xmm7 + paddw xmm6,xmm3 ; xmm6=data0=(00 01 02 03 04 05 06 07) + paddw xmm7,xmm0 ; xmm7=data1=(10 11 12 13 14 15 16 17) + psubw xmm1,xmm3 ; xmm1=data7=(70 71 72 73 74 75 76 77) + psubw xmm5,xmm0 ; xmm5=data6=(60 61 62 63 64 65 66 67) + psubw xmm4,xmm0 ; xmm4=tmp5 + + movdqa xmm3,xmm6 ; transpose coefficients(phase 1) + punpcklwd xmm6,xmm7 ; xmm6=(00 10 01 11 02 12 03 13) + punpckhwd xmm3,xmm7 ; xmm3=(04 14 05 15 06 16 07 17) + movdqa xmm0,xmm5 ; transpose coefficients(phase 1) + punpcklwd xmm5,xmm1 ; xmm5=(60 70 61 71 62 72 63 73) + punpckhwd xmm0,xmm1 ; xmm0=(64 74 65 75 66 76 67 77) + + movdqa xmm7, XMMWORD [wk(0)] ; xmm7=tmp2 + movdqa xmm1, XMMWORD [wk(1)] ; xmm1=tmp3 + + movdqa XMMWORD [wk(0)], xmm5 ; wk(0)=(60 70 61 71 62 72 63 73) + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=(64 74 65 75 66 76 67 77) + + paddw xmm2,xmm4 ; xmm2=tmp4 + movdqa xmm5,xmm7 + movdqa xmm0,xmm1 + paddw xmm7,xmm4 ; xmm7=data2=(20 21 22 23 24 25 26 27) + paddw xmm1,xmm2 ; xmm1=data4=(40 41 42 43 44 45 46 47) + psubw xmm5,xmm4 ; xmm5=data5=(50 51 52 53 54 55 56 57) + psubw xmm0,xmm2 ; xmm0=data3=(30 31 32 33 34 35 36 37) + + movdqa xmm4,xmm7 ; transpose coefficients(phase 1) + punpcklwd xmm7,xmm0 ; xmm7=(20 30 21 31 22 32 23 33) + punpckhwd xmm4,xmm0 ; xmm4=(24 34 25 35 26 36 27 37) + movdqa xmm2,xmm1 ; transpose coefficients(phase 1) + punpcklwd xmm1,xmm5 ; xmm1=(40 50 41 51 42 52 43 53) + punpckhwd xmm2,xmm5 ; xmm2=(44 54 45 55 46 56 47 57) + + movdqa xmm0,xmm3 ; transpose coefficients(phase 2) + punpckldq xmm3,xmm4 ; xmm3=(04 14 24 34 05 15 25 35) + punpckhdq xmm0,xmm4 ; xmm0=(06 16 26 36 07 17 27 37) + movdqa xmm5,xmm6 ; transpose coefficients(phase 2) + punpckldq xmm6,xmm7 ; xmm6=(00 10 20 30 01 11 21 31) + punpckhdq xmm5,xmm7 ; xmm5=(02 12 22 32 03 13 23 33) + + movdqa xmm4, XMMWORD [wk(0)] ; xmm4=(60 70 61 71 62 72 63 73) + movdqa xmm7, XMMWORD [wk(1)] ; xmm7=(64 74 65 75 66 76 67 77) + + movdqa XMMWORD [wk(0)], xmm3 ; wk(0)=(04 14 24 34 05 15 25 35) + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=(06 16 26 36 07 17 27 37) + + movdqa xmm3,xmm1 ; transpose coefficients(phase 2) + punpckldq xmm1,xmm4 ; xmm1=(40 50 60 70 41 51 61 71) + punpckhdq xmm3,xmm4 ; xmm3=(42 52 62 72 43 53 63 73) + movdqa xmm0,xmm2 ; transpose coefficients(phase 2) + punpckldq xmm2,xmm7 ; xmm2=(44 54 64 74 45 55 65 75) + punpckhdq xmm0,xmm7 ; xmm0=(46 56 66 76 47 57 67 77) + + movdqa xmm4,xmm6 ; transpose coefficients(phase 3) + punpcklqdq xmm6,xmm1 ; xmm6=col0=(00 10 20 30 40 50 60 70) + punpckhqdq xmm4,xmm1 ; xmm4=col1=(01 11 21 31 41 51 61 71) + movdqa xmm7,xmm5 ; transpose coefficients(phase 3) + punpcklqdq xmm5,xmm3 ; xmm5=col2=(02 12 22 32 42 52 62 72) + punpckhqdq xmm7,xmm3 ; xmm7=col3=(03 13 23 33 43 53 63 73) + + movdqa xmm1, XMMWORD [wk(0)] ; xmm1=(04 14 24 34 05 15 25 35) + movdqa xmm3, XMMWORD [wk(1)] ; xmm3=(06 16 26 36 07 17 27 37) + + movdqa XMMWORD [wk(0)], xmm4 ; wk(0)=col1 + movdqa XMMWORD [wk(1)], xmm7 ; wk(1)=col3 + + movdqa xmm4,xmm1 ; transpose coefficients(phase 3) + punpcklqdq xmm1,xmm2 ; xmm1=col4=(04 14 24 34 44 54 64 74) + punpckhqdq xmm4,xmm2 ; xmm4=col5=(05 15 25 35 45 55 65 75) + movdqa xmm7,xmm3 ; transpose coefficients(phase 3) + punpcklqdq xmm3,xmm0 ; xmm3=col6=(06 16 26 36 46 56 66 76) + punpckhqdq xmm7,xmm0 ; xmm7=col7=(07 17 27 37 47 57 67 77) +.column_end: + + ; -- Prefetch the next coefficient block + + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32] + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32] + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32] + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32] + + ; ---- Pass 2: process rows from work array, store into output array. + + mov rax, [original_rbp] + mov rdi, r12 ; (JSAMPROW *) + mov rax, r13 + + ; -- Even part + + ; xmm6=col0, xmm5=col2, xmm1=col4, xmm3=col6 + + movdqa xmm2,xmm6 + movdqa xmm0,xmm5 + psubw xmm6,xmm1 ; xmm6=tmp11 + psubw xmm5,xmm3 + paddw xmm2,xmm1 ; xmm2=tmp10 + paddw xmm0,xmm3 ; xmm0=tmp13 + + psllw xmm5,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm5,[rel PW_F1414] + psubw xmm5,xmm0 ; xmm5=tmp12 + + movdqa xmm1,xmm2 + movdqa xmm3,xmm6 + psubw xmm2,xmm0 ; xmm2=tmp3 + psubw xmm6,xmm5 ; xmm6=tmp2 + paddw xmm1,xmm0 ; xmm1=tmp0 + paddw xmm3,xmm5 ; xmm3=tmp1 + + movdqa xmm0, XMMWORD [wk(0)] ; xmm0=col1 + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=col3 + + movdqa XMMWORD [wk(0)], xmm2 ; wk(0)=tmp3 + movdqa XMMWORD [wk(1)], xmm6 ; wk(1)=tmp2 + + ; -- Odd part + + ; xmm0=col1, xmm5=col3, xmm4=col5, xmm7=col7 + + movdqa xmm2,xmm0 + movdqa xmm6,xmm4 + psubw xmm0,xmm7 ; xmm0=z12 + psubw xmm4,xmm5 ; xmm4=z10 + paddw xmm2,xmm7 ; xmm2=z11 + paddw xmm6,xmm5 ; xmm6=z13 + + movdqa xmm7,xmm4 ; xmm7=z10(unscaled) + psllw xmm0,PRE_MULTIPLY_SCALE_BITS + psllw xmm4,PRE_MULTIPLY_SCALE_BITS + + movdqa xmm5,xmm2 + psubw xmm2,xmm6 + paddw xmm5,xmm6 ; xmm5=tmp7 + + psllw xmm2,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm2,[rel PW_F1414] ; xmm2=tmp11 + + ; To avoid overflow... + ; + ; (Original) + ; tmp12 = -2.613125930 * z10 + z5; + ; + ; (This implementation) + ; tmp12 = (-1.613125930 - 1) * z10 + z5; + ; = -1.613125930 * z10 - z10 + z5; + + movdqa xmm6,xmm4 + paddw xmm4,xmm0 + pmulhw xmm4,[rel PW_F1847] ; xmm4=z5 + pmulhw xmm6,[rel PW_MF1613] + pmulhw xmm0,[rel PW_F1082] + psubw xmm6,xmm7 + psubw xmm0,xmm4 ; xmm0=tmp10 + paddw xmm6,xmm4 ; xmm6=tmp12 + + ; -- Final output stage + + psubw xmm6,xmm5 ; xmm6=tmp6 + movdqa xmm7,xmm1 + movdqa xmm4,xmm3 + paddw xmm1,xmm5 ; xmm1=data0=(00 10 20 30 40 50 60 70) + paddw xmm3,xmm6 ; xmm3=data1=(01 11 21 31 41 51 61 71) + psraw xmm1,(PASS1_BITS+3) ; descale + psraw xmm3,(PASS1_BITS+3) ; descale + psubw xmm7,xmm5 ; xmm7=data7=(07 17 27 37 47 57 67 77) + psubw xmm4,xmm6 ; xmm4=data6=(06 16 26 36 46 56 66 76) + psraw xmm7,(PASS1_BITS+3) ; descale + psraw xmm4,(PASS1_BITS+3) ; descale + psubw xmm2,xmm6 ; xmm2=tmp5 + + packsswb xmm1,xmm4 ; xmm1=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76) + packsswb xmm3,xmm7 ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77) + + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=tmp2 + movdqa xmm6, XMMWORD [wk(0)] ; xmm6=tmp3 + + paddw xmm0,xmm2 ; xmm0=tmp4 + movdqa xmm4,xmm5 + movdqa xmm7,xmm6 + paddw xmm5,xmm2 ; xmm5=data2=(02 12 22 32 42 52 62 72) + paddw xmm6,xmm0 ; xmm6=data4=(04 14 24 34 44 54 64 74) + psraw xmm5,(PASS1_BITS+3) ; descale + psraw xmm6,(PASS1_BITS+3) ; descale + psubw xmm4,xmm2 ; xmm4=data5=(05 15 25 35 45 55 65 75) + psubw xmm7,xmm0 ; xmm7=data3=(03 13 23 33 43 53 63 73) + psraw xmm4,(PASS1_BITS+3) ; descale + psraw xmm7,(PASS1_BITS+3) ; descale + + movdqa xmm2,[rel PB_CENTERJSAMP] ; xmm2=[rel PB_CENTERJSAMP] + + packsswb xmm5,xmm6 ; xmm5=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74) + packsswb xmm7,xmm4 ; xmm7=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75) + + paddb xmm1,xmm2 + paddb xmm3,xmm2 + paddb xmm5,xmm2 + paddb xmm7,xmm2 + + movdqa xmm0,xmm1 ; transpose coefficients(phase 1) + punpcklbw xmm1,xmm3 ; xmm1=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71) + punpckhbw xmm0,xmm3 ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77) + movdqa xmm6,xmm5 ; transpose coefficients(phase 1) + punpcklbw xmm5,xmm7 ; xmm5=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73) + punpckhbw xmm6,xmm7 ; xmm6=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75) + + movdqa xmm4,xmm1 ; transpose coefficients(phase 2) + punpcklwd xmm1,xmm5 ; xmm1=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33) + punpckhwd xmm4,xmm5 ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73) + movdqa xmm2,xmm6 ; transpose coefficients(phase 2) + punpcklwd xmm6,xmm0 ; xmm6=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37) + punpckhwd xmm2,xmm0 ; xmm2=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77) + + movdqa xmm3,xmm1 ; transpose coefficients(phase 3) + punpckldq xmm1,xmm6 ; xmm1=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17) + punpckhdq xmm3,xmm6 ; xmm3=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37) + movdqa xmm7,xmm4 ; transpose coefficients(phase 3) + punpckldq xmm4,xmm2 ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57) + punpckhdq xmm7,xmm2 ; xmm7=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77) + + pshufd xmm5,xmm1,0x4E ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07) + pshufd xmm0,xmm3,0x4E ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27) + pshufd xmm6,xmm4,0x4E ; xmm6=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47) + pshufd xmm2,xmm7,0x4E ; xmm2=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67) + + mov rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW] + mov rsi, JSAMPROW [rdi+2*SIZEOF_JSAMPROW] + movq XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm1 + movq XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm3 + mov rdx, JSAMPROW [rdi+4*SIZEOF_JSAMPROW] + mov rsi, JSAMPROW [rdi+6*SIZEOF_JSAMPROW] + movq XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm4 + movq XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm7 + + mov rdx, JSAMPROW [rdi+1*SIZEOF_JSAMPROW] + mov rsi, JSAMPROW [rdi+3*SIZEOF_JSAMPROW] + movq XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm5 + movq XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm0 + mov rdx, JSAMPROW [rdi+5*SIZEOF_JSAMPROW] + mov rsi, JSAMPROW [rdi+7*SIZEOF_JSAMPROW] + movq XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm6 + movq XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm2 + + uncollect_args + mov rsp,rbp ; rsp <- aligned rbp + pop rsp ; rsp <- original rbp + pop rbp + ret + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jidctfst-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jidctfst-sse2.asm new file mode 100644 index 000000000..4658be3ea --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jidctfst-sse2.asm @@ -0,0 +1,502 @@ +; +; jidctfst.asm - fast integer IDCT (SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a fast, not so accurate integer implementation of +; the inverse DCT (Discrete Cosine Transform). The following code is +; based directly on the IJG's original jidctfst.c; see the jidctfst.c +; for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 8 ; 14 is also OK. +%define PASS1_BITS 2 + +%if IFAST_SCALE_BITS != PASS1_BITS +%error "'IFAST_SCALE_BITS' must be equal to 'PASS1_BITS'." +%endif + +%if CONST_BITS == 8 +F_1_082 equ 277 ; FIX(1.082392200) +F_1_414 equ 362 ; FIX(1.414213562) +F_1_847 equ 473 ; FIX(1.847759065) +F_2_613 equ 669 ; FIX(2.613125930) +F_1_613 equ (F_2_613 - 256) ; FIX(2.613125930) - FIX(1) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_1_082 equ DESCALE(1162209775,30-CONST_BITS) ; FIX(1.082392200) +F_1_414 equ DESCALE(1518500249,30-CONST_BITS) ; FIX(1.414213562) +F_1_847 equ DESCALE(1984016188,30-CONST_BITS) ; FIX(1.847759065) +F_2_613 equ DESCALE(2805822602,30-CONST_BITS) ; FIX(2.613125930) +F_1_613 equ (F_2_613 - (1 << CONST_BITS)) ; FIX(2.613125930) - FIX(1) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + +; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow) +; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw) + +%define PRE_MULTIPLY_SCALE_BITS 2 +%define CONST_SHIFT (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS) + + alignz 16 + global EXTN(jconst_idct_ifast_sse2) + +EXTN(jconst_idct_ifast_sse2): + +PW_F1414 times 8 dw F_1_414 << CONST_SHIFT +PW_F1847 times 8 dw F_1_847 << CONST_SHIFT +PW_MF1613 times 8 dw -F_1_613 << CONST_SHIFT +PW_F1082 times 8 dw F_1_082 << CONST_SHIFT +PB_CENTERJSAMP times 16 db CENTERJSAMPLE + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform dequantization and inverse DCT on one block of coefficients. +; +; GLOBAL(void) +; jsimd_idct_ifast_sse2 (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +%define dct_table(b) (b)+8 ; jpeg_component_info * compptr +%define coef_block(b) (b)+12 ; JCOEFPTR coef_block +%define output_buf(b) (b)+16 ; JSAMPARRAY output_buf +%define output_col(b) (b)+20 ; JDIMENSION output_col + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 + + align 16 + global EXTN(jsimd_idct_ifast_sse2) + +EXTN(jsimd_idct_ifast_sse2): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic ebx +; push ecx ; unused +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process columns from input. + +; mov eax, [original_ebp] + mov edx, POINTER [dct_table(eax)] ; quantptr + mov esi, JCOEFPTR [coef_block(eax)] ; inptr + +%ifndef NO_ZERO_COLUMN_TEST_IFAST_SSE2 + mov eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)] + jnz near .columnDCT + + movdqa xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)] + por xmm1, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)] + por xmm1, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)] + por xmm1,xmm0 + packsswb xmm1,xmm1 + packsswb xmm1,xmm1 + movd eax,xmm1 + test eax,eax + jnz short .columnDCT + + ; -- AC terms all zero + + movdqa xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)] + pmullw xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + movdqa xmm7,xmm0 ; xmm0=in0=(00 01 02 03 04 05 06 07) + punpcklwd xmm0,xmm0 ; xmm0=(00 00 01 01 02 02 03 03) + punpckhwd xmm7,xmm7 ; xmm7=(04 04 05 05 06 06 07 07) + + pshufd xmm6,xmm0,0x00 ; xmm6=col0=(00 00 00 00 00 00 00 00) + pshufd xmm2,xmm0,0x55 ; xmm2=col1=(01 01 01 01 01 01 01 01) + pshufd xmm5,xmm0,0xAA ; xmm5=col2=(02 02 02 02 02 02 02 02) + pshufd xmm0,xmm0,0xFF ; xmm0=col3=(03 03 03 03 03 03 03 03) + pshufd xmm1,xmm7,0x00 ; xmm1=col4=(04 04 04 04 04 04 04 04) + pshufd xmm4,xmm7,0x55 ; xmm4=col5=(05 05 05 05 05 05 05 05) + pshufd xmm3,xmm7,0xAA ; xmm3=col6=(06 06 06 06 06 06 06 06) + pshufd xmm7,xmm7,0xFF ; xmm7=col7=(07 07 07 07 07 07 07 07) + + movdqa XMMWORD [wk(0)], xmm2 ; wk(0)=col1 + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=col3 + jmp near .column_end + alignx 16,7 +%endif +.columnDCT: + + ; -- Even part + + movdqa xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)] + pmullw xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_IFAST_MULT_TYPE)] + pmullw xmm1, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_IFAST_MULT_TYPE)] + movdqa xmm2, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_JCOEF)] + movdqa xmm3, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)] + pmullw xmm2, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_IFAST_MULT_TYPE)] + pmullw xmm3, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_IFAST_MULT_TYPE)] + + movdqa xmm4,xmm0 + movdqa xmm5,xmm1 + psubw xmm0,xmm2 ; xmm0=tmp11 + psubw xmm1,xmm3 + paddw xmm4,xmm2 ; xmm4=tmp10 + paddw xmm5,xmm3 ; xmm5=tmp13 + + psllw xmm1,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm1,[GOTOFF(ebx,PW_F1414)] + psubw xmm1,xmm5 ; xmm1=tmp12 + + movdqa xmm6,xmm4 + movdqa xmm7,xmm0 + psubw xmm4,xmm5 ; xmm4=tmp3 + psubw xmm0,xmm1 ; xmm0=tmp2 + paddw xmm6,xmm5 ; xmm6=tmp0 + paddw xmm7,xmm1 ; xmm7=tmp1 + + movdqa XMMWORD [wk(1)], xmm4 ; wk(1)=tmp3 + movdqa XMMWORD [wk(0)], xmm0 ; wk(0)=tmp2 + + ; -- Odd part + + movdqa xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movdqa xmm3, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)] + pmullw xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_IFAST_MULT_TYPE)] + pmullw xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_IFAST_MULT_TYPE)] + movdqa xmm5, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)] + pmullw xmm5, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_IFAST_MULT_TYPE)] + pmullw xmm1, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_IFAST_MULT_TYPE)] + + movdqa xmm4,xmm2 + movdqa xmm0,xmm5 + psubw xmm2,xmm1 ; xmm2=z12 + psubw xmm5,xmm3 ; xmm5=z10 + paddw xmm4,xmm1 ; xmm4=z11 + paddw xmm0,xmm3 ; xmm0=z13 + + movdqa xmm1,xmm5 ; xmm1=z10(unscaled) + psllw xmm2,PRE_MULTIPLY_SCALE_BITS + psllw xmm5,PRE_MULTIPLY_SCALE_BITS + + movdqa xmm3,xmm4 + psubw xmm4,xmm0 + paddw xmm3,xmm0 ; xmm3=tmp7 + + psllw xmm4,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm4,[GOTOFF(ebx,PW_F1414)] ; xmm4=tmp11 + + ; To avoid overflow... + ; + ; (Original) + ; tmp12 = -2.613125930 * z10 + z5; + ; + ; (This implementation) + ; tmp12 = (-1.613125930 - 1) * z10 + z5; + ; = -1.613125930 * z10 - z10 + z5; + + movdqa xmm0,xmm5 + paddw xmm5,xmm2 + pmulhw xmm5,[GOTOFF(ebx,PW_F1847)] ; xmm5=z5 + pmulhw xmm0,[GOTOFF(ebx,PW_MF1613)] + pmulhw xmm2,[GOTOFF(ebx,PW_F1082)] + psubw xmm0,xmm1 + psubw xmm2,xmm5 ; xmm2=tmp10 + paddw xmm0,xmm5 ; xmm0=tmp12 + + ; -- Final output stage + + psubw xmm0,xmm3 ; xmm0=tmp6 + movdqa xmm1,xmm6 + movdqa xmm5,xmm7 + paddw xmm6,xmm3 ; xmm6=data0=(00 01 02 03 04 05 06 07) + paddw xmm7,xmm0 ; xmm7=data1=(10 11 12 13 14 15 16 17) + psubw xmm1,xmm3 ; xmm1=data7=(70 71 72 73 74 75 76 77) + psubw xmm5,xmm0 ; xmm5=data6=(60 61 62 63 64 65 66 67) + psubw xmm4,xmm0 ; xmm4=tmp5 + + movdqa xmm3,xmm6 ; transpose coefficients(phase 1) + punpcklwd xmm6,xmm7 ; xmm6=(00 10 01 11 02 12 03 13) + punpckhwd xmm3,xmm7 ; xmm3=(04 14 05 15 06 16 07 17) + movdqa xmm0,xmm5 ; transpose coefficients(phase 1) + punpcklwd xmm5,xmm1 ; xmm5=(60 70 61 71 62 72 63 73) + punpckhwd xmm0,xmm1 ; xmm0=(64 74 65 75 66 76 67 77) + + movdqa xmm7, XMMWORD [wk(0)] ; xmm7=tmp2 + movdqa xmm1, XMMWORD [wk(1)] ; xmm1=tmp3 + + movdqa XMMWORD [wk(0)], xmm5 ; wk(0)=(60 70 61 71 62 72 63 73) + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=(64 74 65 75 66 76 67 77) + + paddw xmm2,xmm4 ; xmm2=tmp4 + movdqa xmm5,xmm7 + movdqa xmm0,xmm1 + paddw xmm7,xmm4 ; xmm7=data2=(20 21 22 23 24 25 26 27) + paddw xmm1,xmm2 ; xmm1=data4=(40 41 42 43 44 45 46 47) + psubw xmm5,xmm4 ; xmm5=data5=(50 51 52 53 54 55 56 57) + psubw xmm0,xmm2 ; xmm0=data3=(30 31 32 33 34 35 36 37) + + movdqa xmm4,xmm7 ; transpose coefficients(phase 1) + punpcklwd xmm7,xmm0 ; xmm7=(20 30 21 31 22 32 23 33) + punpckhwd xmm4,xmm0 ; xmm4=(24 34 25 35 26 36 27 37) + movdqa xmm2,xmm1 ; transpose coefficients(phase 1) + punpcklwd xmm1,xmm5 ; xmm1=(40 50 41 51 42 52 43 53) + punpckhwd xmm2,xmm5 ; xmm2=(44 54 45 55 46 56 47 57) + + movdqa xmm0,xmm3 ; transpose coefficients(phase 2) + punpckldq xmm3,xmm4 ; xmm3=(04 14 24 34 05 15 25 35) + punpckhdq xmm0,xmm4 ; xmm0=(06 16 26 36 07 17 27 37) + movdqa xmm5,xmm6 ; transpose coefficients(phase 2) + punpckldq xmm6,xmm7 ; xmm6=(00 10 20 30 01 11 21 31) + punpckhdq xmm5,xmm7 ; xmm5=(02 12 22 32 03 13 23 33) + + movdqa xmm4, XMMWORD [wk(0)] ; xmm4=(60 70 61 71 62 72 63 73) + movdqa xmm7, XMMWORD [wk(1)] ; xmm7=(64 74 65 75 66 76 67 77) + + movdqa XMMWORD [wk(0)], xmm3 ; wk(0)=(04 14 24 34 05 15 25 35) + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=(06 16 26 36 07 17 27 37) + + movdqa xmm3,xmm1 ; transpose coefficients(phase 2) + punpckldq xmm1,xmm4 ; xmm1=(40 50 60 70 41 51 61 71) + punpckhdq xmm3,xmm4 ; xmm3=(42 52 62 72 43 53 63 73) + movdqa xmm0,xmm2 ; transpose coefficients(phase 2) + punpckldq xmm2,xmm7 ; xmm2=(44 54 64 74 45 55 65 75) + punpckhdq xmm0,xmm7 ; xmm0=(46 56 66 76 47 57 67 77) + + movdqa xmm4,xmm6 ; transpose coefficients(phase 3) + punpcklqdq xmm6,xmm1 ; xmm6=col0=(00 10 20 30 40 50 60 70) + punpckhqdq xmm4,xmm1 ; xmm4=col1=(01 11 21 31 41 51 61 71) + movdqa xmm7,xmm5 ; transpose coefficients(phase 3) + punpcklqdq xmm5,xmm3 ; xmm5=col2=(02 12 22 32 42 52 62 72) + punpckhqdq xmm7,xmm3 ; xmm7=col3=(03 13 23 33 43 53 63 73) + + movdqa xmm1, XMMWORD [wk(0)] ; xmm1=(04 14 24 34 05 15 25 35) + movdqa xmm3, XMMWORD [wk(1)] ; xmm3=(06 16 26 36 07 17 27 37) + + movdqa XMMWORD [wk(0)], xmm4 ; wk(0)=col1 + movdqa XMMWORD [wk(1)], xmm7 ; wk(1)=col3 + + movdqa xmm4,xmm1 ; transpose coefficients(phase 3) + punpcklqdq xmm1,xmm2 ; xmm1=col4=(04 14 24 34 44 54 64 74) + punpckhqdq xmm4,xmm2 ; xmm4=col5=(05 15 25 35 45 55 65 75) + movdqa xmm7,xmm3 ; transpose coefficients(phase 3) + punpcklqdq xmm3,xmm0 ; xmm3=col6=(06 16 26 36 46 56 66 76) + punpckhqdq xmm7,xmm0 ; xmm7=col7=(07 17 27 37 47 57 67 77) +.column_end: + + ; -- Prefetch the next coefficient block + + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 0*32] + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 1*32] + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 2*32] + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 3*32] + + ; ---- Pass 2: process rows from work array, store into output array. + + mov eax, [original_ebp] + mov edi, JSAMPARRAY [output_buf(eax)] ; (JSAMPROW *) + mov eax, JDIMENSION [output_col(eax)] + + ; -- Even part + + ; xmm6=col0, xmm5=col2, xmm1=col4, xmm3=col6 + + movdqa xmm2,xmm6 + movdqa xmm0,xmm5 + psubw xmm6,xmm1 ; xmm6=tmp11 + psubw xmm5,xmm3 + paddw xmm2,xmm1 ; xmm2=tmp10 + paddw xmm0,xmm3 ; xmm0=tmp13 + + psllw xmm5,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm5,[GOTOFF(ebx,PW_F1414)] + psubw xmm5,xmm0 ; xmm5=tmp12 + + movdqa xmm1,xmm2 + movdqa xmm3,xmm6 + psubw xmm2,xmm0 ; xmm2=tmp3 + psubw xmm6,xmm5 ; xmm6=tmp2 + paddw xmm1,xmm0 ; xmm1=tmp0 + paddw xmm3,xmm5 ; xmm3=tmp1 + + movdqa xmm0, XMMWORD [wk(0)] ; xmm0=col1 + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=col3 + + movdqa XMMWORD [wk(0)], xmm2 ; wk(0)=tmp3 + movdqa XMMWORD [wk(1)], xmm6 ; wk(1)=tmp2 + + ; -- Odd part + + ; xmm0=col1, xmm5=col3, xmm4=col5, xmm7=col7 + + movdqa xmm2,xmm0 + movdqa xmm6,xmm4 + psubw xmm0,xmm7 ; xmm0=z12 + psubw xmm4,xmm5 ; xmm4=z10 + paddw xmm2,xmm7 ; xmm2=z11 + paddw xmm6,xmm5 ; xmm6=z13 + + movdqa xmm7,xmm4 ; xmm7=z10(unscaled) + psllw xmm0,PRE_MULTIPLY_SCALE_BITS + psllw xmm4,PRE_MULTIPLY_SCALE_BITS + + movdqa xmm5,xmm2 + psubw xmm2,xmm6 + paddw xmm5,xmm6 ; xmm5=tmp7 + + psllw xmm2,PRE_MULTIPLY_SCALE_BITS + pmulhw xmm2,[GOTOFF(ebx,PW_F1414)] ; xmm2=tmp11 + + ; To avoid overflow... + ; + ; (Original) + ; tmp12 = -2.613125930 * z10 + z5; + ; + ; (This implementation) + ; tmp12 = (-1.613125930 - 1) * z10 + z5; + ; = -1.613125930 * z10 - z10 + z5; + + movdqa xmm6,xmm4 + paddw xmm4,xmm0 + pmulhw xmm4,[GOTOFF(ebx,PW_F1847)] ; xmm4=z5 + pmulhw xmm6,[GOTOFF(ebx,PW_MF1613)] + pmulhw xmm0,[GOTOFF(ebx,PW_F1082)] + psubw xmm6,xmm7 + psubw xmm0,xmm4 ; xmm0=tmp10 + paddw xmm6,xmm4 ; xmm6=tmp12 + + ; -- Final output stage + + psubw xmm6,xmm5 ; xmm6=tmp6 + movdqa xmm7,xmm1 + movdqa xmm4,xmm3 + paddw xmm1,xmm5 ; xmm1=data0=(00 10 20 30 40 50 60 70) + paddw xmm3,xmm6 ; xmm3=data1=(01 11 21 31 41 51 61 71) + psraw xmm1,(PASS1_BITS+3) ; descale + psraw xmm3,(PASS1_BITS+3) ; descale + psubw xmm7,xmm5 ; xmm7=data7=(07 17 27 37 47 57 67 77) + psubw xmm4,xmm6 ; xmm4=data6=(06 16 26 36 46 56 66 76) + psraw xmm7,(PASS1_BITS+3) ; descale + psraw xmm4,(PASS1_BITS+3) ; descale + psubw xmm2,xmm6 ; xmm2=tmp5 + + packsswb xmm1,xmm4 ; xmm1=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76) + packsswb xmm3,xmm7 ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77) + + movdqa xmm5, XMMWORD [wk(1)] ; xmm5=tmp2 + movdqa xmm6, XMMWORD [wk(0)] ; xmm6=tmp3 + + paddw xmm0,xmm2 ; xmm0=tmp4 + movdqa xmm4,xmm5 + movdqa xmm7,xmm6 + paddw xmm5,xmm2 ; xmm5=data2=(02 12 22 32 42 52 62 72) + paddw xmm6,xmm0 ; xmm6=data4=(04 14 24 34 44 54 64 74) + psraw xmm5,(PASS1_BITS+3) ; descale + psraw xmm6,(PASS1_BITS+3) ; descale + psubw xmm4,xmm2 ; xmm4=data5=(05 15 25 35 45 55 65 75) + psubw xmm7,xmm0 ; xmm7=data3=(03 13 23 33 43 53 63 73) + psraw xmm4,(PASS1_BITS+3) ; descale + psraw xmm7,(PASS1_BITS+3) ; descale + + movdqa xmm2,[GOTOFF(ebx,PB_CENTERJSAMP)] ; xmm2=[PB_CENTERJSAMP] + + packsswb xmm5,xmm6 ; xmm5=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74) + packsswb xmm7,xmm4 ; xmm7=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75) + + paddb xmm1,xmm2 + paddb xmm3,xmm2 + paddb xmm5,xmm2 + paddb xmm7,xmm2 + + movdqa xmm0,xmm1 ; transpose coefficients(phase 1) + punpcklbw xmm1,xmm3 ; xmm1=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71) + punpckhbw xmm0,xmm3 ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77) + movdqa xmm6,xmm5 ; transpose coefficients(phase 1) + punpcklbw xmm5,xmm7 ; xmm5=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73) + punpckhbw xmm6,xmm7 ; xmm6=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75) + + movdqa xmm4,xmm1 ; transpose coefficients(phase 2) + punpcklwd xmm1,xmm5 ; xmm1=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33) + punpckhwd xmm4,xmm5 ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73) + movdqa xmm2,xmm6 ; transpose coefficients(phase 2) + punpcklwd xmm6,xmm0 ; xmm6=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37) + punpckhwd xmm2,xmm0 ; xmm2=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77) + + movdqa xmm3,xmm1 ; transpose coefficients(phase 3) + punpckldq xmm1,xmm6 ; xmm1=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17) + punpckhdq xmm3,xmm6 ; xmm3=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37) + movdqa xmm7,xmm4 ; transpose coefficients(phase 3) + punpckldq xmm4,xmm2 ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57) + punpckhdq xmm7,xmm2 ; xmm7=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77) + + pshufd xmm5,xmm1,0x4E ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07) + pshufd xmm0,xmm3,0x4E ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27) + pshufd xmm6,xmm4,0x4E ; xmm6=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47) + pshufd xmm2,xmm7,0x4E ; xmm2=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67) + + mov edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+2*SIZEOF_JSAMPROW] + movq XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm1 + movq XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm3 + mov edx, JSAMPROW [edi+4*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+6*SIZEOF_JSAMPROW] + movq XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm4 + movq XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm7 + + mov edx, JSAMPROW [edi+1*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+3*SIZEOF_JSAMPROW] + movq XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm5 + movq XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm0 + mov edx, JSAMPROW [edi+5*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+7*SIZEOF_JSAMPROW] + movq XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm6 + movq XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm2 + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; unused + poppic ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jidctint-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jidctint-mmx.asm new file mode 100644 index 000000000..7e25b8203 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jidctint-mmx.asm @@ -0,0 +1,852 @@ +; +; jidctint.asm - accurate integer IDCT (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a slow-but-accurate integer implementation of the +; inverse DCT (Discrete Cosine Transform). The following code is based +; directly on the IJG's original jidctint.c; see the jidctint.c for +; more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 13 +%define PASS1_BITS 2 + +%define DESCALE_P1 (CONST_BITS-PASS1_BITS) +%define DESCALE_P2 (CONST_BITS+PASS1_BITS+3) + +%if CONST_BITS == 13 +F_0_298 equ 2446 ; FIX(0.298631336) +F_0_390 equ 3196 ; FIX(0.390180644) +F_0_541 equ 4433 ; FIX(0.541196100) +F_0_765 equ 6270 ; FIX(0.765366865) +F_0_899 equ 7373 ; FIX(0.899976223) +F_1_175 equ 9633 ; FIX(1.175875602) +F_1_501 equ 12299 ; FIX(1.501321110) +F_1_847 equ 15137 ; FIX(1.847759065) +F_1_961 equ 16069 ; FIX(1.961570560) +F_2_053 equ 16819 ; FIX(2.053119869) +F_2_562 equ 20995 ; FIX(2.562915447) +F_3_072 equ 25172 ; FIX(3.072711026) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_0_298 equ DESCALE( 320652955,30-CONST_BITS) ; FIX(0.298631336) +F_0_390 equ DESCALE( 418953276,30-CONST_BITS) ; FIX(0.390180644) +F_0_541 equ DESCALE( 581104887,30-CONST_BITS) ; FIX(0.541196100) +F_0_765 equ DESCALE( 821806413,30-CONST_BITS) ; FIX(0.765366865) +F_0_899 equ DESCALE( 966342111,30-CONST_BITS) ; FIX(0.899976223) +F_1_175 equ DESCALE(1262586813,30-CONST_BITS) ; FIX(1.175875602) +F_1_501 equ DESCALE(1612031267,30-CONST_BITS) ; FIX(1.501321110) +F_1_847 equ DESCALE(1984016188,30-CONST_BITS) ; FIX(1.847759065) +F_1_961 equ DESCALE(2106220350,30-CONST_BITS) ; FIX(1.961570560) +F_2_053 equ DESCALE(2204520673,30-CONST_BITS) ; FIX(2.053119869) +F_2_562 equ DESCALE(2751909506,30-CONST_BITS) ; FIX(2.562915447) +F_3_072 equ DESCALE(3299298341,30-CONST_BITS) ; FIX(3.072711026) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_idct_islow_mmx) + +EXTN(jconst_idct_islow_mmx): + +PW_F130_F054 times 2 dw (F_0_541+F_0_765), F_0_541 +PW_F054_MF130 times 2 dw F_0_541, (F_0_541-F_1_847) +PW_MF078_F117 times 2 dw (F_1_175-F_1_961), F_1_175 +PW_F117_F078 times 2 dw F_1_175, (F_1_175-F_0_390) +PW_MF060_MF089 times 2 dw (F_0_298-F_0_899),-F_0_899 +PW_MF089_F060 times 2 dw -F_0_899, (F_1_501-F_0_899) +PW_MF050_MF256 times 2 dw (F_2_053-F_2_562),-F_2_562 +PW_MF256_F050 times 2 dw -F_2_562, (F_3_072-F_2_562) +PD_DESCALE_P1 times 2 dd 1 << (DESCALE_P1-1) +PD_DESCALE_P2 times 2 dd 1 << (DESCALE_P2-1) +PB_CENTERJSAMP times 8 db CENTERJSAMPLE + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform dequantization and inverse DCT on one block of coefficients. +; +; GLOBAL(void) +; jsimd_idct_islow_mmx (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +%define dct_table(b) (b)+8 ; jpeg_component_info * compptr +%define coef_block(b) (b)+12 ; JCOEFPTR coef_block +%define output_buf(b) (b)+16 ; JSAMPARRAY output_buf +%define output_col(b) (b)+20 ; JDIMENSION output_col + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_MMWORD ; mmword wk[WK_NUM] +%define WK_NUM 12 +%define workspace wk(0)-DCTSIZE2*SIZEOF_JCOEF + ; JCOEF workspace[DCTSIZE2] + + align 16 + global EXTN(jsimd_idct_islow_mmx) + +EXTN(jsimd_idct_islow_mmx): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_MMWORD) ; align to 64 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [workspace] + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process columns from input, store into work array. + +; mov eax, [original_ebp] + mov edx, POINTER [dct_table(eax)] ; quantptr + mov esi, JCOEFPTR [coef_block(eax)] ; inptr + lea edi, [workspace] ; JCOEF * wsptr + mov ecx, DCTSIZE/4 ; ctr + alignx 16,7 +.columnloop: +%ifndef NO_ZERO_COLUMN_TEST_ISLOW_MMX + mov eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)] + jnz short .columnDCT + + movq mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)] + por mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + por mm1, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)] + por mm0, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + por mm1, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)] + por mm0, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + por mm1,mm0 + packsswb mm1,mm1 + movd eax,mm1 + test eax,eax + jnz short .columnDCT + + ; -- AC terms all zero + + movq mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + pmullw mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + psllw mm0,PASS1_BITS + + movq mm2,mm0 ; mm0=in0=(00 01 02 03) + punpcklwd mm0,mm0 ; mm0=(00 00 01 01) + punpckhwd mm2,mm2 ; mm2=(02 02 03 03) + + movq mm1,mm0 + punpckldq mm0,mm0 ; mm0=(00 00 00 00) + punpckhdq mm1,mm1 ; mm1=(01 01 01 01) + movq mm3,mm2 + punpckldq mm2,mm2 ; mm2=(02 02 02 02) + punpckhdq mm3,mm3 ; mm3=(03 03 03 03) + + movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0 + movq MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm0 + movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm1 + movq MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm1 + movq MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm2 + movq MMWORD [MMBLOCK(2,1,edi,SIZEOF_JCOEF)], mm2 + movq MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm3 + movq MMWORD [MMBLOCK(3,1,edi,SIZEOF_JCOEF)], mm3 + jmp near .nextcolumn + alignx 16,7 +%endif +.columnDCT: + + ; -- Even part + + movq mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)] + pmullw mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw mm1, MMWORD [MMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + movq mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)] + movq mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)] + pmullw mm2, MMWORD [MMBLOCK(4,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw mm3, MMWORD [MMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + ; (Original) + ; z1 = (z2 + z3) * 0.541196100; + ; tmp2 = z1 + z3 * -1.847759065; + ; tmp3 = z1 + z2 * 0.765366865; + ; + ; (This implementation) + ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065); + ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100; + + movq mm4,mm1 ; mm1=in2=z2 + movq mm5,mm1 + punpcklwd mm4,mm3 ; mm3=in6=z3 + punpckhwd mm5,mm3 + movq mm1,mm4 + movq mm3,mm5 + pmaddwd mm4,[GOTOFF(ebx,PW_F130_F054)] ; mm4=tmp3L + pmaddwd mm5,[GOTOFF(ebx,PW_F130_F054)] ; mm5=tmp3H + pmaddwd mm1,[GOTOFF(ebx,PW_F054_MF130)] ; mm1=tmp2L + pmaddwd mm3,[GOTOFF(ebx,PW_F054_MF130)] ; mm3=tmp2H + + movq mm6,mm0 + paddw mm0,mm2 ; mm0=in0+in4 + psubw mm6,mm2 ; mm6=in0-in4 + + pxor mm7,mm7 + pxor mm2,mm2 + punpcklwd mm7,mm0 ; mm7=tmp0L + punpckhwd mm2,mm0 ; mm2=tmp0H + psrad mm7,(16-CONST_BITS) ; psrad mm7,16 & pslld mm7,CONST_BITS + psrad mm2,(16-CONST_BITS) ; psrad mm2,16 & pslld mm2,CONST_BITS + + movq mm0,mm7 + paddd mm7,mm4 ; mm7=tmp10L + psubd mm0,mm4 ; mm0=tmp13L + movq mm4,mm2 + paddd mm2,mm5 ; mm2=tmp10H + psubd mm4,mm5 ; mm4=tmp13H + + movq MMWORD [wk(0)], mm7 ; wk(0)=tmp10L + movq MMWORD [wk(1)], mm2 ; wk(1)=tmp10H + movq MMWORD [wk(2)], mm0 ; wk(2)=tmp13L + movq MMWORD [wk(3)], mm4 ; wk(3)=tmp13H + + pxor mm5,mm5 + pxor mm7,mm7 + punpcklwd mm5,mm6 ; mm5=tmp1L + punpckhwd mm7,mm6 ; mm7=tmp1H + psrad mm5,(16-CONST_BITS) ; psrad mm5,16 & pslld mm5,CONST_BITS + psrad mm7,(16-CONST_BITS) ; psrad mm7,16 & pslld mm7,CONST_BITS + + movq mm2,mm5 + paddd mm5,mm1 ; mm5=tmp11L + psubd mm2,mm1 ; mm2=tmp12L + movq mm0,mm7 + paddd mm7,mm3 ; mm7=tmp11H + psubd mm0,mm3 ; mm0=tmp12H + + movq MMWORD [wk(4)], mm5 ; wk(4)=tmp11L + movq MMWORD [wk(5)], mm7 ; wk(5)=tmp11H + movq MMWORD [wk(6)], mm2 ; wk(6)=tmp12L + movq MMWORD [wk(7)], mm0 ; wk(7)=tmp12H + + ; -- Odd part + + movq mm4, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq mm6, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + pmullw mm4, MMWORD [MMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw mm6, MMWORD [MMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + movq mm1, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movq mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + pmullw mm1, MMWORD [MMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + movq mm5,mm6 + movq mm7,mm4 + paddw mm5,mm3 ; mm5=z3 + paddw mm7,mm1 ; mm7=z4 + + ; (Original) + ; z5 = (z3 + z4) * 1.175875602; + ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; + ; z3 += z5; z4 += z5; + ; + ; (This implementation) + ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; + ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); + + movq mm2,mm5 + movq mm0,mm5 + punpcklwd mm2,mm7 + punpckhwd mm0,mm7 + movq mm5,mm2 + movq mm7,mm0 + pmaddwd mm2,[GOTOFF(ebx,PW_MF078_F117)] ; mm2=z3L + pmaddwd mm0,[GOTOFF(ebx,PW_MF078_F117)] ; mm0=z3H + pmaddwd mm5,[GOTOFF(ebx,PW_F117_F078)] ; mm5=z4L + pmaddwd mm7,[GOTOFF(ebx,PW_F117_F078)] ; mm7=z4H + + movq MMWORD [wk(10)], mm2 ; wk(10)=z3L + movq MMWORD [wk(11)], mm0 ; wk(11)=z3H + + ; (Original) + ; z1 = tmp0 + tmp3; z2 = tmp1 + tmp2; + ; tmp0 = tmp0 * 0.298631336; tmp1 = tmp1 * 2.053119869; + ; tmp2 = tmp2 * 3.072711026; tmp3 = tmp3 * 1.501321110; + ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; + ; tmp0 += z1 + z3; tmp1 += z2 + z4; + ; tmp2 += z2 + z3; tmp3 += z1 + z4; + ; + ; (This implementation) + ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223; + ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447; + ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447); + ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223); + ; tmp0 += z3; tmp1 += z4; + ; tmp2 += z3; tmp3 += z4; + + movq mm2,mm3 + movq mm0,mm3 + punpcklwd mm2,mm4 + punpckhwd mm0,mm4 + movq mm3,mm2 + movq mm4,mm0 + pmaddwd mm2,[GOTOFF(ebx,PW_MF060_MF089)] ; mm2=tmp0L + pmaddwd mm0,[GOTOFF(ebx,PW_MF060_MF089)] ; mm0=tmp0H + pmaddwd mm3,[GOTOFF(ebx,PW_MF089_F060)] ; mm3=tmp3L + pmaddwd mm4,[GOTOFF(ebx,PW_MF089_F060)] ; mm4=tmp3H + + paddd mm2, MMWORD [wk(10)] ; mm2=tmp0L + paddd mm0, MMWORD [wk(11)] ; mm0=tmp0H + paddd mm3,mm5 ; mm3=tmp3L + paddd mm4,mm7 ; mm4=tmp3H + + movq MMWORD [wk(8)], mm2 ; wk(8)=tmp0L + movq MMWORD [wk(9)], mm0 ; wk(9)=tmp0H + + movq mm2,mm1 + movq mm0,mm1 + punpcklwd mm2,mm6 + punpckhwd mm0,mm6 + movq mm1,mm2 + movq mm6,mm0 + pmaddwd mm2,[GOTOFF(ebx,PW_MF050_MF256)] ; mm2=tmp1L + pmaddwd mm0,[GOTOFF(ebx,PW_MF050_MF256)] ; mm0=tmp1H + pmaddwd mm1,[GOTOFF(ebx,PW_MF256_F050)] ; mm1=tmp2L + pmaddwd mm6,[GOTOFF(ebx,PW_MF256_F050)] ; mm6=tmp2H + + paddd mm2,mm5 ; mm2=tmp1L + paddd mm0,mm7 ; mm0=tmp1H + paddd mm1, MMWORD [wk(10)] ; mm1=tmp2L + paddd mm6, MMWORD [wk(11)] ; mm6=tmp2H + + movq MMWORD [wk(10)], mm2 ; wk(10)=tmp1L + movq MMWORD [wk(11)], mm0 ; wk(11)=tmp1H + + ; -- Final output stage + + movq mm5, MMWORD [wk(0)] ; mm5=tmp10L + movq mm7, MMWORD [wk(1)] ; mm7=tmp10H + + movq mm2,mm5 + movq mm0,mm7 + paddd mm5,mm3 ; mm5=data0L + paddd mm7,mm4 ; mm7=data0H + psubd mm2,mm3 ; mm2=data7L + psubd mm0,mm4 ; mm0=data7H + + movq mm3,[GOTOFF(ebx,PD_DESCALE_P1)] ; mm3=[PD_DESCALE_P1] + + paddd mm5,mm3 + paddd mm7,mm3 + psrad mm5,DESCALE_P1 + psrad mm7,DESCALE_P1 + paddd mm2,mm3 + paddd mm0,mm3 + psrad mm2,DESCALE_P1 + psrad mm0,DESCALE_P1 + + packssdw mm5,mm7 ; mm5=data0=(00 01 02 03) + packssdw mm2,mm0 ; mm2=data7=(70 71 72 73) + + movq mm4, MMWORD [wk(4)] ; mm4=tmp11L + movq mm3, MMWORD [wk(5)] ; mm3=tmp11H + + movq mm7,mm4 + movq mm0,mm3 + paddd mm4,mm1 ; mm4=data1L + paddd mm3,mm6 ; mm3=data1H + psubd mm7,mm1 ; mm7=data6L + psubd mm0,mm6 ; mm0=data6H + + movq mm1,[GOTOFF(ebx,PD_DESCALE_P1)] ; mm1=[PD_DESCALE_P1] + + paddd mm4,mm1 + paddd mm3,mm1 + psrad mm4,DESCALE_P1 + psrad mm3,DESCALE_P1 + paddd mm7,mm1 + paddd mm0,mm1 + psrad mm7,DESCALE_P1 + psrad mm0,DESCALE_P1 + + packssdw mm4,mm3 ; mm4=data1=(10 11 12 13) + packssdw mm7,mm0 ; mm7=data6=(60 61 62 63) + + movq mm6,mm5 ; transpose coefficients(phase 1) + punpcklwd mm5,mm4 ; mm5=(00 10 01 11) + punpckhwd mm6,mm4 ; mm6=(02 12 03 13) + movq mm1,mm7 ; transpose coefficients(phase 1) + punpcklwd mm7,mm2 ; mm7=(60 70 61 71) + punpckhwd mm1,mm2 ; mm1=(62 72 63 73) + + movq mm3, MMWORD [wk(6)] ; mm3=tmp12L + movq mm0, MMWORD [wk(7)] ; mm0=tmp12H + movq mm4, MMWORD [wk(10)] ; mm4=tmp1L + movq mm2, MMWORD [wk(11)] ; mm2=tmp1H + + movq MMWORD [wk(0)], mm5 ; wk(0)=(00 10 01 11) + movq MMWORD [wk(1)], mm6 ; wk(1)=(02 12 03 13) + movq MMWORD [wk(4)], mm7 ; wk(4)=(60 70 61 71) + movq MMWORD [wk(5)], mm1 ; wk(5)=(62 72 63 73) + + movq mm5,mm3 + movq mm6,mm0 + paddd mm3,mm4 ; mm3=data2L + paddd mm0,mm2 ; mm0=data2H + psubd mm5,mm4 ; mm5=data5L + psubd mm6,mm2 ; mm6=data5H + + movq mm7,[GOTOFF(ebx,PD_DESCALE_P1)] ; mm7=[PD_DESCALE_P1] + + paddd mm3,mm7 + paddd mm0,mm7 + psrad mm3,DESCALE_P1 + psrad mm0,DESCALE_P1 + paddd mm5,mm7 + paddd mm6,mm7 + psrad mm5,DESCALE_P1 + psrad mm6,DESCALE_P1 + + packssdw mm3,mm0 ; mm3=data2=(20 21 22 23) + packssdw mm5,mm6 ; mm5=data5=(50 51 52 53) + + movq mm1, MMWORD [wk(2)] ; mm1=tmp13L + movq mm4, MMWORD [wk(3)] ; mm4=tmp13H + movq mm2, MMWORD [wk(8)] ; mm2=tmp0L + movq mm7, MMWORD [wk(9)] ; mm7=tmp0H + + movq mm0,mm1 + movq mm6,mm4 + paddd mm1,mm2 ; mm1=data3L + paddd mm4,mm7 ; mm4=data3H + psubd mm0,mm2 ; mm0=data4L + psubd mm6,mm7 ; mm6=data4H + + movq mm2,[GOTOFF(ebx,PD_DESCALE_P1)] ; mm2=[PD_DESCALE_P1] + + paddd mm1,mm2 + paddd mm4,mm2 + psrad mm1,DESCALE_P1 + psrad mm4,DESCALE_P1 + paddd mm0,mm2 + paddd mm6,mm2 + psrad mm0,DESCALE_P1 + psrad mm6,DESCALE_P1 + + packssdw mm1,mm4 ; mm1=data3=(30 31 32 33) + packssdw mm0,mm6 ; mm0=data4=(40 41 42 43) + + movq mm7, MMWORD [wk(0)] ; mm7=(00 10 01 11) + movq mm2, MMWORD [wk(1)] ; mm2=(02 12 03 13) + + movq mm4,mm3 ; transpose coefficients(phase 1) + punpcklwd mm3,mm1 ; mm3=(20 30 21 31) + punpckhwd mm4,mm1 ; mm4=(22 32 23 33) + movq mm6,mm0 ; transpose coefficients(phase 1) + punpcklwd mm0,mm5 ; mm0=(40 50 41 51) + punpckhwd mm6,mm5 ; mm6=(42 52 43 53) + + movq mm1,mm7 ; transpose coefficients(phase 2) + punpckldq mm7,mm3 ; mm7=(00 10 20 30) + punpckhdq mm1,mm3 ; mm1=(01 11 21 31) + movq mm5,mm2 ; transpose coefficients(phase 2) + punpckldq mm2,mm4 ; mm2=(02 12 22 32) + punpckhdq mm5,mm4 ; mm5=(03 13 23 33) + + movq mm3, MMWORD [wk(4)] ; mm3=(60 70 61 71) + movq mm4, MMWORD [wk(5)] ; mm4=(62 72 63 73) + + movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm7 + movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm1 + movq MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm2 + movq MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm5 + + movq mm7,mm0 ; transpose coefficients(phase 2) + punpckldq mm0,mm3 ; mm0=(40 50 60 70) + punpckhdq mm7,mm3 ; mm7=(41 51 61 71) + movq mm1,mm6 ; transpose coefficients(phase 2) + punpckldq mm6,mm4 ; mm6=(42 52 62 72) + punpckhdq mm1,mm4 ; mm1=(43 53 63 73) + + movq MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm0 + movq MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm7 + movq MMWORD [MMBLOCK(2,1,edi,SIZEOF_JCOEF)], mm6 + movq MMWORD [MMBLOCK(3,1,edi,SIZEOF_JCOEF)], mm1 + +.nextcolumn: + add esi, byte 4*SIZEOF_JCOEF ; coef_block + add edx, byte 4*SIZEOF_ISLOW_MULT_TYPE ; quantptr + add edi, byte 4*DCTSIZE*SIZEOF_JCOEF ; wsptr + dec ecx ; ctr + jnz near .columnloop + + ; ---- Pass 2: process rows from work array, store into output array. + + mov eax, [original_ebp] + lea esi, [workspace] ; JCOEF * wsptr + mov edi, JSAMPARRAY [output_buf(eax)] ; (JSAMPROW *) + mov eax, JDIMENSION [output_col(eax)] + mov ecx, DCTSIZE/4 ; ctr + alignx 16,7 +.rowloop: + + ; -- Even part + + movq mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)] + movq mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)] + movq mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)] + + ; (Original) + ; z1 = (z2 + z3) * 0.541196100; + ; tmp2 = z1 + z3 * -1.847759065; + ; tmp3 = z1 + z2 * 0.765366865; + ; + ; (This implementation) + ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065); + ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100; + + movq mm4,mm1 ; mm1=in2=z2 + movq mm5,mm1 + punpcklwd mm4,mm3 ; mm3=in6=z3 + punpckhwd mm5,mm3 + movq mm1,mm4 + movq mm3,mm5 + pmaddwd mm4,[GOTOFF(ebx,PW_F130_F054)] ; mm4=tmp3L + pmaddwd mm5,[GOTOFF(ebx,PW_F130_F054)] ; mm5=tmp3H + pmaddwd mm1,[GOTOFF(ebx,PW_F054_MF130)] ; mm1=tmp2L + pmaddwd mm3,[GOTOFF(ebx,PW_F054_MF130)] ; mm3=tmp2H + + movq mm6,mm0 + paddw mm0,mm2 ; mm0=in0+in4 + psubw mm6,mm2 ; mm6=in0-in4 + + pxor mm7,mm7 + pxor mm2,mm2 + punpcklwd mm7,mm0 ; mm7=tmp0L + punpckhwd mm2,mm0 ; mm2=tmp0H + psrad mm7,(16-CONST_BITS) ; psrad mm7,16 & pslld mm7,CONST_BITS + psrad mm2,(16-CONST_BITS) ; psrad mm2,16 & pslld mm2,CONST_BITS + + movq mm0,mm7 + paddd mm7,mm4 ; mm7=tmp10L + psubd mm0,mm4 ; mm0=tmp13L + movq mm4,mm2 + paddd mm2,mm5 ; mm2=tmp10H + psubd mm4,mm5 ; mm4=tmp13H + + movq MMWORD [wk(0)], mm7 ; wk(0)=tmp10L + movq MMWORD [wk(1)], mm2 ; wk(1)=tmp10H + movq MMWORD [wk(2)], mm0 ; wk(2)=tmp13L + movq MMWORD [wk(3)], mm4 ; wk(3)=tmp13H + + pxor mm5,mm5 + pxor mm7,mm7 + punpcklwd mm5,mm6 ; mm5=tmp1L + punpckhwd mm7,mm6 ; mm7=tmp1H + psrad mm5,(16-CONST_BITS) ; psrad mm5,16 & pslld mm5,CONST_BITS + psrad mm7,(16-CONST_BITS) ; psrad mm7,16 & pslld mm7,CONST_BITS + + movq mm2,mm5 + paddd mm5,mm1 ; mm5=tmp11L + psubd mm2,mm1 ; mm2=tmp12L + movq mm0,mm7 + paddd mm7,mm3 ; mm7=tmp11H + psubd mm0,mm3 ; mm0=tmp12H + + movq MMWORD [wk(4)], mm5 ; wk(4)=tmp11L + movq MMWORD [wk(5)], mm7 ; wk(5)=tmp11H + movq MMWORD [wk(6)], mm2 ; wk(6)=tmp12L + movq MMWORD [wk(7)], mm0 ; wk(7)=tmp12H + + ; -- Odd part + + movq mm4, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq mm6, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movq mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + + movq mm5,mm6 + movq mm7,mm4 + paddw mm5,mm3 ; mm5=z3 + paddw mm7,mm1 ; mm7=z4 + + ; (Original) + ; z5 = (z3 + z4) * 1.175875602; + ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; + ; z3 += z5; z4 += z5; + ; + ; (This implementation) + ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; + ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); + + movq mm2,mm5 + movq mm0,mm5 + punpcklwd mm2,mm7 + punpckhwd mm0,mm7 + movq mm5,mm2 + movq mm7,mm0 + pmaddwd mm2,[GOTOFF(ebx,PW_MF078_F117)] ; mm2=z3L + pmaddwd mm0,[GOTOFF(ebx,PW_MF078_F117)] ; mm0=z3H + pmaddwd mm5,[GOTOFF(ebx,PW_F117_F078)] ; mm5=z4L + pmaddwd mm7,[GOTOFF(ebx,PW_F117_F078)] ; mm7=z4H + + movq MMWORD [wk(10)], mm2 ; wk(10)=z3L + movq MMWORD [wk(11)], mm0 ; wk(11)=z3H + + ; (Original) + ; z1 = tmp0 + tmp3; z2 = tmp1 + tmp2; + ; tmp0 = tmp0 * 0.298631336; tmp1 = tmp1 * 2.053119869; + ; tmp2 = tmp2 * 3.072711026; tmp3 = tmp3 * 1.501321110; + ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; + ; tmp0 += z1 + z3; tmp1 += z2 + z4; + ; tmp2 += z2 + z3; tmp3 += z1 + z4; + ; + ; (This implementation) + ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223; + ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447; + ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447); + ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223); + ; tmp0 += z3; tmp1 += z4; + ; tmp2 += z3; tmp3 += z4; + + movq mm2,mm3 + movq mm0,mm3 + punpcklwd mm2,mm4 + punpckhwd mm0,mm4 + movq mm3,mm2 + movq mm4,mm0 + pmaddwd mm2,[GOTOFF(ebx,PW_MF060_MF089)] ; mm2=tmp0L + pmaddwd mm0,[GOTOFF(ebx,PW_MF060_MF089)] ; mm0=tmp0H + pmaddwd mm3,[GOTOFF(ebx,PW_MF089_F060)] ; mm3=tmp3L + pmaddwd mm4,[GOTOFF(ebx,PW_MF089_F060)] ; mm4=tmp3H + + paddd mm2, MMWORD [wk(10)] ; mm2=tmp0L + paddd mm0, MMWORD [wk(11)] ; mm0=tmp0H + paddd mm3,mm5 ; mm3=tmp3L + paddd mm4,mm7 ; mm4=tmp3H + + movq MMWORD [wk(8)], mm2 ; wk(8)=tmp0L + movq MMWORD [wk(9)], mm0 ; wk(9)=tmp0H + + movq mm2,mm1 + movq mm0,mm1 + punpcklwd mm2,mm6 + punpckhwd mm0,mm6 + movq mm1,mm2 + movq mm6,mm0 + pmaddwd mm2,[GOTOFF(ebx,PW_MF050_MF256)] ; mm2=tmp1L + pmaddwd mm0,[GOTOFF(ebx,PW_MF050_MF256)] ; mm0=tmp1H + pmaddwd mm1,[GOTOFF(ebx,PW_MF256_F050)] ; mm1=tmp2L + pmaddwd mm6,[GOTOFF(ebx,PW_MF256_F050)] ; mm6=tmp2H + + paddd mm2,mm5 ; mm2=tmp1L + paddd mm0,mm7 ; mm0=tmp1H + paddd mm1, MMWORD [wk(10)] ; mm1=tmp2L + paddd mm6, MMWORD [wk(11)] ; mm6=tmp2H + + movq MMWORD [wk(10)], mm2 ; wk(10)=tmp1L + movq MMWORD [wk(11)], mm0 ; wk(11)=tmp1H + + ; -- Final output stage + + movq mm5, MMWORD [wk(0)] ; mm5=tmp10L + movq mm7, MMWORD [wk(1)] ; mm7=tmp10H + + movq mm2,mm5 + movq mm0,mm7 + paddd mm5,mm3 ; mm5=data0L + paddd mm7,mm4 ; mm7=data0H + psubd mm2,mm3 ; mm2=data7L + psubd mm0,mm4 ; mm0=data7H + + movq mm3,[GOTOFF(ebx,PD_DESCALE_P2)] ; mm3=[PD_DESCALE_P2] + + paddd mm5,mm3 + paddd mm7,mm3 + psrad mm5,DESCALE_P2 + psrad mm7,DESCALE_P2 + paddd mm2,mm3 + paddd mm0,mm3 + psrad mm2,DESCALE_P2 + psrad mm0,DESCALE_P2 + + packssdw mm5,mm7 ; mm5=data0=(00 10 20 30) + packssdw mm2,mm0 ; mm2=data7=(07 17 27 37) + + movq mm4, MMWORD [wk(4)] ; mm4=tmp11L + movq mm3, MMWORD [wk(5)] ; mm3=tmp11H + + movq mm7,mm4 + movq mm0,mm3 + paddd mm4,mm1 ; mm4=data1L + paddd mm3,mm6 ; mm3=data1H + psubd mm7,mm1 ; mm7=data6L + psubd mm0,mm6 ; mm0=data6H + + movq mm1,[GOTOFF(ebx,PD_DESCALE_P2)] ; mm1=[PD_DESCALE_P2] + + paddd mm4,mm1 + paddd mm3,mm1 + psrad mm4,DESCALE_P2 + psrad mm3,DESCALE_P2 + paddd mm7,mm1 + paddd mm0,mm1 + psrad mm7,DESCALE_P2 + psrad mm0,DESCALE_P2 + + packssdw mm4,mm3 ; mm4=data1=(01 11 21 31) + packssdw mm7,mm0 ; mm7=data6=(06 16 26 36) + + packsswb mm5,mm7 ; mm5=(00 10 20 30 06 16 26 36) + packsswb mm4,mm2 ; mm4=(01 11 21 31 07 17 27 37) + + movq mm6, MMWORD [wk(6)] ; mm6=tmp12L + movq mm1, MMWORD [wk(7)] ; mm1=tmp12H + movq mm3, MMWORD [wk(10)] ; mm3=tmp1L + movq mm0, MMWORD [wk(11)] ; mm0=tmp1H + + movq MMWORD [wk(0)], mm5 ; wk(0)=(00 10 20 30 06 16 26 36) + movq MMWORD [wk(1)], mm4 ; wk(1)=(01 11 21 31 07 17 27 37) + + movq mm7,mm6 + movq mm2,mm1 + paddd mm6,mm3 ; mm6=data2L + paddd mm1,mm0 ; mm1=data2H + psubd mm7,mm3 ; mm7=data5L + psubd mm2,mm0 ; mm2=data5H + + movq mm5,[GOTOFF(ebx,PD_DESCALE_P2)] ; mm5=[PD_DESCALE_P2] + + paddd mm6,mm5 + paddd mm1,mm5 + psrad mm6,DESCALE_P2 + psrad mm1,DESCALE_P2 + paddd mm7,mm5 + paddd mm2,mm5 + psrad mm7,DESCALE_P2 + psrad mm2,DESCALE_P2 + + packssdw mm6,mm1 ; mm6=data2=(02 12 22 32) + packssdw mm7,mm2 ; mm7=data5=(05 15 25 35) + + movq mm4, MMWORD [wk(2)] ; mm4=tmp13L + movq mm3, MMWORD [wk(3)] ; mm3=tmp13H + movq mm0, MMWORD [wk(8)] ; mm0=tmp0L + movq mm5, MMWORD [wk(9)] ; mm5=tmp0H + + movq mm1,mm4 + movq mm2,mm3 + paddd mm4,mm0 ; mm4=data3L + paddd mm3,mm5 ; mm3=data3H + psubd mm1,mm0 ; mm1=data4L + psubd mm2,mm5 ; mm2=data4H + + movq mm0,[GOTOFF(ebx,PD_DESCALE_P2)] ; mm0=[PD_DESCALE_P2] + + paddd mm4,mm0 + paddd mm3,mm0 + psrad mm4,DESCALE_P2 + psrad mm3,DESCALE_P2 + paddd mm1,mm0 + paddd mm2,mm0 + psrad mm1,DESCALE_P2 + psrad mm2,DESCALE_P2 + + movq mm5,[GOTOFF(ebx,PB_CENTERJSAMP)] ; mm5=[PB_CENTERJSAMP] + + packssdw mm4,mm3 ; mm4=data3=(03 13 23 33) + packssdw mm1,mm2 ; mm1=data4=(04 14 24 34) + + movq mm0, MMWORD [wk(0)] ; mm0=(00 10 20 30 06 16 26 36) + movq mm3, MMWORD [wk(1)] ; mm3=(01 11 21 31 07 17 27 37) + + packsswb mm6,mm1 ; mm6=(02 12 22 32 04 14 24 34) + packsswb mm4,mm7 ; mm4=(03 13 23 33 05 15 25 35) + + paddb mm0,mm5 + paddb mm3,mm5 + paddb mm6,mm5 + paddb mm4,mm5 + + movq mm2,mm0 ; transpose coefficients(phase 1) + punpcklbw mm0,mm3 ; mm0=(00 01 10 11 20 21 30 31) + punpckhbw mm2,mm3 ; mm2=(06 07 16 17 26 27 36 37) + movq mm1,mm6 ; transpose coefficients(phase 1) + punpcklbw mm6,mm4 ; mm6=(02 03 12 13 22 23 32 33) + punpckhbw mm1,mm4 ; mm1=(04 05 14 15 24 25 34 35) + + movq mm7,mm0 ; transpose coefficients(phase 2) + punpcklwd mm0,mm6 ; mm0=(00 01 02 03 10 11 12 13) + punpckhwd mm7,mm6 ; mm7=(20 21 22 23 30 31 32 33) + movq mm5,mm1 ; transpose coefficients(phase 2) + punpcklwd mm1,mm2 ; mm1=(04 05 06 07 14 15 16 17) + punpckhwd mm5,mm2 ; mm5=(24 25 26 27 34 35 36 37) + + movq mm3,mm0 ; transpose coefficients(phase 3) + punpckldq mm0,mm1 ; mm0=(00 01 02 03 04 05 06 07) + punpckhdq mm3,mm1 ; mm3=(10 11 12 13 14 15 16 17) + movq mm4,mm7 ; transpose coefficients(phase 3) + punpckldq mm7,mm5 ; mm7=(20 21 22 23 24 25 26 27) + punpckhdq mm4,mm5 ; mm4=(30 31 32 33 34 35 36 37) + + pushpic ebx ; save GOT address + + mov edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] + mov ebx, JSAMPROW [edi+1*SIZEOF_JSAMPROW] + movq MMWORD [edx+eax*SIZEOF_JSAMPLE], mm0 + movq MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm3 + mov edx, JSAMPROW [edi+2*SIZEOF_JSAMPROW] + mov ebx, JSAMPROW [edi+3*SIZEOF_JSAMPROW] + movq MMWORD [edx+eax*SIZEOF_JSAMPLE], mm7 + movq MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm4 + + poppic ebx ; restore GOT address + + add esi, byte 4*SIZEOF_JCOEF ; wsptr + add edi, byte 4*SIZEOF_JSAMPROW + dec ecx ; ctr + jnz near .rowloop + + emms ; empty MMX state + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jidctint-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jidctint-sse2-64.asm new file mode 100644 index 000000000..32bbfd8d3 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jidctint-sse2-64.asm @@ -0,0 +1,848 @@ +; +; jidctint.asm - accurate integer IDCT (64-bit SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a slow-but-accurate integer implementation of the +; inverse DCT (Discrete Cosine Transform). The following code is based +; directly on the IJG's original jidctint.c; see the jidctint.c for +; more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 13 +%define PASS1_BITS 2 + +%define DESCALE_P1 (CONST_BITS-PASS1_BITS) +%define DESCALE_P2 (CONST_BITS+PASS1_BITS+3) + +%if CONST_BITS == 13 +F_0_298 equ 2446 ; FIX(0.298631336) +F_0_390 equ 3196 ; FIX(0.390180644) +F_0_541 equ 4433 ; FIX(0.541196100) +F_0_765 equ 6270 ; FIX(0.765366865) +F_0_899 equ 7373 ; FIX(0.899976223) +F_1_175 equ 9633 ; FIX(1.175875602) +F_1_501 equ 12299 ; FIX(1.501321110) +F_1_847 equ 15137 ; FIX(1.847759065) +F_1_961 equ 16069 ; FIX(1.961570560) +F_2_053 equ 16819 ; FIX(2.053119869) +F_2_562 equ 20995 ; FIX(2.562915447) +F_3_072 equ 25172 ; FIX(3.072711026) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_0_298 equ DESCALE( 320652955,30-CONST_BITS) ; FIX(0.298631336) +F_0_390 equ DESCALE( 418953276,30-CONST_BITS) ; FIX(0.390180644) +F_0_541 equ DESCALE( 581104887,30-CONST_BITS) ; FIX(0.541196100) +F_0_765 equ DESCALE( 821806413,30-CONST_BITS) ; FIX(0.765366865) +F_0_899 equ DESCALE( 966342111,30-CONST_BITS) ; FIX(0.899976223) +F_1_175 equ DESCALE(1262586813,30-CONST_BITS) ; FIX(1.175875602) +F_1_501 equ DESCALE(1612031267,30-CONST_BITS) ; FIX(1.501321110) +F_1_847 equ DESCALE(1984016188,30-CONST_BITS) ; FIX(1.847759065) +F_1_961 equ DESCALE(2106220350,30-CONST_BITS) ; FIX(1.961570560) +F_2_053 equ DESCALE(2204520673,30-CONST_BITS) ; FIX(2.053119869) +F_2_562 equ DESCALE(2751909506,30-CONST_BITS) ; FIX(2.562915447) +F_3_072 equ DESCALE(3299298341,30-CONST_BITS) ; FIX(3.072711026) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_idct_islow_sse2) + +EXTN(jconst_idct_islow_sse2): + +PW_F130_F054 times 4 dw (F_0_541+F_0_765), F_0_541 +PW_F054_MF130 times 4 dw F_0_541, (F_0_541-F_1_847) +PW_MF078_F117 times 4 dw (F_1_175-F_1_961), F_1_175 +PW_F117_F078 times 4 dw F_1_175, (F_1_175-F_0_390) +PW_MF060_MF089 times 4 dw (F_0_298-F_0_899),-F_0_899 +PW_MF089_F060 times 4 dw -F_0_899, (F_1_501-F_0_899) +PW_MF050_MF256 times 4 dw (F_2_053-F_2_562),-F_2_562 +PW_MF256_F050 times 4 dw -F_2_562, (F_3_072-F_2_562) +PD_DESCALE_P1 times 4 dd 1 << (DESCALE_P1-1) +PD_DESCALE_P2 times 4 dd 1 << (DESCALE_P2-1) +PB_CENTERJSAMP times 16 db CENTERJSAMPLE + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 +; +; Perform dequantization and inverse DCT on one block of coefficients. +; +; GLOBAL(void) +; jsimd_idct_islow_sse2 (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +; r10 = jpeg_component_info * compptr +; r11 = JCOEFPTR coef_block +; r12 = JSAMPARRAY output_buf +; r13 = JDIMENSION output_col + +%define original_rbp rbp+0 +%define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 12 + + align 16 + global EXTN(jsimd_idct_islow_sse2) + +EXTN(jsimd_idct_islow_sse2): + push rbp + mov rax,rsp ; rax = original rbp + sub rsp, byte 4 + and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [rsp],rax + mov rbp,rsp ; rbp = aligned rbp + lea rsp, [wk(0)] + collect_args + + ; ---- Pass 1: process columns from input. + + mov rdx, r10 ; quantptr + mov rsi, r11 ; inptr + +%ifndef NO_ZERO_COLUMN_TEST_ISLOW_SSE2 + mov eax, DWORD [DWBLOCK(1,0,rsi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(2,0,rsi,SIZEOF_JCOEF)] + jnz near .columnDCT + + movdqa xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)] + por xmm1, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)] + por xmm1, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)] + por xmm1,xmm0 + packsswb xmm1,xmm1 + packsswb xmm1,xmm1 + movd eax,xmm1 + test rax,rax + jnz short .columnDCT + + ; -- AC terms all zero + + movdqa xmm5, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)] + pmullw xmm5, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + + psllw xmm5,PASS1_BITS + + movdqa xmm4,xmm5 ; xmm5=in0=(00 01 02 03 04 05 06 07) + punpcklwd xmm5,xmm5 ; xmm5=(00 00 01 01 02 02 03 03) + punpckhwd xmm4,xmm4 ; xmm4=(04 04 05 05 06 06 07 07) + + pshufd xmm7,xmm5,0x00 ; xmm7=col0=(00 00 00 00 00 00 00 00) + pshufd xmm6,xmm5,0x55 ; xmm6=col1=(01 01 01 01 01 01 01 01) + pshufd xmm1,xmm5,0xAA ; xmm1=col2=(02 02 02 02 02 02 02 02) + pshufd xmm5,xmm5,0xFF ; xmm5=col3=(03 03 03 03 03 03 03 03) + pshufd xmm0,xmm4,0x00 ; xmm0=col4=(04 04 04 04 04 04 04 04) + pshufd xmm3,xmm4,0x55 ; xmm3=col5=(05 05 05 05 05 05 05 05) + pshufd xmm2,xmm4,0xAA ; xmm2=col6=(06 06 06 06 06 06 06 06) + pshufd xmm4,xmm4,0xFF ; xmm4=col7=(07 07 07 07 07 07 07 07) + + movdqa XMMWORD [wk(8)], xmm6 ; wk(8)=col1 + movdqa XMMWORD [wk(9)], xmm5 ; wk(9)=col3 + movdqa XMMWORD [wk(10)], xmm3 ; wk(10)=col5 + movdqa XMMWORD [wk(11)], xmm4 ; wk(11)=col7 + jmp near .column_end +%endif +.columnDCT: + + ; -- Even part + + movdqa xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)] + pmullw xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm1, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + movdqa xmm2, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)] + movdqa xmm3, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)] + pmullw xmm2, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm3, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + + ; (Original) + ; z1 = (z2 + z3) * 0.541196100; + ; tmp2 = z1 + z3 * -1.847759065; + ; tmp3 = z1 + z2 * 0.765366865; + ; + ; (This implementation) + ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065); + ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100; + + movdqa xmm4,xmm1 ; xmm1=in2=z2 + movdqa xmm5,xmm1 + punpcklwd xmm4,xmm3 ; xmm3=in6=z3 + punpckhwd xmm5,xmm3 + movdqa xmm1,xmm4 + movdqa xmm3,xmm5 + pmaddwd xmm4,[rel PW_F130_F054] ; xmm4=tmp3L + pmaddwd xmm5,[rel PW_F130_F054] ; xmm5=tmp3H + pmaddwd xmm1,[rel PW_F054_MF130] ; xmm1=tmp2L + pmaddwd xmm3,[rel PW_F054_MF130] ; xmm3=tmp2H + + movdqa xmm6,xmm0 + paddw xmm0,xmm2 ; xmm0=in0+in4 + psubw xmm6,xmm2 ; xmm6=in0-in4 + + pxor xmm7,xmm7 + pxor xmm2,xmm2 + punpcklwd xmm7,xmm0 ; xmm7=tmp0L + punpckhwd xmm2,xmm0 ; xmm2=tmp0H + psrad xmm7,(16-CONST_BITS) ; psrad xmm7,16 & pslld xmm7,CONST_BITS + psrad xmm2,(16-CONST_BITS) ; psrad xmm2,16 & pslld xmm2,CONST_BITS + + movdqa xmm0,xmm7 + paddd xmm7,xmm4 ; xmm7=tmp10L + psubd xmm0,xmm4 ; xmm0=tmp13L + movdqa xmm4,xmm2 + paddd xmm2,xmm5 ; xmm2=tmp10H + psubd xmm4,xmm5 ; xmm4=tmp13H + + movdqa XMMWORD [wk(0)], xmm7 ; wk(0)=tmp10L + movdqa XMMWORD [wk(1)], xmm2 ; wk(1)=tmp10H + movdqa XMMWORD [wk(2)], xmm0 ; wk(2)=tmp13L + movdqa XMMWORD [wk(3)], xmm4 ; wk(3)=tmp13H + + pxor xmm5,xmm5 + pxor xmm7,xmm7 + punpcklwd xmm5,xmm6 ; xmm5=tmp1L + punpckhwd xmm7,xmm6 ; xmm7=tmp1H + psrad xmm5,(16-CONST_BITS) ; psrad xmm5,16 & pslld xmm5,CONST_BITS + psrad xmm7,(16-CONST_BITS) ; psrad xmm7,16 & pslld xmm7,CONST_BITS + + movdqa xmm2,xmm5 + paddd xmm5,xmm1 ; xmm5=tmp11L + psubd xmm2,xmm1 ; xmm2=tmp12L + movdqa xmm0,xmm7 + paddd xmm7,xmm3 ; xmm7=tmp11H + psubd xmm0,xmm3 ; xmm0=tmp12H + + movdqa XMMWORD [wk(4)], xmm5 ; wk(4)=tmp11L + movdqa XMMWORD [wk(5)], xmm7 ; wk(5)=tmp11H + movdqa XMMWORD [wk(6)], xmm2 ; wk(6)=tmp12L + movdqa XMMWORD [wk(7)], xmm0 ; wk(7)=tmp12H + + ; -- Odd part + + movdqa xmm4, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)] + movdqa xmm6, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)] + pmullw xmm4, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm6, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + movdqa xmm1, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)] + movdqa xmm3, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)] + pmullw xmm1, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + + movdqa xmm5,xmm6 + movdqa xmm7,xmm4 + paddw xmm5,xmm3 ; xmm5=z3 + paddw xmm7,xmm1 ; xmm7=z4 + + ; (Original) + ; z5 = (z3 + z4) * 1.175875602; + ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; + ; z3 += z5; z4 += z5; + ; + ; (This implementation) + ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; + ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); + + movdqa xmm2,xmm5 + movdqa xmm0,xmm5 + punpcklwd xmm2,xmm7 + punpckhwd xmm0,xmm7 + movdqa xmm5,xmm2 + movdqa xmm7,xmm0 + pmaddwd xmm2,[rel PW_MF078_F117] ; xmm2=z3L + pmaddwd xmm0,[rel PW_MF078_F117] ; xmm0=z3H + pmaddwd xmm5,[rel PW_F117_F078] ; xmm5=z4L + pmaddwd xmm7,[rel PW_F117_F078] ; xmm7=z4H + + movdqa XMMWORD [wk(10)], xmm2 ; wk(10)=z3L + movdqa XMMWORD [wk(11)], xmm0 ; wk(11)=z3H + + ; (Original) + ; z1 = tmp0 + tmp3; z2 = tmp1 + tmp2; + ; tmp0 = tmp0 * 0.298631336; tmp1 = tmp1 * 2.053119869; + ; tmp2 = tmp2 * 3.072711026; tmp3 = tmp3 * 1.501321110; + ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; + ; tmp0 += z1 + z3; tmp1 += z2 + z4; + ; tmp2 += z2 + z3; tmp3 += z1 + z4; + ; + ; (This implementation) + ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223; + ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447; + ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447); + ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223); + ; tmp0 += z3; tmp1 += z4; + ; tmp2 += z3; tmp3 += z4; + + movdqa xmm2,xmm3 + movdqa xmm0,xmm3 + punpcklwd xmm2,xmm4 + punpckhwd xmm0,xmm4 + movdqa xmm3,xmm2 + movdqa xmm4,xmm0 + pmaddwd xmm2,[rel PW_MF060_MF089] ; xmm2=tmp0L + pmaddwd xmm0,[rel PW_MF060_MF089] ; xmm0=tmp0H + pmaddwd xmm3,[rel PW_MF089_F060] ; xmm3=tmp3L + pmaddwd xmm4,[rel PW_MF089_F060] ; xmm4=tmp3H + + paddd xmm2, XMMWORD [wk(10)] ; xmm2=tmp0L + paddd xmm0, XMMWORD [wk(11)] ; xmm0=tmp0H + paddd xmm3,xmm5 ; xmm3=tmp3L + paddd xmm4,xmm7 ; xmm4=tmp3H + + movdqa XMMWORD [wk(8)], xmm2 ; wk(8)=tmp0L + movdqa XMMWORD [wk(9)], xmm0 ; wk(9)=tmp0H + + movdqa xmm2,xmm1 + movdqa xmm0,xmm1 + punpcklwd xmm2,xmm6 + punpckhwd xmm0,xmm6 + movdqa xmm1,xmm2 + movdqa xmm6,xmm0 + pmaddwd xmm2,[rel PW_MF050_MF256] ; xmm2=tmp1L + pmaddwd xmm0,[rel PW_MF050_MF256] ; xmm0=tmp1H + pmaddwd xmm1,[rel PW_MF256_F050] ; xmm1=tmp2L + pmaddwd xmm6,[rel PW_MF256_F050] ; xmm6=tmp2H + + paddd xmm2,xmm5 ; xmm2=tmp1L + paddd xmm0,xmm7 ; xmm0=tmp1H + paddd xmm1, XMMWORD [wk(10)] ; xmm1=tmp2L + paddd xmm6, XMMWORD [wk(11)] ; xmm6=tmp2H + + movdqa XMMWORD [wk(10)], xmm2 ; wk(10)=tmp1L + movdqa XMMWORD [wk(11)], xmm0 ; wk(11)=tmp1H + + ; -- Final output stage + + movdqa xmm5, XMMWORD [wk(0)] ; xmm5=tmp10L + movdqa xmm7, XMMWORD [wk(1)] ; xmm7=tmp10H + + movdqa xmm2,xmm5 + movdqa xmm0,xmm7 + paddd xmm5,xmm3 ; xmm5=data0L + paddd xmm7,xmm4 ; xmm7=data0H + psubd xmm2,xmm3 ; xmm2=data7L + psubd xmm0,xmm4 ; xmm0=data7H + + movdqa xmm3,[rel PD_DESCALE_P1] ; xmm3=[rel PD_DESCALE_P1] + + paddd xmm5,xmm3 + paddd xmm7,xmm3 + psrad xmm5,DESCALE_P1 + psrad xmm7,DESCALE_P1 + paddd xmm2,xmm3 + paddd xmm0,xmm3 + psrad xmm2,DESCALE_P1 + psrad xmm0,DESCALE_P1 + + packssdw xmm5,xmm7 ; xmm5=data0=(00 01 02 03 04 05 06 07) + packssdw xmm2,xmm0 ; xmm2=data7=(70 71 72 73 74 75 76 77) + + movdqa xmm4, XMMWORD [wk(4)] ; xmm4=tmp11L + movdqa xmm3, XMMWORD [wk(5)] ; xmm3=tmp11H + + movdqa xmm7,xmm4 + movdqa xmm0,xmm3 + paddd xmm4,xmm1 ; xmm4=data1L + paddd xmm3,xmm6 ; xmm3=data1H + psubd xmm7,xmm1 ; xmm7=data6L + psubd xmm0,xmm6 ; xmm0=data6H + + movdqa xmm1,[rel PD_DESCALE_P1] ; xmm1=[rel PD_DESCALE_P1] + + paddd xmm4,xmm1 + paddd xmm3,xmm1 + psrad xmm4,DESCALE_P1 + psrad xmm3,DESCALE_P1 + paddd xmm7,xmm1 + paddd xmm0,xmm1 + psrad xmm7,DESCALE_P1 + psrad xmm0,DESCALE_P1 + + packssdw xmm4,xmm3 ; xmm4=data1=(10 11 12 13 14 15 16 17) + packssdw xmm7,xmm0 ; xmm7=data6=(60 61 62 63 64 65 66 67) + + movdqa xmm6,xmm5 ; transpose coefficients(phase 1) + punpcklwd xmm5,xmm4 ; xmm5=(00 10 01 11 02 12 03 13) + punpckhwd xmm6,xmm4 ; xmm6=(04 14 05 15 06 16 07 17) + movdqa xmm1,xmm7 ; transpose coefficients(phase 1) + punpcklwd xmm7,xmm2 ; xmm7=(60 70 61 71 62 72 63 73) + punpckhwd xmm1,xmm2 ; xmm1=(64 74 65 75 66 76 67 77) + + movdqa xmm3, XMMWORD [wk(6)] ; xmm3=tmp12L + movdqa xmm0, XMMWORD [wk(7)] ; xmm0=tmp12H + movdqa xmm4, XMMWORD [wk(10)] ; xmm4=tmp1L + movdqa xmm2, XMMWORD [wk(11)] ; xmm2=tmp1H + + movdqa XMMWORD [wk(0)], xmm5 ; wk(0)=(00 10 01 11 02 12 03 13) + movdqa XMMWORD [wk(1)], xmm6 ; wk(1)=(04 14 05 15 06 16 07 17) + movdqa XMMWORD [wk(4)], xmm7 ; wk(4)=(60 70 61 71 62 72 63 73) + movdqa XMMWORD [wk(5)], xmm1 ; wk(5)=(64 74 65 75 66 76 67 77) + + movdqa xmm5,xmm3 + movdqa xmm6,xmm0 + paddd xmm3,xmm4 ; xmm3=data2L + paddd xmm0,xmm2 ; xmm0=data2H + psubd xmm5,xmm4 ; xmm5=data5L + psubd xmm6,xmm2 ; xmm6=data5H + + movdqa xmm7,[rel PD_DESCALE_P1] ; xmm7=[rel PD_DESCALE_P1] + + paddd xmm3,xmm7 + paddd xmm0,xmm7 + psrad xmm3,DESCALE_P1 + psrad xmm0,DESCALE_P1 + paddd xmm5,xmm7 + paddd xmm6,xmm7 + psrad xmm5,DESCALE_P1 + psrad xmm6,DESCALE_P1 + + packssdw xmm3,xmm0 ; xmm3=data2=(20 21 22 23 24 25 26 27) + packssdw xmm5,xmm6 ; xmm5=data5=(50 51 52 53 54 55 56 57) + + movdqa xmm1, XMMWORD [wk(2)] ; xmm1=tmp13L + movdqa xmm4, XMMWORD [wk(3)] ; xmm4=tmp13H + movdqa xmm2, XMMWORD [wk(8)] ; xmm2=tmp0L + movdqa xmm7, XMMWORD [wk(9)] ; xmm7=tmp0H + + movdqa xmm0,xmm1 + movdqa xmm6,xmm4 + paddd xmm1,xmm2 ; xmm1=data3L + paddd xmm4,xmm7 ; xmm4=data3H + psubd xmm0,xmm2 ; xmm0=data4L + psubd xmm6,xmm7 ; xmm6=data4H + + movdqa xmm2,[rel PD_DESCALE_P1] ; xmm2=[rel PD_DESCALE_P1] + + paddd xmm1,xmm2 + paddd xmm4,xmm2 + psrad xmm1,DESCALE_P1 + psrad xmm4,DESCALE_P1 + paddd xmm0,xmm2 + paddd xmm6,xmm2 + psrad xmm0,DESCALE_P1 + psrad xmm6,DESCALE_P1 + + packssdw xmm1,xmm4 ; xmm1=data3=(30 31 32 33 34 35 36 37) + packssdw xmm0,xmm6 ; xmm0=data4=(40 41 42 43 44 45 46 47) + + movdqa xmm7, XMMWORD [wk(0)] ; xmm7=(00 10 01 11 02 12 03 13) + movdqa xmm2, XMMWORD [wk(1)] ; xmm2=(04 14 05 15 06 16 07 17) + + movdqa xmm4,xmm3 ; transpose coefficients(phase 1) + punpcklwd xmm3,xmm1 ; xmm3=(20 30 21 31 22 32 23 33) + punpckhwd xmm4,xmm1 ; xmm4=(24 34 25 35 26 36 27 37) + movdqa xmm6,xmm0 ; transpose coefficients(phase 1) + punpcklwd xmm0,xmm5 ; xmm0=(40 50 41 51 42 52 43 53) + punpckhwd xmm6,xmm5 ; xmm6=(44 54 45 55 46 56 47 57) + + movdqa xmm1,xmm7 ; transpose coefficients(phase 2) + punpckldq xmm7,xmm3 ; xmm7=(00 10 20 30 01 11 21 31) + punpckhdq xmm1,xmm3 ; xmm1=(02 12 22 32 03 13 23 33) + movdqa xmm5,xmm2 ; transpose coefficients(phase 2) + punpckldq xmm2,xmm4 ; xmm2=(04 14 24 34 05 15 25 35) + punpckhdq xmm5,xmm4 ; xmm5=(06 16 26 36 07 17 27 37) + + movdqa xmm3, XMMWORD [wk(4)] ; xmm3=(60 70 61 71 62 72 63 73) + movdqa xmm4, XMMWORD [wk(5)] ; xmm4=(64 74 65 75 66 76 67 77) + + movdqa XMMWORD [wk(6)], xmm2 ; wk(6)=(04 14 24 34 05 15 25 35) + movdqa XMMWORD [wk(7)], xmm5 ; wk(7)=(06 16 26 36 07 17 27 37) + + movdqa xmm2,xmm0 ; transpose coefficients(phase 2) + punpckldq xmm0,xmm3 ; xmm0=(40 50 60 70 41 51 61 71) + punpckhdq xmm2,xmm3 ; xmm2=(42 52 62 72 43 53 63 73) + movdqa xmm5,xmm6 ; transpose coefficients(phase 2) + punpckldq xmm6,xmm4 ; xmm6=(44 54 64 74 45 55 65 75) + punpckhdq xmm5,xmm4 ; xmm5=(46 56 66 76 47 57 67 77) + + movdqa xmm3,xmm7 ; transpose coefficients(phase 3) + punpcklqdq xmm7,xmm0 ; xmm7=col0=(00 10 20 30 40 50 60 70) + punpckhqdq xmm3,xmm0 ; xmm3=col1=(01 11 21 31 41 51 61 71) + movdqa xmm4,xmm1 ; transpose coefficients(phase 3) + punpcklqdq xmm1,xmm2 ; xmm1=col2=(02 12 22 32 42 52 62 72) + punpckhqdq xmm4,xmm2 ; xmm4=col3=(03 13 23 33 43 53 63 73) + + movdqa xmm0, XMMWORD [wk(6)] ; xmm0=(04 14 24 34 05 15 25 35) + movdqa xmm2, XMMWORD [wk(7)] ; xmm2=(06 16 26 36 07 17 27 37) + + movdqa XMMWORD [wk(8)], xmm3 ; wk(8)=col1 + movdqa XMMWORD [wk(9)], xmm4 ; wk(9)=col3 + + movdqa xmm3,xmm0 ; transpose coefficients(phase 3) + punpcklqdq xmm0,xmm6 ; xmm0=col4=(04 14 24 34 44 54 64 74) + punpckhqdq xmm3,xmm6 ; xmm3=col5=(05 15 25 35 45 55 65 75) + movdqa xmm4,xmm2 ; transpose coefficients(phase 3) + punpcklqdq xmm2,xmm5 ; xmm2=col6=(06 16 26 36 46 56 66 76) + punpckhqdq xmm4,xmm5 ; xmm4=col7=(07 17 27 37 47 57 67 77) + + movdqa XMMWORD [wk(10)], xmm3 ; wk(10)=col5 + movdqa XMMWORD [wk(11)], xmm4 ; wk(11)=col7 +.column_end: + + ; -- Prefetch the next coefficient block + + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32] + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32] + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32] + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32] + + ; ---- Pass 2: process rows from work array, store into output array. + + mov rax, [original_rbp] + mov rdi, r12 ; (JSAMPROW *) + mov rax, r13 + + ; -- Even part + + ; xmm7=col0, xmm1=col2, xmm0=col4, xmm2=col6 + + ; (Original) + ; z1 = (z2 + z3) * 0.541196100; + ; tmp2 = z1 + z3 * -1.847759065; + ; tmp3 = z1 + z2 * 0.765366865; + ; + ; (This implementation) + ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065); + ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100; + + movdqa xmm6,xmm1 ; xmm1=in2=z2 + movdqa xmm5,xmm1 + punpcklwd xmm6,xmm2 ; xmm2=in6=z3 + punpckhwd xmm5,xmm2 + movdqa xmm1,xmm6 + movdqa xmm2,xmm5 + pmaddwd xmm6,[rel PW_F130_F054] ; xmm6=tmp3L + pmaddwd xmm5,[rel PW_F130_F054] ; xmm5=tmp3H + pmaddwd xmm1,[rel PW_F054_MF130] ; xmm1=tmp2L + pmaddwd xmm2,[rel PW_F054_MF130] ; xmm2=tmp2H + + movdqa xmm3,xmm7 + paddw xmm7,xmm0 ; xmm7=in0+in4 + psubw xmm3,xmm0 ; xmm3=in0-in4 + + pxor xmm4,xmm4 + pxor xmm0,xmm0 + punpcklwd xmm4,xmm7 ; xmm4=tmp0L + punpckhwd xmm0,xmm7 ; xmm0=tmp0H + psrad xmm4,(16-CONST_BITS) ; psrad xmm4,16 & pslld xmm4,CONST_BITS + psrad xmm0,(16-CONST_BITS) ; psrad xmm0,16 & pslld xmm0,CONST_BITS + + movdqa xmm7,xmm4 + paddd xmm4,xmm6 ; xmm4=tmp10L + psubd xmm7,xmm6 ; xmm7=tmp13L + movdqa xmm6,xmm0 + paddd xmm0,xmm5 ; xmm0=tmp10H + psubd xmm6,xmm5 ; xmm6=tmp13H + + movdqa XMMWORD [wk(0)], xmm4 ; wk(0)=tmp10L + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=tmp10H + movdqa XMMWORD [wk(2)], xmm7 ; wk(2)=tmp13L + movdqa XMMWORD [wk(3)], xmm6 ; wk(3)=tmp13H + + pxor xmm5,xmm5 + pxor xmm4,xmm4 + punpcklwd xmm5,xmm3 ; xmm5=tmp1L + punpckhwd xmm4,xmm3 ; xmm4=tmp1H + psrad xmm5,(16-CONST_BITS) ; psrad xmm5,16 & pslld xmm5,CONST_BITS + psrad xmm4,(16-CONST_BITS) ; psrad xmm4,16 & pslld xmm4,CONST_BITS + + movdqa xmm0,xmm5 + paddd xmm5,xmm1 ; xmm5=tmp11L + psubd xmm0,xmm1 ; xmm0=tmp12L + movdqa xmm7,xmm4 + paddd xmm4,xmm2 ; xmm4=tmp11H + psubd xmm7,xmm2 ; xmm7=tmp12H + + movdqa XMMWORD [wk(4)], xmm5 ; wk(4)=tmp11L + movdqa XMMWORD [wk(5)], xmm4 ; wk(5)=tmp11H + movdqa XMMWORD [wk(6)], xmm0 ; wk(6)=tmp12L + movdqa XMMWORD [wk(7)], xmm7 ; wk(7)=tmp12H + + ; -- Odd part + + movdqa xmm6, XMMWORD [wk(9)] ; xmm6=col3 + movdqa xmm3, XMMWORD [wk(8)] ; xmm3=col1 + movdqa xmm1, XMMWORD [wk(11)] ; xmm1=col7 + movdqa xmm2, XMMWORD [wk(10)] ; xmm2=col5 + + movdqa xmm5,xmm6 + movdqa xmm4,xmm3 + paddw xmm5,xmm1 ; xmm5=z3 + paddw xmm4,xmm2 ; xmm4=z4 + + ; (Original) + ; z5 = (z3 + z4) * 1.175875602; + ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; + ; z3 += z5; z4 += z5; + ; + ; (This implementation) + ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; + ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); + + movdqa xmm0,xmm5 + movdqa xmm7,xmm5 + punpcklwd xmm0,xmm4 + punpckhwd xmm7,xmm4 + movdqa xmm5,xmm0 + movdqa xmm4,xmm7 + pmaddwd xmm0,[rel PW_MF078_F117] ; xmm0=z3L + pmaddwd xmm7,[rel PW_MF078_F117] ; xmm7=z3H + pmaddwd xmm5,[rel PW_F117_F078] ; xmm5=z4L + pmaddwd xmm4,[rel PW_F117_F078] ; xmm4=z4H + + movdqa XMMWORD [wk(10)], xmm0 ; wk(10)=z3L + movdqa XMMWORD [wk(11)], xmm7 ; wk(11)=z3H + + ; (Original) + ; z1 = tmp0 + tmp3; z2 = tmp1 + tmp2; + ; tmp0 = tmp0 * 0.298631336; tmp1 = tmp1 * 2.053119869; + ; tmp2 = tmp2 * 3.072711026; tmp3 = tmp3 * 1.501321110; + ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; + ; tmp0 += z1 + z3; tmp1 += z2 + z4; + ; tmp2 += z2 + z3; tmp3 += z1 + z4; + ; + ; (This implementation) + ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223; + ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447; + ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447); + ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223); + ; tmp0 += z3; tmp1 += z4; + ; tmp2 += z3; tmp3 += z4; + + movdqa xmm0,xmm1 + movdqa xmm7,xmm1 + punpcklwd xmm0,xmm3 + punpckhwd xmm7,xmm3 + movdqa xmm1,xmm0 + movdqa xmm3,xmm7 + pmaddwd xmm0,[rel PW_MF060_MF089] ; xmm0=tmp0L + pmaddwd xmm7,[rel PW_MF060_MF089] ; xmm7=tmp0H + pmaddwd xmm1,[rel PW_MF089_F060] ; xmm1=tmp3L + pmaddwd xmm3,[rel PW_MF089_F060] ; xmm3=tmp3H + + paddd xmm0, XMMWORD [wk(10)] ; xmm0=tmp0L + paddd xmm7, XMMWORD [wk(11)] ; xmm7=tmp0H + paddd xmm1,xmm5 ; xmm1=tmp3L + paddd xmm3,xmm4 ; xmm3=tmp3H + + movdqa XMMWORD [wk(8)], xmm0 ; wk(8)=tmp0L + movdqa XMMWORD [wk(9)], xmm7 ; wk(9)=tmp0H + + movdqa xmm0,xmm2 + movdqa xmm7,xmm2 + punpcklwd xmm0,xmm6 + punpckhwd xmm7,xmm6 + movdqa xmm2,xmm0 + movdqa xmm6,xmm7 + pmaddwd xmm0,[rel PW_MF050_MF256] ; xmm0=tmp1L + pmaddwd xmm7,[rel PW_MF050_MF256] ; xmm7=tmp1H + pmaddwd xmm2,[rel PW_MF256_F050] ; xmm2=tmp2L + pmaddwd xmm6,[rel PW_MF256_F050] ; xmm6=tmp2H + + paddd xmm0,xmm5 ; xmm0=tmp1L + paddd xmm7,xmm4 ; xmm7=tmp1H + paddd xmm2, XMMWORD [wk(10)] ; xmm2=tmp2L + paddd xmm6, XMMWORD [wk(11)] ; xmm6=tmp2H + + movdqa XMMWORD [wk(10)], xmm0 ; wk(10)=tmp1L + movdqa XMMWORD [wk(11)], xmm7 ; wk(11)=tmp1H + + ; -- Final output stage + + movdqa xmm5, XMMWORD [wk(0)] ; xmm5=tmp10L + movdqa xmm4, XMMWORD [wk(1)] ; xmm4=tmp10H + + movdqa xmm0,xmm5 + movdqa xmm7,xmm4 + paddd xmm5,xmm1 ; xmm5=data0L + paddd xmm4,xmm3 ; xmm4=data0H + psubd xmm0,xmm1 ; xmm0=data7L + psubd xmm7,xmm3 ; xmm7=data7H + + movdqa xmm1,[rel PD_DESCALE_P2] ; xmm1=[rel PD_DESCALE_P2] + + paddd xmm5,xmm1 + paddd xmm4,xmm1 + psrad xmm5,DESCALE_P2 + psrad xmm4,DESCALE_P2 + paddd xmm0,xmm1 + paddd xmm7,xmm1 + psrad xmm0,DESCALE_P2 + psrad xmm7,DESCALE_P2 + + packssdw xmm5,xmm4 ; xmm5=data0=(00 10 20 30 40 50 60 70) + packssdw xmm0,xmm7 ; xmm0=data7=(07 17 27 37 47 57 67 77) + + movdqa xmm3, XMMWORD [wk(4)] ; xmm3=tmp11L + movdqa xmm1, XMMWORD [wk(5)] ; xmm1=tmp11H + + movdqa xmm4,xmm3 + movdqa xmm7,xmm1 + paddd xmm3,xmm2 ; xmm3=data1L + paddd xmm1,xmm6 ; xmm1=data1H + psubd xmm4,xmm2 ; xmm4=data6L + psubd xmm7,xmm6 ; xmm7=data6H + + movdqa xmm2,[rel PD_DESCALE_P2] ; xmm2=[rel PD_DESCALE_P2] + + paddd xmm3,xmm2 + paddd xmm1,xmm2 + psrad xmm3,DESCALE_P2 + psrad xmm1,DESCALE_P2 + paddd xmm4,xmm2 + paddd xmm7,xmm2 + psrad xmm4,DESCALE_P2 + psrad xmm7,DESCALE_P2 + + packssdw xmm3,xmm1 ; xmm3=data1=(01 11 21 31 41 51 61 71) + packssdw xmm4,xmm7 ; xmm4=data6=(06 16 26 36 46 56 66 76) + + packsswb xmm5,xmm4 ; xmm5=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76) + packsswb xmm3,xmm0 ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77) + + movdqa xmm6, XMMWORD [wk(6)] ; xmm6=tmp12L + movdqa xmm2, XMMWORD [wk(7)] ; xmm2=tmp12H + movdqa xmm1, XMMWORD [wk(10)] ; xmm1=tmp1L + movdqa xmm7, XMMWORD [wk(11)] ; xmm7=tmp1H + + movdqa XMMWORD [wk(0)], xmm5 ; wk(0)=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76) + movdqa XMMWORD [wk(1)], xmm3 ; wk(1)=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77) + + movdqa xmm4,xmm6 + movdqa xmm0,xmm2 + paddd xmm6,xmm1 ; xmm6=data2L + paddd xmm2,xmm7 ; xmm2=data2H + psubd xmm4,xmm1 ; xmm4=data5L + psubd xmm0,xmm7 ; xmm0=data5H + + movdqa xmm5,[rel PD_DESCALE_P2] ; xmm5=[rel PD_DESCALE_P2] + + paddd xmm6,xmm5 + paddd xmm2,xmm5 + psrad xmm6,DESCALE_P2 + psrad xmm2,DESCALE_P2 + paddd xmm4,xmm5 + paddd xmm0,xmm5 + psrad xmm4,DESCALE_P2 + psrad xmm0,DESCALE_P2 + + packssdw xmm6,xmm2 ; xmm6=data2=(02 12 22 32 42 52 62 72) + packssdw xmm4,xmm0 ; xmm4=data5=(05 15 25 35 45 55 65 75) + + movdqa xmm3, XMMWORD [wk(2)] ; xmm3=tmp13L + movdqa xmm1, XMMWORD [wk(3)] ; xmm1=tmp13H + movdqa xmm7, XMMWORD [wk(8)] ; xmm7=tmp0L + movdqa xmm5, XMMWORD [wk(9)] ; xmm5=tmp0H + + movdqa xmm2,xmm3 + movdqa xmm0,xmm1 + paddd xmm3,xmm7 ; xmm3=data3L + paddd xmm1,xmm5 ; xmm1=data3H + psubd xmm2,xmm7 ; xmm2=data4L + psubd xmm0,xmm5 ; xmm0=data4H + + movdqa xmm7,[rel PD_DESCALE_P2] ; xmm7=[rel PD_DESCALE_P2] + + paddd xmm3,xmm7 + paddd xmm1,xmm7 + psrad xmm3,DESCALE_P2 + psrad xmm1,DESCALE_P2 + paddd xmm2,xmm7 + paddd xmm0,xmm7 + psrad xmm2,DESCALE_P2 + psrad xmm0,DESCALE_P2 + + movdqa xmm5,[rel PB_CENTERJSAMP] ; xmm5=[rel PB_CENTERJSAMP] + + packssdw xmm3,xmm1 ; xmm3=data3=(03 13 23 33 43 53 63 73) + packssdw xmm2,xmm0 ; xmm2=data4=(04 14 24 34 44 54 64 74) + + movdqa xmm7, XMMWORD [wk(0)] ; xmm7=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76) + movdqa xmm1, XMMWORD [wk(1)] ; xmm1=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77) + + packsswb xmm6,xmm2 ; xmm6=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74) + packsswb xmm3,xmm4 ; xmm3=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75) + + paddb xmm7,xmm5 + paddb xmm1,xmm5 + paddb xmm6,xmm5 + paddb xmm3,xmm5 + + movdqa xmm0,xmm7 ; transpose coefficients(phase 1) + punpcklbw xmm7,xmm1 ; xmm7=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71) + punpckhbw xmm0,xmm1 ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77) + movdqa xmm2,xmm6 ; transpose coefficients(phase 1) + punpcklbw xmm6,xmm3 ; xmm6=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73) + punpckhbw xmm2,xmm3 ; xmm2=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75) + + movdqa xmm4,xmm7 ; transpose coefficients(phase 2) + punpcklwd xmm7,xmm6 ; xmm7=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33) + punpckhwd xmm4,xmm6 ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73) + movdqa xmm5,xmm2 ; transpose coefficients(phase 2) + punpcklwd xmm2,xmm0 ; xmm2=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37) + punpckhwd xmm5,xmm0 ; xmm5=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77) + + movdqa xmm1,xmm7 ; transpose coefficients(phase 3) + punpckldq xmm7,xmm2 ; xmm7=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17) + punpckhdq xmm1,xmm2 ; xmm1=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37) + movdqa xmm3,xmm4 ; transpose coefficients(phase 3) + punpckldq xmm4,xmm5 ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57) + punpckhdq xmm3,xmm5 ; xmm3=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77) + + pshufd xmm6,xmm7,0x4E ; xmm6=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07) + pshufd xmm0,xmm1,0x4E ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27) + pshufd xmm2,xmm4,0x4E ; xmm2=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47) + pshufd xmm5,xmm3,0x4E ; xmm5=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67) + + mov rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW] + mov rsi, JSAMPROW [rdi+2*SIZEOF_JSAMPROW] + movq XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm7 + movq XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm1 + mov rdx, JSAMPROW [rdi+4*SIZEOF_JSAMPROW] + mov rsi, JSAMPROW [rdi+6*SIZEOF_JSAMPROW] + movq XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm4 + movq XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm3 + + mov rdx, JSAMPROW [rdi+1*SIZEOF_JSAMPROW] + mov rsi, JSAMPROW [rdi+3*SIZEOF_JSAMPROW] + movq XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm6 + movq XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm0 + mov rdx, JSAMPROW [rdi+5*SIZEOF_JSAMPROW] + mov rsi, JSAMPROW [rdi+7*SIZEOF_JSAMPROW] + movq XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm2 + movq XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm5 + + uncollect_args + mov rsp,rbp ; rsp <- aligned rbp + pop rsp ; rsp <- original rbp + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jidctint-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jidctint-sse2.asm new file mode 100644 index 000000000..4a35f3dff --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jidctint-sse2.asm @@ -0,0 +1,859 @@ +; +; jidctint.asm - accurate integer IDCT (SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains a slow-but-accurate integer implementation of the +; inverse DCT (Discrete Cosine Transform). The following code is based +; directly on the IJG's original jidctint.c; see the jidctint.c for +; more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 13 +%define PASS1_BITS 2 + +%define DESCALE_P1 (CONST_BITS-PASS1_BITS) +%define DESCALE_P2 (CONST_BITS+PASS1_BITS+3) + +%if CONST_BITS == 13 +F_0_298 equ 2446 ; FIX(0.298631336) +F_0_390 equ 3196 ; FIX(0.390180644) +F_0_541 equ 4433 ; FIX(0.541196100) +F_0_765 equ 6270 ; FIX(0.765366865) +F_0_899 equ 7373 ; FIX(0.899976223) +F_1_175 equ 9633 ; FIX(1.175875602) +F_1_501 equ 12299 ; FIX(1.501321110) +F_1_847 equ 15137 ; FIX(1.847759065) +F_1_961 equ 16069 ; FIX(1.961570560) +F_2_053 equ 16819 ; FIX(2.053119869) +F_2_562 equ 20995 ; FIX(2.562915447) +F_3_072 equ 25172 ; FIX(3.072711026) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_0_298 equ DESCALE( 320652955,30-CONST_BITS) ; FIX(0.298631336) +F_0_390 equ DESCALE( 418953276,30-CONST_BITS) ; FIX(0.390180644) +F_0_541 equ DESCALE( 581104887,30-CONST_BITS) ; FIX(0.541196100) +F_0_765 equ DESCALE( 821806413,30-CONST_BITS) ; FIX(0.765366865) +F_0_899 equ DESCALE( 966342111,30-CONST_BITS) ; FIX(0.899976223) +F_1_175 equ DESCALE(1262586813,30-CONST_BITS) ; FIX(1.175875602) +F_1_501 equ DESCALE(1612031267,30-CONST_BITS) ; FIX(1.501321110) +F_1_847 equ DESCALE(1984016188,30-CONST_BITS) ; FIX(1.847759065) +F_1_961 equ DESCALE(2106220350,30-CONST_BITS) ; FIX(1.961570560) +F_2_053 equ DESCALE(2204520673,30-CONST_BITS) ; FIX(2.053119869) +F_2_562 equ DESCALE(2751909506,30-CONST_BITS) ; FIX(2.562915447) +F_3_072 equ DESCALE(3299298341,30-CONST_BITS) ; FIX(3.072711026) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_idct_islow_sse2) + +EXTN(jconst_idct_islow_sse2): + +PW_F130_F054 times 4 dw (F_0_541+F_0_765), F_0_541 +PW_F054_MF130 times 4 dw F_0_541, (F_0_541-F_1_847) +PW_MF078_F117 times 4 dw (F_1_175-F_1_961), F_1_175 +PW_F117_F078 times 4 dw F_1_175, (F_1_175-F_0_390) +PW_MF060_MF089 times 4 dw (F_0_298-F_0_899),-F_0_899 +PW_MF089_F060 times 4 dw -F_0_899, (F_1_501-F_0_899) +PW_MF050_MF256 times 4 dw (F_2_053-F_2_562),-F_2_562 +PW_MF256_F050 times 4 dw -F_2_562, (F_3_072-F_2_562) +PD_DESCALE_P1 times 4 dd 1 << (DESCALE_P1-1) +PD_DESCALE_P2 times 4 dd 1 << (DESCALE_P2-1) +PB_CENTERJSAMP times 16 db CENTERJSAMPLE + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform dequantization and inverse DCT on one block of coefficients. +; +; GLOBAL(void) +; jsimd_idct_islow_sse2 (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +%define dct_table(b) (b)+8 ; jpeg_component_info * compptr +%define coef_block(b) (b)+12 ; JCOEFPTR coef_block +%define output_buf(b) (b)+16 ; JSAMPARRAY output_buf +%define output_col(b) (b)+20 ; JDIMENSION output_col + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 12 + + align 16 + global EXTN(jsimd_idct_islow_sse2) + +EXTN(jsimd_idct_islow_sse2): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic ebx +; push ecx ; unused +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process columns from input. + +; mov eax, [original_ebp] + mov edx, POINTER [dct_table(eax)] ; quantptr + mov esi, JCOEFPTR [coef_block(eax)] ; inptr + +%ifndef NO_ZERO_COLUMN_TEST_ISLOW_SSE2 + mov eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)] + jnz near .columnDCT + + movdqa xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)] + por xmm1, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)] + por xmm1, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)] + por xmm1,xmm0 + packsswb xmm1,xmm1 + packsswb xmm1,xmm1 + movd eax,xmm1 + test eax,eax + jnz short .columnDCT + + ; -- AC terms all zero + + movdqa xmm5, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)] + pmullw xmm5, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + psllw xmm5,PASS1_BITS + + movdqa xmm4,xmm5 ; xmm5=in0=(00 01 02 03 04 05 06 07) + punpcklwd xmm5,xmm5 ; xmm5=(00 00 01 01 02 02 03 03) + punpckhwd xmm4,xmm4 ; xmm4=(04 04 05 05 06 06 07 07) + + pshufd xmm7,xmm5,0x00 ; xmm7=col0=(00 00 00 00 00 00 00 00) + pshufd xmm6,xmm5,0x55 ; xmm6=col1=(01 01 01 01 01 01 01 01) + pshufd xmm1,xmm5,0xAA ; xmm1=col2=(02 02 02 02 02 02 02 02) + pshufd xmm5,xmm5,0xFF ; xmm5=col3=(03 03 03 03 03 03 03 03) + pshufd xmm0,xmm4,0x00 ; xmm0=col4=(04 04 04 04 04 04 04 04) + pshufd xmm3,xmm4,0x55 ; xmm3=col5=(05 05 05 05 05 05 05 05) + pshufd xmm2,xmm4,0xAA ; xmm2=col6=(06 06 06 06 06 06 06 06) + pshufd xmm4,xmm4,0xFF ; xmm4=col7=(07 07 07 07 07 07 07 07) + + movdqa XMMWORD [wk(8)], xmm6 ; wk(8)=col1 + movdqa XMMWORD [wk(9)], xmm5 ; wk(9)=col3 + movdqa XMMWORD [wk(10)], xmm3 ; wk(10)=col5 + movdqa XMMWORD [wk(11)], xmm4 ; wk(11)=col7 + jmp near .column_end + alignx 16,7 +%endif +.columnDCT: + + ; -- Even part + + movdqa xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)] + pmullw xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm1, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + movdqa xmm2, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_JCOEF)] + movdqa xmm3, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)] + pmullw xmm2, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm3, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + ; (Original) + ; z1 = (z2 + z3) * 0.541196100; + ; tmp2 = z1 + z3 * -1.847759065; + ; tmp3 = z1 + z2 * 0.765366865; + ; + ; (This implementation) + ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065); + ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100; + + movdqa xmm4,xmm1 ; xmm1=in2=z2 + movdqa xmm5,xmm1 + punpcklwd xmm4,xmm3 ; xmm3=in6=z3 + punpckhwd xmm5,xmm3 + movdqa xmm1,xmm4 + movdqa xmm3,xmm5 + pmaddwd xmm4,[GOTOFF(ebx,PW_F130_F054)] ; xmm4=tmp3L + pmaddwd xmm5,[GOTOFF(ebx,PW_F130_F054)] ; xmm5=tmp3H + pmaddwd xmm1,[GOTOFF(ebx,PW_F054_MF130)] ; xmm1=tmp2L + pmaddwd xmm3,[GOTOFF(ebx,PW_F054_MF130)] ; xmm3=tmp2H + + movdqa xmm6,xmm0 + paddw xmm0,xmm2 ; xmm0=in0+in4 + psubw xmm6,xmm2 ; xmm6=in0-in4 + + pxor xmm7,xmm7 + pxor xmm2,xmm2 + punpcklwd xmm7,xmm0 ; xmm7=tmp0L + punpckhwd xmm2,xmm0 ; xmm2=tmp0H + psrad xmm7,(16-CONST_BITS) ; psrad xmm7,16 & pslld xmm7,CONST_BITS + psrad xmm2,(16-CONST_BITS) ; psrad xmm2,16 & pslld xmm2,CONST_BITS + + movdqa xmm0,xmm7 + paddd xmm7,xmm4 ; xmm7=tmp10L + psubd xmm0,xmm4 ; xmm0=tmp13L + movdqa xmm4,xmm2 + paddd xmm2,xmm5 ; xmm2=tmp10H + psubd xmm4,xmm5 ; xmm4=tmp13H + + movdqa XMMWORD [wk(0)], xmm7 ; wk(0)=tmp10L + movdqa XMMWORD [wk(1)], xmm2 ; wk(1)=tmp10H + movdqa XMMWORD [wk(2)], xmm0 ; wk(2)=tmp13L + movdqa XMMWORD [wk(3)], xmm4 ; wk(3)=tmp13H + + pxor xmm5,xmm5 + pxor xmm7,xmm7 + punpcklwd xmm5,xmm6 ; xmm5=tmp1L + punpckhwd xmm7,xmm6 ; xmm7=tmp1H + psrad xmm5,(16-CONST_BITS) ; psrad xmm5,16 & pslld xmm5,CONST_BITS + psrad xmm7,(16-CONST_BITS) ; psrad xmm7,16 & pslld xmm7,CONST_BITS + + movdqa xmm2,xmm5 + paddd xmm5,xmm1 ; xmm5=tmp11L + psubd xmm2,xmm1 ; xmm2=tmp12L + movdqa xmm0,xmm7 + paddd xmm7,xmm3 ; xmm7=tmp11H + psubd xmm0,xmm3 ; xmm0=tmp12H + + movdqa XMMWORD [wk(4)], xmm5 ; wk(4)=tmp11L + movdqa XMMWORD [wk(5)], xmm7 ; wk(5)=tmp11H + movdqa XMMWORD [wk(6)], xmm2 ; wk(6)=tmp12L + movdqa XMMWORD [wk(7)], xmm0 ; wk(7)=tmp12H + + ; -- Odd part + + movdqa xmm4, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movdqa xmm6, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)] + pmullw xmm4, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm6, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + movdqa xmm1, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movdqa xmm3, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)] + pmullw xmm1, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + movdqa xmm5,xmm6 + movdqa xmm7,xmm4 + paddw xmm5,xmm3 ; xmm5=z3 + paddw xmm7,xmm1 ; xmm7=z4 + + ; (Original) + ; z5 = (z3 + z4) * 1.175875602; + ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; + ; z3 += z5; z4 += z5; + ; + ; (This implementation) + ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; + ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); + + movdqa xmm2,xmm5 + movdqa xmm0,xmm5 + punpcklwd xmm2,xmm7 + punpckhwd xmm0,xmm7 + movdqa xmm5,xmm2 + movdqa xmm7,xmm0 + pmaddwd xmm2,[GOTOFF(ebx,PW_MF078_F117)] ; xmm2=z3L + pmaddwd xmm0,[GOTOFF(ebx,PW_MF078_F117)] ; xmm0=z3H + pmaddwd xmm5,[GOTOFF(ebx,PW_F117_F078)] ; xmm5=z4L + pmaddwd xmm7,[GOTOFF(ebx,PW_F117_F078)] ; xmm7=z4H + + movdqa XMMWORD [wk(10)], xmm2 ; wk(10)=z3L + movdqa XMMWORD [wk(11)], xmm0 ; wk(11)=z3H + + ; (Original) + ; z1 = tmp0 + tmp3; z2 = tmp1 + tmp2; + ; tmp0 = tmp0 * 0.298631336; tmp1 = tmp1 * 2.053119869; + ; tmp2 = tmp2 * 3.072711026; tmp3 = tmp3 * 1.501321110; + ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; + ; tmp0 += z1 + z3; tmp1 += z2 + z4; + ; tmp2 += z2 + z3; tmp3 += z1 + z4; + ; + ; (This implementation) + ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223; + ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447; + ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447); + ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223); + ; tmp0 += z3; tmp1 += z4; + ; tmp2 += z3; tmp3 += z4; + + movdqa xmm2,xmm3 + movdqa xmm0,xmm3 + punpcklwd xmm2,xmm4 + punpckhwd xmm0,xmm4 + movdqa xmm3,xmm2 + movdqa xmm4,xmm0 + pmaddwd xmm2,[GOTOFF(ebx,PW_MF060_MF089)] ; xmm2=tmp0L + pmaddwd xmm0,[GOTOFF(ebx,PW_MF060_MF089)] ; xmm0=tmp0H + pmaddwd xmm3,[GOTOFF(ebx,PW_MF089_F060)] ; xmm3=tmp3L + pmaddwd xmm4,[GOTOFF(ebx,PW_MF089_F060)] ; xmm4=tmp3H + + paddd xmm2, XMMWORD [wk(10)] ; xmm2=tmp0L + paddd xmm0, XMMWORD [wk(11)] ; xmm0=tmp0H + paddd xmm3,xmm5 ; xmm3=tmp3L + paddd xmm4,xmm7 ; xmm4=tmp3H + + movdqa XMMWORD [wk(8)], xmm2 ; wk(8)=tmp0L + movdqa XMMWORD [wk(9)], xmm0 ; wk(9)=tmp0H + + movdqa xmm2,xmm1 + movdqa xmm0,xmm1 + punpcklwd xmm2,xmm6 + punpckhwd xmm0,xmm6 + movdqa xmm1,xmm2 + movdqa xmm6,xmm0 + pmaddwd xmm2,[GOTOFF(ebx,PW_MF050_MF256)] ; xmm2=tmp1L + pmaddwd xmm0,[GOTOFF(ebx,PW_MF050_MF256)] ; xmm0=tmp1H + pmaddwd xmm1,[GOTOFF(ebx,PW_MF256_F050)] ; xmm1=tmp2L + pmaddwd xmm6,[GOTOFF(ebx,PW_MF256_F050)] ; xmm6=tmp2H + + paddd xmm2,xmm5 ; xmm2=tmp1L + paddd xmm0,xmm7 ; xmm0=tmp1H + paddd xmm1, XMMWORD [wk(10)] ; xmm1=tmp2L + paddd xmm6, XMMWORD [wk(11)] ; xmm6=tmp2H + + movdqa XMMWORD [wk(10)], xmm2 ; wk(10)=tmp1L + movdqa XMMWORD [wk(11)], xmm0 ; wk(11)=tmp1H + + ; -- Final output stage + + movdqa xmm5, XMMWORD [wk(0)] ; xmm5=tmp10L + movdqa xmm7, XMMWORD [wk(1)] ; xmm7=tmp10H + + movdqa xmm2,xmm5 + movdqa xmm0,xmm7 + paddd xmm5,xmm3 ; xmm5=data0L + paddd xmm7,xmm4 ; xmm7=data0H + psubd xmm2,xmm3 ; xmm2=data7L + psubd xmm0,xmm4 ; xmm0=data7H + + movdqa xmm3,[GOTOFF(ebx,PD_DESCALE_P1)] ; xmm3=[PD_DESCALE_P1] + + paddd xmm5,xmm3 + paddd xmm7,xmm3 + psrad xmm5,DESCALE_P1 + psrad xmm7,DESCALE_P1 + paddd xmm2,xmm3 + paddd xmm0,xmm3 + psrad xmm2,DESCALE_P1 + psrad xmm0,DESCALE_P1 + + packssdw xmm5,xmm7 ; xmm5=data0=(00 01 02 03 04 05 06 07) + packssdw xmm2,xmm0 ; xmm2=data7=(70 71 72 73 74 75 76 77) + + movdqa xmm4, XMMWORD [wk(4)] ; xmm4=tmp11L + movdqa xmm3, XMMWORD [wk(5)] ; xmm3=tmp11H + + movdqa xmm7,xmm4 + movdqa xmm0,xmm3 + paddd xmm4,xmm1 ; xmm4=data1L + paddd xmm3,xmm6 ; xmm3=data1H + psubd xmm7,xmm1 ; xmm7=data6L + psubd xmm0,xmm6 ; xmm0=data6H + + movdqa xmm1,[GOTOFF(ebx,PD_DESCALE_P1)] ; xmm1=[PD_DESCALE_P1] + + paddd xmm4,xmm1 + paddd xmm3,xmm1 + psrad xmm4,DESCALE_P1 + psrad xmm3,DESCALE_P1 + paddd xmm7,xmm1 + paddd xmm0,xmm1 + psrad xmm7,DESCALE_P1 + psrad xmm0,DESCALE_P1 + + packssdw xmm4,xmm3 ; xmm4=data1=(10 11 12 13 14 15 16 17) + packssdw xmm7,xmm0 ; xmm7=data6=(60 61 62 63 64 65 66 67) + + movdqa xmm6,xmm5 ; transpose coefficients(phase 1) + punpcklwd xmm5,xmm4 ; xmm5=(00 10 01 11 02 12 03 13) + punpckhwd xmm6,xmm4 ; xmm6=(04 14 05 15 06 16 07 17) + movdqa xmm1,xmm7 ; transpose coefficients(phase 1) + punpcklwd xmm7,xmm2 ; xmm7=(60 70 61 71 62 72 63 73) + punpckhwd xmm1,xmm2 ; xmm1=(64 74 65 75 66 76 67 77) + + movdqa xmm3, XMMWORD [wk(6)] ; xmm3=tmp12L + movdqa xmm0, XMMWORD [wk(7)] ; xmm0=tmp12H + movdqa xmm4, XMMWORD [wk(10)] ; xmm4=tmp1L + movdqa xmm2, XMMWORD [wk(11)] ; xmm2=tmp1H + + movdqa XMMWORD [wk(0)], xmm5 ; wk(0)=(00 10 01 11 02 12 03 13) + movdqa XMMWORD [wk(1)], xmm6 ; wk(1)=(04 14 05 15 06 16 07 17) + movdqa XMMWORD [wk(4)], xmm7 ; wk(4)=(60 70 61 71 62 72 63 73) + movdqa XMMWORD [wk(5)], xmm1 ; wk(5)=(64 74 65 75 66 76 67 77) + + movdqa xmm5,xmm3 + movdqa xmm6,xmm0 + paddd xmm3,xmm4 ; xmm3=data2L + paddd xmm0,xmm2 ; xmm0=data2H + psubd xmm5,xmm4 ; xmm5=data5L + psubd xmm6,xmm2 ; xmm6=data5H + + movdqa xmm7,[GOTOFF(ebx,PD_DESCALE_P1)] ; xmm7=[PD_DESCALE_P1] + + paddd xmm3,xmm7 + paddd xmm0,xmm7 + psrad xmm3,DESCALE_P1 + psrad xmm0,DESCALE_P1 + paddd xmm5,xmm7 + paddd xmm6,xmm7 + psrad xmm5,DESCALE_P1 + psrad xmm6,DESCALE_P1 + + packssdw xmm3,xmm0 ; xmm3=data2=(20 21 22 23 24 25 26 27) + packssdw xmm5,xmm6 ; xmm5=data5=(50 51 52 53 54 55 56 57) + + movdqa xmm1, XMMWORD [wk(2)] ; xmm1=tmp13L + movdqa xmm4, XMMWORD [wk(3)] ; xmm4=tmp13H + movdqa xmm2, XMMWORD [wk(8)] ; xmm2=tmp0L + movdqa xmm7, XMMWORD [wk(9)] ; xmm7=tmp0H + + movdqa xmm0,xmm1 + movdqa xmm6,xmm4 + paddd xmm1,xmm2 ; xmm1=data3L + paddd xmm4,xmm7 ; xmm4=data3H + psubd xmm0,xmm2 ; xmm0=data4L + psubd xmm6,xmm7 ; xmm6=data4H + + movdqa xmm2,[GOTOFF(ebx,PD_DESCALE_P1)] ; xmm2=[PD_DESCALE_P1] + + paddd xmm1,xmm2 + paddd xmm4,xmm2 + psrad xmm1,DESCALE_P1 + psrad xmm4,DESCALE_P1 + paddd xmm0,xmm2 + paddd xmm6,xmm2 + psrad xmm0,DESCALE_P1 + psrad xmm6,DESCALE_P1 + + packssdw xmm1,xmm4 ; xmm1=data3=(30 31 32 33 34 35 36 37) + packssdw xmm0,xmm6 ; xmm0=data4=(40 41 42 43 44 45 46 47) + + movdqa xmm7, XMMWORD [wk(0)] ; xmm7=(00 10 01 11 02 12 03 13) + movdqa xmm2, XMMWORD [wk(1)] ; xmm2=(04 14 05 15 06 16 07 17) + + movdqa xmm4,xmm3 ; transpose coefficients(phase 1) + punpcklwd xmm3,xmm1 ; xmm3=(20 30 21 31 22 32 23 33) + punpckhwd xmm4,xmm1 ; xmm4=(24 34 25 35 26 36 27 37) + movdqa xmm6,xmm0 ; transpose coefficients(phase 1) + punpcklwd xmm0,xmm5 ; xmm0=(40 50 41 51 42 52 43 53) + punpckhwd xmm6,xmm5 ; xmm6=(44 54 45 55 46 56 47 57) + + movdqa xmm1,xmm7 ; transpose coefficients(phase 2) + punpckldq xmm7,xmm3 ; xmm7=(00 10 20 30 01 11 21 31) + punpckhdq xmm1,xmm3 ; xmm1=(02 12 22 32 03 13 23 33) + movdqa xmm5,xmm2 ; transpose coefficients(phase 2) + punpckldq xmm2,xmm4 ; xmm2=(04 14 24 34 05 15 25 35) + punpckhdq xmm5,xmm4 ; xmm5=(06 16 26 36 07 17 27 37) + + movdqa xmm3, XMMWORD [wk(4)] ; xmm3=(60 70 61 71 62 72 63 73) + movdqa xmm4, XMMWORD [wk(5)] ; xmm4=(64 74 65 75 66 76 67 77) + + movdqa XMMWORD [wk(6)], xmm2 ; wk(6)=(04 14 24 34 05 15 25 35) + movdqa XMMWORD [wk(7)], xmm5 ; wk(7)=(06 16 26 36 07 17 27 37) + + movdqa xmm2,xmm0 ; transpose coefficients(phase 2) + punpckldq xmm0,xmm3 ; xmm0=(40 50 60 70 41 51 61 71) + punpckhdq xmm2,xmm3 ; xmm2=(42 52 62 72 43 53 63 73) + movdqa xmm5,xmm6 ; transpose coefficients(phase 2) + punpckldq xmm6,xmm4 ; xmm6=(44 54 64 74 45 55 65 75) + punpckhdq xmm5,xmm4 ; xmm5=(46 56 66 76 47 57 67 77) + + movdqa xmm3,xmm7 ; transpose coefficients(phase 3) + punpcklqdq xmm7,xmm0 ; xmm7=col0=(00 10 20 30 40 50 60 70) + punpckhqdq xmm3,xmm0 ; xmm3=col1=(01 11 21 31 41 51 61 71) + movdqa xmm4,xmm1 ; transpose coefficients(phase 3) + punpcklqdq xmm1,xmm2 ; xmm1=col2=(02 12 22 32 42 52 62 72) + punpckhqdq xmm4,xmm2 ; xmm4=col3=(03 13 23 33 43 53 63 73) + + movdqa xmm0, XMMWORD [wk(6)] ; xmm0=(04 14 24 34 05 15 25 35) + movdqa xmm2, XMMWORD [wk(7)] ; xmm2=(06 16 26 36 07 17 27 37) + + movdqa XMMWORD [wk(8)], xmm3 ; wk(8)=col1 + movdqa XMMWORD [wk(9)], xmm4 ; wk(9)=col3 + + movdqa xmm3,xmm0 ; transpose coefficients(phase 3) + punpcklqdq xmm0,xmm6 ; xmm0=col4=(04 14 24 34 44 54 64 74) + punpckhqdq xmm3,xmm6 ; xmm3=col5=(05 15 25 35 45 55 65 75) + movdqa xmm4,xmm2 ; transpose coefficients(phase 3) + punpcklqdq xmm2,xmm5 ; xmm2=col6=(06 16 26 36 46 56 66 76) + punpckhqdq xmm4,xmm5 ; xmm4=col7=(07 17 27 37 47 57 67 77) + + movdqa XMMWORD [wk(10)], xmm3 ; wk(10)=col5 + movdqa XMMWORD [wk(11)], xmm4 ; wk(11)=col7 +.column_end: + + ; -- Prefetch the next coefficient block + + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 0*32] + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 1*32] + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 2*32] + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 3*32] + + ; ---- Pass 2: process rows from work array, store into output array. + + mov eax, [original_ebp] + mov edi, JSAMPARRAY [output_buf(eax)] ; (JSAMPROW *) + mov eax, JDIMENSION [output_col(eax)] + + ; -- Even part + + ; xmm7=col0, xmm1=col2, xmm0=col4, xmm2=col6 + + ; (Original) + ; z1 = (z2 + z3) * 0.541196100; + ; tmp2 = z1 + z3 * -1.847759065; + ; tmp3 = z1 + z2 * 0.765366865; + ; + ; (This implementation) + ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065); + ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100; + + movdqa xmm6,xmm1 ; xmm1=in2=z2 + movdqa xmm5,xmm1 + punpcklwd xmm6,xmm2 ; xmm2=in6=z3 + punpckhwd xmm5,xmm2 + movdqa xmm1,xmm6 + movdqa xmm2,xmm5 + pmaddwd xmm6,[GOTOFF(ebx,PW_F130_F054)] ; xmm6=tmp3L + pmaddwd xmm5,[GOTOFF(ebx,PW_F130_F054)] ; xmm5=tmp3H + pmaddwd xmm1,[GOTOFF(ebx,PW_F054_MF130)] ; xmm1=tmp2L + pmaddwd xmm2,[GOTOFF(ebx,PW_F054_MF130)] ; xmm2=tmp2H + + movdqa xmm3,xmm7 + paddw xmm7,xmm0 ; xmm7=in0+in4 + psubw xmm3,xmm0 ; xmm3=in0-in4 + + pxor xmm4,xmm4 + pxor xmm0,xmm0 + punpcklwd xmm4,xmm7 ; xmm4=tmp0L + punpckhwd xmm0,xmm7 ; xmm0=tmp0H + psrad xmm4,(16-CONST_BITS) ; psrad xmm4,16 & pslld xmm4,CONST_BITS + psrad xmm0,(16-CONST_BITS) ; psrad xmm0,16 & pslld xmm0,CONST_BITS + + movdqa xmm7,xmm4 + paddd xmm4,xmm6 ; xmm4=tmp10L + psubd xmm7,xmm6 ; xmm7=tmp13L + movdqa xmm6,xmm0 + paddd xmm0,xmm5 ; xmm0=tmp10H + psubd xmm6,xmm5 ; xmm6=tmp13H + + movdqa XMMWORD [wk(0)], xmm4 ; wk(0)=tmp10L + movdqa XMMWORD [wk(1)], xmm0 ; wk(1)=tmp10H + movdqa XMMWORD [wk(2)], xmm7 ; wk(2)=tmp13L + movdqa XMMWORD [wk(3)], xmm6 ; wk(3)=tmp13H + + pxor xmm5,xmm5 + pxor xmm4,xmm4 + punpcklwd xmm5,xmm3 ; xmm5=tmp1L + punpckhwd xmm4,xmm3 ; xmm4=tmp1H + psrad xmm5,(16-CONST_BITS) ; psrad xmm5,16 & pslld xmm5,CONST_BITS + psrad xmm4,(16-CONST_BITS) ; psrad xmm4,16 & pslld xmm4,CONST_BITS + + movdqa xmm0,xmm5 + paddd xmm5,xmm1 ; xmm5=tmp11L + psubd xmm0,xmm1 ; xmm0=tmp12L + movdqa xmm7,xmm4 + paddd xmm4,xmm2 ; xmm4=tmp11H + psubd xmm7,xmm2 ; xmm7=tmp12H + + movdqa XMMWORD [wk(4)], xmm5 ; wk(4)=tmp11L + movdqa XMMWORD [wk(5)], xmm4 ; wk(5)=tmp11H + movdqa XMMWORD [wk(6)], xmm0 ; wk(6)=tmp12L + movdqa XMMWORD [wk(7)], xmm7 ; wk(7)=tmp12H + + ; -- Odd part + + movdqa xmm6, XMMWORD [wk(9)] ; xmm6=col3 + movdqa xmm3, XMMWORD [wk(8)] ; xmm3=col1 + movdqa xmm1, XMMWORD [wk(11)] ; xmm1=col7 + movdqa xmm2, XMMWORD [wk(10)] ; xmm2=col5 + + movdqa xmm5,xmm6 + movdqa xmm4,xmm3 + paddw xmm5,xmm1 ; xmm5=z3 + paddw xmm4,xmm2 ; xmm4=z4 + + ; (Original) + ; z5 = (z3 + z4) * 1.175875602; + ; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; + ; z3 += z5; z4 += z5; + ; + ; (This implementation) + ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; + ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); + + movdqa xmm0,xmm5 + movdqa xmm7,xmm5 + punpcklwd xmm0,xmm4 + punpckhwd xmm7,xmm4 + movdqa xmm5,xmm0 + movdqa xmm4,xmm7 + pmaddwd xmm0,[GOTOFF(ebx,PW_MF078_F117)] ; xmm0=z3L + pmaddwd xmm7,[GOTOFF(ebx,PW_MF078_F117)] ; xmm7=z3H + pmaddwd xmm5,[GOTOFF(ebx,PW_F117_F078)] ; xmm5=z4L + pmaddwd xmm4,[GOTOFF(ebx,PW_F117_F078)] ; xmm4=z4H + + movdqa XMMWORD [wk(10)], xmm0 ; wk(10)=z3L + movdqa XMMWORD [wk(11)], xmm7 ; wk(11)=z3H + + ; (Original) + ; z1 = tmp0 + tmp3; z2 = tmp1 + tmp2; + ; tmp0 = tmp0 * 0.298631336; tmp1 = tmp1 * 2.053119869; + ; tmp2 = tmp2 * 3.072711026; tmp3 = tmp3 * 1.501321110; + ; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; + ; tmp0 += z1 + z3; tmp1 += z2 + z4; + ; tmp2 += z2 + z3; tmp3 += z1 + z4; + ; + ; (This implementation) + ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223; + ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447; + ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447); + ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223); + ; tmp0 += z3; tmp1 += z4; + ; tmp2 += z3; tmp3 += z4; + + movdqa xmm0,xmm1 + movdqa xmm7,xmm1 + punpcklwd xmm0,xmm3 + punpckhwd xmm7,xmm3 + movdqa xmm1,xmm0 + movdqa xmm3,xmm7 + pmaddwd xmm0,[GOTOFF(ebx,PW_MF060_MF089)] ; xmm0=tmp0L + pmaddwd xmm7,[GOTOFF(ebx,PW_MF060_MF089)] ; xmm7=tmp0H + pmaddwd xmm1,[GOTOFF(ebx,PW_MF089_F060)] ; xmm1=tmp3L + pmaddwd xmm3,[GOTOFF(ebx,PW_MF089_F060)] ; xmm3=tmp3H + + paddd xmm0, XMMWORD [wk(10)] ; xmm0=tmp0L + paddd xmm7, XMMWORD [wk(11)] ; xmm7=tmp0H + paddd xmm1,xmm5 ; xmm1=tmp3L + paddd xmm3,xmm4 ; xmm3=tmp3H + + movdqa XMMWORD [wk(8)], xmm0 ; wk(8)=tmp0L + movdqa XMMWORD [wk(9)], xmm7 ; wk(9)=tmp0H + + movdqa xmm0,xmm2 + movdqa xmm7,xmm2 + punpcklwd xmm0,xmm6 + punpckhwd xmm7,xmm6 + movdqa xmm2,xmm0 + movdqa xmm6,xmm7 + pmaddwd xmm0,[GOTOFF(ebx,PW_MF050_MF256)] ; xmm0=tmp1L + pmaddwd xmm7,[GOTOFF(ebx,PW_MF050_MF256)] ; xmm7=tmp1H + pmaddwd xmm2,[GOTOFF(ebx,PW_MF256_F050)] ; xmm2=tmp2L + pmaddwd xmm6,[GOTOFF(ebx,PW_MF256_F050)] ; xmm6=tmp2H + + paddd xmm0,xmm5 ; xmm0=tmp1L + paddd xmm7,xmm4 ; xmm7=tmp1H + paddd xmm2, XMMWORD [wk(10)] ; xmm2=tmp2L + paddd xmm6, XMMWORD [wk(11)] ; xmm6=tmp2H + + movdqa XMMWORD [wk(10)], xmm0 ; wk(10)=tmp1L + movdqa XMMWORD [wk(11)], xmm7 ; wk(11)=tmp1H + + ; -- Final output stage + + movdqa xmm5, XMMWORD [wk(0)] ; xmm5=tmp10L + movdqa xmm4, XMMWORD [wk(1)] ; xmm4=tmp10H + + movdqa xmm0,xmm5 + movdqa xmm7,xmm4 + paddd xmm5,xmm1 ; xmm5=data0L + paddd xmm4,xmm3 ; xmm4=data0H + psubd xmm0,xmm1 ; xmm0=data7L + psubd xmm7,xmm3 ; xmm7=data7H + + movdqa xmm1,[GOTOFF(ebx,PD_DESCALE_P2)] ; xmm1=[PD_DESCALE_P2] + + paddd xmm5,xmm1 + paddd xmm4,xmm1 + psrad xmm5,DESCALE_P2 + psrad xmm4,DESCALE_P2 + paddd xmm0,xmm1 + paddd xmm7,xmm1 + psrad xmm0,DESCALE_P2 + psrad xmm7,DESCALE_P2 + + packssdw xmm5,xmm4 ; xmm5=data0=(00 10 20 30 40 50 60 70) + packssdw xmm0,xmm7 ; xmm0=data7=(07 17 27 37 47 57 67 77) + + movdqa xmm3, XMMWORD [wk(4)] ; xmm3=tmp11L + movdqa xmm1, XMMWORD [wk(5)] ; xmm1=tmp11H + + movdqa xmm4,xmm3 + movdqa xmm7,xmm1 + paddd xmm3,xmm2 ; xmm3=data1L + paddd xmm1,xmm6 ; xmm1=data1H + psubd xmm4,xmm2 ; xmm4=data6L + psubd xmm7,xmm6 ; xmm7=data6H + + movdqa xmm2,[GOTOFF(ebx,PD_DESCALE_P2)] ; xmm2=[PD_DESCALE_P2] + + paddd xmm3,xmm2 + paddd xmm1,xmm2 + psrad xmm3,DESCALE_P2 + psrad xmm1,DESCALE_P2 + paddd xmm4,xmm2 + paddd xmm7,xmm2 + psrad xmm4,DESCALE_P2 + psrad xmm7,DESCALE_P2 + + packssdw xmm3,xmm1 ; xmm3=data1=(01 11 21 31 41 51 61 71) + packssdw xmm4,xmm7 ; xmm4=data6=(06 16 26 36 46 56 66 76) + + packsswb xmm5,xmm4 ; xmm5=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76) + packsswb xmm3,xmm0 ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77) + + movdqa xmm6, XMMWORD [wk(6)] ; xmm6=tmp12L + movdqa xmm2, XMMWORD [wk(7)] ; xmm2=tmp12H + movdqa xmm1, XMMWORD [wk(10)] ; xmm1=tmp1L + movdqa xmm7, XMMWORD [wk(11)] ; xmm7=tmp1H + + movdqa XMMWORD [wk(0)], xmm5 ; wk(0)=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76) + movdqa XMMWORD [wk(1)], xmm3 ; wk(1)=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77) + + movdqa xmm4,xmm6 + movdqa xmm0,xmm2 + paddd xmm6,xmm1 ; xmm6=data2L + paddd xmm2,xmm7 ; xmm2=data2H + psubd xmm4,xmm1 ; xmm4=data5L + psubd xmm0,xmm7 ; xmm0=data5H + + movdqa xmm5,[GOTOFF(ebx,PD_DESCALE_P2)] ; xmm5=[PD_DESCALE_P2] + + paddd xmm6,xmm5 + paddd xmm2,xmm5 + psrad xmm6,DESCALE_P2 + psrad xmm2,DESCALE_P2 + paddd xmm4,xmm5 + paddd xmm0,xmm5 + psrad xmm4,DESCALE_P2 + psrad xmm0,DESCALE_P2 + + packssdw xmm6,xmm2 ; xmm6=data2=(02 12 22 32 42 52 62 72) + packssdw xmm4,xmm0 ; xmm4=data5=(05 15 25 35 45 55 65 75) + + movdqa xmm3, XMMWORD [wk(2)] ; xmm3=tmp13L + movdqa xmm1, XMMWORD [wk(3)] ; xmm1=tmp13H + movdqa xmm7, XMMWORD [wk(8)] ; xmm7=tmp0L + movdqa xmm5, XMMWORD [wk(9)] ; xmm5=tmp0H + + movdqa xmm2,xmm3 + movdqa xmm0,xmm1 + paddd xmm3,xmm7 ; xmm3=data3L + paddd xmm1,xmm5 ; xmm1=data3H + psubd xmm2,xmm7 ; xmm2=data4L + psubd xmm0,xmm5 ; xmm0=data4H + + movdqa xmm7,[GOTOFF(ebx,PD_DESCALE_P2)] ; xmm7=[PD_DESCALE_P2] + + paddd xmm3,xmm7 + paddd xmm1,xmm7 + psrad xmm3,DESCALE_P2 + psrad xmm1,DESCALE_P2 + paddd xmm2,xmm7 + paddd xmm0,xmm7 + psrad xmm2,DESCALE_P2 + psrad xmm0,DESCALE_P2 + + movdqa xmm5,[GOTOFF(ebx,PB_CENTERJSAMP)] ; xmm5=[PB_CENTERJSAMP] + + packssdw xmm3,xmm1 ; xmm3=data3=(03 13 23 33 43 53 63 73) + packssdw xmm2,xmm0 ; xmm2=data4=(04 14 24 34 44 54 64 74) + + movdqa xmm7, XMMWORD [wk(0)] ; xmm7=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76) + movdqa xmm1, XMMWORD [wk(1)] ; xmm1=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77) + + packsswb xmm6,xmm2 ; xmm6=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74) + packsswb xmm3,xmm4 ; xmm3=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75) + + paddb xmm7,xmm5 + paddb xmm1,xmm5 + paddb xmm6,xmm5 + paddb xmm3,xmm5 + + movdqa xmm0,xmm7 ; transpose coefficients(phase 1) + punpcklbw xmm7,xmm1 ; xmm7=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71) + punpckhbw xmm0,xmm1 ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77) + movdqa xmm2,xmm6 ; transpose coefficients(phase 1) + punpcklbw xmm6,xmm3 ; xmm6=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73) + punpckhbw xmm2,xmm3 ; xmm2=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75) + + movdqa xmm4,xmm7 ; transpose coefficients(phase 2) + punpcklwd xmm7,xmm6 ; xmm7=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33) + punpckhwd xmm4,xmm6 ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73) + movdqa xmm5,xmm2 ; transpose coefficients(phase 2) + punpcklwd xmm2,xmm0 ; xmm2=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37) + punpckhwd xmm5,xmm0 ; xmm5=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77) + + movdqa xmm1,xmm7 ; transpose coefficients(phase 3) + punpckldq xmm7,xmm2 ; xmm7=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17) + punpckhdq xmm1,xmm2 ; xmm1=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37) + movdqa xmm3,xmm4 ; transpose coefficients(phase 3) + punpckldq xmm4,xmm5 ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57) + punpckhdq xmm3,xmm5 ; xmm3=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77) + + pshufd xmm6,xmm7,0x4E ; xmm6=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07) + pshufd xmm0,xmm1,0x4E ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27) + pshufd xmm2,xmm4,0x4E ; xmm2=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47) + pshufd xmm5,xmm3,0x4E ; xmm5=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67) + + mov edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+2*SIZEOF_JSAMPROW] + movq XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm7 + movq XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm1 + mov edx, JSAMPROW [edi+4*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+6*SIZEOF_JSAMPROW] + movq XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm4 + movq XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm3 + + mov edx, JSAMPROW [edi+1*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+3*SIZEOF_JSAMPROW] + movq XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm6 + movq XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm0 + mov edx, JSAMPROW [edi+5*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+7*SIZEOF_JSAMPROW] + movq XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm2 + movq XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm5 + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; unused + poppic ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jidctred-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jidctred-mmx.asm new file mode 100644 index 000000000..1c939013c --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jidctred-mmx.asm @@ -0,0 +1,706 @@ +; +; jidctred.asm - reduced-size IDCT (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains inverse-DCT routines that produce reduced-size +; output: either 4x4 or 2x2 pixels from an 8x8 DCT block. +; The following code is based directly on the IJG's original jidctred.c; +; see the jidctred.c for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 13 +%define PASS1_BITS 2 + +%define DESCALE_P1_4 (CONST_BITS-PASS1_BITS+1) +%define DESCALE_P2_4 (CONST_BITS+PASS1_BITS+3+1) +%define DESCALE_P1_2 (CONST_BITS-PASS1_BITS+2) +%define DESCALE_P2_2 (CONST_BITS+PASS1_BITS+3+2) + +%if CONST_BITS == 13 +F_0_211 equ 1730 ; FIX(0.211164243) +F_0_509 equ 4176 ; FIX(0.509795579) +F_0_601 equ 4926 ; FIX(0.601344887) +F_0_720 equ 5906 ; FIX(0.720959822) +F_0_765 equ 6270 ; FIX(0.765366865) +F_0_850 equ 6967 ; FIX(0.850430095) +F_0_899 equ 7373 ; FIX(0.899976223) +F_1_061 equ 8697 ; FIX(1.061594337) +F_1_272 equ 10426 ; FIX(1.272758580) +F_1_451 equ 11893 ; FIX(1.451774981) +F_1_847 equ 15137 ; FIX(1.847759065) +F_2_172 equ 17799 ; FIX(2.172734803) +F_2_562 equ 20995 ; FIX(2.562915447) +F_3_624 equ 29692 ; FIX(3.624509785) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_0_211 equ DESCALE( 226735879,30-CONST_BITS) ; FIX(0.211164243) +F_0_509 equ DESCALE( 547388834,30-CONST_BITS) ; FIX(0.509795579) +F_0_601 equ DESCALE( 645689155,30-CONST_BITS) ; FIX(0.601344887) +F_0_720 equ DESCALE( 774124714,30-CONST_BITS) ; FIX(0.720959822) +F_0_765 equ DESCALE( 821806413,30-CONST_BITS) ; FIX(0.765366865) +F_0_850 equ DESCALE( 913142361,30-CONST_BITS) ; FIX(0.850430095) +F_0_899 equ DESCALE( 966342111,30-CONST_BITS) ; FIX(0.899976223) +F_1_061 equ DESCALE(1139878239,30-CONST_BITS) ; FIX(1.061594337) +F_1_272 equ DESCALE(1366614119,30-CONST_BITS) ; FIX(1.272758580) +F_1_451 equ DESCALE(1558831516,30-CONST_BITS) ; FIX(1.451774981) +F_1_847 equ DESCALE(1984016188,30-CONST_BITS) ; FIX(1.847759065) +F_2_172 equ DESCALE(2332956230,30-CONST_BITS) ; FIX(2.172734803) +F_2_562 equ DESCALE(2751909506,30-CONST_BITS) ; FIX(2.562915447) +F_3_624 equ DESCALE(3891787747,30-CONST_BITS) ; FIX(3.624509785) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_idct_red_mmx) + +EXTN(jconst_idct_red_mmx): + +PW_F184_MF076 times 2 dw F_1_847,-F_0_765 +PW_F256_F089 times 2 dw F_2_562, F_0_899 +PW_F106_MF217 times 2 dw F_1_061,-F_2_172 +PW_MF060_MF050 times 2 dw -F_0_601,-F_0_509 +PW_F145_MF021 times 2 dw F_1_451,-F_0_211 +PW_F362_MF127 times 2 dw F_3_624,-F_1_272 +PW_F085_MF072 times 2 dw F_0_850,-F_0_720 +PD_DESCALE_P1_4 times 2 dd 1 << (DESCALE_P1_4-1) +PD_DESCALE_P2_4 times 2 dd 1 << (DESCALE_P2_4-1) +PD_DESCALE_P1_2 times 2 dd 1 << (DESCALE_P1_2-1) +PD_DESCALE_P2_2 times 2 dd 1 << (DESCALE_P2_2-1) +PB_CENTERJSAMP times 8 db CENTERJSAMPLE + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform dequantization and inverse DCT on one block of coefficients, +; producing a reduced-size 4x4 output block. +; +; GLOBAL(void) +; jsimd_idct_4x4_mmx (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +%define dct_table(b) (b)+8 ; void * dct_table +%define coef_block(b) (b)+12 ; JCOEFPTR coef_block +%define output_buf(b) (b)+16 ; JSAMPARRAY output_buf +%define output_col(b) (b)+20 ; JDIMENSION output_col + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_MMWORD ; mmword wk[WK_NUM] +%define WK_NUM 2 +%define workspace wk(0)-DCTSIZE2*SIZEOF_JCOEF + ; JCOEF workspace[DCTSIZE2] + + align 16 + global EXTN(jsimd_idct_4x4_mmx) + +EXTN(jsimd_idct_4x4_mmx): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_MMWORD) ; align to 64 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [workspace] + pushpic ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process columns from input, store into work array. + +; mov eax, [original_ebp] + mov edx, POINTER [dct_table(eax)] ; quantptr + mov esi, JCOEFPTR [coef_block(eax)] ; inptr + lea edi, [workspace] ; JCOEF * wsptr + mov ecx, DCTSIZE/4 ; ctr + alignx 16,7 +.columnloop: +%ifndef NO_ZERO_COLUMN_TEST_4X4_MMX + mov eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)] + jnz short .columnDCT + + movq mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)] + por mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + por mm1, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + por mm0, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)] + por mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + por mm0,mm1 + packsswb mm0,mm0 + movd eax,mm0 + test eax,eax + jnz short .columnDCT + + ; -- AC terms all zero + + movq mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + pmullw mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + psllw mm0,PASS1_BITS + + movq mm2,mm0 ; mm0=in0=(00 01 02 03) + punpcklwd mm0,mm0 ; mm0=(00 00 01 01) + punpckhwd mm2,mm2 ; mm2=(02 02 03 03) + + movq mm1,mm0 + punpckldq mm0,mm0 ; mm0=(00 00 00 00) + punpckhdq mm1,mm1 ; mm1=(01 01 01 01) + movq mm3,mm2 + punpckldq mm2,mm2 ; mm2=(02 02 02 02) + punpckhdq mm3,mm3 ; mm3=(03 03 03 03) + + movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0 + movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm1 + movq MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm2 + movq MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm3 + jmp near .nextcolumn + alignx 16,7 +%endif +.columnDCT: + + ; -- Odd part + + movq mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + pmullw mm0, MMWORD [MMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + movq mm2, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movq mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + pmullw mm2, MMWORD [MMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + movq mm4,mm0 + movq mm5,mm0 + punpcklwd mm4,mm1 + punpckhwd mm5,mm1 + movq mm0,mm4 + movq mm1,mm5 + pmaddwd mm4,[GOTOFF(ebx,PW_F256_F089)] ; mm4=(tmp2L) + pmaddwd mm5,[GOTOFF(ebx,PW_F256_F089)] ; mm5=(tmp2H) + pmaddwd mm0,[GOTOFF(ebx,PW_F106_MF217)] ; mm0=(tmp0L) + pmaddwd mm1,[GOTOFF(ebx,PW_F106_MF217)] ; mm1=(tmp0H) + + movq mm6,mm2 + movq mm7,mm2 + punpcklwd mm6,mm3 + punpckhwd mm7,mm3 + movq mm2,mm6 + movq mm3,mm7 + pmaddwd mm6,[GOTOFF(ebx,PW_MF060_MF050)] ; mm6=(tmp2L) + pmaddwd mm7,[GOTOFF(ebx,PW_MF060_MF050)] ; mm7=(tmp2H) + pmaddwd mm2,[GOTOFF(ebx,PW_F145_MF021)] ; mm2=(tmp0L) + pmaddwd mm3,[GOTOFF(ebx,PW_F145_MF021)] ; mm3=(tmp0H) + + paddd mm6,mm4 ; mm6=tmp2L + paddd mm7,mm5 ; mm7=tmp2H + paddd mm2,mm0 ; mm2=tmp0L + paddd mm3,mm1 ; mm3=tmp0H + + movq MMWORD [wk(0)], mm2 ; wk(0)=tmp0L + movq MMWORD [wk(1)], mm3 ; wk(1)=tmp0H + + ; -- Even part + + movq mm4, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + movq mm5, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)] + movq mm0, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)] + pmullw mm4, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw mm5, MMWORD [MMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw mm0, MMWORD [MMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + pxor mm1,mm1 + pxor mm2,mm2 + punpcklwd mm1,mm4 ; mm1=tmp0L + punpckhwd mm2,mm4 ; mm2=tmp0H + psrad mm1,(16-CONST_BITS-1) ; psrad mm1,16 & pslld mm1,CONST_BITS+1 + psrad mm2,(16-CONST_BITS-1) ; psrad mm2,16 & pslld mm2,CONST_BITS+1 + + movq mm3,mm5 ; mm5=in2=z2 + punpcklwd mm5,mm0 ; mm0=in6=z3 + punpckhwd mm3,mm0 + pmaddwd mm5,[GOTOFF(ebx,PW_F184_MF076)] ; mm5=tmp2L + pmaddwd mm3,[GOTOFF(ebx,PW_F184_MF076)] ; mm3=tmp2H + + movq mm4,mm1 + movq mm0,mm2 + paddd mm1,mm5 ; mm1=tmp10L + paddd mm2,mm3 ; mm2=tmp10H + psubd mm4,mm5 ; mm4=tmp12L + psubd mm0,mm3 ; mm0=tmp12H + + ; -- Final output stage + + movq mm5,mm1 + movq mm3,mm2 + paddd mm1,mm6 ; mm1=data0L + paddd mm2,mm7 ; mm2=data0H + psubd mm5,mm6 ; mm5=data3L + psubd mm3,mm7 ; mm3=data3H + + movq mm6,[GOTOFF(ebx,PD_DESCALE_P1_4)] ; mm6=[PD_DESCALE_P1_4] + + paddd mm1,mm6 + paddd mm2,mm6 + psrad mm1,DESCALE_P1_4 + psrad mm2,DESCALE_P1_4 + paddd mm5,mm6 + paddd mm3,mm6 + psrad mm5,DESCALE_P1_4 + psrad mm3,DESCALE_P1_4 + + packssdw mm1,mm2 ; mm1=data0=(00 01 02 03) + packssdw mm5,mm3 ; mm5=data3=(30 31 32 33) + + movq mm7, MMWORD [wk(0)] ; mm7=tmp0L + movq mm6, MMWORD [wk(1)] ; mm6=tmp0H + + movq mm2,mm4 + movq mm3,mm0 + paddd mm4,mm7 ; mm4=data1L + paddd mm0,mm6 ; mm0=data1H + psubd mm2,mm7 ; mm2=data2L + psubd mm3,mm6 ; mm3=data2H + + movq mm7,[GOTOFF(ebx,PD_DESCALE_P1_4)] ; mm7=[PD_DESCALE_P1_4] + + paddd mm4,mm7 + paddd mm0,mm7 + psrad mm4,DESCALE_P1_4 + psrad mm0,DESCALE_P1_4 + paddd mm2,mm7 + paddd mm3,mm7 + psrad mm2,DESCALE_P1_4 + psrad mm3,DESCALE_P1_4 + + packssdw mm4,mm0 ; mm4=data1=(10 11 12 13) + packssdw mm2,mm3 ; mm2=data2=(20 21 22 23) + + movq mm6,mm1 ; transpose coefficients(phase 1) + punpcklwd mm1,mm4 ; mm1=(00 10 01 11) + punpckhwd mm6,mm4 ; mm6=(02 12 03 13) + movq mm7,mm2 ; transpose coefficients(phase 1) + punpcklwd mm2,mm5 ; mm2=(20 30 21 31) + punpckhwd mm7,mm5 ; mm7=(22 32 23 33) + + movq mm0,mm1 ; transpose coefficients(phase 2) + punpckldq mm1,mm2 ; mm1=(00 10 20 30) + punpckhdq mm0,mm2 ; mm0=(01 11 21 31) + movq mm3,mm6 ; transpose coefficients(phase 2) + punpckldq mm6,mm7 ; mm6=(02 12 22 32) + punpckhdq mm3,mm7 ; mm3=(03 13 23 33) + + movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm1 + movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm0 + movq MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm6 + movq MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm3 + +.nextcolumn: + add esi, byte 4*SIZEOF_JCOEF ; coef_block + add edx, byte 4*SIZEOF_ISLOW_MULT_TYPE ; quantptr + add edi, byte 4*DCTSIZE*SIZEOF_JCOEF ; wsptr + dec ecx ; ctr + jnz near .columnloop + + ; ---- Pass 2: process rows from work array, store into output array. + + mov eax, [original_ebp] + lea esi, [workspace] ; JCOEF * wsptr + mov edi, JSAMPARRAY [output_buf(eax)] ; (JSAMPROW *) + mov eax, JDIMENSION [output_col(eax)] + + ; -- Odd part + + movq mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + movq mm2, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movq mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + + movq mm4,mm0 + movq mm5,mm0 + punpcklwd mm4,mm1 + punpckhwd mm5,mm1 + movq mm0,mm4 + movq mm1,mm5 + pmaddwd mm4,[GOTOFF(ebx,PW_F256_F089)] ; mm4=(tmp2L) + pmaddwd mm5,[GOTOFF(ebx,PW_F256_F089)] ; mm5=(tmp2H) + pmaddwd mm0,[GOTOFF(ebx,PW_F106_MF217)] ; mm0=(tmp0L) + pmaddwd mm1,[GOTOFF(ebx,PW_F106_MF217)] ; mm1=(tmp0H) + + movq mm6,mm2 + movq mm7,mm2 + punpcklwd mm6,mm3 + punpckhwd mm7,mm3 + movq mm2,mm6 + movq mm3,mm7 + pmaddwd mm6,[GOTOFF(ebx,PW_MF060_MF050)] ; mm6=(tmp2L) + pmaddwd mm7,[GOTOFF(ebx,PW_MF060_MF050)] ; mm7=(tmp2H) + pmaddwd mm2,[GOTOFF(ebx,PW_F145_MF021)] ; mm2=(tmp0L) + pmaddwd mm3,[GOTOFF(ebx,PW_F145_MF021)] ; mm3=(tmp0H) + + paddd mm6,mm4 ; mm6=tmp2L + paddd mm7,mm5 ; mm7=tmp2H + paddd mm2,mm0 ; mm2=tmp0L + paddd mm3,mm1 ; mm3=tmp0H + + movq MMWORD [wk(0)], mm2 ; wk(0)=tmp0L + movq MMWORD [wk(1)], mm3 ; wk(1)=tmp0H + + ; -- Even part + + movq mm4, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + movq mm5, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)] + movq mm0, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)] + + pxor mm1,mm1 + pxor mm2,mm2 + punpcklwd mm1,mm4 ; mm1=tmp0L + punpckhwd mm2,mm4 ; mm2=tmp0H + psrad mm1,(16-CONST_BITS-1) ; psrad mm1,16 & pslld mm1,CONST_BITS+1 + psrad mm2,(16-CONST_BITS-1) ; psrad mm2,16 & pslld mm2,CONST_BITS+1 + + movq mm3,mm5 ; mm5=in2=z2 + punpcklwd mm5,mm0 ; mm0=in6=z3 + punpckhwd mm3,mm0 + pmaddwd mm5,[GOTOFF(ebx,PW_F184_MF076)] ; mm5=tmp2L + pmaddwd mm3,[GOTOFF(ebx,PW_F184_MF076)] ; mm3=tmp2H + + movq mm4,mm1 + movq mm0,mm2 + paddd mm1,mm5 ; mm1=tmp10L + paddd mm2,mm3 ; mm2=tmp10H + psubd mm4,mm5 ; mm4=tmp12L + psubd mm0,mm3 ; mm0=tmp12H + + ; -- Final output stage + + movq mm5,mm1 + movq mm3,mm2 + paddd mm1,mm6 ; mm1=data0L + paddd mm2,mm7 ; mm2=data0H + psubd mm5,mm6 ; mm5=data3L + psubd mm3,mm7 ; mm3=data3H + + movq mm6,[GOTOFF(ebx,PD_DESCALE_P2_4)] ; mm6=[PD_DESCALE_P2_4] + + paddd mm1,mm6 + paddd mm2,mm6 + psrad mm1,DESCALE_P2_4 + psrad mm2,DESCALE_P2_4 + paddd mm5,mm6 + paddd mm3,mm6 + psrad mm5,DESCALE_P2_4 + psrad mm3,DESCALE_P2_4 + + packssdw mm1,mm2 ; mm1=data0=(00 10 20 30) + packssdw mm5,mm3 ; mm5=data3=(03 13 23 33) + + movq mm7, MMWORD [wk(0)] ; mm7=tmp0L + movq mm6, MMWORD [wk(1)] ; mm6=tmp0H + + movq mm2,mm4 + movq mm3,mm0 + paddd mm4,mm7 ; mm4=data1L + paddd mm0,mm6 ; mm0=data1H + psubd mm2,mm7 ; mm2=data2L + psubd mm3,mm6 ; mm3=data2H + + movq mm7,[GOTOFF(ebx,PD_DESCALE_P2_4)] ; mm7=[PD_DESCALE_P2_4] + + paddd mm4,mm7 + paddd mm0,mm7 + psrad mm4,DESCALE_P2_4 + psrad mm0,DESCALE_P2_4 + paddd mm2,mm7 + paddd mm3,mm7 + psrad mm2,DESCALE_P2_4 + psrad mm3,DESCALE_P2_4 + + packssdw mm4,mm0 ; mm4=data1=(01 11 21 31) + packssdw mm2,mm3 ; mm2=data2=(02 12 22 32) + + movq mm6,[GOTOFF(ebx,PB_CENTERJSAMP)] ; mm6=[PB_CENTERJSAMP] + + packsswb mm1,mm2 ; mm1=(00 10 20 30 02 12 22 32) + packsswb mm4,mm5 ; mm4=(01 11 21 31 03 13 23 33) + paddb mm1,mm6 + paddb mm4,mm6 + + movq mm7,mm1 ; transpose coefficients(phase 1) + punpcklbw mm1,mm4 ; mm1=(00 01 10 11 20 21 30 31) + punpckhbw mm7,mm4 ; mm7=(02 03 12 13 22 23 32 33) + + movq mm0,mm1 ; transpose coefficients(phase 2) + punpcklwd mm1,mm7 ; mm1=(00 01 02 03 10 11 12 13) + punpckhwd mm0,mm7 ; mm0=(20 21 22 23 30 31 32 33) + + mov edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+2*SIZEOF_JSAMPROW] + movd DWORD [edx+eax*SIZEOF_JSAMPLE], mm1 + movd DWORD [esi+eax*SIZEOF_JSAMPLE], mm0 + + psrlq mm1,4*BYTE_BIT + psrlq mm0,4*BYTE_BIT + + mov edx, JSAMPROW [edi+1*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+3*SIZEOF_JSAMPROW] + movd DWORD [edx+eax*SIZEOF_JSAMPLE], mm1 + movd DWORD [esi+eax*SIZEOF_JSAMPLE], mm0 + + emms ; empty MMX state + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + poppic ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + + +; -------------------------------------------------------------------------- +; +; Perform dequantization and inverse DCT on one block of coefficients, +; producing a reduced-size 2x2 output block. +; +; GLOBAL(void) +; jsimd_idct_2x2_mmx (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +%define dct_table(b) (b)+8 ; void * dct_table +%define coef_block(b) (b)+12 ; JCOEFPTR coef_block +%define output_buf(b) (b)+16 ; JSAMPARRAY output_buf +%define output_col(b) (b)+20 ; JDIMENSION output_col + + align 16 + global EXTN(jsimd_idct_2x2_mmx) + +EXTN(jsimd_idct_2x2_mmx): + push ebp + mov ebp,esp + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process columns from input. + + mov edx, POINTER [dct_table(ebp)] ; quantptr + mov esi, JCOEFPTR [coef_block(ebp)] ; inptr + + ; | input: | result: | + ; | 00 01 ** 03 ** 05 ** 07 | | + ; | 10 11 ** 13 ** 15 ** 17 | | + ; | ** ** ** ** ** ** ** ** | | + ; | 30 31 ** 33 ** 35 ** 37 | A0 A1 A3 A5 A7 | + ; | ** ** ** ** ** ** ** ** | B0 B1 B3 B5 B7 | + ; | 50 51 ** 53 ** 55 ** 57 | | + ; | ** ** ** ** ** ** ** ** | | + ; | 70 71 ** 73 ** 75 ** 77 | | + + ; -- Odd part + + movq mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)] + pmullw mm0, MMWORD [MMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + movq mm2, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movq mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)] + pmullw mm2, MMWORD [MMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + ; mm0=(10 11 ** 13), mm1=(30 31 ** 33) + ; mm2=(50 51 ** 53), mm3=(70 71 ** 73) + + pcmpeqd mm7,mm7 + pslld mm7,WORD_BIT ; mm7={0x0000 0xFFFF 0x0000 0xFFFF} + + movq mm4,mm0 ; mm4=(10 11 ** 13) + movq mm5,mm2 ; mm5=(50 51 ** 53) + punpcklwd mm4,mm1 ; mm4=(10 30 11 31) + punpcklwd mm5,mm3 ; mm5=(50 70 51 71) + pmaddwd mm4,[GOTOFF(ebx,PW_F362_MF127)] + pmaddwd mm5,[GOTOFF(ebx,PW_F085_MF072)] + + psrld mm0,WORD_BIT ; mm0=(11 -- 13 --) + pand mm1,mm7 ; mm1=(-- 31 -- 33) + psrld mm2,WORD_BIT ; mm2=(51 -- 53 --) + pand mm3,mm7 ; mm3=(-- 71 -- 73) + por mm0,mm1 ; mm0=(11 31 13 33) + por mm2,mm3 ; mm2=(51 71 53 73) + pmaddwd mm0,[GOTOFF(ebx,PW_F362_MF127)] + pmaddwd mm2,[GOTOFF(ebx,PW_F085_MF072)] + + paddd mm4,mm5 ; mm4=tmp0[col0 col1] + + movq mm6, MMWORD [MMBLOCK(1,1,esi,SIZEOF_JCOEF)] + movq mm1, MMWORD [MMBLOCK(3,1,esi,SIZEOF_JCOEF)] + pmullw mm6, MMWORD [MMBLOCK(1,1,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw mm1, MMWORD [MMBLOCK(3,1,edx,SIZEOF_ISLOW_MULT_TYPE)] + movq mm3, MMWORD [MMBLOCK(5,1,esi,SIZEOF_JCOEF)] + movq mm5, MMWORD [MMBLOCK(7,1,esi,SIZEOF_JCOEF)] + pmullw mm3, MMWORD [MMBLOCK(5,1,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw mm5, MMWORD [MMBLOCK(7,1,edx,SIZEOF_ISLOW_MULT_TYPE)] + + ; mm6=(** 15 ** 17), mm1=(** 35 ** 37) + ; mm3=(** 55 ** 57), mm5=(** 75 ** 77) + + psrld mm6,WORD_BIT ; mm6=(15 -- 17 --) + pand mm1,mm7 ; mm1=(-- 35 -- 37) + psrld mm3,WORD_BIT ; mm3=(55 -- 57 --) + pand mm5,mm7 ; mm5=(-- 75 -- 77) + por mm6,mm1 ; mm6=(15 35 17 37) + por mm3,mm5 ; mm3=(55 75 57 77) + pmaddwd mm6,[GOTOFF(ebx,PW_F362_MF127)] + pmaddwd mm3,[GOTOFF(ebx,PW_F085_MF072)] + + paddd mm0,mm2 ; mm0=tmp0[col1 col3] + paddd mm6,mm3 ; mm6=tmp0[col5 col7] + + ; -- Even part + + movq mm1, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)] + movq mm5, MMWORD [MMBLOCK(0,1,esi,SIZEOF_JCOEF)] + pmullw mm1, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw mm5, MMWORD [MMBLOCK(0,1,edx,SIZEOF_ISLOW_MULT_TYPE)] + + ; mm1=(00 01 ** 03), mm5=(** 05 ** 07) + + movq mm2,mm1 ; mm2=(00 01 ** 03) + pslld mm1,WORD_BIT ; mm1=(-- 00 -- **) + psrad mm1,(WORD_BIT-CONST_BITS-2) ; mm1=tmp10[col0 ****] + + pand mm2,mm7 ; mm2=(-- 01 -- 03) + pand mm5,mm7 ; mm5=(-- 05 -- 07) + psrad mm2,(WORD_BIT-CONST_BITS-2) ; mm2=tmp10[col1 col3] + psrad mm5,(WORD_BIT-CONST_BITS-2) ; mm5=tmp10[col5 col7] + + ; -- Final output stage + + movq mm3,mm1 + paddd mm1,mm4 ; mm1=data0[col0 ****]=(A0 **) + psubd mm3,mm4 ; mm3=data1[col0 ****]=(B0 **) + punpckldq mm1,mm3 ; mm1=(A0 B0) + + movq mm7,[GOTOFF(ebx,PD_DESCALE_P1_2)] ; mm7=[PD_DESCALE_P1_2] + + movq mm4,mm2 + movq mm3,mm5 + paddd mm2,mm0 ; mm2=data0[col1 col3]=(A1 A3) + paddd mm5,mm6 ; mm5=data0[col5 col7]=(A5 A7) + psubd mm4,mm0 ; mm4=data1[col1 col3]=(B1 B3) + psubd mm3,mm6 ; mm3=data1[col5 col7]=(B5 B7) + + paddd mm1,mm7 + psrad mm1,DESCALE_P1_2 + + paddd mm2,mm7 + paddd mm5,mm7 + psrad mm2,DESCALE_P1_2 + psrad mm5,DESCALE_P1_2 + paddd mm4,mm7 + paddd mm3,mm7 + psrad mm4,DESCALE_P1_2 + psrad mm3,DESCALE_P1_2 + + ; ---- Pass 2: process rows, store into output array. + + mov edi, JSAMPARRAY [output_buf(ebp)] ; (JSAMPROW *) + mov eax, JDIMENSION [output_col(ebp)] + + ; | input:| result:| + ; | A0 B0 | | + ; | A1 B1 | C0 C1 | + ; | A3 B3 | D0 D1 | + ; | A5 B5 | | + ; | A7 B7 | | + + ; -- Odd part + + packssdw mm2,mm4 ; mm2=(A1 A3 B1 B3) + packssdw mm5,mm3 ; mm5=(A5 A7 B5 B7) + pmaddwd mm2,[GOTOFF(ebx,PW_F362_MF127)] + pmaddwd mm5,[GOTOFF(ebx,PW_F085_MF072)] + + paddd mm2,mm5 ; mm2=tmp0[row0 row1] + + ; -- Even part + + pslld mm1,(CONST_BITS+2) ; mm1=tmp10[row0 row1] + + ; -- Final output stage + + movq mm0,[GOTOFF(ebx,PD_DESCALE_P2_2)] ; mm0=[PD_DESCALE_P2_2] + + movq mm6,mm1 + paddd mm1,mm2 ; mm1=data0[row0 row1]=(C0 C1) + psubd mm6,mm2 ; mm6=data1[row0 row1]=(D0 D1) + + paddd mm1,mm0 + paddd mm6,mm0 + psrad mm1,DESCALE_P2_2 + psrad mm6,DESCALE_P2_2 + + movq mm7,mm1 ; transpose coefficients + punpckldq mm1,mm6 ; mm1=(C0 D0) + punpckhdq mm7,mm6 ; mm7=(C1 D1) + + packssdw mm1,mm7 ; mm1=(C0 D0 C1 D1) + packsswb mm1,mm1 ; mm1=(C0 D0 C1 D1 C0 D0 C1 D1) + paddb mm1,[GOTOFF(ebx,PB_CENTERJSAMP)] + + movd ecx,mm1 + movd ebx,mm1 ; ebx=(C0 D0 C1 D1) + shr ecx,2*BYTE_BIT ; ecx=(C1 D1 -- --) + + mov edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+1*SIZEOF_JSAMPROW] + mov WORD [edx+eax*SIZEOF_JSAMPLE], bx + mov WORD [esi+eax*SIZEOF_JSAMPLE], cx + + emms ; empty MMX state + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jidctred-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jidctred-sse2-64.asm new file mode 100644 index 000000000..dad43d905 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jidctred-sse2-64.asm @@ -0,0 +1,576 @@ +; +; jidctred.asm - reduced-size IDCT (64-bit SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains inverse-DCT routines that produce reduced-size +; output: either 4x4 or 2x2 pixels from an 8x8 DCT block. +; The following code is based directly on the IJG's original jidctred.c; +; see the jidctred.c for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 13 +%define PASS1_BITS 2 + +%define DESCALE_P1_4 (CONST_BITS-PASS1_BITS+1) +%define DESCALE_P2_4 (CONST_BITS+PASS1_BITS+3+1) +%define DESCALE_P1_2 (CONST_BITS-PASS1_BITS+2) +%define DESCALE_P2_2 (CONST_BITS+PASS1_BITS+3+2) + +%if CONST_BITS == 13 +F_0_211 equ 1730 ; FIX(0.211164243) +F_0_509 equ 4176 ; FIX(0.509795579) +F_0_601 equ 4926 ; FIX(0.601344887) +F_0_720 equ 5906 ; FIX(0.720959822) +F_0_765 equ 6270 ; FIX(0.765366865) +F_0_850 equ 6967 ; FIX(0.850430095) +F_0_899 equ 7373 ; FIX(0.899976223) +F_1_061 equ 8697 ; FIX(1.061594337) +F_1_272 equ 10426 ; FIX(1.272758580) +F_1_451 equ 11893 ; FIX(1.451774981) +F_1_847 equ 15137 ; FIX(1.847759065) +F_2_172 equ 17799 ; FIX(2.172734803) +F_2_562 equ 20995 ; FIX(2.562915447) +F_3_624 equ 29692 ; FIX(3.624509785) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_0_211 equ DESCALE( 226735879,30-CONST_BITS) ; FIX(0.211164243) +F_0_509 equ DESCALE( 547388834,30-CONST_BITS) ; FIX(0.509795579) +F_0_601 equ DESCALE( 645689155,30-CONST_BITS) ; FIX(0.601344887) +F_0_720 equ DESCALE( 774124714,30-CONST_BITS) ; FIX(0.720959822) +F_0_765 equ DESCALE( 821806413,30-CONST_BITS) ; FIX(0.765366865) +F_0_850 equ DESCALE( 913142361,30-CONST_BITS) ; FIX(0.850430095) +F_0_899 equ DESCALE( 966342111,30-CONST_BITS) ; FIX(0.899976223) +F_1_061 equ DESCALE(1139878239,30-CONST_BITS) ; FIX(1.061594337) +F_1_272 equ DESCALE(1366614119,30-CONST_BITS) ; FIX(1.272758580) +F_1_451 equ DESCALE(1558831516,30-CONST_BITS) ; FIX(1.451774981) +F_1_847 equ DESCALE(1984016188,30-CONST_BITS) ; FIX(1.847759065) +F_2_172 equ DESCALE(2332956230,30-CONST_BITS) ; FIX(2.172734803) +F_2_562 equ DESCALE(2751909506,30-CONST_BITS) ; FIX(2.562915447) +F_3_624 equ DESCALE(3891787747,30-CONST_BITS) ; FIX(3.624509785) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_idct_red_sse2) + +EXTN(jconst_idct_red_sse2): + +PW_F184_MF076 times 4 dw F_1_847,-F_0_765 +PW_F256_F089 times 4 dw F_2_562, F_0_899 +PW_F106_MF217 times 4 dw F_1_061,-F_2_172 +PW_MF060_MF050 times 4 dw -F_0_601,-F_0_509 +PW_F145_MF021 times 4 dw F_1_451,-F_0_211 +PW_F362_MF127 times 4 dw F_3_624,-F_1_272 +PW_F085_MF072 times 4 dw F_0_850,-F_0_720 +PD_DESCALE_P1_4 times 4 dd 1 << (DESCALE_P1_4-1) +PD_DESCALE_P2_4 times 4 dd 1 << (DESCALE_P2_4-1) +PD_DESCALE_P1_2 times 4 dd 1 << (DESCALE_P1_2-1) +PD_DESCALE_P2_2 times 4 dd 1 << (DESCALE_P2_2-1) +PB_CENTERJSAMP times 16 db CENTERJSAMPLE + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 +; +; Perform dequantization and inverse DCT on one block of coefficients, +; producing a reduced-size 4x4 output block. +; +; GLOBAL(void) +; jsimd_idct_4x4_sse2 (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +; r10 = void * dct_table +; r11 = JCOEFPTR coef_block +; r12 = JSAMPARRAY output_buf +; r13 = JDIMENSION output_col + +%define original_rbp rbp+0 +%define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 + + align 16 + global EXTN(jsimd_idct_4x4_sse2) + +EXTN(jsimd_idct_4x4_sse2): + push rbp + mov rax,rsp ; rax = original rbp + sub rsp, byte 4 + and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [rsp],rax + mov rbp,rsp ; rbp = aligned rbp + lea rsp, [wk(0)] + collect_args + + ; ---- Pass 1: process columns from input. + + mov rdx, r10 ; quantptr + mov rsi, r11 ; inptr + +%ifndef NO_ZERO_COLUMN_TEST_4X4_SSE2 + mov eax, DWORD [DWBLOCK(1,0,rsi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(2,0,rsi,SIZEOF_JCOEF)] + jnz short .columnDCT + + movdqa xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)] + por xmm1, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)] + por xmm1, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)] + por xmm0,xmm1 + packsswb xmm0,xmm0 + packsswb xmm0,xmm0 + movd eax,xmm0 + test rax,rax + jnz short .columnDCT + + ; -- AC terms all zero + + movdqa xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)] + pmullw xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + + psllw xmm0,PASS1_BITS + + movdqa xmm3,xmm0 ; xmm0=in0=(00 01 02 03 04 05 06 07) + punpcklwd xmm0,xmm0 ; xmm0=(00 00 01 01 02 02 03 03) + punpckhwd xmm3,xmm3 ; xmm3=(04 04 05 05 06 06 07 07) + + pshufd xmm1,xmm0,0x50 ; xmm1=[col0 col1]=(00 00 00 00 01 01 01 01) + pshufd xmm0,xmm0,0xFA ; xmm0=[col2 col3]=(02 02 02 02 03 03 03 03) + pshufd xmm6,xmm3,0x50 ; xmm6=[col4 col5]=(04 04 04 04 05 05 05 05) + pshufd xmm3,xmm3,0xFA ; xmm3=[col6 col7]=(06 06 06 06 07 07 07 07) + + jmp near .column_end +%endif +.columnDCT: + + ; -- Odd part + + movdqa xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)] + pmullw xmm0, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm1, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + movdqa xmm2, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)] + movdqa xmm3, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)] + pmullw xmm2, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + + movdqa xmm4,xmm0 + movdqa xmm5,xmm0 + punpcklwd xmm4,xmm1 + punpckhwd xmm5,xmm1 + movdqa xmm0,xmm4 + movdqa xmm1,xmm5 + pmaddwd xmm4,[rel PW_F256_F089] ; xmm4=(tmp2L) + pmaddwd xmm5,[rel PW_F256_F089] ; xmm5=(tmp2H) + pmaddwd xmm0,[rel PW_F106_MF217] ; xmm0=(tmp0L) + pmaddwd xmm1,[rel PW_F106_MF217] ; xmm1=(tmp0H) + + movdqa xmm6,xmm2 + movdqa xmm7,xmm2 + punpcklwd xmm6,xmm3 + punpckhwd xmm7,xmm3 + movdqa xmm2,xmm6 + movdqa xmm3,xmm7 + pmaddwd xmm6,[rel PW_MF060_MF050] ; xmm6=(tmp2L) + pmaddwd xmm7,[rel PW_MF060_MF050] ; xmm7=(tmp2H) + pmaddwd xmm2,[rel PW_F145_MF021] ; xmm2=(tmp0L) + pmaddwd xmm3,[rel PW_F145_MF021] ; xmm3=(tmp0H) + + paddd xmm6,xmm4 ; xmm6=tmp2L + paddd xmm7,xmm5 ; xmm7=tmp2H + paddd xmm2,xmm0 ; xmm2=tmp0L + paddd xmm3,xmm1 ; xmm3=tmp0H + + movdqa XMMWORD [wk(0)], xmm2 ; wk(0)=tmp0L + movdqa XMMWORD [wk(1)], xmm3 ; wk(1)=tmp0H + + ; -- Even part + + movdqa xmm4, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)] + movdqa xmm5, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)] + movdqa xmm0, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)] + pmullw xmm4, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm5, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm0, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + + pxor xmm1,xmm1 + pxor xmm2,xmm2 + punpcklwd xmm1,xmm4 ; xmm1=tmp0L + punpckhwd xmm2,xmm4 ; xmm2=tmp0H + psrad xmm1,(16-CONST_BITS-1) ; psrad xmm1,16 & pslld xmm1,CONST_BITS+1 + psrad xmm2,(16-CONST_BITS-1) ; psrad xmm2,16 & pslld xmm2,CONST_BITS+1 + + movdqa xmm3,xmm5 ; xmm5=in2=z2 + punpcklwd xmm5,xmm0 ; xmm0=in6=z3 + punpckhwd xmm3,xmm0 + pmaddwd xmm5,[rel PW_F184_MF076] ; xmm5=tmp2L + pmaddwd xmm3,[rel PW_F184_MF076] ; xmm3=tmp2H + + movdqa xmm4,xmm1 + movdqa xmm0,xmm2 + paddd xmm1,xmm5 ; xmm1=tmp10L + paddd xmm2,xmm3 ; xmm2=tmp10H + psubd xmm4,xmm5 ; xmm4=tmp12L + psubd xmm0,xmm3 ; xmm0=tmp12H + + ; -- Final output stage + + movdqa xmm5,xmm1 + movdqa xmm3,xmm2 + paddd xmm1,xmm6 ; xmm1=data0L + paddd xmm2,xmm7 ; xmm2=data0H + psubd xmm5,xmm6 ; xmm5=data3L + psubd xmm3,xmm7 ; xmm3=data3H + + movdqa xmm6,[rel PD_DESCALE_P1_4] ; xmm6=[rel PD_DESCALE_P1_4] + + paddd xmm1,xmm6 + paddd xmm2,xmm6 + psrad xmm1,DESCALE_P1_4 + psrad xmm2,DESCALE_P1_4 + paddd xmm5,xmm6 + paddd xmm3,xmm6 + psrad xmm5,DESCALE_P1_4 + psrad xmm3,DESCALE_P1_4 + + packssdw xmm1,xmm2 ; xmm1=data0=(00 01 02 03 04 05 06 07) + packssdw xmm5,xmm3 ; xmm5=data3=(30 31 32 33 34 35 36 37) + + movdqa xmm7, XMMWORD [wk(0)] ; xmm7=tmp0L + movdqa xmm6, XMMWORD [wk(1)] ; xmm6=tmp0H + + movdqa xmm2,xmm4 + movdqa xmm3,xmm0 + paddd xmm4,xmm7 ; xmm4=data1L + paddd xmm0,xmm6 ; xmm0=data1H + psubd xmm2,xmm7 ; xmm2=data2L + psubd xmm3,xmm6 ; xmm3=data2H + + movdqa xmm7,[rel PD_DESCALE_P1_4] ; xmm7=[rel PD_DESCALE_P1_4] + + paddd xmm4,xmm7 + paddd xmm0,xmm7 + psrad xmm4,DESCALE_P1_4 + psrad xmm0,DESCALE_P1_4 + paddd xmm2,xmm7 + paddd xmm3,xmm7 + psrad xmm2,DESCALE_P1_4 + psrad xmm3,DESCALE_P1_4 + + packssdw xmm4,xmm0 ; xmm4=data1=(10 11 12 13 14 15 16 17) + packssdw xmm2,xmm3 ; xmm2=data2=(20 21 22 23 24 25 26 27) + + movdqa xmm6,xmm1 ; transpose coefficients(phase 1) + punpcklwd xmm1,xmm4 ; xmm1=(00 10 01 11 02 12 03 13) + punpckhwd xmm6,xmm4 ; xmm6=(04 14 05 15 06 16 07 17) + movdqa xmm7,xmm2 ; transpose coefficients(phase 1) + punpcklwd xmm2,xmm5 ; xmm2=(20 30 21 31 22 32 23 33) + punpckhwd xmm7,xmm5 ; xmm7=(24 34 25 35 26 36 27 37) + + movdqa xmm0,xmm1 ; transpose coefficients(phase 2) + punpckldq xmm1,xmm2 ; xmm1=[col0 col1]=(00 10 20 30 01 11 21 31) + punpckhdq xmm0,xmm2 ; xmm0=[col2 col3]=(02 12 22 32 03 13 23 33) + movdqa xmm3,xmm6 ; transpose coefficients(phase 2) + punpckldq xmm6,xmm7 ; xmm6=[col4 col5]=(04 14 24 34 05 15 25 35) + punpckhdq xmm3,xmm7 ; xmm3=[col6 col7]=(06 16 26 36 07 17 27 37) +.column_end: + + ; -- Prefetch the next coefficient block + + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32] + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32] + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32] + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32] + + ; ---- Pass 2: process rows, store into output array. + + mov rax, [original_rbp] + mov rdi, r12 ; (JSAMPROW *) + mov rax, r13 + + ; -- Even part + + pxor xmm4,xmm4 + punpcklwd xmm4,xmm1 ; xmm4=tmp0 + psrad xmm4,(16-CONST_BITS-1) ; psrad xmm4,16 & pslld xmm4,CONST_BITS+1 + + ; -- Odd part + + punpckhwd xmm1,xmm0 + punpckhwd xmm6,xmm3 + movdqa xmm5,xmm1 + movdqa xmm2,xmm6 + pmaddwd xmm1,[rel PW_F256_F089] ; xmm1=(tmp2) + pmaddwd xmm6,[rel PW_MF060_MF050] ; xmm6=(tmp2) + pmaddwd xmm5,[rel PW_F106_MF217] ; xmm5=(tmp0) + pmaddwd xmm2,[rel PW_F145_MF021] ; xmm2=(tmp0) + + paddd xmm6,xmm1 ; xmm6=tmp2 + paddd xmm2,xmm5 ; xmm2=tmp0 + + ; -- Even part + + punpcklwd xmm0,xmm3 + pmaddwd xmm0,[rel PW_F184_MF076] ; xmm0=tmp2 + + movdqa xmm7,xmm4 + paddd xmm4,xmm0 ; xmm4=tmp10 + psubd xmm7,xmm0 ; xmm7=tmp12 + + ; -- Final output stage + + movdqa xmm1,[rel PD_DESCALE_P2_4] ; xmm1=[rel PD_DESCALE_P2_4] + + movdqa xmm5,xmm4 + movdqa xmm3,xmm7 + paddd xmm4,xmm6 ; xmm4=data0=(00 10 20 30) + paddd xmm7,xmm2 ; xmm7=data1=(01 11 21 31) + psubd xmm5,xmm6 ; xmm5=data3=(03 13 23 33) + psubd xmm3,xmm2 ; xmm3=data2=(02 12 22 32) + + paddd xmm4,xmm1 + paddd xmm7,xmm1 + psrad xmm4,DESCALE_P2_4 + psrad xmm7,DESCALE_P2_4 + paddd xmm5,xmm1 + paddd xmm3,xmm1 + psrad xmm5,DESCALE_P2_4 + psrad xmm3,DESCALE_P2_4 + + packssdw xmm4,xmm3 ; xmm4=(00 10 20 30 02 12 22 32) + packssdw xmm7,xmm5 ; xmm7=(01 11 21 31 03 13 23 33) + + movdqa xmm0,xmm4 ; transpose coefficients(phase 1) + punpcklwd xmm4,xmm7 ; xmm4=(00 01 10 11 20 21 30 31) + punpckhwd xmm0,xmm7 ; xmm0=(02 03 12 13 22 23 32 33) + + movdqa xmm6,xmm4 ; transpose coefficients(phase 2) + punpckldq xmm4,xmm0 ; xmm4=(00 01 02 03 10 11 12 13) + punpckhdq xmm6,xmm0 ; xmm6=(20 21 22 23 30 31 32 33) + + packsswb xmm4,xmm6 ; xmm4=(00 01 02 03 10 11 12 13 20 ..) + paddb xmm4,[rel PB_CENTERJSAMP] + + pshufd xmm2,xmm4,0x39 ; xmm2=(10 11 12 13 20 21 22 23 30 ..) + pshufd xmm1,xmm4,0x4E ; xmm1=(20 21 22 23 30 31 32 33 00 ..) + pshufd xmm3,xmm4,0x93 ; xmm3=(30 31 32 33 00 01 02 03 10 ..) + + mov rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW] + mov rsi, JSAMPROW [rdi+1*SIZEOF_JSAMPROW] + movd XMM_DWORD [rdx+rax*SIZEOF_JSAMPLE], xmm4 + movd XMM_DWORD [rsi+rax*SIZEOF_JSAMPLE], xmm2 + mov rdx, JSAMPROW [rdi+2*SIZEOF_JSAMPROW] + mov rsi, JSAMPROW [rdi+3*SIZEOF_JSAMPROW] + movd XMM_DWORD [rdx+rax*SIZEOF_JSAMPLE], xmm1 + movd XMM_DWORD [rsi+rax*SIZEOF_JSAMPLE], xmm3 + + uncollect_args + mov rsp,rbp ; rsp <- aligned rbp + pop rsp ; rsp <- original rbp + pop rbp + ret + + +; -------------------------------------------------------------------------- +; +; Perform dequantization and inverse DCT on one block of coefficients, +; producing a reduced-size 2x2 output block. +; +; GLOBAL(void) +; jsimd_idct_2x2_sse2 (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +; r10 = void * dct_table +; r11 = JCOEFPTR coef_block +; r12 = JSAMPARRAY output_buf +; r13 = JDIMENSION output_col + + align 16 + global EXTN(jsimd_idct_2x2_sse2) + +EXTN(jsimd_idct_2x2_sse2): + push rbp + mov rax,rsp + mov rbp,rsp + collect_args + push rbx + + ; ---- Pass 1: process columns from input. + + mov rdx, r10 ; quantptr + mov rsi, r11 ; inptr + + ; | input: | result: | + ; | 00 01 ** 03 ** 05 ** 07 | | + ; | 10 11 ** 13 ** 15 ** 17 | | + ; | ** ** ** ** ** ** ** ** | | + ; | 30 31 ** 33 ** 35 ** 37 | A0 A1 A3 A5 A7 | + ; | ** ** ** ** ** ** ** ** | B0 B1 B3 B5 B7 | + ; | 50 51 ** 53 ** 55 ** 57 | | + ; | ** ** ** ** ** ** ** ** | | + ; | 70 71 ** 73 ** 75 ** 77 | | + + ; -- Odd part + + movdqa xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)] + pmullw xmm0, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm1, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + movdqa xmm2, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)] + movdqa xmm3, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)] + pmullw xmm2, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + + ; xmm0=(10 11 ** 13 ** 15 ** 17), xmm1=(30 31 ** 33 ** 35 ** 37) + ; xmm2=(50 51 ** 53 ** 55 ** 57), xmm3=(70 71 ** 73 ** 75 ** 77) + + pcmpeqd xmm7,xmm7 + pslld xmm7,WORD_BIT ; xmm7={0x0000 0xFFFF 0x0000 0xFFFF ..} + + movdqa xmm4,xmm0 ; xmm4=(10 11 ** 13 ** 15 ** 17) + movdqa xmm5,xmm2 ; xmm5=(50 51 ** 53 ** 55 ** 57) + punpcklwd xmm4,xmm1 ; xmm4=(10 30 11 31 ** ** 13 33) + punpcklwd xmm5,xmm3 ; xmm5=(50 70 51 71 ** ** 53 73) + pmaddwd xmm4,[rel PW_F362_MF127] + pmaddwd xmm5,[rel PW_F085_MF072] + + psrld xmm0,WORD_BIT ; xmm0=(11 -- 13 -- 15 -- 17 --) + pand xmm1,xmm7 ; xmm1=(-- 31 -- 33 -- 35 -- 37) + psrld xmm2,WORD_BIT ; xmm2=(51 -- 53 -- 55 -- 57 --) + pand xmm3,xmm7 ; xmm3=(-- 71 -- 73 -- 75 -- 77) + por xmm0,xmm1 ; xmm0=(11 31 13 33 15 35 17 37) + por xmm2,xmm3 ; xmm2=(51 71 53 73 55 75 57 77) + pmaddwd xmm0,[rel PW_F362_MF127] + pmaddwd xmm2,[rel PW_F085_MF072] + + paddd xmm4,xmm5 ; xmm4=tmp0[col0 col1 **** col3] + paddd xmm0,xmm2 ; xmm0=tmp0[col1 col3 col5 col7] + + ; -- Even part + + movdqa xmm6, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)] + pmullw xmm6, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)] + + ; xmm6=(00 01 ** 03 ** 05 ** 07) + + movdqa xmm1,xmm6 ; xmm1=(00 01 ** 03 ** 05 ** 07) + pslld xmm6,WORD_BIT ; xmm6=(-- 00 -- ** -- ** -- **) + pand xmm1,xmm7 ; xmm1=(-- 01 -- 03 -- 05 -- 07) + psrad xmm6,(WORD_BIT-CONST_BITS-2) ; xmm6=tmp10[col0 **** **** ****] + psrad xmm1,(WORD_BIT-CONST_BITS-2) ; xmm1=tmp10[col1 col3 col5 col7] + + ; -- Final output stage + + movdqa xmm3,xmm6 + movdqa xmm5,xmm1 + paddd xmm6,xmm4 ; xmm6=data0[col0 **** **** ****]=(A0 ** ** **) + paddd xmm1,xmm0 ; xmm1=data0[col1 col3 col5 col7]=(A1 A3 A5 A7) + psubd xmm3,xmm4 ; xmm3=data1[col0 **** **** ****]=(B0 ** ** **) + psubd xmm5,xmm0 ; xmm5=data1[col1 col3 col5 col7]=(B1 B3 B5 B7) + + movdqa xmm2,[rel PD_DESCALE_P1_2] ; xmm2=[rel PD_DESCALE_P1_2] + + punpckldq xmm6,xmm3 ; xmm6=(A0 B0 ** **) + + movdqa xmm7,xmm1 + punpcklqdq xmm1,xmm5 ; xmm1=(A1 A3 B1 B3) + punpckhqdq xmm7,xmm5 ; xmm7=(A5 A7 B5 B7) + + paddd xmm6,xmm2 + psrad xmm6,DESCALE_P1_2 + + paddd xmm1,xmm2 + paddd xmm7,xmm2 + psrad xmm1,DESCALE_P1_2 + psrad xmm7,DESCALE_P1_2 + + ; -- Prefetch the next coefficient block + + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32] + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32] + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32] + prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32] + + ; ---- Pass 2: process rows, store into output array. + + mov rdi, r12 ; (JSAMPROW *) + mov rax, r13 + + ; | input:| result:| + ; | A0 B0 | | + ; | A1 B1 | C0 C1 | + ; | A3 B3 | D0 D1 | + ; | A5 B5 | | + ; | A7 B7 | | + + ; -- Odd part + + packssdw xmm1,xmm1 ; xmm1=(A1 A3 B1 B3 A1 A3 B1 B3) + packssdw xmm7,xmm7 ; xmm7=(A5 A7 B5 B7 A5 A7 B5 B7) + pmaddwd xmm1,[rel PW_F362_MF127] + pmaddwd xmm7,[rel PW_F085_MF072] + + paddd xmm1,xmm7 ; xmm1=tmp0[row0 row1 row0 row1] + + ; -- Even part + + pslld xmm6,(CONST_BITS+2) ; xmm6=tmp10[row0 row1 **** ****] + + ; -- Final output stage + + movdqa xmm4,xmm6 + paddd xmm6,xmm1 ; xmm6=data0[row0 row1 **** ****]=(C0 C1 ** **) + psubd xmm4,xmm1 ; xmm4=data1[row0 row1 **** ****]=(D0 D1 ** **) + + punpckldq xmm6,xmm4 ; xmm6=(C0 D0 C1 D1) + + paddd xmm6,[rel PD_DESCALE_P2_2] + psrad xmm6,DESCALE_P2_2 + + packssdw xmm6,xmm6 ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1) + packsswb xmm6,xmm6 ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1 ..) + paddb xmm6,[rel PB_CENTERJSAMP] + + pextrw ebx,xmm6,0x00 ; ebx=(C0 D0 -- --) + pextrw ecx,xmm6,0x01 ; ecx=(C1 D1 -- --) + + mov rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW] + mov rsi, JSAMPROW [rdi+1*SIZEOF_JSAMPROW] + mov WORD [rdx+rax*SIZEOF_JSAMPLE], bx + mov WORD [rsi+rax*SIZEOF_JSAMPLE], cx + + pop rbx + uncollect_args + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jidctred-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jidctred-sse2.asm new file mode 100644 index 000000000..06dade8a4 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jidctred-sse2.asm @@ -0,0 +1,594 @@ +; +; jidctred.asm - reduced-size IDCT (SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; This file contains inverse-DCT routines that produce reduced-size +; output: either 4x4 or 2x2 pixels from an 8x8 DCT block. +; The following code is based directly on the IJG's original jidctred.c; +; see the jidctred.c for more details. +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + +%define CONST_BITS 13 +%define PASS1_BITS 2 + +%define DESCALE_P1_4 (CONST_BITS-PASS1_BITS+1) +%define DESCALE_P2_4 (CONST_BITS+PASS1_BITS+3+1) +%define DESCALE_P1_2 (CONST_BITS-PASS1_BITS+2) +%define DESCALE_P2_2 (CONST_BITS+PASS1_BITS+3+2) + +%if CONST_BITS == 13 +F_0_211 equ 1730 ; FIX(0.211164243) +F_0_509 equ 4176 ; FIX(0.509795579) +F_0_601 equ 4926 ; FIX(0.601344887) +F_0_720 equ 5906 ; FIX(0.720959822) +F_0_765 equ 6270 ; FIX(0.765366865) +F_0_850 equ 6967 ; FIX(0.850430095) +F_0_899 equ 7373 ; FIX(0.899976223) +F_1_061 equ 8697 ; FIX(1.061594337) +F_1_272 equ 10426 ; FIX(1.272758580) +F_1_451 equ 11893 ; FIX(1.451774981) +F_1_847 equ 15137 ; FIX(1.847759065) +F_2_172 equ 17799 ; FIX(2.172734803) +F_2_562 equ 20995 ; FIX(2.562915447) +F_3_624 equ 29692 ; FIX(3.624509785) +%else +; NASM cannot do compile-time arithmetic on floating-point constants. +%define DESCALE(x,n) (((x)+(1<<((n)-1)))>>(n)) +F_0_211 equ DESCALE( 226735879,30-CONST_BITS) ; FIX(0.211164243) +F_0_509 equ DESCALE( 547388834,30-CONST_BITS) ; FIX(0.509795579) +F_0_601 equ DESCALE( 645689155,30-CONST_BITS) ; FIX(0.601344887) +F_0_720 equ DESCALE( 774124714,30-CONST_BITS) ; FIX(0.720959822) +F_0_765 equ DESCALE( 821806413,30-CONST_BITS) ; FIX(0.765366865) +F_0_850 equ DESCALE( 913142361,30-CONST_BITS) ; FIX(0.850430095) +F_0_899 equ DESCALE( 966342111,30-CONST_BITS) ; FIX(0.899976223) +F_1_061 equ DESCALE(1139878239,30-CONST_BITS) ; FIX(1.061594337) +F_1_272 equ DESCALE(1366614119,30-CONST_BITS) ; FIX(1.272758580) +F_1_451 equ DESCALE(1558831516,30-CONST_BITS) ; FIX(1.451774981) +F_1_847 equ DESCALE(1984016188,30-CONST_BITS) ; FIX(1.847759065) +F_2_172 equ DESCALE(2332956230,30-CONST_BITS) ; FIX(2.172734803) +F_2_562 equ DESCALE(2751909506,30-CONST_BITS) ; FIX(2.562915447) +F_3_624 equ DESCALE(3891787747,30-CONST_BITS) ; FIX(3.624509785) +%endif + +; -------------------------------------------------------------------------- + SECTION SEG_CONST + + alignz 16 + global EXTN(jconst_idct_red_sse2) + +EXTN(jconst_idct_red_sse2): + +PW_F184_MF076 times 4 dw F_1_847,-F_0_765 +PW_F256_F089 times 4 dw F_2_562, F_0_899 +PW_F106_MF217 times 4 dw F_1_061,-F_2_172 +PW_MF060_MF050 times 4 dw -F_0_601,-F_0_509 +PW_F145_MF021 times 4 dw F_1_451,-F_0_211 +PW_F362_MF127 times 4 dw F_3_624,-F_1_272 +PW_F085_MF072 times 4 dw F_0_850,-F_0_720 +PD_DESCALE_P1_4 times 4 dd 1 << (DESCALE_P1_4-1) +PD_DESCALE_P2_4 times 4 dd 1 << (DESCALE_P2_4-1) +PD_DESCALE_P1_2 times 4 dd 1 << (DESCALE_P1_2-1) +PD_DESCALE_P2_2 times 4 dd 1 << (DESCALE_P2_2-1) +PB_CENTERJSAMP times 16 db CENTERJSAMPLE + + alignz 16 + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Perform dequantization and inverse DCT on one block of coefficients, +; producing a reduced-size 4x4 output block. +; +; GLOBAL(void) +; jsimd_idct_4x4_sse2 (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +%define dct_table(b) (b)+8 ; void * dct_table +%define coef_block(b) (b)+12 ; JCOEFPTR coef_block +%define output_buf(b) (b)+16 ; JSAMPARRAY output_buf +%define output_col(b) (b)+20 ; JDIMENSION output_col + +%define original_ebp ebp+0 +%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] +%define WK_NUM 2 + + align 16 + global EXTN(jsimd_idct_4x4_sse2) + +EXTN(jsimd_idct_4x4_sse2): + push ebp + mov eax,esp ; eax = original ebp + sub esp, byte 4 + and esp, byte (-SIZEOF_XMMWORD) ; align to 128 bits + mov [esp],eax + mov ebp,esp ; ebp = aligned ebp + lea esp, [wk(0)] + pushpic ebx +; push ecx ; unused +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process columns from input. + +; mov eax, [original_ebp] + mov edx, POINTER [dct_table(eax)] ; quantptr + mov esi, JCOEFPTR [coef_block(eax)] ; inptr + +%ifndef NO_ZERO_COLUMN_TEST_4X4_SSE2 + mov eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)] + or eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)] + jnz short .columnDCT + + movdqa xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)] + por xmm1, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)] + por xmm0, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)] + por xmm1, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)] + por xmm0,xmm1 + packsswb xmm0,xmm0 + packsswb xmm0,xmm0 + movd eax,xmm0 + test eax,eax + jnz short .columnDCT + + ; -- AC terms all zero + + movdqa xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)] + pmullw xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + psllw xmm0,PASS1_BITS + + movdqa xmm3,xmm0 ; xmm0=in0=(00 01 02 03 04 05 06 07) + punpcklwd xmm0,xmm0 ; xmm0=(00 00 01 01 02 02 03 03) + punpckhwd xmm3,xmm3 ; xmm3=(04 04 05 05 06 06 07 07) + + pshufd xmm1,xmm0,0x50 ; xmm1=[col0 col1]=(00 00 00 00 01 01 01 01) + pshufd xmm0,xmm0,0xFA ; xmm0=[col2 col3]=(02 02 02 02 03 03 03 03) + pshufd xmm6,xmm3,0x50 ; xmm6=[col4 col5]=(04 04 04 04 05 05 05 05) + pshufd xmm3,xmm3,0xFA ; xmm3=[col6 col7]=(06 06 06 06 07 07 07 07) + + jmp near .column_end + alignx 16,7 +%endif +.columnDCT: + + ; -- Odd part + + movdqa xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)] + pmullw xmm0, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm1, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + movdqa xmm2, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movdqa xmm3, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)] + pmullw xmm2, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + movdqa xmm4,xmm0 + movdqa xmm5,xmm0 + punpcklwd xmm4,xmm1 + punpckhwd xmm5,xmm1 + movdqa xmm0,xmm4 + movdqa xmm1,xmm5 + pmaddwd xmm4,[GOTOFF(ebx,PW_F256_F089)] ; xmm4=(tmp2L) + pmaddwd xmm5,[GOTOFF(ebx,PW_F256_F089)] ; xmm5=(tmp2H) + pmaddwd xmm0,[GOTOFF(ebx,PW_F106_MF217)] ; xmm0=(tmp0L) + pmaddwd xmm1,[GOTOFF(ebx,PW_F106_MF217)] ; xmm1=(tmp0H) + + movdqa xmm6,xmm2 + movdqa xmm7,xmm2 + punpcklwd xmm6,xmm3 + punpckhwd xmm7,xmm3 + movdqa xmm2,xmm6 + movdqa xmm3,xmm7 + pmaddwd xmm6,[GOTOFF(ebx,PW_MF060_MF050)] ; xmm6=(tmp2L) + pmaddwd xmm7,[GOTOFF(ebx,PW_MF060_MF050)] ; xmm7=(tmp2H) + pmaddwd xmm2,[GOTOFF(ebx,PW_F145_MF021)] ; xmm2=(tmp0L) + pmaddwd xmm3,[GOTOFF(ebx,PW_F145_MF021)] ; xmm3=(tmp0H) + + paddd xmm6,xmm4 ; xmm6=tmp2L + paddd xmm7,xmm5 ; xmm7=tmp2H + paddd xmm2,xmm0 ; xmm2=tmp0L + paddd xmm3,xmm1 ; xmm3=tmp0H + + movdqa XMMWORD [wk(0)], xmm2 ; wk(0)=tmp0L + movdqa XMMWORD [wk(1)], xmm3 ; wk(1)=tmp0H + + ; -- Even part + + movdqa xmm4, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)] + movdqa xmm5, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)] + movdqa xmm0, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)] + pmullw xmm4, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm5, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm0, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + pxor xmm1,xmm1 + pxor xmm2,xmm2 + punpcklwd xmm1,xmm4 ; xmm1=tmp0L + punpckhwd xmm2,xmm4 ; xmm2=tmp0H + psrad xmm1,(16-CONST_BITS-1) ; psrad xmm1,16 & pslld xmm1,CONST_BITS+1 + psrad xmm2,(16-CONST_BITS-1) ; psrad xmm2,16 & pslld xmm2,CONST_BITS+1 + + movdqa xmm3,xmm5 ; xmm5=in2=z2 + punpcklwd xmm5,xmm0 ; xmm0=in6=z3 + punpckhwd xmm3,xmm0 + pmaddwd xmm5,[GOTOFF(ebx,PW_F184_MF076)] ; xmm5=tmp2L + pmaddwd xmm3,[GOTOFF(ebx,PW_F184_MF076)] ; xmm3=tmp2H + + movdqa xmm4,xmm1 + movdqa xmm0,xmm2 + paddd xmm1,xmm5 ; xmm1=tmp10L + paddd xmm2,xmm3 ; xmm2=tmp10H + psubd xmm4,xmm5 ; xmm4=tmp12L + psubd xmm0,xmm3 ; xmm0=tmp12H + + ; -- Final output stage + + movdqa xmm5,xmm1 + movdqa xmm3,xmm2 + paddd xmm1,xmm6 ; xmm1=data0L + paddd xmm2,xmm7 ; xmm2=data0H + psubd xmm5,xmm6 ; xmm5=data3L + psubd xmm3,xmm7 ; xmm3=data3H + + movdqa xmm6,[GOTOFF(ebx,PD_DESCALE_P1_4)] ; xmm6=[PD_DESCALE_P1_4] + + paddd xmm1,xmm6 + paddd xmm2,xmm6 + psrad xmm1,DESCALE_P1_4 + psrad xmm2,DESCALE_P1_4 + paddd xmm5,xmm6 + paddd xmm3,xmm6 + psrad xmm5,DESCALE_P1_4 + psrad xmm3,DESCALE_P1_4 + + packssdw xmm1,xmm2 ; xmm1=data0=(00 01 02 03 04 05 06 07) + packssdw xmm5,xmm3 ; xmm5=data3=(30 31 32 33 34 35 36 37) + + movdqa xmm7, XMMWORD [wk(0)] ; xmm7=tmp0L + movdqa xmm6, XMMWORD [wk(1)] ; xmm6=tmp0H + + movdqa xmm2,xmm4 + movdqa xmm3,xmm0 + paddd xmm4,xmm7 ; xmm4=data1L + paddd xmm0,xmm6 ; xmm0=data1H + psubd xmm2,xmm7 ; xmm2=data2L + psubd xmm3,xmm6 ; xmm3=data2H + + movdqa xmm7,[GOTOFF(ebx,PD_DESCALE_P1_4)] ; xmm7=[PD_DESCALE_P1_4] + + paddd xmm4,xmm7 + paddd xmm0,xmm7 + psrad xmm4,DESCALE_P1_4 + psrad xmm0,DESCALE_P1_4 + paddd xmm2,xmm7 + paddd xmm3,xmm7 + psrad xmm2,DESCALE_P1_4 + psrad xmm3,DESCALE_P1_4 + + packssdw xmm4,xmm0 ; xmm4=data1=(10 11 12 13 14 15 16 17) + packssdw xmm2,xmm3 ; xmm2=data2=(20 21 22 23 24 25 26 27) + + movdqa xmm6,xmm1 ; transpose coefficients(phase 1) + punpcklwd xmm1,xmm4 ; xmm1=(00 10 01 11 02 12 03 13) + punpckhwd xmm6,xmm4 ; xmm6=(04 14 05 15 06 16 07 17) + movdqa xmm7,xmm2 ; transpose coefficients(phase 1) + punpcklwd xmm2,xmm5 ; xmm2=(20 30 21 31 22 32 23 33) + punpckhwd xmm7,xmm5 ; xmm7=(24 34 25 35 26 36 27 37) + + movdqa xmm0,xmm1 ; transpose coefficients(phase 2) + punpckldq xmm1,xmm2 ; xmm1=[col0 col1]=(00 10 20 30 01 11 21 31) + punpckhdq xmm0,xmm2 ; xmm0=[col2 col3]=(02 12 22 32 03 13 23 33) + movdqa xmm3,xmm6 ; transpose coefficients(phase 2) + punpckldq xmm6,xmm7 ; xmm6=[col4 col5]=(04 14 24 34 05 15 25 35) + punpckhdq xmm3,xmm7 ; xmm3=[col6 col7]=(06 16 26 36 07 17 27 37) +.column_end: + + ; -- Prefetch the next coefficient block + + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 0*32] + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 1*32] + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 2*32] + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 3*32] + + ; ---- Pass 2: process rows, store into output array. + + mov eax, [original_ebp] + mov edi, JSAMPARRAY [output_buf(eax)] ; (JSAMPROW *) + mov eax, JDIMENSION [output_col(eax)] + + ; -- Even part + + pxor xmm4,xmm4 + punpcklwd xmm4,xmm1 ; xmm4=tmp0 + psrad xmm4,(16-CONST_BITS-1) ; psrad xmm4,16 & pslld xmm4,CONST_BITS+1 + + ; -- Odd part + + punpckhwd xmm1,xmm0 + punpckhwd xmm6,xmm3 + movdqa xmm5,xmm1 + movdqa xmm2,xmm6 + pmaddwd xmm1,[GOTOFF(ebx,PW_F256_F089)] ; xmm1=(tmp2) + pmaddwd xmm6,[GOTOFF(ebx,PW_MF060_MF050)] ; xmm6=(tmp2) + pmaddwd xmm5,[GOTOFF(ebx,PW_F106_MF217)] ; xmm5=(tmp0) + pmaddwd xmm2,[GOTOFF(ebx,PW_F145_MF021)] ; xmm2=(tmp0) + + paddd xmm6,xmm1 ; xmm6=tmp2 + paddd xmm2,xmm5 ; xmm2=tmp0 + + ; -- Even part + + punpcklwd xmm0,xmm3 + pmaddwd xmm0,[GOTOFF(ebx,PW_F184_MF076)] ; xmm0=tmp2 + + movdqa xmm7,xmm4 + paddd xmm4,xmm0 ; xmm4=tmp10 + psubd xmm7,xmm0 ; xmm7=tmp12 + + ; -- Final output stage + + movdqa xmm1,[GOTOFF(ebx,PD_DESCALE_P2_4)] ; xmm1=[PD_DESCALE_P2_4] + + movdqa xmm5,xmm4 + movdqa xmm3,xmm7 + paddd xmm4,xmm6 ; xmm4=data0=(00 10 20 30) + paddd xmm7,xmm2 ; xmm7=data1=(01 11 21 31) + psubd xmm5,xmm6 ; xmm5=data3=(03 13 23 33) + psubd xmm3,xmm2 ; xmm3=data2=(02 12 22 32) + + paddd xmm4,xmm1 + paddd xmm7,xmm1 + psrad xmm4,DESCALE_P2_4 + psrad xmm7,DESCALE_P2_4 + paddd xmm5,xmm1 + paddd xmm3,xmm1 + psrad xmm5,DESCALE_P2_4 + psrad xmm3,DESCALE_P2_4 + + packssdw xmm4,xmm3 ; xmm4=(00 10 20 30 02 12 22 32) + packssdw xmm7,xmm5 ; xmm7=(01 11 21 31 03 13 23 33) + + movdqa xmm0,xmm4 ; transpose coefficients(phase 1) + punpcklwd xmm4,xmm7 ; xmm4=(00 01 10 11 20 21 30 31) + punpckhwd xmm0,xmm7 ; xmm0=(02 03 12 13 22 23 32 33) + + movdqa xmm6,xmm4 ; transpose coefficients(phase 2) + punpckldq xmm4,xmm0 ; xmm4=(00 01 02 03 10 11 12 13) + punpckhdq xmm6,xmm0 ; xmm6=(20 21 22 23 30 31 32 33) + + packsswb xmm4,xmm6 ; xmm4=(00 01 02 03 10 11 12 13 20 ..) + paddb xmm4,[GOTOFF(ebx,PB_CENTERJSAMP)] + + pshufd xmm2,xmm4,0x39 ; xmm2=(10 11 12 13 20 21 22 23 30 ..) + pshufd xmm1,xmm4,0x4E ; xmm1=(20 21 22 23 30 31 32 33 00 ..) + pshufd xmm3,xmm4,0x93 ; xmm3=(30 31 32 33 00 01 02 03 10 ..) + + mov edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+1*SIZEOF_JSAMPROW] + movd XMM_DWORD [edx+eax*SIZEOF_JSAMPLE], xmm4 + movd XMM_DWORD [esi+eax*SIZEOF_JSAMPLE], xmm2 + mov edx, JSAMPROW [edi+2*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+3*SIZEOF_JSAMPROW] + movd XMM_DWORD [edx+eax*SIZEOF_JSAMPLE], xmm1 + movd XMM_DWORD [esi+eax*SIZEOF_JSAMPLE], xmm3 + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; unused + poppic ebx + mov esp,ebp ; esp <- aligned ebp + pop esp ; esp <- original ebp + pop ebp + ret + + +; -------------------------------------------------------------------------- +; +; Perform dequantization and inverse DCT on one block of coefficients, +; producing a reduced-size 2x2 output block. +; +; GLOBAL(void) +; jsimd_idct_2x2_sse2 (void * dct_table, JCOEFPTR coef_block, +; JSAMPARRAY output_buf, JDIMENSION output_col) +; + +%define dct_table(b) (b)+8 ; void * dct_table +%define coef_block(b) (b)+12 ; JCOEFPTR coef_block +%define output_buf(b) (b)+16 ; JSAMPARRAY output_buf +%define output_col(b) (b)+20 ; JDIMENSION output_col + + align 16 + global EXTN(jsimd_idct_2x2_sse2) + +EXTN(jsimd_idct_2x2_sse2): + push ebp + mov ebp,esp + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + get_GOT ebx ; get GOT address + + ; ---- Pass 1: process columns from input. + + mov edx, POINTER [dct_table(ebp)] ; quantptr + mov esi, JCOEFPTR [coef_block(ebp)] ; inptr + + ; | input: | result: | + ; | 00 01 ** 03 ** 05 ** 07 | | + ; | 10 11 ** 13 ** 15 ** 17 | | + ; | ** ** ** ** ** ** ** ** | | + ; | 30 31 ** 33 ** 35 ** 37 | A0 A1 A3 A5 A7 | + ; | ** ** ** ** ** ** ** ** | B0 B1 B3 B5 B7 | + ; | 50 51 ** 53 ** 55 ** 57 | | + ; | ** ** ** ** ** ** ** ** | | + ; | 70 71 ** 73 ** 75 ** 77 | | + + ; -- Odd part + + movdqa xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)] + movdqa xmm1, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)] + pmullw xmm0, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm1, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + movdqa xmm2, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)] + movdqa xmm3, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)] + pmullw xmm2, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + pmullw xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + ; xmm0=(10 11 ** 13 ** 15 ** 17), xmm1=(30 31 ** 33 ** 35 ** 37) + ; xmm2=(50 51 ** 53 ** 55 ** 57), xmm3=(70 71 ** 73 ** 75 ** 77) + + pcmpeqd xmm7,xmm7 + pslld xmm7,WORD_BIT ; xmm7={0x0000 0xFFFF 0x0000 0xFFFF ..} + + movdqa xmm4,xmm0 ; xmm4=(10 11 ** 13 ** 15 ** 17) + movdqa xmm5,xmm2 ; xmm5=(50 51 ** 53 ** 55 ** 57) + punpcklwd xmm4,xmm1 ; xmm4=(10 30 11 31 ** ** 13 33) + punpcklwd xmm5,xmm3 ; xmm5=(50 70 51 71 ** ** 53 73) + pmaddwd xmm4,[GOTOFF(ebx,PW_F362_MF127)] + pmaddwd xmm5,[GOTOFF(ebx,PW_F085_MF072)] + + psrld xmm0,WORD_BIT ; xmm0=(11 -- 13 -- 15 -- 17 --) + pand xmm1,xmm7 ; xmm1=(-- 31 -- 33 -- 35 -- 37) + psrld xmm2,WORD_BIT ; xmm2=(51 -- 53 -- 55 -- 57 --) + pand xmm3,xmm7 ; xmm3=(-- 71 -- 73 -- 75 -- 77) + por xmm0,xmm1 ; xmm0=(11 31 13 33 15 35 17 37) + por xmm2,xmm3 ; xmm2=(51 71 53 73 55 75 57 77) + pmaddwd xmm0,[GOTOFF(ebx,PW_F362_MF127)] + pmaddwd xmm2,[GOTOFF(ebx,PW_F085_MF072)] + + paddd xmm4,xmm5 ; xmm4=tmp0[col0 col1 **** col3] + paddd xmm0,xmm2 ; xmm0=tmp0[col1 col3 col5 col7] + + ; -- Even part + + movdqa xmm6, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)] + pmullw xmm6, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)] + + ; xmm6=(00 01 ** 03 ** 05 ** 07) + + movdqa xmm1,xmm6 ; xmm1=(00 01 ** 03 ** 05 ** 07) + pslld xmm6,WORD_BIT ; xmm6=(-- 00 -- ** -- ** -- **) + pand xmm1,xmm7 ; xmm1=(-- 01 -- 03 -- 05 -- 07) + psrad xmm6,(WORD_BIT-CONST_BITS-2) ; xmm6=tmp10[col0 **** **** ****] + psrad xmm1,(WORD_BIT-CONST_BITS-2) ; xmm1=tmp10[col1 col3 col5 col7] + + ; -- Final output stage + + movdqa xmm3,xmm6 + movdqa xmm5,xmm1 + paddd xmm6,xmm4 ; xmm6=data0[col0 **** **** ****]=(A0 ** ** **) + paddd xmm1,xmm0 ; xmm1=data0[col1 col3 col5 col7]=(A1 A3 A5 A7) + psubd xmm3,xmm4 ; xmm3=data1[col0 **** **** ****]=(B0 ** ** **) + psubd xmm5,xmm0 ; xmm5=data1[col1 col3 col5 col7]=(B1 B3 B5 B7) + + movdqa xmm2,[GOTOFF(ebx,PD_DESCALE_P1_2)] ; xmm2=[PD_DESCALE_P1_2] + + punpckldq xmm6,xmm3 ; xmm6=(A0 B0 ** **) + + movdqa xmm7,xmm1 + punpcklqdq xmm1,xmm5 ; xmm1=(A1 A3 B1 B3) + punpckhqdq xmm7,xmm5 ; xmm7=(A5 A7 B5 B7) + + paddd xmm6,xmm2 + psrad xmm6,DESCALE_P1_2 + + paddd xmm1,xmm2 + paddd xmm7,xmm2 + psrad xmm1,DESCALE_P1_2 + psrad xmm7,DESCALE_P1_2 + + ; -- Prefetch the next coefficient block + + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 0*32] + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 1*32] + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 2*32] + prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 3*32] + + ; ---- Pass 2: process rows, store into output array. + + mov edi, JSAMPARRAY [output_buf(ebp)] ; (JSAMPROW *) + mov eax, JDIMENSION [output_col(ebp)] + + ; | input:| result:| + ; | A0 B0 | | + ; | A1 B1 | C0 C1 | + ; | A3 B3 | D0 D1 | + ; | A5 B5 | | + ; | A7 B7 | | + + ; -- Odd part + + packssdw xmm1,xmm1 ; xmm1=(A1 A3 B1 B3 A1 A3 B1 B3) + packssdw xmm7,xmm7 ; xmm7=(A5 A7 B5 B7 A5 A7 B5 B7) + pmaddwd xmm1,[GOTOFF(ebx,PW_F362_MF127)] + pmaddwd xmm7,[GOTOFF(ebx,PW_F085_MF072)] + + paddd xmm1,xmm7 ; xmm1=tmp0[row0 row1 row0 row1] + + ; -- Even part + + pslld xmm6,(CONST_BITS+2) ; xmm6=tmp10[row0 row1 **** ****] + + ; -- Final output stage + + movdqa xmm4,xmm6 + paddd xmm6,xmm1 ; xmm6=data0[row0 row1 **** ****]=(C0 C1 ** **) + psubd xmm4,xmm1 ; xmm4=data1[row0 row1 **** ****]=(D0 D1 ** **) + + punpckldq xmm6,xmm4 ; xmm6=(C0 D0 C1 D1) + + paddd xmm6,[GOTOFF(ebx,PD_DESCALE_P2_2)] + psrad xmm6,DESCALE_P2_2 + + packssdw xmm6,xmm6 ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1) + packsswb xmm6,xmm6 ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1 ..) + paddb xmm6,[GOTOFF(ebx,PB_CENTERJSAMP)] + + pextrw ebx,xmm6,0x00 ; ebx=(C0 D0 -- --) + pextrw ecx,xmm6,0x01 ; ecx=(C1 D1 -- --) + + mov edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW] + mov esi, JSAMPROW [edi+1*SIZEOF_JSAMPROW] + mov WORD [edx+eax*SIZEOF_JSAMPLE], bx + mov WORD [esi+eax*SIZEOF_JSAMPLE], cx + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jquant-3dn.asm b/app/src/main/jni/libjpeg-turbo/simd/jquant-3dn.asm new file mode 100644 index 000000000..76e19f70f --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jquant-3dn.asm @@ -0,0 +1,233 @@ +; +; jquant.asm - sample data conversion and quantization (3DNow! & MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Load data into workspace, applying unsigned->signed conversion +; +; GLOBAL(void) +; jsimd_convsamp_float_3dnow (JSAMPARRAY sample_data, JDIMENSION start_col, +; FAST_FLOAT * workspace); +; + +%define sample_data ebp+8 ; JSAMPARRAY sample_data +%define start_col ebp+12 ; JDIMENSION start_col +%define workspace ebp+16 ; FAST_FLOAT * workspace + + align 16 + global EXTN(jsimd_convsamp_float_3dnow) + +EXTN(jsimd_convsamp_float_3dnow): + push ebp + mov ebp,esp + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + pcmpeqw mm7,mm7 + psllw mm7,7 + packsswb mm7,mm7 ; mm7 = PB_CENTERJSAMPLE (0x808080..) + + mov esi, JSAMPARRAY [sample_data] ; (JSAMPROW *) + mov eax, JDIMENSION [start_col] + mov edi, POINTER [workspace] ; (DCTELEM *) + mov ecx, DCTSIZE/2 + alignx 16,7 +.convloop: + mov ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW] ; (JSAMPLE *) + mov edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW] ; (JSAMPLE *) + + movq mm0, MMWORD [ebx+eax*SIZEOF_JSAMPLE] + movq mm1, MMWORD [edx+eax*SIZEOF_JSAMPLE] + + psubb mm0,mm7 ; mm0=(01234567) + psubb mm1,mm7 ; mm1=(89ABCDEF) + + punpcklbw mm2,mm0 ; mm2=(*0*1*2*3) + punpckhbw mm0,mm0 ; mm0=(*4*5*6*7) + punpcklbw mm3,mm1 ; mm3=(*8*9*A*B) + punpckhbw mm1,mm1 ; mm1=(*C*D*E*F) + + punpcklwd mm4,mm2 ; mm4=(***0***1) + punpckhwd mm2,mm2 ; mm2=(***2***3) + punpcklwd mm5,mm0 ; mm5=(***4***5) + punpckhwd mm0,mm0 ; mm0=(***6***7) + + psrad mm4,(DWORD_BIT-BYTE_BIT) ; mm4=(01) + psrad mm2,(DWORD_BIT-BYTE_BIT) ; mm2=(23) + pi2fd mm4,mm4 + pi2fd mm2,mm2 + psrad mm5,(DWORD_BIT-BYTE_BIT) ; mm5=(45) + psrad mm0,(DWORD_BIT-BYTE_BIT) ; mm0=(67) + pi2fd mm5,mm5 + pi2fd mm0,mm0 + + movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], mm4 + movq MMWORD [MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], mm2 + movq MMWORD [MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT)], mm5 + movq MMWORD [MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT)], mm0 + + punpcklwd mm6,mm3 ; mm6=(***8***9) + punpckhwd mm3,mm3 ; mm3=(***A***B) + punpcklwd mm4,mm1 ; mm4=(***C***D) + punpckhwd mm1,mm1 ; mm1=(***E***F) + + psrad mm6,(DWORD_BIT-BYTE_BIT) ; mm6=(89) + psrad mm3,(DWORD_BIT-BYTE_BIT) ; mm3=(AB) + pi2fd mm6,mm6 + pi2fd mm3,mm3 + psrad mm4,(DWORD_BIT-BYTE_BIT) ; mm4=(CD) + psrad mm1,(DWORD_BIT-BYTE_BIT) ; mm1=(EF) + pi2fd mm4,mm4 + pi2fd mm1,mm1 + + movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], mm6 + movq MMWORD [MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], mm3 + movq MMWORD [MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT)], mm4 + movq MMWORD [MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT)], mm1 + + add esi, byte 2*SIZEOF_JSAMPROW + add edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT + dec ecx + jnz near .convloop + + femms ; empty MMX/3DNow! state + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + pop ebp + ret + + +; -------------------------------------------------------------------------- +; +; Quantize/descale the coefficients, and store into coef_block +; +; GLOBAL(void) +; jsimd_quantize_float_3dnow (JCOEFPTR coef_block, FAST_FLOAT * divisors, +; FAST_FLOAT * workspace); +; + +%define coef_block ebp+8 ; JCOEFPTR coef_block +%define divisors ebp+12 ; FAST_FLOAT * divisors +%define workspace ebp+16 ; FAST_FLOAT * workspace + + align 16 + global EXTN(jsimd_quantize_float_3dnow) + +EXTN(jsimd_quantize_float_3dnow): + push ebp + mov ebp,esp +; push ebx ; unused +; push ecx ; unused +; push edx ; need not be preserved + push esi + push edi + + mov eax, 0x4B400000 ; (float)0x00C00000 (rndint_magic) + movd mm7,eax + punpckldq mm7,mm7 ; mm7={12582912.0F 12582912.0F} + + mov esi, POINTER [workspace] + mov edx, POINTER [divisors] + mov edi, JCOEFPTR [coef_block] + mov eax, DCTSIZE2/16 + alignx 16,7 +.quantloop: + movq mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)] + movq mm1, MMWORD [MMBLOCK(0,1,esi,SIZEOF_FAST_FLOAT)] + pfmul mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)] + pfmul mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)] + movq mm2, MMWORD [MMBLOCK(0,2,esi,SIZEOF_FAST_FLOAT)] + movq mm3, MMWORD [MMBLOCK(0,3,esi,SIZEOF_FAST_FLOAT)] + pfmul mm2, MMWORD [MMBLOCK(0,2,edx,SIZEOF_FAST_FLOAT)] + pfmul mm3, MMWORD [MMBLOCK(0,3,edx,SIZEOF_FAST_FLOAT)] + + pfadd mm0,mm7 ; mm0=(00 ** 01 **) + pfadd mm1,mm7 ; mm1=(02 ** 03 **) + pfadd mm2,mm7 ; mm0=(04 ** 05 **) + pfadd mm3,mm7 ; mm1=(06 ** 07 **) + + movq mm4,mm0 + punpcklwd mm0,mm1 ; mm0=(00 02 ** **) + punpckhwd mm4,mm1 ; mm4=(01 03 ** **) + movq mm5,mm2 + punpcklwd mm2,mm3 ; mm2=(04 06 ** **) + punpckhwd mm5,mm3 ; mm5=(05 07 ** **) + + punpcklwd mm0,mm4 ; mm0=(00 01 02 03) + punpcklwd mm2,mm5 ; mm2=(04 05 06 07) + + movq mm6, MMWORD [MMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)] + movq mm1, MMWORD [MMBLOCK(1,1,esi,SIZEOF_FAST_FLOAT)] + pfmul mm6, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)] + pfmul mm1, MMWORD [MMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)] + movq mm3, MMWORD [MMBLOCK(1,2,esi,SIZEOF_FAST_FLOAT)] + movq mm4, MMWORD [MMBLOCK(1,3,esi,SIZEOF_FAST_FLOAT)] + pfmul mm3, MMWORD [MMBLOCK(1,2,edx,SIZEOF_FAST_FLOAT)] + pfmul mm4, MMWORD [MMBLOCK(1,3,edx,SIZEOF_FAST_FLOAT)] + + pfadd mm6,mm7 ; mm0=(10 ** 11 **) + pfadd mm1,mm7 ; mm4=(12 ** 13 **) + pfadd mm3,mm7 ; mm0=(14 ** 15 **) + pfadd mm4,mm7 ; mm4=(16 ** 17 **) + + movq mm5,mm6 + punpcklwd mm6,mm1 ; mm6=(10 12 ** **) + punpckhwd mm5,mm1 ; mm5=(11 13 ** **) + movq mm1,mm3 + punpcklwd mm3,mm4 ; mm3=(14 16 ** **) + punpckhwd mm1,mm4 ; mm1=(15 17 ** **) + + punpcklwd mm6,mm5 ; mm6=(10 11 12 13) + punpcklwd mm3,mm1 ; mm3=(14 15 16 17) + + movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0 + movq MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm2 + movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm6 + movq MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm3 + + add esi, byte 16*SIZEOF_FAST_FLOAT + add edx, byte 16*SIZEOF_FAST_FLOAT + add edi, byte 16*SIZEOF_JCOEF + dec eax + jnz near .quantloop + + femms ; empty MMX/3DNow! state + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; unused +; pop ebx ; unused + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jquant-mmx.asm b/app/src/main/jni/libjpeg-turbo/simd/jquant-mmx.asm new file mode 100644 index 000000000..822c7ee57 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jquant-mmx.asm @@ -0,0 +1,274 @@ +; +; jquant.asm - sample data conversion and quantization (MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Load data into workspace, applying unsigned->signed conversion +; +; GLOBAL(void) +; jsimd_convsamp_mmx (JSAMPARRAY sample_data, JDIMENSION start_col, +; DCTELEM * workspace); +; + +%define sample_data ebp+8 ; JSAMPARRAY sample_data +%define start_col ebp+12 ; JDIMENSION start_col +%define workspace ebp+16 ; DCTELEM * workspace + + align 16 + global EXTN(jsimd_convsamp_mmx) + +EXTN(jsimd_convsamp_mmx): + push ebp + mov ebp,esp + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + pxor mm6,mm6 ; mm6=(all 0's) + pcmpeqw mm7,mm7 + psllw mm7,7 ; mm7={0xFF80 0xFF80 0xFF80 0xFF80} + + mov esi, JSAMPARRAY [sample_data] ; (JSAMPROW *) + mov eax, JDIMENSION [start_col] + mov edi, POINTER [workspace] ; (DCTELEM *) + mov ecx, DCTSIZE/4 + alignx 16,7 +.convloop: + mov ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW] ; (JSAMPLE *) + mov edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW] ; (JSAMPLE *) + + movq mm0, MMWORD [ebx+eax*SIZEOF_JSAMPLE] ; mm0=(01234567) + movq mm1, MMWORD [edx+eax*SIZEOF_JSAMPLE] ; mm1=(89ABCDEF) + + mov ebx, JSAMPROW [esi+2*SIZEOF_JSAMPROW] ; (JSAMPLE *) + mov edx, JSAMPROW [esi+3*SIZEOF_JSAMPROW] ; (JSAMPLE *) + + movq mm2, MMWORD [ebx+eax*SIZEOF_JSAMPLE] ; mm2=(GHIJKLMN) + movq mm3, MMWORD [edx+eax*SIZEOF_JSAMPLE] ; mm3=(OPQRSTUV) + + movq mm4,mm0 + punpcklbw mm0,mm6 ; mm0=(0123) + punpckhbw mm4,mm6 ; mm4=(4567) + movq mm5,mm1 + punpcklbw mm1,mm6 ; mm1=(89AB) + punpckhbw mm5,mm6 ; mm5=(CDEF) + + paddw mm0,mm7 + paddw mm4,mm7 + paddw mm1,mm7 + paddw mm5,mm7 + + movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_DCTELEM)], mm0 + movq MMWORD [MMBLOCK(0,1,edi,SIZEOF_DCTELEM)], mm4 + movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_DCTELEM)], mm1 + movq MMWORD [MMBLOCK(1,1,edi,SIZEOF_DCTELEM)], mm5 + + movq mm0,mm2 + punpcklbw mm2,mm6 ; mm2=(GHIJ) + punpckhbw mm0,mm6 ; mm0=(KLMN) + movq mm4,mm3 + punpcklbw mm3,mm6 ; mm3=(OPQR) + punpckhbw mm4,mm6 ; mm4=(STUV) + + paddw mm2,mm7 + paddw mm0,mm7 + paddw mm3,mm7 + paddw mm4,mm7 + + movq MMWORD [MMBLOCK(2,0,edi,SIZEOF_DCTELEM)], mm2 + movq MMWORD [MMBLOCK(2,1,edi,SIZEOF_DCTELEM)], mm0 + movq MMWORD [MMBLOCK(3,0,edi,SIZEOF_DCTELEM)], mm3 + movq MMWORD [MMBLOCK(3,1,edi,SIZEOF_DCTELEM)], mm4 + + add esi, byte 4*SIZEOF_JSAMPROW + add edi, byte 4*DCTSIZE*SIZEOF_DCTELEM + dec ecx + jnz short .convloop + + emms ; empty MMX state + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + pop ebp + ret + +; -------------------------------------------------------------------------- +; +; Quantize/descale the coefficients, and store into coef_block +; +; This implementation is based on an algorithm described in +; "How to optimize for the Pentium family of microprocessors" +; (http://www.agner.org/assem/). +; +; GLOBAL(void) +; jsimd_quantize_mmx (JCOEFPTR coef_block, DCTELEM * divisors, +; DCTELEM * workspace); +; + +%define RECIPROCAL(m,n,b) MMBLOCK(DCTSIZE*0+(m),(n),(b),SIZEOF_DCTELEM) +%define CORRECTION(m,n,b) MMBLOCK(DCTSIZE*1+(m),(n),(b),SIZEOF_DCTELEM) +%define SCALE(m,n,b) MMBLOCK(DCTSIZE*2+(m),(n),(b),SIZEOF_DCTELEM) +%define SHIFT(m,n,b) MMBLOCK(DCTSIZE*3+(m),(n),(b),SIZEOF_DCTELEM) + +%define coef_block ebp+8 ; JCOEFPTR coef_block +%define divisors ebp+12 ; DCTELEM * divisors +%define workspace ebp+16 ; DCTELEM * workspace + + align 16 + global EXTN(jsimd_quantize_mmx) + +EXTN(jsimd_quantize_mmx): + push ebp + mov ebp,esp +; push ebx ; unused +; push ecx ; unused +; push edx ; need not be preserved + push esi + push edi + + mov esi, POINTER [workspace] + mov edx, POINTER [divisors] + mov edi, JCOEFPTR [coef_block] + mov ah, 2 + alignx 16,7 +.quantloop1: + mov al, DCTSIZE2/8/2 + alignx 16,7 +.quantloop2: + movq mm2, MMWORD [MMBLOCK(0,0,esi,SIZEOF_DCTELEM)] + movq mm3, MMWORD [MMBLOCK(0,1,esi,SIZEOF_DCTELEM)] + + movq mm0,mm2 + movq mm1,mm3 + + psraw mm2,(WORD_BIT-1) ; -1 if value < 0, 0 otherwise + psraw mm3,(WORD_BIT-1) + + pxor mm0,mm2 ; val = -val + pxor mm1,mm3 + psubw mm0,mm2 + psubw mm1,mm3 + + ; + ; MMX is an annoyingly crappy instruction set. It has two + ; misfeatures that are causing problems here: + ; + ; - All multiplications are signed. + ; + ; - The second operand for the shifts is not treated as packed. + ; + ; + ; We work around the first problem by implementing this algorithm: + ; + ; unsigned long unsigned_multiply(unsigned short x, unsigned short y) + ; { + ; enum { SHORT_BIT = 16 }; + ; signed short sx = (signed short) x; + ; signed short sy = (signed short) y; + ; signed long sz; + ; + ; sz = (long) sx * (long) sy; /* signed multiply */ + ; + ; if (sx < 0) sz += (long) sy << SHORT_BIT; + ; if (sy < 0) sz += (long) sx << SHORT_BIT; + ; + ; return (unsigned long) sz; + ; } + ; + ; (note that a negative sx adds _sy_ and vice versa) + ; + ; For the second problem, we replace the shift by a multiplication. + ; Unfortunately that means we have to deal with the signed issue again. + ; + + paddw mm0, MMWORD [CORRECTION(0,0,edx)] ; correction + roundfactor + paddw mm1, MMWORD [CORRECTION(0,1,edx)] + + movq mm4,mm0 ; store current value for later + movq mm5,mm1 + pmulhw mm0, MMWORD [RECIPROCAL(0,0,edx)] ; reciprocal + pmulhw mm1, MMWORD [RECIPROCAL(0,1,edx)] + paddw mm0,mm4 ; reciprocal is always negative (MSB=1), + paddw mm1,mm5 ; so we always need to add the initial value + ; (input value is never negative as we + ; inverted it at the start of this routine) + + ; here it gets a bit tricky as both scale + ; and mm0/mm1 can be negative + movq mm6, MMWORD [SCALE(0,0,edx)] ; scale + movq mm7, MMWORD [SCALE(0,1,edx)] + movq mm4,mm0 + movq mm5,mm1 + pmulhw mm0,mm6 + pmulhw mm1,mm7 + + psraw mm6,(WORD_BIT-1) ; determine if scale is negative + psraw mm7,(WORD_BIT-1) + + pand mm6,mm4 ; and add input if it is + pand mm7,mm5 + paddw mm0,mm6 + paddw mm1,mm7 + + psraw mm4,(WORD_BIT-1) ; then check if negative input + psraw mm5,(WORD_BIT-1) + + pand mm4, MMWORD [SCALE(0,0,edx)] ; and add scale if it is + pand mm5, MMWORD [SCALE(0,1,edx)] + paddw mm0,mm4 + paddw mm1,mm5 + + pxor mm0,mm2 ; val = -val + pxor mm1,mm3 + psubw mm0,mm2 + psubw mm1,mm3 + + movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_DCTELEM)], mm0 + movq MMWORD [MMBLOCK(0,1,edi,SIZEOF_DCTELEM)], mm1 + + add esi, byte 8*SIZEOF_DCTELEM + add edx, byte 8*SIZEOF_DCTELEM + add edi, byte 8*SIZEOF_JCOEF + dec al + jnz near .quantloop2 + dec ah + jnz near .quantloop1 ; to avoid branch misprediction + + emms ; empty MMX state + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; unused +; pop ebx ; unused + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jquant-sse.asm b/app/src/main/jni/libjpeg-turbo/simd/jquant-sse.asm new file mode 100644 index 000000000..3f7fa5d31 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jquant-sse.asm @@ -0,0 +1,211 @@ +; +; jquant.asm - sample data conversion and quantization (SSE & MMX) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Load data into workspace, applying unsigned->signed conversion +; +; GLOBAL(void) +; jsimd_convsamp_float_sse (JSAMPARRAY sample_data, JDIMENSION start_col, +; FAST_FLOAT * workspace); +; + +%define sample_data ebp+8 ; JSAMPARRAY sample_data +%define start_col ebp+12 ; JDIMENSION start_col +%define workspace ebp+16 ; FAST_FLOAT * workspace + + align 16 + global EXTN(jsimd_convsamp_float_sse) + +EXTN(jsimd_convsamp_float_sse): + push ebp + mov ebp,esp + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + pcmpeqw mm7,mm7 + psllw mm7,7 + packsswb mm7,mm7 ; mm7 = PB_CENTERJSAMPLE (0x808080..) + + mov esi, JSAMPARRAY [sample_data] ; (JSAMPROW *) + mov eax, JDIMENSION [start_col] + mov edi, POINTER [workspace] ; (DCTELEM *) + mov ecx, DCTSIZE/2 + alignx 16,7 +.convloop: + mov ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW] ; (JSAMPLE *) + mov edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW] ; (JSAMPLE *) + + movq mm0, MMWORD [ebx+eax*SIZEOF_JSAMPLE] + movq mm1, MMWORD [edx+eax*SIZEOF_JSAMPLE] + + psubb mm0,mm7 ; mm0=(01234567) + psubb mm1,mm7 ; mm1=(89ABCDEF) + + punpcklbw mm2,mm0 ; mm2=(*0*1*2*3) + punpckhbw mm0,mm0 ; mm0=(*4*5*6*7) + punpcklbw mm3,mm1 ; mm3=(*8*9*A*B) + punpckhbw mm1,mm1 ; mm1=(*C*D*E*F) + + punpcklwd mm4,mm2 ; mm4=(***0***1) + punpckhwd mm2,mm2 ; mm2=(***2***3) + punpcklwd mm5,mm0 ; mm5=(***4***5) + punpckhwd mm0,mm0 ; mm0=(***6***7) + + psrad mm4,(DWORD_BIT-BYTE_BIT) ; mm4=(01) + psrad mm2,(DWORD_BIT-BYTE_BIT) ; mm2=(23) + cvtpi2ps xmm0,mm4 ; xmm0=(01**) + cvtpi2ps xmm1,mm2 ; xmm1=(23**) + psrad mm5,(DWORD_BIT-BYTE_BIT) ; mm5=(45) + psrad mm0,(DWORD_BIT-BYTE_BIT) ; mm0=(67) + cvtpi2ps xmm2,mm5 ; xmm2=(45**) + cvtpi2ps xmm3,mm0 ; xmm3=(67**) + + punpcklwd mm6,mm3 ; mm6=(***8***9) + punpckhwd mm3,mm3 ; mm3=(***A***B) + punpcklwd mm4,mm1 ; mm4=(***C***D) + punpckhwd mm1,mm1 ; mm1=(***E***F) + + psrad mm6,(DWORD_BIT-BYTE_BIT) ; mm6=(89) + psrad mm3,(DWORD_BIT-BYTE_BIT) ; mm3=(AB) + cvtpi2ps xmm4,mm6 ; xmm4=(89**) + cvtpi2ps xmm5,mm3 ; xmm5=(AB**) + psrad mm4,(DWORD_BIT-BYTE_BIT) ; mm4=(CD) + psrad mm1,(DWORD_BIT-BYTE_BIT) ; mm1=(EF) + cvtpi2ps xmm6,mm4 ; xmm6=(CD**) + cvtpi2ps xmm7,mm1 ; xmm7=(EF**) + + movlhps xmm0,xmm1 ; xmm0=(0123) + movlhps xmm2,xmm3 ; xmm2=(4567) + movlhps xmm4,xmm5 ; xmm4=(89AB) + movlhps xmm6,xmm7 ; xmm6=(CDEF) + + movaps XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm0 + movaps XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm2 + movaps XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm4 + movaps XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm6 + + add esi, byte 2*SIZEOF_JSAMPROW + add edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT + dec ecx + jnz near .convloop + + emms ; empty MMX state + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + pop ebp + ret + + +; -------------------------------------------------------------------------- +; +; Quantize/descale the coefficients, and store into coef_block +; +; GLOBAL(void) +; jsimd_quantize_float_sse (JCOEFPTR coef_block, FAST_FLOAT * divisors, +; FAST_FLOAT * workspace); +; + +%define coef_block ebp+8 ; JCOEFPTR coef_block +%define divisors ebp+12 ; FAST_FLOAT * divisors +%define workspace ebp+16 ; FAST_FLOAT * workspace + + align 16 + global EXTN(jsimd_quantize_float_sse) + +EXTN(jsimd_quantize_float_sse): + push ebp + mov ebp,esp +; push ebx ; unused +; push ecx ; unused +; push edx ; need not be preserved + push esi + push edi + + mov esi, POINTER [workspace] + mov edx, POINTER [divisors] + mov edi, JCOEFPTR [coef_block] + mov eax, DCTSIZE2/16 + alignx 16,7 +.quantloop: + movaps xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(0,1,esi,SIZEOF_FAST_FLOAT)] + mulps xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)] + mulps xmm1, XMMWORD [XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)] + movaps xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(1,1,esi,SIZEOF_FAST_FLOAT)] + mulps xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)] + mulps xmm3, XMMWORD [XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)] + + movhlps xmm4,xmm0 + movhlps xmm5,xmm1 + + cvtps2pi mm0,xmm0 + cvtps2pi mm1,xmm1 + cvtps2pi mm4,xmm4 + cvtps2pi mm5,xmm5 + + movhlps xmm6,xmm2 + movhlps xmm7,xmm3 + + cvtps2pi mm2,xmm2 + cvtps2pi mm3,xmm3 + cvtps2pi mm6,xmm6 + cvtps2pi mm7,xmm7 + + packssdw mm0,mm4 + packssdw mm1,mm5 + packssdw mm2,mm6 + packssdw mm3,mm7 + + movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0 + movq MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm1 + movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm2 + movq MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm3 + + add esi, byte 16*SIZEOF_FAST_FLOAT + add edx, byte 16*SIZEOF_FAST_FLOAT + add edi, byte 16*SIZEOF_JCOEF + dec eax + jnz short .quantloop + + emms ; empty MMX state + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; unused +; pop ebx ; unused + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jquantf-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jquantf-sse2-64.asm new file mode 100644 index 000000000..20e815f7d --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jquantf-sse2-64.asm @@ -0,0 +1,158 @@ +; +; jquantf.asm - sample data conversion and quantization (64-bit SSE & SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 +; +; Load data into workspace, applying unsigned->signed conversion +; +; GLOBAL(void) +; jsimd_convsamp_float_sse2 (JSAMPARRAY sample_data, JDIMENSION start_col, +; FAST_FLOAT * workspace); +; + +; r10 = JSAMPARRAY sample_data +; r11 = JDIMENSION start_col +; r12 = FAST_FLOAT * workspace + + align 16 + global EXTN(jsimd_convsamp_float_sse2) + +EXTN(jsimd_convsamp_float_sse2): + push rbp + mov rax,rsp + mov rbp,rsp + collect_args + push rbx + + pcmpeqw xmm7,xmm7 + psllw xmm7,7 + packsswb xmm7,xmm7 ; xmm7 = PB_CENTERJSAMPLE (0x808080..) + + mov rsi, r10 + mov rax, r11 + mov rdi, r12 + mov rcx, DCTSIZE/2 +.convloop: + mov rbx, JSAMPROW [rsi+0*SIZEOF_JSAMPROW] ; (JSAMPLE *) + mov rdx, JSAMPROW [rsi+1*SIZEOF_JSAMPROW] ; (JSAMPLE *) + + movq xmm0, XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE] + movq xmm1, XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE] + + psubb xmm0,xmm7 ; xmm0=(01234567) + psubb xmm1,xmm7 ; xmm1=(89ABCDEF) + + punpcklbw xmm0,xmm0 ; xmm0=(*0*1*2*3*4*5*6*7) + punpcklbw xmm1,xmm1 ; xmm1=(*8*9*A*B*C*D*E*F) + + punpcklwd xmm2,xmm0 ; xmm2=(***0***1***2***3) + punpckhwd xmm0,xmm0 ; xmm0=(***4***5***6***7) + punpcklwd xmm3,xmm1 ; xmm3=(***8***9***A***B) + punpckhwd xmm1,xmm1 ; xmm1=(***C***D***E***F) + + psrad xmm2,(DWORD_BIT-BYTE_BIT) ; xmm2=(0123) + psrad xmm0,(DWORD_BIT-BYTE_BIT) ; xmm0=(4567) + cvtdq2ps xmm2,xmm2 ; xmm2=(0123) + cvtdq2ps xmm0,xmm0 ; xmm0=(4567) + psrad xmm3,(DWORD_BIT-BYTE_BIT) ; xmm3=(89AB) + psrad xmm1,(DWORD_BIT-BYTE_BIT) ; xmm1=(CDEF) + cvtdq2ps xmm3,xmm3 ; xmm3=(89AB) + cvtdq2ps xmm1,xmm1 ; xmm1=(CDEF) + + movaps XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_FAST_FLOAT)], xmm2 + movaps XMMWORD [XMMBLOCK(0,1,rdi,SIZEOF_FAST_FLOAT)], xmm0 + movaps XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_FAST_FLOAT)], xmm3 + movaps XMMWORD [XMMBLOCK(1,1,rdi,SIZEOF_FAST_FLOAT)], xmm1 + + add rsi, byte 2*SIZEOF_JSAMPROW + add rdi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT + dec rcx + jnz short .convloop + + pop rbx + uncollect_args + pop rbp + ret + + +; -------------------------------------------------------------------------- +; +; Quantize/descale the coefficients, and store into coef_block +; +; GLOBAL(void) +; jsimd_quantize_float_sse2 (JCOEFPTR coef_block, FAST_FLOAT * divisors, +; FAST_FLOAT * workspace); +; + +; r10 = JCOEFPTR coef_block +; r11 = FAST_FLOAT * divisors +; r12 = FAST_FLOAT * workspace + + align 16 + global EXTN(jsimd_quantize_float_sse2) + +EXTN(jsimd_quantize_float_sse2): + push rbp + mov rax,rsp + mov rbp,rsp + collect_args + + mov rsi, r12 + mov rdx, r11 + mov rdi, r10 + mov rax, DCTSIZE2/16 +.quantloop: + movaps xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(0,1,rsi,SIZEOF_FAST_FLOAT)] + mulps xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)] + mulps xmm1, XMMWORD [XMMBLOCK(0,1,rdx,SIZEOF_FAST_FLOAT)] + movaps xmm2, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(1,1,rsi,SIZEOF_FAST_FLOAT)] + mulps xmm2, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)] + mulps xmm3, XMMWORD [XMMBLOCK(1,1,rdx,SIZEOF_FAST_FLOAT)] + + cvtps2dq xmm0,xmm0 + cvtps2dq xmm1,xmm1 + cvtps2dq xmm2,xmm2 + cvtps2dq xmm3,xmm3 + + packssdw xmm0,xmm1 + packssdw xmm2,xmm3 + + movdqa XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_JCOEF)], xmm0 + movdqa XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_JCOEF)], xmm2 + + add rsi, byte 16*SIZEOF_FAST_FLOAT + add rdx, byte 16*SIZEOF_FAST_FLOAT + add rdi, byte 16*SIZEOF_JCOEF + dec rax + jnz short .quantloop + + uncollect_args + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jquantf-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jquantf-sse2.asm new file mode 100644 index 000000000..b1d3efcca --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jquantf-sse2.asm @@ -0,0 +1,171 @@ +; +; jquantf.asm - sample data conversion and quantization (SSE & SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Load data into workspace, applying unsigned->signed conversion +; +; GLOBAL(void) +; jsimd_convsamp_float_sse2 (JSAMPARRAY sample_data, JDIMENSION start_col, +; FAST_FLOAT * workspace); +; + +%define sample_data ebp+8 ; JSAMPARRAY sample_data +%define start_col ebp+12 ; JDIMENSION start_col +%define workspace ebp+16 ; FAST_FLOAT * workspace + + align 16 + global EXTN(jsimd_convsamp_float_sse2) + +EXTN(jsimd_convsamp_float_sse2): + push ebp + mov ebp,esp + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + pcmpeqw xmm7,xmm7 + psllw xmm7,7 + packsswb xmm7,xmm7 ; xmm7 = PB_CENTERJSAMPLE (0x808080..) + + mov esi, JSAMPARRAY [sample_data] ; (JSAMPROW *) + mov eax, JDIMENSION [start_col] + mov edi, POINTER [workspace] ; (DCTELEM *) + mov ecx, DCTSIZE/2 + alignx 16,7 +.convloop: + mov ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW] ; (JSAMPLE *) + mov edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW] ; (JSAMPLE *) + + movq xmm0, XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE] + movq xmm1, XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE] + + psubb xmm0,xmm7 ; xmm0=(01234567) + psubb xmm1,xmm7 ; xmm1=(89ABCDEF) + + punpcklbw xmm0,xmm0 ; xmm0=(*0*1*2*3*4*5*6*7) + punpcklbw xmm1,xmm1 ; xmm1=(*8*9*A*B*C*D*E*F) + + punpcklwd xmm2,xmm0 ; xmm2=(***0***1***2***3) + punpckhwd xmm0,xmm0 ; xmm0=(***4***5***6***7) + punpcklwd xmm3,xmm1 ; xmm3=(***8***9***A***B) + punpckhwd xmm1,xmm1 ; xmm1=(***C***D***E***F) + + psrad xmm2,(DWORD_BIT-BYTE_BIT) ; xmm2=(0123) + psrad xmm0,(DWORD_BIT-BYTE_BIT) ; xmm0=(4567) + cvtdq2ps xmm2,xmm2 ; xmm2=(0123) + cvtdq2ps xmm0,xmm0 ; xmm0=(4567) + psrad xmm3,(DWORD_BIT-BYTE_BIT) ; xmm3=(89AB) + psrad xmm1,(DWORD_BIT-BYTE_BIT) ; xmm1=(CDEF) + cvtdq2ps xmm3,xmm3 ; xmm3=(89AB) + cvtdq2ps xmm1,xmm1 ; xmm1=(CDEF) + + movaps XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm2 + movaps XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm0 + movaps XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm3 + movaps XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm1 + + add esi, byte 2*SIZEOF_JSAMPROW + add edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT + dec ecx + jnz short .convloop + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + pop ebp + ret + + +; -------------------------------------------------------------------------- +; +; Quantize/descale the coefficients, and store into coef_block +; +; GLOBAL(void) +; jsimd_quantize_float_sse2 (JCOEFPTR coef_block, FAST_FLOAT * divisors, +; FAST_FLOAT * workspace); +; + +%define coef_block ebp+8 ; JCOEFPTR coef_block +%define divisors ebp+12 ; FAST_FLOAT * divisors +%define workspace ebp+16 ; FAST_FLOAT * workspace + + align 16 + global EXTN(jsimd_quantize_float_sse2) + +EXTN(jsimd_quantize_float_sse2): + push ebp + mov ebp,esp +; push ebx ; unused +; push ecx ; unused +; push edx ; need not be preserved + push esi + push edi + + mov esi, POINTER [workspace] + mov edx, POINTER [divisors] + mov edi, JCOEFPTR [coef_block] + mov eax, DCTSIZE2/16 + alignx 16,7 +.quantloop: + movaps xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm1, XMMWORD [XMMBLOCK(0,1,esi,SIZEOF_FAST_FLOAT)] + mulps xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)] + mulps xmm1, XMMWORD [XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)] + movaps xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)] + movaps xmm3, XMMWORD [XMMBLOCK(1,1,esi,SIZEOF_FAST_FLOAT)] + mulps xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)] + mulps xmm3, XMMWORD [XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)] + + cvtps2dq xmm0,xmm0 + cvtps2dq xmm1,xmm1 + cvtps2dq xmm2,xmm2 + cvtps2dq xmm3,xmm3 + + packssdw xmm0,xmm1 + packssdw xmm2,xmm3 + + movdqa XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_JCOEF)], xmm0 + movdqa XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_JCOEF)], xmm2 + + add esi, byte 16*SIZEOF_FAST_FLOAT + add edx, byte 16*SIZEOF_FAST_FLOAT + add edi, byte 16*SIZEOF_JCOEF + dec eax + jnz short .quantloop + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; unused +; pop ebx ; unused + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jquanti-sse2-64.asm b/app/src/main/jni/libjpeg-turbo/simd/jquanti-sse2-64.asm new file mode 100644 index 000000000..50b8dce7a --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jquanti-sse2-64.asm @@ -0,0 +1,187 @@ +; +; jquanti.asm - sample data conversion and quantization (64-bit SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2009 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 64 +; +; Load data into workspace, applying unsigned->signed conversion +; +; GLOBAL(void) +; jsimd_convsamp_sse2 (JSAMPARRAY sample_data, JDIMENSION start_col, +; DCTELEM * workspace); +; + +; r10 = JSAMPARRAY sample_data +; r11 = JDIMENSION start_col +; r12 = DCTELEM * workspace + + align 16 + global EXTN(jsimd_convsamp_sse2) + +EXTN(jsimd_convsamp_sse2): + push rbp + mov rax,rsp + mov rbp,rsp + collect_args + push rbx + + pxor xmm6,xmm6 ; xmm6=(all 0's) + pcmpeqw xmm7,xmm7 + psllw xmm7,7 ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..} + + mov rsi, r10 + mov rax, r11 + mov rdi, r12 + mov rcx, DCTSIZE/4 +.convloop: + mov rbx, JSAMPROW [rsi+0*SIZEOF_JSAMPROW] ; (JSAMPLE *) + mov rdx, JSAMPROW [rsi+1*SIZEOF_JSAMPROW] ; (JSAMPLE *) + + movq xmm0, XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE] ; xmm0=(01234567) + movq xmm1, XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE] ; xmm1=(89ABCDEF) + + mov rbx, JSAMPROW [rsi+2*SIZEOF_JSAMPROW] ; (JSAMPLE *) + mov rdx, JSAMPROW [rsi+3*SIZEOF_JSAMPROW] ; (JSAMPLE *) + + movq xmm2, XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE] ; xmm2=(GHIJKLMN) + movq xmm3, XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE] ; xmm3=(OPQRSTUV) + + punpcklbw xmm0,xmm6 ; xmm0=(01234567) + punpcklbw xmm1,xmm6 ; xmm1=(89ABCDEF) + paddw xmm0,xmm7 + paddw xmm1,xmm7 + punpcklbw xmm2,xmm6 ; xmm2=(GHIJKLMN) + punpcklbw xmm3,xmm6 ; xmm3=(OPQRSTUV) + paddw xmm2,xmm7 + paddw xmm3,xmm7 + + movdqa XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_DCTELEM)], xmm0 + movdqa XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_DCTELEM)], xmm1 + movdqa XMMWORD [XMMBLOCK(2,0,rdi,SIZEOF_DCTELEM)], xmm2 + movdqa XMMWORD [XMMBLOCK(3,0,rdi,SIZEOF_DCTELEM)], xmm3 + + add rsi, byte 4*SIZEOF_JSAMPROW + add rdi, byte 4*DCTSIZE*SIZEOF_DCTELEM + dec rcx + jnz short .convloop + + pop rbx + uncollect_args + pop rbp + ret + +; -------------------------------------------------------------------------- +; +; Quantize/descale the coefficients, and store into coef_block +; +; This implementation is based on an algorithm described in +; "How to optimize for the Pentium family of microprocessors" +; (http://www.agner.org/assem/). +; +; GLOBAL(void) +; jsimd_quantize_sse2 (JCOEFPTR coef_block, DCTELEM * divisors, +; DCTELEM * workspace); +; + +%define RECIPROCAL(m,n,b) XMMBLOCK(DCTSIZE*0+(m),(n),(b),SIZEOF_DCTELEM) +%define CORRECTION(m,n,b) XMMBLOCK(DCTSIZE*1+(m),(n),(b),SIZEOF_DCTELEM) +%define SCALE(m,n,b) XMMBLOCK(DCTSIZE*2+(m),(n),(b),SIZEOF_DCTELEM) + +; r10 = JCOEFPTR coef_block +; r11 = DCTELEM * divisors +; r12 = DCTELEM * workspace + + align 16 + global EXTN(jsimd_quantize_sse2) + +EXTN(jsimd_quantize_sse2): + push rbp + mov rax,rsp + mov rbp,rsp + collect_args + + mov rsi, r12 + mov rdx, r11 + mov rdi, r10 + mov rax, DCTSIZE2/32 +.quantloop: + movdqa xmm4, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_DCTELEM)] + movdqa xmm5, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_DCTELEM)] + movdqa xmm6, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_DCTELEM)] + movdqa xmm7, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_DCTELEM)] + movdqa xmm0,xmm4 + movdqa xmm1,xmm5 + movdqa xmm2,xmm6 + movdqa xmm3,xmm7 + psraw xmm4,(WORD_BIT-1) + psraw xmm5,(WORD_BIT-1) + psraw xmm6,(WORD_BIT-1) + psraw xmm7,(WORD_BIT-1) + pxor xmm0,xmm4 + pxor xmm1,xmm5 + pxor xmm2,xmm6 + pxor xmm3,xmm7 + psubw xmm0,xmm4 ; if (xmm0 < 0) xmm0 = -xmm0; + psubw xmm1,xmm5 ; if (xmm1 < 0) xmm1 = -xmm1; + psubw xmm2,xmm6 ; if (xmm2 < 0) xmm2 = -xmm2; + psubw xmm3,xmm7 ; if (xmm3 < 0) xmm3 = -xmm3; + + paddw xmm0, XMMWORD [CORRECTION(0,0,rdx)] ; correction + roundfactor + paddw xmm1, XMMWORD [CORRECTION(1,0,rdx)] + paddw xmm2, XMMWORD [CORRECTION(2,0,rdx)] + paddw xmm3, XMMWORD [CORRECTION(3,0,rdx)] + pmulhuw xmm0, XMMWORD [RECIPROCAL(0,0,rdx)] ; reciprocal + pmulhuw xmm1, XMMWORD [RECIPROCAL(1,0,rdx)] + pmulhuw xmm2, XMMWORD [RECIPROCAL(2,0,rdx)] + pmulhuw xmm3, XMMWORD [RECIPROCAL(3,0,rdx)] + pmulhuw xmm0, XMMWORD [SCALE(0,0,rdx)] ; scale + pmulhuw xmm1, XMMWORD [SCALE(1,0,rdx)] + pmulhuw xmm2, XMMWORD [SCALE(2,0,rdx)] + pmulhuw xmm3, XMMWORD [SCALE(3,0,rdx)] + + pxor xmm0,xmm4 + pxor xmm1,xmm5 + pxor xmm2,xmm6 + pxor xmm3,xmm7 + psubw xmm0,xmm4 + psubw xmm1,xmm5 + psubw xmm2,xmm6 + psubw xmm3,xmm7 + movdqa XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_DCTELEM)], xmm0 + movdqa XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_DCTELEM)], xmm1 + movdqa XMMWORD [XMMBLOCK(2,0,rdi,SIZEOF_DCTELEM)], xmm2 + movdqa XMMWORD [XMMBLOCK(3,0,rdi,SIZEOF_DCTELEM)], xmm3 + + add rsi, byte 32*SIZEOF_DCTELEM + add rdx, byte 32*SIZEOF_DCTELEM + add rdi, byte 32*SIZEOF_JCOEF + dec rax + jnz near .quantloop + + uncollect_args + pop rbp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jquanti-sse2.asm b/app/src/main/jni/libjpeg-turbo/simd/jquanti-sse2.asm new file mode 100644 index 000000000..79ca3db42 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jquanti-sse2.asm @@ -0,0 +1,200 @@ +; +; jquanti.asm - sample data conversion and quantization (SSE2) +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" +%include "jdct.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Load data into workspace, applying unsigned->signed conversion +; +; GLOBAL(void) +; jsimd_convsamp_sse2 (JSAMPARRAY sample_data, JDIMENSION start_col, +; DCTELEM * workspace); +; + +%define sample_data ebp+8 ; JSAMPARRAY sample_data +%define start_col ebp+12 ; JDIMENSION start_col +%define workspace ebp+16 ; DCTELEM * workspace + + align 16 + global EXTN(jsimd_convsamp_sse2) + +EXTN(jsimd_convsamp_sse2): + push ebp + mov ebp,esp + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved + push esi + push edi + + pxor xmm6,xmm6 ; xmm6=(all 0's) + pcmpeqw xmm7,xmm7 + psllw xmm7,7 ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..} + + mov esi, JSAMPARRAY [sample_data] ; (JSAMPROW *) + mov eax, JDIMENSION [start_col] + mov edi, POINTER [workspace] ; (DCTELEM *) + mov ecx, DCTSIZE/4 + alignx 16,7 +.convloop: + mov ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW] ; (JSAMPLE *) + mov edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW] ; (JSAMPLE *) + + movq xmm0, XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE] ; xmm0=(01234567) + movq xmm1, XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE] ; xmm1=(89ABCDEF) + + mov ebx, JSAMPROW [esi+2*SIZEOF_JSAMPROW] ; (JSAMPLE *) + mov edx, JSAMPROW [esi+3*SIZEOF_JSAMPROW] ; (JSAMPLE *) + + movq xmm2, XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE] ; xmm2=(GHIJKLMN) + movq xmm3, XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE] ; xmm3=(OPQRSTUV) + + punpcklbw xmm0,xmm6 ; xmm0=(01234567) + punpcklbw xmm1,xmm6 ; xmm1=(89ABCDEF) + paddw xmm0,xmm7 + paddw xmm1,xmm7 + punpcklbw xmm2,xmm6 ; xmm2=(GHIJKLMN) + punpcklbw xmm3,xmm6 ; xmm3=(OPQRSTUV) + paddw xmm2,xmm7 + paddw xmm3,xmm7 + + movdqa XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_DCTELEM)], xmm0 + movdqa XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_DCTELEM)], xmm1 + movdqa XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_DCTELEM)], xmm2 + movdqa XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_DCTELEM)], xmm3 + + add esi, byte 4*SIZEOF_JSAMPROW + add edi, byte 4*DCTSIZE*SIZEOF_DCTELEM + dec ecx + jnz short .convloop + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + pop ebp + ret + +; -------------------------------------------------------------------------- +; +; Quantize/descale the coefficients, and store into coef_block +; +; This implementation is based on an algorithm described in +; "How to optimize for the Pentium family of microprocessors" +; (http://www.agner.org/assem/). +; +; GLOBAL(void) +; jsimd_quantize_sse2 (JCOEFPTR coef_block, DCTELEM * divisors, +; DCTELEM * workspace); +; + +%define RECIPROCAL(m,n,b) XMMBLOCK(DCTSIZE*0+(m),(n),(b),SIZEOF_DCTELEM) +%define CORRECTION(m,n,b) XMMBLOCK(DCTSIZE*1+(m),(n),(b),SIZEOF_DCTELEM) +%define SCALE(m,n,b) XMMBLOCK(DCTSIZE*2+(m),(n),(b),SIZEOF_DCTELEM) + +%define coef_block ebp+8 ; JCOEFPTR coef_block +%define divisors ebp+12 ; DCTELEM * divisors +%define workspace ebp+16 ; DCTELEM * workspace + + align 16 + global EXTN(jsimd_quantize_sse2) + +EXTN(jsimd_quantize_sse2): + push ebp + mov ebp,esp +; push ebx ; unused +; push ecx ; unused +; push edx ; need not be preserved + push esi + push edi + + mov esi, POINTER [workspace] + mov edx, POINTER [divisors] + mov edi, JCOEFPTR [coef_block] + mov eax, DCTSIZE2/32 + alignx 16,7 +.quantloop: + movdqa xmm4, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_DCTELEM)] + movdqa xmm5, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_DCTELEM)] + movdqa xmm6, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_DCTELEM)] + movdqa xmm7, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_DCTELEM)] + movdqa xmm0,xmm4 + movdqa xmm1,xmm5 + movdqa xmm2,xmm6 + movdqa xmm3,xmm7 + psraw xmm4,(WORD_BIT-1) + psraw xmm5,(WORD_BIT-1) + psraw xmm6,(WORD_BIT-1) + psraw xmm7,(WORD_BIT-1) + pxor xmm0,xmm4 + pxor xmm1,xmm5 + pxor xmm2,xmm6 + pxor xmm3,xmm7 + psubw xmm0,xmm4 ; if (xmm0 < 0) xmm0 = -xmm0; + psubw xmm1,xmm5 ; if (xmm1 < 0) xmm1 = -xmm1; + psubw xmm2,xmm6 ; if (xmm2 < 0) xmm2 = -xmm2; + psubw xmm3,xmm7 ; if (xmm3 < 0) xmm3 = -xmm3; + + paddw xmm0, XMMWORD [CORRECTION(0,0,edx)] ; correction + roundfactor + paddw xmm1, XMMWORD [CORRECTION(1,0,edx)] + paddw xmm2, XMMWORD [CORRECTION(2,0,edx)] + paddw xmm3, XMMWORD [CORRECTION(3,0,edx)] + pmulhuw xmm0, XMMWORD [RECIPROCAL(0,0,edx)] ; reciprocal + pmulhuw xmm1, XMMWORD [RECIPROCAL(1,0,edx)] + pmulhuw xmm2, XMMWORD [RECIPROCAL(2,0,edx)] + pmulhuw xmm3, XMMWORD [RECIPROCAL(3,0,edx)] + pmulhuw xmm0, XMMWORD [SCALE(0,0,edx)] ; scale + pmulhuw xmm1, XMMWORD [SCALE(1,0,edx)] + pmulhuw xmm2, XMMWORD [SCALE(2,0,edx)] + pmulhuw xmm3, XMMWORD [SCALE(3,0,edx)] + + pxor xmm0,xmm4 + pxor xmm1,xmm5 + pxor xmm2,xmm6 + pxor xmm3,xmm7 + psubw xmm0,xmm4 + psubw xmm1,xmm5 + psubw xmm2,xmm6 + psubw xmm3,xmm7 + movdqa XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_DCTELEM)], xmm0 + movdqa XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_DCTELEM)], xmm1 + movdqa XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_DCTELEM)], xmm2 + movdqa XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_DCTELEM)], xmm3 + + add esi, byte 32*SIZEOF_DCTELEM + add edx, byte 32*SIZEOF_DCTELEM + add edi, byte 32*SIZEOF_JCOEF + dec eax + jnz near .quantloop + + pop edi + pop esi +; pop edx ; need not be preserved +; pop ecx ; unused +; pop ebx ; unused + pop ebp + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jsimd.h b/app/src/main/jni/libjpeg-turbo/simd/jsimd.h new file mode 100644 index 000000000..c5abd4584 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jsimd.h @@ -0,0 +1,678 @@ +/* + * simd/jsimd.h + * + * Copyright 2009 Pierre Ossman for Cendio AB + * Copyright 2011 D. R. Commander + * Copyright (C) 2013-2014, MIPS Technologies, Inc., California + * Copyright (C) 2014 Linaro Limited + * + * Based on the x86 SIMD extension for IJG JPEG library, + * Copyright (C) 1999-2006, MIYASAKA Masaru. + * For conditions of distribution and use, see copyright notice in jsimdext.inc + * + */ + +/* Bitmask for supported acceleration methods */ + +#define JSIMD_NONE 0x00 +#define JSIMD_MMX 0x01 +#define JSIMD_3DNOW 0x02 +#define JSIMD_SSE 0x04 +#define JSIMD_SSE2 0x08 +#define JSIMD_ARM_NEON 0x10 +#define JSIMD_MIPS_DSPR2 0x20 + +/* SIMD Ext: retrieve SIMD/CPU information */ +EXTERN(unsigned int) jpeg_simd_cpu_support (void); + +/* RGB & extended RGB --> YCC Colorspace Conversion */ +EXTERN(void) jsimd_rgb_ycc_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgb_ycc_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgbx_ycc_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgr_ycc_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgrx_ycc_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxbgr_ycc_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxrgb_ycc_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); + +extern const int jconst_rgb_ycc_convert_sse2[]; +EXTERN(void) jsimd_rgb_ycc_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgb_ycc_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgbx_ycc_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgr_ycc_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgrx_ycc_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxbgr_ycc_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxrgb_ycc_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); + +EXTERN(void) jsimd_rgb_ycc_convert_neon + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgb_ycc_convert_neon + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgbx_ycc_convert_neon + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgr_ycc_convert_neon + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgrx_ycc_convert_neon + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxbgr_ycc_convert_neon + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxrgb_ycc_convert_neon + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); + +EXTERN(void) jsimd_rgb_ycc_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgb_ycc_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgbx_ycc_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgr_ycc_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgrx_ycc_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxbgr_ycc_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxrgb_ycc_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); + +/* RGB & extended RGB --> Grayscale Colorspace Conversion */ +EXTERN(void) jsimd_rgb_gray_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgb_gray_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgbx_gray_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgr_gray_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgrx_gray_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxbgr_gray_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxrgb_gray_convert_mmx + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); + +extern const int jconst_rgb_gray_convert_sse2[]; +EXTERN(void) jsimd_rgb_gray_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgb_gray_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgbx_gray_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgr_gray_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgrx_gray_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxbgr_gray_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxrgb_gray_convert_sse2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); + +EXTERN(void) jsimd_rgb_gray_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgb_gray_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extrgbx_gray_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgr_gray_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extbgrx_gray_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxbgr_gray_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); +EXTERN(void) jsimd_extxrgb_gray_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows); + +/* YCC --> RGB & extended RGB Colorspace Conversion */ +EXTERN(void) jsimd_ycc_rgb_convert_mmx + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extrgb_convert_mmx + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extrgbx_convert_mmx + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extbgr_convert_mmx + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extbgrx_convert_mmx + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extxbgr_convert_mmx + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extxrgb_convert_mmx + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); + +extern const int jconst_ycc_rgb_convert_sse2[]; +EXTERN(void) jsimd_ycc_rgb_convert_sse2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extrgb_convert_sse2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extrgbx_convert_sse2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extbgr_convert_sse2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extbgrx_convert_sse2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extxbgr_convert_sse2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extxrgb_convert_sse2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); + +EXTERN(void) jsimd_ycc_rgb_convert_neon + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extrgb_convert_neon + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extrgbx_convert_neon + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extbgr_convert_neon + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extbgrx_convert_neon + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extxbgr_convert_neon + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extxrgb_convert_neon + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_rgb565_convert_neon + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); + +EXTERN(void) jsimd_ycc_rgb_convert_mips_dspr2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extrgb_convert_mips_dspr2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extrgbx_convert_mips_dspr2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extbgr_convert_mips_dspr2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extbgrx_convert_mips_dspr2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extxbgr_convert_mips_dspr2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); +EXTERN(void) jsimd_ycc_extxrgb_convert_mips_dspr2 + (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows); + +/* NULL Colorspace Conversion */ +EXTERN(void) jsimd_c_null_convert_mips_dspr2 + (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows, int num_components); + +/* h2v1 Downsampling */ +EXTERN(void) jsimd_h2v1_downsample_mmx + (JDIMENSION image_width, int max_v_samp_factor, + JDIMENSION v_samp_factor, JDIMENSION width_blocks, + JSAMPARRAY input_data, JSAMPARRAY output_data); + +EXTERN(void) jsimd_h2v1_downsample_sse2 + (JDIMENSION image_width, int max_v_samp_factor, + JDIMENSION v_samp_factor, JDIMENSION width_blocks, + JSAMPARRAY input_data, JSAMPARRAY output_data); + +EXTERN(void) jsimd_h2v1_downsample_mips_dspr2 + (JDIMENSION image_width, int max_v_samp_factor, + JDIMENSION v_samp_factor, JDIMENSION width_blocks, + JSAMPARRAY input_data, JSAMPARRAY output_data); + +/* h2v2 Downsampling */ +EXTERN(void) jsimd_h2v2_downsample_mmx + (JDIMENSION image_width, int max_v_samp_factor, + JDIMENSION v_samp_factor, JDIMENSION width_blocks, + JSAMPARRAY input_data, JSAMPARRAY output_data); + +EXTERN(void) jsimd_h2v2_downsample_sse2 + (JDIMENSION image_width, int max_v_samp_factor, + JDIMENSION v_samp_factor, JDIMENSION width_blocks, + JSAMPARRAY input_data, JSAMPARRAY output_data); + +EXTERN(void) jsimd_h2v2_downsample_mips_dspr2 + (JDIMENSION image_width, int max_v_samp_factor, + JDIMENSION v_samp_factor, JDIMENSION width_blocks, + JSAMPARRAY input_data, JSAMPARRAY output_data); + +/* h2v2 Smooth Downsampling */ +EXTERN(void) jsimd_h2v2_smooth_downsample_mips_dspr2 + (JSAMPARRAY input_data, JSAMPARRAY output_data, + JDIMENSION v_samp_factor, int max_v_samp_factor, + int smoothing_factor, JDIMENSION width_blocks, + JDIMENSION image_width); + + +/* Upsampling */ +EXTERN(void) jsimd_h2v1_upsample_mmx + (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr); +EXTERN(void) jsimd_h2v2_upsample_mmx + (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr); + +EXTERN(void) jsimd_h2v1_upsample_sse2 + (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr); +EXTERN(void) jsimd_h2v2_upsample_sse2 + (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr); + +EXTERN(void) jsimd_h2v1_upsample_mips_dspr2 + (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr); +EXTERN(void) jsimd_h2v2_upsample_mips_dspr2 + (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr); + +EXTERN(void) jsimd_int_upsample_mips_dspr2 + (UINT8 h_expand, UINT8 v_expand, JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr, JDIMENSION output_width, + int max_v_samp_factor); + + +/* Fancy Upsampling */ +EXTERN(void) jsimd_h2v1_fancy_upsample_mmx + (int max_v_samp_factor, JDIMENSION downsampled_width, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr); +EXTERN(void) jsimd_h2v2_fancy_upsample_mmx + (int max_v_samp_factor, JDIMENSION downsampled_width, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr); + +extern const int jconst_fancy_upsample_sse2[]; +EXTERN(void) jsimd_h2v1_fancy_upsample_sse2 + (int max_v_samp_factor, JDIMENSION downsampled_width, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr); +EXTERN(void) jsimd_h2v2_fancy_upsample_sse2 + (int max_v_samp_factor, JDIMENSION downsampled_width, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr); + +EXTERN(void) jsimd_h2v1_fancy_upsample_neon + (int max_v_samp_factor, JDIMENSION downsampled_width, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr); + +EXTERN(void) jsimd_h2v1_fancy_upsample_mips_dspr2 + (int max_v_samp_factor, JDIMENSION downsampled_width, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr); +EXTERN(void) jsimd_h2v2_fancy_upsample_mips_dspr2 + (int max_v_samp_factor, JDIMENSION downsampled_width, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr); + +/* Merged Upsampling */ +EXTERN(void) jsimd_h2v1_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); + +EXTERN(void) jsimd_h2v2_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_mmx + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); + +extern const int jconst_merged_upsample_sse2[]; +EXTERN(void) jsimd_h2v1_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); + +EXTERN(void) jsimd_h2v2_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); +EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_sse2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf); + +EXTERN(void) jsimd_h2v1_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); +EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); +EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); +EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); +EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); +EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); +EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); + +EXTERN(void) jsimd_h2v2_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); +EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); +EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); +EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); +EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); +EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); +EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_mips_dspr2 + (JDIMENSION output_width, JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range); + +/* Sample Conversion */ +EXTERN(void) jsimd_convsamp_mmx + (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM * workspace); + +EXTERN(void) jsimd_convsamp_sse2 + (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM * workspace); + +EXTERN(void) jsimd_convsamp_neon + (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM * workspace); + +EXTERN(void) jsimd_convsamp_mips_dspr2 + (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM * workspace); + +/* Floating Point Sample Conversion */ +EXTERN(void) jsimd_convsamp_float_3dnow + (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT * workspace); + +EXTERN(void) jsimd_convsamp_float_sse + (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT * workspace); + +EXTERN(void) jsimd_convsamp_float_sse2 + (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT * workspace); + +EXTERN(void) jsimd_convsamp_float_mips_dspr2 + (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT * workspace); + +/* Slow Integer Forward DCT */ +EXTERN(void) jsimd_fdct_islow_mmx (DCTELEM * data); + +extern const int jconst_fdct_islow_sse2[]; +EXTERN(void) jsimd_fdct_islow_sse2 (DCTELEM * data); + +EXTERN(void) jsimd_fdct_islow_mips_dspr2 (DCTELEM * data); + +/* Fast Integer Forward DCT */ +EXTERN(void) jsimd_fdct_ifast_mmx (DCTELEM * data); + +extern const int jconst_fdct_ifast_sse2[]; +EXTERN(void) jsimd_fdct_ifast_sse2 (DCTELEM * data); + +EXTERN(void) jsimd_fdct_ifast_neon (DCTELEM * data); + +EXTERN(void) jsimd_fdct_ifast_mips_dspr2 (DCTELEM * data); + +/* Floating Point Forward DCT */ +EXTERN(void) jsimd_fdct_float_3dnow (FAST_FLOAT * data); + +extern const int jconst_fdct_float_sse[]; +EXTERN(void) jsimd_fdct_float_sse (FAST_FLOAT * data); + +/* Quantization */ +EXTERN(void) jsimd_quantize_mmx + (JCOEFPTR coef_block, DCTELEM * divisors, DCTELEM * workspace); + +EXTERN(void) jsimd_quantize_sse2 + (JCOEFPTR coef_block, DCTELEM * divisors, DCTELEM * workspace); + +EXTERN(void) jsimd_quantize_neon + (JCOEFPTR coef_block, DCTELEM * divisors, DCTELEM * workspace); + +EXTERN(void) jsimd_quantize_mips_dspr2 + (JCOEFPTR coef_block, DCTELEM * divisors, DCTELEM * workspace); + +/* Floating Point Quantization */ +EXTERN(void) jsimd_quantize_float_3dnow + (JCOEFPTR coef_block, FAST_FLOAT * divisors, FAST_FLOAT * workspace); + +EXTERN(void) jsimd_quantize_float_sse + (JCOEFPTR coef_block, FAST_FLOAT * divisors, FAST_FLOAT * workspace); + +EXTERN(void) jsimd_quantize_float_sse2 + (JCOEFPTR coef_block, FAST_FLOAT * divisors, FAST_FLOAT * workspace); + +EXTERN(void) jsimd_quantize_float_mips_dspr2 + (JCOEFPTR coef_block, FAST_FLOAT * divisors, FAST_FLOAT * workspace); + +/* Scaled Inverse DCT */ +EXTERN(void) jsimd_idct_2x2_mmx + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); +EXTERN(void) jsimd_idct_4x4_mmx + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); + +extern const int jconst_idct_red_sse2[]; +EXTERN(void) jsimd_idct_2x2_sse2 + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); +EXTERN(void) jsimd_idct_4x4_sse2 + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); + +EXTERN(void) jsimd_idct_2x2_neon + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); +EXTERN(void) jsimd_idct_4x4_neon + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); + +EXTERN(void) jsimd_idct_2x2_mips_dspr2 + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); +EXTERN(void) jsimd_idct_4x4_mips_dspr2 + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col, int * workspace); +EXTERN(void) jsimd_idct_6x6_mips_dspr2 + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); +EXTERN(void) jsimd_idct_12x12_pass1_mips_dspr2 + (JCOEFPTR coef_block, void * dct_table, int * workspace); +EXTERN(void) jsimd_idct_12x12_pass2_mips_dspr2 + (int * workspace, int * output); + +/* Slow Integer Inverse DCT */ +EXTERN(void) jsimd_idct_islow_mmx + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); + +extern const int jconst_idct_islow_sse2[]; +EXTERN(void) jsimd_idct_islow_sse2 + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); + +EXTERN(void) jsimd_idct_islow_neon + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); + +EXTERN(void) jsimd_idct_islow_mips_dspr2 + (void * dct_table, JCOEFPTR coef_block, int * output_buf, + JSAMPLE * output_col); + +/* Fast Integer Inverse DCT */ +EXTERN(void) jsimd_idct_ifast_mmx + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); + +extern const int jconst_idct_ifast_sse2[]; +EXTERN(void) jsimd_idct_ifast_sse2 + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); + +EXTERN(void) jsimd_idct_ifast_neon + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); + +EXTERN(void) jsimd_idct_ifast_cols_mips_dspr2 + (JCOEF * inptr, IFAST_MULT_TYPE * quantptr, DCTELEM * wsptr, + const int * idct_coefs); +EXTERN(void) jsimd_idct_ifast_rows_mips_dspr2 + (DCTELEM * wsptr, JSAMPARRAY output_buf, JDIMENSION output_col, + const int * idct_coefs); + +/* Floating Point Inverse DCT */ +EXTERN(void) jsimd_idct_float_3dnow + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); + +extern const int jconst_idct_float_sse[]; +EXTERN(void) jsimd_idct_float_sse + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); + +extern const int jconst_idct_float_sse2[]; +EXTERN(void) jsimd_idct_float_sse2 + (void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col); diff --git a/app/src/main/jni/libjpeg-turbo/simd/jsimd_arm.c b/app/src/main/jni/libjpeg-turbo/simd/jsimd_arm.c new file mode 100644 index 000000000..4cbcf2d57 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jsimd_arm.c @@ -0,0 +1,708 @@ +/* + * jsimd_arm.c + * + * Copyright 2009 Pierre Ossman for Cendio AB + * Copyright 2009-2011, 2013-2014 D. R. Commander + * + * Based on the x86 SIMD extension for IJG JPEG library, + * Copyright (C) 1999-2006, MIYASAKA Masaru. + * For conditions of distribution and use, see copyright notice in jsimdext.inc + * + * This file contains the interface between the "normal" portions + * of the library and the SIMD implementations when running on a + * 32-bit ARM architecture. + */ + +#define JPEG_INTERNALS +#include "../jinclude.h" +#include "../jpeglib.h" +#include "../jsimd.h" +#include "../jdct.h" +#include "../jsimddct.h" +#include "jsimd.h" + +#include +#include +#include + +static unsigned int simd_support = ~0; + +#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__) + +#define SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT (1024 * 1024) + +LOCAL(int) +check_feature (char *buffer, char *feature) +{ + char *p; + if (*feature == 0) + return 0; + if (strncmp(buffer, "Features", 8) != 0) + return 0; + buffer += 8; + while (isspace(*buffer)) + buffer++; + + /* Check if 'feature' is present in the buffer as a separate word */ + while ((p = strstr(buffer, feature))) { + if (p > buffer && !isspace(*(p - 1))) { + buffer++; + continue; + } + p += strlen(feature); + if (*p != 0 && !isspace(*p)) { + buffer++; + continue; + } + return 1; + } + return 0; +} + +LOCAL(int) +parse_proc_cpuinfo (int bufsize) +{ + char *buffer = (char *)malloc(bufsize); + FILE *fd; + simd_support = 0; + + if (!buffer) + return 0; + + fd = fopen("/proc/cpuinfo", "r"); + if (fd) { + while (fgets(buffer, bufsize, fd)) { + if (!strchr(buffer, '\n') && !feof(fd)) { + /* "impossible" happened - insufficient size of the buffer! */ + fclose(fd); + free(buffer); + return 0; + } + if (check_feature(buffer, "neon")) + simd_support |= JSIMD_ARM_NEON; + } + fclose(fd); + } + free(buffer); + return 1; +} + +#endif + +/* + * Check what SIMD accelerations are supported. + * + * FIXME: This code is racy under a multi-threaded environment. + */ +LOCAL(void) +init_simd (void) +{ + char *env = NULL; +#if !defined(__ARM_NEON__) && defined(__linux__) || defined(ANDROID) || defined(__ANDROID__) + int bufsize = 1024; /* an initial guess for the line buffer size limit */ +#endif + + if (simd_support != ~0U) + return; + + simd_support = 0; + +#if defined(__ARM_NEON__) + simd_support |= JSIMD_ARM_NEON; +#elif defined(__linux__) || defined(ANDROID) || defined(__ANDROID__) + /* We still have a chance to use NEON regardless of globally used + * -mcpu/-mfpu options passed to gcc by performing runtime detection via + * /proc/cpuinfo parsing on linux/android */ + while (!parse_proc_cpuinfo(bufsize)) { + bufsize *= 2; + if (bufsize > SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT) + break; + } +#endif + + /* Force different settings through environment variables */ + env = getenv("JSIMD_FORCENEON"); + if ((env != NULL) && (strcmp(env, "1") == 0)) + simd_support &= JSIMD_ARM_NEON; + env = getenv("JSIMD_FORCENONE"); + if ((env != NULL) && (strcmp(env, "1") == 0)) + simd_support = 0; +} + +GLOBAL(int) +jsimd_can_rgb_ycc (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_rgb_gray (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_ycc_rgb (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_ycc_rgb565 (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_rgb_ycc_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ + void (*neonfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int); + + switch(cinfo->in_color_space) { + case JCS_EXT_RGB: + neonfct=jsimd_extrgb_ycc_convert_neon; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + neonfct=jsimd_extrgbx_ycc_convert_neon; + break; + case JCS_EXT_BGR: + neonfct=jsimd_extbgr_ycc_convert_neon; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + neonfct=jsimd_extbgrx_ycc_convert_neon; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + neonfct=jsimd_extxbgr_ycc_convert_neon; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + neonfct=jsimd_extxrgb_ycc_convert_neon; + break; + default: + neonfct=jsimd_extrgb_ycc_convert_neon; + break; + } + + if (simd_support & JSIMD_ARM_NEON) + neonfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows); +} + +GLOBAL(void) +jsimd_rgb_gray_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ +} + +GLOBAL(void) +jsimd_ycc_rgb_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + void (*neonfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int); + + switch(cinfo->out_color_space) { + case JCS_EXT_RGB: + neonfct=jsimd_ycc_extrgb_convert_neon; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + neonfct=jsimd_ycc_extrgbx_convert_neon; + break; + case JCS_EXT_BGR: + neonfct=jsimd_ycc_extbgr_convert_neon; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + neonfct=jsimd_ycc_extbgrx_convert_neon; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + neonfct=jsimd_ycc_extxbgr_convert_neon; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + neonfct=jsimd_ycc_extxrgb_convert_neon; + break; + default: + neonfct=jsimd_ycc_extrgb_convert_neon; + break; + } + + if (simd_support & JSIMD_ARM_NEON) + neonfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows); +} + +GLOBAL(void) +jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_ycc_rgb565_convert_neon(cinfo->output_width, input_buf, input_row, + output_buf, num_rows); +} + +GLOBAL(int) +jsimd_can_h2v2_downsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_downsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ +} + +GLOBAL(void) +jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ +} + +GLOBAL(int) +jsimd_can_h2v2_upsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_upsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ +} + +GLOBAL(void) +jsimd_h2v1_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ +} + +GLOBAL(int) +jsimd_can_h2v2_fancy_upsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_fancy_upsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ +} + +GLOBAL(void) +jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_h2v1_fancy_upsample_neon(cinfo->max_v_samp_factor, + compptr->downsampled_width, input_data, + output_data_ptr); +} + +GLOBAL(int) +jsimd_can_h2v2_merged_upsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_merged_upsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ +} + +GLOBAL(void) +jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ +} + +GLOBAL(int) +jsimd_can_convsamp (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_convsamp_float (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col, + DCTELEM * workspace) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_convsamp_neon(sample_data, start_col, workspace); +} + +GLOBAL(void) +jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col, + FAST_FLOAT * workspace) +{ +} + +GLOBAL(int) +jsimd_can_fdct_islow (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_fdct_ifast (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_fdct_float (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_fdct_islow (DCTELEM * data) +{ +} + +GLOBAL(void) +jsimd_fdct_ifast (DCTELEM * data) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_fdct_ifast_neon(data); +} + +GLOBAL(void) +jsimd_fdct_float (FAST_FLOAT * data) +{ +} + +GLOBAL(int) +jsimd_can_quantize (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_quantize_float (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_quantize (JCOEFPTR coef_block, DCTELEM * divisors, + DCTELEM * workspace) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_quantize_neon(coef_block, divisors, workspace); +} + +GLOBAL(void) +jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT * divisors, + FAST_FLOAT * workspace) +{ +} + +GLOBAL(int) +jsimd_can_idct_2x2 (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_4x4 (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_idct_2x2_neon(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(void) +jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_idct_4x4_neon(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(int) +jsimd_can_idct_islow (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_ifast (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(IFAST_MULT_TYPE) != 2) + return 0; + if (IFAST_SCALE_BITS != 2) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_float (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_idct_islow_neon(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(void) +jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_idct_ifast_neon(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(void) +jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ +} diff --git a/app/src/main/jni/libjpeg-turbo/simd/jsimd_arm64.c b/app/src/main/jni/libjpeg-turbo/simd/jsimd_arm64.c new file mode 100644 index 000000000..65724cb9f --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jsimd_arm64.c @@ -0,0 +1,544 @@ +/* + * jsimd_arm64.c + * + * Copyright 2009 Pierre Ossman for Cendio AB + * Copyright 2009-2011, 2013-2014 D. R. Commander + * + * Based on the x86 SIMD extension for IJG JPEG library, + * Copyright (C) 1999-2006, MIYASAKA Masaru. + * For conditions of distribution and use, see copyright notice in jsimdext.inc + * + * This file contains the interface between the "normal" portions + * of the library and the SIMD implementations when running on a + * 64-bit ARM architecture. + */ + +#define JPEG_INTERNALS +#include "../jinclude.h" +#include "../jpeglib.h" +#include "../jsimd.h" +#include "../jdct.h" +#include "../jsimddct.h" +#include "jsimd.h" + +#include +#include +#include + +static unsigned int simd_support = ~0; + +/* + * Check what SIMD accelerations are supported. + * + * FIXME: This code is racy under a multi-threaded environment. + */ + +/* + * ARMv8 architectures support NEON extensions by default. + * It is no longer optional as it was with ARMv7. + */ + + +LOCAL(void) +init_simd (void) +{ + char *env = NULL; + + if (simd_support != ~0U) + return; + + simd_support = 0; + + simd_support |= JSIMD_ARM_NEON; + + /* Force different settings through environment variables */ + env = getenv("JSIMD_FORCENEON"); + if ((env != NULL) && (strcmp(env, "1") == 0)) + simd_support &= JSIMD_ARM_NEON; + env = getenv("JSIMD_FORCENONE"); + if ((env != NULL) && (strcmp(env, "1") == 0)) + simd_support = 0; +} + +GLOBAL(int) +jsimd_can_rgb_ycc (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_rgb_gray (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_ycc_rgb (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_ycc_rgb565 (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_rgb_ycc_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ +} + +GLOBAL(void) +jsimd_rgb_gray_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ +} + +GLOBAL(void) +jsimd_ycc_rgb_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + void (*neonfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int); + + switch(cinfo->out_color_space) { + case JCS_EXT_RGB: + neonfct=jsimd_ycc_extrgb_convert_neon; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + neonfct=jsimd_ycc_extrgbx_convert_neon; + break; + case JCS_EXT_BGR: + neonfct=jsimd_ycc_extbgr_convert_neon; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + neonfct=jsimd_ycc_extbgrx_convert_neon; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + neonfct=jsimd_ycc_extxbgr_convert_neon; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + neonfct=jsimd_ycc_extxrgb_convert_neon; + break; + default: + neonfct=jsimd_ycc_extrgb_convert_neon; + break; + } + + if (simd_support & JSIMD_ARM_NEON) + neonfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows); +} + +GLOBAL(void) +jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_ycc_rgb565_convert_neon(cinfo->output_width, input_buf, input_row, + output_buf, num_rows); +} + +GLOBAL(int) +jsimd_can_h2v2_downsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_downsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ +} + +GLOBAL(void) +jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ +} + +GLOBAL(int) +jsimd_can_h2v2_upsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_upsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ +} + +GLOBAL(void) +jsimd_h2v1_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ +} + +GLOBAL(int) +jsimd_can_h2v2_fancy_upsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_fancy_upsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ +} + +GLOBAL(void) +jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ +} + +GLOBAL(int) +jsimd_can_h2v2_merged_upsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_merged_upsample (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ +} + +GLOBAL(void) +jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ +} + +GLOBAL(int) +jsimd_can_convsamp (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_convsamp_float (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col, + DCTELEM * workspace) +{ +} + +GLOBAL(void) +jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col, + FAST_FLOAT * workspace) +{ +} + +GLOBAL(int) +jsimd_can_fdct_islow (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_fdct_ifast (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_fdct_float (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_fdct_islow (DCTELEM * data) +{ +} + +GLOBAL(void) +jsimd_fdct_ifast (DCTELEM * data) +{ +} + +GLOBAL(void) +jsimd_fdct_float (FAST_FLOAT * data) +{ +} + +GLOBAL(int) +jsimd_can_quantize (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(int) +jsimd_can_quantize_float (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_quantize (JCOEFPTR coef_block, DCTELEM * divisors, + DCTELEM * workspace) +{ +} + +GLOBAL(void) +jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT * divisors, + FAST_FLOAT * workspace) +{ +} + +GLOBAL(int) +jsimd_can_idct_2x2 (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_4x4 (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_idct_2x2_neon(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(void) +jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_idct_4x4_neon(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(int) +jsimd_can_idct_islow (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_ifast (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(IFAST_MULT_TYPE) != 2) + return 0; + if (IFAST_SCALE_BITS != 2) + return 0; + + if (simd_support & JSIMD_ARM_NEON) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_float (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_idct_islow_neon(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(void) +jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if (simd_support & JSIMD_ARM_NEON) + jsimd_idct_ifast_neon(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(void) +jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ +} diff --git a/app/src/main/jni/libjpeg-turbo/simd/jsimd_arm64_neon.S b/app/src/main/jni/libjpeg-turbo/simd/jsimd_arm64_neon.S new file mode 100644 index 000000000..2186f2485 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jsimd_arm64_neon.S @@ -0,0 +1,1862 @@ +/* + * ARMv8 NEON optimizations for libjpeg-turbo + * + * Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies). + * All rights reserved. + * Author: Siarhei Siamashka + * Copyright (C) 2013-2014, Linaro Limited + * Author: Ragesh Radhakrishnan + * Copyright (C) 2014, D. R. Commander. All rights reserved. + * + * This software is provided 'as-is', without any express or implied + * warranty. In no event will the authors be held liable for any damages + * arising from the use of this software. + * + * Permission is granted to anyone to use this software for any purpose, + * including commercial applications, and to alter it and redistribute it + * freely, subject to the following restrictions: + * + * 1. The origin of this software must not be misrepresented; you must not + * claim that you wrote the original software. If you use this software + * in a product, an acknowledgment in the product documentation would be + * appreciated but is not required. + * 2. Altered source versions must be plainly marked as such, and must not be + * misrepresented as being the original software. + * 3. This notice may not be removed or altered from any source distribution. + */ + +#if defined(__linux__) && defined(__ELF__) +.section .note.GNU-stack,"",%progbits /* mark stack as non-executable */ +#endif + +.text +.arch armv8-a+fp+simd + + +#define RESPECT_STRICT_ALIGNMENT 1 + + +/*****************************************************************************/ + +/* Supplementary macro for setting function attributes */ +.macro asm_function fname +#ifdef __APPLE__ + .globl _\fname +_\fname: +#else + .global \fname +#ifdef __ELF__ + .hidden \fname + .type \fname, %function +#endif +\fname: +#endif +.endm + +/* Transpose elements of single 128 bit registers */ +.macro transpose_single x0,x1,xi,xilen,literal + ins \xi\xilen[0], \x0\xilen[0] + ins \x1\xilen[0], \x0\xilen[1] + trn1 \x0\literal, \x0\literal, \x1\literal + trn2 \x1\literal, \xi\literal, \x1\literal +.endm + +/* Transpose elements of 2 differnet registers */ +.macro transpose x0,x1,xi,xilen,literal + mov \xi\xilen, \x0\xilen + trn1 \x0\literal, \x0\literal, \x1\literal + trn2 \x1\literal, \xi\literal, \x1\literal +.endm + +/* Transpose a block of 4x4 coefficients in four 64-bit registers */ +.macro transpose_4x4_32 x0,x0len x1,x1len x2,x2len x3,x3len,xi,xilen + mov \xi\xilen, \x0\xilen + trn1 \x0\x0len, \x0\x0len, \x2\x2len + trn2 \x2\x2len, \xi\x0len, \x2\x2len + mov \xi\xilen, \x1\xilen + trn1 \x1\x1len, \x1\x1len, \x3\x3len + trn2 \x3\x3len, \xi\x1len, \x3\x3len +.endm + +.macro transpose_4x4_16 x0,x0len x1,x1len, x2,x2len, x3,x3len,xi,xilen + mov \xi\xilen, \x0\xilen + trn1 \x0\x0len, \x0\x0len, \x1\x1len + trn2 \x1\x2len, \xi\x0len, \x1\x2len + mov \xi\xilen, \x2\xilen + trn1 \x2\x2len, \x2\x2len, \x3\x3len + trn2 \x3\x2len, \xi\x1len, \x3\x3len +.endm + +.macro transpose_4x4 x0, x1, x2, x3,x5 + transpose_4x4_16 \x0,.4h, \x1,.4h, \x2,.4h,\x3,.4h,\x5,.16b + transpose_4x4_32 \x0,.2s, \x1,.2s, \x2,.2s,\x3,.2s,\x5,.16b +.endm + + +#define CENTERJSAMPLE 128 + +/*****************************************************************************/ + +/* + * Perform dequantization and inverse DCT on one block of coefficients. + * + * GLOBAL(void) + * jsimd_idct_islow_neon (void * dct_table, JCOEFPTR coef_block, + * JSAMPARRAY output_buf, JDIMENSION output_col) + */ + +#define FIX_0_298631336 (2446) +#define FIX_0_390180644 (3196) +#define FIX_0_541196100 (4433) +#define FIX_0_765366865 (6270) +#define FIX_0_899976223 (7373) +#define FIX_1_175875602 (9633) +#define FIX_1_501321110 (12299) +#define FIX_1_847759065 (15137) +#define FIX_1_961570560 (16069) +#define FIX_2_053119869 (16819) +#define FIX_2_562915447 (20995) +#define FIX_3_072711026 (25172) + +#define FIX_1_175875602_MINUS_1_961570560 (FIX_1_175875602 - FIX_1_961570560) +#define FIX_1_175875602_MINUS_0_390180644 (FIX_1_175875602 - FIX_0_390180644) +#define FIX_0_541196100_MINUS_1_847759065 (FIX_0_541196100 - FIX_1_847759065) +#define FIX_3_072711026_MINUS_2_562915447 (FIX_3_072711026 - FIX_2_562915447) +#define FIX_0_298631336_MINUS_0_899976223 (FIX_0_298631336 - FIX_0_899976223) +#define FIX_1_501321110_MINUS_0_899976223 (FIX_1_501321110 - FIX_0_899976223) +#define FIX_2_053119869_MINUS_2_562915447 (FIX_2_053119869 - FIX_2_562915447) +#define FIX_0_541196100_PLUS_0_765366865 (FIX_0_541196100 + FIX_0_765366865) + +/* + * Reference SIMD-friendly 1-D ISLOW iDCT C implementation. + * Uses some ideas from the comments in 'simd/jiss2int-64.asm' + */ +#define REF_1D_IDCT(xrow0, xrow1, xrow2, xrow3, xrow4, xrow5, xrow6, xrow7) \ +{ \ + DCTELEM row0, row1, row2, row3, row4, row5, row6, row7; \ + INT32 q1, q2, q3, q4, q5, q6, q7; \ + INT32 tmp11_plus_tmp2, tmp11_minus_tmp2; \ + \ + /* 1-D iDCT input data */ \ + row0 = xrow0; \ + row1 = xrow1; \ + row2 = xrow2; \ + row3 = xrow3; \ + row4 = xrow4; \ + row5 = xrow5; \ + row6 = xrow6; \ + row7 = xrow7; \ + \ + q5 = row7 + row3; \ + q4 = row5 + row1; \ + q6 = MULTIPLY(q5, FIX_1_175875602_MINUS_1_961570560) + \ + MULTIPLY(q4, FIX_1_175875602); \ + q7 = MULTIPLY(q5, FIX_1_175875602) + \ + MULTIPLY(q4, FIX_1_175875602_MINUS_0_390180644); \ + q2 = MULTIPLY(row2, FIX_0_541196100) + \ + MULTIPLY(row6, FIX_0_541196100_MINUS_1_847759065); \ + q4 = q6; \ + q3 = ((INT32) row0 - (INT32) row4) << 13; \ + q6 += MULTIPLY(row5, -FIX_2_562915447) + \ + MULTIPLY(row3, FIX_3_072711026_MINUS_2_562915447); \ + /* now we can use q1 (reloadable constants have been used up) */ \ + q1 = q3 + q2; \ + q4 += MULTIPLY(row7, FIX_0_298631336_MINUS_0_899976223) + \ + MULTIPLY(row1, -FIX_0_899976223); \ + q5 = q7; \ + q1 = q1 + q6; \ + q7 += MULTIPLY(row7, -FIX_0_899976223) + \ + MULTIPLY(row1, FIX_1_501321110_MINUS_0_899976223); \ + \ + /* (tmp11 + tmp2) has been calculated (out_row1 before descale) */ \ + tmp11_plus_tmp2 = q1; \ + row1 = 0; \ + \ + q1 = q1 - q6; \ + q5 += MULTIPLY(row5, FIX_2_053119869_MINUS_2_562915447) + \ + MULTIPLY(row3, -FIX_2_562915447); \ + q1 = q1 - q6; \ + q6 = MULTIPLY(row2, FIX_0_541196100_PLUS_0_765366865) + \ + MULTIPLY(row6, FIX_0_541196100); \ + q3 = q3 - q2; \ + \ + /* (tmp11 - tmp2) has been calculated (out_row6 before descale) */ \ + tmp11_minus_tmp2 = q1; \ + \ + q1 = ((INT32) row0 + (INT32) row4) << 13; \ + q2 = q1 + q6; \ + q1 = q1 - q6; \ + \ + /* pick up the results */ \ + tmp0 = q4; \ + tmp1 = q5; \ + tmp2 = (tmp11_plus_tmp2 - tmp11_minus_tmp2) / 2; \ + tmp3 = q7; \ + tmp10 = q2; \ + tmp11 = (tmp11_plus_tmp2 + tmp11_minus_tmp2) / 2; \ + tmp12 = q3; \ + tmp13 = q1; \ +} + +#define XFIX_0_899976223 v0.h[0] +#define XFIX_0_541196100 v0.h[1] +#define XFIX_2_562915447 v0.h[2] +#define XFIX_0_298631336_MINUS_0_899976223 v0.h[3] +#define XFIX_1_501321110_MINUS_0_899976223 v1.h[0] +#define XFIX_2_053119869_MINUS_2_562915447 v1.h[1] +#define XFIX_0_541196100_PLUS_0_765366865 v1.h[2] +#define XFIX_1_175875602 v1.h[3] +#define XFIX_1_175875602_MINUS_0_390180644 v2.h[0] +#define XFIX_0_541196100_MINUS_1_847759065 v2.h[1] +#define XFIX_3_072711026_MINUS_2_562915447 v2.h[2] +#define XFIX_1_175875602_MINUS_1_961570560 v2.h[3] + +.balign 16 +Ljsimd_idct_islow_neon_consts: + .short FIX_0_899976223 /* d0[0] */ + .short FIX_0_541196100 /* d0[1] */ + .short FIX_2_562915447 /* d0[2] */ + .short FIX_0_298631336_MINUS_0_899976223 /* d0[3] */ + .short FIX_1_501321110_MINUS_0_899976223 /* d1[0] */ + .short FIX_2_053119869_MINUS_2_562915447 /* d1[1] */ + .short FIX_0_541196100_PLUS_0_765366865 /* d1[2] */ + .short FIX_1_175875602 /* d1[3] */ + /* reloadable constants */ + .short FIX_1_175875602_MINUS_0_390180644 /* d2[0] */ + .short FIX_0_541196100_MINUS_1_847759065 /* d2[1] */ + .short FIX_3_072711026_MINUS_2_562915447 /* d2[2] */ + .short FIX_1_175875602_MINUS_1_961570560 /* d2[3] */ + +asm_function jsimd_idct_islow_neon + + DCT_TABLE .req x0 + COEF_BLOCK .req x1 + OUTPUT_BUF .req x2 + OUTPUT_COL .req x3 + TMP1 .req x0 + TMP2 .req x1 + TMP3 .req x2 + TMP4 .req x15 + + ROW0L .req v16 + ROW0R .req v17 + ROW1L .req v18 + ROW1R .req v19 + ROW2L .req v20 + ROW2R .req v21 + ROW3L .req v22 + ROW3R .req v23 + ROW4L .req v24 + ROW4R .req v25 + ROW5L .req v26 + ROW5R .req v27 + ROW6L .req v28 + ROW6R .req v29 + ROW7L .req v30 + ROW7R .req v31 + /* Save all NEON registers and x15 (32 NEON registers * 8 bytes + 16) */ + sub sp, sp, 272 + str x15, [sp], 16 + adr x15, Ljsimd_idct_islow_neon_consts + st1 {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32 + st1 {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32 + st1 {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32 + st1 {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32 + st1 {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32 + st1 {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32 + st1 {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32 + st1 {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32 + ld1 {v16.4h, v17.4h, v18.4h, v19.4h}, [COEF_BLOCK], 32 + ld1 {v0.4h, v1.4h, v2.4h, v3.4h}, [DCT_TABLE], 32 + ld1 {v20.4h, v21.4h, v22.4h, v23.4h}, [COEF_BLOCK], 32 + mul v16.4h, v16.4h, v0.4h + mul v17.4h, v17.4h, v1.4h + ins v16.d[1], v17.d[0] /* 128 bit q8 */ + ld1 {v4.4h, v5.4h, v6.4h, v7.4h}, [DCT_TABLE], 32 + mul v18.4h, v18.4h, v2.4h + mul v19.4h, v19.4h, v3.4h + ins v18.d[1], v19.d[0] /* 128 bit q9 */ + ld1 {v24.4h, v25.4h, v26.4h, v27.4h}, [COEF_BLOCK], 32 + mul v20.4h, v20.4h, v4.4h + mul v21.4h, v21.4h, v5.4h + ins v20.d[1], v21.d[0] /* 128 bit q10 */ + ld1 {v0.4h, v1.4h, v2.4h, v3.4h}, [DCT_TABLE], 32 + mul v22.4h, v22.4h, v6.4h + mul v23.4h, v23.4h, v7.4h + ins v22.d[1], v23.d[0] /* 128 bit q11 */ + ld1 {v28.4h, v29.4h, v30.4h, v31.4h}, [COEF_BLOCK] + mul v24.4h, v24.4h, v0.4h + mul v25.4h, v25.4h, v1.4h + ins v24.d[1], v25.d[0] /* 128 bit q12 */ + ld1 {v4.4h, v5.4h, v6.4h, v7.4h}, [DCT_TABLE], 32 + mul v28.4h, v28.4h, v4.4h + mul v29.4h, v29.4h, v5.4h + ins v28.d[1], v29.d[0] /* 128 bit q14 */ + mul v26.4h, v26.4h, v2.4h + mul v27.4h, v27.4h, v3.4h + ins v26.d[1], v27.d[0] /* 128 bit q13 */ + ld1 {v0.4h, v1.4h, v2.4h, v3.4h}, [x15] /* load constants */ + add x15, x15, #16 + mul v30.4h, v30.4h, v6.4h + mul v31.4h, v31.4h, v7.4h + ins v30.d[1], v31.d[0] /* 128 bit q15 */ + /* Go to the bottom of the stack */ + sub sp, sp, 352 + stp x4, x5, [sp], 16 + st1 {v8.4h, v9.4h, v10.4h, v11.4h}, [sp], 32 /* save NEON registers */ + st1 {v12.4h, v13.4h, v14.4h, v15.4h}, [sp], 32 + /* 1-D IDCT, pass 1, left 4x8 half */ + add v4.4h, ROW7L.4h, ROW3L.4h + add v5.4h, ROW5L.4h, ROW1L.4h + smull v12.4s, v4.4h, XFIX_1_175875602_MINUS_1_961570560 + smlal v12.4s, v5.4h, XFIX_1_175875602 + smull v14.4s, v4.4h, XFIX_1_175875602 + /* Check for the zero coefficients in the right 4x8 half */ + smlal v14.4s, v5.4h, XFIX_1_175875602_MINUS_0_390180644 + ssubl v6.4s, ROW0L.4h, ROW4L.4h + ldp w4, w5, [COEF_BLOCK, #(-96 + 2 * (4 + 1 * 8))] + smull v4.4s, ROW2L.4h, XFIX_0_541196100 + smlal v4.4s, ROW6L.4h, XFIX_0_541196100_MINUS_1_847759065 + orr x0, x4, x5 + mov v8.16b, v12.16b + smlsl v12.4s, ROW5L.4h, XFIX_2_562915447 + ldp w4, w5, [COEF_BLOCK, #(-96 + 2 * (4 + 2 * 8))] + smlal v12.4s, ROW3L.4h, XFIX_3_072711026_MINUS_2_562915447 + shl v6.4s, v6.4s, #13 + orr x0, x0, x4 + smlsl v8.4s, ROW1L.4h, XFIX_0_899976223 + orr x0, x0 , x5 + add v2.4s, v6.4s, v4.4s + ldp w4, w5, [COEF_BLOCK, #(-96 + 2 * (4 + 3 * 8))] + mov v10.16b, v14.16b + add v2.4s, v2.4s, v12.4s + orr x0, x0, x4 + smlsl v14.4s, ROW7L.4h, XFIX_0_899976223 + orr x0, x0, x5 + smlal v14.4s, ROW1L.4h, XFIX_1_501321110_MINUS_0_899976223 + rshrn ROW1L.4h, v2.4s, #11 + ldp w4, w5, [COEF_BLOCK, #(-96 + 2 * (4 + 4 * 8))] + sub v2.4s, v2.4s, v12.4s + smlal v10.4s, ROW5L.4h, XFIX_2_053119869_MINUS_2_562915447 + orr x0, x0, x4 + smlsl v10.4s, ROW3L.4h, XFIX_2_562915447 + orr x0, x0, x5 + sub v2.4s, v2.4s, v12.4s + smull v12.4s, ROW2L.4h, XFIX_0_541196100_PLUS_0_765366865 + ldp w4, w5, [COEF_BLOCK, #(-96 + 2 * (4 + 5 * 8))] + smlal v12.4s, ROW6L.4h, XFIX_0_541196100 + sub v6.4s, v6.4s, v4.4s + orr x0, x0, x4 + rshrn ROW6L.4h, v2.4s, #11 + orr x0, x0, x5 + add v2.4s, v6.4s, v10.4s + ldp w4, w5, [COEF_BLOCK, #(-96 + 2 * (4 + 6 * 8))] + sub v6.4s, v6.4s, v10.4s + saddl v10.4s, ROW0L.4h, ROW4L.4h + orr x0, x0, x4 + rshrn ROW2L.4h, v2.4s, #11 + orr x0, x0, x5 + rshrn ROW5L.4h, v6.4s, #11 + ldp w4, w5, [COEF_BLOCK, #(-96 + 2 * (4 + 7 * 8))] + shl v10.4s, v10.4s, #13 + smlal v8.4s, ROW7L.4h, XFIX_0_298631336_MINUS_0_899976223 + orr x0, x0, x4 + add v4.4s, v10.4s, v12.4s + orr x0, x0, x5 + cmp x0, #0 /* orrs instruction removed */ + sub v2.4s, v10.4s, v12.4s + add v12.4s, v4.4s, v14.4s + ldp w4, w5, [COEF_BLOCK, #(-96 + 2 * (4 + 0 * 8))] + sub v4.4s, v4.4s, v14.4s + add v10.4s, v2.4s, v8.4s + orr x0, x4, x5 + sub v6.4s, v2.4s, v8.4s + /* pop {x4, x5} */ + sub sp, sp, 80 + ldp x4, x5, [sp], 16 + rshrn ROW7L.4h, v4.4s, #11 + rshrn ROW3L.4h, v10.4s, #11 + rshrn ROW0L.4h, v12.4s, #11 + rshrn ROW4L.4h, v6.4s, #11 + + b.eq 3f /* Go to do some special handling for the sparse right 4x8 half */ + + /* 1-D IDCT, pass 1, right 4x8 half */ + ld1 {v2.4h}, [x15] /* reload constants */ + add v10.4h, ROW7R.4h, ROW3R.4h + add v8.4h, ROW5R.4h, ROW1R.4h + /* Transpose ROW6L <-> ROW7L (v3 available free register) */ + transpose ROW6L, ROW7L, v3, .16b, .4h + smull v12.4s, v10.4h, XFIX_1_175875602_MINUS_1_961570560 + smlal v12.4s, v8.4h, XFIX_1_175875602 + /* Transpose ROW2L <-> ROW3L (v3 available free register) */ + transpose ROW2L, ROW3L, v3, .16b, .4h + smull v14.4s, v10.4h, XFIX_1_175875602 + smlal v14.4s, v8.4h, XFIX_1_175875602_MINUS_0_390180644 + /* Transpose ROW0L <-> ROW1L (v3 available free register) */ + transpose ROW0L, ROW1L, v3, .16b, .4h + ssubl v6.4s, ROW0R.4h, ROW4R.4h + smull v4.4s, ROW2R.4h, XFIX_0_541196100 + smlal v4.4s, ROW6R.4h, XFIX_0_541196100_MINUS_1_847759065 + /* Transpose ROW4L <-> ROW5L (v3 available free register) */ + transpose ROW4L, ROW5L, v3, .16b, .4h + mov v8.16b, v12.16b + smlsl v12.4s, ROW5R.4h, XFIX_2_562915447 + smlal v12.4s, ROW3R.4h, XFIX_3_072711026_MINUS_2_562915447 + /* Transpose ROW1L <-> ROW3L (v3 available free register) */ + transpose ROW1L, ROW3L, v3, .16b, .2s + shl v6.4s, v6.4s, #13 + smlsl v8.4s, ROW1R.4h, XFIX_0_899976223 + /* Transpose ROW4L <-> ROW6L (v3 available free register) */ + transpose ROW4L, ROW6L, v3, .16b, .2s + add v2.4s, v6.4s, v4.4s + mov v10.16b, v14.16b + add v2.4s, v2.4s, v12.4s + /* Transpose ROW0L <-> ROW2L (v3 available free register) */ + transpose ROW0L, ROW2L, v3, .16b, .2s + smlsl v14.4s, ROW7R.4h, XFIX_0_899976223 + smlal v14.4s, ROW1R.4h, XFIX_1_501321110_MINUS_0_899976223 + rshrn ROW1R.4h, v2.4s, #11 + /* Transpose ROW5L <-> ROW7L (v3 available free register) */ + transpose ROW5L, ROW7L, v3, .16b, .2s + sub v2.4s, v2.4s, v12.4s + smlal v10.4s, ROW5R.4h, XFIX_2_053119869_MINUS_2_562915447 + smlsl v10.4s, ROW3R.4h, XFIX_2_562915447 + sub v2.4s, v2.4s, v12.4s + smull v12.4s, ROW2R.4h, XFIX_0_541196100_PLUS_0_765366865 + smlal v12.4s, ROW6R.4h, XFIX_0_541196100 + sub v6.4s, v6.4s, v4.4s + rshrn ROW6R.4h, v2.4s, #11 + add v2.4s, v6.4s, v10.4s + sub v6.4s, v6.4s, v10.4s + saddl v10.4s, ROW0R.4h, ROW4R.4h + rshrn ROW2R.4h, v2.4s, #11 + rshrn ROW5R.4h, v6.4s, #11 + shl v10.4s, v10.4s, #13 + smlal v8.4s, ROW7R.4h, XFIX_0_298631336_MINUS_0_899976223 + add v4.4s, v10.4s, v12.4s + sub v2.4s, v10.4s, v12.4s + add v12.4s, v4.4s, v14.4s + sub v4.4s, v4.4s, v14.4s + add v10.4s, v2.4s, v8.4s + sub v6.4s, v2.4s, v8.4s + rshrn ROW7R.4h, v4.4s, #11 + rshrn ROW3R.4h, v10.4s, #11 + rshrn ROW0R.4h, v12.4s, #11 + rshrn ROW4R.4h, v6.4s, #11 + /* Transpose right 4x8 half */ + transpose ROW6R, ROW7R, v3, .16b, .4h + transpose ROW2R, ROW3R, v3, .16b, .4h + transpose ROW0R, ROW1R, v3, .16b, .4h + transpose ROW4R, ROW5R, v3, .16b, .4h + transpose ROW1R, ROW3R, v3, .16b, .2s + transpose ROW4R, ROW6R, v3, .16b, .2s + transpose ROW0R, ROW2R, v3, .16b, .2s + transpose ROW5R, ROW7R, v3, .16b, .2s + +1: /* 1-D IDCT, pass 2 (normal variant), left 4x8 half */ + ld1 {v2.4h}, [x15] /* reload constants */ + smull v12.4S, ROW1R.4h, XFIX_1_175875602 /* ROW5L.4h <-> ROW1R.4h */ + smlal v12.4s, ROW1L.4h, XFIX_1_175875602 + smlal v12.4s, ROW3R.4h, XFIX_1_175875602_MINUS_1_961570560 /* ROW7L.4h <-> ROW3R.4h */ + smlal v12.4s, ROW3L.4h, XFIX_1_175875602_MINUS_1_961570560 + smull v14.4s, ROW3R.4h, XFIX_1_175875602 /* ROW7L.4h <-> ROW3R.4h */ + smlal v14.4s, ROW3L.4h, XFIX_1_175875602 + smlal v14.4s, ROW1R.4h, XFIX_1_175875602_MINUS_0_390180644 /* ROW5L.4h <-> ROW1R.4h */ + smlal v14.4s, ROW1L.4h, XFIX_1_175875602_MINUS_0_390180644 + ssubl v6.4s, ROW0L.4h, ROW0R.4h /* ROW4L.4h <-> ROW0R.4h */ + smull v4.4s, ROW2L.4h, XFIX_0_541196100 + smlal v4.4s, ROW2R.4h, XFIX_0_541196100_MINUS_1_847759065 /* ROW6L.4h <-> ROW2R.4h */ + mov v8.16b, v12.16b + smlsl v12.4s, ROW1R.4h, XFIX_2_562915447 /* ROW5L.4h <-> ROW1R.4h */ + smlal v12.4s, ROW3L.4h, XFIX_3_072711026_MINUS_2_562915447 + shl v6.4s, v6.4s, #13 + smlsl v8.4s, ROW1L.4h, XFIX_0_899976223 + add v2.4s, v6.4s, v4.4s + mov v10.16b, v14.16b + add v2.4s, v2.4s, v12.4s + smlsl v14.4s, ROW3R.4h, XFIX_0_899976223 /* ROW7L.4h <-> ROW3R.4h */ + smlal v14.4s, ROW1L.4h, XFIX_1_501321110_MINUS_0_899976223 + shrn ROW1L.4h, v2.4s, #16 + sub v2.4s, v2.4s, v12.4s + smlal v10.4s, ROW1R.4h, XFIX_2_053119869_MINUS_2_562915447 /* ROW5L.4h <-> ROW1R.4h */ + smlsl v10.4s, ROW3L.4h, XFIX_2_562915447 + sub v2.4s, v2.4s, v12.4s + smull v12.4s, ROW2L.4h, XFIX_0_541196100_PLUS_0_765366865 + smlal v12.4s, ROW2R.4h, XFIX_0_541196100 /* ROW6L.4h <-> ROW2R.4h */ + sub v6.4s, v6.4s, v4.4s + shrn ROW2R.4h, v2.4s, #16 /* ROW6L.4h <-> ROW2R.4h */ + add v2.4s, v6.4s, v10.4s + sub v6.4s, v6.4s, v10.4s + saddl v10.4s, ROW0L.4h, ROW0R.4h /* ROW4L.4h <-> ROW0R.4h */ + shrn ROW2L.4h, v2.4s, #16 + shrn ROW1R.4h, v6.4s, #16 /* ROW5L.4h <-> ROW1R.4h */ + shl v10.4s, v10.4s, #13 + smlal v8.4s, ROW3R.4h, XFIX_0_298631336_MINUS_0_899976223 /* ROW7L.4h <-> ROW3R.4h */ + add v4.4s, v10.4s, v12.4s + sub v2.4s, v10.4s, v12.4s + add v12.4s, v4.4s, v14.4s + sub v4.4s, v4.4s, v14.4s + add v10.4s, v2.4s, v8.4s + sub v6.4s, v2.4s, v8.4s + shrn ROW3R.4h, v4.4s, #16 /* ROW7L.4h <-> ROW3R.4h */ + shrn ROW3L.4h, v10.4s, #16 + shrn ROW0L.4h, v12.4s, #16 + shrn ROW0R.4h, v6.4s, #16 /* ROW4L.4h <-> ROW0R.4h */ + /* 1-D IDCT, pass 2, right 4x8 half */ + ld1 {v2.4h}, [x15] /* reload constants */ + smull v12.4s, ROW5R.4h, XFIX_1_175875602 + smlal v12.4s, ROW5L.4h, XFIX_1_175875602 /* ROW5L.4h <-> ROW1R.4h */ + smlal v12.4s, ROW7R.4h, XFIX_1_175875602_MINUS_1_961570560 + smlal v12.4s, ROW7L.4h, XFIX_1_175875602_MINUS_1_961570560 /* ROW7L.4h <-> ROW3R.4h */ + smull v14.4s, ROW7R.4h, XFIX_1_175875602 + smlal v14.4s, ROW7L.4h, XFIX_1_175875602 /* ROW7L.4h <-> ROW3R.4h */ + smlal v14.4s, ROW5R.4h, XFIX_1_175875602_MINUS_0_390180644 + smlal v14.4s, ROW5L.4h, XFIX_1_175875602_MINUS_0_390180644 /* ROW5L.4h <-> ROW1R.4h */ + ssubl v6.4s, ROW4L.4h, ROW4R.4h /* ROW4L.4h <-> ROW0R.4h */ + smull v4.4s, ROW6L.4h, XFIX_0_541196100 /* ROW6L.4h <-> ROW2R.4h */ + smlal v4.4s, ROW6R.4h, XFIX_0_541196100_MINUS_1_847759065 + mov v8.16b, v12.16b + smlsl v12.4s, ROW5R.4h, XFIX_2_562915447 + smlal v12.4s, ROW7L.4h, XFIX_3_072711026_MINUS_2_562915447 /* ROW7L.4h <-> ROW3R.4h */ + shl v6.4s, v6.4s, #13 + smlsl v8.4s, ROW5L.4h, XFIX_0_899976223 /* ROW5L.4h <-> ROW1R.4h */ + add v2.4s, v6.4s, v4.4s + mov v10.16b, v14.16b + add v2.4s, v2.4s, v12.4s + smlsl v14.4s, ROW7R.4h, XFIX_0_899976223 + smlal v14.4s, ROW5L.4h, XFIX_1_501321110_MINUS_0_899976223 /* ROW5L.4h <-> ROW1R.4h */ + shrn ROW5L.4h, v2.4s, #16 /* ROW5L.4h <-> ROW1R.4h */ + sub v2.4s, v2.4s, v12.4s + smlal v10.4s, ROW5R.4h, XFIX_2_053119869_MINUS_2_562915447 + smlsl v10.4s, ROW7L.4h, XFIX_2_562915447 /* ROW7L.4h <-> ROW3R.4h */ + sub v2.4s, v2.4s, v12.4s + smull v12.4s, ROW6L.4h, XFIX_0_541196100_PLUS_0_765366865 /* ROW6L.4h <-> ROW2R.4h */ + smlal v12.4s, ROW6R.4h, XFIX_0_541196100 + sub v6.4s, v6.4s, v4.4s + shrn ROW6R.4h, v2.4s, #16 + add v2.4s, v6.4s, v10.4s + sub v6.4s, v6.4s, v10.4s + saddl v10.4s, ROW4L.4h, ROW4R.4h /* ROW4L.4h <-> ROW0R.4h */ + shrn ROW6L.4h, v2.4s, #16 /* ROW6L.4h <-> ROW2R.4h */ + shrn ROW5R.4h, v6.4s, #16 + shl v10.4s, v10.4s, #13 + smlal v8.4s, ROW7R.4h, XFIX_0_298631336_MINUS_0_899976223 + add v4.4s, v10.4s, v12.4s + sub v2.4s, v10.4s, v12.4s + add v12.4s, v4.4s, v14.4s + sub v4.4s, v4.4s, v14.4s + add v10.4s, v2.4s, v8.4s + sub v6.4s, v2.4s, v8.4s + shrn ROW7R.4h, v4.4s, #16 + shrn ROW7L.4h, v10.4s, #16 /* ROW7L.4h <-> ROW3R.4h */ + shrn ROW4L.4h, v12.4s, #16 /* ROW4L.4h <-> ROW0R.4h */ + shrn ROW4R.4h, v6.4s, #16 + +2: /* Descale to 8-bit and range limit */ + ins v16.d[1], v17.d[0] + ins v18.d[1], v19.d[0] + ins v20.d[1], v21.d[0] + ins v22.d[1], v23.d[0] + sqrshrn v16.8b, v16.8h, #2 + sqrshrn2 v16.16b, v18.8h, #2 + sqrshrn v18.8b, v20.8h, #2 + sqrshrn2 v18.16b, v22.8h, #2 + + /* vpop {v8.4h - d15.4h} */ /* restore NEON registers */ + ld1 {v8.4h, v9.4h, v10.4h, v11.4h}, [sp], 32 + ld1 {v12.4h, v13.4h, v14.4h, v15.4h}, [sp], 32 + ins v24.d[1], v25.d[0] + + sqrshrn v20.8b, v24.8h, #2 + /* Transpose the final 8-bit samples and do signed->unsigned conversion */ + /* trn1 v16.8h, v16.8h, v18.8h */ + transpose v16, v18, v3, .16b, .8h + ins v26.d[1], v27.d[0] + ins v28.d[1], v29.d[0] + ins v30.d[1], v31.d[0] + sqrshrn2 v20.16b, v26.8h, #2 + sqrshrn v22.8b, v28.8h, #2 + movi v0.16b, #(CENTERJSAMPLE) + sqrshrn2 v22.16b, v30.8h, #2 + transpose_single v16, v17, v3, .d, .8b + transpose_single v18, v19, v3, .d, .8b + add v16.8b, v16.8b, v0.8b + add v17.8b, v17.8b, v0.8b + add v18.8b, v18.8b, v0.8b + add v19.8b, v19.8b, v0.8b + transpose v20, v22, v3, .16b, .8h + /* Store results to the output buffer */ + ldp TMP1, TMP2, [OUTPUT_BUF], 16 + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + st1 {v16.8b}, [TMP1] + transpose_single v20, v21, v3, .d, .8b + st1 {v17.8b}, [TMP2] + ldp TMP1, TMP2, [OUTPUT_BUF], 16 + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + st1 {v18.8b}, [TMP1] + add v20.8b, v20.8b, v0.8b + add v21.8b, v21.8b, v0.8b + st1 {v19.8b}, [TMP2] + ldp TMP1, TMP2, [OUTPUT_BUF], 16 + ldp TMP3, TMP4, [OUTPUT_BUF] + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + add TMP3, TMP3, OUTPUT_COL + add TMP4, TMP4, OUTPUT_COL + transpose_single v22, v23, v3, .d, .8b + st1 {v20.8b}, [TMP1] + add v22.8b, v22.8b, v0.8b + add v23.8b, v23.8b, v0.8b + st1 {v21.8b}, [TMP2] + st1 {v22.8b}, [TMP3] + st1 {v23.8b}, [TMP4] + ldr x15, [sp], 16 + ld1 {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32 + ld1 {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32 + ld1 {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32 + ld1 {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32 + ld1 {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32 + ld1 {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32 + ld1 {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32 + ld1 {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32 + blr x30 + +3: /* Left 4x8 half is done, right 4x8 half contains mostly zeros */ + + /* Transpose left 4x8 half */ + transpose ROW6L, ROW7L, v3, .16b, .4h + transpose ROW2L, ROW3L, v3, .16b, .4h + transpose ROW0L, ROW1L, v3, .16b, .4h + transpose ROW4L, ROW5L, v3, .16b, .4h + shl ROW0R.4h, ROW0R.4h, #2 /* PASS1_BITS */ + transpose ROW1L, ROW3L, v3, .16b, .2s + transpose ROW4L, ROW6L, v3, .16b, .2s + transpose ROW0L, ROW2L, v3, .16b, .2s + transpose ROW5L, ROW7L, v3, .16b, .2s + cmp x0, #0 + b.eq 4f /* Right 4x8 half has all zeros, go to 'sparse' second pass */ + + /* Only row 0 is non-zero for the right 4x8 half */ + dup ROW1R.4h, ROW0R.h[1] + dup ROW2R.4h, ROW0R.h[2] + dup ROW3R.4h, ROW0R.h[3] + dup ROW4R.4h, ROW0R.h[0] + dup ROW5R.4h, ROW0R.h[1] + dup ROW6R.4h, ROW0R.h[2] + dup ROW7R.4h, ROW0R.h[3] + dup ROW0R.4h, ROW0R.h[0] + b 1b /* Go to 'normal' second pass */ + +4: /* 1-D IDCT, pass 2 (sparse variant with zero rows 4-7), left 4x8 half */ + ld1 {v2.4h}, [x15] /* reload constants */ + smull v12.4s, ROW1L.4h, XFIX_1_175875602 + smlal v12.4s, ROW3L.4h, XFIX_1_175875602_MINUS_1_961570560 + smull v14.4s, ROW3L.4h, XFIX_1_175875602 + smlal v14.4s, ROW1L.4h, XFIX_1_175875602_MINUS_0_390180644 + smull v4.4s, ROW2L.4h, XFIX_0_541196100 + sshll v6.4s, ROW0L.4h, #13 + mov v8.16b, v12.16b + smlal v12.4s, ROW3L.4h, XFIX_3_072711026_MINUS_2_562915447 + smlsl v8.4s, ROW1L.4h, XFIX_0_899976223 + add v2.4s, v6.4s, v4.4s + mov v10.16b, v14.16b + smlal v14.4s, ROW1L.4h, XFIX_1_501321110_MINUS_0_899976223 + add v2.4s, v2.4s, v12.4s + add v12.4s, v12.4s, v12.4s + smlsl v10.4s, ROW3L.4h, XFIX_2_562915447 + shrn ROW1L.4h, v2.4s, #16 + sub v2.4s, v2.4s, v12.4s + smull v12.4s, ROW2L.4h, XFIX_0_541196100_PLUS_0_765366865 + sub v6.4s, v6.4s, v4.4s + shrn ROW2R.4h, v2.4s, #16 /* ROW6L.4h <-> ROW2R.4h */ + add v2.4s, v6.4s, v10.4s + sub v6.4s, v6.4s, v10.4s + sshll v10.4s, ROW0L.4h, #13 + shrn ROW2L.4h, v2.4s, #16 + shrn ROW1R.4h, v6.4s, #16 /* ROW5L.4h <-> ROW1R.4h */ + add v4.4s, v10.4s, v12.4s + sub v2.4s, v10.4s, v12.4s + add v12.4s, v4.4s, v14.4s + sub v4.4s, v4.4s, v14.4s + add v10.4s, v2.4s, v8.4s + sub v6.4s, v2.4s, v8.4s + shrn ROW3R.4h, v4.4s, #16 /* ROW7L.4h <-> ROW3R.4h */ + shrn ROW3L.4h, v10.4s, #16 + shrn ROW0L.4h, v12.4s, #16 + shrn ROW0R.4h, v6.4s, #16 /* ROW4L.4h <-> ROW0R.4h */ + /* 1-D IDCT, pass 2 (sparse variant with zero rows 4-7), right 4x8 half */ + ld1 {v2.4h}, [x15] /* reload constants */ + smull v12.4s, ROW5L.4h, XFIX_1_175875602 + smlal v12.4s, ROW7L.4h, XFIX_1_175875602_MINUS_1_961570560 + smull v14.4s, ROW7L.4h, XFIX_1_175875602 + smlal v14.4s, ROW5L.4h, XFIX_1_175875602_MINUS_0_390180644 + smull v4.4s, ROW6L.4h, XFIX_0_541196100 + sshll v6.4s, ROW4L.4h, #13 + mov v8.16b, v12.16b + smlal v12.4s, ROW7L.4h, XFIX_3_072711026_MINUS_2_562915447 + smlsl v8.4s, ROW5L.4h, XFIX_0_899976223 + add v2.4s, v6.4s, v4.4s + mov v10.16b, v14.16b + smlal v14.4s, ROW5L.4h, XFIX_1_501321110_MINUS_0_899976223 + add v2.4s, v2.4s, v12.4s + add v12.4s, v12.4s, v12.4s + smlsl v10.4s, ROW7L.4h, XFIX_2_562915447 + shrn ROW5L.4h, v2.4s, #16 /* ROW5L.4h <-> ROW1R.4h */ + sub v2.4s, v2.4s, v12.4s + smull v12.4s, ROW6L.4h, XFIX_0_541196100_PLUS_0_765366865 + sub v6.4s, v6.4s, v4.4s + shrn ROW6R.4h, v2.4s, #16 + add v2.4s, v6.4s, v10.4s + sub v6.4s, v6.4s, v10.4s + sshll v10.4s, ROW4L.4h, #13 + shrn ROW6L.4h, v2.4s, #16 /* ROW6L.4h <-> ROW2R.4h */ + shrn ROW5R.4h, v6.4s, #16 + add v4.4s, v10.4s, v12.4s + sub v2.4s, v10.4s, v12.4s + add v12.4s, v4.4s, v14.4s + sub v4.4s, v4.4s, v14.4s + add v10.4s, v2.4s, v8.4s + sub v6.4s, v2.4s, v8.4s + shrn ROW7R.4h, v4.4s, #16 + shrn ROW7L.4h, v10.4s, #16 /* ROW7L.4h <-> ROW3R.4h */ + shrn ROW4L.4h, v12.4s, #16 /* ROW4L.4h <-> ROW0R.4h */ + shrn ROW4R.4h, v6.4s, #16 + b 2b /* Go to epilogue */ + + .unreq DCT_TABLE + .unreq COEF_BLOCK + .unreq OUTPUT_BUF + .unreq OUTPUT_COL + .unreq TMP1 + .unreq TMP2 + .unreq TMP3 + .unreq TMP4 + + .unreq ROW0L + .unreq ROW0R + .unreq ROW1L + .unreq ROW1R + .unreq ROW2L + .unreq ROW2R + .unreq ROW3L + .unreq ROW3R + .unreq ROW4L + .unreq ROW4R + .unreq ROW5L + .unreq ROW5R + .unreq ROW6L + .unreq ROW6R + .unreq ROW7L + .unreq ROW7R + + +/*****************************************************************************/ + +/* + * jsimd_idct_ifast_neon + * + * This function contains a fast, not so accurate integer implementation of + * the inverse DCT (Discrete Cosine Transform). It uses the same calculations + * and produces exactly the same output as IJG's original 'jpeg_idct_ifast' + * function from jidctfst.c + * + * Normally 1-D AAN DCT needs 5 multiplications and 29 additions. + * But in ARM NEON case some extra additions are required because VQDMULH + * instruction can't handle the constants larger than 1. So the expressions + * like "x * 1.082392200" have to be converted to "x * 0.082392200 + x", + * which introduces an extra addition. Overall, there are 6 extra additions + * per 1-D IDCT pass, totalling to 5 VQDMULH and 35 VADD/VSUB instructions. + */ + +#define XFIX_1_082392200 v0.h[0] +#define XFIX_1_414213562 v0.h[1] +#define XFIX_1_847759065 v0.h[2] +#define XFIX_2_613125930 v0.h[3] + +.balign 16 +Ljsimd_idct_ifast_neon_consts: + .short (277 * 128 - 256 * 128) /* XFIX_1_082392200 */ + .short (362 * 128 - 256 * 128) /* XFIX_1_414213562 */ + .short (473 * 128 - 256 * 128) /* XFIX_1_847759065 */ + .short (669 * 128 - 512 * 128) /* XFIX_2_613125930 */ + +asm_function jsimd_idct_ifast_neon + + DCT_TABLE .req x0 + COEF_BLOCK .req x1 + OUTPUT_BUF .req x2 + OUTPUT_COL .req x3 + TMP1 .req x0 + TMP2 .req x1 + TMP3 .req x2 + TMP4 .req x22 + TMP5 .req x23 + + /* Load and dequantize coefficients into NEON registers + * with the following allocation: + * 0 1 2 3 | 4 5 6 7 + * ---------+-------- + * 0 | d16 | d17 ( v8.8h ) + * 1 | d18 | d19 ( v9.8h ) + * 2 | d20 | d21 ( v10.8h ) + * 3 | d22 | d23 ( v11.8h ) + * 4 | d24 | d25 ( v12.8h ) + * 5 | d26 | d27 ( v13.8h ) + * 6 | d28 | d29 ( v14.8h ) + * 7 | d30 | d31 ( v15.8h ) + */ + /* Save NEON registers used in fast IDCT */ + sub sp, sp, #176 + stp x22, x23, [sp], 16 + adr x23, Ljsimd_idct_ifast_neon_consts + st1 {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32 + st1 {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32 + st1 {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32 + st1 {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32 + st1 {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32 + ld1 {v8.8h, v9.8h}, [COEF_BLOCK], 32 + ld1 {v0.8h, v1.8h}, [DCT_TABLE], 32 + ld1 {v10.8h, v11.8h}, [COEF_BLOCK], 32 + mul v8.8h, v8.8h, v0.8h + ld1 {v2.8h, v3.8h}, [DCT_TABLE], 32 + mul v9.8h, v9.8h, v1.8h + ld1 {v12.8h, v13.8h}, [COEF_BLOCK], 32 + mul v10.8h, v10.8h, v2.8h + ld1 {v0.8h, v1.8h}, [DCT_TABLE], 32 + mul v11.8h, v11.8h, v3.8h + ld1 {v14.8h, v15.8h}, [COEF_BLOCK], 32 + mul v12.8h, v12.8h, v0.8h + ld1 {v2.8h, v3.8h}, [DCT_TABLE], 32 + mul v14.8h, v14.8h, v2.8h + mul v13.8h, v13.8h, v1.8h + ld1 {v0.4h}, [x23] /* load constants */ + mul v15.8h, v15.8h, v3.8h + + /* 1-D IDCT, pass 1 */ + sub v2.8h, v10.8h, v14.8h + add v14.8h, v10.8h, v14.8h + sub v1.8h, v11.8h, v13.8h + add v13.8h, v11.8h, v13.8h + sub v5.8h, v9.8h, v15.8h + add v15.8h, v9.8h, v15.8h + sqdmulh v4.8h, v2.8h, XFIX_1_414213562 + sqdmulh v6.8h, v1.8h, XFIX_2_613125930 + add v3.8h, v1.8h, v1.8h + sub v1.8h, v5.8h, v1.8h + add v10.8h, v2.8h, v4.8h + sqdmulh v4.8h, v1.8h, XFIX_1_847759065 + sub v2.8h, v15.8h, v13.8h + add v3.8h, v3.8h, v6.8h + sqdmulh v6.8h, v2.8h, XFIX_1_414213562 + add v1.8h, v1.8h, v4.8h + sqdmulh v4.8h, v5.8h, XFIX_1_082392200 + sub v10.8h, v10.8h, v14.8h + add v2.8h, v2.8h, v6.8h + sub v6.8h, v8.8h, v12.8h + add v12.8h, v8.8h, v12.8h + add v9.8h, v5.8h, v4.8h + add v5.8h, v6.8h, v10.8h + sub v10.8h, v6.8h, v10.8h + add v6.8h, v15.8h, v13.8h + add v8.8h, v12.8h, v14.8h + sub v3.8h, v6.8h, v3.8h + sub v12.8h, v12.8h, v14.8h + sub v3.8h, v3.8h, v1.8h + sub v1.8h, v9.8h, v1.8h + add v2.8h, v3.8h, v2.8h + sub v15.8h, v8.8h, v6.8h + add v1.8h, v1.8h, v2.8h + add v8.8h, v8.8h, v6.8h + add v14.8h, v5.8h, v3.8h + sub v9.8h, v5.8h, v3.8h + sub v13.8h, v10.8h, v2.8h + add v10.8h, v10.8h, v2.8h + /* Transpose q8-q9 */ + mov v18.16b, v8.16b + trn1 v8.8h, v8.8h, v9.8h + trn2 v9.8h, v18.8h, v9.8h + sub v11.8h, v12.8h, v1.8h + /* Transpose q14-q15 */ + mov v18.16b, v14.16b + trn1 v14.8h, v14.8h, v15.8h + trn2 v15.8h, v18.8h, v15.8h + add v12.8h, v12.8h, v1.8h + /* Transpose q10-q11 */ + mov v18.16b, v10.16b + trn1 v10.8h, v10.8h, v11.8h + trn2 v11.8h, v18.8h, v11.8h + /* Transpose q12-q13 */ + mov v18.16b, v12.16b + trn1 v12.8h, v12.8h, v13.8h + trn2 v13.8h, v18.8h, v13.8h + /* Transpose q9-q11 */ + mov v18.16b, v9.16b + trn1 v9.4s, v9.4s, v11.4s + trn2 v11.4s, v18.4s, v11.4s + /* Transpose q12-q14 */ + mov v18.16b, v12.16b + trn1 v12.4s, v12.4s, v14.4s + trn2 v14.4s, v18.4s, v14.4s + /* Transpose q8-q10 */ + mov v18.16b, v8.16b + trn1 v8.4s, v8.4s, v10.4s + trn2 v10.4s, v18.4s, v10.4s + /* Transpose q13-q15 */ + mov v18.16b, v13.16b + trn1 v13.4s, v13.4s, v15.4s + trn2 v15.4s, v18.4s, v15.4s + /* vswp v14.4h, v10-MSB.4h */ + umov x22, v14.d[0] + ins v14.d[0], v10.d[1] + ins v10.d[1], x22 + /* vswp v13.4h, v9MSB.4h */ + + umov x22, v13.d[0] + ins v13.d[0], v9.d[1] + ins v9.d[1], x22 + /* 1-D IDCT, pass 2 */ + sub v2.8h, v10.8h, v14.8h + /* vswp v15.4h, v11MSB.4h */ + umov x22, v15.d[0] + ins v15.d[0], v11.d[1] + ins v11.d[1], x22 + add v14.8h, v10.8h, v14.8h + /* vswp v12.4h, v8-MSB.4h */ + umov x22, v12.d[0] + ins v12.d[0], v8.d[1] + ins v8.d[1], x22 + sub v1.8h, v11.8h, v13.8h + add v13.8h, v11.8h, v13.8h + sub v5.8h, v9.8h, v15.8h + add v15.8h, v9.8h, v15.8h + sqdmulh v4.8h, v2.8h, XFIX_1_414213562 + sqdmulh v6.8h, v1.8h, XFIX_2_613125930 + add v3.8h, v1.8h, v1.8h + sub v1.8h, v5.8h, v1.8h + add v10.8h, v2.8h, v4.8h + sqdmulh v4.8h, v1.8h, XFIX_1_847759065 + sub v2.8h, v15.8h, v13.8h + add v3.8h, v3.8h, v6.8h + sqdmulh v6.8h, v2.8h, XFIX_1_414213562 + add v1.8h, v1.8h, v4.8h + sqdmulh v4.8h, v5.8h, XFIX_1_082392200 + sub v10.8h, v10.8h, v14.8h + add v2.8h, v2.8h, v6.8h + sub v6.8h, v8.8h, v12.8h + add v12.8h, v8.8h, v12.8h + add v9.8h, v5.8h, v4.8h + add v5.8h, v6.8h, v10.8h + sub v10.8h, v6.8h, v10.8h + add v6.8h, v15.8h, v13.8h + add v8.8h, v12.8h, v14.8h + sub v3.8h, v6.8h, v3.8h + sub v12.8h, v12.8h, v14.8h + sub v3.8h, v3.8h, v1.8h + sub v1.8h, v9.8h, v1.8h + add v2.8h, v3.8h, v2.8h + sub v15.8h, v8.8h, v6.8h + add v1.8h, v1.8h, v2.8h + add v8.8h, v8.8h, v6.8h + add v14.8h, v5.8h, v3.8h + sub v9.8h, v5.8h, v3.8h + sub v13.8h, v10.8h, v2.8h + add v10.8h, v10.8h, v2.8h + sub v11.8h, v12.8h, v1.8h + add v12.8h, v12.8h, v1.8h + /* Descale to 8-bit and range limit */ + movi v0.16b, #0x80 + sqshrn v8.8b, v8.8h, #5 + sqshrn2 v8.16b, v9.8h, #5 + sqshrn v9.8b, v10.8h, #5 + sqshrn2 v9.16b, v11.8h, #5 + sqshrn v10.8b, v12.8h, #5 + sqshrn2 v10.16b, v13.8h, #5 + sqshrn v11.8b, v14.8h, #5 + sqshrn2 v11.16b, v15.8h, #5 + add v8.16b, v8.16b, v0.16b + add v9.16b, v9.16b, v0.16b + add v10.16b, v10.16b, v0.16b + add v11.16b, v11.16b, v0.16b + /* Transpose the final 8-bit samples */ + /* Transpose q8-q9 */ + mov v18.16b, v8.16b + trn1 v8.8h, v8.8h, v9.8h + trn2 v9.8h, v18.8h, v9.8h + /* Transpose q10-q11 */ + mov v18.16b, v10.16b + trn1 v10.8h, v10.8h, v11.8h + trn2 v11.8h, v18.8h, v11.8h + /* Transpose q8-q10 */ + mov v18.16b, v8.16b + trn1 v8.4s, v8.4s, v10.4s + trn2 v10.4s, v18.4s, v10.4s + /* Transpose q9-q11 */ + mov v18.16b, v9.16b + trn1 v9.4s, v9.4s, v11.4s + trn2 v11.4s, v18.4s, v11.4s + /* make copy */ + ins v17.d[0], v8.d[1] + /* Transpose d16-d17-msb */ + mov v18.16b, v8.16b + trn1 v8.8b, v8.8b, v17.8b + trn2 v17.8b, v18.8b, v17.8b + /* make copy */ + ins v19.d[0], v9.d[1] + mov v18.16b, v9.16b + trn1 v9.8b, v9.8b, v19.8b + trn2 v19.8b, v18.8b, v19.8b + /* Store results to the output buffer */ + ldp TMP1, TMP2, [OUTPUT_BUF], 16 + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + st1 {v8.8b}, [TMP1] + st1 {v17.8b}, [TMP2] + ldp TMP1, TMP2, [OUTPUT_BUF], 16 + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + st1 {v9.8b}, [TMP1] + /* make copy */ + ins v7.d[0], v10.d[1] + mov v18.16b, v10.16b + trn1 v10.8b, v10.8b, v7.8b + trn2 v7.8b, v18.8b, v7.8b + st1 {v19.8b}, [TMP2] + ldp TMP1, TMP2, [OUTPUT_BUF], 16 + ldp TMP4, TMP5, [OUTPUT_BUF], 16 + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + add TMP4, TMP4, OUTPUT_COL + add TMP5, TMP5, OUTPUT_COL + st1 {v10.8b}, [TMP1] + /* make copy */ + ins v16.d[0], v11.d[1] + mov v18.16b, v11.16b + trn1 v11.8b, v11.8b, v16.8b + trn2 v16.8b, v18.8b, v16.8b + st1 {v7.8b}, [TMP2] + st1 {v11.8b}, [TMP4] + st1 {v16.8b}, [TMP5] + sub sp, sp, #176 + ldp x22, x23, [sp], 16 + ld1 {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32 + ld1 {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32 + ld1 {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32 + ld1 {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32 + ld1 {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32 + blr x30 + + .unreq DCT_TABLE + .unreq COEF_BLOCK + .unreq OUTPUT_BUF + .unreq OUTPUT_COL + .unreq TMP1 + .unreq TMP2 + .unreq TMP3 + .unreq TMP4 + + +/*****************************************************************************/ + +/* + * jsimd_idct_4x4_neon + * + * This function contains inverse-DCT code for getting reduced-size + * 4x4 pixels output from an 8x8 DCT block. It uses the same calculations + * and produces exactly the same output as IJG's original 'jpeg_idct_4x4' + * function from jpeg-6b (jidctred.c). + * + * NOTE: jpeg-8 has an improved implementation of 4x4 inverse-DCT, which + * requires much less arithmetic operations and hence should be faster. + * The primary purpose of this particular NEON optimized function is + * bit exact compatibility with jpeg-6b. + * + * TODO: a bit better instructions scheduling can be achieved by expanding + * idct_helper/transpose_4x4 macros and reordering instructions, + * but readability will suffer somewhat. + */ + +#define CONST_BITS 13 + +#define FIX_0_211164243 (1730) /* FIX(0.211164243) */ +#define FIX_0_509795579 (4176) /* FIX(0.509795579) */ +#define FIX_0_601344887 (4926) /* FIX(0.601344887) */ +#define FIX_0_720959822 (5906) /* FIX(0.720959822) */ +#define FIX_0_765366865 (6270) /* FIX(0.765366865) */ +#define FIX_0_850430095 (6967) /* FIX(0.850430095) */ +#define FIX_0_899976223 (7373) /* FIX(0.899976223) */ +#define FIX_1_061594337 (8697) /* FIX(1.061594337) */ +#define FIX_1_272758580 (10426) /* FIX(1.272758580) */ +#define FIX_1_451774981 (11893) /* FIX(1.451774981) */ +#define FIX_1_847759065 (15137) /* FIX(1.847759065) */ +#define FIX_2_172734803 (17799) /* FIX(2.172734803) */ +#define FIX_2_562915447 (20995) /* FIX(2.562915447) */ +#define FIX_3_624509785 (29692) /* FIX(3.624509785) */ + +.balign 16 +Ljsimd_idct_4x4_neon_consts: + .short FIX_1_847759065 /* v0.h[0] */ + .short -FIX_0_765366865 /* v0.h[1] */ + .short -FIX_0_211164243 /* v0.h[2] */ + .short FIX_1_451774981 /* v0.h[3] */ + .short -FIX_2_172734803 /* d1[0] */ + .short FIX_1_061594337 /* d1[1] */ + .short -FIX_0_509795579 /* d1[2] */ + .short -FIX_0_601344887 /* d1[3] */ + .short FIX_0_899976223 /* v2.h[0] */ + .short FIX_2_562915447 /* v2.h[1] */ + .short 1 << (CONST_BITS+1) /* v2.h[2] */ + .short 0 /* v2.h[3] */ + +.macro idct_helper x4, x6, x8, x10, x12, x14, x16, shift, y26, y27, y28, y29 + smull v28.4s, \x4, v2.h[2] + smlal v28.4s, \x8, v0.h[0] + smlal v28.4s, \x14, v0.h[1] + + smull v26.4s, \x16, v1.h[2] + smlal v26.4s, \x12, v1.h[3] + smlal v26.4s, \x10, v2.h[0] + smlal v26.4s, \x6, v2.h[1] + + smull v30.4s, \x4, v2.h[2] + smlsl v30.4s, \x8, v0.h[0] + smlsl v30.4s, \x14, v0.h[1] + + smull v24.4s, \x16, v0.h[2] + smlal v24.4s, \x12, v0.h[3] + smlal v24.4s, \x10, v1.h[0] + smlal v24.4s, \x6, v1.h[1] + + add v20.4s, v28.4s, v26.4s + sub v28.4s, v28.4s, v26.4s + +.if \shift > 16 + srshr v20.4s, v20.4s, #\shift + srshr v28.4s, v28.4s, #\shift + xtn \y26, v20.4s + xtn \y29, v28.4s +.else + rshrn \y26, v20.4s, #\shift + rshrn \y29, v28.4s, #\shift +.endif + + add v20.4s, v30.4s, v24.4s + sub v30.4s, v30.4s, v24.4s + +.if \shift > 16 + srshr v20.4s, v20.4s, #\shift + srshr v30.4s, v30.4s, #\shift + xtn \y27, v20.4s + xtn \y28, v30.4s +.else + rshrn \y27, v20.4s, #\shift + rshrn \y28, v30.4s, #\shift +.endif + +.endm + +asm_function jsimd_idct_4x4_neon + + DCT_TABLE .req x0 + COEF_BLOCK .req x1 + OUTPUT_BUF .req x2 + OUTPUT_COL .req x3 + TMP1 .req x0 + TMP2 .req x1 + TMP3 .req x2 + TMP4 .req x15 + + /* Save all used NEON registers */ + sub sp, sp, 272 + str x15, [sp], 16 + /* Load constants (v3.4h is just used for padding) */ + adr TMP4, Ljsimd_idct_4x4_neon_consts + st1 {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32 + st1 {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32 + st1 {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32 + st1 {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32 + st1 {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32 + st1 {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32 + st1 {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32 + st1 {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32 + ld1 {v0.4h, v1.4h, v2.4h, v3.4h}, [TMP4] + + /* Load all COEF_BLOCK into NEON registers with the following allocation: + * 0 1 2 3 | 4 5 6 7 + * ---------+-------- + * 0 | v4.4h | v5.4h + * 1 | v6.4h | v7.4h + * 2 | v8.4h | v9.4h + * 3 | v10.4h | v11.4h + * 4 | - | - + * 5 | v12.4h | v13.4h + * 6 | v14.4h | v15.4h + * 7 | v16.4h | v17.4h + */ + ld1 {v4.4h, v5.4h, v6.4h, v7.4h}, [COEF_BLOCK], 32 + ld1 {v8.4h, v9.4h, v10.4h, v11.4h}, [COEF_BLOCK], 32 + add COEF_BLOCK, COEF_BLOCK, #16 + ld1 {v12.4h, v13.4h, v14.4h, v15.4h}, [COEF_BLOCK], 32 + ld1 {v16.4h, v17.4h}, [COEF_BLOCK], 16 + /* dequantize */ + ld1 {v18.4h, v19.4h, v20.4h, v21.4h}, [DCT_TABLE], 32 + mul v4.4h, v4.4h, v18.4h + mul v5.4h, v5.4h, v19.4h + ins v4.d[1], v5.d[0] /* 128 bit q4 */ + ld1 {v22.4h, v23.4h, v24.4h, v25.4h}, [DCT_TABLE], 32 + mul v6.4h, v6.4h, v20.4h + mul v7.4h, v7.4h, v21.4h + ins v6.d[1], v7.d[0] /* 128 bit q6 */ + mul v8.4h, v8.4h, v22.4h + mul v9.4h, v9.4h, v23.4h + ins v8.d[1], v9.d[0] /* 128 bit q8 */ + add DCT_TABLE, DCT_TABLE, #16 + ld1 {v26.4h, v27.4h, v28.4h, v29.4h}, [DCT_TABLE], 32 + mul v10.4h, v10.4h, v24.4h + mul v11.4h, v11.4h, v25.4h + ins v10.d[1], v11.d[0] /* 128 bit q10 */ + mul v12.4h, v12.4h, v26.4h + mul v13.4h, v13.4h, v27.4h + ins v12.d[1], v13.d[0] /* 128 bit q12 */ + ld1 {v30.4h, v31.4h}, [DCT_TABLE], 16 + mul v14.4h, v14.4h, v28.4h + mul v15.4h, v15.4h, v29.4h + ins v14.d[1], v15.d[0] /* 128 bit q14 */ + mul v16.4h, v16.4h, v30.4h + mul v17.4h, v17.4h, v31.4h + ins v16.d[1], v17.d[0] /* 128 bit q16 */ + + /* Pass 1 */ + idct_helper v4.4h, v6.4h, v8.4h, v10.4h, v12.4h, v14.4h, v16.4h, 12, v4.4h, v6.4h, v8.4h, v10.4h + transpose_4x4 v4, v6, v8, v10, v3 + ins v10.d[1], v11.d[0] + idct_helper v5.4h, v7.4h, v9.4h, v11.4h, v13.4h, v15.4h, v17.4h, 12, v5.4h, v7.4h, v9.4h, v11.4h + transpose_4x4 v5, v7, v9, v11, v3 + ins v10.d[1], v11.d[0] + /* Pass 2 */ + idct_helper v4.4h, v6.4h, v8.4h, v10.4h, v7.4h, v9.4h, v11.4h, 19, v26.4h, v27.4h, v28.4h, v29.4h + transpose_4x4 v26, v27, v28, v29, v3 + + /* Range limit */ + movi v30.8h, #0x80 + ins v26.d[1], v27.d[0] + ins v28.d[1], v29.d[0] + add v26.8h, v26.8h, v30.8h + add v28.8h, v28.8h, v30.8h + sqxtun v26.8b, v26.8h + sqxtun v27.8b, v28.8h + + /* Store results to the output buffer */ + ldp TMP1, TMP2, [OUTPUT_BUF], 16 + ldp TMP3, TMP4, [OUTPUT_BUF] + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + add TMP3, TMP3, OUTPUT_COL + add TMP4, TMP4, OUTPUT_COL + +#if defined(__ARMEL__) && !RESPECT_STRICT_ALIGNMENT + /* We can use much less instructions on little endian systems if the + * OS kernel is not configured to trap unaligned memory accesses + */ + st1 {v26.s}[0], [TMP1], 4 + st1 {v27.s}[0], [TMP3], 4 + st1 {v26.s}[1], [TMP2], 4 + st1 {v27.s}[1], [TMP4], 4 +#else + st1 {v26.b}[0], [TMP1], 1 + st1 {v27.b}[0], [TMP3], 1 + st1 {v26.b}[1], [TMP1], 1 + st1 {v27.b}[1], [TMP3], 1 + st1 {v26.b}[2], [TMP1], 1 + st1 {v27.b}[2], [TMP3], 1 + st1 {v26.b}[3], [TMP1], 1 + st1 {v27.b}[3], [TMP3], 1 + + st1 {v26.b}[4], [TMP2], 1 + st1 {v27.b}[4], [TMP4], 1 + st1 {v26.b}[5], [TMP2], 1 + st1 {v27.b}[5], [TMP4], 1 + st1 {v26.b}[6], [TMP2], 1 + st1 {v27.b}[6], [TMP4], 1 + st1 {v26.b}[7], [TMP2], 1 + st1 {v27.b}[7], [TMP4], 1 +#endif + + /* vpop {v8.4h - v15.4h} ;not available */ + sub sp, sp, #272 + ldr x15, [sp], 16 + ld1 {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32 + ld1 {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32 + ld1 {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32 + ld1 {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32 + ld1 {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32 + ld1 {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32 + ld1 {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32 + ld1 {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32 + blr x30 + + .unreq DCT_TABLE + .unreq COEF_BLOCK + .unreq OUTPUT_BUF + .unreq OUTPUT_COL + .unreq TMP1 + .unreq TMP2 + .unreq TMP3 + .unreq TMP4 + +.purgem idct_helper + + +/*****************************************************************************/ + +/* + * jsimd_idct_2x2_neon + * + * This function contains inverse-DCT code for getting reduced-size + * 2x2 pixels output from an 8x8 DCT block. It uses the same calculations + * and produces exactly the same output as IJG's original 'jpeg_idct_2x2' + * function from jpeg-6b (jidctred.c). + * + * NOTE: jpeg-8 has an improved implementation of 2x2 inverse-DCT, which + * requires much less arithmetic operations and hence should be faster. + * The primary purpose of this particular NEON optimized function is + * bit exact compatibility with jpeg-6b. + */ + +.balign 8 +Ljsimd_idct_2x2_neon_consts: + .short -FIX_0_720959822 /* v14[0] */ + .short FIX_0_850430095 /* v14[1] */ + .short -FIX_1_272758580 /* v14[2] */ + .short FIX_3_624509785 /* v14[3] */ + +.macro idct_helper x4, x6, x10, x12, x16, shift, y26, y27 + sshll v15.4s, \x4, #15 + smull v26.4s, \x6, v14.h[3] + smlal v26.4s, \x10, v14.h[2] + smlal v26.4s, \x12, v14.h[1] + smlal v26.4s, \x16, v14.h[0] + + add v20.4s, v15.4s, v26.4s + sub v15.4s, v15.4s, v26.4s + +.if \shift > 16 + srshr v20.4s, v20.4s, #\shift + srshr v15.4s, v15.4s, #\shift + xtn \y26, v20.4s + xtn \y27, v15.4s +.else + rshrn \y26, v20.4s, #\shift + rshrn \y27, v15.4s, #\shift +.endif + +.endm + +asm_function jsimd_idct_2x2_neon + + DCT_TABLE .req x0 + COEF_BLOCK .req x1 + OUTPUT_BUF .req x2 + OUTPUT_COL .req x3 + TMP1 .req x0 + TMP2 .req x15 + + /* vpush {v8.4h - v15.4h} ; not available */ + sub sp, sp, 208 + str x15, [sp], 16 + + /* Load constants */ + adr TMP2, Ljsimd_idct_2x2_neon_consts + st1 {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32 + st1 {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32 + st1 {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32 + st1 {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32 + st1 {v21.8b, v22.8b}, [sp], 16 + st1 {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32 + st1 {v30.8b, v31.8b}, [sp], 16 + ld1 {v14.4h}, [TMP2] + + /* Load all COEF_BLOCK into NEON registers with the following allocation: + * 0 1 2 3 | 4 5 6 7 + * ---------+-------- + * 0 | v4.4h | v5.4h + * 1 | v6.4h | v7.4h + * 2 | - | - + * 3 | v10.4h | v11.4h + * 4 | - | - + * 5 | v12.4h | v13.4h + * 6 | - | - + * 7 | v16.4h | v17.4h + */ + ld1 {v4.4h, v5.4h, v6.4h, v7.4h}, [COEF_BLOCK], 32 + add COEF_BLOCK, COEF_BLOCK, #16 + ld1 {v10.4h, v11.4h}, [COEF_BLOCK], 16 + add COEF_BLOCK, COEF_BLOCK, #16 + ld1 {v12.4h, v13.4h}, [COEF_BLOCK], 16 + add COEF_BLOCK, COEF_BLOCK, #16 + ld1 {v16.4h, v17.4h}, [COEF_BLOCK], 16 + /* Dequantize */ + ld1 {v18.4h, v19.4h, v20.4h, v21.4h}, [DCT_TABLE], 32 + mul v4.4h, v4.4h, v18.4h + mul v5.4h, v5.4h, v19.4h + ins v4.d[1], v5.d[0] + mul v6.4h, v6.4h, v20.4h + mul v7.4h, v7.4h, v21.4h + ins v6.d[1], v7.d[0] + add DCT_TABLE, DCT_TABLE, #16 + ld1 {v24.4h, v25.4h}, [DCT_TABLE], 16 + mul v10.4h, v10.4h, v24.4h + mul v11.4h, v11.4h, v25.4h + ins v10.d[1], v11.d[0] + add DCT_TABLE, DCT_TABLE, #16 + ld1 {v26.4h, v27.4h}, [DCT_TABLE], 16 + mul v12.4h, v12.4h, v26.4h + mul v13.4h, v13.4h, v27.4h + ins v12.d[1], v13.d[0] + add DCT_TABLE, DCT_TABLE, #16 + ld1 {v30.4h, v31.4h}, [DCT_TABLE], 16 + mul v16.4h, v16.4h, v30.4h + mul v17.4h, v17.4h, v31.4h + ins v16.d[1], v17.d[0] + + /* Pass 1 */ +#if 0 + idct_helper v4.4h, v6.4h, v10.4h, v12.4h, v16.4h, 13, v4.4h, v6.4h + transpose_4x4 v4.4h, v6.4h, v8.4h, v10.4h + idct_helper v5.4h, v7.4h, v11.4h, v13.4h, v17.4h, 13, v5.4h, v7.4h + transpose_4x4 v5.4h, v7.4h, v9.4h, v11.4h +#else + smull v26.4s, v6.4h, v14.h[3] + smlal v26.4s, v10.4h, v14.h[2] + smlal v26.4s, v12.4h, v14.h[1] + smlal v26.4s, v16.4h, v14.h[0] + smull v24.4s, v7.4h, v14.h[3] + smlal v24.4s, v11.4h, v14.h[2] + smlal v24.4s, v13.4h, v14.h[1] + smlal v24.4s, v17.4h, v14.h[0] + sshll v15.4s, v4.4h, #15 + sshll v30.4s, v5.4h, #15 + add v20.4s, v15.4s, v26.4s + sub v15.4s, v15.4s, v26.4s + rshrn v4.4h, v20.4s, #13 + rshrn v6.4h, v15.4s, #13 + add v20.4s, v30.4s, v24.4s + sub v15.4s, v30.4s, v24.4s + rshrn v5.4h, v20.4s, #13 + rshrn v7.4h, v15.4s, #13 + ins v4.d[1], v5.d[0] + ins v6.d[1], v7.d[0] + transpose v4, v6, v3, .16b, .8h + transpose v6, v10, v3, .16b, .4s + ins v11.d[0], v10.d[1] + ins v7.d[0], v6.d[1] +#endif + + /* Pass 2 */ + idct_helper v4.4h, v6.4h, v10.4h, v7.4h, v11.4h, 20, v26.4h, v27.4h + + /* Range limit */ + movi v30.8h, #0x80 + ins v26.d[1], v27.d[0] + add v26.8h, v26.8h, v30.8h + sqxtun v30.8b, v26.8h + ins v26.d[0], v30.d[0] + sqxtun v27.8b, v26.8h + + /* Store results to the output buffer */ + ldp TMP1, TMP2, [OUTPUT_BUF] + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + + st1 {v26.b}[0], [TMP1], 1 + st1 {v27.b}[4], [TMP1], 1 + st1 {v26.b}[1], [TMP2], 1 + st1 {v27.b}[5], [TMP2], 1 + + sub sp, sp, #208 + ldr x15, [sp], 16 + ld1 {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32 + ld1 {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32 + ld1 {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32 + ld1 {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32 + ld1 {v21.8b, v22.8b}, [sp], 16 + ld1 {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32 + ld1 {v30.8b, v31.8b}, [sp], 16 + blr x30 + + .unreq DCT_TABLE + .unreq COEF_BLOCK + .unreq OUTPUT_BUF + .unreq OUTPUT_COL + .unreq TMP1 + .unreq TMP2 + +.purgem idct_helper + + +/*****************************************************************************/ + +/* + * jsimd_ycc_extrgb_convert_neon + * jsimd_ycc_extbgr_convert_neon + * jsimd_ycc_extrgbx_convert_neon + * jsimd_ycc_extbgrx_convert_neon + * jsimd_ycc_extxbgr_convert_neon + * jsimd_ycc_extxrgb_convert_neon + * + * Colorspace conversion YCbCr -> RGB + */ + + +.macro do_load size + .if \size == 8 + ld1 {v4.8b}, [U], 8 + ld1 {v5.8b}, [V], 8 + ld1 {v0.8b}, [Y], 8 + prfm pldl1keep, [U, #64] + prfm pldl1keep, [V, #64] + prfm pldl1keep, [Y, #64] + .elseif \size == 4 + ld1 {v4.b}[0], [U], 1 + ld1 {v4.b}[1], [U], 1 + ld1 {v4.b}[2], [U], 1 + ld1 {v4.b}[3], [U], 1 + ld1 {v5.b}[0], [V], 1 + ld1 {v5.b}[1], [V], 1 + ld1 {v5.b}[2], [V], 1 + ld1 {v5.b}[3], [V], 1 + ld1 {v0.b}[0], [Y], 1 + ld1 {v0.b}[1], [Y], 1 + ld1 {v0.b}[2], [Y], 1 + ld1 {v0.b}[3], [Y], 1 + .elseif \size == 2 + ld1 {v4.b}[4], [U], 1 + ld1 {v4.b}[5], [U], 1 + ld1 {v5.b}[4], [V], 1 + ld1 {v5.b}[5], [V], 1 + ld1 {v0.b}[4], [Y], 1 + ld1 {v0.b}[5], [Y], 1 + .elseif \size == 1 + ld1 {v4.b}[6], [U], 1 + ld1 {v5.b}[6], [V], 1 + ld1 {v0.b}[6], [Y], 1 + .else + .error unsupported macroblock size + .endif +.endm + +.macro do_store bpp, size + .if \bpp == 24 + .if \size == 8 + st3 {v10.8b, v11.8b, v12.8b}, [RGB], 24 + .elseif \size == 4 + st3 {v10.b, v11.b, v12.b}[0], [RGB], 3 + st3 {v10.b, v11.b, v12.b}[1], [RGB], 3 + st3 {v10.b, v11.b, v12.b}[2], [RGB], 3 + st3 {v10.b, v11.b, v12.b}[3], [RGB], 3 + .elseif \size == 2 + st3 {v10.b, v11.b, v12.b}[4], [RGB], 3 + st3 {v10.b, v11.b, v12.b}[5], [RGB], 3 + .elseif \size == 1 + st3 {v10.b, v11.b, v12.b}[6], [RGB], 3 + .else + .error unsupported macroblock size + .endif + .elseif \bpp == 32 + .if \size == 8 + st4 {v10.8b, v11.8b, v12.8b, v13.8b}, [RGB], 32 + .elseif \size == 4 + st4 {v10.b, v11.b, v12.b, v13.b}[0], [RGB], 4 + st4 {v10.b, v11.b, v12.b, v13.b}[1], [RGB], 4 + st4 {v10.b, v11.b, v12.b, v13.b}[2], [RGB], 4 + st4 {v10.b, v11.b, v12.b, v13.b}[3], [RGB], 4 + .elseif \size == 2 + st4 {v10.b, v11.b, v12.b, v13.b}[4], [RGB], 4 + st4 {v10.b, v11.b, v12.b, v13.b}[5], [RGB], 4 + .elseif \size == 1 + st4 {v10.b, v11.b, v12.b, v13.b}[6], [RGB], 4 + .else + .error unsupported macroblock size + .endif + .elseif \bpp==16 + .if \size == 8 + st1 {v25.8h}, [RGB],16 + .elseif \size == 4 + st1 {v25.4h}, [RGB],8 + .elseif \size == 2 + st1 {v25.h}[4], [RGB],2 + st1 {v25.h}[5], [RGB],2 + .elseif \size == 1 + st1 {v25.h}[6], [RGB],2 + .else + .error unsupported macroblock size + .endif + .else + .error unsupported bpp + .endif +.endm + +.macro generate_jsimd_ycc_rgb_convert_neon colorid, bpp, r_offs, rsize, g_offs, gsize, b_offs, bsize, defsize + +/* + * 2-stage pipelined YCbCr->RGB conversion + */ + +.macro do_yuv_to_rgb_stage1 + uaddw v6.8h, v2.8h, v4.8b /* q3 = u - 128 */ + uaddw v8.8h, v2.8h, v5.8b /* q2 = v - 128 */ + smull v20.4s, v6.4h, v1.h[1] /* multiply by -11277 */ + smlal v20.4s, v8.4h, v1.h[2] /* multiply by -23401 */ + smull2 v22.4s, v6.8h, v1.h[1] /* multiply by -11277 */ + smlal2 v22.4s, v8.8h, v1.h[2] /* multiply by -23401 */ + smull v24.4s, v8.4h, v1.h[0] /* multiply by 22971 */ + smull2 v26.4s, v8.8h, v1.h[0] /* multiply by 22971 */ + smull v28.4s, v6.4h, v1.h[3] /* multiply by 29033 */ + smull2 v30.4s, v6.8h, v1.h[3] /* multiply by 29033 */ +.endm + +.macro do_yuv_to_rgb_stage2 + rshrn v20.4h, v20.4s, #15 + rshrn2 v20.8h, v22.4s, #15 + rshrn v24.4h, v24.4s, #14 + rshrn2 v24.8h, v26.4s, #14 + rshrn v28.4h, v28.4s, #14 + rshrn2 v28.8h, v30.4s, #14 + uaddw v20.8h, v20.8h, v0.8b + uaddw v24.8h, v24.8h, v0.8b + uaddw v28.8h, v28.8h, v0.8b +.if \bpp != 16 + sqxtun v1\g_offs\defsize, v20.8h + sqxtun v1\r_offs\defsize, v24.8h + sqxtun v1\b_offs\defsize, v28.8h +.else + sqshlu v21.8h, v20.8h, #8 + sqshlu v25.8h, v24.8h, #8 + sqshlu v29.8h, v28.8h, #8 + sri v25.8h, v21.8h, #5 + sri v25.8h, v29.8h, #11 +.endif + +.endm + +.macro do_yuv_to_rgb_stage2_store_load_stage1 + rshrn v20.4h, v20.4s, #15 + rshrn v24.4h, v24.4s, #14 + rshrn v28.4h, v28.4s, #14 + ld1 {v4.8b}, [U], 8 + rshrn2 v20.8h, v22.4s, #15 + rshrn2 v24.8h, v26.4s, #14 + rshrn2 v28.8h, v30.4s, #14 + ld1 {v5.8b}, [V], 8 + uaddw v20.8h, v20.8h, v0.8b + uaddw v24.8h, v24.8h, v0.8b + uaddw v28.8h, v28.8h, v0.8b +.if \bpp != 16 /**************** rgb24/rgb32 *********************************/ + sqxtun v1\g_offs\defsize, v20.8h + ld1 {v0.8b}, [Y], 8 + sqxtun v1\r_offs\defsize, v24.8h + prfm pldl1keep, [U, #64] + prfm pldl1keep, [V, #64] + prfm pldl1keep, [Y, #64] + sqxtun v1\b_offs\defsize, v28.8h + uaddw v6.8h, v2.8h, v4.8b /* v6.16b = u - 128 */ + uaddw v8.8h, v2.8h, v5.8b /* q2 = v - 128 */ + smull v20.4s, v6.4h, v1.h[1] /* multiply by -11277 */ + smlal v20.4s, v8.4h, v1.h[2] /* multiply by -23401 */ + smull2 v22.4s, v6.8h, v1.h[1] /* multiply by -11277 */ + smlal2 v22.4s, v8.8h, v1.h[2] /* multiply by -23401 */ + smull v24.4s, v8.4h, v1.h[0] /* multiply by 22971 */ + smull2 v26.4s, v8.8h, v1.h[0] /* multiply by 22971 */ +.else /**************************** rgb565 ***********************************/ + sqshlu v21.8h, v20.8h, #8 + sqshlu v25.8h, v24.8h, #8 + sqshlu v29.8h, v28.8h, #8 + uaddw v6.8h, v2.8h, v4.8b /* v6.16b = u - 128 */ + uaddw v8.8h, v2.8h, v5.8b /* q2 = v - 128 */ + ld1 {v0.8b}, [Y], 8 + smull v20.4s, v6.4h, v1.h[1] /* multiply by -11277 */ + smlal v20.4s, v8.4h, v1.h[2] /* multiply by -23401 */ + smull2 v22.4s, v6.8h, v1.h[1] /* multiply by -11277 */ + smlal2 v22.4s, v8.8h, v1.h[2] /* multiply by -23401 */ + sri v25.8h, v21.8h, #5 + smull v24.4s, v8.4h, v1.h[0] /* multiply by 22971 */ + smull2 v26.4s, v8.8h, v1.h[0] /* multiply by 22971 */ + prfm pldl1keep, [U, #64] + prfm pldl1keep, [V, #64] + prfm pldl1keep, [Y, #64] + sri v25.8h, v29.8h, #11 +.endif + do_store \bpp, 8 + smull v28.4s, v6.4h, v1.h[3] /* multiply by 29033 */ + smull2 v30.4s, v6.8h, v1.h[3] /* multiply by 29033 */ +.endm + +.macro do_yuv_to_rgb + do_yuv_to_rgb_stage1 + do_yuv_to_rgb_stage2 +.endm + +/* Apple gas crashes on adrl, work around that by using adr. + * But this requires a copy of these constants for each function. + */ + +.balign 16 +Ljsimd_ycc_\colorid\()_neon_consts: + .short 0, 0, 0, 0 + .short 22971, -11277, -23401, 29033 + .short -128, -128, -128, -128 + .short -128, -128, -128, -128 + +asm_function jsimd_ycc_\colorid\()_convert_neon + OUTPUT_WIDTH .req x0 + INPUT_BUF .req x1 + INPUT_ROW .req x2 + OUTPUT_BUF .req x3 + NUM_ROWS .req x4 + + INPUT_BUF0 .req x5 + INPUT_BUF1 .req x6 + INPUT_BUF2 .req x1 + + RGB .req x7 + Y .req x8 + U .req x9 + V .req x10 + N .req x15 + + sub sp, sp, 336 + str x15, [sp], 16 + /* Load constants to d1, d2, d3 (v0.4h is just used for padding) */ + adr x15, Ljsimd_ycc_\colorid\()_neon_consts + /* Save NEON registers */ + st1 {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32 + st1 {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32 + st1 {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32 + st1 {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32 + st1 {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32 + st1 {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32 + st1 {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32 + st1 {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32 + ld1 {v0.4h, v1.4h}, [x15], 16 + ld1 {v2.8h}, [x15] + + /* Save ARM registers and handle input arguments */ + /* push {x4, x5, x6, x7, x8, x9, x10, x30} */ + stp x4, x5, [sp], 16 + stp x6, x7, [sp], 16 + stp x8, x9, [sp], 16 + stp x10, x30, [sp], 16 + ldr INPUT_BUF0, [INPUT_BUF] + ldr INPUT_BUF1, [INPUT_BUF, #8] + ldr INPUT_BUF2, [INPUT_BUF, #16] + .unreq INPUT_BUF + + /* Initially set v10, v11.4h, v12.8b, d13 to 0xFF */ + movi v10.16b, #255 + movi v13.16b, #255 + + /* Outer loop over scanlines */ + cmp NUM_ROWS, #1 + b.lt 9f +0: + lsl x16, INPUT_ROW, #3 + ldr Y, [INPUT_BUF0, x16] + ldr U, [INPUT_BUF1, x16] + mov N, OUTPUT_WIDTH + ldr V, [INPUT_BUF2, x16] + add INPUT_ROW, INPUT_ROW, #1 + ldr RGB, [OUTPUT_BUF], #8 + + /* Inner loop over pixels */ + subs N, N, #8 + b.lt 3f + do_load 8 + do_yuv_to_rgb_stage1 + subs N, N, #8 + b.lt 2f +1: + do_yuv_to_rgb_stage2_store_load_stage1 + subs N, N, #8 + b.ge 1b +2: + do_yuv_to_rgb_stage2 + do_store \bpp, 8 + tst N, #7 + b.eq 8f +3: + tst N, #4 + b.eq 3f + do_load 4 +3: + tst N, #2 + b.eq 4f + do_load 2 +4: + tst N, #1 + b.eq 5f + do_load 1 +5: + do_yuv_to_rgb + tst N, #4 + b.eq 6f + do_store \bpp, 4 +6: + tst N, #2 + b.eq 7f + do_store \bpp, 2 +7: + tst N, #1 + b.eq 8f + do_store \bpp, 1 +8: + subs NUM_ROWS, NUM_ROWS, #1 + b.gt 0b +9: + /* Restore all registers and return */ + sub sp, sp, #336 + ldr x15, [sp], 16 + ld1 {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32 + ld1 {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32 + ld1 {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32 + ld1 {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32 + ld1 {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32 + ld1 {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32 + ld1 {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32 + ld1 {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32 + /* pop {r4, r5, r6, r7, r8, r9, r10, pc} */ + ldp x4, x5, [sp], 16 + ldp x6, x7, [sp], 16 + ldp x8, x9, [sp], 16 + ldp x10, x30, [sp], 16 + br x30 + .unreq OUTPUT_WIDTH + .unreq INPUT_ROW + .unreq OUTPUT_BUF + .unreq NUM_ROWS + .unreq INPUT_BUF0 + .unreq INPUT_BUF1 + .unreq INPUT_BUF2 + .unreq RGB + .unreq Y + .unreq U + .unreq V + .unreq N + +.purgem do_yuv_to_rgb +.purgem do_yuv_to_rgb_stage1 +.purgem do_yuv_to_rgb_stage2 +.purgem do_yuv_to_rgb_stage2_store_load_stage1 +.endm + +/*--------------------------------- id ----- bpp R rsize G gsize B bsize defsize */ +generate_jsimd_ycc_rgb_convert_neon extrgb, 24, 0, .4h, 1, .4h, 2, .4h, .8b +generate_jsimd_ycc_rgb_convert_neon extbgr, 24, 2, .4h, 1, .4h, 0, .4h, .8b +generate_jsimd_ycc_rgb_convert_neon extrgbx, 32, 0, .4h, 1, .4h, 2, .4h, .8b +generate_jsimd_ycc_rgb_convert_neon extbgrx, 32, 2, .4h, 1, .4h, 0, .4h, .8b +generate_jsimd_ycc_rgb_convert_neon extxbgr, 32, 3, .4h, 2, .4h, 1, .4h, .8b +generate_jsimd_ycc_rgb_convert_neon extxrgb, 32, 1, .4h, 2, .4h, 3, .4h, .8b +generate_jsimd_ycc_rgb_convert_neon rgb565, 16, 0, .4h, 0, .4h, 0, .4h, .8b +.purgem do_load +.purgem do_store diff --git a/app/src/main/jni/libjpeg-turbo/simd/jsimd_arm_neon.S b/app/src/main/jni/libjpeg-turbo/simd/jsimd_arm_neon.S new file mode 100644 index 000000000..4d9685b06 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jsimd_arm_neon.S @@ -0,0 +1,2439 @@ +/* + * ARMv7 NEON optimizations for libjpeg-turbo + * + * Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies). + * All rights reserved. + * Author: Siarhei Siamashka + * Copyright (C) 2014 Siarhei Siamashka. All Rights Reserved. + * Copyright (C) 2014 Linaro Limited. All Rights Reserved. + * + * This software is provided 'as-is', without any express or implied + * warranty. In no event will the authors be held liable for any damages + * arising from the use of this software. + * + * Permission is granted to anyone to use this software for any purpose, + * including commercial applications, and to alter it and redistribute it + * freely, subject to the following restrictions: + * + * 1. The origin of this software must not be misrepresented; you must not + * claim that you wrote the original software. If you use this software + * in a product, an acknowledgment in the product documentation would be + * appreciated but is not required. + * 2. Altered source versions must be plainly marked as such, and must not be + * misrepresented as being the original software. + * 3. This notice may not be removed or altered from any source distribution. + */ + +#if defined(__linux__) && defined(__ELF__) +.section .note.GNU-stack,"",%progbits /* mark stack as non-executable */ +#endif + +.text +.fpu neon +.arch armv7a +.object_arch armv4 +.arm + + +#define RESPECT_STRICT_ALIGNMENT 1 + + +/*****************************************************************************/ + +/* Supplementary macro for setting function attributes */ +.macro asm_function fname +#ifdef __APPLE__ + .globl _\fname +_\fname: +#else + .global \fname +#ifdef __ELF__ + .hidden \fname + .type \fname, %function +#endif +\fname: +#endif +.endm + +/* Transpose a block of 4x4 coefficients in four 64-bit registers */ +.macro transpose_4x4 x0, x1, x2, x3 + vtrn.16 \x0, \x1 + vtrn.16 \x2, \x3 + vtrn.32 \x0, \x2 + vtrn.32 \x1, \x3 +.endm + + +#define CENTERJSAMPLE 128 + +/*****************************************************************************/ + +/* + * Perform dequantization and inverse DCT on one block of coefficients. + * + * GLOBAL(void) + * jsimd_idct_islow_neon (void * dct_table, JCOEFPTR coef_block, + * JSAMPARRAY output_buf, JDIMENSION output_col) + */ + +#define FIX_0_298631336 (2446) +#define FIX_0_390180644 (3196) +#define FIX_0_541196100 (4433) +#define FIX_0_765366865 (6270) +#define FIX_0_899976223 (7373) +#define FIX_1_175875602 (9633) +#define FIX_1_501321110 (12299) +#define FIX_1_847759065 (15137) +#define FIX_1_961570560 (16069) +#define FIX_2_053119869 (16819) +#define FIX_2_562915447 (20995) +#define FIX_3_072711026 (25172) + +#define FIX_1_175875602_MINUS_1_961570560 (FIX_1_175875602 - FIX_1_961570560) +#define FIX_1_175875602_MINUS_0_390180644 (FIX_1_175875602 - FIX_0_390180644) +#define FIX_0_541196100_MINUS_1_847759065 (FIX_0_541196100 - FIX_1_847759065) +#define FIX_3_072711026_MINUS_2_562915447 (FIX_3_072711026 - FIX_2_562915447) +#define FIX_0_298631336_MINUS_0_899976223 (FIX_0_298631336 - FIX_0_899976223) +#define FIX_1_501321110_MINUS_0_899976223 (FIX_1_501321110 - FIX_0_899976223) +#define FIX_2_053119869_MINUS_2_562915447 (FIX_2_053119869 - FIX_2_562915447) +#define FIX_0_541196100_PLUS_0_765366865 (FIX_0_541196100 + FIX_0_765366865) + +/* + * Reference SIMD-friendly 1-D ISLOW iDCT C implementation. + * Uses some ideas from the comments in 'simd/jiss2int-64.asm' + */ +#define REF_1D_IDCT(xrow0, xrow1, xrow2, xrow3, xrow4, xrow5, xrow6, xrow7) \ +{ \ + DCTELEM row0, row1, row2, row3, row4, row5, row6, row7; \ + INT32 q1, q2, q3, q4, q5, q6, q7; \ + INT32 tmp11_plus_tmp2, tmp11_minus_tmp2; \ + \ + /* 1-D iDCT input data */ \ + row0 = xrow0; \ + row1 = xrow1; \ + row2 = xrow2; \ + row3 = xrow3; \ + row4 = xrow4; \ + row5 = xrow5; \ + row6 = xrow6; \ + row7 = xrow7; \ + \ + q5 = row7 + row3; \ + q4 = row5 + row1; \ + q6 = MULTIPLY(q5, FIX_1_175875602_MINUS_1_961570560) + \ + MULTIPLY(q4, FIX_1_175875602); \ + q7 = MULTIPLY(q5, FIX_1_175875602) + \ + MULTIPLY(q4, FIX_1_175875602_MINUS_0_390180644); \ + q2 = MULTIPLY(row2, FIX_0_541196100) + \ + MULTIPLY(row6, FIX_0_541196100_MINUS_1_847759065); \ + q4 = q6; \ + q3 = ((INT32) row0 - (INT32) row4) << 13; \ + q6 += MULTIPLY(row5, -FIX_2_562915447) + \ + MULTIPLY(row3, FIX_3_072711026_MINUS_2_562915447); \ + /* now we can use q1 (reloadable constants have been used up) */ \ + q1 = q3 + q2; \ + q4 += MULTIPLY(row7, FIX_0_298631336_MINUS_0_899976223) + \ + MULTIPLY(row1, -FIX_0_899976223); \ + q5 = q7; \ + q1 = q1 + q6; \ + q7 += MULTIPLY(row7, -FIX_0_899976223) + \ + MULTIPLY(row1, FIX_1_501321110_MINUS_0_899976223); \ + \ + /* (tmp11 + tmp2) has been calculated (out_row1 before descale) */ \ + tmp11_plus_tmp2 = q1; \ + row1 = 0; \ + \ + q1 = q1 - q6; \ + q5 += MULTIPLY(row5, FIX_2_053119869_MINUS_2_562915447) + \ + MULTIPLY(row3, -FIX_2_562915447); \ + q1 = q1 - q6; \ + q6 = MULTIPLY(row2, FIX_0_541196100_PLUS_0_765366865) + \ + MULTIPLY(row6, FIX_0_541196100); \ + q3 = q3 - q2; \ + \ + /* (tmp11 - tmp2) has been calculated (out_row6 before descale) */ \ + tmp11_minus_tmp2 = q1; \ + \ + q1 = ((INT32) row0 + (INT32) row4) << 13; \ + q2 = q1 + q6; \ + q1 = q1 - q6; \ + \ + /* pick up the results */ \ + tmp0 = q4; \ + tmp1 = q5; \ + tmp2 = (tmp11_plus_tmp2 - tmp11_minus_tmp2) / 2; \ + tmp3 = q7; \ + tmp10 = q2; \ + tmp11 = (tmp11_plus_tmp2 + tmp11_minus_tmp2) / 2; \ + tmp12 = q3; \ + tmp13 = q1; \ +} + +#define XFIX_0_899976223 d0[0] +#define XFIX_0_541196100 d0[1] +#define XFIX_2_562915447 d0[2] +#define XFIX_0_298631336_MINUS_0_899976223 d0[3] +#define XFIX_1_501321110_MINUS_0_899976223 d1[0] +#define XFIX_2_053119869_MINUS_2_562915447 d1[1] +#define XFIX_0_541196100_PLUS_0_765366865 d1[2] +#define XFIX_1_175875602 d1[3] +#define XFIX_1_175875602_MINUS_0_390180644 d2[0] +#define XFIX_0_541196100_MINUS_1_847759065 d2[1] +#define XFIX_3_072711026_MINUS_2_562915447 d2[2] +#define XFIX_1_175875602_MINUS_1_961570560 d2[3] + +.balign 16 +jsimd_idct_islow_neon_consts: + .short FIX_0_899976223 /* d0[0] */ + .short FIX_0_541196100 /* d0[1] */ + .short FIX_2_562915447 /* d0[2] */ + .short FIX_0_298631336_MINUS_0_899976223 /* d0[3] */ + .short FIX_1_501321110_MINUS_0_899976223 /* d1[0] */ + .short FIX_2_053119869_MINUS_2_562915447 /* d1[1] */ + .short FIX_0_541196100_PLUS_0_765366865 /* d1[2] */ + .short FIX_1_175875602 /* d1[3] */ + /* reloadable constants */ + .short FIX_1_175875602_MINUS_0_390180644 /* d2[0] */ + .short FIX_0_541196100_MINUS_1_847759065 /* d2[1] */ + .short FIX_3_072711026_MINUS_2_562915447 /* d2[2] */ + .short FIX_1_175875602_MINUS_1_961570560 /* d2[3] */ + +asm_function jsimd_idct_islow_neon + + DCT_TABLE .req r0 + COEF_BLOCK .req r1 + OUTPUT_BUF .req r2 + OUTPUT_COL .req r3 + TMP1 .req r0 + TMP2 .req r1 + TMP3 .req r2 + TMP4 .req ip + + ROW0L .req d16 + ROW0R .req d17 + ROW1L .req d18 + ROW1R .req d19 + ROW2L .req d20 + ROW2R .req d21 + ROW3L .req d22 + ROW3R .req d23 + ROW4L .req d24 + ROW4R .req d25 + ROW5L .req d26 + ROW5R .req d27 + ROW6L .req d28 + ROW6R .req d29 + ROW7L .req d30 + ROW7R .req d31 + + /* Load and dequantize coefficients into NEON registers + * with the following allocation: + * 0 1 2 3 | 4 5 6 7 + * ---------+-------- + * 0 | d16 | d17 ( q8 ) + * 1 | d18 | d19 ( q9 ) + * 2 | d20 | d21 ( q10 ) + * 3 | d22 | d23 ( q11 ) + * 4 | d24 | d25 ( q12 ) + * 5 | d26 | d27 ( q13 ) + * 6 | d28 | d29 ( q14 ) + * 7 | d30 | d31 ( q15 ) + */ + adr ip, jsimd_idct_islow_neon_consts + vld1.16 {d16, d17, d18, d19}, [COEF_BLOCK, :128]! + vld1.16 {d0, d1, d2, d3}, [DCT_TABLE, :128]! + vld1.16 {d20, d21, d22, d23}, [COEF_BLOCK, :128]! + vmul.s16 q8, q8, q0 + vld1.16 {d4, d5, d6, d7}, [DCT_TABLE, :128]! + vmul.s16 q9, q9, q1 + vld1.16 {d24, d25, d26, d27}, [COEF_BLOCK, :128]! + vmul.s16 q10, q10, q2 + vld1.16 {d0, d1, d2, d3}, [DCT_TABLE, :128]! + vmul.s16 q11, q11, q3 + vld1.16 {d28, d29, d30, d31}, [COEF_BLOCK, :128] + vmul.s16 q12, q12, q0 + vld1.16 {d4, d5, d6, d7}, [DCT_TABLE, :128]! + vmul.s16 q14, q14, q2 + vmul.s16 q13, q13, q1 + vld1.16 {d0, d1, d2, d3}, [ip, :128] /* load constants */ + add ip, ip, #16 + vmul.s16 q15, q15, q3 + vpush {d8-d15} /* save NEON registers */ + /* 1-D IDCT, pass 1, left 4x8 half */ + vadd.s16 d4, ROW7L, ROW3L + vadd.s16 d5, ROW5L, ROW1L + vmull.s16 q6, d4, XFIX_1_175875602_MINUS_1_961570560 + vmlal.s16 q6, d5, XFIX_1_175875602 + vmull.s16 q7, d4, XFIX_1_175875602 + /* Check for the zero coefficients in the right 4x8 half */ + push {r4, r5} + vmlal.s16 q7, d5, XFIX_1_175875602_MINUS_0_390180644 + vsubl.s16 q3, ROW0L, ROW4L + ldrd r4, [COEF_BLOCK, #(-96 + 2 * (4 + 1 * 8))] + vmull.s16 q2, ROW2L, XFIX_0_541196100 + vmlal.s16 q2, ROW6L, XFIX_0_541196100_MINUS_1_847759065 + orr r0, r4, r5 + vmov q4, q6 + vmlsl.s16 q6, ROW5L, XFIX_2_562915447 + ldrd r4, [COEF_BLOCK, #(-96 + 2 * (4 + 2 * 8))] + vmlal.s16 q6, ROW3L, XFIX_3_072711026_MINUS_2_562915447 + vshl.s32 q3, q3, #13 + orr r0, r0, r4 + vmlsl.s16 q4, ROW1L, XFIX_0_899976223 + orr r0, r0, r5 + vadd.s32 q1, q3, q2 + ldrd r4, [COEF_BLOCK, #(-96 + 2 * (4 + 3 * 8))] + vmov q5, q7 + vadd.s32 q1, q1, q6 + orr r0, r0, r4 + vmlsl.s16 q7, ROW7L, XFIX_0_899976223 + orr r0, r0, r5 + vmlal.s16 q7, ROW1L, XFIX_1_501321110_MINUS_0_899976223 + vrshrn.s32 ROW1L, q1, #11 + ldrd r4, [COEF_BLOCK, #(-96 + 2 * (4 + 4 * 8))] + vsub.s32 q1, q1, q6 + vmlal.s16 q5, ROW5L, XFIX_2_053119869_MINUS_2_562915447 + orr r0, r0, r4 + vmlsl.s16 q5, ROW3L, XFIX_2_562915447 + orr r0, r0, r5 + vsub.s32 q1, q1, q6 + vmull.s16 q6, ROW2L, XFIX_0_541196100_PLUS_0_765366865 + ldrd r4, [COEF_BLOCK, #(-96 + 2 * (4 + 5 * 8))] + vmlal.s16 q6, ROW6L, XFIX_0_541196100 + vsub.s32 q3, q3, q2 + orr r0, r0, r4 + vrshrn.s32 ROW6L, q1, #11 + orr r0, r0, r5 + vadd.s32 q1, q3, q5 + ldrd r4, [COEF_BLOCK, #(-96 + 2 * (4 + 6 * 8))] + vsub.s32 q3, q3, q5 + vaddl.s16 q5, ROW0L, ROW4L + orr r0, r0, r4 + vrshrn.s32 ROW2L, q1, #11 + orr r0, r0, r5 + vrshrn.s32 ROW5L, q3, #11 + ldrd r4, [COEF_BLOCK, #(-96 + 2 * (4 + 7 * 8))] + vshl.s32 q5, q5, #13 + vmlal.s16 q4, ROW7L, XFIX_0_298631336_MINUS_0_899976223 + orr r0, r0, r4 + vadd.s32 q2, q5, q6 + orrs r0, r0, r5 + vsub.s32 q1, q5, q6 + vadd.s32 q6, q2, q7 + ldrd r4, [COEF_BLOCK, #(-96 + 2 * (4 + 0 * 8))] + vsub.s32 q2, q2, q7 + vadd.s32 q5, q1, q4 + orr r0, r4, r5 + vsub.s32 q3, q1, q4 + pop {r4, r5} + vrshrn.s32 ROW7L, q2, #11 + vrshrn.s32 ROW3L, q5, #11 + vrshrn.s32 ROW0L, q6, #11 + vrshrn.s32 ROW4L, q3, #11 + + beq 3f /* Go to do some special handling for the sparse right 4x8 half */ + + /* 1-D IDCT, pass 1, right 4x8 half */ + vld1.s16 {d2}, [ip, :64] /* reload constants */ + vadd.s16 d10, ROW7R, ROW3R + vadd.s16 d8, ROW5R, ROW1R + /* Transpose left 4x8 half */ + vtrn.16 ROW6L, ROW7L + vmull.s16 q6, d10, XFIX_1_175875602_MINUS_1_961570560 + vmlal.s16 q6, d8, XFIX_1_175875602 + vtrn.16 ROW2L, ROW3L + vmull.s16 q7, d10, XFIX_1_175875602 + vmlal.s16 q7, d8, XFIX_1_175875602_MINUS_0_390180644 + vtrn.16 ROW0L, ROW1L + vsubl.s16 q3, ROW0R, ROW4R + vmull.s16 q2, ROW2R, XFIX_0_541196100 + vmlal.s16 q2, ROW6R, XFIX_0_541196100_MINUS_1_847759065 + vtrn.16 ROW4L, ROW5L + vmov q4, q6 + vmlsl.s16 q6, ROW5R, XFIX_2_562915447 + vmlal.s16 q6, ROW3R, XFIX_3_072711026_MINUS_2_562915447 + vtrn.32 ROW1L, ROW3L + vshl.s32 q3, q3, #13 + vmlsl.s16 q4, ROW1R, XFIX_0_899976223 + vtrn.32 ROW4L, ROW6L + vadd.s32 q1, q3, q2 + vmov q5, q7 + vadd.s32 q1, q1, q6 + vtrn.32 ROW0L, ROW2L + vmlsl.s16 q7, ROW7R, XFIX_0_899976223 + vmlal.s16 q7, ROW1R, XFIX_1_501321110_MINUS_0_899976223 + vrshrn.s32 ROW1R, q1, #11 + vtrn.32 ROW5L, ROW7L + vsub.s32 q1, q1, q6 + vmlal.s16 q5, ROW5R, XFIX_2_053119869_MINUS_2_562915447 + vmlsl.s16 q5, ROW3R, XFIX_2_562915447 + vsub.s32 q1, q1, q6 + vmull.s16 q6, ROW2R, XFIX_0_541196100_PLUS_0_765366865 + vmlal.s16 q6, ROW6R, XFIX_0_541196100 + vsub.s32 q3, q3, q2 + vrshrn.s32 ROW6R, q1, #11 + vadd.s32 q1, q3, q5 + vsub.s32 q3, q3, q5 + vaddl.s16 q5, ROW0R, ROW4R + vrshrn.s32 ROW2R, q1, #11 + vrshrn.s32 ROW5R, q3, #11 + vshl.s32 q5, q5, #13 + vmlal.s16 q4, ROW7R, XFIX_0_298631336_MINUS_0_899976223 + vadd.s32 q2, q5, q6 + vsub.s32 q1, q5, q6 + vadd.s32 q6, q2, q7 + vsub.s32 q2, q2, q7 + vadd.s32 q5, q1, q4 + vsub.s32 q3, q1, q4 + vrshrn.s32 ROW7R, q2, #11 + vrshrn.s32 ROW3R, q5, #11 + vrshrn.s32 ROW0R, q6, #11 + vrshrn.s32 ROW4R, q3, #11 + /* Transpose right 4x8 half */ + vtrn.16 ROW6R, ROW7R + vtrn.16 ROW2R, ROW3R + vtrn.16 ROW0R, ROW1R + vtrn.16 ROW4R, ROW5R + vtrn.32 ROW1R, ROW3R + vtrn.32 ROW4R, ROW6R + vtrn.32 ROW0R, ROW2R + vtrn.32 ROW5R, ROW7R + +1: /* 1-D IDCT, pass 2 (normal variant), left 4x8 half */ + vld1.s16 {d2}, [ip, :64] /* reload constants */ + vmull.s16 q6, ROW1R, XFIX_1_175875602 /* ROW5L <-> ROW1R */ + vmlal.s16 q6, ROW1L, XFIX_1_175875602 + vmlal.s16 q6, ROW3R, XFIX_1_175875602_MINUS_1_961570560 /* ROW7L <-> ROW3R */ + vmlal.s16 q6, ROW3L, XFIX_1_175875602_MINUS_1_961570560 + vmull.s16 q7, ROW3R, XFIX_1_175875602 /* ROW7L <-> ROW3R */ + vmlal.s16 q7, ROW3L, XFIX_1_175875602 + vmlal.s16 q7, ROW1R, XFIX_1_175875602_MINUS_0_390180644 /* ROW5L <-> ROW1R */ + vmlal.s16 q7, ROW1L, XFIX_1_175875602_MINUS_0_390180644 + vsubl.s16 q3, ROW0L, ROW0R /* ROW4L <-> ROW0R */ + vmull.s16 q2, ROW2L, XFIX_0_541196100 + vmlal.s16 q2, ROW2R, XFIX_0_541196100_MINUS_1_847759065 /* ROW6L <-> ROW2R */ + vmov q4, q6 + vmlsl.s16 q6, ROW1R, XFIX_2_562915447 /* ROW5L <-> ROW1R */ + vmlal.s16 q6, ROW3L, XFIX_3_072711026_MINUS_2_562915447 + vshl.s32 q3, q3, #13 + vmlsl.s16 q4, ROW1L, XFIX_0_899976223 + vadd.s32 q1, q3, q2 + vmov q5, q7 + vadd.s32 q1, q1, q6 + vmlsl.s16 q7, ROW3R, XFIX_0_899976223 /* ROW7L <-> ROW3R */ + vmlal.s16 q7, ROW1L, XFIX_1_501321110_MINUS_0_899976223 + vshrn.s32 ROW1L, q1, #16 + vsub.s32 q1, q1, q6 + vmlal.s16 q5, ROW1R, XFIX_2_053119869_MINUS_2_562915447 /* ROW5L <-> ROW1R */ + vmlsl.s16 q5, ROW3L, XFIX_2_562915447 + vsub.s32 q1, q1, q6 + vmull.s16 q6, ROW2L, XFIX_0_541196100_PLUS_0_765366865 + vmlal.s16 q6, ROW2R, XFIX_0_541196100 /* ROW6L <-> ROW2R */ + vsub.s32 q3, q3, q2 + vshrn.s32 ROW2R, q1, #16 /* ROW6L <-> ROW2R */ + vadd.s32 q1, q3, q5 + vsub.s32 q3, q3, q5 + vaddl.s16 q5, ROW0L, ROW0R /* ROW4L <-> ROW0R */ + vshrn.s32 ROW2L, q1, #16 + vshrn.s32 ROW1R, q3, #16 /* ROW5L <-> ROW1R */ + vshl.s32 q5, q5, #13 + vmlal.s16 q4, ROW3R, XFIX_0_298631336_MINUS_0_899976223 /* ROW7L <-> ROW3R */ + vadd.s32 q2, q5, q6 + vsub.s32 q1, q5, q6 + vadd.s32 q6, q2, q7 + vsub.s32 q2, q2, q7 + vadd.s32 q5, q1, q4 + vsub.s32 q3, q1, q4 + vshrn.s32 ROW3R, q2, #16 /* ROW7L <-> ROW3R */ + vshrn.s32 ROW3L, q5, #16 + vshrn.s32 ROW0L, q6, #16 + vshrn.s32 ROW0R, q3, #16 /* ROW4L <-> ROW0R */ + /* 1-D IDCT, pass 2, right 4x8 half */ + vld1.s16 {d2}, [ip, :64] /* reload constants */ + vmull.s16 q6, ROW5R, XFIX_1_175875602 + vmlal.s16 q6, ROW5L, XFIX_1_175875602 /* ROW5L <-> ROW1R */ + vmlal.s16 q6, ROW7R, XFIX_1_175875602_MINUS_1_961570560 + vmlal.s16 q6, ROW7L, XFIX_1_175875602_MINUS_1_961570560 /* ROW7L <-> ROW3R */ + vmull.s16 q7, ROW7R, XFIX_1_175875602 + vmlal.s16 q7, ROW7L, XFIX_1_175875602 /* ROW7L <-> ROW3R */ + vmlal.s16 q7, ROW5R, XFIX_1_175875602_MINUS_0_390180644 + vmlal.s16 q7, ROW5L, XFIX_1_175875602_MINUS_0_390180644 /* ROW5L <-> ROW1R */ + vsubl.s16 q3, ROW4L, ROW4R /* ROW4L <-> ROW0R */ + vmull.s16 q2, ROW6L, XFIX_0_541196100 /* ROW6L <-> ROW2R */ + vmlal.s16 q2, ROW6R, XFIX_0_541196100_MINUS_1_847759065 + vmov q4, q6 + vmlsl.s16 q6, ROW5R, XFIX_2_562915447 + vmlal.s16 q6, ROW7L, XFIX_3_072711026_MINUS_2_562915447 /* ROW7L <-> ROW3R */ + vshl.s32 q3, q3, #13 + vmlsl.s16 q4, ROW5L, XFIX_0_899976223 /* ROW5L <-> ROW1R */ + vadd.s32 q1, q3, q2 + vmov q5, q7 + vadd.s32 q1, q1, q6 + vmlsl.s16 q7, ROW7R, XFIX_0_899976223 + vmlal.s16 q7, ROW5L, XFIX_1_501321110_MINUS_0_899976223 /* ROW5L <-> ROW1R */ + vshrn.s32 ROW5L, q1, #16 /* ROW5L <-> ROW1R */ + vsub.s32 q1, q1, q6 + vmlal.s16 q5, ROW5R, XFIX_2_053119869_MINUS_2_562915447 + vmlsl.s16 q5, ROW7L, XFIX_2_562915447 /* ROW7L <-> ROW3R */ + vsub.s32 q1, q1, q6 + vmull.s16 q6, ROW6L, XFIX_0_541196100_PLUS_0_765366865 /* ROW6L <-> ROW2R */ + vmlal.s16 q6, ROW6R, XFIX_0_541196100 + vsub.s32 q3, q3, q2 + vshrn.s32 ROW6R, q1, #16 + vadd.s32 q1, q3, q5 + vsub.s32 q3, q3, q5 + vaddl.s16 q5, ROW4L, ROW4R /* ROW4L <-> ROW0R */ + vshrn.s32 ROW6L, q1, #16 /* ROW6L <-> ROW2R */ + vshrn.s32 ROW5R, q3, #16 + vshl.s32 q5, q5, #13 + vmlal.s16 q4, ROW7R, XFIX_0_298631336_MINUS_0_899976223 + vadd.s32 q2, q5, q6 + vsub.s32 q1, q5, q6 + vadd.s32 q6, q2, q7 + vsub.s32 q2, q2, q7 + vadd.s32 q5, q1, q4 + vsub.s32 q3, q1, q4 + vshrn.s32 ROW7R, q2, #16 + vshrn.s32 ROW7L, q5, #16 /* ROW7L <-> ROW3R */ + vshrn.s32 ROW4L, q6, #16 /* ROW4L <-> ROW0R */ + vshrn.s32 ROW4R, q3, #16 + +2: /* Descale to 8-bit and range limit */ + vqrshrn.s16 d16, q8, #2 + vqrshrn.s16 d17, q9, #2 + vqrshrn.s16 d18, q10, #2 + vqrshrn.s16 d19, q11, #2 + vpop {d8-d15} /* restore NEON registers */ + vqrshrn.s16 d20, q12, #2 + /* Transpose the final 8-bit samples and do signed->unsigned conversion */ + vtrn.16 q8, q9 + vqrshrn.s16 d21, q13, #2 + vqrshrn.s16 d22, q14, #2 + vmov.u8 q0, #(CENTERJSAMPLE) + vqrshrn.s16 d23, q15, #2 + vtrn.8 d16, d17 + vtrn.8 d18, d19 + vadd.u8 q8, q8, q0 + vadd.u8 q9, q9, q0 + vtrn.16 q10, q11 + /* Store results to the output buffer */ + ldmia OUTPUT_BUF!, {TMP1, TMP2} + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + vst1.8 {d16}, [TMP1] + vtrn.8 d20, d21 + vst1.8 {d17}, [TMP2] + ldmia OUTPUT_BUF!, {TMP1, TMP2} + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + vst1.8 {d18}, [TMP1] + vadd.u8 q10, q10, q0 + vst1.8 {d19}, [TMP2] + ldmia OUTPUT_BUF, {TMP1, TMP2, TMP3, TMP4} + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + add TMP3, TMP3, OUTPUT_COL + add TMP4, TMP4, OUTPUT_COL + vtrn.8 d22, d23 + vst1.8 {d20}, [TMP1] + vadd.u8 q11, q11, q0 + vst1.8 {d21}, [TMP2] + vst1.8 {d22}, [TMP3] + vst1.8 {d23}, [TMP4] + bx lr + +3: /* Left 4x8 half is done, right 4x8 half contains mostly zeros */ + + /* Transpose left 4x8 half */ + vtrn.16 ROW6L, ROW7L + vtrn.16 ROW2L, ROW3L + vtrn.16 ROW0L, ROW1L + vtrn.16 ROW4L, ROW5L + vshl.s16 ROW0R, ROW0R, #2 /* PASS1_BITS */ + vtrn.32 ROW1L, ROW3L + vtrn.32 ROW4L, ROW6L + vtrn.32 ROW0L, ROW2L + vtrn.32 ROW5L, ROW7L + + cmp r0, #0 + beq 4f /* Right 4x8 half has all zeros, go to 'sparse' second pass */ + + /* Only row 0 is non-zero for the right 4x8 half */ + vdup.s16 ROW1R, ROW0R[1] + vdup.s16 ROW2R, ROW0R[2] + vdup.s16 ROW3R, ROW0R[3] + vdup.s16 ROW4R, ROW0R[0] + vdup.s16 ROW5R, ROW0R[1] + vdup.s16 ROW6R, ROW0R[2] + vdup.s16 ROW7R, ROW0R[3] + vdup.s16 ROW0R, ROW0R[0] + b 1b /* Go to 'normal' second pass */ + +4: /* 1-D IDCT, pass 2 (sparse variant with zero rows 4-7), left 4x8 half */ + vld1.s16 {d2}, [ip, :64] /* reload constants */ + vmull.s16 q6, ROW1L, XFIX_1_175875602 + vmlal.s16 q6, ROW3L, XFIX_1_175875602_MINUS_1_961570560 + vmull.s16 q7, ROW3L, XFIX_1_175875602 + vmlal.s16 q7, ROW1L, XFIX_1_175875602_MINUS_0_390180644 + vmull.s16 q2, ROW2L, XFIX_0_541196100 + vshll.s16 q3, ROW0L, #13 + vmov q4, q6 + vmlal.s16 q6, ROW3L, XFIX_3_072711026_MINUS_2_562915447 + vmlsl.s16 q4, ROW1L, XFIX_0_899976223 + vadd.s32 q1, q3, q2 + vmov q5, q7 + vmlal.s16 q7, ROW1L, XFIX_1_501321110_MINUS_0_899976223 + vadd.s32 q1, q1, q6 + vadd.s32 q6, q6, q6 + vmlsl.s16 q5, ROW3L, XFIX_2_562915447 + vshrn.s32 ROW1L, q1, #16 + vsub.s32 q1, q1, q6 + vmull.s16 q6, ROW2L, XFIX_0_541196100_PLUS_0_765366865 + vsub.s32 q3, q3, q2 + vshrn.s32 ROW2R, q1, #16 /* ROW6L <-> ROW2R */ + vadd.s32 q1, q3, q5 + vsub.s32 q3, q3, q5 + vshll.s16 q5, ROW0L, #13 + vshrn.s32 ROW2L, q1, #16 + vshrn.s32 ROW1R, q3, #16 /* ROW5L <-> ROW1R */ + vadd.s32 q2, q5, q6 + vsub.s32 q1, q5, q6 + vadd.s32 q6, q2, q7 + vsub.s32 q2, q2, q7 + vadd.s32 q5, q1, q4 + vsub.s32 q3, q1, q4 + vshrn.s32 ROW3R, q2, #16 /* ROW7L <-> ROW3R */ + vshrn.s32 ROW3L, q5, #16 + vshrn.s32 ROW0L, q6, #16 + vshrn.s32 ROW0R, q3, #16 /* ROW4L <-> ROW0R */ + /* 1-D IDCT, pass 2 (sparse variant with zero rows 4-7), right 4x8 half */ + vld1.s16 {d2}, [ip, :64] /* reload constants */ + vmull.s16 q6, ROW5L, XFIX_1_175875602 + vmlal.s16 q6, ROW7L, XFIX_1_175875602_MINUS_1_961570560 + vmull.s16 q7, ROW7L, XFIX_1_175875602 + vmlal.s16 q7, ROW5L, XFIX_1_175875602_MINUS_0_390180644 + vmull.s16 q2, ROW6L, XFIX_0_541196100 + vshll.s16 q3, ROW4L, #13 + vmov q4, q6 + vmlal.s16 q6, ROW7L, XFIX_3_072711026_MINUS_2_562915447 + vmlsl.s16 q4, ROW5L, XFIX_0_899976223 + vadd.s32 q1, q3, q2 + vmov q5, q7 + vmlal.s16 q7, ROW5L, XFIX_1_501321110_MINUS_0_899976223 + vadd.s32 q1, q1, q6 + vadd.s32 q6, q6, q6 + vmlsl.s16 q5, ROW7L, XFIX_2_562915447 + vshrn.s32 ROW5L, q1, #16 /* ROW5L <-> ROW1R */ + vsub.s32 q1, q1, q6 + vmull.s16 q6, ROW6L, XFIX_0_541196100_PLUS_0_765366865 + vsub.s32 q3, q3, q2 + vshrn.s32 ROW6R, q1, #16 + vadd.s32 q1, q3, q5 + vsub.s32 q3, q3, q5 + vshll.s16 q5, ROW4L, #13 + vshrn.s32 ROW6L, q1, #16 /* ROW6L <-> ROW2R */ + vshrn.s32 ROW5R, q3, #16 + vadd.s32 q2, q5, q6 + vsub.s32 q1, q5, q6 + vadd.s32 q6, q2, q7 + vsub.s32 q2, q2, q7 + vadd.s32 q5, q1, q4 + vsub.s32 q3, q1, q4 + vshrn.s32 ROW7R, q2, #16 + vshrn.s32 ROW7L, q5, #16 /* ROW7L <-> ROW3R */ + vshrn.s32 ROW4L, q6, #16 /* ROW4L <-> ROW0R */ + vshrn.s32 ROW4R, q3, #16 + b 2b /* Go to epilogue */ + + .unreq DCT_TABLE + .unreq COEF_BLOCK + .unreq OUTPUT_BUF + .unreq OUTPUT_COL + .unreq TMP1 + .unreq TMP2 + .unreq TMP3 + .unreq TMP4 + + .unreq ROW0L + .unreq ROW0R + .unreq ROW1L + .unreq ROW1R + .unreq ROW2L + .unreq ROW2R + .unreq ROW3L + .unreq ROW3R + .unreq ROW4L + .unreq ROW4R + .unreq ROW5L + .unreq ROW5R + .unreq ROW6L + .unreq ROW6R + .unreq ROW7L + .unreq ROW7R + + +/*****************************************************************************/ + +/* + * jsimd_idct_ifast_neon + * + * This function contains a fast, not so accurate integer implementation of + * the inverse DCT (Discrete Cosine Transform). It uses the same calculations + * and produces exactly the same output as IJG's original 'jpeg_idct_ifast' + * function from jidctfst.c + * + * Normally 1-D AAN DCT needs 5 multiplications and 29 additions. + * But in ARM NEON case some extra additions are required because VQDMULH + * instruction can't handle the constants larger than 1. So the expressions + * like "x * 1.082392200" have to be converted to "x * 0.082392200 + x", + * which introduces an extra addition. Overall, there are 6 extra additions + * per 1-D IDCT pass, totalling to 5 VQDMULH and 35 VADD/VSUB instructions. + */ + +#define XFIX_1_082392200 d0[0] +#define XFIX_1_414213562 d0[1] +#define XFIX_1_847759065 d0[2] +#define XFIX_2_613125930 d0[3] + +.balign 16 +jsimd_idct_ifast_neon_consts: + .short (277 * 128 - 256 * 128) /* XFIX_1_082392200 */ + .short (362 * 128 - 256 * 128) /* XFIX_1_414213562 */ + .short (473 * 128 - 256 * 128) /* XFIX_1_847759065 */ + .short (669 * 128 - 512 * 128) /* XFIX_2_613125930 */ + +asm_function jsimd_idct_ifast_neon + + DCT_TABLE .req r0 + COEF_BLOCK .req r1 + OUTPUT_BUF .req r2 + OUTPUT_COL .req r3 + TMP1 .req r0 + TMP2 .req r1 + TMP3 .req r2 + TMP4 .req ip + + /* Load and dequantize coefficients into NEON registers + * with the following allocation: + * 0 1 2 3 | 4 5 6 7 + * ---------+-------- + * 0 | d16 | d17 ( q8 ) + * 1 | d18 | d19 ( q9 ) + * 2 | d20 | d21 ( q10 ) + * 3 | d22 | d23 ( q11 ) + * 4 | d24 | d25 ( q12 ) + * 5 | d26 | d27 ( q13 ) + * 6 | d28 | d29 ( q14 ) + * 7 | d30 | d31 ( q15 ) + */ + adr ip, jsimd_idct_ifast_neon_consts + vld1.16 {d16, d17, d18, d19}, [COEF_BLOCK, :128]! + vld1.16 {d0, d1, d2, d3}, [DCT_TABLE, :128]! + vld1.16 {d20, d21, d22, d23}, [COEF_BLOCK, :128]! + vmul.s16 q8, q8, q0 + vld1.16 {d4, d5, d6, d7}, [DCT_TABLE, :128]! + vmul.s16 q9, q9, q1 + vld1.16 {d24, d25, d26, d27}, [COEF_BLOCK, :128]! + vmul.s16 q10, q10, q2 + vld1.16 {d0, d1, d2, d3}, [DCT_TABLE, :128]! + vmul.s16 q11, q11, q3 + vld1.16 {d28, d29, d30, d31}, [COEF_BLOCK, :128] + vmul.s16 q12, q12, q0 + vld1.16 {d4, d5, d6, d7}, [DCT_TABLE, :128]! + vmul.s16 q14, q14, q2 + vmul.s16 q13, q13, q1 + vld1.16 {d0}, [ip, :64] /* load constants */ + vmul.s16 q15, q15, q3 + vpush {d8-d13} /* save NEON registers */ + /* 1-D IDCT, pass 1 */ + vsub.s16 q2, q10, q14 + vadd.s16 q14, q10, q14 + vsub.s16 q1, q11, q13 + vadd.s16 q13, q11, q13 + vsub.s16 q5, q9, q15 + vadd.s16 q15, q9, q15 + vqdmulh.s16 q4, q2, XFIX_1_414213562 + vqdmulh.s16 q6, q1, XFIX_2_613125930 + vadd.s16 q3, q1, q1 + vsub.s16 q1, q5, q1 + vadd.s16 q10, q2, q4 + vqdmulh.s16 q4, q1, XFIX_1_847759065 + vsub.s16 q2, q15, q13 + vadd.s16 q3, q3, q6 + vqdmulh.s16 q6, q2, XFIX_1_414213562 + vadd.s16 q1, q1, q4 + vqdmulh.s16 q4, q5, XFIX_1_082392200 + vsub.s16 q10, q10, q14 + vadd.s16 q2, q2, q6 + vsub.s16 q6, q8, q12 + vadd.s16 q12, q8, q12 + vadd.s16 q9, q5, q4 + vadd.s16 q5, q6, q10 + vsub.s16 q10, q6, q10 + vadd.s16 q6, q15, q13 + vadd.s16 q8, q12, q14 + vsub.s16 q3, q6, q3 + vsub.s16 q12, q12, q14 + vsub.s16 q3, q3, q1 + vsub.s16 q1, q9, q1 + vadd.s16 q2, q3, q2 + vsub.s16 q15, q8, q6 + vadd.s16 q1, q1, q2 + vadd.s16 q8, q8, q6 + vadd.s16 q14, q5, q3 + vsub.s16 q9, q5, q3 + vsub.s16 q13, q10, q2 + vadd.s16 q10, q10, q2 + /* Transpose */ + vtrn.16 q8, q9 + vsub.s16 q11, q12, q1 + vtrn.16 q14, q15 + vadd.s16 q12, q12, q1 + vtrn.16 q10, q11 + vtrn.16 q12, q13 + vtrn.32 q9, q11 + vtrn.32 q12, q14 + vtrn.32 q8, q10 + vtrn.32 q13, q15 + vswp d28, d21 + vswp d26, d19 + /* 1-D IDCT, pass 2 */ + vsub.s16 q2, q10, q14 + vswp d30, d23 + vadd.s16 q14, q10, q14 + vswp d24, d17 + vsub.s16 q1, q11, q13 + vadd.s16 q13, q11, q13 + vsub.s16 q5, q9, q15 + vadd.s16 q15, q9, q15 + vqdmulh.s16 q4, q2, XFIX_1_414213562 + vqdmulh.s16 q6, q1, XFIX_2_613125930 + vadd.s16 q3, q1, q1 + vsub.s16 q1, q5, q1 + vadd.s16 q10, q2, q4 + vqdmulh.s16 q4, q1, XFIX_1_847759065 + vsub.s16 q2, q15, q13 + vadd.s16 q3, q3, q6 + vqdmulh.s16 q6, q2, XFIX_1_414213562 + vadd.s16 q1, q1, q4 + vqdmulh.s16 q4, q5, XFIX_1_082392200 + vsub.s16 q10, q10, q14 + vadd.s16 q2, q2, q6 + vsub.s16 q6, q8, q12 + vadd.s16 q12, q8, q12 + vadd.s16 q9, q5, q4 + vadd.s16 q5, q6, q10 + vsub.s16 q10, q6, q10 + vadd.s16 q6, q15, q13 + vadd.s16 q8, q12, q14 + vsub.s16 q3, q6, q3 + vsub.s16 q12, q12, q14 + vsub.s16 q3, q3, q1 + vsub.s16 q1, q9, q1 + vadd.s16 q2, q3, q2 + vsub.s16 q15, q8, q6 + vadd.s16 q1, q1, q2 + vadd.s16 q8, q8, q6 + vadd.s16 q14, q5, q3 + vsub.s16 q9, q5, q3 + vsub.s16 q13, q10, q2 + vpop {d8-d13} /* restore NEON registers */ + vadd.s16 q10, q10, q2 + vsub.s16 q11, q12, q1 + vadd.s16 q12, q12, q1 + /* Descale to 8-bit and range limit */ + vmov.u8 q0, #0x80 + vqshrn.s16 d16, q8, #5 + vqshrn.s16 d17, q9, #5 + vqshrn.s16 d18, q10, #5 + vqshrn.s16 d19, q11, #5 + vqshrn.s16 d20, q12, #5 + vqshrn.s16 d21, q13, #5 + vqshrn.s16 d22, q14, #5 + vqshrn.s16 d23, q15, #5 + vadd.u8 q8, q8, q0 + vadd.u8 q9, q9, q0 + vadd.u8 q10, q10, q0 + vadd.u8 q11, q11, q0 + /* Transpose the final 8-bit samples */ + vtrn.16 q8, q9 + vtrn.16 q10, q11 + vtrn.32 q8, q10 + vtrn.32 q9, q11 + vtrn.8 d16, d17 + vtrn.8 d18, d19 + /* Store results to the output buffer */ + ldmia OUTPUT_BUF!, {TMP1, TMP2} + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + vst1.8 {d16}, [TMP1] + vst1.8 {d17}, [TMP2] + ldmia OUTPUT_BUF!, {TMP1, TMP2} + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + vst1.8 {d18}, [TMP1] + vtrn.8 d20, d21 + vst1.8 {d19}, [TMP2] + ldmia OUTPUT_BUF, {TMP1, TMP2, TMP3, TMP4} + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + add TMP3, TMP3, OUTPUT_COL + add TMP4, TMP4, OUTPUT_COL + vst1.8 {d20}, [TMP1] + vtrn.8 d22, d23 + vst1.8 {d21}, [TMP2] + vst1.8 {d22}, [TMP3] + vst1.8 {d23}, [TMP4] + bx lr + + .unreq DCT_TABLE + .unreq COEF_BLOCK + .unreq OUTPUT_BUF + .unreq OUTPUT_COL + .unreq TMP1 + .unreq TMP2 + .unreq TMP3 + .unreq TMP4 + + +/*****************************************************************************/ + +/* + * jsimd_idct_4x4_neon + * + * This function contains inverse-DCT code for getting reduced-size + * 4x4 pixels output from an 8x8 DCT block. It uses the same calculations + * and produces exactly the same output as IJG's original 'jpeg_idct_4x4' + * function from jpeg-6b (jidctred.c). + * + * NOTE: jpeg-8 has an improved implementation of 4x4 inverse-DCT, which + * requires much less arithmetic operations and hence should be faster. + * The primary purpose of this particular NEON optimized function is + * bit exact compatibility with jpeg-6b. + * + * TODO: a bit better instructions scheduling can be achieved by expanding + * idct_helper/transpose_4x4 macros and reordering instructions, + * but readability will suffer somewhat. + */ + +#define CONST_BITS 13 + +#define FIX_0_211164243 (1730) /* FIX(0.211164243) */ +#define FIX_0_509795579 (4176) /* FIX(0.509795579) */ +#define FIX_0_601344887 (4926) /* FIX(0.601344887) */ +#define FIX_0_720959822 (5906) /* FIX(0.720959822) */ +#define FIX_0_765366865 (6270) /* FIX(0.765366865) */ +#define FIX_0_850430095 (6967) /* FIX(0.850430095) */ +#define FIX_0_899976223 (7373) /* FIX(0.899976223) */ +#define FIX_1_061594337 (8697) /* FIX(1.061594337) */ +#define FIX_1_272758580 (10426) /* FIX(1.272758580) */ +#define FIX_1_451774981 (11893) /* FIX(1.451774981) */ +#define FIX_1_847759065 (15137) /* FIX(1.847759065) */ +#define FIX_2_172734803 (17799) /* FIX(2.172734803) */ +#define FIX_2_562915447 (20995) /* FIX(2.562915447) */ +#define FIX_3_624509785 (29692) /* FIX(3.624509785) */ + +.balign 16 +jsimd_idct_4x4_neon_consts: + .short FIX_1_847759065 /* d0[0] */ + .short -FIX_0_765366865 /* d0[1] */ + .short -FIX_0_211164243 /* d0[2] */ + .short FIX_1_451774981 /* d0[3] */ + .short -FIX_2_172734803 /* d1[0] */ + .short FIX_1_061594337 /* d1[1] */ + .short -FIX_0_509795579 /* d1[2] */ + .short -FIX_0_601344887 /* d1[3] */ + .short FIX_0_899976223 /* d2[0] */ + .short FIX_2_562915447 /* d2[1] */ + .short 1 << (CONST_BITS+1) /* d2[2] */ + .short 0 /* d2[3] */ + +.macro idct_helper x4, x6, x8, x10, x12, x14, x16, shift, y26, y27, y28, y29 + vmull.s16 q14, \x4, d2[2] + vmlal.s16 q14, \x8, d0[0] + vmlal.s16 q14, \x14, d0[1] + + vmull.s16 q13, \x16, d1[2] + vmlal.s16 q13, \x12, d1[3] + vmlal.s16 q13, \x10, d2[0] + vmlal.s16 q13, \x6, d2[1] + + vmull.s16 q15, \x4, d2[2] + vmlsl.s16 q15, \x8, d0[0] + vmlsl.s16 q15, \x14, d0[1] + + vmull.s16 q12, \x16, d0[2] + vmlal.s16 q12, \x12, d0[3] + vmlal.s16 q12, \x10, d1[0] + vmlal.s16 q12, \x6, d1[1] + + vadd.s32 q10, q14, q13 + vsub.s32 q14, q14, q13 + +.if \shift > 16 + vrshr.s32 q10, q10, #\shift + vrshr.s32 q14, q14, #\shift + vmovn.s32 \y26, q10 + vmovn.s32 \y29, q14 +.else + vrshrn.s32 \y26, q10, #\shift + vrshrn.s32 \y29, q14, #\shift +.endif + + vadd.s32 q10, q15, q12 + vsub.s32 q15, q15, q12 + +.if \shift > 16 + vrshr.s32 q10, q10, #\shift + vrshr.s32 q15, q15, #\shift + vmovn.s32 \y27, q10 + vmovn.s32 \y28, q15 +.else + vrshrn.s32 \y27, q10, #\shift + vrshrn.s32 \y28, q15, #\shift +.endif + +.endm + +asm_function jsimd_idct_4x4_neon + + DCT_TABLE .req r0 + COEF_BLOCK .req r1 + OUTPUT_BUF .req r2 + OUTPUT_COL .req r3 + TMP1 .req r0 + TMP2 .req r1 + TMP3 .req r2 + TMP4 .req ip + + vpush {d8-d15} + + /* Load constants (d3 is just used for padding) */ + adr TMP4, jsimd_idct_4x4_neon_consts + vld1.16 {d0, d1, d2, d3}, [TMP4, :128] + + /* Load all COEF_BLOCK into NEON registers with the following allocation: + * 0 1 2 3 | 4 5 6 7 + * ---------+-------- + * 0 | d4 | d5 + * 1 | d6 | d7 + * 2 | d8 | d9 + * 3 | d10 | d11 + * 4 | - | - + * 5 | d12 | d13 + * 6 | d14 | d15 + * 7 | d16 | d17 + */ + vld1.16 {d4, d5, d6, d7}, [COEF_BLOCK, :128]! + vld1.16 {d8, d9, d10, d11}, [COEF_BLOCK, :128]! + add COEF_BLOCK, COEF_BLOCK, #16 + vld1.16 {d12, d13, d14, d15}, [COEF_BLOCK, :128]! + vld1.16 {d16, d17}, [COEF_BLOCK, :128]! + /* dequantize */ + vld1.16 {d18, d19, d20, d21}, [DCT_TABLE, :128]! + vmul.s16 q2, q2, q9 + vld1.16 {d22, d23, d24, d25}, [DCT_TABLE, :128]! + vmul.s16 q3, q3, q10 + vmul.s16 q4, q4, q11 + add DCT_TABLE, DCT_TABLE, #16 + vld1.16 {d26, d27, d28, d29}, [DCT_TABLE, :128]! + vmul.s16 q5, q5, q12 + vmul.s16 q6, q6, q13 + vld1.16 {d30, d31}, [DCT_TABLE, :128]! + vmul.s16 q7, q7, q14 + vmul.s16 q8, q8, q15 + + /* Pass 1 */ + idct_helper d4, d6, d8, d10, d12, d14, d16, 12, d4, d6, d8, d10 + transpose_4x4 d4, d6, d8, d10 + idct_helper d5, d7, d9, d11, d13, d15, d17, 12, d5, d7, d9, d11 + transpose_4x4 d5, d7, d9, d11 + + /* Pass 2 */ + idct_helper d4, d6, d8, d10, d7, d9, d11, 19, d26, d27, d28, d29 + transpose_4x4 d26, d27, d28, d29 + + /* Range limit */ + vmov.u16 q15, #0x80 + vadd.s16 q13, q13, q15 + vadd.s16 q14, q14, q15 + vqmovun.s16 d26, q13 + vqmovun.s16 d27, q14 + + /* Store results to the output buffer */ + ldmia OUTPUT_BUF, {TMP1, TMP2, TMP3, TMP4} + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + add TMP3, TMP3, OUTPUT_COL + add TMP4, TMP4, OUTPUT_COL + +#if defined(__ARMEL__) && !RESPECT_STRICT_ALIGNMENT + /* We can use much less instructions on little endian systems if the + * OS kernel is not configured to trap unaligned memory accesses + */ + vst1.32 {d26[0]}, [TMP1]! + vst1.32 {d27[0]}, [TMP3]! + vst1.32 {d26[1]}, [TMP2]! + vst1.32 {d27[1]}, [TMP4]! +#else + vst1.8 {d26[0]}, [TMP1]! + vst1.8 {d27[0]}, [TMP3]! + vst1.8 {d26[1]}, [TMP1]! + vst1.8 {d27[1]}, [TMP3]! + vst1.8 {d26[2]}, [TMP1]! + vst1.8 {d27[2]}, [TMP3]! + vst1.8 {d26[3]}, [TMP1]! + vst1.8 {d27[3]}, [TMP3]! + + vst1.8 {d26[4]}, [TMP2]! + vst1.8 {d27[4]}, [TMP4]! + vst1.8 {d26[5]}, [TMP2]! + vst1.8 {d27[5]}, [TMP4]! + vst1.8 {d26[6]}, [TMP2]! + vst1.8 {d27[6]}, [TMP4]! + vst1.8 {d26[7]}, [TMP2]! + vst1.8 {d27[7]}, [TMP4]! +#endif + + vpop {d8-d15} + bx lr + + .unreq DCT_TABLE + .unreq COEF_BLOCK + .unreq OUTPUT_BUF + .unreq OUTPUT_COL + .unreq TMP1 + .unreq TMP2 + .unreq TMP3 + .unreq TMP4 + +.purgem idct_helper + + +/*****************************************************************************/ + +/* + * jsimd_idct_2x2_neon + * + * This function contains inverse-DCT code for getting reduced-size + * 2x2 pixels output from an 8x8 DCT block. It uses the same calculations + * and produces exactly the same output as IJG's original 'jpeg_idct_2x2' + * function from jpeg-6b (jidctred.c). + * + * NOTE: jpeg-8 has an improved implementation of 2x2 inverse-DCT, which + * requires much less arithmetic operations and hence should be faster. + * The primary purpose of this particular NEON optimized function is + * bit exact compatibility with jpeg-6b. + */ + +.balign 8 +jsimd_idct_2x2_neon_consts: + .short -FIX_0_720959822 /* d0[0] */ + .short FIX_0_850430095 /* d0[1] */ + .short -FIX_1_272758580 /* d0[2] */ + .short FIX_3_624509785 /* d0[3] */ + +.macro idct_helper x4, x6, x10, x12, x16, shift, y26, y27 + vshll.s16 q14, \x4, #15 + vmull.s16 q13, \x6, d0[3] + vmlal.s16 q13, \x10, d0[2] + vmlal.s16 q13, \x12, d0[1] + vmlal.s16 q13, \x16, d0[0] + + vadd.s32 q10, q14, q13 + vsub.s32 q14, q14, q13 + +.if \shift > 16 + vrshr.s32 q10, q10, #\shift + vrshr.s32 q14, q14, #\shift + vmovn.s32 \y26, q10 + vmovn.s32 \y27, q14 +.else + vrshrn.s32 \y26, q10, #\shift + vrshrn.s32 \y27, q14, #\shift +.endif + +.endm + +asm_function jsimd_idct_2x2_neon + + DCT_TABLE .req r0 + COEF_BLOCK .req r1 + OUTPUT_BUF .req r2 + OUTPUT_COL .req r3 + TMP1 .req r0 + TMP2 .req ip + + vpush {d8-d15} + + /* Load constants */ + adr TMP2, jsimd_idct_2x2_neon_consts + vld1.16 {d0}, [TMP2, :64] + + /* Load all COEF_BLOCK into NEON registers with the following allocation: + * 0 1 2 3 | 4 5 6 7 + * ---------+-------- + * 0 | d4 | d5 + * 1 | d6 | d7 + * 2 | - | - + * 3 | d10 | d11 + * 4 | - | - + * 5 | d12 | d13 + * 6 | - | - + * 7 | d16 | d17 + */ + vld1.16 {d4, d5, d6, d7}, [COEF_BLOCK, :128]! + add COEF_BLOCK, COEF_BLOCK, #16 + vld1.16 {d10, d11}, [COEF_BLOCK, :128]! + add COEF_BLOCK, COEF_BLOCK, #16 + vld1.16 {d12, d13}, [COEF_BLOCK, :128]! + add COEF_BLOCK, COEF_BLOCK, #16 + vld1.16 {d16, d17}, [COEF_BLOCK, :128]! + /* Dequantize */ + vld1.16 {d18, d19, d20, d21}, [DCT_TABLE, :128]! + vmul.s16 q2, q2, q9 + vmul.s16 q3, q3, q10 + add DCT_TABLE, DCT_TABLE, #16 + vld1.16 {d24, d25}, [DCT_TABLE, :128]! + vmul.s16 q5, q5, q12 + add DCT_TABLE, DCT_TABLE, #16 + vld1.16 {d26, d27}, [DCT_TABLE, :128]! + vmul.s16 q6, q6, q13 + add DCT_TABLE, DCT_TABLE, #16 + vld1.16 {d30, d31}, [DCT_TABLE, :128]! + vmul.s16 q8, q8, q15 + + /* Pass 1 */ +#if 0 + idct_helper d4, d6, d10, d12, d16, 13, d4, d6 + transpose_4x4 d4, d6, d8, d10 + idct_helper d5, d7, d11, d13, d17, 13, d5, d7 + transpose_4x4 d5, d7, d9, d11 +#else + vmull.s16 q13, d6, d0[3] + vmlal.s16 q13, d10, d0[2] + vmlal.s16 q13, d12, d0[1] + vmlal.s16 q13, d16, d0[0] + vmull.s16 q12, d7, d0[3] + vmlal.s16 q12, d11, d0[2] + vmlal.s16 q12, d13, d0[1] + vmlal.s16 q12, d17, d0[0] + vshll.s16 q14, d4, #15 + vshll.s16 q15, d5, #15 + vadd.s32 q10, q14, q13 + vsub.s32 q14, q14, q13 + vrshrn.s32 d4, q10, #13 + vrshrn.s32 d6, q14, #13 + vadd.s32 q10, q15, q12 + vsub.s32 q14, q15, q12 + vrshrn.s32 d5, q10, #13 + vrshrn.s32 d7, q14, #13 + vtrn.16 q2, q3 + vtrn.32 q3, q5 +#endif + + /* Pass 2 */ + idct_helper d4, d6, d10, d7, d11, 20, d26, d27 + + /* Range limit */ + vmov.u16 q15, #0x80 + vadd.s16 q13, q13, q15 + vqmovun.s16 d26, q13 + vqmovun.s16 d27, q13 + + /* Store results to the output buffer */ + ldmia OUTPUT_BUF, {TMP1, TMP2} + add TMP1, TMP1, OUTPUT_COL + add TMP2, TMP2, OUTPUT_COL + + vst1.8 {d26[0]}, [TMP1]! + vst1.8 {d27[4]}, [TMP1]! + vst1.8 {d26[1]}, [TMP2]! + vst1.8 {d27[5]}, [TMP2]! + + vpop {d8-d15} + bx lr + + .unreq DCT_TABLE + .unreq COEF_BLOCK + .unreq OUTPUT_BUF + .unreq OUTPUT_COL + .unreq TMP1 + .unreq TMP2 + +.purgem idct_helper + + +/*****************************************************************************/ + +/* + * jsimd_ycc_extrgb_convert_neon + * jsimd_ycc_extbgr_convert_neon + * jsimd_ycc_extrgbx_convert_neon + * jsimd_ycc_extbgrx_convert_neon + * jsimd_ycc_extxbgr_convert_neon + * jsimd_ycc_extxrgb_convert_neon + * + * Colorspace conversion YCbCr -> RGB + */ + + +.macro do_load size + .if \size == 8 + vld1.8 {d4}, [U, :64]! + vld1.8 {d5}, [V, :64]! + vld1.8 {d0}, [Y, :64]! + pld [U, #64] + pld [V, #64] + pld [Y, #64] + .elseif \size == 4 + vld1.8 {d4[0]}, [U]! + vld1.8 {d4[1]}, [U]! + vld1.8 {d4[2]}, [U]! + vld1.8 {d4[3]}, [U]! + vld1.8 {d5[0]}, [V]! + vld1.8 {d5[1]}, [V]! + vld1.8 {d5[2]}, [V]! + vld1.8 {d5[3]}, [V]! + vld1.8 {d0[0]}, [Y]! + vld1.8 {d0[1]}, [Y]! + vld1.8 {d0[2]}, [Y]! + vld1.8 {d0[3]}, [Y]! + .elseif \size == 2 + vld1.8 {d4[4]}, [U]! + vld1.8 {d4[5]}, [U]! + vld1.8 {d5[4]}, [V]! + vld1.8 {d5[5]}, [V]! + vld1.8 {d0[4]}, [Y]! + vld1.8 {d0[5]}, [Y]! + .elseif \size == 1 + vld1.8 {d4[6]}, [U]! + vld1.8 {d5[6]}, [V]! + vld1.8 {d0[6]}, [Y]! + .else + .error unsupported macroblock size + .endif +.endm + +.macro do_store bpp, size + .if \bpp == 24 + .if \size == 8 + vst3.8 {d10, d11, d12}, [RGB]! + .elseif \size == 4 + vst3.8 {d10[0], d11[0], d12[0]}, [RGB]! + vst3.8 {d10[1], d11[1], d12[1]}, [RGB]! + vst3.8 {d10[2], d11[2], d12[2]}, [RGB]! + vst3.8 {d10[3], d11[3], d12[3]}, [RGB]! + .elseif \size == 2 + vst3.8 {d10[4], d11[4], d12[4]}, [RGB]! + vst3.8 {d10[5], d11[5], d12[5]}, [RGB]! + .elseif \size == 1 + vst3.8 {d10[6], d11[6], d12[6]}, [RGB]! + .else + .error unsupported macroblock size + .endif + .elseif \bpp == 32 + .if \size == 8 + vst4.8 {d10, d11, d12, d13}, [RGB]! + .elseif \size == 4 + vst4.8 {d10[0], d11[0], d12[0], d13[0]}, [RGB]! + vst4.8 {d10[1], d11[1], d12[1], d13[1]}, [RGB]! + vst4.8 {d10[2], d11[2], d12[2], d13[2]}, [RGB]! + vst4.8 {d10[3], d11[3], d12[3], d13[3]}, [RGB]! + .elseif \size == 2 + vst4.8 {d10[4], d11[4], d12[4], d13[4]}, [RGB]! + vst4.8 {d10[5], d11[5], d12[5], d13[5]}, [RGB]! + .elseif \size == 1 + vst4.8 {d10[6], d11[6], d12[6], d13[6]}, [RGB]! + .else + .error unsupported macroblock size + .endif + .elseif \bpp == 16 + .if \size == 8 + vst1.16 {q15}, [RGB]! + .elseif \size == 4 + vst1.16 {d30}, [RGB]! + .elseif \size == 2 + vst1.16 {d31[0]}, [RGB]! + vst1.16 {d31[1]}, [RGB]! + .elseif \size == 1 + vst1.16 {d31[2]}, [RGB]! + .else + .error unsupported macroblock size + .endif + .else + .error unsupported bpp + .endif +.endm + +.macro generate_jsimd_ycc_rgb_convert_neon colorid, bpp, r_offs, g_offs, b_offs + +/* + * 2 stage pipelined YCbCr->RGB conversion + */ + +.macro do_yuv_to_rgb_stage1 + vaddw.u8 q3, q1, d4 /* q3 = u - 128 */ + vaddw.u8 q4, q1, d5 /* q2 = v - 128 */ + vmull.s16 q10, d6, d1[1] /* multiply by -11277 */ + vmlal.s16 q10, d8, d1[2] /* multiply by -23401 */ + vmull.s16 q11, d7, d1[1] /* multiply by -11277 */ + vmlal.s16 q11, d9, d1[2] /* multiply by -23401 */ + vmull.s16 q12, d8, d1[0] /* multiply by 22971 */ + vmull.s16 q13, d9, d1[0] /* multiply by 22971 */ + vmull.s16 q14, d6, d1[3] /* multiply by 29033 */ + vmull.s16 q15, d7, d1[3] /* multiply by 29033 */ +.endm + +.macro do_yuv_to_rgb_stage2 + vrshrn.s32 d20, q10, #15 + vrshrn.s32 d21, q11, #15 + vrshrn.s32 d24, q12, #14 + vrshrn.s32 d25, q13, #14 + vrshrn.s32 d28, q14, #14 + vrshrn.s32 d29, q15, #14 + vaddw.u8 q11, q10, d0 + vaddw.u8 q12, q12, d0 + vaddw.u8 q14, q14, d0 +.if \bpp != 16 + vqmovun.s16 d1\g_offs, q11 + vqmovun.s16 d1\r_offs, q12 + vqmovun.s16 d1\b_offs, q14 +.else /* rgb565 */ + vqshlu.s16 q13, q11, #8 + vqshlu.s16 q15, q12, #8 + vqshlu.s16 q14, q14, #8 + vsri.u16 q15, q13, #5 + vsri.u16 q15, q14, #11 +.endif +.endm + +.macro do_yuv_to_rgb_stage2_store_load_stage1 + /* "do_yuv_to_rgb_stage2" and "store" */ + vrshrn.s32 d20, q10, #15 + /* "load" and "do_yuv_to_rgb_stage1" */ + pld [U, #64] + vrshrn.s32 d21, q11, #15 + pld [V, #64] + vrshrn.s32 d24, q12, #14 + vrshrn.s32 d25, q13, #14 + vld1.8 {d4}, [U, :64]! + vrshrn.s32 d28, q14, #14 + vld1.8 {d5}, [V, :64]! + vrshrn.s32 d29, q15, #14 + vaddw.u8 q3, q1, d4 /* q3 = u - 128 */ + vaddw.u8 q4, q1, d5 /* q2 = v - 128 */ + vaddw.u8 q11, q10, d0 + vmull.s16 q10, d6, d1[1] /* multiply by -11277 */ + vmlal.s16 q10, d8, d1[2] /* multiply by -23401 */ + vaddw.u8 q12, q12, d0 + vaddw.u8 q14, q14, d0 +.if \bpp != 16 /**************** rgb24/rgb32 *********************************/ + vqmovun.s16 d1\g_offs, q11 + pld [Y, #64] + vqmovun.s16 d1\r_offs, q12 + vld1.8 {d0}, [Y, :64]! + vqmovun.s16 d1\b_offs, q14 + vmull.s16 q11, d7, d1[1] /* multiply by -11277 */ + vmlal.s16 q11, d9, d1[2] /* multiply by -23401 */ + do_store \bpp, 8 + vmull.s16 q12, d8, d1[0] /* multiply by 22971 */ + vmull.s16 q13, d9, d1[0] /* multiply by 22971 */ + vmull.s16 q14, d6, d1[3] /* multiply by 29033 */ + vmull.s16 q15, d7, d1[3] /* multiply by 29033 */ +.else /**************************** rgb565 ***********************************/ + vqshlu.s16 q13, q11, #8 + pld [Y, #64] + vqshlu.s16 q15, q12, #8 + vqshlu.s16 q14, q14, #8 + vld1.8 {d0}, [Y, :64]! + vmull.s16 q11, d7, d1[1] + vmlal.s16 q11, d9, d1[2] + vsri.u16 q15, q13, #5 + vmull.s16 q12, d8, d1[0] + vsri.u16 q15, q14, #11 + vmull.s16 q13, d9, d1[0] + vmull.s16 q14, d6, d1[3] + do_store \bpp, 8 + vmull.s16 q15, d7, d1[3] +.endif +.endm + +.macro do_yuv_to_rgb + do_yuv_to_rgb_stage1 + do_yuv_to_rgb_stage2 +.endm + +/* Apple gas crashes on adrl, work around that by using adr. + * But this requires a copy of these constants for each function. + */ + +.balign 16 +jsimd_ycc_\colorid\()_neon_consts: + .short 0, 0, 0, 0 + .short 22971, -11277, -23401, 29033 + .short -128, -128, -128, -128 + .short -128, -128, -128, -128 + +asm_function jsimd_ycc_\colorid\()_convert_neon + OUTPUT_WIDTH .req r0 + INPUT_BUF .req r1 + INPUT_ROW .req r2 + OUTPUT_BUF .req r3 + NUM_ROWS .req r4 + + INPUT_BUF0 .req r5 + INPUT_BUF1 .req r6 + INPUT_BUF2 .req INPUT_BUF + + RGB .req r7 + Y .req r8 + U .req r9 + V .req r10 + N .req ip + + /* Load constants to d1, d2, d3 (d0 is just used for padding) */ + adr ip, jsimd_ycc_\colorid\()_neon_consts + vld1.16 {d0, d1, d2, d3}, [ip, :128] + + /* Save ARM registers and handle input arguments */ + push {r4, r5, r6, r7, r8, r9, r10, lr} + ldr NUM_ROWS, [sp, #(4 * 8)] + ldr INPUT_BUF0, [INPUT_BUF] + ldr INPUT_BUF1, [INPUT_BUF, #4] + ldr INPUT_BUF2, [INPUT_BUF, #8] + .unreq INPUT_BUF + + /* Save NEON registers */ + vpush {d8-d15} + + /* Initially set d10, d11, d12, d13 to 0xFF */ + vmov.u8 q5, #255 + vmov.u8 q6, #255 + + /* Outer loop over scanlines */ + cmp NUM_ROWS, #1 + blt 9f +0: + ldr Y, [INPUT_BUF0, INPUT_ROW, lsl #2] + ldr U, [INPUT_BUF1, INPUT_ROW, lsl #2] + mov N, OUTPUT_WIDTH + ldr V, [INPUT_BUF2, INPUT_ROW, lsl #2] + add INPUT_ROW, INPUT_ROW, #1 + ldr RGB, [OUTPUT_BUF], #4 + + /* Inner loop over pixels */ + subs N, N, #8 + blt 3f + do_load 8 + do_yuv_to_rgb_stage1 + subs N, N, #8 + blt 2f +1: + do_yuv_to_rgb_stage2_store_load_stage1 + subs N, N, #8 + bge 1b +2: + do_yuv_to_rgb_stage2 + do_store \bpp, 8 + tst N, #7 + beq 8f +3: + tst N, #4 + beq 3f + do_load 4 +3: + tst N, #2 + beq 4f + do_load 2 +4: + tst N, #1 + beq 5f + do_load 1 +5: + do_yuv_to_rgb + tst N, #4 + beq 6f + do_store \bpp, 4 +6: + tst N, #2 + beq 7f + do_store \bpp, 2 +7: + tst N, #1 + beq 8f + do_store \bpp, 1 +8: + subs NUM_ROWS, NUM_ROWS, #1 + bgt 0b +9: + /* Restore all registers and return */ + vpop {d8-d15} + pop {r4, r5, r6, r7, r8, r9, r10, pc} + + .unreq OUTPUT_WIDTH + .unreq INPUT_ROW + .unreq OUTPUT_BUF + .unreq NUM_ROWS + .unreq INPUT_BUF0 + .unreq INPUT_BUF1 + .unreq INPUT_BUF2 + .unreq RGB + .unreq Y + .unreq U + .unreq V + .unreq N + +.purgem do_yuv_to_rgb +.purgem do_yuv_to_rgb_stage1 +.purgem do_yuv_to_rgb_stage2 +.purgem do_yuv_to_rgb_stage2_store_load_stage1 + +.endm + +/*--------------------------------- id ----- bpp R G B */ +generate_jsimd_ycc_rgb_convert_neon extrgb, 24, 0, 1, 2 +generate_jsimd_ycc_rgb_convert_neon extbgr, 24, 2, 1, 0 +generate_jsimd_ycc_rgb_convert_neon extrgbx, 32, 0, 1, 2 +generate_jsimd_ycc_rgb_convert_neon extbgrx, 32, 2, 1, 0 +generate_jsimd_ycc_rgb_convert_neon extxbgr, 32, 3, 2, 1 +generate_jsimd_ycc_rgb_convert_neon extxrgb, 32, 1, 2, 3 +generate_jsimd_ycc_rgb_convert_neon rgb565, 16, 0, 0, 0 + +.purgem do_load +.purgem do_store + + +/*****************************************************************************/ + +/* + * jsimd_extrgb_ycc_convert_neon + * jsimd_extbgr_ycc_convert_neon + * jsimd_extrgbx_ycc_convert_neon + * jsimd_extbgrx_ycc_convert_neon + * jsimd_extxbgr_ycc_convert_neon + * jsimd_extxrgb_ycc_convert_neon + * + * Colorspace conversion RGB -> YCbCr + */ + +.macro do_store size + .if \size == 8 + vst1.8 {d20}, [Y]! + vst1.8 {d21}, [U]! + vst1.8 {d22}, [V]! + .elseif \size == 4 + vst1.8 {d20[0]}, [Y]! + vst1.8 {d20[1]}, [Y]! + vst1.8 {d20[2]}, [Y]! + vst1.8 {d20[3]}, [Y]! + vst1.8 {d21[0]}, [U]! + vst1.8 {d21[1]}, [U]! + vst1.8 {d21[2]}, [U]! + vst1.8 {d21[3]}, [U]! + vst1.8 {d22[0]}, [V]! + vst1.8 {d22[1]}, [V]! + vst1.8 {d22[2]}, [V]! + vst1.8 {d22[3]}, [V]! + .elseif \size == 2 + vst1.8 {d20[4]}, [Y]! + vst1.8 {d20[5]}, [Y]! + vst1.8 {d21[4]}, [U]! + vst1.8 {d21[5]}, [U]! + vst1.8 {d22[4]}, [V]! + vst1.8 {d22[5]}, [V]! + .elseif \size == 1 + vst1.8 {d20[6]}, [Y]! + vst1.8 {d21[6]}, [U]! + vst1.8 {d22[6]}, [V]! + .else + .error unsupported macroblock size + .endif +.endm + +.macro do_load bpp, size + .if \bpp == 24 + .if \size == 8 + vld3.8 {d10, d11, d12}, [RGB]! + pld [RGB, #128] + .elseif \size == 4 + vld3.8 {d10[0], d11[0], d12[0]}, [RGB]! + vld3.8 {d10[1], d11[1], d12[1]}, [RGB]! + vld3.8 {d10[2], d11[2], d12[2]}, [RGB]! + vld3.8 {d10[3], d11[3], d12[3]}, [RGB]! + .elseif \size == 2 + vld3.8 {d10[4], d11[4], d12[4]}, [RGB]! + vld3.8 {d10[5], d11[5], d12[5]}, [RGB]! + .elseif \size == 1 + vld3.8 {d10[6], d11[6], d12[6]}, [RGB]! + .else + .error unsupported macroblock size + .endif + .elseif \bpp == 32 + .if \size == 8 + vld4.8 {d10, d11, d12, d13}, [RGB]! + pld [RGB, #128] + .elseif \size == 4 + vld4.8 {d10[0], d11[0], d12[0], d13[0]}, [RGB]! + vld4.8 {d10[1], d11[1], d12[1], d13[1]}, [RGB]! + vld4.8 {d10[2], d11[2], d12[2], d13[2]}, [RGB]! + vld4.8 {d10[3], d11[3], d12[3], d13[3]}, [RGB]! + .elseif \size == 2 + vld4.8 {d10[4], d11[4], d12[4], d13[4]}, [RGB]! + vld4.8 {d10[5], d11[5], d12[5], d13[5]}, [RGB]! + .elseif \size == 1 + vld4.8 {d10[6], d11[6], d12[6], d13[6]}, [RGB]! + .else + .error unsupported macroblock size + .endif + .else + .error unsupported bpp + .endif +.endm + +.macro generate_jsimd_rgb_ycc_convert_neon colorid, bpp, r_offs, g_offs, b_offs + +/* + * 2 stage pipelined RGB->YCbCr conversion + */ + +.macro do_rgb_to_yuv_stage1 + vmovl.u8 q2, d1\r_offs /* r = { d4, d5 } */ + vmovl.u8 q3, d1\g_offs /* g = { d6, d7 } */ + vmovl.u8 q4, d1\b_offs /* b = { d8, d9 } */ + vmull.u16 q7, d4, d0[0] + vmlal.u16 q7, d6, d0[1] + vmlal.u16 q7, d8, d0[2] + vmull.u16 q8, d5, d0[0] + vmlal.u16 q8, d7, d0[1] + vmlal.u16 q8, d9, d0[2] + vrev64.32 q9, q1 + vrev64.32 q13, q1 + vmlsl.u16 q9, d4, d0[3] + vmlsl.u16 q9, d6, d1[0] + vmlal.u16 q9, d8, d1[1] + vmlsl.u16 q13, d5, d0[3] + vmlsl.u16 q13, d7, d1[0] + vmlal.u16 q13, d9, d1[1] + vrev64.32 q14, q1 + vrev64.32 q15, q1 + vmlal.u16 q14, d4, d1[1] + vmlsl.u16 q14, d6, d1[2] + vmlsl.u16 q14, d8, d1[3] + vmlal.u16 q15, d5, d1[1] + vmlsl.u16 q15, d7, d1[2] + vmlsl.u16 q15, d9, d1[3] +.endm + +.macro do_rgb_to_yuv_stage2 + vrshrn.u32 d20, q7, #16 + vrshrn.u32 d21, q8, #16 + vshrn.u32 d22, q9, #16 + vshrn.u32 d23, q13, #16 + vshrn.u32 d24, q14, #16 + vshrn.u32 d25, q15, #16 + vmovn.u16 d20, q10 /* d20 = y */ + vmovn.u16 d21, q11 /* d21 = u */ + vmovn.u16 d22, q12 /* d22 = v */ +.endm + +.macro do_rgb_to_yuv + do_rgb_to_yuv_stage1 + do_rgb_to_yuv_stage2 +.endm + +.macro do_rgb_to_yuv_stage2_store_load_stage1 + vrshrn.u32 d20, q7, #16 + vrshrn.u32 d21, q8, #16 + vshrn.u32 d22, q9, #16 + vrev64.32 q9, q1 + vshrn.u32 d23, q13, #16 + vrev64.32 q13, q1 + vshrn.u32 d24, q14, #16 + vshrn.u32 d25, q15, #16 + do_load \bpp, 8 + vmovn.u16 d20, q10 /* d20 = y */ + vmovl.u8 q2, d1\r_offs /* r = { d4, d5 } */ + vmovn.u16 d21, q11 /* d21 = u */ + vmovl.u8 q3, d1\g_offs /* g = { d6, d7 } */ + vmovn.u16 d22, q12 /* d22 = v */ + vmovl.u8 q4, d1\b_offs /* b = { d8, d9 } */ + vmull.u16 q7, d4, d0[0] + vmlal.u16 q7, d6, d0[1] + vmlal.u16 q7, d8, d0[2] + vst1.8 {d20}, [Y]! + vmull.u16 q8, d5, d0[0] + vmlal.u16 q8, d7, d0[1] + vmlal.u16 q8, d9, d0[2] + vmlsl.u16 q9, d4, d0[3] + vmlsl.u16 q9, d6, d1[0] + vmlal.u16 q9, d8, d1[1] + vst1.8 {d21}, [U]! + vmlsl.u16 q13, d5, d0[3] + vmlsl.u16 q13, d7, d1[0] + vmlal.u16 q13, d9, d1[1] + vrev64.32 q14, q1 + vrev64.32 q15, q1 + vmlal.u16 q14, d4, d1[1] + vmlsl.u16 q14, d6, d1[2] + vmlsl.u16 q14, d8, d1[3] + vst1.8 {d22}, [V]! + vmlal.u16 q15, d5, d1[1] + vmlsl.u16 q15, d7, d1[2] + vmlsl.u16 q15, d9, d1[3] +.endm + +.balign 16 +jsimd_\colorid\()_ycc_neon_consts: + .short 19595, 38470, 7471, 11059 + .short 21709, 32768, 27439, 5329 + .short 32767, 128, 32767, 128 + .short 32767, 128, 32767, 128 + +asm_function jsimd_\colorid\()_ycc_convert_neon + OUTPUT_WIDTH .req r0 + INPUT_BUF .req r1 + OUTPUT_BUF .req r2 + OUTPUT_ROW .req r3 + NUM_ROWS .req r4 + + OUTPUT_BUF0 .req r5 + OUTPUT_BUF1 .req r6 + OUTPUT_BUF2 .req OUTPUT_BUF + + RGB .req r7 + Y .req r8 + U .req r9 + V .req r10 + N .req ip + + /* Load constants to d0, d1, d2, d3 */ + adr ip, jsimd_\colorid\()_ycc_neon_consts + vld1.16 {d0, d1, d2, d3}, [ip, :128] + + /* Save ARM registers and handle input arguments */ + push {r4, r5, r6, r7, r8, r9, r10, lr} + ldr NUM_ROWS, [sp, #(4 * 8)] + ldr OUTPUT_BUF0, [OUTPUT_BUF] + ldr OUTPUT_BUF1, [OUTPUT_BUF, #4] + ldr OUTPUT_BUF2, [OUTPUT_BUF, #8] + .unreq OUTPUT_BUF + + /* Save NEON registers */ + vpush {d8-d15} + + /* Outer loop over scanlines */ + cmp NUM_ROWS, #1 + blt 9f +0: + ldr Y, [OUTPUT_BUF0, OUTPUT_ROW, lsl #2] + ldr U, [OUTPUT_BUF1, OUTPUT_ROW, lsl #2] + mov N, OUTPUT_WIDTH + ldr V, [OUTPUT_BUF2, OUTPUT_ROW, lsl #2] + add OUTPUT_ROW, OUTPUT_ROW, #1 + ldr RGB, [INPUT_BUF], #4 + + /* Inner loop over pixels */ + subs N, N, #8 + blt 3f + do_load \bpp, 8 + do_rgb_to_yuv_stage1 + subs N, N, #8 + blt 2f +1: + do_rgb_to_yuv_stage2_store_load_stage1 + subs N, N, #8 + bge 1b +2: + do_rgb_to_yuv_stage2 + do_store 8 + tst N, #7 + beq 8f +3: + tst N, #4 + beq 3f + do_load \bpp, 4 +3: + tst N, #2 + beq 4f + do_load \bpp, 2 +4: + tst N, #1 + beq 5f + do_load \bpp, 1 +5: + do_rgb_to_yuv + tst N, #4 + beq 6f + do_store 4 +6: + tst N, #2 + beq 7f + do_store 2 +7: + tst N, #1 + beq 8f + do_store 1 +8: + subs NUM_ROWS, NUM_ROWS, #1 + bgt 0b +9: + /* Restore all registers and return */ + vpop {d8-d15} + pop {r4, r5, r6, r7, r8, r9, r10, pc} + + .unreq OUTPUT_WIDTH + .unreq OUTPUT_ROW + .unreq INPUT_BUF + .unreq NUM_ROWS + .unreq OUTPUT_BUF0 + .unreq OUTPUT_BUF1 + .unreq OUTPUT_BUF2 + .unreq RGB + .unreq Y + .unreq U + .unreq V + .unreq N + +.purgem do_rgb_to_yuv +.purgem do_rgb_to_yuv_stage1 +.purgem do_rgb_to_yuv_stage2 +.purgem do_rgb_to_yuv_stage2_store_load_stage1 + +.endm + +/*--------------------------------- id ----- bpp R G B */ +generate_jsimd_rgb_ycc_convert_neon extrgb, 24, 0, 1, 2 +generate_jsimd_rgb_ycc_convert_neon extbgr, 24, 2, 1, 0 +generate_jsimd_rgb_ycc_convert_neon extrgbx, 32, 0, 1, 2 +generate_jsimd_rgb_ycc_convert_neon extbgrx, 32, 2, 1, 0 +generate_jsimd_rgb_ycc_convert_neon extxbgr, 32, 3, 2, 1 +generate_jsimd_rgb_ycc_convert_neon extxrgb, 32, 1, 2, 3 + +.purgem do_load +.purgem do_store + + +/*****************************************************************************/ + +/* + * Load data into workspace, applying unsigned->signed conversion + * + * TODO: can be combined with 'jsimd_fdct_ifast_neon' to get + * rid of VST1.16 instructions + */ + +asm_function jsimd_convsamp_neon + SAMPLE_DATA .req r0 + START_COL .req r1 + WORKSPACE .req r2 + TMP1 .req r3 + TMP2 .req r4 + TMP3 .req r5 + TMP4 .req ip + + push {r4, r5} + vmov.u8 d0, #128 + + ldmia SAMPLE_DATA!, {TMP1, TMP2, TMP3, TMP4} + add TMP1, TMP1, START_COL + add TMP2, TMP2, START_COL + add TMP3, TMP3, START_COL + add TMP4, TMP4, START_COL + vld1.8 {d16}, [TMP1] + vsubl.u8 q8, d16, d0 + vld1.8 {d18}, [TMP2] + vsubl.u8 q9, d18, d0 + vld1.8 {d20}, [TMP3] + vsubl.u8 q10, d20, d0 + vld1.8 {d22}, [TMP4] + ldmia SAMPLE_DATA!, {TMP1, TMP2, TMP3, TMP4} + vsubl.u8 q11, d22, d0 + vst1.16 {d16, d17, d18, d19}, [WORKSPACE, :128]! + add TMP1, TMP1, START_COL + add TMP2, TMP2, START_COL + vst1.16 {d20, d21, d22, d23}, [WORKSPACE, :128]! + add TMP3, TMP3, START_COL + add TMP4, TMP4, START_COL + vld1.8 {d24}, [TMP1] + vsubl.u8 q12, d24, d0 + vld1.8 {d26}, [TMP2] + vsubl.u8 q13, d26, d0 + vld1.8 {d28}, [TMP3] + vsubl.u8 q14, d28, d0 + vld1.8 {d30}, [TMP4] + vsubl.u8 q15, d30, d0 + vst1.16 {d24, d25, d26, d27}, [WORKSPACE, :128]! + vst1.16 {d28, d29, d30, d31}, [WORKSPACE, :128]! + pop {r4, r5} + bx lr + + .unreq SAMPLE_DATA + .unreq START_COL + .unreq WORKSPACE + .unreq TMP1 + .unreq TMP2 + .unreq TMP3 + .unreq TMP4 + + +/*****************************************************************************/ + +/* + * jsimd_fdct_ifast_neon + * + * This function contains a fast, not so accurate integer implementation of + * the forward DCT (Discrete Cosine Transform). It uses the same calculations + * and produces exactly the same output as IJG's original 'jpeg_fdct_ifast' + * function from jfdctfst.c + * + * TODO: can be combined with 'jsimd_convsamp_neon' to get + * rid of a bunch of VLD1.16 instructions + */ + +#define XFIX_0_382683433 d0[0] +#define XFIX_0_541196100 d0[1] +#define XFIX_0_707106781 d0[2] +#define XFIX_1_306562965 d0[3] + +.balign 16 +jsimd_fdct_ifast_neon_consts: + .short (98 * 128) /* XFIX_0_382683433 */ + .short (139 * 128) /* XFIX_0_541196100 */ + .short (181 * 128) /* XFIX_0_707106781 */ + .short (334 * 128 - 256 * 128) /* XFIX_1_306562965 */ + +asm_function jsimd_fdct_ifast_neon + + DATA .req r0 + TMP .req ip + + vpush {d8-d15} + + /* Load constants */ + adr TMP, jsimd_fdct_ifast_neon_consts + vld1.16 {d0}, [TMP, :64] + + /* Load all DATA into NEON registers with the following allocation: + * 0 1 2 3 | 4 5 6 7 + * ---------+-------- + * 0 | d16 | d17 | q8 + * 1 | d18 | d19 | q9 + * 2 | d20 | d21 | q10 + * 3 | d22 | d23 | q11 + * 4 | d24 | d25 | q12 + * 5 | d26 | d27 | q13 + * 6 | d28 | d29 | q14 + * 7 | d30 | d31 | q15 + */ + + vld1.16 {d16, d17, d18, d19}, [DATA, :128]! + vld1.16 {d20, d21, d22, d23}, [DATA, :128]! + vld1.16 {d24, d25, d26, d27}, [DATA, :128]! + vld1.16 {d28, d29, d30, d31}, [DATA, :128] + sub DATA, DATA, #(128 - 32) + + mov TMP, #2 +1: + /* Transpose */ + vtrn.16 q12, q13 + vtrn.16 q10, q11 + vtrn.16 q8, q9 + vtrn.16 q14, q15 + vtrn.32 q9, q11 + vtrn.32 q13, q15 + vtrn.32 q8, q10 + vtrn.32 q12, q14 + vswp d30, d23 + vswp d24, d17 + vswp d26, d19 + /* 1-D FDCT */ + vadd.s16 q2, q11, q12 + vswp d28, d21 + vsub.s16 q12, q11, q12 + vsub.s16 q6, q10, q13 + vadd.s16 q10, q10, q13 + vsub.s16 q7, q9, q14 + vadd.s16 q9, q9, q14 + vsub.s16 q1, q8, q15 + vadd.s16 q8, q8, q15 + vsub.s16 q4, q9, q10 + vsub.s16 q5, q8, q2 + vadd.s16 q3, q9, q10 + vadd.s16 q4, q4, q5 + vadd.s16 q2, q8, q2 + vqdmulh.s16 q4, q4, XFIX_0_707106781 + vadd.s16 q11, q12, q6 + vadd.s16 q8, q2, q3 + vsub.s16 q12, q2, q3 + vadd.s16 q3, q6, q7 + vadd.s16 q7, q7, q1 + vqdmulh.s16 q3, q3, XFIX_0_707106781 + vsub.s16 q6, q11, q7 + vadd.s16 q10, q5, q4 + vqdmulh.s16 q6, q6, XFIX_0_382683433 + vsub.s16 q14, q5, q4 + vqdmulh.s16 q11, q11, XFIX_0_541196100 + vqdmulh.s16 q5, q7, XFIX_1_306562965 + vadd.s16 q4, q1, q3 + vsub.s16 q3, q1, q3 + vadd.s16 q7, q7, q6 + vadd.s16 q11, q11, q6 + vadd.s16 q7, q7, q5 + vadd.s16 q13, q3, q11 + vsub.s16 q11, q3, q11 + vadd.s16 q9, q4, q7 + vsub.s16 q15, q4, q7 + subs TMP, TMP, #1 + bne 1b + + /* store results */ + vst1.16 {d16, d17, d18, d19}, [DATA, :128]! + vst1.16 {d20, d21, d22, d23}, [DATA, :128]! + vst1.16 {d24, d25, d26, d27}, [DATA, :128]! + vst1.16 {d28, d29, d30, d31}, [DATA, :128] + + vpop {d8-d15} + bx lr + + .unreq DATA + .unreq TMP + + +/*****************************************************************************/ + +/* + * GLOBAL(void) + * jsimd_quantize_neon (JCOEFPTR coef_block, DCTELEM * divisors, + * DCTELEM * workspace); + * + * Note: the code uses 2 stage pipelining in order to improve instructions + * scheduling and eliminate stalls (this provides ~15% better + * performance for this function on both ARM Cortex-A8 and + * ARM Cortex-A9 when compared to the non-pipelined variant). + * The instructions which belong to the second stage use different + * indentation for better readiability. + */ +asm_function jsimd_quantize_neon + + COEF_BLOCK .req r0 + DIVISORS .req r1 + WORKSPACE .req r2 + + RECIPROCAL .req DIVISORS + CORRECTION .req r3 + SHIFT .req ip + LOOP_COUNT .req r4 + + vld1.16 {d0, d1, d2, d3}, [WORKSPACE, :128]! + vabs.s16 q12, q0 + add CORRECTION, DIVISORS, #(64 * 2) + add SHIFT, DIVISORS, #(64 * 6) + vld1.16 {d20, d21, d22, d23}, [CORRECTION, :128]! + vabs.s16 q13, q1 + vld1.16 {d16, d17, d18, d19}, [RECIPROCAL, :128]! + vadd.u16 q12, q12, q10 /* add correction */ + vadd.u16 q13, q13, q11 + vmull.u16 q10, d24, d16 /* multiply by reciprocal */ + vmull.u16 q11, d25, d17 + vmull.u16 q8, d26, d18 + vmull.u16 q9, d27, d19 + vld1.16 {d24, d25, d26, d27}, [SHIFT, :128]! + vshrn.u32 d20, q10, #16 + vshrn.u32 d21, q11, #16 + vshrn.u32 d22, q8, #16 + vshrn.u32 d23, q9, #16 + vneg.s16 q12, q12 + vneg.s16 q13, q13 + vshr.s16 q2, q0, #15 /* extract sign */ + vshr.s16 q3, q1, #15 + vshl.u16 q14, q10, q12 /* shift */ + vshl.u16 q15, q11, q13 + + push {r4, r5} + mov LOOP_COUNT, #3 +1: + vld1.16 {d0, d1, d2, d3}, [WORKSPACE, :128]! + veor.u16 q14, q14, q2 /* restore sign */ + vabs.s16 q12, q0 + vld1.16 {d20, d21, d22, d23}, [CORRECTION, :128]! + vabs.s16 q13, q1 + veor.u16 q15, q15, q3 + vld1.16 {d16, d17, d18, d19}, [RECIPROCAL, :128]! + vadd.u16 q12, q12, q10 /* add correction */ + vadd.u16 q13, q13, q11 + vmull.u16 q10, d24, d16 /* multiply by reciprocal */ + vmull.u16 q11, d25, d17 + vmull.u16 q8, d26, d18 + vmull.u16 q9, d27, d19 + vsub.u16 q14, q14, q2 + vld1.16 {d24, d25, d26, d27}, [SHIFT, :128]! + vsub.u16 q15, q15, q3 + vshrn.u32 d20, q10, #16 + vshrn.u32 d21, q11, #16 + vst1.16 {d28, d29, d30, d31}, [COEF_BLOCK, :128]! + vshrn.u32 d22, q8, #16 + vshrn.u32 d23, q9, #16 + vneg.s16 q12, q12 + vneg.s16 q13, q13 + vshr.s16 q2, q0, #15 /* extract sign */ + vshr.s16 q3, q1, #15 + vshl.u16 q14, q10, q12 /* shift */ + vshl.u16 q15, q11, q13 + subs LOOP_COUNT, LOOP_COUNT, #1 + bne 1b + pop {r4, r5} + + veor.u16 q14, q14, q2 /* restore sign */ + veor.u16 q15, q15, q3 + vsub.u16 q14, q14, q2 + vsub.u16 q15, q15, q3 + vst1.16 {d28, d29, d30, d31}, [COEF_BLOCK, :128]! + + bx lr /* return */ + + .unreq COEF_BLOCK + .unreq DIVISORS + .unreq WORKSPACE + .unreq RECIPROCAL + .unreq CORRECTION + .unreq SHIFT + .unreq LOOP_COUNT + + +/*****************************************************************************/ + +/* + * GLOBAL(void) + * jsimd_h2v1_fancy_upsample_neon (int max_v_samp_factor, + * JDIMENSION downsampled_width, + * JSAMPARRAY input_data, + * JSAMPARRAY * output_data_ptr); + * + * Note: the use of unaligned writes is the main remaining bottleneck in + * this code, which can be potentially solved to get up to tens + * of percents performance improvement on Cortex-A8/Cortex-A9. + */ + +/* + * Upsample 16 source pixels to 32 destination pixels. The new 16 source + * pixels are loaded to q0. The previous 16 source pixels are in q1. The + * shifted-by-one source pixels are constructed in q2 by using q0 and q1. + * Register d28 is used for multiplication by 3. Register q15 is used + * for adding +1 bias. + */ +.macro upsample16 OUTPTR, INPTR + vld1.8 {q0}, [\INPTR]! + vmovl.u8 q8, d0 + vext.8 q2, q1, q0, #15 + vmovl.u8 q9, d1 + vaddw.u8 q10, q15, d4 + vaddw.u8 q11, q15, d5 + vmlal.u8 q8, d4, d28 + vmlal.u8 q9, d5, d28 + vmlal.u8 q10, d0, d28 + vmlal.u8 q11, d1, d28 + vmov q1, q0 /* backup source pixels to q1 */ + vrshrn.u16 d6, q8, #2 + vrshrn.u16 d7, q9, #2 + vshrn.u16 d8, q10, #2 + vshrn.u16 d9, q11, #2 + vst2.8 {d6, d7, d8, d9}, [\OUTPTR]! +.endm + +/* + * Upsample 32 source pixels to 64 destination pixels. Compared to 'usample16' + * macro, the roles of q0 and q1 registers are reversed for even and odd + * groups of 16 pixels, that's why "vmov q1, q0" instructions are not needed. + * Also this unrolling allows to reorder loads and stores to compensate + * multiplication latency and reduce stalls. + */ +.macro upsample32 OUTPTR, INPTR + /* even 16 pixels group */ + vld1.8 {q0}, [\INPTR]! + vmovl.u8 q8, d0 + vext.8 q2, q1, q0, #15 + vmovl.u8 q9, d1 + vaddw.u8 q10, q15, d4 + vaddw.u8 q11, q15, d5 + vmlal.u8 q8, d4, d28 + vmlal.u8 q9, d5, d28 + vmlal.u8 q10, d0, d28 + vmlal.u8 q11, d1, d28 + /* odd 16 pixels group */ + vld1.8 {q1}, [\INPTR]! + vrshrn.u16 d6, q8, #2 + vrshrn.u16 d7, q9, #2 + vshrn.u16 d8, q10, #2 + vshrn.u16 d9, q11, #2 + vmovl.u8 q8, d2 + vext.8 q2, q0, q1, #15 + vmovl.u8 q9, d3 + vaddw.u8 q10, q15, d4 + vaddw.u8 q11, q15, d5 + vmlal.u8 q8, d4, d28 + vmlal.u8 q9, d5, d28 + vmlal.u8 q10, d2, d28 + vmlal.u8 q11, d3, d28 + vst2.8 {d6, d7, d8, d9}, [\OUTPTR]! + vrshrn.u16 d6, q8, #2 + vrshrn.u16 d7, q9, #2 + vshrn.u16 d8, q10, #2 + vshrn.u16 d9, q11, #2 + vst2.8 {d6, d7, d8, d9}, [\OUTPTR]! +.endm + +/* + * Upsample a row of WIDTH pixels from INPTR to OUTPTR. + */ +.macro upsample_row OUTPTR, INPTR, WIDTH, TMP1 + /* special case for the first and last pixels */ + sub \WIDTH, \WIDTH, #1 + add \OUTPTR, \OUTPTR, #1 + ldrb \TMP1, [\INPTR, \WIDTH] + strb \TMP1, [\OUTPTR, \WIDTH, asl #1] + ldrb \TMP1, [\INPTR], #1 + strb \TMP1, [\OUTPTR, #-1] + vmov.8 d3[7], \TMP1 + + subs \WIDTH, \WIDTH, #32 + blt 5f +0: /* process 32 pixels per iteration */ + upsample32 \OUTPTR, \INPTR + subs \WIDTH, \WIDTH, #32 + bge 0b +5: + adds \WIDTH, \WIDTH, #16 + blt 1f +0: /* process 16 pixels if needed */ + upsample16 \OUTPTR, \INPTR + subs \WIDTH, \WIDTH, #16 +1: + adds \WIDTH, \WIDTH, #16 + beq 9f + + /* load the remaining 1-15 pixels */ + add \INPTR, \INPTR, \WIDTH + tst \WIDTH, #1 + beq 2f + sub \INPTR, \INPTR, #1 + vld1.8 {d0[0]}, [\INPTR] +2: + tst \WIDTH, #2 + beq 2f + vext.8 d0, d0, d0, #6 + sub \INPTR, \INPTR, #1 + vld1.8 {d0[1]}, [\INPTR] + sub \INPTR, \INPTR, #1 + vld1.8 {d0[0]}, [\INPTR] +2: + tst \WIDTH, #4 + beq 2f + vrev64.32 d0, d0 + sub \INPTR, \INPTR, #1 + vld1.8 {d0[3]}, [\INPTR] + sub \INPTR, \INPTR, #1 + vld1.8 {d0[2]}, [\INPTR] + sub \INPTR, \INPTR, #1 + vld1.8 {d0[1]}, [\INPTR] + sub \INPTR, \INPTR, #1 + vld1.8 {d0[0]}, [\INPTR] +2: + tst \WIDTH, #8 + beq 2f + vmov d1, d0 + sub \INPTR, \INPTR, #8 + vld1.8 {d0}, [\INPTR] +2: /* upsample the remaining pixels */ + vmovl.u8 q8, d0 + vext.8 q2, q1, q0, #15 + vmovl.u8 q9, d1 + vaddw.u8 q10, q15, d4 + vaddw.u8 q11, q15, d5 + vmlal.u8 q8, d4, d28 + vmlal.u8 q9, d5, d28 + vmlal.u8 q10, d0, d28 + vmlal.u8 q11, d1, d28 + vrshrn.u16 d10, q8, #2 + vrshrn.u16 d12, q9, #2 + vshrn.u16 d11, q10, #2 + vshrn.u16 d13, q11, #2 + vzip.8 d10, d11 + vzip.8 d12, d13 + /* store the remaining pixels */ + tst \WIDTH, #8 + beq 2f + vst1.8 {d10, d11}, [\OUTPTR]! + vmov q5, q6 +2: + tst \WIDTH, #4 + beq 2f + vst1.8 {d10}, [\OUTPTR]! + vmov d10, d11 +2: + tst \WIDTH, #2 + beq 2f + vst1.8 {d10[0]}, [\OUTPTR]! + vst1.8 {d10[1]}, [\OUTPTR]! + vst1.8 {d10[2]}, [\OUTPTR]! + vst1.8 {d10[3]}, [\OUTPTR]! + vext.8 d10, d10, d10, #4 +2: + tst \WIDTH, #1 + beq 2f + vst1.8 {d10[0]}, [\OUTPTR]! + vst1.8 {d10[1]}, [\OUTPTR]! +2: +9: +.endm + +asm_function jsimd_h2v1_fancy_upsample_neon + + MAX_V_SAMP_FACTOR .req r0 + DOWNSAMPLED_WIDTH .req r1 + INPUT_DATA .req r2 + OUTPUT_DATA_PTR .req r3 + OUTPUT_DATA .req OUTPUT_DATA_PTR + + OUTPTR .req r4 + INPTR .req r5 + WIDTH .req ip + TMP .req lr + + push {r4, r5, r6, lr} + vpush {d8-d15} + + ldr OUTPUT_DATA, [OUTPUT_DATA_PTR] + cmp MAX_V_SAMP_FACTOR, #0 + ble 99f + + /* initialize constants */ + vmov.u8 d28, #3 + vmov.u16 q15, #1 +11: + ldr INPTR, [INPUT_DATA], #4 + ldr OUTPTR, [OUTPUT_DATA], #4 + mov WIDTH, DOWNSAMPLED_WIDTH + upsample_row OUTPTR, INPTR, WIDTH, TMP + subs MAX_V_SAMP_FACTOR, MAX_V_SAMP_FACTOR, #1 + bgt 11b + +99: + vpop {d8-d15} + pop {r4, r5, r6, pc} + + .unreq MAX_V_SAMP_FACTOR + .unreq DOWNSAMPLED_WIDTH + .unreq INPUT_DATA + .unreq OUTPUT_DATA_PTR + .unreq OUTPUT_DATA + + .unreq OUTPTR + .unreq INPTR + .unreq WIDTH + .unreq TMP + + +.purgem upsample16 +.purgem upsample32 +.purgem upsample_row diff --git a/app/src/main/jni/libjpeg-turbo/simd/jsimd_i386.c b/app/src/main/jni/libjpeg-turbo/simd/jsimd_i386.c new file mode 100644 index 000000000..25d06974d --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jsimd_i386.c @@ -0,0 +1,1061 @@ +/* + * jsimd_i386.c + * + * Copyright 2009 Pierre Ossman for Cendio AB + * Copyright 2009-2011, 2013-2014 D. R. Commander + * + * Based on the x86 SIMD extension for IJG JPEG library, + * Copyright (C) 1999-2006, MIYASAKA Masaru. + * For conditions of distribution and use, see copyright notice in jsimdext.inc + * + * This file contains the interface between the "normal" portions + * of the library and the SIMD implementations when running on a + * 32-bit x86 architecture. + */ + +#define JPEG_INTERNALS +#include "../jinclude.h" +#include "../jpeglib.h" +#include "../jsimd.h" +#include "../jdct.h" +#include "../jsimddct.h" +#include "jsimd.h" + +/* + * In the PIC cases, we have no guarantee that constants will keep + * their alignment. This macro allows us to verify it at runtime. + */ +#define IS_ALIGNED(ptr, order) (((unsigned)ptr & ((1 << order) - 1)) == 0) + +#define IS_ALIGNED_SSE(ptr) (IS_ALIGNED(ptr, 4)) /* 16 byte alignment */ + +static unsigned int simd_support = ~0; + +/* + * Check what SIMD accelerations are supported. + * + * FIXME: This code is racy under a multi-threaded environment. + */ +LOCAL(void) +init_simd (void) +{ + char *env = NULL; + + if (simd_support != ~0U) + return; + + simd_support = jpeg_simd_cpu_support(); + + /* Force different settings through environment variables */ + env = getenv("JSIMD_FORCEMMX"); + if ((env != NULL) && (strcmp(env, "1") == 0)) + simd_support &= JSIMD_MMX; + env = getenv("JSIMD_FORCE3DNOW"); + if ((env != NULL) && (strcmp(env, "1") == 0)) + simd_support &= JSIMD_3DNOW|JSIMD_MMX; + env = getenv("JSIMD_FORCESSE"); + if ((env != NULL) && (strcmp(env, "1") == 0)) + simd_support &= JSIMD_SSE|JSIMD_MMX; + env = getenv("JSIMD_FORCESSE2"); + if ((env != NULL) && (strcmp(env, "1") == 0)) + simd_support &= JSIMD_SSE2; + env = getenv("JSIMD_FORCENONE"); + if ((env != NULL) && (strcmp(env, "1") == 0)) + simd_support = 0; +} + +GLOBAL(int) +jsimd_can_rgb_ycc (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) + return 0; + + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_rgb_ycc_convert_sse2)) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_rgb_gray (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) + return 0; + + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2)) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_ycc_rgb (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) + return 0; + + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_ycc_rgb_convert_sse2)) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_ycc_rgb565 (void) +{ + return 0; +} + +GLOBAL(void) +jsimd_rgb_ycc_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ + void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int); + void (*mmxfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int); + + switch(cinfo->in_color_space) { + case JCS_EXT_RGB: + sse2fct=jsimd_extrgb_ycc_convert_sse2; + mmxfct=jsimd_extrgb_ycc_convert_mmx; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + sse2fct=jsimd_extrgbx_ycc_convert_sse2; + mmxfct=jsimd_extrgbx_ycc_convert_mmx; + break; + case JCS_EXT_BGR: + sse2fct=jsimd_extbgr_ycc_convert_sse2; + mmxfct=jsimd_extbgr_ycc_convert_mmx; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + sse2fct=jsimd_extbgrx_ycc_convert_sse2; + mmxfct=jsimd_extbgrx_ycc_convert_mmx; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + sse2fct=jsimd_extxbgr_ycc_convert_sse2; + mmxfct=jsimd_extxbgr_ycc_convert_mmx; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + sse2fct=jsimd_extxrgb_ycc_convert_sse2; + mmxfct=jsimd_extxrgb_ycc_convert_mmx; + break; + default: + sse2fct=jsimd_rgb_ycc_convert_sse2; + mmxfct=jsimd_rgb_ycc_convert_mmx; + break; + } + + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_rgb_ycc_convert_sse2)) + sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows); + else if (simd_support & JSIMD_MMX) + mmxfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows); +} + +GLOBAL(void) +jsimd_rgb_gray_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ + void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int); + void (*mmxfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int); + + switch(cinfo->in_color_space) { + case JCS_EXT_RGB: + sse2fct=jsimd_extrgb_gray_convert_sse2; + mmxfct=jsimd_extrgb_gray_convert_mmx; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + sse2fct=jsimd_extrgbx_gray_convert_sse2; + mmxfct=jsimd_extrgbx_gray_convert_mmx; + break; + case JCS_EXT_BGR: + sse2fct=jsimd_extbgr_gray_convert_sse2; + mmxfct=jsimd_extbgr_gray_convert_mmx; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + sse2fct=jsimd_extbgrx_gray_convert_sse2; + mmxfct=jsimd_extbgrx_gray_convert_mmx; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + sse2fct=jsimd_extxbgr_gray_convert_sse2; + mmxfct=jsimd_extxbgr_gray_convert_mmx; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + sse2fct=jsimd_extxrgb_gray_convert_sse2; + mmxfct=jsimd_extxrgb_gray_convert_mmx; + break; + default: + sse2fct=jsimd_rgb_gray_convert_sse2; + mmxfct=jsimd_rgb_gray_convert_mmx; + break; + } + + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2)) + sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows); + else if (simd_support & JSIMD_MMX) + mmxfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows); +} + +GLOBAL(void) +jsimd_ycc_rgb_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int); + void (*mmxfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int); + + switch(cinfo->out_color_space) { + case JCS_EXT_RGB: + sse2fct=jsimd_ycc_extrgb_convert_sse2; + mmxfct=jsimd_ycc_extrgb_convert_mmx; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + sse2fct=jsimd_ycc_extrgbx_convert_sse2; + mmxfct=jsimd_ycc_extrgbx_convert_mmx; + break; + case JCS_EXT_BGR: + sse2fct=jsimd_ycc_extbgr_convert_sse2; + mmxfct=jsimd_ycc_extbgr_convert_mmx; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + sse2fct=jsimd_ycc_extbgrx_convert_sse2; + mmxfct=jsimd_ycc_extbgrx_convert_mmx; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + sse2fct=jsimd_ycc_extxbgr_convert_sse2; + mmxfct=jsimd_ycc_extxbgr_convert_mmx; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + sse2fct=jsimd_ycc_extxrgb_convert_sse2; + mmxfct=jsimd_ycc_extxrgb_convert_mmx; + break; + default: + sse2fct=jsimd_ycc_rgb_convert_sse2; + mmxfct=jsimd_ycc_rgb_convert_mmx; + break; + } + + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_ycc_rgb_convert_sse2)) + sse2fct(cinfo->output_width, input_buf, input_row, output_buf, num_rows); + else if (simd_support & JSIMD_MMX) + mmxfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows); +} + +GLOBAL(void) +jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ +} + +GLOBAL(int) +jsimd_can_h2v2_downsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_SSE2) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_downsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_SSE2) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ + if (simd_support & JSIMD_SSE2) + jsimd_h2v2_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor, + compptr->v_samp_factor, + compptr->width_in_blocks, input_data, + output_data); + else if (simd_support & JSIMD_MMX) + jsimd_h2v2_downsample_mmx(cinfo->image_width, cinfo->max_v_samp_factor, + compptr->v_samp_factor, compptr->width_in_blocks, + input_data, output_data); +} + +GLOBAL(void) +jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ + if (simd_support & JSIMD_SSE2) + jsimd_h2v1_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor, + compptr->v_samp_factor, + compptr->width_in_blocks, input_data, + output_data); + else if (simd_support & JSIMD_MMX) + jsimd_h2v1_downsample_mmx(cinfo->image_width, cinfo->max_v_samp_factor, + compptr->v_samp_factor, compptr->width_in_blocks, + input_data, output_data); +} + +GLOBAL(int) +jsimd_can_h2v2_upsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_SSE2) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_upsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_SSE2) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ + if (simd_support & JSIMD_SSE2) + jsimd_h2v2_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width, + input_data, output_data_ptr); + else if (simd_support & JSIMD_MMX) + jsimd_h2v2_upsample_mmx(cinfo->max_v_samp_factor, cinfo->output_width, + input_data, output_data_ptr); +} + +GLOBAL(void) +jsimd_h2v1_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ + if (simd_support & JSIMD_SSE2) + jsimd_h2v1_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width, + input_data, output_data_ptr); + else if (simd_support & JSIMD_MMX) + jsimd_h2v1_upsample_mmx(cinfo->max_v_samp_factor, cinfo->output_width, + input_data, output_data_ptr); +} + +GLOBAL(int) +jsimd_can_h2v2_fancy_upsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_fancy_upsample_sse2)) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_fancy_upsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_fancy_upsample_sse2)) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_fancy_upsample_sse2)) + jsimd_h2v2_fancy_upsample_sse2(cinfo->max_v_samp_factor, + compptr->downsampled_width, input_data, + output_data_ptr); + else if (simd_support & JSIMD_MMX) + jsimd_h2v2_fancy_upsample_mmx(cinfo->max_v_samp_factor, + compptr->downsampled_width, input_data, + output_data_ptr); +} + +GLOBAL(void) +jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_fancy_upsample_sse2)) + jsimd_h2v1_fancy_upsample_sse2(cinfo->max_v_samp_factor, + compptr->downsampled_width, input_data, + output_data_ptr); + else if (simd_support & JSIMD_MMX) + jsimd_h2v1_fancy_upsample_mmx(cinfo->max_v_samp_factor, + compptr->downsampled_width, input_data, + output_data_ptr); +} + +GLOBAL(int) +jsimd_can_h2v2_merged_upsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_merged_upsample_sse2)) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_merged_upsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_merged_upsample_sse2)) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY); + void (*mmxfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY); + + switch(cinfo->out_color_space) { + case JCS_EXT_RGB: + sse2fct=jsimd_h2v2_extrgb_merged_upsample_sse2; + mmxfct=jsimd_h2v2_extrgb_merged_upsample_mmx; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + sse2fct=jsimd_h2v2_extrgbx_merged_upsample_sse2; + mmxfct=jsimd_h2v2_extrgbx_merged_upsample_mmx; + break; + case JCS_EXT_BGR: + sse2fct=jsimd_h2v2_extbgr_merged_upsample_sse2; + mmxfct=jsimd_h2v2_extbgr_merged_upsample_mmx; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + sse2fct=jsimd_h2v2_extbgrx_merged_upsample_sse2; + mmxfct=jsimd_h2v2_extbgrx_merged_upsample_mmx; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + sse2fct=jsimd_h2v2_extxbgr_merged_upsample_sse2; + mmxfct=jsimd_h2v2_extxbgr_merged_upsample_mmx; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + sse2fct=jsimd_h2v2_extxrgb_merged_upsample_sse2; + mmxfct=jsimd_h2v2_extxrgb_merged_upsample_mmx; + break; + default: + sse2fct=jsimd_h2v2_merged_upsample_sse2; + mmxfct=jsimd_h2v2_merged_upsample_mmx; + break; + } + + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_merged_upsample_sse2)) + sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf); + else if (simd_support & JSIMD_MMX) + mmxfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf); +} + +GLOBAL(void) +jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY); + void (*mmxfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY); + + switch(cinfo->out_color_space) { + case JCS_EXT_RGB: + sse2fct=jsimd_h2v1_extrgb_merged_upsample_sse2; + mmxfct=jsimd_h2v1_extrgb_merged_upsample_mmx; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + sse2fct=jsimd_h2v1_extrgbx_merged_upsample_sse2; + mmxfct=jsimd_h2v1_extrgbx_merged_upsample_mmx; + break; + case JCS_EXT_BGR: + sse2fct=jsimd_h2v1_extbgr_merged_upsample_sse2; + mmxfct=jsimd_h2v1_extbgr_merged_upsample_mmx; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + sse2fct=jsimd_h2v1_extbgrx_merged_upsample_sse2; + mmxfct=jsimd_h2v1_extbgrx_merged_upsample_mmx; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + sse2fct=jsimd_h2v1_extxbgr_merged_upsample_sse2; + mmxfct=jsimd_h2v1_extxbgr_merged_upsample_mmx; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + sse2fct=jsimd_h2v1_extxrgb_merged_upsample_sse2; + mmxfct=jsimd_h2v1_extxrgb_merged_upsample_mmx; + break; + default: + sse2fct=jsimd_h2v1_merged_upsample_sse2; + mmxfct=jsimd_h2v1_merged_upsample_mmx; + break; + } + + if ((simd_support & JSIMD_SSE2) && + IS_ALIGNED_SSE(jconst_merged_upsample_sse2)) + sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf); + else if (simd_support & JSIMD_MMX) + mmxfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf); +} + +GLOBAL(int) +jsimd_can_convsamp (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + if (simd_support & JSIMD_SSE2) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_convsamp_float (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(FAST_FLOAT) != 4) + return 0; + + if (simd_support & JSIMD_SSE2) + return 1; + if (simd_support & JSIMD_SSE) + return 1; + if (simd_support & JSIMD_3DNOW) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col, + DCTELEM * workspace) +{ + if (simd_support & JSIMD_SSE2) + jsimd_convsamp_sse2(sample_data, start_col, workspace); + else if (simd_support & JSIMD_MMX) + jsimd_convsamp_mmx(sample_data, start_col, workspace); +} + +GLOBAL(void) +jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col, + FAST_FLOAT * workspace) +{ + if (simd_support & JSIMD_SSE2) + jsimd_convsamp_float_sse2(sample_data, start_col, workspace); + else if (simd_support & JSIMD_SSE) + jsimd_convsamp_float_sse(sample_data, start_col, workspace); + else if (simd_support & JSIMD_3DNOW) + jsimd_convsamp_float_3dnow(sample_data, start_col, workspace); +} + +GLOBAL(int) +jsimd_can_fdct_islow (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2)) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_fdct_ifast (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_ifast_sse2)) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_fdct_float (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(FAST_FLOAT) != 4) + return 0; + + if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_fdct_float_sse)) + return 1; + if (simd_support & JSIMD_3DNOW) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_fdct_islow (DCTELEM * data) +{ + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2)) + jsimd_fdct_islow_sse2(data); + else if (simd_support & JSIMD_MMX) + jsimd_fdct_islow_mmx(data); +} + +GLOBAL(void) +jsimd_fdct_ifast (DCTELEM * data) +{ + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2)) + jsimd_fdct_ifast_sse2(data); + else if (simd_support & JSIMD_MMX) + jsimd_fdct_ifast_mmx(data); +} + +GLOBAL(void) +jsimd_fdct_float (FAST_FLOAT * data) +{ + if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_fdct_float_sse)) + jsimd_fdct_float_sse(data); + else if (simd_support & JSIMD_3DNOW) + jsimd_fdct_float_3dnow(data); +} + +GLOBAL(int) +jsimd_can_quantize (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + if (simd_support & JSIMD_SSE2) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_quantize_float (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (sizeof(FAST_FLOAT) != 4) + return 0; + + if (simd_support & JSIMD_SSE2) + return 1; + if (simd_support & JSIMD_SSE) + return 1; + if (simd_support & JSIMD_3DNOW) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_quantize (JCOEFPTR coef_block, DCTELEM * divisors, + DCTELEM * workspace) +{ + if (simd_support & JSIMD_SSE2) + jsimd_quantize_sse2(coef_block, divisors, workspace); + else if (simd_support & JSIMD_MMX) + jsimd_quantize_mmx(coef_block, divisors, workspace); +} + +GLOBAL(void) +jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT * divisors, + FAST_FLOAT * workspace) +{ + if (simd_support & JSIMD_SSE2) + jsimd_quantize_float_sse2(coef_block, divisors, workspace); + else if (simd_support & JSIMD_SSE) + jsimd_quantize_float_sse(coef_block, divisors, workspace); + else if (simd_support & JSIMD_3DNOW) + jsimd_quantize_float_3dnow(coef_block, divisors, workspace); +} + +GLOBAL(int) +jsimd_can_idct_2x2 (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2)) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_4x4 (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2)) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2)) + jsimd_idct_2x2_sse2(compptr->dct_table, coef_block, output_buf, + output_col); + else if (simd_support & JSIMD_MMX) + jsimd_idct_2x2_mmx(compptr->dct_table, coef_block, output_buf, output_col); +} + +GLOBAL(void) +jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2)) + jsimd_idct_4x4_sse2(compptr->dct_table, coef_block, output_buf, + output_col); + else if (simd_support & JSIMD_MMX) + jsimd_idct_4x4_mmx(compptr->dct_table, coef_block, output_buf, output_col); +} + +GLOBAL(int) +jsimd_can_idct_islow (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_islow_sse2)) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_ifast (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(IFAST_MULT_TYPE) != 2) + return 0; + if (IFAST_SCALE_BITS != 2) + return 0; + + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_ifast_sse2)) + return 1; + if (simd_support & JSIMD_MMX) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_float (void) +{ + init_simd(); + + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(FAST_FLOAT) != 4) + return 0; + if (sizeof(FLOAT_MULT_TYPE) != 4) + return 0; + + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_float_sse2)) + return 1; + if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_idct_float_sse)) + return 1; + if (simd_support & JSIMD_3DNOW) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_islow_sse2)) + jsimd_idct_islow_sse2(compptr->dct_table, coef_block, output_buf, + output_col); + else if (simd_support & JSIMD_MMX) + jsimd_idct_islow_mmx(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(void) +jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_ifast_sse2)) + jsimd_idct_ifast_sse2(compptr->dct_table, coef_block, output_buf, + output_col); + else if (simd_support & JSIMD_MMX) + jsimd_idct_ifast_mmx(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(void) +jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_float_sse2)) + jsimd_idct_float_sse2(compptr->dct_table, coef_block, output_buf, + output_col); + else if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_idct_float_sse)) + jsimd_idct_float_sse(compptr->dct_table, coef_block, output_buf, + output_col); + else if (simd_support & JSIMD_3DNOW) + jsimd_idct_float_3dnow(compptr->dct_table, coef_block, output_buf, + output_col); +} + diff --git a/app/src/main/jni/libjpeg-turbo/simd/jsimd_mips.c b/app/src/main/jni/libjpeg-turbo/simd/jsimd_mips.c new file mode 100644 index 000000000..abcd19f54 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jsimd_mips.c @@ -0,0 +1,1111 @@ +/* + * jsimd_mips.c + * + * Copyright 2009 Pierre Ossman for Cendio AB + * Copyright 2009-2011, 2014 D. R. Commander + * Copyright (C) 2013-2014, MIPS Technologies, Inc., California + * + * Based on the x86 SIMD extension for IJG JPEG library, + * Copyright (C) 1999-2006, MIYASAKA Masaru. + * For conditions of distribution and use, see copyright notice in jsimdext.inc + * + * This file contains the interface between the "normal" portions + * of the library and the SIMD implementations when running on a + * MIPS architecture. + */ + +#define JPEG_INTERNALS +#include "../jinclude.h" +#include "../jpeglib.h" +#include "../jsimd.h" +#include "../jdct.h" +#include "../jsimddct.h" +#include "jsimd.h" + +#include +#include +#include + +static unsigned int simd_support = ~0; + +#if defined(__linux__) + +LOCAL(int) +parse_proc_cpuinfo(const char* search_string) +{ + const char* file_name = "/proc/cpuinfo"; + char cpuinfo_line[256]; + FILE* f = NULL; + simd_support = 0; + + if ((f = fopen(file_name, "r")) != NULL) { + while (fgets(cpuinfo_line, sizeof(cpuinfo_line), f) != NULL) { + if (strstr(cpuinfo_line, search_string) != NULL) { + fclose(f); + simd_support |= JSIMD_MIPS_DSPR2; + return 1; + } + } + fclose(f); + } + /* Did not find string in the proc file, or not Linux ELF. */ + return 0; +} + +#endif + +/* + * Check what SIMD accelerations are supported. + * + * FIXME: This code is racy under a multi-threaded environment. + */ +LOCAL(void) +init_simd (void) +{ + if (simd_support != ~0U) + return; + + simd_support = 0; + +#if defined(__MIPSEL__) && defined(__mips_dsp) && (__mips_dsp_rev >= 2) + simd_support |= JSIMD_MIPS_DSPR2; +#elif defined(__linux__) + /* We still have a chance to use MIPS DSPR2 regardless of globally used + * -mdspr2 options passed to gcc by performing runtime detection via + * /proc/cpuinfo parsing on linux */ + if (!parse_proc_cpuinfo("MIPS 74K")) + return; +#endif +} + +static const int mips_idct_ifast_coefs[4] = { + 0x45404540, // FIX( 1.082392200 / 2) = 17734 = 0x4546 + 0x5A805A80, // FIX( 1.414213562 / 2) = 23170 = 0x5A82 + 0x76407640, // FIX( 1.847759065 / 2) = 30274 = 0x7642 + 0xAC60AC60 // FIX(-2.613125930 / 4) = -21407 = 0xAC61 +}; + +/* The following struct is borrowed from jdsample.c */ +typedef void (*upsample1_ptr) (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr); + +typedef struct { + struct jpeg_upsampler pub; + JSAMPARRAY color_buf[MAX_COMPONENTS]; + upsample1_ptr methods[MAX_COMPONENTS]; + int next_row_out; + JDIMENSION rows_to_go; + int rowgroup_height[MAX_COMPONENTS]; + UINT8 h_expand[MAX_COMPONENTS]; + UINT8 v_expand[MAX_COMPONENTS]; +} my_upsampler; + +typedef my_upsampler * my_upsample_ptr; + +GLOBAL(int) +jsimd_can_rgb_ycc (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_rgb_gray (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_ycc_rgb (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_ycc_rgb565 (void) +{ + return 0; +} + +GLOBAL(int) +jsimd_c_can_null_convert (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_rgb_ycc_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ + void (*mipsdspr2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int); + + switch(cinfo->in_color_space) { + case JCS_EXT_RGB: + mipsdspr2fct=jsimd_extrgb_ycc_convert_mips_dspr2; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + mipsdspr2fct=jsimd_extrgbx_ycc_convert_mips_dspr2; + break; + case JCS_EXT_BGR: + mipsdspr2fct=jsimd_extbgr_ycc_convert_mips_dspr2; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + mipsdspr2fct=jsimd_extbgrx_ycc_convert_mips_dspr2; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + mipsdspr2fct=jsimd_extxbgr_ycc_convert_mips_dspr2; + + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + mipsdspr2fct=jsimd_extxrgb_ycc_convert_mips_dspr2; + break; + default: + mipsdspr2fct=jsimd_extrgb_ycc_convert_mips_dspr2; + break; + } + + if (simd_support & JSIMD_MIPS_DSPR2) + mipsdspr2fct(cinfo->image_width, input_buf, output_buf, output_row, + num_rows); +} + +GLOBAL(void) +jsimd_rgb_gray_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ + void (*mipsdspr2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int); + + switch(cinfo->in_color_space) { + case JCS_EXT_RGB: + mipsdspr2fct=jsimd_extrgb_gray_convert_mips_dspr2; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + mipsdspr2fct=jsimd_extrgbx_gray_convert_mips_dspr2; + break; + case JCS_EXT_BGR: + mipsdspr2fct=jsimd_extbgr_gray_convert_mips_dspr2; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + mipsdspr2fct=jsimd_extbgrx_gray_convert_mips_dspr2; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + mipsdspr2fct=jsimd_extxbgr_gray_convert_mips_dspr2; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + mipsdspr2fct=jsimd_extxrgb_gray_convert_mips_dspr2; + break; + default: + mipsdspr2fct=jsimd_extrgb_gray_convert_mips_dspr2; + break; + } + + if (simd_support & JSIMD_MIPS_DSPR2) + mipsdspr2fct(cinfo->image_width, input_buf, output_buf, output_row, + num_rows); +} + +GLOBAL(void) +jsimd_ycc_rgb_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + void (*mipsdspr2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int); + + switch(cinfo->out_color_space) { + case JCS_EXT_RGB: + mipsdspr2fct=jsimd_ycc_extrgb_convert_mips_dspr2; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + mipsdspr2fct=jsimd_ycc_extrgbx_convert_mips_dspr2; + break; + case JCS_EXT_BGR: + mipsdspr2fct=jsimd_ycc_extbgr_convert_mips_dspr2; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + mipsdspr2fct=jsimd_ycc_extbgrx_convert_mips_dspr2; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + mipsdspr2fct=jsimd_ycc_extxbgr_convert_mips_dspr2; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + mipsdspr2fct=jsimd_ycc_extxrgb_convert_mips_dspr2; + break; + default: + mipsdspr2fct=jsimd_ycc_extrgb_convert_mips_dspr2; + break; + } + + if (simd_support & JSIMD_MIPS_DSPR2) + mipsdspr2fct(cinfo->output_width, input_buf, input_row, output_buf, + num_rows); +} + +GLOBAL(void) +jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ +} + +GLOBAL(void) +jsimd_c_null_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_c_null_convert_mips_dspr2(cinfo->image_width, input_buf, + output_buf, output_row, num_rows, + cinfo->num_components); +} + +GLOBAL(int) +jsimd_can_h2v2_downsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v2_smooth_downsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if(DCTSIZE != 8) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_downsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_h2v2_downsample_mips_dspr2(cinfo->image_width, + cinfo->max_v_samp_factor, + compptr->v_samp_factor, + compptr->width_in_blocks, input_data, + output_data); +} + +GLOBAL(void) +jsimd_h2v2_smooth_downsample (j_compress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ + jsimd_h2v2_smooth_downsample_mips_dspr2(input_data, output_data, + compptr->v_samp_factor, + cinfo->max_v_samp_factor, + cinfo->smoothing_factor, + compptr->width_in_blocks, + cinfo->image_width); +} + +GLOBAL(void) +jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_h2v1_downsample_mips_dspr2(cinfo->image_width, + cinfo->max_v_samp_factor, + compptr->v_samp_factor, + compptr->width_in_blocks, + input_data, output_data); +} + +GLOBAL(int) +jsimd_can_h2v2_upsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_upsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_int_upsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_h2v2_upsample_mips_dspr2(cinfo->max_v_samp_factor, + cinfo->output_width, input_data, + output_data_ptr); +} + +GLOBAL(void) +jsimd_h2v1_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_h2v1_upsample_mips_dspr2(cinfo->max_v_samp_factor, + cinfo->output_width, input_data, + output_data_ptr); +} + +GLOBAL(void) +jsimd_int_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr) +{ + my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample; + + jsimd_int_upsample_mips_dspr2(upsample->h_expand[compptr->component_index], + upsample->v_expand[compptr->component_index], + input_data, output_data_ptr, + cinfo->output_width, + cinfo->max_v_samp_factor); +} + +GLOBAL(int) +jsimd_can_h2v2_fancy_upsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_fancy_upsample (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_h2v2_fancy_upsample_mips_dspr2(cinfo->max_v_samp_factor, + compptr->downsampled_width, + input_data, output_data_ptr); +} + +GLOBAL(void) +jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_h2v1_fancy_upsample_mips_dspr2(cinfo->max_v_samp_factor, + compptr->downsampled_width, + input_data, output_data_ptr); +} + +GLOBAL(int) +jsimd_can_h2v2_merged_upsample (void) +{ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_h2v1_merged_upsample (void) +{ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + void (*mipsdspr2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, + JSAMPLE *); + + switch(cinfo->out_color_space) { + case JCS_EXT_RGB: + mipsdspr2fct=jsimd_h2v2_extrgb_merged_upsample_mips_dspr2; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + mipsdspr2fct=jsimd_h2v2_extrgbx_merged_upsample_mips_dspr2; + break; + case JCS_EXT_BGR: + mipsdspr2fct=jsimd_h2v2_extbgr_merged_upsample_mips_dspr2; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + mipsdspr2fct=jsimd_h2v2_extbgrx_merged_upsample_mips_dspr2; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + mipsdspr2fct=jsimd_h2v2_extxbgr_merged_upsample_mips_dspr2; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + mipsdspr2fct=jsimd_h2v2_extxrgb_merged_upsample_mips_dspr2; + break; + default: + mipsdspr2fct=jsimd_h2v2_extrgb_merged_upsample_mips_dspr2; + break; + } + + mipsdspr2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf, + cinfo->sample_range_limit); +} + +GLOBAL(void) +jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + void (*mipsdspr2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, + JSAMPLE *); + + switch(cinfo->out_color_space) { + case JCS_EXT_RGB: + mipsdspr2fct=jsimd_h2v1_extrgb_merged_upsample_mips_dspr2; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + mipsdspr2fct=jsimd_h2v1_extrgbx_merged_upsample_mips_dspr2; + break; + case JCS_EXT_BGR: + mipsdspr2fct=jsimd_h2v1_extbgr_merged_upsample_mips_dspr2; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + mipsdspr2fct=jsimd_h2v1_extbgrx_merged_upsample_mips_dspr2; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + mipsdspr2fct=jsimd_h2v1_extxbgr_merged_upsample_mips_dspr2; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + mipsdspr2fct=jsimd_h2v1_extxrgb_merged_upsample_mips_dspr2; + break; + default: + mipsdspr2fct=jsimd_h2v1_extrgb_merged_upsample_mips_dspr2; + break; + } + + mipsdspr2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf, + cinfo->sample_range_limit); +} + +GLOBAL(int) +jsimd_can_convsamp (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_convsamp_float (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col, + DCTELEM * workspace) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_convsamp_mips_dspr2(sample_data, start_col, workspace); +} + +GLOBAL(void) +jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col, + FAST_FLOAT * workspace) +{ + if ((simd_support & JSIMD_MIPS_DSPR2)) + jsimd_convsamp_float_mips_dspr2(sample_data, start_col, workspace); +} + +GLOBAL(int) +jsimd_can_fdct_islow (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_fdct_ifast (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_fdct_float (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_fdct_islow (DCTELEM * data) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_fdct_islow_mips_dspr2(data); +} + +GLOBAL(void) +jsimd_fdct_ifast (DCTELEM * data) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_fdct_ifast_mips_dspr2(data); +} + +GLOBAL(void) +jsimd_fdct_float (FAST_FLOAT * data) +{ +} + +GLOBAL(int) +jsimd_can_quantize (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_quantize_float (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_quantize (JCOEFPTR coef_block, DCTELEM * divisors, + DCTELEM * workspace) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_quantize_mips_dspr2(coef_block, divisors, workspace); +} + +GLOBAL(void) +jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT * divisors, + FAST_FLOAT * workspace) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_quantize_float_mips_dspr2(coef_block, divisors, workspace); +} + +GLOBAL(int) +jsimd_can_idct_2x2 (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_4x4 (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_6x6 (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_12x12 (void) +{ + init_simd(); + + if (BITS_IN_JSAMPLE != 8) + return 0; + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(void) +jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_idct_2x2_mips_dspr2(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(void) +jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if (simd_support & JSIMD_MIPS_DSPR2) { + int workspace[DCTSIZE*4]; /* buffers data between passes */ + jsimd_idct_4x4_mips_dspr2(compptr->dct_table, coef_block, output_buf, + output_col, workspace); + } +} + +GLOBAL(void) +jsimd_idct_6x6 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if (simd_support & JSIMD_MIPS_DSPR2) + jsimd_idct_6x6_mips_dspr2(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(void) +jsimd_idct_12x12 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, + JSAMPARRAY output_buf, JDIMENSION output_col) +{ + if (simd_support & JSIMD_MIPS_DSPR2) { + int workspace[96]; + int output[12] = { + (int)(output_buf[0] + output_col), + (int)(output_buf[1] + output_col), + (int)(output_buf[2] + output_col), + (int)(output_buf[3] + output_col), + (int)(output_buf[4] + output_col), + (int)(output_buf[5] + output_col), + (int)(output_buf[6] + output_col), + (int)(output_buf[7] + output_col), + (int)(output_buf[8] + output_col), + (int)(output_buf[9] + output_col), + (int)(output_buf[10] + output_col), + (int)(output_buf[11] + output_col), + }; + jsimd_idct_12x12_pass1_mips_dspr2(coef_block, compptr->dct_table, + workspace); + jsimd_idct_12x12_pass2_mips_dspr2(workspace, output); + } +} + +GLOBAL(int) +jsimd_can_idct_islow (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_ifast (void) +{ + init_simd(); + + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(IFAST_MULT_TYPE) != 2) + return 0; + if (IFAST_SCALE_BITS != 2) + return 0; + + if (simd_support & JSIMD_MIPS_DSPR2) + return 1; + + return 0; +} + +GLOBAL(int) +jsimd_can_idct_float (void) +{ + init_simd(); + + return 0; +} + +GLOBAL(void) +jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if (simd_support & JSIMD_MIPS_DSPR2) { + int output[8] = { + (int)(output_buf[0] + output_col), + (int)(output_buf[1] + output_col), + (int)(output_buf[2] + output_col), + (int)(output_buf[3] + output_col), + (int)(output_buf[4] + output_col), + (int)(output_buf[5] + output_col), + (int)(output_buf[6] + output_col), + (int)(output_buf[7] + output_col), + }; + + jsimd_idct_islow_mips_dspr2(coef_block, compptr->dct_table, + output, IDCT_range_limit(cinfo)); + } +} + +GLOBAL(void) +jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + if (simd_support & JSIMD_MIPS_DSPR2) { + JCOEFPTR inptr; + IFAST_MULT_TYPE * quantptr; + DCTELEM workspace[DCTSIZE2]; /* buffers data between passes */ + + /* Pass 1: process columns from input, store into work array. */ + + inptr = coef_block; + quantptr = (IFAST_MULT_TYPE *) compptr->dct_table; + + jsimd_idct_ifast_cols_mips_dspr2(inptr, quantptr, + workspace, mips_idct_ifast_coefs); + + /* Pass 2: process rows from work array, store into output array. */ + /* Note that we must descale the results by a factor of 8 == 2**3, */ + /* and also undo the PASS1_BITS scaling. */ + + jsimd_idct_ifast_rows_mips_dspr2(workspace, output_buf, + output_col, mips_idct_ifast_coefs); + } +} + +GLOBAL(void) +jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ +} diff --git a/app/src/main/jni/libjpeg-turbo/simd/jsimd_mips_dspr2.S b/app/src/main/jni/libjpeg-turbo/simd/jsimd_mips_dspr2.S new file mode 100644 index 000000000..4572a51fa --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jsimd_mips_dspr2.S @@ -0,0 +1,4485 @@ +/* + * MIPS DSPr2 optimizations for libjpeg-turbo + * + * Copyright (C) 2013-2014, MIPS Technologies, Inc., California. + * All rights reserved. + * Authors: Teodora Novkovic (teodora.novkovic@imgtec.com) + * Darko Laus (darko.laus@imgtec.com) + * This software is provided 'as-is', without any express or implied + * warranty. In no event will the authors be held liable for any damages + * arising from the use of this software. + * + * Permission is granted to anyone to use this software for any purpose, + * including commercial applications, and to alter it and redistribute it + * freely, subject to the following restrictions: + * + * 1. The origin of this software must not be misrepresented; you must not + * claim that you wrote the original software. If you use this software + * in a product, an acknowledgment in the product documentation would be + * appreciated but is not required. + * 2. Altered source versions must be plainly marked as such, and must not be + * misrepresented as being the original software. + * 3. This notice may not be removed or altered from any source distribution. + */ + +#include "jsimd_mips_dspr2_asm.h" + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_c_null_convert_mips_dspr2) +/* + * a0 - cinfo->image_width + * a1 - input_buf + * a2 - output_buf + * a3 - output_row + * 16(sp) - num_rows + * 20(sp) - cinfo->num_components + * + * Null conversion for compression + */ + + SAVE_REGS_ON_STACK 8, s0, s1 + + lw t9, 24(sp) // t9 = num_rows + lw s0, 28(sp) // s0 = cinfo->num_components + andi t0, a0, 3 // t0 = cinfo->image_width & 3 + beqz t0, 4f // no residual + nop +0: + addiu t9, t9, -1 + bltz t9, 7f + li t1, 0 +1: + sll t3, t1, 2 + lwx t5, t3(a2) // t5 = outptr = output_buf[ci] + lw t2, 0(a1) // t2 = inptr = *input_buf + sll t4, a3, 2 + lwx t5, t4(t5) // t5 = outptr = output_buf[ci][output_row] + addu t2, t2, t1 + addu s1, t5, a0 + addu t6, t5, t0 +2: + lbu t3, 0(t2) + addiu t5, t5, 1 + sb t3, -1(t5) + bne t6, t5, 2b + addu t2, t2, s0 +3: + lbu t3, 0(t2) + addu t4, t2, s0 + addu t7, t4, s0 + addu t8, t7, s0 + addu t2, t8, s0 + lbu t4, 0(t4) + lbu t7, 0(t7) + lbu t8, 0(t8) + addiu t5, t5, 4 + sb t3, -4(t5) + sb t4, -3(t5) + sb t7, -2(t5) + bne s1, t5, 3b + sb t8, -1(t5) + addiu t1, t1, 1 + bne t1, s0, 1b + nop + addiu a1, a1, 4 + bgez t9, 0b + addiu a3, a3, 1 + b 7f + nop +4: + addiu t9, t9, -1 + bltz t9, 7f + li t1, 0 +5: + sll t3, t1, 2 + lwx t5, t3(a2) // t5 = outptr = output_buf[ci] + lw t2, 0(a1) // t2 = inptr = *input_buf + sll t4, a3, 2 + lwx t5, t4(t5) // t5 = outptr = output_buf[ci][output_row] + addu t2, t2, t1 + addu s1, t5, a0 + addu t6, t5, t0 +6: + lbu t3, 0(t2) + addu t4, t2, s0 + addu t7, t4, s0 + addu t8, t7, s0 + addu t2, t8, s0 + lbu t4, 0(t4) + lbu t7, 0(t7) + lbu t8, 0(t8) + addiu t5, t5, 4 + sb t3, -4(t5) + sb t4, -3(t5) + sb t7, -2(t5) + bne s1, t5, 6b + sb t8, -1(t5) + addiu t1, t1, 1 + bne t1, s0, 5b + nop + addiu a1, a1, 4 + bgez t9, 4b + addiu a3, a3, 1 +7: + RESTORE_REGS_FROM_STACK 8, s0, s1 + + j ra + nop + +END(jsimd_c_null_convert_mips_dspr2) + +/*****************************************************************************/ +/* + * jsimd_extrgb_ycc_convert_mips_dspr2 + * jsimd_extbgr_ycc_convert_mips_dspr2 + * jsimd_extrgbx_ycc_convert_mips_dspr2 + * jsimd_extbgrx_ycc_convert_mips_dspr2 + * jsimd_extxbgr_ycc_convert_mips_dspr2 + * jsimd_extxrgb_ycc_convert_mips_dspr2 + * + * Colorspace conversion RGB -> YCbCr + */ + +.macro GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 colorid, pixel_size, r_offs, g_offs, b_offs + +.macro DO_RGB_TO_YCC r, \ + g, \ + b, \ + inptr + lbu \r, \r_offs(\inptr) + lbu \g, \g_offs(\inptr) + lbu \b, \b_offs(\inptr) + addiu \inptr, \pixel_size +.endm + +LEAF_MIPS_DSPR2(jsimd_\colorid\()_ycc_convert_mips_dspr2) +/* + * a0 - cinfo->image_width + * a1 - input_buf + * a2 - output_buf + * a3 - output_row + * 16(sp) - num_rows + */ + + SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + lw t7, 48(sp) // t7 = num_rows + li s0, 0x4c8b // FIX(0.29900) + li s1, 0x9646 // FIX(0.58700) + li s2, 0x1d2f // FIX(0.11400) + li s3, 0xffffd4cd // -FIX(0.16874) + li s4, 0xffffab33 // -FIX(0.33126) + li s5, 0x8000 // FIX(0.50000) + li s6, 0xffff94d1 // -FIX(0.41869) + li s7, 0xffffeb2f // -FIX(0.08131) + li t8, 0x807fff // CBCR_OFFSET + ONE_HALF-1 + +0: + addiu t7, -1 // --num_rows + lw t6, 0(a1) // t6 = input_buf[0] + lw t0, 0(a2) + lw t1, 4(a2) + lw t2, 8(a2) + sll t3, a3, 2 + lwx t0, t3(t0) // t0 = output_buf[0][output_row] + lwx t1, t3(t1) // t1 = output_buf[1][output_row] + lwx t2, t3(t2) // t2 = output_buf[2][output_row] + + addu t9, t2, a0 // t9 = end address + addiu a3, 1 + +1: + DO_RGB_TO_YCC t3, t4, t5, t6 + + mtlo s5, $ac0 + mtlo t8, $ac1 + mtlo t8, $ac2 + maddu $ac0, s2, t5 + maddu $ac1, s5, t5 + maddu $ac2, s5, t3 + maddu $ac0, s0, t3 + maddu $ac1, s3, t3 + maddu $ac2, s6, t4 + maddu $ac0, s1, t4 + maddu $ac1, s4, t4 + maddu $ac2, s7, t5 + extr.w t3, $ac0, 16 + extr.w t4, $ac1, 16 + extr.w t5, $ac2, 16 + sb t3, 0(t0) + sb t4, 0(t1) + sb t5, 0(t2) + addiu t0, 1 + addiu t2, 1 + bne t2, t9, 1b + addiu t1, 1 + bgtz t7, 0b + addiu a1, 4 + + RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + j ra + nop +END(jsimd_\colorid\()_ycc_convert_mips_dspr2) + +.purgem DO_RGB_TO_YCC + +.endm + +/*------------------------------------------id -- pix R G B */ +GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extrgb, 3, 0, 1, 2 +GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extbgr, 3, 2, 1, 0 +GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extrgbx, 4, 0, 1, 2 +GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extbgrx, 4, 2, 1, 0 +GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extxbgr, 4, 3, 2, 1 +GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extxrgb, 4, 1, 2, 3 + +/*****************************************************************************/ +/* + * jsimd_ycc_extrgb_convert_mips_dspr2 + * jsimd_ycc_extbgr_convert_mips_dspr2 + * jsimd_ycc_extrgbx_convert_mips_dspr2 + * jsimd_ycc_extbgrx_convert_mips_dspr2 + * jsimd_ycc_extxbgr_convert_mips_dspr2 + * jsimd_ycc_extxrgb_convert_mips_dspr2 + * + * Colorspace conversion YCbCr -> RGB + */ + +.macro GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 colorid, pixel_size, r_offs, g_offs, b_offs, a_offs + +.macro STORE_YCC_TO_RGB scratch0 \ + scratch1 \ + scratch2 \ + outptr + sb \scratch0, \r_offs(\outptr) + sb \scratch1, \g_offs(\outptr) + sb \scratch2, \b_offs(\outptr) +.if (\pixel_size == 4) + li t0, 0xFF + sb t0, \a_offs(\outptr) +.endif + addiu \outptr, \pixel_size +.endm + +LEAF_MIPS_DSPR2(jsimd_ycc_\colorid\()_convert_mips_dspr2) +/* + * a0 - cinfo->image_width + * a1 - input_buf + * a2 - input_row + * a3 - output_buf + * 16(sp) - num_rows + */ + + SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + lw s1, 48(sp) + li t3, 0x8000 + li t4, 0x166e9 // FIX(1.40200) + li t5, 0x1c5a2 // FIX(1.77200) + li t6, 0xffff492e // -FIX(0.71414) + li t7, 0xffffa7e6 // -FIX(0.34414) + repl.ph t8, 128 + +0: + lw s0, 0(a3) + lw t0, 0(a1) + lw t1, 4(a1) + lw t2, 8(a1) + sll s5, a2, 2 + addiu s1, -1 + lwx s2, s5(t0) + lwx s3, s5(t1) + lwx s4, s5(t2) + addu t9, s2, a0 + addiu a2, 1 + +1: + lbu s7, 0(s4) // cr + lbu s6, 0(s3) // cb + lbu s5, 0(s2) // y + addiu s2, 1 + addiu s4, 1 + addiu s7, -128 + addiu s6, -128 + mul t2, t7, s6 + mul t0, t6, s7 // Crgtab[cr] + sll s7, 15 + mulq_rs.w t1, t4, s7 // Crrtab[cr] + sll s6, 15 + addu t2, t3 // Cbgtab[cb] + addu t2, t0 + + mulq_rs.w t0, t5, s6 // Cbbtab[cb] + sra t2, 16 + addu t1, s5 + addu t2, s5 // add y + ins t2, t1, 16, 16 + subu.ph t2, t2, t8 + addu t0, s5 + shll_s.ph t2, t2, 8 + subu t0, 128 + shra.ph t2, t2, 8 + shll_s.w t0, t0, 24 + addu.ph t2, t2, t8 // clip & store + sra t0, t0, 24 + sra t1, t2, 16 + addiu t0, 128 + + STORE_YCC_TO_RGB t1, t2, t0, s0 + + bne s2, t9, 1b + addiu s3, 1 + bgtz s1, 0b + addiu a3, 4 + + RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + j ra + nop +END(jsimd_ycc_\colorid\()_convert_mips_dspr2) + +.purgem STORE_YCC_TO_RGB + +.endm + +/*------------------------------------------id -- pix R G B A */ +GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extrgb, 3, 0, 1, 2, 3 +GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extbgr, 3, 2, 1, 0, 3 +GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extrgbx, 4, 0, 1, 2, 3 +GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extbgrx, 4, 2, 1, 0, 3 +GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extxbgr, 4, 3, 2, 1, 0 +GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extxrgb, 4, 1, 2, 3, 0 + +/*****************************************************************************/ +/* + * jsimd_extrgb_gray_convert_mips_dspr2 + * jsimd_extbgr_gray_convert_mips_dspr2 + * jsimd_extrgbx_gray_convert_mips_dspr2 + * jsimd_extbgrx_gray_convert_mips_dspr2 + * jsimd_extxbgr_gray_convert_mips_dspr2 + * jsimd_extxrgb_gray_convert_mips_dspr2 + * + * Colorspace conversion RGB -> GRAY + */ + +.macro GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 colorid, pixel_size, r_offs, g_offs, b_offs + +.macro DO_RGB_TO_GRAY r, \ + g, \ + b, \ + inptr + lbu \r, \r_offs(\inptr) + lbu \g, \g_offs(\inptr) + lbu \b, \b_offs(\inptr) + addiu \inptr, \pixel_size +.endm + +LEAF_MIPS_DSPR2(jsimd_\colorid\()_gray_convert_mips_dspr2) +/* + * a0 - cinfo->image_width + * a1 - input_buf + * a2 - output_buf + * a3 - output_row + * 16(sp) - num_rows + */ + + SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + li s0, 0x4c8b // s0 = FIX(0.29900) + li s1, 0x9646 // s1 = FIX(0.58700) + li s2, 0x1d2f // s2 = FIX(0.11400) + li s7, 0x8000 // s7 = FIX(0.50000) + lw s6, 48(sp) + andi t7, a0, 3 + +0: + addiu s6, -1 // s6 = num_rows + lw t0, 0(a1) + lw t1, 0(a2) + sll t3, a3, 2 + lwx t1, t3(t1) + addiu a3, 1 + addu t9, t1, a0 + subu t8, t9, t7 + beq t1, t8, 2f + nop + +1: + DO_RGB_TO_GRAY t3, t4, t5, t0 + DO_RGB_TO_GRAY s3, s4, s5, t0 + + mtlo s7, $ac0 + maddu $ac0, s2, t5 + maddu $ac0, s1, t4 + maddu $ac0, s0, t3 + mtlo s7, $ac1 + maddu $ac1, s2, s5 + maddu $ac1, s1, s4 + maddu $ac1, s0, s3 + extr.w t6, $ac0, 16 + + DO_RGB_TO_GRAY t3, t4, t5, t0 + DO_RGB_TO_GRAY s3, s4, s5, t0 + + mtlo s7, $ac0 + maddu $ac0, s2, t5 + maddu $ac0, s1, t4 + extr.w t2, $ac1, 16 + maddu $ac0, s0, t3 + mtlo s7, $ac1 + maddu $ac1, s2, s5 + maddu $ac1, s1, s4 + maddu $ac1, s0, s3 + extr.w t5, $ac0, 16 + sb t6, 0(t1) + sb t2, 1(t1) + extr.w t3, $ac1, 16 + addiu t1, 4 + sb t5, -2(t1) + sb t3, -1(t1) + bne t1, t8, 1b + nop + +2: + beqz t7, 4f + nop + +3: + DO_RGB_TO_GRAY t3, t4, t5, t0 + + mtlo s7, $ac0 + maddu $ac0, s2, t5 + maddu $ac0, s1, t4 + maddu $ac0, s0, t3 + extr.w t6, $ac0, 16 + sb t6, 0(t1) + addiu t1, 1 + bne t1, t9, 3b + nop + +4: + bgtz s6, 0b + addiu a1, 4 + + RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + j ra + nop +END(jsimd_\colorid\()_gray_convert_mips_dspr2) + +.purgem DO_RGB_TO_GRAY + +.endm + +/*------------------------------------------id -- pix R G B */ +GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extrgb, 3, 0, 1, 2 +GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extbgr, 3, 2, 1, 0 +GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extrgbx, 4, 0, 1, 2 +GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extbgrx, 4, 2, 1, 0 +GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extxbgr, 4, 3, 2, 1 +GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extxrgb, 4, 1, 2, 3 +/*****************************************************************************/ +/* + * jsimd_h2v2_merged_upsample_mips_dspr2 + * jsimd_h2v2_extrgb_merged_upsample_mips_dspr2 + * jsimd_h2v2_extrgbx_merged_upsample_mips_dspr2 + * jsimd_h2v2_extbgr_merged_upsample_mips_dspr2 + * jsimd_h2v2_extbgrx_merged_upsample_mips_dspr2 + * jsimd_h2v2_extxbgr_merged_upsample_mips_dspr2 + * jsimd_h2v2_extxrgb_merged_upsample_mips_dspr2 + * + * Merged h2v2 upsample routines + */ +.macro GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 colorid, \ + pixel_size, \ + r1_offs, \ + g1_offs, \ + b1_offs, \ + a1_offs, \ + r2_offs, \ + g2_offs, \ + b2_offs, \ + a2_offs + +.macro STORE_H2V2_2_PIXELS scratch0 \ + scratch1 \ + scratch2 \ + scratch3 \ + scratch4 \ + scratch5 \ + outptr + sb \scratch0, \r1_offs(\outptr) + sb \scratch1, \g1_offs(\outptr) + sb \scratch2, \b1_offs(\outptr) + sb \scratch3, \r2_offs(\outptr) + sb \scratch4, \g2_offs(\outptr) + sb \scratch5, \b2_offs(\outptr) +.if (\pixel_size == 8) + li \scratch0, 0xFF + sb \scratch0, \a1_offs(\outptr) + sb \scratch0, \a2_offs(\outptr) +.endif + addiu \outptr, \pixel_size +.endm + +.macro STORE_H2V2_1_PIXEL scratch0 \ + scratch1 \ + scratch2 \ + outptr + sb \scratch0, \r1_offs(\outptr) + sb \scratch1, \g1_offs(\outptr) + sb \scratch2, \b1_offs(\outptr) + +.if (\pixel_size == 8) + li t0, 0xFF + sb t0, \a1_offs(\outptr) +.endif +.endm + +LEAF_MIPS_DSPR2(jsimd_h2v2_\colorid\()_merged_upsample_mips_dspr2) +/* + * a0 - cinfo->output_width + * a1 - input_buf + * a2 - in_row_group_ctr + * a3 - output_buf + * 16(sp) - cinfo->sample_range_limit + */ + + SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra + + lw t9, 56(sp) // cinfo->sample_range_limit + lw v0, 0(a1) + lw v1, 4(a1) + lw t0, 8(a1) + sll t1, a2, 3 + addiu t2, t1, 4 + sll t3, a2, 2 + lw t4, 0(a3) // t4 = output_buf[0] + lwx t1, t1(v0) // t1 = input_buf[0][in_row_group_ctr*2] + lwx t2, t2(v0) // t2 = input_buf[0][in_row_group_ctr*2 + 1] + lwx t5, t3(v1) // t5 = input_buf[1][in_row_group_ctr] + lwx t6, t3(t0) // t6 = input_buf[2][in_row_group_ctr] + lw t7, 4(a3) // t7 = output_buf[1] + li s1, 0xe6ea + addiu t8, s1, 0x7fff // t8 = 0x166e9 [FIX(1.40200)] + addiu s0, t8, 0x5eb9 // s0 = 0x1c5a2 [FIX(1.77200)] + addiu s1, zero, 0xa7e6 // s4 = 0xffffa7e6 [-FIX(0.34414)] + xori s2, s1, 0xeec8 // s3 = 0xffff492e [-FIX(0.71414)] + srl t3, a0, 1 + blez t3, 2f + addu t0, t5, t3 // t0 = end address + 1: + lbu t3, 0(t5) + lbu s3, 0(t6) + addiu t5, t5, 1 + addiu t3, t3, -128 // (cb - 128) + addiu s3, s3, -128 // (cr - 128) + mult $ac1, s1, t3 + madd $ac1, s2, s3 + sll s3, s3, 15 + sll t3, t3, 15 + mulq_rs.w s4, t8, s3 // s4 = (C1 * cr + ONE_HALF)>> SCALEBITS + extr_r.w s5, $ac1, 16 + mulq_rs.w s6, s0, t3 // s6 = (C2 * cb + ONE_HALF)>> SCALEBITS + lbu v0, 0(t1) + addiu t6, t6, 1 + addiu t1, t1, 2 + addu t3, v0, s4 // y+cred + addu s3, v0, s5 // y+cgreen + addu v1, v0, s6 // y+cblue + addu t3, t9, t3 // y+cred + addu s3, t9, s3 // y+cgreen + addu v1, t9, v1 // y+cblue + lbu AT, 0(t3) + lbu s7, 0(s3) + lbu ra, 0(v1) + lbu v0, -1(t1) + addu t3, v0, s4 // y+cred + addu s3, v0, s5 // y+cgreen + addu v1, v0, s6 // y+cblue + addu t3, t9, t3 // y+cred + addu s3, t9, s3 // y+cgreen + addu v1, t9, v1 // y+cblue + lbu t3, 0(t3) + lbu s3, 0(s3) + lbu v1, 0(v1) + lbu v0, 0(t2) + + STORE_H2V2_2_PIXELS AT, s7, ra, t3, s3, v1, t4 + + addu t3, v0, s4 // y+cred + addu s3, v0, s5 // y+cgreen + addu v1, v0, s6 // y+cblue + addu t3, t9, t3 // y+cred + addu s3, t9, s3 // y+cgreen + addu v1, t9, v1 // y+cblue + lbu AT, 0(t3) + lbu s7, 0(s3) + lbu ra, 0(v1) + lbu v0, 1(t2) + addiu t2, t2, 2 + addu t3, v0, s4 // y+cred + addu s3, v0, s5 // y+cgreen + addu v1, v0, s6 // y+cblue + addu t3, t9, t3 // y+cred + addu s3, t9, s3 // y+cgreen + addu v1, t9, v1 // y+cblue + lbu t3, 0(t3) + lbu s3, 0(s3) + lbu v1, 0(v1) + + STORE_H2V2_2_PIXELS AT, s7, ra, t3, s3, v1, t7 + + bne t0, t5, 1b + nop +2: + andi t0, a0, 1 + beqz t0, 4f + lbu t3, 0(t5) + lbu s3, 0(t6) + addiu t3, t3, -128 // (cb - 128) + addiu s3, s3, -128 // (cr - 128) + mult $ac1, s1, t3 + madd $ac1, s2, s3 + sll s3, s3, 15 + sll t3, t3, 15 + lbu v0, 0(t1) + extr_r.w s5, $ac1, 16 + mulq_rs.w s4, t8, s3 // s4 = (C1 * cr + ONE_HALF)>> SCALEBITS + mulq_rs.w s6, s0, t3 // s6 = (C2 * cb + ONE_HALF)>> SCALEBITS + addu t3, v0, s4 // y+cred + addu s3, v0, s5 // y+cgreen + addu v1, v0, s6 // y+cblue + addu t3, t9, t3 // y+cred + addu s3, t9, s3 // y+cgreen + addu v1, t9, v1 // y+cblue + lbu t3, 0(t3) + lbu s3, 0(s3) + lbu v1, 0(v1) + lbu v0, 0(t2) + + STORE_H2V2_1_PIXEL t3, s3, v1, t4 + + addu t3, v0, s4 // y+cred + addu s3, v0, s5 // y+cgreen + addu v1, v0, s6 // y+cblue + addu t3, t9, t3 // y+cred + addu s3, t9, s3 // y+cgreen + addu v1, t9, v1 // y+cblue + lbu t3, 0(t3) + lbu s3, 0(s3) + lbu v1, 0(v1) + + STORE_H2V2_1_PIXEL t3, s3, v1, t7 +4: + RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra + + j ra + nop + +END(jsimd_h2v2_\colorid\()_merged_upsample_mips_dspr2) + +.purgem STORE_H2V2_1_PIXEL +.purgem STORE_H2V2_2_PIXELS +.endm + +/*-----------------------------------------id -- pix R1 G1 B1 A1 R2 G2 B2 A2 */ +GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extrgb, 6, 0, 1, 2, 6, 3, 4, 5, 6 +GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extbgr, 6, 2, 1, 0, 3, 5, 4, 3, 6 +GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extrgbx, 8, 0, 1, 2, 3, 4, 5, 6, 7 +GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extbgrx, 8, 2, 1, 0, 3, 6, 5, 4, 7 +GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extxbgr, 8, 3, 2, 1, 0, 7, 6, 5, 4 +GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extxrgb, 8, 1, 2, 3, 0, 5, 6, 7, 4 +/*****************************************************************************/ +/* + * jsimd_h2v1_merged_upsample_mips_dspr2 + * jsimd_h2v1_extrgb_merged_upsample_mips_dspr2 + * jsimd_h2v1_extrgbx_merged_upsample_mips_dspr2 + * jsimd_h2v1_extbgr_merged_upsample_mips_dspr2 + * jsimd_h2v1_extbgrx_merged_upsample_mips_dspr2 + * jsimd_h2v1_extxbgr_merged_upsample_mips_dspr2 + * jsimd_h2v1_extxrgb_merged_upsample_mips_dspr2 + * + * Merged h2v1 upsample routines + */ + +.macro GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 colorid, \ + pixel_size, \ + r1_offs, \ + g1_offs, \ + b1_offs, \ + a1_offs, \ + r2_offs, \ + g2_offs, \ + b2_offs, \ + a2_offs + +.macro STORE_H2V1_2_PIXELS scratch0 \ + scratch1 \ + scratch2 \ + scratch3 \ + scratch4 \ + scratch5 \ + outptr + sb \scratch0, \r1_offs(\outptr) + sb \scratch1, \g1_offs(\outptr) + sb \scratch2, \b1_offs(\outptr) + sb \scratch3, \r2_offs(\outptr) + sb \scratch4, \g2_offs(\outptr) + sb \scratch5, \b2_offs(\outptr) +.if (\pixel_size == 8) + li t0, 0xFF + sb t0, \a1_offs(\outptr) + sb t0, \a2_offs(\outptr) +.endif + addiu \outptr, \pixel_size +.endm + +.macro STORE_H2V1_1_PIXEL scratch0 \ + scratch1 \ + scratch2 \ + outptr + sb \scratch0, \r1_offs(\outptr) + sb \scratch1, \g1_offs(\outptr) + sb \scratch2, \b1_offs(\outptr) +.if (\pixel_size == 8) + li t0, 0xFF + sb t0, \a1_offs(\outptr) +.endif +.endm + +LEAF_MIPS_DSPR2(jsimd_h2v1_\colorid\()_merged_upsample_mips_dspr2) +/* + * a0 - cinfo->output_width + * a1 - input_buf + * a2 - in_row_group_ctr + * a3 - output_buf + * 16(sp) - range_limit + */ + + SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra + + li t0, 0xe6ea + lw t1, 0(a1) // t1 = input_buf[0] + lw t2, 4(a1) // t2 = input_buf[1] + lw t3, 8(a1) // t3 = input_buf[2] + lw t8, 56(sp) // t8 = range_limit + addiu s1, t0, 0x7fff // s1 = 0x166e9 [FIX(1.40200)] + addiu s2, s1, 0x5eb9 // s2 = 0x1c5a2 [FIX(1.77200)] + addiu s0, t0, 0x9916 // s0 = 0x8000 + addiu s4, zero, 0xa7e6 // s4 = 0xffffa7e6 [-FIX(0.34414)] + xori s3, s4, 0xeec8 // s3 = 0xffff492e [-FIX(0.71414)] + srl t0, a0, 1 + sll t4, a2, 2 + lwx s5, t4(t1) // s5 = inptr0 + lwx s6, t4(t2) // s6 = inptr1 + lwx s7, t4(t3) // s7 = inptr2 + lw t7, 0(a3) // t7 = outptr + blez t0, 2f + addu t9, s6, t0 // t9 = end address +1: + lbu t2, 0(s6) // t2 = cb + lbu t0, 0(s7) // t0 = cr + lbu t1, 0(s5) // t1 = y + addiu t2, t2, -128 // t2 = cb - 128 + addiu t0, t0, -128 // t0 = cr - 128 + mult $ac1, s4, t2 + madd $ac1, s3, t0 + sll t0, t0, 15 + sll t2, t2, 15 + mulq_rs.w t0, s1, t0 // t0 = (C1*cr + ONE_HALF)>> SCALEBITS + extr_r.w t5, $ac1, 16 + mulq_rs.w t6, s2, t2 // t6 = (C2*cb + ONE_HALF)>> SCALEBITS + addiu s7, s7, 1 + addiu s6, s6, 1 + addu t2, t1, t0 // t2 = y + cred + addu t3, t1, t5 // t3 = y + cgreen + addu t4, t1, t6 // t4 = y + cblue + addu t2, t8, t2 + addu t3, t8, t3 + addu t4, t8, t4 + lbu t1, 1(s5) + lbu v0, 0(t2) + lbu v1, 0(t3) + lbu ra, 0(t4) + addu t2, t1, t0 + addu t3, t1, t5 + addu t4, t1, t6 + addu t2, t8, t2 + addu t3, t8, t3 + addu t4, t8, t4 + lbu t2, 0(t2) + lbu t3, 0(t3) + lbu t4, 0(t4) + + STORE_H2V1_2_PIXELS v0, v1, ra, t2, t3, t4, t7 + + bne t9, s6, 1b + addiu s5, s5, 2 +2: + andi t0, a0, 1 + beqz t0, 4f + nop +3: + lbu t2, 0(s6) + lbu t0, 0(s7) + lbu t1, 0(s5) + addiu t2, t2, -128 //(cb - 128) + addiu t0, t0, -128 //(cr - 128) + mul t3, s4, t2 + mul t4, s3, t0 + sll t0, t0, 15 + sll t2, t2, 15 + mulq_rs.w t0, s1, t0 // (C1*cr + ONE_HALF)>> SCALEBITS + mulq_rs.w t6, s2, t2 // (C2*cb + ONE_HALF)>> SCALEBITS + addu t3, t3, s0 + addu t3, t4, t3 + sra t5, t3, 16 // (C4*cb + ONE_HALF + C3*cr)>> SCALEBITS + addu t2, t1, t0 // y + cred + addu t3, t1, t5 // y + cgreen + addu t4, t1, t6 // y + cblue + addu t2, t8, t2 + addu t3, t8, t3 + addu t4, t8, t4 + lbu t2, 0(t2) + lbu t3, 0(t3) + lbu t4, 0(t4) + + STORE_H2V1_1_PIXEL t2, t3, t4, t7 +4: + RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra + + j ra + nop + +END(jsimd_h2v1_\colorid\()_merged_upsample_mips_dspr2) + +.purgem STORE_H2V1_1_PIXEL +.purgem STORE_H2V1_2_PIXELS +.endm + +/*-----------------------------------------id -- pix R1 G1 B1 A1 R2 G2 B2 A2 */ +GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extrgb, 6, 0, 1, 2, 6, 3, 4, 5, 6 +GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extbgr, 6, 2, 1, 0, 3, 5, 4, 3, 6 +GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extrgbx, 8, 0, 1, 2, 3, 4, 5, 6, 7 +GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extbgrx, 8, 2, 1, 0, 3, 6, 5, 4, 7 +GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extxbgr, 8, 3, 2, 1, 0, 7, 6, 5, 4 +GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extxrgb, 8, 1, 2, 3, 0, 5, 6, 7, 4 +/*****************************************************************************/ +/* + * jsimd_h2v2_fancy_upsample_mips_dspr2 + * + * Fancy processing for the common case of 2:1 horizontal and 2:1 vertical. + */ +LEAF_MIPS_DSPR2(jsimd_h2v2_fancy_upsample_mips_dspr2) +/* + * a0 - cinfo->max_v_samp_factor + * a1 - downsampled_width + * a2 - input_data + * a3 - output_data_ptr + */ + + SAVE_REGS_ON_STACK 24, s0, s1, s2, s3, s4, s5 + + li s4, 0 + lw s2, 0(a3) // s2 = *output_data_ptr +0: + li t9, 2 + lw s1, -4(a2) // s1 = inptr1 + +1: + lw s0, 0(a2) // s0 = inptr0 + lwx s3, s4(s2) + addiu s5, a1, -2 // s5 = downsampled_width - 2 + srl t4, s5, 1 + sll t4, t4, 1 + lbu t0, 0(s0) + lbu t1, 1(s0) + lbu t2, 0(s1) + lbu t3, 1(s1) + addiu s0, 2 + addiu s1, 2 + addu t8, s0, t4 // t8 = end address + andi s5, s5, 1 // s5 = residual + sll t4, t0, 1 + sll t6, t1, 1 + addu t0, t0, t4 // t0 = (*inptr0++) * 3 + addu t1, t1, t6 // t1 = (*inptr0++) * 3 + addu t7, t0, t2 // t7 = thiscolsum + addu t6, t1, t3 // t5 = nextcolsum + sll t0, t7, 2 // t0 = thiscolsum * 4 + subu t1, t0, t7 // t1 = thiscolsum * 3 + shra_r.w t0, t0, 4 + addiu t1, 7 + addu t1, t1, t6 + srl t1, t1, 4 + sb t0, 0(s3) + sb t1, 1(s3) + addiu s3, 2 +2: + lh t0, 0(s0) // t0 = A3|A2 + lh t2, 0(s1) // t2 = B3|B2 + addiu s0, 2 + addiu s1, 2 + preceu.ph.qbr t0, t0 // t0 = 0|A3|0|A2 + preceu.ph.qbr t2, t2 // t2 = 0|B3|0|B2 + shll.ph t1, t0, 1 + sll t3, t6, 1 + addu.ph t0, t1, t0 // t0 = A3*3|A2*3 + addu t3, t3, t6 // t3 = this * 3 + addu.ph t0, t0, t2 // t0 = next2|next1 + addu t1, t3, t7 + andi t7, t0, 0xFFFF // t7 = next1 + sll t2, t7, 1 + addu t2, t7, t2 // t2 = next1*3 + addu t4, t2, t6 + srl t6, t0, 16 // t6 = next2 + shra_r.w t1, t1, 4 // t1 = (this*3 + last + 8) >> 4 + addu t0, t3, t7 + addiu t0, 7 + srl t0, t0, 4 // t0 = (this*3 + next1 + 7) >> 4 + shra_r.w t4, t4, 4 // t3 = (next1*3 + this + 8) >> 4 + addu t2, t2, t6 + addiu t2, 7 + srl t2, t2, 4 // t2 = (next1*3 + next2 + 7) >> 4 + sb t1, 0(s3) + sb t0, 1(s3) + sb t4, 2(s3) + sb t2, 3(s3) + bne t8, s0, 2b + addiu s3, 4 + beqz s5, 4f + addu t8, s0, s5 +3: + lbu t0, 0(s0) + lbu t2, 0(s1) + addiu s0, 1 + addiu s1, 1 + sll t3, t6, 1 + sll t1, t0, 1 + addu t1, t0, t1 // t1 = inptr0 * 3 + addu t3, t3, t6 // t3 = thiscolsum * 3 + addu t5, t1, t2 + addu t1, t3, t7 + shra_r.w t1, t1, 4 + addu t0, t3, t5 + addiu t0, 7 + srl t0, t0, 4 + sb t1, 0(s3) + sb t0, 1(s3) + addiu s3, 2 + move t7, t6 + bne t8, s0, 3b + move t6, t5 +4: + sll t0, t6, 2 // t0 = thiscolsum * 4 + subu t1, t0, t6 // t1 = thiscolsum * 3 + addu t1, t1, t7 + addiu s4, 4 + shra_r.w t1, t1, 4 + addiu t0, 7 + srl t0, t0, 4 + sb t1, 0(s3) + sb t0, 1(s3) + addiu t9, -1 + addiu s3, 2 + bnez t9, 1b + lw s1, 4(a2) + srl t0, s4, 2 + subu t0, a0, t0 + bgtz t0, 0b + addiu a2, 4 + + RESTORE_REGS_FROM_STACK 24, s0, s1, s2, s3, s4, s5 + + j ra + nop +END(jsimd_h2v2_fancy_upsample_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_h2v1_fancy_upsample_mips_dspr2) +/* + * a0 - cinfo->max_v_samp_factor + * a1 - downsampled_width + * a2 - input_data + * a3 - output_data_ptr + */ + + SAVE_REGS_ON_STACK 16, s0, s1, s2, s3 + + .set at + + beqz a0, 3f + sll t0, a0, 2 + lw s1, 0(a3) + li s3, 0x10001 + addu s0, s1, t0 +0: + addiu t8, a1, -2 + srl t9, t8, 2 + lw t7, 0(a2) + lw s2, 0(s1) + lbu t0, 0(t7) + lbu t1, 1(t7) // t1 = inptr[1] + sll t2, t0, 1 + addu t2, t2, t0 // t2 = invalue*3 + addu t2, t2, t1 + shra_r.w t2, t2, 2 + sb t0, 0(s2) + sb t2, 1(s2) + beqz t9, 11f + addiu s2, 2 +1: + ulw t0, 0(t7) // t0 = |P3|P2|P1|P0| + ulw t1, 1(t7) + ulh t2, 4(t7) // t2 = |0|0|P5|P4| + preceu.ph.qbl t3, t0 // t3 = |0|P3|0|P2| + preceu.ph.qbr t0, t0 // t0 = |0|P1|0|P0| + preceu.ph.qbr t2, t2 // t2 = |0|P5|0|P4| + preceu.ph.qbl t4, t1 // t4 = |0|P4|0|P3| + preceu.ph.qbr t1, t1 // t1 = |0|P2|0|P1| + shll.ph t5, t4, 1 + shll.ph t6, t1, 1 + addu.ph t5, t5, t4 // t5 = |P4*3|P3*3| + addu.ph t6, t6, t1 // t6 = |P2*3|P1*3| + addu.ph t4, t3, s3 + addu.ph t0, t0, s3 + addu.ph t4, t4, t5 + addu.ph t0, t0, t6 + shrl.ph t4, t4, 2 // t4 = |0|P3|0|P2| + shrl.ph t0, t0, 2 // t0 = |0|P1|0|P0| + addu.ph t2, t2, t5 + addu.ph t3, t3, t6 + shra_r.ph t2, t2, 2 // t2 = |0|P5|0|P4| + shra_r.ph t3, t3, 2 // t3 = |0|P3|0|P2| + shll.ph t2, t2, 8 + shll.ph t3, t3, 8 + or t2, t4, t2 + or t3, t3, t0 + addiu t9, -1 + usw t3, 0(s2) + usw t2, 4(s2) + addiu s2, 8 + bgtz t9, 1b + addiu t7, 4 +11: + andi t8, 3 + beqz t8, 22f + addiu t7, 1 + +2: + lbu t0, 0(t7) + addiu t7, 1 + sll t1, t0, 1 + addu t2, t0, t1 // t2 = invalue + lbu t3, -2(t7) + lbu t4, 0(t7) + addiu t3, 1 + addiu t4, 2 + addu t3, t3, t2 + addu t4, t4, t2 + srl t3, 2 + srl t4, 2 + sb t3, 0(s2) + sb t4, 1(s2) + addiu t8, -1 + bgtz t8, 2b + addiu s2, 2 + +22: + lbu t0, 0(t7) + lbu t2, -1(t7) + sll t1, t0, 1 + addu t1, t1, t0 // t1 = invalue * 3 + addu t1, t1, t2 + addiu t1, 1 + srl t1, t1, 2 + sb t1, 0(s2) + sb t0, 1(s2) + addiu s1, 4 + bne s1, s0, 0b + addiu a2, 4 +3: + RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3 + + j ra + nop +END(jsimd_h2v1_fancy_upsample_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_h2v1_downsample_mips_dspr2) +/* + * a0 - cinfo->image_width + * a1 - cinfo->max_v_samp_factor + * a2 - compptr->v_samp_factor + * a3 - compptr->width_in_blocks + * 16(sp) - input_data + * 20(sp) - output_data + */ + .set at + + SAVE_REGS_ON_STACK 24, s0, s1, s2, s3, s4 + + beqz a2, 7f + lw s1, 44(sp) // s1 = output_data + lw s0, 40(sp) // s0 = input_data + srl s2, a0, 2 + andi t9, a0, 2 + srl t7, t9, 1 + addu s2, t7, s2 + sll t0, a3, 3 // t0 = width_in_blocks*DCT + srl t7, t0, 1 + subu s2, t7, s2 +0: + andi t6, a0, 1 // t6 = temp_index + addiu t6, -1 + lw t4, 0(s1) // t4 = outptr + lw t5, 0(s0) // t5 = inptr0 + li s3, 0 // s3 = bias + srl t7, a0, 1 // t7 = image_width1 + srl s4, t7, 2 + andi t8, t7, 3 +1: + ulhu t0, 0(t5) + ulhu t1, 2(t5) + ulhu t2, 4(t5) + ulhu t3, 6(t5) + raddu.w.qb t0, t0 + raddu.w.qb t1, t1 + raddu.w.qb t2, t2 + raddu.w.qb t3, t3 + shra.ph t0, t0, 1 + shra_r.ph t1, t1, 1 + shra.ph t2, t2, 1 + shra_r.ph t3, t3, 1 + sb t0, 0(t4) + sb t1, 1(t4) + sb t2, 2(t4) + sb t3, 3(t4) + addiu s4, -1 + addiu t4, 4 + bgtz s4, 1b + addiu t5, 8 + beqz t8, 3f + addu s4, t4, t8 +2: + ulhu t0, 0(t5) + raddu.w.qb t0, t0 + addqh.w t0, t0, s3 + xori s3, s3, 1 + sb t0, 0(t4) + addiu t4, 1 + bne t4, s4, 2b + addiu t5, 2 +3: + lbux t1, t6(t5) + sll t1, 1 + addqh.w t2, t1, s3 // t2 = pixval1 + xori s3, s3, 1 + addqh.w t3, t1, s3 // t3 = pixval2 + blez s2, 5f + append t3, t2, 8 + addu t5, t4, s2 // t5 = loop_end2 +4: + ush t3, 0(t4) + addiu s2, -1 + bgtz s2, 4b + addiu t4, 2 +5: + beqz t9, 6f + nop + sb t2, 0(t4) +6: + addiu s1, 4 + addiu a2, -1 + bnez a2, 0b + addiu s0, 4 +7: + RESTORE_REGS_FROM_STACK 24, s0, s1, s2, s3, s4 + + j ra + nop +END(jsimd_h2v1_downsample_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_h2v2_downsample_mips_dspr2) + +/* + * a0 - cinfo->image_width + * a1 - cinfo->max_v_samp_factor + * a2 - compptr->v_samp_factor + * a3 - compptr->width_in_blocks + * 16(sp) - input_data + * 20(sp) - output_data + */ + .set at + SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + beqz a2, 8f + lw s1, 52(sp) // s1 = output_data + lw s0, 48(sp) // s0 = input_data + + andi t6, a0, 1 // t6 = temp_index + addiu t6, -1 + srl t7, a0, 1 // t7 = image_width1 + srl s4, t7, 2 + andi t8, t7, 3 + andi t9, a0, 2 + srl s2, a0, 2 + srl t7, t9, 1 + addu s2, t7, s2 + sll t0, a3, 3 // s2 = width_in_blocks*DCT + srl t7, t0, 1 + subu s2, t7, s2 +0: + lw t4, 0(s1) // t4 = outptr + lw t5, 0(s0) // t5 = inptr0 + lw s7, 4(s0) // s7 = inptr1 + li s6, 1 // s6 = bias +2: + ulw t0, 0(t5) // t0 = |P3|P2|P1|P0| + ulw t1, 0(s7) // t1 = |Q3|Q2|Q1|Q0| + ulw t2, 4(t5) + ulw t3, 4(s7) + precrq.ph.w t7, t0, t1 // t2 = |P3|P2|Q3|Q2| + ins t0, t1, 16, 16 // t0 = |Q1|Q0|P1|P0| + raddu.w.qb t1, t7 + raddu.w.qb t0, t0 + shra_r.w t1, t1, 2 + addiu t0, 1 + srl t0, 2 + precrq.ph.w t7, t2, t3 + ins t2, t3, 16, 16 + raddu.w.qb t7, t7 + raddu.w.qb t2, t2 + shra_r.w t7, t7, 2 + addiu t2, 1 + srl t2, 2 + sb t0, 0(t4) + sb t1, 1(t4) + sb t2, 2(t4) + sb t7, 3(t4) + addiu t4, 4 + addiu t5, 8 + addiu s4, s4, -1 + bgtz s4, 2b + addiu s7, 8 + beqz t8, 4f + addu t8, t4, t8 +3: + ulhu t0, 0(t5) + ulhu t1, 0(s7) + ins t0, t1, 16, 16 + raddu.w.qb t0, t0 + addu t0, t0, s6 + srl t0, 2 + xori s6, s6, 3 + sb t0, 0(t4) + addiu t5, 2 + addiu t4, 1 + bne t8, t4, 3b + addiu s7, 2 +4: + lbux t1, t6(t5) + sll t1, 1 + lbux t0, t6(s7) + sll t0, 1 + addu t1, t1, t0 + addu t3, t1, s6 + srl t0, t3, 2 // t2 = pixval1 + xori s6, s6, 3 + addu t2, t1, s6 + srl t1, t2, 2 // t3 = pixval2 + blez s2, 6f + append t1, t0, 8 +5: + ush t1, 0(t4) + addiu s2, -1 + bgtz s2, 5b + addiu t4, 2 +6: + beqz t9, 7f + nop + sb t0, 0(t4) +7: + addiu s1, 4 + addiu a2, -1 + bnez a2, 0b + addiu s0, 8 +8: + RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + j ra + nop +END(jsimd_h2v2_downsample_mips_dspr2) +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_h2v2_smooth_downsample_mips_dspr2) +/* + * a0 - input_data + * a1 - output_data + * a2 - compptr->v_samp_factor + * a3 - cinfo->max_v_samp_factor + * 16(sp) - cinfo->smoothing_factor + * 20(sp) - compptr->width_in_blocks + * 24(sp) - cinfo->image_width + */ + + .set at + + SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + lw s7, 52(sp) // compptr->width_in_blocks + lw s0, 56(sp) // cinfo->image_width + lw s6, 48(sp) // cinfo->smoothing_factor + sll s7, 3 // output_cols = width_in_blocks * DCTSIZE + sll v0, s7, 1 + subu v0, v0, s0 + blez v0, 2f + move v1, zero + addiu t0, a3, 2 // t0 = cinfo->max_v_samp_factor + 2 +0: + addiu t1, a0, -4 + sll t2, v1, 2 + lwx t1, t2(t1) + move t3, v0 + addu t1, t1, s0 + lbu t2, -1(t1) +1: + addiu t3, t3, -1 + sb t2, 0(t1) + bgtz t3, 1b + addiu t1, t1, 1 + addiu v1, v1, 1 + bne v1, t0, 0b + nop +2: + li v0, 80 + mul v0, s6, v0 + li v1, 16384 + move t4, zero + move t5, zero + subu t6, v1, v0 // t6 = 16384 - tmp_smoot_f * 80 + sll t7, s6, 4 // t7 = tmp_smoot_f * 16 +3: +/* Special case for first column: pretend column -1 is same as column 0 */ + sll v0, t4, 2 + lwx t8, v0(a1) // outptr = output_data[outrow] + sll v1, t5, 2 + addiu t9, v1, 4 + addiu s0, v1, -4 + addiu s1, v1, 8 + lwx s2, v1(a0) // inptr0 = input_data[inrow] + lwx t9, t9(a0) // inptr1 = input_data[inrow+1] + lwx s0, s0(a0) // above_ptr = input_data[inrow-1] + lwx s1, s1(a0) // below_ptr = input_data[inrow+2] + lh v0, 0(s2) + lh v1, 0(t9) + lh t0, 0(s0) + lh t1, 0(s1) + ins v0, v1, 16, 16 + ins t0, t1, 16, 16 + raddu.w.qb t2, v0 + raddu.w.qb s3, t0 + lbu v0, 0(s2) + lbu v1, 2(s2) + lbu t0, 0(t9) + lbu t1, 2(t9) + addu v0, v0, v1 + mult $ac1,t2, t6 + addu t0, t0, t1 + lbu t2, 2(s0) + addu t0, t0, v0 + lbu t3, 2(s1) + addu s3, t0, s3 + lbu v0, 0(s0) + lbu t0, 0(s1) + sll s3, s3, 1 + addu v0, v0, t2 + addu t0, t0, t3 + addu t0, t0, v0 + addu s3, t0, s3 + madd $ac1,s3, t7 + extr_r.w v0, $ac1, 16 + addiu t8, t8, 1 + addiu s2, s2, 2 + addiu t9, t9, 2 + addiu s0, s0, 2 + addiu s1, s1, 2 + sb v0, -1(t8) + addiu s4, s7, -2 + and s4, s4, 3 + addu s5, s4, t8 //end adress +4: + lh v0, 0(s2) + lh v1, 0(t9) + lh t0, 0(s0) + lh t1, 0(s1) + ins v0, v1, 16, 16 + ins t0, t1, 16, 16 + raddu.w.qb t2, v0 + raddu.w.qb s3, t0 + lbu v0, -1(s2) + lbu v1, 2(s2) + lbu t0, -1(t9) + lbu t1, 2(t9) + addu v0, v0, v1 + mult $ac1, t2, t6 + addu t0, t0, t1 + lbu t2, 2(s0) + addu t0, t0, v0 + lbu t3, 2(s1) + addu s3, t0, s3 + lbu v0, -1(s0) + lbu t0, -1(s1) + sll s3, s3, 1 + addu v0, v0, t2 + addu t0, t0, t3 + addu t0, t0, v0 + addu s3, t0, s3 + madd $ac1, s3, t7 + extr_r.w t2, $ac1, 16 + addiu t8, t8, 1 + addiu s2, s2, 2 + addiu t9, t9, 2 + addiu s0, s0, 2 + sb t2, -1(t8) + bne s5, t8, 4b + addiu s1, s1, 2 + addiu s5, s7, -2 + subu s5, s5, s4 + addu s5, s5, t8 //end adress +5: + lh v0, 0(s2) + lh v1, 0(t9) + lh t0, 0(s0) + lh t1, 0(s1) + ins v0, v1, 16, 16 + ins t0, t1, 16, 16 + raddu.w.qb t2, v0 + raddu.w.qb s3, t0 + lbu v0, -1(s2) + lbu v1, 2(s2) + lbu t0, -1(t9) + lbu t1, 2(t9) + addu v0, v0, v1 + mult $ac1, t2, t6 + addu t0, t0, t1 + lbu t2, 2(s0) + addu t0, t0, v0 + lbu t3, 2(s1) + addu s3, t0, s3 + lbu v0, -1(s0) + lbu t0, -1(s1) + sll s3, s3, 1 + addu v0, v0, t2 + addu t0, t0, t3 + lh v1, 2(t9) + addu t0, t0, v0 + lh v0, 2(s2) + addu s3, t0, s3 + lh t0, 2(s0) + lh t1, 2(s1) + madd $ac1, s3, t7 + extr_r.w t2, $ac1, 16 + ins t0, t1, 16, 16 + ins v0, v1, 16, 16 + raddu.w.qb s3, t0 + lbu v1, 4(s2) + lbu t0, 1(t9) + lbu t1, 4(t9) + sb t2, 0(t8) + raddu.w.qb t3, v0 + lbu v0, 1(s2) + addu t0, t0, t1 + mult $ac1, t3, t6 + addu v0, v0, v1 + lbu t2, 4(s0) + addu t0, t0, v0 + lbu v0, 1(s0) + addu s3, t0, s3 + lbu t0, 1(s1) + lbu t3, 4(s1) + addu v0, v0, t2 + sll s3, s3, 1 + addu t0, t0, t3 + lh v1, 4(t9) + addu t0, t0, v0 + lh v0, 4(s2) + addu s3, t0, s3 + lh t0, 4(s0) + lh t1, 4(s1) + madd $ac1, s3, t7 + extr_r.w t2, $ac1, 16 + ins t0, t1, 16, 16 + ins v0, v1, 16, 16 + raddu.w.qb s3, t0 + lbu v1, 6(s2) + lbu t0, 3(t9) + lbu t1, 6(t9) + sb t2, 1(t8) + raddu.w.qb t3, v0 + lbu v0, 3(s2) + addu t0, t0,t1 + mult $ac1, t3, t6 + addu v0, v0, v1 + lbu t2, 6(s0) + addu t0, t0, v0 + lbu v0, 3(s0) + addu s3, t0, s3 + lbu t0, 3(s1) + lbu t3, 6(s1) + addu v0, v0, t2 + sll s3, s3, 1 + addu t0, t0, t3 + lh v1, 6(t9) + addu t0, t0, v0 + lh v0, 6(s2) + addu s3, t0, s3 + lh t0, 6(s0) + lh t1, 6(s1) + madd $ac1, s3, t7 + extr_r.w t3, $ac1, 16 + ins t0, t1, 16, 16 + ins v0, v1, 16, 16 + raddu.w.qb s3, t0 + lbu v1, 8(s2) + lbu t0, 5(t9) + lbu t1, 8(t9) + sb t3, 2(t8) + raddu.w.qb t2, v0 + lbu v0, 5(s2) + addu t0, t0, t1 + mult $ac1, t2, t6 + addu v0, v0, v1 + lbu t2, 8(s0) + addu t0, t0, v0 + lbu v0, 5(s0) + addu s3, t0, s3 + lbu t0, 5(s1) + lbu t3, 8(s1) + addu v0, v0, t2 + sll s3, s3, 1 + addu t0, t0, t3 + addiu t8, t8, 4 + addu t0, t0, v0 + addiu s2, s2, 8 + addu s3, t0, s3 + addiu t9, t9, 8 + madd $ac1, s3, t7 + extr_r.w t1, $ac1, 16 + addiu s0, s0, 8 + addiu s1, s1, 8 + bne s5, t8, 5b + sb t1, -1(t8) +/* Special case for last column */ + lh v0, 0(s2) + lh v1, 0(t9) + lh t0, 0(s0) + lh t1, 0(s1) + ins v0, v1, 16, 16 + ins t0, t1, 16, 16 + raddu.w.qb t2, v0 + raddu.w.qb s3, t0 + lbu v0, -1(s2) + lbu v1, 1(s2) + lbu t0, -1(t9) + lbu t1, 1(t9) + addu v0, v0, v1 + mult $ac1, t2, t6 + addu t0, t0, t1 + lbu t2, 1(s0) + addu t0, t0, v0 + lbu t3, 1(s1) + addu s3, t0, s3 + lbu v0, -1(s0) + lbu t0, -1(s1) + sll s3, s3, 1 + addu v0, v0, t2 + addu t0, t0, t3 + addu t0, t0, v0 + addu s3, t0, s3 + madd $ac1, s3, t7 + extr_r.w t0, $ac1, 16 + addiu t5, t5, 2 + sb t0, 0(t8) + addiu t4, t4, 1 + bne t4, a2, 3b + addiu t5, t5, 2 + + RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + j ra + nop + +END(jsimd_h2v2_smooth_downsample_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_int_upsample_mips_dspr2) +/* + * a0 - upsample->h_expand[compptr->component_index] + * a1 - upsample->v_expand[compptr->component_index] + * a2 - input_data + * a3 - output_data_ptr + * 16(sp) - cinfo->output_width + * 20(sp) - cinfo->max_v_samp_factor + */ + .set at + + SAVE_REGS_ON_STACK 16, s0, s1, s2, s3 + + lw s0, 0(a3) // s0 = output_data + lw s1, 32(sp) // s1 = cinfo->output_width + lw s2, 36(sp) // s2 = cinfo->max_v_samp_factor + li t6, 0 // t6 = inrow + beqz s2, 10f + li s3, 0 // s3 = outrow +0: + addu t0, a2, t6 + addu t7, s0, s3 + lw t3, 0(t0) // t3 = inptr + lw t8, 0(t7) // t8 = outptr + beqz s1, 4f + addu t5, t8, s1 // t5 = outend +1: + lb t2, 0(t3) // t2 = invalue = *inptr++ + addiu t3, 1 + beqz a0, 3f + move t0, a0 // t0 = h_expand +2: + sb t2, 0(t8) + addiu t0, -1 + bgtz t0, 2b + addiu t8, 1 +3: + bgt t5, t8, 1b + nop +4: + addiu t9, a1, -1 // t9 = v_expand - 1 + blez t9, 9f + nop +5: + lw t3, 0(s0) + lw t4, 4(s0) + subu t0, s1, 0xF + blez t0, 7f + addu t5, t3, s1 // t5 = end address + andi t7, s1, 0xF // t7 = residual + subu t8, t5, t7 +6: + ulw t0, 0(t3) + ulw t1, 4(t3) + ulw t2, 8(t3) + usw t0, 0(t4) + ulw t0, 12(t3) + usw t1, 4(t4) + usw t2, 8(t4) + usw t0, 12(t4) + addiu t3, 16 + bne t3, t8, 6b + addiu t4, 16 + beqz t7, 8f + nop +7: + lbu t0, 0(t3) + sb t0, 0(t4) + addiu t3, 1 + bne t3, t5, 7b + addiu t4, 1 +8: + addiu t9, -1 + bgtz t9, 5b + addiu s0, 8 +9: + addu s3, s3, a1 + bne s3, s2, 0b + addiu t6, 1 +10: + RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3 + + j ra + nop +END(jsimd_int_upsample_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_h2v1_upsample_mips_dspr2) +/* + * a0 - cinfo->max_v_samp_factor + * a1 - cinfo->output_width + * a2 - input_data + * a3 - output_data_ptr + */ + lw t7, 0(a3) // t7 = output_data + andi t8, a1, 0xf // t8 = residual + sll t0, a0, 2 + blez a0, 4f + addu t9, t7, t0 // t9 = output_data end address +0: + lw t5, 0(t7) // t5 = outptr + lw t6, 0(a2) // t6 = inptr + addu t3, t5, a1 // t3 = outptr + output_width (end address) + subu t3, t8 // t3 = end address - residual + beq t5, t3, 2f + move t4, t8 +1: + ulw t0, 0(t6) // t0 = |P3|P2|P1|P0| + ulw t2, 4(t6) // t2 = |P7|P6|P5|P4| + srl t1, t0, 16 // t1 = |X|X|P3|P2| + ins t0, t0, 16, 16 // t0 = |P1|P0|P1|P0| + ins t1, t1, 16, 16 // t1 = |P3|P2|P3|P2| + ins t0, t0, 8, 16 // t0 = |P1|P1|P0|P0| + ins t1, t1, 8, 16 // t1 = |P3|P3|P2|P2| + usw t0, 0(t5) + usw t1, 4(t5) + srl t0, t2, 16 // t0 = |X|X|P7|P6| + ins t2, t2, 16, 16 // t2 = |P5|P4|P5|P4| + ins t0, t0, 16, 16 // t0 = |P7|P6|P7|P6| + ins t2, t2, 8, 16 // t2 = |P5|P5|P4|P4| + ins t0, t0, 8, 16 // t0 = |P7|P7|P6|P6| + usw t2, 8(t5) + usw t0, 12(t5) + addiu t5, 16 + bne t5, t3, 1b + addiu t6, 8 + beqz t8, 3f + move t4, t8 +2: + lbu t1, 0(t6) + sb t1, 0(t5) + sb t1, 1(t5) + addiu t4, -2 + addiu t6, 1 + bgtz t4, 2b + addiu t5, 2 +3: + addiu t7, 4 + bne t9, t7, 0b + addiu a2, 4 +4: + j ra + nop +END(jsimd_h2v1_upsample_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_h2v2_upsample_mips_dspr2) +/* + * a0 - cinfo->max_v_samp_factor + * a1 - cinfo->output_width + * a2 - input_data + * a3 - output_data_ptr + */ + lw t7, 0(a3) + blez a0, 7f + andi t9, a1, 0xf // t9 = residual +0: + lw t6, 0(a2) // t6 = inptr + lw t5, 0(t7) // t5 = outptr + addu t8, t5, a1 // t8 = outptr end address + subu t8, t9 // t8 = end address - residual + beq t5, t8, 2f + move t4, t9 +1: + ulw t0, 0(t6) + srl t1, t0, 16 + ins t0, t0, 16, 16 + ins t0, t0, 8, 16 + ins t1, t1, 16, 16 + ins t1, t1, 8, 16 + ulw t2, 4(t6) + usw t0, 0(t5) + usw t1, 4(t5) + srl t3, t2, 16 + ins t2, t2, 16, 16 + ins t2, t2, 8, 16 + ins t3, t3, 16, 16 + ins t3, t3, 8, 16 + usw t2, 8(t5) + usw t3, 12(t5) + addiu t5, 16 + bne t5, t8, 1b + addiu t6, 8 + beqz t9, 3f + move t4, t9 +2: + lbu t0, 0(t6) + sb t0, 0(t5) + sb t0, 1(t5) + addiu t4, -2 + addiu t6, 1 + bgtz t4, 2b + addiu t5, 2 +3: + ulw t6, 0(t7) // t6 = outptr + ulw t5, 4(t7) // t5 = outptr[1] + addu t4, t6, a1 // t4 = new end address + subu t8, t4, t9 + beqz t8, 5f + nop +4: + ulw t0, 0(t6) + ulw t1, 4(t6) + ulw t2, 8(t6) + usw t0, 0(t5) + ulw t0, 12(t6) + usw t1, 4(t5) + usw t2, 8(t5) + usw t0, 12(t5) + addiu t6, 16 + bne t6, t8, 4b + addiu t5, 16 + beqz t9, 6f + nop +5: + lbu t0, 0(t6) + sb t0, 0(t5) + addiu t6, 1 + bne t6, t4, 5b + addiu t5, 1 +6: + addiu t7, 8 + addiu a0, -2 + bgtz a0, 0b + addiu a2, 4 +7: + j ra + nop +END(jsimd_h2v2_upsample_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_idct_islow_mips_dspr2) +/* + * a0 - coef_block + * a1 - compptr->dcttable + * a2 - output + * a3 - range_limit + */ + + SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + addiu sp, sp, -256 + move v0, sp + addiu v1, zero, 8 // v1 = DCTSIZE = 8 +1: + lh s4, 32(a0) // s4 = inptr[16] + lh s5, 64(a0) // s5 = inptr[32] + lh s6, 96(a0) // s6 = inptr[48] + lh t1, 112(a0) // t1 = inptr[56] + lh t7, 16(a0) // t7 = inptr[8] + lh t5, 80(a0) // t5 = inptr[40] + lh t3, 48(a0) // t3 = inptr[24] + or s4, s4, t1 + or s4, s4, t3 + or s4, s4, t5 + or s4, s4, t7 + or s4, s4, s5 + or s4, s4, s6 + bnez s4, 2f + addiu v1, v1, -1 + lh s5, 0(a1) // quantptr[DCTSIZE*0] + lh s6, 0(a0) // inptr[DCTSIZE*0] + mul s5, s5, s6 // DEQUANTIZE(inptr[0], quantptr[0]) + sll s5, s5, 2 + sw s5, 0(v0) + sw s5, 32(v0) + sw s5, 64(v0) + sw s5, 96(v0) + sw s5, 128(v0) + sw s5, 160(v0) + sw s5, 192(v0) + b 3f + sw s5, 224(v0) +2: + lh t0, 112(a1) + lh t2, 48(a1) + lh t4, 80(a1) + lh t6, 16(a1) + mul t0, t0, t1 // DEQUANTIZE(inptr[DCTSIZE*7],quant[DCTSIZE*7]) + mul t1, t2, t3 // DEQUANTIZE(inptr[DCTSIZE*3],quant[DCTSIZE*3]) + mul t2, t4, t5 // DEQUANTIZE(inptr[DCTSIZE*5],quant[DCTSIZE*5]) + mul t3, t6, t7 // DEQUANTIZE(inptr[DCTSIZE*1],quant[DCTSIZE*1]) + lh t4, 32(a1) + lh t5, 32(a0) + lh t6, 96(a1) + lh t7, 96(a0) + addu s0, t0, t1 // z3 = tmp0 + tmp2 + addu s1, t1, t2 // z2 = tmp1 + tmp2 + addu s2, t2, t3 // z4 = tmp1 + tmp3 + addu s3, s0, s2 // z3 + z4 + addiu t9, zero, 9633 // FIX_1_175875602 + mul s3, s3, t9 // z5 = MULTIPLY(z3 + z4, FIX_1_175875602) + addu t8, t0, t3 // z1 = tmp0 + tmp3 + addiu t9, zero, 2446 // FIX_0_298631336 + mul t0, t0, t9 // tmp0 = MULTIPLY(tmp0, FIX_0_298631336) + addiu t9, zero, 16819 // FIX_2_053119869 + mul t2, t2, t9 // tmp1 = MULTIPLY(tmp1, FIX_2_053119869) + addiu t9, zero, 25172 // FIX_3_072711026 + mul t1, t1, t9 // tmp2 = MULTIPLY(tmp2, FIX_3_072711026) + addiu t9, zero, 12299 // FIX_1_501321110 + mul t3, t3, t9 // tmp3 = MULTIPLY(tmp3, FIX_1_501321110) + addiu t9, zero, 16069 // FIX_1_961570560 + mul s0, s0, t9 // -z3 = MULTIPLY(z3, FIX_1_961570560) + addiu t9, zero, 3196 // FIX_0_390180644 + mul s2, s2, t9 // -z4 = MULTIPLY(z4, FIX_0_390180644) + addiu t9, zero, 7373 // FIX_0_899976223 + mul t8, t8, t9 // -z1 = MULTIPLY(z1, FIX_0_899976223) + addiu t9, zero, 20995 // FIX_2_562915447 + mul s1, s1, t9 // -z2 = MULTIPLY(z2, FIX_2_562915447) + subu s0, s3, s0 // z3 += z5 + addu t0, t0, s0 // tmp0 += z3 + addu t1, t1, s0 // tmp2 += z3 + subu s2, s3, s2 // z4 += z5 + addu t2, t2, s2 // tmp1 += z4 + addu t3, t3, s2 // tmp3 += z4 + subu t0, t0, t8 // tmp0 += z1 + subu t1, t1, s1 // tmp2 += z2 + subu t2, t2, s1 // tmp1 += z2 + subu t3, t3, t8 // tmp3 += z1 + mul s0, t4, t5 // DEQUANTIZE(inptr[DCTSIZE*2],quant[DCTSIZE*2]) + addiu t9, zero, 6270 // FIX_0_765366865 + mul s1, t6, t7 // DEQUANTIZE(inptr[DCTSIZE*6],quant[DCTSIZE*6]) + lh t4, 0(a1) + lh t5, 0(a0) + lh t6, 64(a1) + lh t7, 64(a0) + mul s2, t9, s0 // MULTIPLY(z2, FIX_0_765366865) + mul t5, t4, t5 // DEQUANTIZE(inptr[DCTSIZE*0],quant[DCTSIZE*0]) + mul t6, t6, t7 // DEQUANTIZE(inptr[DCTSIZE*4],quant[DCTSIZE*4]) + addiu t9, zero, 4433 // FIX_0_541196100 + addu s3, s0, s1 // z2 + z3 + mul s3, s3, t9 // z1 = MULTIPLY(z2 + z3, FIX_0_541196100) + addiu t9, zero, 15137 // FIX_1_847759065 + mul t8, s1, t9 // MULTIPLY(z3, FIX_1_847759065) + addu t4, t5, t6 + subu t5, t5, t6 + sll t4, t4, 13 // tmp0 = (z2 + z3) << CONST_BITS + sll t5, t5, 13 // tmp1 = (z2 - z3) << CONST_BITS + addu t7, s3, s2 // tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865) + subu t6, s3, t8 // tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065) + addu s0, t4, t7 + subu s1, t4, t7 + addu s2, t5, t6 + subu s3, t5, t6 + addu t4, s0, t3 + subu s0, s0, t3 + addu t3, s2, t1 + subu s2, s2, t1 + addu t1, s3, t2 + subu s3, s3, t2 + addu t2, s1, t0 + subu s1, s1, t0 + shra_r.w t4, t4, 11 + shra_r.w t3, t3, 11 + shra_r.w t1, t1, 11 + shra_r.w t2, t2, 11 + shra_r.w s1, s1, 11 + shra_r.w s3, s3, 11 + shra_r.w s2, s2, 11 + shra_r.w s0, s0, 11 + sw t4, 0(v0) + sw t3, 32(v0) + sw t1, 64(v0) + sw t2, 96(v0) + sw s1, 128(v0) + sw s3, 160(v0) + sw s2, 192(v0) + sw s0, 224(v0) +3: + addiu a1, a1, 2 + addiu a0, a0, 2 + bgtz v1, 1b + addiu v0, v0, 4 + move v0, sp + addiu v1, zero, 8 +4: + lw t0, 8(v0) // z2 = (INT32) wsptr[2] + lw t1, 24(v0) // z3 = (INT32) wsptr[6] + lw t2, 0(v0) // (INT32) wsptr[0] + lw t3, 16(v0) // (INT32) wsptr[4] + lw s4, 4(v0) // (INT32) wsptr[1] + lw s5, 12(v0) // (INT32) wsptr[3] + lw s6, 20(v0) // (INT32) wsptr[5] + lw s7, 28(v0) // (INT32) wsptr[7] + or s4, s4, t0 + or s4, s4, t1 + or s4, s4, t3 + or s4, s4, s7 + or s4, s4, s5 + or s4, s4, s6 + bnez s4, 5f + addiu v1, v1, -1 + shra_r.w s5, t2, 5 + andi s5, s5, 0x3ff + lbux s5, s5(a3) + lw s1, 0(a2) + replv.qb s5, s5 + usw s5, 0(s1) + usw s5, 4(s1) + b 6f + nop +5: + addu t4, t0, t1 // z2 + z3 + addiu t8, zero, 4433 // FIX_0_541196100 + mul t5, t4, t8 // z1 = MULTIPLY(z2 + z3, FIX_0_541196100) + addiu t8, zero, 15137 // FIX_1_847759065 + mul t1, t1, t8 // MULTIPLY(z3, FIX_1_847759065) + addiu t8, zero, 6270 // FIX_0_765366865 + mul t0, t0, t8 // MULTIPLY(z2, FIX_0_765366865) + addu t4, t2, t3 // (INT32) wsptr[0] + (INT32) wsptr[4] + subu t2, t2, t3 // (INT32) wsptr[0] - (INT32) wsptr[4] + sll t4, t4, 13 // tmp0 = ((wsptr[0] + wsptr[4]) << CONST_BITS + sll t2, t2, 13 // tmp1 = ((wsptr[0] - wsptr[4]) << CONST_BITS + subu t1, t5, t1 // tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065) + subu t3, t2, t1 // tmp12 = tmp1 - tmp2 + addu t2, t2, t1 // tmp11 = tmp1 + tmp2 + addu t5, t5, t0 // tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865) + subu t1, t4, t5 // tmp13 = tmp0 - tmp3 + addu t0, t4, t5 // tmp10 = tmp0 + tmp3 + lw t4, 28(v0) // tmp0 = (INT32) wsptr[7] + lw t6, 12(v0) // tmp2 = (INT32) wsptr[3] + lw t5, 20(v0) // tmp1 = (INT32) wsptr[5] + lw t7, 4(v0) // tmp3 = (INT32) wsptr[1] + addu s0, t4, t6 // z3 = tmp0 + tmp2 + addiu t8, zero, 9633 // FIX_1_175875602 + addu s1, t5, t7 // z4 = tmp1 + tmp3 + addu s2, s0, s1 // z3 + z4 + mul s2, s2, t8 // z5 = MULTIPLY(z3 + z4, FIX_1_175875602) + addu s3, t4, t7 // z1 = tmp0 + tmp3 + addu t9, t5, t6 // z2 = tmp1 + tmp2 + addiu t8, zero, 16069 // FIX_1_961570560 + mul s0, s0, t8 // -z3 = MULTIPLY(z3, FIX_1_961570560) + addiu t8, zero, 3196 // FIX_0_390180644 + mul s1, s1, t8 // -z4 = MULTIPLY(z4, FIX_0_390180644) + addiu t8, zero, 2446 // FIX_0_298631336 + mul t4, t4, t8 // tmp0 = MULTIPLY(tmp0, FIX_0_298631336) + addiu t8, zero, 7373 // FIX_0_899976223 + mul s3, s3, t8 // -z1 = MULTIPLY(z1, FIX_0_899976223) + addiu t8, zero, 16819 // FIX_2_053119869 + mul t5, t5, t8 // tmp1 = MULTIPLY(tmp1, FIX_2_053119869) + addiu t8, zero, 20995 // FIX_2_562915447 + mul t9, t9, t8 // -z2 = MULTIPLY(z2, FIX_2_562915447) + addiu t8, zero, 25172 // FIX_3_072711026 + mul t6, t6, t8 // tmp2 = MULTIPLY(tmp2, FIX_3_072711026) + addiu t8, zero, 12299 // FIX_1_501321110 + mul t7, t7, t8 // tmp3 = MULTIPLY(tmp3, FIX_1_501321110) + subu s0, s2, s0 // z3 += z5 + subu s1, s2, s1 // z4 += z5 + addu t4, t4, s0 + subu t4, t4, s3 // tmp0 + addu t5, t5, s1 + subu t5, t5, t9 // tmp1 + addu t6, t6, s0 + subu t6, t6, t9 // tmp2 + addu t7, t7, s1 + subu t7, t7, s3 // tmp3 + addu s0, t0, t7 + subu t0, t0, t7 + addu t7, t2, t6 + subu t2, t2, t6 + addu t6, t3, t5 + subu t3, t3, t5 + addu t5, t1, t4 + subu t1, t1, t4 + shra_r.w s0, s0, 18 + shra_r.w t7, t7, 18 + shra_r.w t6, t6, 18 + shra_r.w t5, t5, 18 + shra_r.w t1, t1, 18 + shra_r.w t3, t3, 18 + shra_r.w t2, t2, 18 + shra_r.w t0, t0, 18 + andi s0, s0, 0x3ff + andi t7, t7, 0x3ff + andi t6, t6, 0x3ff + andi t5, t5, 0x3ff + andi t1, t1, 0x3ff + andi t3, t3, 0x3ff + andi t2, t2, 0x3ff + andi t0, t0, 0x3ff + lw s1, 0(a2) + lbux s0, s0(a3) + lbux t7, t7(a3) + lbux t6, t6(a3) + lbux t5, t5(a3) + lbux t1, t1(a3) + lbux t3, t3(a3) + lbux t2, t2(a3) + lbux t0, t0(a3) + sb s0, 0(s1) + sb t7, 1(s1) + sb t6, 2(s1) + sb t5, 3(s1) + sb t1, 4(s1) + sb t3, 5(s1) + sb t2, 6(s1) + sb t0, 7(s1) +6: + addiu v0, v0, 32 + bgtz v1, 4b + addiu a2, a2, 4 + addiu sp, sp, 256 + + RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + j ra + nop + +END(jsimd_idct_islow_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_idct_ifast_cols_mips_dspr2) +/* + * a0 - inptr + * a1 - quantptr + * a2 - wsptr + * a3 - mips_idct_ifast_coefs + */ + + SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + addiu t9, a0, 16 // end address + or AT, a3, zero + +0: + lw s0, 0(a1) // quantptr[DCTSIZE*0] + lw t0, 0(a0) // inptr[DCTSIZE*0] + lw t1, 16(a0) // inptr[DCTSIZE*1] + muleq_s.w.phl v0, t0, s0 // tmp0 ... + lw t2, 32(a0) // inptr[DCTSIZE*2] + lw t3, 48(a0) // inptr[DCTSIZE*3] + lw t4, 64(a0) // inptr[DCTSIZE*4] + lw t5, 80(a0) // inptr[DCTSIZE*5] + muleq_s.w.phr t0, t0, s0 // ... tmp0 ... + lw t6, 96(a0) // inptr[DCTSIZE*6] + lw t7, 112(a0) // inptr[DCTSIZE*7] + or s4, t1, t2 + or s5, t3, t4 + bnez s4, 1f + ins t0, v0, 16, 16 // ... tmp0 + bnez s5, 1f + or s6, t5, t6 + or s6, s6, t7 + bnez s6, 1f + sw t0, 0(a2) // wsptr[DCTSIZE*0] + sw t0, 16(a2) // wsptr[DCTSIZE*1] + sw t0, 32(a2) // wsptr[DCTSIZE*2] + sw t0, 48(a2) // wsptr[DCTSIZE*3] + sw t0, 64(a2) // wsptr[DCTSIZE*4] + sw t0, 80(a2) // wsptr[DCTSIZE*5] + sw t0, 96(a2) // wsptr[DCTSIZE*6] + sw t0, 112(a2) // wsptr[DCTSIZE*7] + addiu a0, a0, 4 + b 2f + addiu a1, a1, 4 + +1: + lw s1, 32(a1) // quantptr[DCTSIZE*2] + lw s2, 64(a1) // quantptr[DCTSIZE*4] + muleq_s.w.phl v0, t2, s1 // tmp1 ... + muleq_s.w.phr t2, t2, s1 // ... tmp1 ... + lw s0, 16(a1) // quantptr[DCTSIZE*1] + lw s1, 48(a1) // quantptr[DCTSIZE*3] + lw s3, 96(a1) // quantptr[DCTSIZE*6] + muleq_s.w.phl v1, t4, s2 // tmp2 ... + muleq_s.w.phr t4, t4, s2 // ... tmp2 ... + lw s2, 80(a1) // quantptr[DCTSIZE*5] + lw t8, 4(AT) // FIX(1.414213562) + ins t2, v0, 16, 16 // ... tmp1 + muleq_s.w.phl v0, t6, s3 // tmp3 ... + muleq_s.w.phr t6, t6, s3 // ... tmp3 ... + ins t4, v1, 16, 16 // ... tmp2 + addq.ph s4, t0, t4 // tmp10 + subq.ph s5, t0, t4 // tmp11 + ins t6, v0, 16, 16 // ... tmp3 + subq.ph s6, t2, t6 // tmp12 ... + addq.ph s7, t2, t6 // tmp13 + mulq_s.ph s6, s6, t8 // ... tmp12 ... + addq.ph t0, s4, s7 // tmp0 + subq.ph t6, s4, s7 // tmp3 + muleq_s.w.phl v0, t1, s0 // tmp4 ... + muleq_s.w.phr t1, t1, s0 // ... tmp4 ... + shll_s.ph s6, s6, 1 // x2 + lw s3, 112(a1) // quantptr[DCTSIZE*7] + subq.ph s6, s6, s7 // ... tmp12 + muleq_s.w.phl v1, t7, s3 // tmp7 ... + muleq_s.w.phr t7, t7, s3 // ... tmp7 ... + ins t1, v0, 16, 16 // ... tmp4 + addq.ph t2, s5, s6 // tmp1 + subq.ph t4, s5, s6 // tmp2 + muleq_s.w.phl v0, t5, s2 // tmp6 ... + muleq_s.w.phr t5, t5, s2 // ... tmp6 ... + ins t7, v1, 16, 16 // ... tmp7 + addq.ph s5, t1, t7 // z11 + subq.ph s6, t1, t7 // z12 + muleq_s.w.phl v1, t3, s1 // tmp5 ... + muleq_s.w.phr t3, t3, s1 // ... tmp5 ... + ins t5, v0, 16, 16 // ... tmp6 + ins t3, v1, 16, 16 // ... tmp5 + addq.ph s7, t5, t3 // z13 + subq.ph v0, t5, t3 // z10 + addq.ph t7, s5, s7 // tmp7 + subq.ph s5, s5, s7 // tmp11 ... + addq.ph v1, v0, s6 // z5 ... + mulq_s.ph s5, s5, t8 // ... tmp11 + lw t8, 8(AT) // FIX(1.847759065) + lw s4, 0(AT) // FIX(1.082392200) + addq.ph s0, t0, t7 + subq.ph s1, t0, t7 + mulq_s.ph v1, v1, t8 // ... z5 + shll_s.ph s5, s5, 1 // x2 + lw t8, 12(AT) // FIX(-2.613125930) + sw s0, 0(a2) // wsptr[DCTSIZE*0] + shll_s.ph v0, v0, 1 // x4 + mulq_s.ph v0, v0, t8 // tmp12 ... + mulq_s.ph s4, s6, s4 // tmp10 ... + shll_s.ph v1, v1, 1 // x2 + addiu a0, a0, 4 + addiu a1, a1, 4 + sw s1, 112(a2) // wsptr[DCTSIZE*7] + shll_s.ph s6, v0, 1 // x4 + shll_s.ph s4, s4, 1 // x2 + addq.ph s6, s6, v1 // ... tmp12 + subq.ph t5, s6, t7 // tmp6 + subq.ph s4, s4, v1 // ... tmp10 + subq.ph t3, s5, t5 // tmp5 + addq.ph s2, t2, t5 + addq.ph t1, s4, t3 // tmp4 + subq.ph s3, t2, t5 + sw s2, 16(a2) // wsptr[DCTSIZE*1] + sw s3, 96(a2) // wsptr[DCTSIZE*6] + addq.ph v0, t4, t3 + subq.ph v1, t4, t3 + sw v0, 32(a2) // wsptr[DCTSIZE*2] + sw v1, 80(a2) // wsptr[DCTSIZE*5] + addq.ph v0, t6, t1 + subq.ph v1, t6, t1 + sw v0, 64(a2) // wsptr[DCTSIZE*4] + sw v1, 48(a2) // wsptr[DCTSIZE*3] + +2: + bne a0, t9, 0b + addiu a2, a2, 4 + + RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + j ra + nop + +END(jsimd_idct_ifast_cols_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_idct_ifast_rows_mips_dspr2) +/* + * a0 - wsptr + * a1 - output_buf + * a2 - output_col + * a3 - mips_idct_ifast_coefs + */ + + SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8, a3 + + addiu t9, a0, 128 // end address + lui s8, 0x8080 + ori s8, s8, 0x8080 + +0: + lw AT, 36(sp) // restore $a3 (mips_idct_ifast_coefs) + lw t0, 0(a0) // wsptr[DCTSIZE*0+0/1] b a + lw s0, 16(a0) // wsptr[DCTSIZE*1+0/1] B A + lw t2, 4(a0) // wsptr[DCTSIZE*0+2/3] d c + lw s2, 20(a0) // wsptr[DCTSIZE*1+2/3] D C + lw t4, 8(a0) // wsptr[DCTSIZE*0+4/5] f e + lw s4, 24(a0) // wsptr[DCTSIZE*1+4/5] F E + lw t6, 12(a0) // wsptr[DCTSIZE*0+6/7] h g + lw s6, 28(a0) // wsptr[DCTSIZE*1+6/7] H G + precrq.ph.w t1, s0, t0 // B b + ins t0, s0, 16, 16 // A a + bnez t1, 1f + or s0, t2, s2 + bnez s0, 1f + or s0, t4, s4 + bnez s0, 1f + or s0, t6, s6 + bnez s0, 1f + shll_s.ph s0, t0, 2 // A a + lw a3, 0(a1) + lw AT, 4(a1) + precrq.ph.w t0, s0, s0 // A A + ins s0, s0, 16, 16 // a a + addu a3, a3, a2 + addu AT, AT, a2 + precrq.qb.ph t0, t0, t0 // A A A A + precrq.qb.ph s0, s0, s0 // a a a a + addu.qb s0, s0, s8 + addu.qb t0, t0, s8 + sw s0, 0(a3) + sw s0, 4(a3) + sw t0, 0(AT) + sw t0, 4(AT) + addiu a0, a0, 32 + bne a0, t9, 0b + addiu a1, a1, 8 + b 2f + nop + +1: + precrq.ph.w t3, s2, t2 + ins t2, s2, 16, 16 + precrq.ph.w t5, s4, t4 + ins t4, s4, 16, 16 + precrq.ph.w t7, s6, t6 + ins t6, s6, 16, 16 + lw t8, 4(AT) // FIX(1.414213562) + addq.ph s4, t0, t4 // tmp10 + subq.ph s5, t0, t4 // tmp11 + subq.ph s6, t2, t6 // tmp12 ... + addq.ph s7, t2, t6 // tmp13 + mulq_s.ph s6, s6, t8 // ... tmp12 ... + addq.ph t0, s4, s7 // tmp0 + subq.ph t6, s4, s7 // tmp3 + shll_s.ph s6, s6, 1 // x2 + subq.ph s6, s6, s7 // ... tmp12 + addq.ph t2, s5, s6 // tmp1 + subq.ph t4, s5, s6 // tmp2 + addq.ph s5, t1, t7 // z11 + subq.ph s6, t1, t7 // z12 + addq.ph s7, t5, t3 // z13 + subq.ph v0, t5, t3 // z10 + addq.ph t7, s5, s7 // tmp7 + subq.ph s5, s5, s7 // tmp11 ... + addq.ph v1, v0, s6 // z5 ... + mulq_s.ph s5, s5, t8 // ... tmp11 + lw t8, 8(AT) // FIX(1.847759065) + lw s4, 0(AT) // FIX(1.082392200) + addq.ph s0, t0, t7 // tmp0 + tmp7 + subq.ph s7, t0, t7 // tmp0 - tmp7 + mulq_s.ph v1, v1, t8 // ... z5 + lw a3, 0(a1) + lw t8, 12(AT) // FIX(-2.613125930) + shll_s.ph s5, s5, 1 // x2 + addu a3, a3, a2 + shll_s.ph v0, v0, 1 // x4 + mulq_s.ph v0, v0, t8 // tmp12 ... + mulq_s.ph s4, s6, s4 // tmp10 ... + shll_s.ph v1, v1, 1 // x2 + addiu a0, a0, 32 + addiu a1, a1, 8 + shll_s.ph s6, v0, 1 // x4 + shll_s.ph s4, s4, 1 // x2 + addq.ph s6, s6, v1 // ... tmp12 + shll_s.ph s0, s0, 2 + subq.ph t5, s6, t7 // tmp6 + subq.ph s4, s4, v1 // ... tmp10 + subq.ph t3, s5, t5 // tmp5 + shll_s.ph s7, s7, 2 + addq.ph t1, s4, t3 // tmp4 + addq.ph s1, t2, t5 // tmp1 + tmp6 + subq.ph s6, t2, t5 // tmp1 - tmp6 + addq.ph s2, t4, t3 // tmp2 + tmp5 + subq.ph s5, t4, t3 // tmp2 - tmp5 + addq.ph s4, t6, t1 // tmp3 + tmp4 + subq.ph s3, t6, t1 // tmp3 - tmp4 + shll_s.ph s1, s1, 2 + shll_s.ph s2, s2, 2 + shll_s.ph s3, s3, 2 + shll_s.ph s4, s4, 2 + shll_s.ph s5, s5, 2 + shll_s.ph s6, s6, 2 + precrq.ph.w t0, s1, s0 // B A + ins s0, s1, 16, 16 // b a + precrq.ph.w t2, s3, s2 // D C + ins s2, s3, 16, 16 // d c + precrq.ph.w t4, s5, s4 // F E + ins s4, s5, 16, 16 // f e + precrq.ph.w t6, s7, s6 // H G + ins s6, s7, 16, 16 // h g + precrq.qb.ph t0, t2, t0 // D C B A + precrq.qb.ph s0, s2, s0 // d c b a + precrq.qb.ph t4, t6, t4 // H G F E + precrq.qb.ph s4, s6, s4 // h g f e + addu.qb s0, s0, s8 + addu.qb s4, s4, s8 + sw s0, 0(a3) // outptr[0/1/2/3] d c b a + sw s4, 4(a3) // outptr[4/5/6/7] h g f e + lw a3, -4(a1) + addu.qb t0, t0, s8 + addu a3, a3, a2 + addu.qb t4, t4, s8 + sw t0, 0(a3) // outptr[0/1/2/3] D C B A + bne a0, t9, 0b + sw t4, 4(a3) // outptr[4/5/6/7] H G F E + +2: + + RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8, a3 + + j ra + nop + +END(jsimd_idct_ifast_rows_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_fdct_islow_mips_dspr2) +/* + * a0 - data + */ + + SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8 + + lui t0, 6437 + ori t0, 2260 + lui t1, 9633 + ori t1, 11363 + lui t2, 0xd39e + ori t2, 0xe6dc + lui t3, 0xf72d + ori t3, 9633 + lui t4, 2261 + ori t4, 9633 + lui t5, 0xd39e + ori t5, 6437 + lui t6, 9633 + ori t6, 0xd39d + lui t7, 0xe6dc + ori t7, 2260 + lui t8, 4433 + ori t8, 10703 + lui t9, 0xd630 + ori t9, 4433 + li s8, 8 + move a1, a0 +1: + lw s0, 0(a1) // tmp0 = 1|0 + lw s1, 4(a1) // tmp1 = 3|2 + lw s2, 8(a1) // tmp2 = 5|4 + lw s3, 12(a1) // tmp3 = 7|6 + packrl.ph s1, s1, s1 // tmp1 = 2|3 + packrl.ph s3, s3, s3 // tmp3 = 6|7 + subq.ph s7, s1, s2 // tmp7 = 2-5|3-4 = t5|t4 + subq.ph s5, s0, s3 // tmp5 = 1-6|0-7 = t6|t7 + mult $0, $0 // ac0 = 0 + dpa.w.ph $ac0, s7, t0 // ac0 += t5* 6437 + t4* 2260 + dpa.w.ph $ac0, s5, t1 // ac0 += t6* 9633 + t7* 11363 + mult $ac1, $0, $0 // ac1 = 0 + dpa.w.ph $ac1, s7, t2 // ac1 += t5*-11362 + t4* -6436 + dpa.w.ph $ac1, s5, t3 // ac1 += t6* -2259 + t7* 9633 + mult $ac2, $0, $0 // ac2 = 0 + dpa.w.ph $ac2, s7, t4 // ac2 += t5* 2261 + t4* 9633 + dpa.w.ph $ac2, s5, t5 // ac2 += t6*-11362 + t7* 6437 + mult $ac3, $0, $0 // ac3 = 0 + dpa.w.ph $ac3, s7, t6 // ac3 += t5* 9633 + t4*-11363 + dpa.w.ph $ac3, s5, t7 // ac3 += t6* -6436 + t7* 2260 + addq.ph s6, s1, s2 // tmp6 = 2+5|3+4 = t2|t3 + addq.ph s4, s0, s3 // tmp4 = 1+6|0+7 = t1|t0 + extr_r.w s0, $ac0, 11 // tmp0 = (ac0 + 1024) >> 11 + extr_r.w s1, $ac1, 11 // tmp1 = (ac1 + 1024) >> 11 + extr_r.w s2, $ac2, 11 // tmp2 = (ac2 + 1024) >> 11 + extr_r.w s3, $ac3, 11 // tmp3 = (ac3 + 1024) >> 11 + addq.ph s5, s4, s6 // tmp5 = t1+t2|t0+t3 = t11|t10 + subq.ph s7, s4, s6 // tmp7 = t1-t2|t0-t3 = t12|t13 + sh s0, 2(a1) + sh s1, 6(a1) + sh s2, 10(a1) + sh s3, 14(a1) + mult $0, $0 // ac0 = 0 + dpa.w.ph $ac0, s7, t8 // ac0 += t12* 4433 + t13* 10703 + mult $ac1, $0, $0 // ac1 = 0 + dpa.w.ph $ac1, s7, t9 // ac1 += t12*-10704 + t13* 4433 + sra s4, s5, 16 // tmp4 = t11 + addiu a1, a1, 16 + addiu s8, s8, -1 + extr_r.w s0, $ac0, 11 // tmp0 = (ac0 + 1024) >> 11 + extr_r.w s1, $ac1, 11 // tmp1 = (ac1 + 1024) >> 11 + addu s2, s5, s4 // tmp2 = t10 + t11 + subu s3, s5, s4 // tmp3 = t10 - t11 + sll s2, s2, 2 // tmp2 = (t10 + t11) << 2 + sll s3, s3, 2 // tmp3 = (t10 - t11) << 2 + sh s2, -16(a1) + sh s3, -8(a1) + sh s0, -12(a1) + bgtz s8, 1b + sh s1, -4(a1) + li t0, 2260 + li t1, 11363 + li t2, 9633 + li t3, 6436 + li t4, 6437 + li t5, 2261 + li t6, 11362 + li t7, 2259 + li t8, 4433 + li t9, 10703 + li a1, 10704 + li s8, 8 + +2: + lh a2, 0(a0) // 0 + lh a3, 16(a0) // 8 + lh v0, 32(a0) // 16 + lh v1, 48(a0) // 24 + lh s4, 64(a0) // 32 + lh s5, 80(a0) // 40 + lh s6, 96(a0) // 48 + lh s7, 112(a0) // 56 + addu s2, v0, s5 // tmp2 = 16 + 40 + subu s5, v0, s5 // tmp5 = 16 - 40 + addu s3, v1, s4 // tmp3 = 24 + 32 + subu s4, v1, s4 // tmp4 = 24 - 32 + addu s0, a2, s7 // tmp0 = 0 + 56 + subu s7, a2, s7 // tmp7 = 0 - 56 + addu s1, a3, s6 // tmp1 = 8 + 48 + subu s6, a3, s6 // tmp6 = 8 - 48 + addu a2, s0, s3 // tmp10 = tmp0 + tmp3 + subu v1, s0, s3 // tmp13 = tmp0 - tmp3 + addu a3, s1, s2 // tmp11 = tmp1 + tmp2 + subu v0, s1, s2 // tmp12 = tmp1 - tmp2 + mult s7, t1 // ac0 = tmp7 * c1 + madd s4, t0 // ac0 += tmp4 * c0 + madd s5, t4 // ac0 += tmp5 * c4 + madd s6, t2 // ac0 += tmp6 * c2 + mult $ac1, s7, t2 // ac1 = tmp7 * c2 + msub $ac1, s4, t3 // ac1 -= tmp4 * c3 + msub $ac1, s5, t6 // ac1 -= tmp5 * c6 + msub $ac1, s6, t7 // ac1 -= tmp6 * c7 + mult $ac2, s7, t4 // ac2 = tmp7 * c4 + madd $ac2, s4, t2 // ac2 += tmp4 * c2 + madd $ac2, s5, t5 // ac2 += tmp5 * c5 + msub $ac2, s6, t6 // ac2 -= tmp6 * c6 + mult $ac3, s7, t0 // ac3 = tmp7 * c0 + msub $ac3, s4, t1 // ac3 -= tmp4 * c1 + madd $ac3, s5, t2 // ac3 += tmp5 * c2 + msub $ac3, s6, t3 // ac3 -= tmp6 * c3 + extr_r.w s0, $ac0, 15 // tmp0 = (ac0 + 16384) >> 15 + extr_r.w s1, $ac1, 15 // tmp1 = (ac1 + 16384) >> 15 + extr_r.w s2, $ac2, 15 // tmp2 = (ac2 + 16384) >> 15 + extr_r.w s3, $ac3, 15 // tmp3 = (ac3 + 16384) >> 15 + addiu s8, s8, -1 + addu s4, a2, a3 // tmp4 = tmp10 + tmp11 + subu s5, a2, a3 // tmp5 = tmp10 - tmp11 + sh s0, 16(a0) + sh s1, 48(a0) + sh s2, 80(a0) + sh s3, 112(a0) + mult v0, t8 // ac0 = tmp12 * c8 + madd v1, t9 // ac0 += tmp13 * c9 + mult $ac1, v1, t8 // ac1 = tmp13 * c8 + msub $ac1, v0, a1 // ac1 -= tmp12 * c10 + addiu a0, a0, 2 + extr_r.w s6, $ac0, 15 // tmp6 = (ac0 + 16384) >> 15 + extr_r.w s7, $ac1, 15 // tmp7 = (ac1 + 16384) >> 15 + shra_r.w s4, s4, 2 // tmp4 = (tmp4 + 2) >> 2 + shra_r.w s5, s5, 2 // tmp5 = (tmp5 + 2) >> 2 + sh s4, -2(a0) + sh s5, 62(a0) + sh s6, 30(a0) + bgtz s8, 2b + sh s7, 94(a0) + + RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8 + + jr ra + nop + +END(jsimd_fdct_islow_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_fdct_ifast_mips_dspr2) +/* + * a0 - data + */ + .set at + SAVE_REGS_ON_STACK 8, s0, s1 + li a1, 0x014e014e // FIX_1_306562965 (334 << 16)|(334 & 0xffff) + li a2, 0x008b008b // FIX_0_541196100 (139 << 16)|(139 & 0xffff) + li a3, 0x00620062 // FIX_0_382683433 (98 << 16) |(98 & 0xffff) + li s1, 0x00b500b5 // FIX_0_707106781 (181 << 16)|(181 & 0xffff) + + move v0, a0 + addiu v1, v0, 128 // end address + +0: + lw t0, 0(v0) // tmp0 = 1|0 + lw t1, 4(v0) // tmp1 = 3|2 + lw t2, 8(v0) // tmp2 = 5|4 + lw t3, 12(v0) // tmp3 = 7|6 + packrl.ph t1, t1, t1 // tmp1 = 2|3 + packrl.ph t3, t3, t3 // tmp3 = 6|7 + subq.ph t7, t1, t2 // tmp7 = 2-5|3-4 = t5|t4 + subq.ph t5, t0, t3 // tmp5 = 1-6|0-7 = t6|t7 + addq.ph t6, t1, t2 // tmp6 = 2+5|3+4 = t2|t3 + addq.ph t4, t0, t3 // tmp4 = 1+6|0+7 = t1|t0 + addq.ph t8, t4, t6 // tmp5 = t1+t2|t0+t3 = t11|t10 + subq.ph t9, t4, t6 // tmp7 = t1-t2|t0-t3 = t12|t13 + sra t4, t8, 16 // tmp4 = t11 + mult $0, $0 // ac0 = 0 + dpa.w.ph $ac0, t9, s1 + mult $ac1, $0, $0 // ac1 = 0 + dpa.w.ph $ac1, t7, a3 // ac1 += t4*98 + t5*98 + dpsx.w.ph $ac1, t5, a3 // ac1 += t6*98 + t7*98 + mult $ac2, $0, $0 // ac2 = 0 + dpa.w.ph $ac2, t7, a2 // ac2 += t4*139 + t5*139 + mult $ac3, $0, $0 // ac3 = 0 + dpa.w.ph $ac3, t5, a1 // ac3 += t6*334 + t7*334 + precrq.ph.w t0, t5, t7 // t0 = t5|t6 + addq.ph t2, t8, t4 // tmp2 = t10 + t11 + subq.ph t3, t8, t4 // tmp3 = t10 - t11 + extr.w t4, $ac0, 8 + mult $0, $0 // ac0 = 0 + dpa.w.ph $ac0, t0, s1 // ac0 += t5*181 + t6*181 + extr.w t0, $ac1, 8 // t0 = z5 + extr.w t1, $ac2, 8 // t1 = MULTIPLY(tmp10, 139) + extr.w t7, $ac3, 8 // t2 = MULTIPLY(tmp12, 334) + extr.w t8, $ac0, 8 // t8 = z3 = MULTIPLY(tmp11, 181) + add t6, t1, t0 // t6 = z2 + add t7, t7, t0 // t7 = z4 + subq.ph t0, t5, t8 // t0 = z13 = tmp7 - z3 + addq.ph t8, t5, t8 // t9 = z11 = tmp7 + z3 + addq.ph t1, t0, t6 // t1 = z13 + z2 + subq.ph t6, t0, t6 // t6 = z13 - z2 + addq.ph t0, t8, t7 // t0 = z11 + z4 + subq.ph t7, t8, t7 // t7 = z11 - z4 + addq.ph t5, t4, t9 + subq.ph t4, t9, t4 + sh t2, 0(v0) + sh t5, 4(v0) + sh t3, 8(v0) + sh t4, 12(v0) + sh t1, 10(v0) + sh t6, 6(v0) + sh t0, 2(v0) + sh t7, 14(v0) + addiu v0, 16 + bne v1, v0, 0b + nop + move v0, a0 + addiu v1, v0, 16 + +1: + lh t0, 0(v0) // 0 + lh t1, 16(v0) // 8 + lh t2, 32(v0) // 16 + lh t3, 48(v0) // 24 + lh t4, 64(v0) // 32 + lh t5, 80(v0) // 40 + lh t6, 96(v0) // 48 + lh t7, 112(v0) // 56 + add t8, t0, t7 // t8 = tmp0 + sub t7, t0, t7 // t7 = tmp7 + add t0, t1, t6 // t0 = tmp1 + sub t1, t1, t6 // t1 = tmp6 + add t6, t2, t5 // t6 = tmp2 + sub t5, t2, t5 // t5 = tmp5 + add t2, t3, t4 // t2 = tmp3 + sub t3, t3, t4 // t3 = tmp4 + add t4, t8, t2 // t4 = tmp10 = tmp0 + tmp3 + sub t8, t8, t2 // t8 = tmp13 = tmp0 - tmp3 + sub s0, t0, t6 // s0 = tmp12 = tmp1 - tmp2 + ins t8, s0, 16, 16 // t8 = tmp12|tmp13 + add t2, t0, t6 // t2 = tmp11 = tmp1 + tmp2 + mult $0, $0 // ac0 = 0 + dpa.w.ph $ac0, t8, s1 // ac0 += t12*181 + t13*181 + add s0, t4, t2 // t8 = tmp10+tmp11 + sub t4, t4, t2 // t4 = tmp10-tmp11 + sh s0, 0(v0) + sh t4, 64(v0) + extr.w t2, $ac0, 8 // z1 = MULTIPLY(tmp12+tmp13,FIX_0_707106781) + addq.ph t4, t8, t2 // t9 = tmp13 + z1 + subq.ph t8, t8, t2 // t2 = tmp13 - z1 + sh t4, 32(v0) + sh t8, 96(v0) + add t3, t3, t5 // t3 = tmp10 = tmp4 + tmp5 + add t0, t5, t1 // t0 = tmp11 = tmp5 + tmp6 + add t1, t1, t7 // t1 = tmp12 = tmp6 + tmp7 + andi t4, a1, 0xffff + mul s0, t1, t4 + sra s0, s0, 8 // s0 = z4 = MULTIPLY(tmp12, FIX_1_306562965) + ins t1, t3, 16, 16 // t1 = tmp10|tmp12 + mult $0, $0 // ac0 = 0 + mulsa.w.ph $ac0, t1, a3 // ac0 += t10*98 - t12*98 + extr.w t8, $ac0, 8 // z5 = MULTIPLY(tmp10-tmp12,FIX_0_382683433) + add t2, t7, t8 // t2 = tmp7 + z5 + sub t7, t7, t8 // t7 = tmp7 - z5 + andi t4, a2, 0xffff + mul t8, t3, t4 + sra t8, t8, 8 // t8 = z2 = MULTIPLY(tmp10, FIX_0_541196100) + andi t4, s1, 0xffff + mul t6, t0, t4 + sra t6, t6, 8 // t6 = z3 = MULTIPLY(tmp11, FIX_0_707106781) + add t0, t6, t8 // t0 = z3 + z2 + sub t1, t6, t8 // t1 = z3 - z2 + add t3, t6, s0 // t3 = z3 + z4 + sub t4, t6, s0 // t4 = z3 - z4 + sub t5, t2, t1 // t5 = dataptr[5] + sub t6, t7, t0 // t6 = dataptr[3] + add t3, t2, t3 // t3 = dataptr[1] + add t4, t7, t4 // t4 = dataptr[7] + sh t5, 80(v0) + sh t6, 48(v0) + sh t3, 16(v0) + sh t4, 112(v0) + addiu v0, 2 + bne v0, v1, 1b + nop + + RESTORE_REGS_FROM_STACK 8, s0, s1 + + j ra + nop +END(jsimd_fdct_ifast_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_quantize_mips_dspr2) +/* + * a0 - coef_block + * a1 - divisors + * a2 - workspace + */ + + .set at + + SAVE_REGS_ON_STACK 16, s0, s1, s2 + + addiu v0, a2, 124 // v0 = workspace_end + lh t0, 0(a2) + lh t1, 0(a1) + lh t2, 128(a1) + sra t3, t0, 15 + sll t3, t3, 1 + addiu t3, t3, 1 + mul t0, t0, t3 + lh t4, 384(a1) + lh t5, 130(a1) + lh t6, 2(a2) + lh t7, 2(a1) + lh t8, 386(a1) + +1: + andi t1, 0xffff + add t9, t0, t2 + andi t9, 0xffff + mul v1, t9, t1 + sra s0, t6, 15 + sll s0, s0, 1 + addiu s0, s0, 1 + addiu t9, t4, 16 + srav v1, v1, t9 + mul v1, v1, t3 + mul t6, t6, s0 + andi t7, 0xffff + addiu a2, a2, 4 + addiu a1, a1, 4 + add s1, t6, t5 + andi s1, 0xffff + sh v1, 0(a0) + + mul s2, s1, t7 + addiu s1, t8, 16 + srav s2, s2, s1 + mul s2,s2, s0 + lh t0, 0(a2) + lh t1, 0(a1) + sra t3, t0, 15 + sll t3, t3, 1 + addiu t3, t3, 1 + mul t0, t0, t3 + lh t2, 128(a1) + lh t4, 384(a1) + lh t5, 130(a1) + lh t8, 386(a1) + lh t6, 2(a2) + lh t7, 2(a1) + sh s2, 2(a0) + lh t0, 0(a2) + sra t3, t0, 15 + sll t3, t3, 1 + addiu t3, t3, 1 + mul t0, t0,t3 + bne a2, v0, 1b + addiu a0, a0, 4 + + andi t1, 0xffff + add t9, t0, t2 + andi t9, 0xffff + mul v1, t9, t1 + sra s0, t6, 15 + sll s0, s0, 1 + addiu s0, s0, 1 + addiu t9, t4, 16 + srav v1, v1, t9 + mul v1, v1, t3 + mul t6, t6, s0 + andi t7, 0xffff + sh v1, 0(a0) + add s1, t6, t5 + andi s1, 0xffff + mul s2, s1, t7 + addiu s1, t8, 16 + addiu a2, a2, 4 + addiu a1, a1, 4 + srav s2, s2, s1 + mul s2, s2, s0 + sh s2, 2(a0) + + RESTORE_REGS_FROM_STACK 16, s0, s1, s2 + + j ra + nop + +END(jsimd_quantize_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_quantize_float_mips_dspr2) +/* + * a0 - coef_block + * a1 - divisors + * a2 - workspace + */ + + .set at + + li t1, 0x46800100 //integer representation 16384.5 + mtc1 t1, f0 + li t0, 63 +0: + lwc1 f1, 0(a2) + lwc1 f5, 0(a1) + lwc1 f2, 4(a2) + lwc1 f6, 4(a1) + lwc1 f3, 8(a2) + lwc1 f7, 8(a1) + lwc1 f4, 12(a2) + lwc1 f8, 12(a1) + madd.s f1, f0, f1, f5 + madd.s f2, f0, f2, f6 + madd.s f3, f0, f3, f7 + madd.s f4, f0, f4, f8 + lwc1 f5, 16(a1) + lwc1 f6, 20(a1) + trunc.w.s f1, f1 + trunc.w.s f2, f2 + trunc.w.s f3, f3 + trunc.w.s f4, f4 + lwc1 f7, 24(a1) + lwc1 f8, 28(a1) + mfc1 t1, f1 + mfc1 t2, f2 + mfc1 t3, f3 + mfc1 t4, f4 + lwc1 f1, 16(a2) + lwc1 f2, 20(a2) + lwc1 f3, 24(a2) + lwc1 f4, 28(a2) + madd.s f1, f0, f1, f5 + madd.s f2, f0, f2, f6 + madd.s f3, f0, f3, f7 + madd.s f4, f0, f4, f8 + addiu t1, t1, -16384 + addiu t2, t2, -16384 + addiu t3, t3, -16384 + addiu t4, t4, -16384 + trunc.w.s f1, f1 + trunc.w.s f2, f2 + trunc.w.s f3, f3 + trunc.w.s f4, f4 + sh t1, 0(a0) + sh t2, 2(a0) + sh t3, 4(a0) + sh t4, 6(a0) + mfc1 t1, f1 + mfc1 t2, f2 + mfc1 t3, f3 + mfc1 t4, f4 + addiu t0, t0, -8 + addiu a2, a2, 32 + addiu a1, a1, 32 + addiu t1, t1, -16384 + addiu t2, t2, -16384 + addiu t3, t3, -16384 + addiu t4, t4, -16384 + sh t1, 8(a0) + sh t2, 10(a0) + sh t3, 12(a0) + sh t4, 14(a0) + bgez t0, 0b + addiu a0, a0, 16 + + j ra + nop + +END(jsimd_quantize_float_mips_dspr2) +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_idct_2x2_mips_dspr2) +/* + * a0 - compptr->dct_table + * a1 - coef_block + * a2 - output_buf + * a3 - output_col + */ + .set at + + SAVE_REGS_ON_STACK 24, s0, s1, s2, s3, s4, s5 + + addiu sp, sp, -40 + move v0, sp + addiu s2, zero, 29692 + addiu s3, zero, -10426 + addiu s4, zero, 6967 + addiu s5, zero, -5906 + lh t0, 0(a1) // t0 = inptr[DCTSIZE*0] + lh t5, 0(a0) // t5 = quantptr[DCTSIZE*0] + lh t1, 48(a1) // t1 = inptr[DCTSIZE*3] + lh t6, 48(a0) // t6 = quantptr[DCTSIZE*3] + mul t4, t5, t0 + lh t0, 16(a1) // t0 = inptr[DCTSIZE*1] + lh t5, 16(a0) // t5 = quantptr[DCTSIZE*1] + mul t6, t6, t1 + mul t5, t5, t0 + lh t2, 80(a1) // t2 = inptr[DCTSIZE*5] + lh t7, 80(a0) // t7 = quantptr[DCTSIZE*5] + lh t3, 112(a1) // t3 = inptr[DCTSIZE*7] + lh t8, 112(a0) // t8 = quantptr[DCTSIZE*7] + mul t7, t7, t2 + mult zero, zero + mul t8, t8, t3 + li s0, 0x73FCD746 // s0 = (29692 << 16) | (-10426 & 0xffff) + li s1, 0x1B37E8EE // s1 = (6967 << 16) | (-5906 & 0xffff) + ins t6, t5, 16, 16 // t6 = t5|t6 + sll t4, t4, 15 + dpa.w.ph $ac0, t6, s0 + lh t1, 2(a1) + lh t6, 2(a0) + ins t8, t7, 16, 16 // t8 = t7|t8 + dpa.w.ph $ac0, t8, s1 + mflo t0, $ac0 + mul t5, t6, t1 + lh t1, 18(a1) + lh t6, 18(a0) + lh t2, 50(a1) + lh t7, 50(a0) + mul t6, t6, t1 + subu t8, t4, t0 + mul t7, t7, t2 + addu t0, t4, t0 + shra_r.w t0, t0, 13 + lh t1, 82(a1) + lh t2, 82(a0) + lh t3, 114(a1) + lh t4, 114(a0) + shra_r.w t8, t8, 13 + mul t1, t1, t2 + mul t3, t3, t4 + sw t0, 0(v0) + sw t8, 20(v0) + sll t4, t5, 15 + ins t7, t6, 16, 16 + mult zero, zero + dpa.w.ph $ac0, t7, s0 + ins t3, t1, 16, 16 + lh t1, 6(a1) + lh t6, 6(a0) + dpa.w.ph $ac0, t3, s1 + mflo t0, $ac0 + mul t5, t6, t1 + lh t1, 22(a1) + lh t6, 22(a0) + lh t2, 54(a1) + lh t7, 54(a0) + mul t6, t6, t1 + subu t8, t4, t0 + mul t7, t7, t2 + addu t0, t4, t0 + shra_r.w t0, t0, 13 + lh t1, 86(a1) + lh t2, 86(a0) + lh t3, 118(a1) + lh t4, 118(a0) + shra_r.w t8, t8, 13 + mul t1, t1, t2 + mul t3, t3, t4 + sw t0, 4(v0) + sw t8, 24(v0) + sll t4, t5, 15 + ins t7, t6, 16, 16 + mult zero, zero + dpa.w.ph $ac0, t7, s0 + ins t3, t1, 16, 16 + lh t1, 10(a1) + lh t6, 10(a0) + dpa.w.ph $ac0, t3, s1 + mflo t0, $ac0 + mul t5, t6, t1 + lh t1, 26(a1) + lh t6, 26(a0) + lh t2, 58(a1) + lh t7, 58(a0) + mul t6, t6, t1 + subu t8, t4, t0 + mul t7, t7, t2 + addu t0, t4, t0 + shra_r.w t0, t0, 13 + lh t1, 90(a1) + lh t2, 90(a0) + lh t3, 122(a1) + lh t4, 122(a0) + shra_r.w t8, t8, 13 + mul t1, t1, t2 + mul t3, t3, t4 + sw t0, 8(v0) + sw t8, 28(v0) + sll t4, t5, 15 + ins t7, t6, 16, 16 + mult zero, zero + dpa.w.ph $ac0, t7, s0 + ins t3, t1, 16, 16 + lh t1, 14(a1) + lh t6, 14(a0) + dpa.w.ph $ac0, t3, s1 + mflo t0, $ac0 + mul t5, t6, t1 + lh t1, 30(a1) + lh t6, 30(a0) + lh t2, 62(a1) + lh t7, 62(a0) + mul t6, t6, t1 + subu t8, t4, t0 + mul t7, t7, t2 + addu t0, t4, t0 + shra_r.w t0, t0, 13 + lh t1, 94(a1) + lh t2, 94(a0) + lh t3, 126(a1) + lh t4, 126(a0) + shra_r.w t8, t8, 13 + mul t1, t1, t2 + mul t3, t3, t4 + sw t0, 12(v0) + sw t8, 32(v0) + sll t4, t5, 15 + ins t7, t6, 16, 16 + mult zero, zero + dpa.w.ph $ac0, t7, s0 + ins t3, t1, 16, 16 + dpa.w.ph $ac0, t3, s1 + mflo t0, $ac0 + lw t9, 0(a2) + lw t3, 0(v0) + lw t7, 4(v0) + lw t1, 8(v0) + addu t9, t9, a3 + sll t3, t3, 15 + subu t8, t4, t0 + addu t0, t4, t0 + shra_r.w t0, t0, 13 + shra_r.w t8, t8, 13 + sw t0, 16(v0) + sw t8, 36(v0) + lw t5, 12(v0) + lw t6, 16(v0) + mult t7, s2 + madd t1, s3 + madd t5, s4 + madd t6, s5 + lw t5, 24(v0) + lw t7, 28(v0) + mflo t0, $ac0 + lw t8, 32(v0) + lw t2, 36(v0) + mult $ac1, t5, s2 + madd $ac1, t7, s3 + madd $ac1, t8, s4 + madd $ac1, t2, s5 + addu t1, t3, t0 + subu t6, t3, t0 + shra_r.w t1, t1, 20 + shra_r.w t6, t6, 20 + mflo t4, $ac1 + shll_s.w t1, t1, 24 + shll_s.w t6, t6, 24 + sra t1, t1, 24 + sra t6, t6, 24 + addiu t1, t1, 128 + addiu t6, t6, 128 + lw t0, 20(v0) + sb t1, 0(t9) + sb t6, 1(t9) + sll t0, t0, 15 + lw t9, 4(a2) + addu t1, t0, t4 + subu t6, t0, t4 + addu t9, t9, a3 + shra_r.w t1, t1, 20 + shra_r.w t6, t6, 20 + shll_s.w t1, t1, 24 + shll_s.w t6, t6, 24 + sra t1, t1, 24 + sra t6, t6, 24 + addiu t1, t1, 128 + addiu t6, t6, 128 + sb t1, 0(t9) + sb t6, 1(t9) + addiu sp, sp, 40 + + RESTORE_REGS_FROM_STACK 24, s0, s1, s2, s3, s4, s5 + + j ra + nop + +END(jsimd_idct_2x2_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_idct_4x4_mips_dspr2) +/* + * a0 - compptr->dct_table + * a1 - coef_block + * a2 - output_buf + * a3 - output_col + * 16(sp) - workspace[DCTSIZE*4]; // buffers data between passes + */ + + .set at + SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + lw v1, 48(sp) + move t0, a1 + move t1, v1 + li t9, 4 + li s0, 0x2e75f93e + li s1, 0x21f9ba79 + li s2, 0xecc2efb0 + li s3, 0x52031ccd + +0: + lh s6, 32(t0) // inptr[DCTSIZE*2] + lh t6, 32(a0) // quantptr[DCTSIZE*2] + lh s7, 96(t0) // inptr[DCTSIZE*6] + lh t7, 96(a0) // quantptr[DCTSIZE*6] + mul t6, s6, t6 // z2 = (inptr[DCTSIZE*2] * quantptr[DCTSIZE*2]) + lh s4, 0(t0) // inptr[DCTSIZE*0] + mul t7, s7, t7 // z3 = (inptr[DCTSIZE*6] * quantptr[DCTSIZE*6]) + lh s5, 0(a0) // quantptr[0] + li s6, 15137 + li s7, 6270 + mul t2, s4, s5 // tmp0 = (inptr[0] * quantptr[0]) + mul t6, s6, t6 // z2 = (inptr[DCTSIZE*2] * quantptr[DCTSIZE*2]) + lh t5, 112(t0) // inptr[DCTSIZE*7] + mul t7, s7, t7 // z3 = (inptr[DCTSIZE*6] * quantptr[DCTSIZE*6]) + lh s4, 112(a0) // quantptr[DCTSIZE*7] + lh v0, 80(t0) // inptr[DCTSIZE*5] + lh s5, 80(a0) // quantptr[DCTSIZE*5] + lh s6, 48(a0) // quantptr[DCTSIZE*3] + sll t2, t2, 14 // tmp0 <<= (CONST_BITS+1) + lh s7, 16(a0) // quantptr[DCTSIZE*1] + lh t8, 16(t0) // inptr[DCTSIZE*1] + subu t6, t6, t7 // tmp2 = MULTIPLY(z2, t5) - MULTIPLY(z3, t6) + lh t7, 48(t0) // inptr[DCTSIZE*3] + mul t5, s4, t5 // z1 = (inptr[DCTSIZE*7] * quantptr[DCTSIZE*7]) + mul v0, s5, v0 // z2 = (inptr[DCTSIZE*5] * quantptr[DCTSIZE*5]) + mul t7, s6, t7 // z3 = (inptr[DCTSIZE*3] * quantptr[DCTSIZE*3]) + mul t8, s7, t8 // z4 = (inptr[DCTSIZE*1] * quantptr[DCTSIZE*1]) + addu t3, t2, t6 // tmp10 = tmp0 + z2 + subu t4, t2, t6 // tmp10 = tmp0 - z2 + mult $ac0, zero, zero + mult $ac1, zero, zero + ins t5, v0, 16, 16 + ins t7, t8, 16, 16 + addiu t9, t9, -1 + dpa.w.ph $ac0, t5, s0 + dpa.w.ph $ac0, t7, s1 + dpa.w.ph $ac1, t5, s2 + dpa.w.ph $ac1, t7, s3 + mflo s4, $ac0 + mflo s5, $ac1 + addiu a0, a0, 2 + addiu t1, t1, 4 + addiu t0, t0, 2 + addu t6, t4, s4 + subu t5, t4, s4 + addu s6, t3, s5 + subu s7, t3, s5 + shra_r.w t6, t6, 12 // DESCALE(tmp12 + temp1, 12) + shra_r.w t5, t5, 12 // DESCALE(tmp12 - temp1, 12) + shra_r.w s6, s6, 12 // DESCALE(tmp10 + temp2, 12) + shra_r.w s7, s7, 12 // DESCALE(tmp10 - temp2, 12) + sw t6, 28(t1) + sw t5, 60(t1) + sw s6, -4(t1) + bgtz t9, 0b + sw s7, 92(t1) + // second loop three pass + li t9, 3 +1: + lh s6, 34(t0) // inptr[DCTSIZE*2] + lh t6, 34(a0) // quantptr[DCTSIZE*2] + lh s7, 98(t0) // inptr[DCTSIZE*6] + lh t7, 98(a0) // quantptr[DCTSIZE*6] + mul t6, s6, t6 // z2 = (inptr[DCTSIZE*2] * quantptr[DCTSIZE*2]) + lh s4, 2(t0) // inptr[DCTSIZE*0] + mul t7, s7, t7 // z3 = (inptr[DCTSIZE*6] * quantptr[DCTSIZE*6]) + lh s5, 2(a0) // quantptr[DCTSIZE*0] + li s6, 15137 + li s7, 6270 + mul t2, s4, s5 // tmp0 = (inptr[0] * quantptr[0]) + mul v0, s6, t6 // z2 = (inptr[DCTSIZE*2] * quantptr[DCTSIZE*2]) + lh t5, 114(t0) // inptr[DCTSIZE*7] + mul t7, s7, t7 // z3 = (inptr[DCTSIZE*6] * quantptr[DCTSIZE*6]) + lh s4, 114(a0) // quantptr[DCTSIZE*7] + lh s5, 82(a0) // quantptr[DCTSIZE*5] + lh t6, 82(t0) // inptr[DCTSIZE*5] + sll t2, t2, 14 // tmp0 <<= (CONST_BITS+1) + lh s6, 50(a0) // quantptr[DCTSIZE*3] + lh t8, 18(t0) // inptr[DCTSIZE*1] + subu v0, v0, t7 // tmp2 = MULTIPLY(z2, t5) - MULTIPLY(z3, t6) + lh t7, 50(t0) // inptr[DCTSIZE*3] + lh s7, 18(a0) // quantptr[DCTSIZE*1] + mul t5, s4, t5 // z1 = (inptr[DCTSIZE*7] * quantptr[DCTSIZE*7]) + mul t6, s5, t6 // z2 = (inptr[DCTSIZE*5] * quantptr[DCTSIZE*5]) + mul t7, s6, t7 // z3 = (inptr[DCTSIZE*3] * quantptr[DCTSIZE*3]) + mul t8, s7, t8 // z4 = (inptr[DCTSIZE*1] * quantptr[DCTSIZE*1]) + addu t3, t2, v0 // tmp10 = tmp0 + z2 + subu t4, t2, v0 // tmp10 = tmp0 - z2 + mult $ac0, zero, zero + mult $ac1, zero, zero + ins t5, t6, 16, 16 + ins t7, t8, 16, 16 + dpa.w.ph $ac0, t5, s0 + dpa.w.ph $ac0, t7, s1 + dpa.w.ph $ac1, t5, s2 + dpa.w.ph $ac1, t7, s3 + mflo t5, $ac0 + mflo t6, $ac1 + addiu t9, t9, -1 + addiu t0, t0, 2 + addiu a0, a0, 2 + addiu t1, t1, 4 + addu s5, t4, t5 + subu s4, t4, t5 + addu s6, t3, t6 + subu s7, t3, t6 + shra_r.w s5, s5, 12 // DESCALE(tmp12 + temp1, 12) + shra_r.w s4, s4, 12 // DESCALE(tmp12 - temp1, 12) + shra_r.w s6, s6, 12 // DESCALE(tmp10 + temp2, 12) + shra_r.w s7, s7, 12 // DESCALE(tmp10 - temp2, 12) + sw s5, 32(t1) + sw s4, 64(t1) + sw s6, 0(t1) + bgtz t9, 1b + sw s7, 96(t1) + move t1, v1 + li s4, 15137 + lw s6, 8(t1) // wsptr[2] + li s5, 6270 + lw s7, 24(t1) // wsptr[6] + mul s4, s4, s6 // MULTIPLY((INT32) wsptr[2], FIX_1_847759065) + lw t2, 0(t1) // wsptr[0] + mul s5, s5, s7 // MULTIPLY((INT32) wsptr[6], - FIX_0_765366865) + lh t5, 28(t1) // wsptr[7] + lh t6, 20(t1) // wsptr[5] + lh t7, 12(t1) // wsptr[3] + lh t8, 4(t1) // wsptr[1] + ins t5, t6, 16, 16 + ins t7, t8, 16, 16 + mult $ac0, zero, zero + dpa.w.ph $ac0, t5, s0 + dpa.w.ph $ac0, t7, s1 + mult $ac1, zero, zero + dpa.w.ph $ac1, t5, s2 + dpa.w.ph $ac1, t7, s3 + sll t2, t2, 14 // tmp0 = ((INT32) wsptr[0]) << (CONST_BITS+1) + mflo s6, $ac0 + // MULTIPLY(wsptr[2], FIX_1_847759065 + MULTIPLY(wsptr[6], -FIX_0_765366865) + subu s4, s4, s5 + addu t3, t2, s4 // tmp10 = tmp0 + z2 + mflo s7, $ac1 + subu t4, t2, s4 // tmp10 = tmp0 - z2 + addu t7, t4, s6 + subu t8, t4, s6 + addu t5, t3, s7 + subu t6, t3, s7 + shra_r.w t5, t5, 19 // DESCALE(tmp10 + temp2, 19) + shra_r.w t6, t6, 19 // DESCALE(tmp10 - temp2, 19) + shra_r.w t7, t7, 19 // DESCALE(tmp12 + temp1, 19) + shra_r.w t8, t8, 19 // DESCALE(tmp12 - temp1, 19) + sll s4, t9, 2 + lw v0, 0(a2) // output_buf[ctr] + shll_s.w t5, t5, 24 + shll_s.w t6, t6, 24 + shll_s.w t7, t7, 24 + shll_s.w t8, t8, 24 + sra t5, t5, 24 + sra t6, t6, 24 + sra t7, t7, 24 + sra t8, t8, 24 + addu v0, v0, a3 // outptr = output_buf[ctr] + output_col + addiu t5, t5, 128 + addiu t6, t6, 128 + addiu t7, t7, 128 + addiu t8, t8, 128 + sb t5, 0(v0) + sb t7, 1(v0) + sb t8, 2(v0) + sb t6, 3(v0) + // 2 + li s4, 15137 + lw s6, 40(t1) // wsptr[2] + li s5, 6270 + lw s7, 56(t1) // wsptr[6] + mul s4, s4, s6 // MULTIPLY((INT32) wsptr[2], FIX_1_847759065) + lw t2, 32(t1) // wsptr[0] + mul s5, s5, s7 // MULTIPLY((INT32) wsptr[6], - FIX_0_765366865) + lh t5, 60(t1) // wsptr[7] + lh t6, 52(t1) // wsptr[5] + lh t7, 44(t1) // wsptr[3] + lh t8, 36(t1) // wsptr[1] + ins t5, t6, 16, 16 + ins t7, t8, 16, 16 + mult $ac0, zero, zero + dpa.w.ph $ac0, t5, s0 + dpa.w.ph $ac0, t7, s1 + mult $ac1, zero, zero + dpa.w.ph $ac1, t5, s2 + dpa.w.ph $ac1, t7, s3 + sll t2, t2, 14 // tmp0 = ((INT32) wsptr[0]) << (CONST_BITS+1) + mflo s6, $ac0 + // MULTIPLY(wsptr[2], FIX_1_847759065 + MULTIPLY(wsptr[6], -FIX_0_765366865) + subu s4, s4, s5 + addu t3, t2, s4 // tmp10 = tmp0 + z2 + mflo s7, $ac1 + subu t4, t2, s4 // tmp10 = tmp0 - z2 + addu t7, t4, s6 + subu t8, t4, s6 + addu t5, t3, s7 + subu t6, t3, s7 + shra_r.w t5, t5, 19 // DESCALE(tmp10 + temp2, CONST_BITS-PASS1_BITS+1) + shra_r.w t6, t6, 19 // DESCALE(tmp10 - temp2, CONST_BITS-PASS1_BITS+1) + shra_r.w t7, t7, 19 // DESCALE(tmp12 + temp1, CONST_BITS-PASS1_BITS+1) + shra_r.w t8, t8, 19 // DESCALE(tmp12 - temp1, CONST_BITS-PASS1_BITS+1) + sll s4, t9, 2 + lw v0, 4(a2) // output_buf[ctr] + shll_s.w t5, t5, 24 + shll_s.w t6, t6, 24 + shll_s.w t7, t7, 24 + shll_s.w t8, t8, 24 + sra t5, t5, 24 + sra t6, t6, 24 + sra t7, t7, 24 + sra t8, t8, 24 + addu v0, v0, a3 // outptr = output_buf[ctr] + output_col + addiu t5, t5, 128 + addiu t6, t6, 128 + addiu t7, t7, 128 + addiu t8, t8, 128 + sb t5, 0(v0) + sb t7, 1(v0) + sb t8, 2(v0) + sb t6, 3(v0) + // 3 + li s4, 15137 + lw s6, 72(t1) // wsptr[2] + li s5, 6270 + lw s7, 88(t1) // wsptr[6] + mul s4, s4, s6 // MULTIPLY((INT32) wsptr[2], FIX_1_847759065) + lw t2, 64(t1) // wsptr[0] + mul s5, s5, s7 // MULTIPLY((INT32) wsptr[6], - FIX_0_765366865) + lh t5, 92(t1) // wsptr[7] + lh t6, 84(t1) // wsptr[5] + lh t7, 76(t1) // wsptr[3] + lh t8, 68(t1) // wsptr[1] + ins t5, t6, 16, 16 + ins t7, t8, 16, 16 + mult $ac0, zero, zero + dpa.w.ph $ac0, t5, s0 + dpa.w.ph $ac0, t7, s1 + mult $ac1, zero, zero + dpa.w.ph $ac1, t5, s2 + dpa.w.ph $ac1, t7, s3 + sll t2, t2, 14 // tmp0 = ((INT32) wsptr[0]) << (CONST_BITS+1) + mflo s6, $ac0 + // MULTIPLY(wsptr[2], FIX_1_847759065 + MULTIPLY(wsptr[6], -FIX_0_765366865) + subu s4, s4, s5 + addu t3, t2, s4 // tmp10 = tmp0 + z2 + mflo s7, $ac1 + subu t4, t2, s4 // tmp10 = tmp0 - z2 + addu t7, t4, s6 + subu t8, t4, s6 + addu t5, t3, s7 + subu t6, t3, s7 + shra_r.w t5, t5, 19 // DESCALE(tmp10 + temp2, 19) + shra_r.w t6, t6, 19 // DESCALE(tmp10 - temp2, 19) + shra_r.w t7, t7, 19 // DESCALE(tmp12 + temp1, 19) + shra_r.w t8, t8, 19 // DESCALE(tmp12 - temp1, 19) + sll s4, t9, 2 + lw v0, 8(a2) // output_buf[ctr] + shll_s.w t5, t5, 24 + shll_s.w t6, t6, 24 + shll_s.w t7, t7, 24 + shll_s.w t8, t8, 24 + sra t5, t5, 24 + sra t6, t6, 24 + sra t7, t7, 24 + sra t8, t8, 24 + addu v0, v0, a3 // outptr = output_buf[ctr] + output_col + addiu t5, t5, 128 + addiu t6, t6, 128 + addiu t7, t7, 128 + addiu t8, t8, 128 + sb t5, 0(v0) + sb t7, 1(v0) + sb t8, 2(v0) + sb t6, 3(v0) + li s4, 15137 + lw s6, 104(t1) // wsptr[2] + li s5, 6270 + lw s7, 120(t1) // wsptr[6] + mul s4, s4, s6 // MULTIPLY((INT32) wsptr[2], FIX_1_847759065) + lw t2, 96(t1) // wsptr[0] + mul s5, s5, s7 // MULTIPLY((INT32) wsptr[6], -FIX_0_765366865) + lh t5, 124(t1) // wsptr[7] + lh t6, 116(t1) // wsptr[5] + lh t7, 108(t1) // wsptr[3] + lh t8, 100(t1) // wsptr[1] + ins t5, t6, 16, 16 + ins t7, t8, 16, 16 + mult $ac0, zero, zero + dpa.w.ph $ac0, t5, s0 + dpa.w.ph $ac0, t7, s1 + mult $ac1, zero, zero + dpa.w.ph $ac1, t5, s2 + dpa.w.ph $ac1, t7, s3 + sll t2, t2, 14 // tmp0 = ((INT32) wsptr[0]) << (CONST_BITS+1) + mflo s6, $ac0 + // MULTIPLY(wsptr[2], FIX_1_847759065 + MULTIPLY(wsptr[6], -FIX_0_765366865) + subu s4, s4, s5 + addu t3, t2, s4 // tmp10 = tmp0 + z2; + mflo s7, $ac1 + subu t4, t2, s4 // tmp10 = tmp0 - z2; + addu t7, t4, s6 + subu t8, t4, s6 + addu t5, t3, s7 + subu t6, t3, s7 + shra_r.w t5, t5, 19 // DESCALE(tmp10 + temp2, 19) + shra_r.w t6, t6, 19 // DESCALE(tmp10 - temp2, 19) + shra_r.w t7, t7, 19 // DESCALE(tmp12 + temp1, 19) + shra_r.w t8, t8, 19 // DESCALE(tmp12 - temp1, 19) + sll s4, t9, 2 + lw v0, 12(a2) // output_buf[ctr] + shll_s.w t5, t5, 24 + shll_s.w t6, t6, 24 + shll_s.w t7, t7, 24 + shll_s.w t8, t8, 24 + sra t5, t5, 24 + sra t6, t6, 24 + sra t7, t7, 24 + sra t8, t8, 24 + addu v0, v0, a3 // outptr = output_buf[ctr] + output_col + addiu t5, t5, 128 + addiu t6, t6, 128 + addiu t7, t7, 128 + addiu t8, t8, 128 + sb t5, 0(v0) + sb t7, 1(v0) + sb t8, 2(v0) + sb t6, 3(v0) + + RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + j ra + nop +END(jsimd_idct_4x4_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_idct_6x6_mips_dspr2) +/* + * a0 - compptr->dct_table + * a1 - coef_block + * a2 - output_buf + * a3 - output_col + */ + .set at + + SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + addiu sp, sp, -144 + move v0, sp + addiu v1, v0, 24 + addiu t9, zero, 5793 + addiu s0, zero, 10033 + addiu s1, zero, 2998 + +1: + lh s2, 0(a0) // q0 = quantptr[ 0] + lh s3, 32(a0) // q1 = quantptr[16] + lh s4, 64(a0) // q2 = quantptr[32] + lh t2, 64(a1) // tmp2 = inptr[32] + lh t1, 32(a1) // tmp1 = inptr[16] + lh t0, 0(a1) // tmp0 = inptr[ 0] + mul t2, t2, s4 // tmp2 = tmp2 * q2 + mul t1, t1, s3 // tmp1 = tmp1 * q1 + mul t0, t0, s2 // tmp0 = tmp0 * q0 + lh t6, 16(a1) // z1 = inptr[ 8] + lh t8, 80(a1) // z3 = inptr[40] + lh t7, 48(a1) // z2 = inptr[24] + lh s2, 16(a0) // q0 = quantptr[ 8] + lh s4, 80(a0) // q2 = quantptr[40] + lh s3, 48(a0) // q1 = quantptr[24] + mul t2, t2, t9 // tmp2 = tmp2 * 5793 + mul t1, t1, s0 // tmp1 = tmp1 * 10033 + sll t0, t0, 13 // tmp0 = tmp0 << 13 + mul t6, t6, s2 // z1 = z1 * q0 + mul t8, t8, s4 // z3 = z3 * q2 + mul t7, t7, s3 // z2 = z2 * q1 + addu t3, t0, t2 // tmp10 = tmp0 + tmp2 + sll t2, t2, 1 // tmp2 = tmp2 << 2 + subu t4, t0, t2 // tmp11 = tmp0 - tmp2; + subu t5, t3, t1 // tmp12 = tmp10 - tmp1 + addu t3, t3, t1 // tmp10 = tmp10 + tmp1 + addu t1, t6, t8 // tmp1 = z1 + z3 + mul t1, t1, s1 // tmp1 = tmp1 * 2998 + shra_r.w t4, t4, 11 // tmp11 = (tmp11 + 1024) >> 11 + subu t2, t6, t8 // tmp2 = z1 - z3 + subu t2, t2, t7 // tmp2 = tmp2 - z2 + sll t2, t2, 2 // tmp2 = tmp2 << 2 + addu t0, t6, t7 // tmp0 = z1 + z2 + sll t0, t0, 13 // tmp0 = tmp0 << 13 + subu s2, t8, t7 // q0 = z3 - z2 + sll s2, s2, 13 // q0 = q0 << 13 + addu t0, t0, t1 // tmp0 = tmp0 + tmp1 + addu t1, s2, t1 // tmp1 = q0 + tmp1 + addu s2, t4, t2 // q0 = tmp11 + tmp2 + subu s3, t4, t2 // q1 = tmp11 - tmp2 + addu t6, t3, t0 // z1 = tmp10 + tmp0 + subu t7, t3, t0 // z2 = tmp10 - tmp0 + addu t4, t5, t1 // tmp11 = tmp12 + tmp1 + subu t5, t5, t1 // tmp12 = tmp12 - tmp1 + shra_r.w t6, t6, 11 // z1 = (z1 + 1024) >> 11 + shra_r.w t7, t7, 11 // z2 = (z2 + 1024) >> 11 + shra_r.w t4, t4, 11 // tmp11 = (tmp11 + 1024) >> 11 + shra_r.w t5, t5, 11 // tmp12 = (tmp12 + 1024) >> 11 + sw s2, 24(v0) + sw s3, 96(v0) + sw t6, 0(v0) + sw t7, 120(v0) + sw t4, 48(v0) + sw t5, 72(v0) + addiu v0, v0, 4 + addiu a1, a1, 2 + bne v0, v1, 1b + addiu a0, a0, 2 + + /* Pass 2: process 6 rows from work array, store into output array. */ + move v0, sp + addiu v1, v0, 144 + +2: + lw t0, 0(v0) + lw t2, 16(v0) + lw s5, 0(a2) + addiu t0, t0, 16 + sll t0, t0, 13 + mul t3, t2, t9 + lw t6, 4(v0) + lw t8, 20(v0) + lw t7, 12(v0) + addu s5, s5, a3 + addu s6, t6, t8 + mul s6, s6, s1 + addu t1, t0, t3 + subu t4, t0, t3 + subu t4, t4, t3 + lw t3, 8(v0) + mul t0, t3, s0 + addu s7, t6, t7 + sll s7, s7, 13 + addu s7, s6, s7 + subu t2, t8, t7 + sll t2, t2, 13 + addu t2, s6, t2 + subu s6, t6, t7 + subu s6, s6, t8 + sll s6, s6, 13 + addu t3, t1, t0 + subu t5, t1, t0 + addu t6, t3, s7 + subu t3, t3, s7 + addu t7, t4, s6 + subu t4, t4, s6 + addu t8, t5, t2 + subu t5, t5, t2 + shll_s.w t6, t6, 6 + shll_s.w t3, t3, 6 + shll_s.w t7, t7, 6 + shll_s.w t4, t4, 6 + shll_s.w t8, t8, 6 + shll_s.w t5, t5, 6 + sra t6, t6, 24 + addiu t6, t6, 128 + sra t3, t3, 24 + addiu t3, t3, 128 + sb t6, 0(s5) + sra t7, t7, 24 + addiu t7, t7, 128 + sb t3, 5(s5) + sra t4, t4, 24 + addiu t4, t4, 128 + sb t7, 1(s5) + sra t8, t8, 24 + addiu t8, t8, 128 + sb t4, 4(s5) + addiu v0, v0, 24 + sra t5, t5, 24 + addiu t5, t5, 128 + sb t8, 2(s5) + addiu a2, a2, 4 + bne v0, v1, 2b + sb t5, 3(s5) + + addiu sp, sp, 144 + + RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7 + + j ra + nop + +END(jsimd_idct_6x6_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_idct_12x12_pass1_mips_dspr2) +/* + * a0 - compptr->dct_table + * a1 - coef_block + * a2 - workspace + */ + + SAVE_REGS_ON_STACK 16, s0, s1, s2, s3 + + li a3, 8 + +1: + // odd part + lh t0, 48(a1) + lh t1, 48(a0) + lh t2, 16(a1) + lh t3, 16(a0) + lh t4, 80(a1) + lh t5, 80(a0) + lh t6, 112(a1) + lh t7, 112(a0) + mul t0, t0, t1 // z2 + mul t1, t2, t3 // z1 + mul t2, t4, t5 // z3 + mul t3, t6, t7 // z4 + li t4, 10703 // FIX(1.306562965) + li t5, 4433 // FIX_0_541196100 + li t6, 7053 // FIX(0.860918669) + mul t4, t0,t4 // tmp11 + mul t5, t0,t5 // -tmp14 + addu t7, t1,t2 // tmp10 + addu t8, t7,t3 // tmp10 + z4 + mul t6, t6, t8 // tmp15 + li t8, 2139 // FIX(0.261052384) + mul t8, t7, t8 // MULTIPLY(tmp10, FIX(0.261052384)) + li t7, 2295 // FIX(0.280143716) + mul t7, t1, t7 // MULTIPLY(z1, FIX(0.280143716)) + addu t9, t2, t3 // z3 + z4 + li s0, 8565 // FIX(1.045510580) + mul t9, t9, s0 // -tmp13 + li s0, 12112 // FIX(1.478575242) + mul s0, t2, s0 // MULTIPLY(z3, FIX(1.478575242) + li s1, 12998 // FIX(1.586706681) + mul s1, t3, s1 // MULTIPLY(z4, FIX(1.586706681)) + li s2, 5540 // FIX(0.676326758) + mul s2, t1, s2 // MULTIPLY(z1, FIX(0.676326758)) + li s3, 16244 // FIX(1.982889723) + mul s3, t3, s3 // MULTIPLY(z4, FIX(1.982889723)) + subu t1, t1, t3 // z1-=z4 + subu t0, t0, t2 // z2-=z3 + addu t2, t0, t1 // z1+z2 + li t3, 4433 // FIX_0_541196100 + mul t2, t2, t3 // z3 + li t3, 6270 // FIX_0_765366865 + mul t1, t1, t3 // MULTIPLY(z1, FIX_0_765366865) + li t3, 15137 // FIX_0_765366865 + mul t0, t0, t3 // MULTIPLY(z2, FIX_1_847759065) + addu t8, t6, t8 // tmp12 + addu t3, t8, t4 // tmp12 + tmp11 + addu t3, t3, t7 // tmp10 + subu t8, t8, t9 // tmp12 + tmp13 + addu s0, t5, s0 + subu t8, t8, s0 // tmp12 + subu t9, t6, t9 + subu s1, s1, t4 + addu t9, t9, s1 // tmp13 + subu t6, t6, t5 + subu t6, t6, s2 + subu t6, t6, s3 // tmp15 + // even part start + lh t4, 64(a1) + lh t5, 64(a0) + lh t7, 32(a1) + lh s0, 32(a0) + lh s1, 0(a1) + lh s2, 0(a0) + lh s3, 96(a1) + lh v0, 96(a0) + mul t4, t4, t5 // DEQUANTIZE(inptr[DCTSIZE*4],quantptr[DCTSIZE*4]) + mul t5, t7, s0 // DEQUANTIZE(inptr[DCTSIZE*2],quantptr[DCTSIZE*2]) + mul t7, s1, s2 // DEQUANTIZE(inptr[DCTSIZE*0],quantptr[DCTSIZE*0]) + mul s0, s3, v0 // DEQUANTIZE(inptr[DCTSIZE*6],quantptr[DCTSIZE*6]) + // odd part end + addu t1, t2, t1 // tmp11 + subu t0, t2, t0 // tmp14 + // update counter and pointers + addiu a3, a3, -1 + addiu a0, a0, 2 + addiu a1, a1, 2 + // even part rest + li s1, 10033 + li s2, 11190 + mul t4, t4, s1 // z4 + mul s1, t5, s2 // z4 + sll t5, t5, 13 // z1 + sll t7, t7, 13 + addiu t7, t7, 1024 // z3 + sll s0, s0, 13 // z2 + addu s2, t7, t4 // tmp10 + subu t4, t7, t4 // tmp11 + subu s3, t5, s0 // tmp12 + addu t2, t7, s3 // tmp21 + subu s3, t7, s3 // tmp24 + addu t7, s1, s0 // tmp12 + addu v0, s2, t7 // tmp20 + subu s2, s2, t7 // tmp25 + subu s1, s1, t5 // z4 - z1 + subu s1, s1, s0 // tmp12 + addu s0, t4, s1 // tmp22 + subu t4, t4, s1 // tmp23 + // final output stage + addu t5, v0, t3 + subu v0, v0, t3 + addu t3, t2, t1 + subu t2, t2, t1 + addu t1, s0, t8 + subu s0, s0, t8 + addu t8, t4, t9 + subu t4, t4, t9 + addu t9, s3, t0 + subu s3, s3, t0 + addu t0, s2, t6 + subu s2, s2, t6 + sra t5, t5, 11 + sra t3, t3, 11 + sra t1, t1, 11 + sra t8, t8, 11 + sra t9, t9, 11 + sra t0, t0, 11 + sra s2, s2, 11 + sra s3, s3, 11 + sra t4, t4, 11 + sra s0, s0, 11 + sra t2, t2, 11 + sra v0, v0, 11 + sw t5, 0(a2) + sw t3, 32(a2) + sw t1, 64(a2) + sw t8, 96(a2) + sw t9, 128(a2) + sw t0, 160(a2) + sw s2, 192(a2) + sw s3, 224(a2) + sw t4, 256(a2) + sw s0, 288(a2) + sw t2, 320(a2) + sw v0, 352(a2) + bgtz a3, 1b + addiu a2, a2, 4 + + RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3 + + j ra + nop + +END(jsimd_idct_12x12_pass1_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_idct_12x12_pass2_mips_dspr2) +/* + * a0 - workspace + * a1 - output + */ + + SAVE_REGS_ON_STACK 16, s0, s1, s2, s3 + + li a3, 12 + +1: + // Odd part + lw t0, 12(a0) + lw t1, 4(a0) + lw t2, 20(a0) + lw t3, 28(a0) + li t4, 10703 // FIX(1.306562965) + li t5, 4433 // FIX_0_541196100 + mul t4, t0, t4 // tmp11 + mul t5, t0, t5 // -tmp14 + addu t6, t1, t2 // tmp10 + li t7, 2139 // FIX(0.261052384) + mul t7, t6, t7 // MULTIPLY(tmp10, FIX(0.261052384)) + addu t6, t6, t3 // tmp10 + z4 + li t8, 7053 // FIX(0.860918669) + mul t6, t6, t8 // tmp15 + li t8, 2295 // FIX(0.280143716) + mul t8, t1, t8 // MULTIPLY(z1, FIX(0.280143716)) + addu t9, t2, t3 // z3 + z4 + li s0, 8565 // FIX(1.045510580) + mul t9, t9, s0 // -tmp13 + li s0, 12112 // FIX(1.478575242) + mul s0, t2, s0 // MULTIPLY(z3, FIX(1.478575242)) + li s1, 12998 // FIX(1.586706681) + mul s1, t3, s1 // MULTIPLY(z4, FIX(1.586706681)) + li s2, 5540 // FIX(0.676326758) + mul s2, t1, s2 // MULTIPLY(z1, FIX(0.676326758)) + li s3, 16244 // FIX(1.982889723) + mul s3, t3, s3 // MULTIPLY(z4, FIX(1.982889723)) + subu t1, t1, t3 // z1 -= z4 + subu t0, t0, t2 // z2 -= z3 + addu t2, t1, t0 // z1 + z2 + li t3, 4433 // FIX_0_541196100 + mul t2, t2, t3 // z3 + li t3, 6270 // FIX_0_765366865 + mul t1, t1, t3 // MULTIPLY(z1, FIX_0_765366865) + li t3, 15137 // FIX_1_847759065 + mul t0, t0, t3 // MULTIPLY(z2, FIX_1_847759065) + addu t3, t6, t7 // tmp12 + addu t7, t3, t4 + addu t7, t7, t8 // tmp10 + subu t3, t3, t9 + subu t3, t3, t5 + subu t3, t3, s0 // tmp12 + subu t9, t6, t9 + subu t9, t9, t4 + addu t9, t9, s1 // tmp13 + subu t6, t6, t5 + subu t6, t6, s2 + subu t6, t6, s3 // tmp15 + addu t1, t2, t1 // tmp11 + subu t0, t2, t0 // tmp14 + // even part + lw t2, 16(a0) // z4 + lw t4, 8(a0) // z1 + lw t5, 0(a0) // z3 + lw t8, 24(a0) // z2 + li s0, 10033 // FIX(1.224744871) + li s1, 11190 // FIX(1.366025404) + mul t2, t2, s0 // z4 + mul s0, t4, s1 // z4 + addiu t5, t5, 0x10 + sll t5, t5, 13 // z3 + sll t4, t4, 13 // z1 + sll t8, t8, 13 // z2 + subu s1, t4, t8 // tmp12 + addu s2, t5, t2 // tmp10 + subu t2, t5, t2 // tmp11 + addu s3, t5, s1 // tmp21 + subu s1, t5, s1 // tmp24 + addu t5, s0, t8 // tmp12 + addu v0, s2, t5 // tmp20 + subu t5, s2, t5 // tmp25 + subu t4, s0, t4 + subu t4, t4, t8 // tmp12 + addu t8, t2, t4 // tmp22 + subu t2, t2, t4 // tmp23 + // increment counter and pointers + addiu a3, a3, -1 + addiu a0, a0, 32 + // Final stage + addu t4, v0, t7 + subu v0, v0, t7 + addu t7, s3, t1 + subu s3, s3, t1 + addu t1, t8, t3 + subu t8, t8, t3 + addu t3, t2, t9 + subu t2, t2, t9 + addu t9, s1, t0 + subu s1, s1, t0 + addu t0, t5, t6 + subu t5, t5, t6 + sll t4, t4, 4 + sll t7, t7, 4 + sll t1, t1, 4 + sll t3, t3, 4 + sll t9, t9, 4 + sll t0, t0, 4 + sll t5, t5, 4 + sll s1, s1, 4 + sll t2, t2, 4 + sll t8, t8, 4 + sll s3, s3, 4 + sll v0, v0, 4 + shll_s.w t4, t4, 2 + shll_s.w t7, t7, 2 + shll_s.w t1, t1, 2 + shll_s.w t3, t3, 2 + shll_s.w t9, t9, 2 + shll_s.w t0, t0, 2 + shll_s.w t5, t5, 2 + shll_s.w s1, s1, 2 + shll_s.w t2, t2, 2 + shll_s.w t8, t8, 2 + shll_s.w s3, s3, 2 + shll_s.w v0, v0, 2 + srl t4, t4, 24 + srl t7, t7, 24 + srl t1, t1, 24 + srl t3, t3, 24 + srl t9, t9, 24 + srl t0, t0, 24 + srl t5, t5, 24 + srl s1, s1, 24 + srl t2, t2, 24 + srl t8, t8, 24 + srl s3, s3, 24 + srl v0, v0, 24 + lw t6, 0(a1) + addiu t4, t4, 0x80 + addiu t7, t7, 0x80 + addiu t1, t1, 0x80 + addiu t3, t3, 0x80 + addiu t9, t9, 0x80 + addiu t0, t0, 0x80 + addiu t5, t5, 0x80 + addiu s1, s1, 0x80 + addiu t2, t2, 0x80 + addiu t8, t8, 0x80 + addiu s3, s3, 0x80 + addiu v0, v0, 0x80 + sb t4, 0(t6) + sb t7, 1(t6) + sb t1, 2(t6) + sb t3, 3(t6) + sb t9, 4(t6) + sb t0, 5(t6) + sb t5, 6(t6) + sb s1, 7(t6) + sb t2, 8(t6) + sb t8, 9(t6) + sb s3, 10(t6) + sb v0, 11(t6) + bgtz a3, 1b + addiu a1, a1, 4 + + RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3 + + jr ra + nop + +END(jsimd_idct_12x12_pass2_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_convsamp_mips_dspr2) +/* + * a0 - sample_data + * a1 - start_col + * a2 - workspace + */ + + lw t0, 0(a0) + li t7, 0xff80ff80 + addu t0, t0, a1 + ulw t1, 0(t0) + ulw t2, 4(t0) + preceu.ph.qbr t3, t1 + preceu.ph.qbl t4, t1 + lw t0, 4(a0) + preceu.ph.qbr t5, t2 + preceu.ph.qbl t6, t2 + addu t0, t0, a1 + addu.ph t3, t3, t7 + addu.ph t4, t4, t7 + ulw t1, 0(t0) + ulw t2, 4(t0) + addu.ph t5, t5, t7 + addu.ph t6, t6, t7 + usw t3, 0(a2) + usw t4, 4(a2) + preceu.ph.qbr t3, t1 + preceu.ph.qbl t4, t1 + usw t5, 8(a2) + usw t6, 12(a2) + + lw t0, 8(a0) + preceu.ph.qbr t5, t2 + preceu.ph.qbl t6, t2 + addu t0, t0, a1 + addu.ph t3, t3, t7 + addu.ph t4, t4, t7 + ulw t1, 0(t0) + ulw t2, 4(t0) + addu.ph t5, t5, t7 + addu.ph t6, t6, t7 + usw t3, 16(a2) + usw t4, 20(a2) + preceu.ph.qbr t3, t1 + preceu.ph.qbl t4, t1 + usw t5, 24(a2) + usw t6, 28(a2) + + lw t0, 12(a0) + preceu.ph.qbr t5, t2 + preceu.ph.qbl t6, t2 + addu t0, t0, a1 + addu.ph t3, t3, t7 + addu.ph t4, t4, t7 + ulw t1, 0(t0) + ulw t2, 4(t0) + addu.ph t5, t5, t7 + addu.ph t6, t6, t7 + usw t3, 32(a2) + usw t4, 36(a2) + preceu.ph.qbr t3, t1 + preceu.ph.qbl t4, t1 + usw t5, 40(a2) + usw t6, 44(a2) + + lw t0, 16(a0) + preceu.ph.qbr t5, t2 + preceu.ph.qbl t6, t2 + addu t0, t0, a1 + addu.ph t3, t3, t7 + addu.ph t4, t4, t7 + ulw t1, 0(t0) + ulw t2, 4(t0) + addu.ph t5, t5, t7 + addu.ph t6, t6, t7 + usw t3, 48(a2) + usw t4, 52(a2) + preceu.ph.qbr t3, t1 + preceu.ph.qbl t4, t1 + usw t5, 56(a2) + usw t6, 60(a2) + + lw t0, 20(a0) + preceu.ph.qbr t5, t2 + preceu.ph.qbl t6, t2 + addu t0, t0, a1 + addu.ph t3, t3, t7 + addu.ph t4, t4, t7 + ulw t1, 0(t0) + ulw t2, 4(t0) + addu.ph t5, t5, t7 + addu.ph t6, t6, t7 + usw t3, 64(a2) + usw t4, 68(a2) + preceu.ph.qbr t3, t1 + preceu.ph.qbl t4, t1 + usw t5, 72(a2) + usw t6, 76(a2) + + lw t0, 24(a0) + preceu.ph.qbr t5, t2 + preceu.ph.qbl t6, t2 + addu t0, t0, a1 + addu.ph t3, t3, t7 + addu.ph t4, t4, t7 + ulw t1, 0(t0) + ulw t2, 4(t0) + addu.ph t5, t5, t7 + addu.ph t6, t6, t7 + usw t3, 80(a2) + usw t4, 84(a2) + preceu.ph.qbr t3, t1 + preceu.ph.qbl t4, t1 + usw t5, 88(a2) + usw t6, 92(a2) + + lw t0, 28(a0) + preceu.ph.qbr t5, t2 + preceu.ph.qbl t6, t2 + addu t0, t0, a1 + addu.ph t3, t3, t7 + addu.ph t4, t4, t7 + ulw t1, 0(t0) + ulw t2, 4(t0) + addu.ph t5, t5, t7 + addu.ph t6, t6, t7 + usw t3, 96(a2) + usw t4, 100(a2) + preceu.ph.qbr t3, t1 + preceu.ph.qbl t4, t1 + usw t5, 104(a2) + usw t6, 108(a2) + preceu.ph.qbr t5, t2 + preceu.ph.qbl t6, t2 + addu.ph t3, t3, t7 + addu.ph t4, t4, t7 + addu.ph t5, t5, t7 + addu.ph t6, t6, t7 + usw t3, 112(a2) + usw t4, 116(a2) + usw t5, 120(a2) + usw t6, 124(a2) + + j ra + nop + +END(jsimd_convsamp_mips_dspr2) + +/*****************************************************************************/ +LEAF_MIPS_DSPR2(jsimd_convsamp_float_mips_dspr2) +/* + * a0 - sample_data + * a1 - start_col + * a2 - workspace + */ + + .set at + + lw t0, 0(a0) + addu t0, t0, a1 + lbu t1, 0(t0) + lbu t2, 1(t0) + lbu t3, 2(t0) + lbu t4, 3(t0) + lbu t5, 4(t0) + lbu t6, 5(t0) + lbu t7, 6(t0) + lbu t8, 7(t0) + addiu t1, t1, -128 + addiu t2, t2, -128 + addiu t3, t3, -128 + addiu t4, t4, -128 + addiu t5, t5, -128 + addiu t6, t6, -128 + addiu t7, t7, -128 + addiu t8, t8, -128 + mtc1 t1, f1 + mtc1 t2, f2 + mtc1 t3, f3 + mtc1 t4, f4 + mtc1 t5, f5 + mtc1 t6, f6 + mtc1 t7, f7 + mtc1 t8, f8 + cvt.s.w f1, f1 + cvt.s.w f2, f2 + cvt.s.w f3, f3 + cvt.s.w f4, f4 + cvt.s.w f5, f5 + cvt.s.w f6, f6 + cvt.s.w f7, f7 + cvt.s.w f8, f8 + lw t0, 4(a0) + swc1 f1, 0(a2) + swc1 f2, 4(a2) + swc1 f3, 8(a2) + addu t0, t0, a1 + swc1 f4, 12(a2) + swc1 f5, 16(a2) + swc1 f6, 20(a2) + swc1 f7, 24(a2) + swc1 f8, 28(a2) + //elemr 1 + lbu t1, 0(t0) + lbu t2, 1(t0) + lbu t3, 2(t0) + lbu t4, 3(t0) + lbu t5, 4(t0) + lbu t6, 5(t0) + lbu t7, 6(t0) + lbu t8, 7(t0) + addiu t1, t1, -128 + addiu t2, t2, -128 + addiu t3, t3, -128 + addiu t4, t4, -128 + addiu t5, t5, -128 + addiu t6, t6, -128 + addiu t7, t7, -128 + addiu t8, t8, -128 + mtc1 t1, f1 + mtc1 t2, f2 + mtc1 t3, f3 + mtc1 t4, f4 + mtc1 t5, f5 + mtc1 t6, f6 + mtc1 t7, f7 + mtc1 t8, f8 + cvt.s.w f1, f1 + cvt.s.w f2, f2 + cvt.s.w f3, f3 + cvt.s.w f4, f4 + cvt.s.w f5, f5 + cvt.s.w f6, f6 + cvt.s.w f7, f7 + cvt.s.w f8, f8 + lw t0, 8(a0) + swc1 f1, 32(a2) + swc1 f2, 36(a2) + swc1 f3, 40(a2) + addu t0, t0, a1 + swc1 f4, 44(a2) + swc1 f5, 48(a2) + swc1 f6, 52(a2) + swc1 f7, 56(a2) + swc1 f8, 60(a2) + //elemr 2 + lbu t1, 0(t0) + lbu t2, 1(t0) + lbu t3, 2(t0) + lbu t4, 3(t0) + lbu t5, 4(t0) + lbu t6, 5(t0) + lbu t7, 6(t0) + lbu t8, 7(t0) + addiu t1, t1, -128 + addiu t2, t2, -128 + addiu t3, t3, -128 + addiu t4, t4, -128 + addiu t5, t5, -128 + addiu t6, t6, -128 + addiu t7, t7, -128 + addiu t8, t8, -128 + mtc1 t1, f1 + mtc1 t2, f2 + mtc1 t3, f3 + mtc1 t4, f4 + mtc1 t5, f5 + mtc1 t6, f6 + mtc1 t7, f7 + mtc1 t8, f8 + cvt.s.w f1, f1 + cvt.s.w f2, f2 + cvt.s.w f3, f3 + cvt.s.w f4, f4 + cvt.s.w f5, f5 + cvt.s.w f6, f6 + cvt.s.w f7, f7 + cvt.s.w f8, f8 + lw t0, 12(a0) + swc1 f1, 64(a2) + swc1 f2, 68(a2) + swc1 f3, 72(a2) + addu t0, t0, a1 + swc1 f4, 76(a2) + swc1 f5, 80(a2) + swc1 f6, 84(a2) + swc1 f7, 88(a2) + swc1 f8, 92(a2) + //elemr 3 + lbu t1, 0(t0) + lbu t2, 1(t0) + lbu t3, 2(t0) + lbu t4, 3(t0) + lbu t5, 4(t0) + lbu t6, 5(t0) + lbu t7, 6(t0) + lbu t8, 7(t0) + addiu t1, t1, -128 + addiu t2, t2, -128 + addiu t3, t3, -128 + addiu t4, t4, -128 + addiu t5, t5, -128 + addiu t6, t6, -128 + addiu t7, t7, -128 + addiu t8, t8, -128 + mtc1 t1, f1 + mtc1 t2, f2 + mtc1 t3, f3 + mtc1 t4, f4 + mtc1 t5, f5 + mtc1 t6, f6 + mtc1 t7, f7 + mtc1 t8, f8 + cvt.s.w f1, f1 + cvt.s.w f2, f2 + cvt.s.w f3, f3 + cvt.s.w f4, f4 + cvt.s.w f5, f5 + cvt.s.w f6, f6 + cvt.s.w f7, f7 + cvt.s.w f8, f8 + lw t0, 16(a0) + swc1 f1, 96(a2) + swc1 f2, 100(a2) + swc1 f3, 104(a2) + addu t0, t0, a1 + swc1 f4, 108(a2) + swc1 f5, 112(a2) + swc1 f6, 116(a2) + swc1 f7, 120(a2) + swc1 f8, 124(a2) + //elemr 4 + lbu t1, 0(t0) + lbu t2, 1(t0) + lbu t3, 2(t0) + lbu t4, 3(t0) + lbu t5, 4(t0) + lbu t6, 5(t0) + lbu t7, 6(t0) + lbu t8, 7(t0) + addiu t1, t1, -128 + addiu t2, t2, -128 + addiu t3, t3, -128 + addiu t4, t4, -128 + addiu t5, t5, -128 + addiu t6, t6, -128 + addiu t7, t7, -128 + addiu t8, t8, -128 + mtc1 t1, f1 + mtc1 t2, f2 + mtc1 t3, f3 + mtc1 t4, f4 + mtc1 t5, f5 + mtc1 t6, f6 + mtc1 t7, f7 + mtc1 t8, f8 + cvt.s.w f1, f1 + cvt.s.w f2, f2 + cvt.s.w f3, f3 + cvt.s.w f4, f4 + cvt.s.w f5, f5 + cvt.s.w f6, f6 + cvt.s.w f7, f7 + cvt.s.w f8, f8 + lw t0, 20(a0) + swc1 f1, 128(a2) + swc1 f2, 132(a2) + swc1 f3, 136(a2) + addu t0, t0, a1 + swc1 f4, 140(a2) + swc1 f5, 144(a2) + swc1 f6, 148(a2) + swc1 f7, 152(a2) + swc1 f8, 156(a2) + //elemr 5 + lbu t1, 0(t0) + lbu t2, 1(t0) + lbu t3, 2(t0) + lbu t4, 3(t0) + lbu t5, 4(t0) + lbu t6, 5(t0) + lbu t7, 6(t0) + lbu t8, 7(t0) + addiu t1, t1, -128 + addiu t2, t2, -128 + addiu t3, t3, -128 + addiu t4, t4, -128 + addiu t5, t5, -128 + addiu t6, t6, -128 + addiu t7, t7, -128 + addiu t8, t8, -128 + mtc1 t1, f1 + mtc1 t2, f2 + mtc1 t3, f3 + mtc1 t4, f4 + mtc1 t5, f5 + mtc1 t6, f6 + mtc1 t7, f7 + mtc1 t8, f8 + cvt.s.w f1, f1 + cvt.s.w f2, f2 + cvt.s.w f3, f3 + cvt.s.w f4, f4 + cvt.s.w f5, f5 + cvt.s.w f6, f6 + cvt.s.w f7, f7 + cvt.s.w f8, f8 + lw t0, 24(a0) + swc1 f1, 160(a2) + swc1 f2, 164(a2) + swc1 f3, 168(a2) + addu t0, t0, a1 + swc1 f4, 172(a2) + swc1 f5, 176(a2) + swc1 f6, 180(a2) + swc1 f7, 184(a2) + swc1 f8, 188(a2) + //elemr 6 + lbu t1, 0(t0) + lbu t2, 1(t0) + lbu t3, 2(t0) + lbu t4, 3(t0) + lbu t5, 4(t0) + lbu t6, 5(t0) + lbu t7, 6(t0) + lbu t8, 7(t0) + addiu t1, t1, -128 + addiu t2, t2, -128 + addiu t3, t3, -128 + addiu t4, t4, -128 + addiu t5, t5, -128 + addiu t6, t6, -128 + addiu t7, t7, -128 + addiu t8, t8, -128 + mtc1 t1, f1 + mtc1 t2, f2 + mtc1 t3, f3 + mtc1 t4, f4 + mtc1 t5, f5 + mtc1 t6, f6 + mtc1 t7, f7 + mtc1 t8, f8 + cvt.s.w f1, f1 + cvt.s.w f2, f2 + cvt.s.w f3, f3 + cvt.s.w f4, f4 + cvt.s.w f5, f5 + cvt.s.w f6, f6 + cvt.s.w f7, f7 + cvt.s.w f8, f8 + lw t0, 28(a0) + swc1 f1, 192(a2) + swc1 f2, 196(a2) + swc1 f3, 200(a2) + addu t0, t0, a1 + swc1 f4, 204(a2) + swc1 f5, 208(a2) + swc1 f6, 212(a2) + swc1 f7, 216(a2) + swc1 f8, 220(a2) + //elemr 7 + lbu t1, 0(t0) + lbu t2, 1(t0) + lbu t3, 2(t0) + lbu t4, 3(t0) + lbu t5, 4(t0) + lbu t6, 5(t0) + lbu t7, 6(t0) + lbu t8, 7(t0) + addiu t1, t1, -128 + addiu t2, t2, -128 + addiu t3, t3, -128 + addiu t4, t4, -128 + addiu t5, t5, -128 + addiu t6, t6, -128 + addiu t7, t7, -128 + addiu t8, t8, -128 + mtc1 t1, f1 + mtc1 t2, f2 + mtc1 t3, f3 + mtc1 t4, f4 + mtc1 t5, f5 + mtc1 t6, f6 + mtc1 t7, f7 + mtc1 t8, f8 + cvt.s.w f1, f1 + cvt.s.w f2, f2 + cvt.s.w f3, f3 + cvt.s.w f4, f4 + cvt.s.w f5, f5 + cvt.s.w f6, f6 + cvt.s.w f7, f7 + cvt.s.w f8, f8 + swc1 f1, 224(a2) + swc1 f2, 228(a2) + swc1 f3, 232(a2) + swc1 f4, 236(a2) + swc1 f5, 240(a2) + swc1 f6, 244(a2) + swc1 f7, 248(a2) + swc1 f8, 252(a2) + + j ra + nop + +END(jsimd_convsamp_float_mips_dspr2) + +/*****************************************************************************/ + diff --git a/app/src/main/jni/libjpeg-turbo/simd/jsimd_mips_dspr2_asm.h b/app/src/main/jni/libjpeg-turbo/simd/jsimd_mips_dspr2_asm.h new file mode 100644 index 000000000..50ec31bfb --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jsimd_mips_dspr2_asm.h @@ -0,0 +1,285 @@ +/* + * MIPS DSPr2 optimizations for libjpeg-turbo + * + * Copyright (C) 2013, MIPS Technologies, Inc., California. + * All rights reserved. + * Authors: Teodora Novkovic (teodora.novkovic@imgtec.com) + * Darko Laus (darko.laus@imgtec.com) + * This software is provided 'as-is', without any express or implied + * warranty. In no event will the authors be held liable for any damages + * arising from the use of this software. + * + * Permission is granted to anyone to use this software for any purpose, + * including commercial applications, and to alter it and redistribute it + * freely, subject to the following restrictions: + * + * 1. The origin of this software must not be misrepresented; you must not + * claim that you wrote the original software. If you use this software + * in a product, an acknowledgment in the product documentation would be + * appreciated but is not required. + * 2. Altered source versions must be plainly marked as such, and must not be + * misrepresented as being the original software. + * 3. This notice may not be removed or altered from any source distribution. + */ + +#define zero $0 +#define AT $1 +#define v0 $2 +#define v1 $3 +#define a0 $4 +#define a1 $5 +#define a2 $6 +#define a3 $7 +#define t0 $8 +#define t1 $9 +#define t2 $10 +#define t3 $11 +#define t4 $12 +#define t5 $13 +#define t6 $14 +#define t7 $15 +#define s0 $16 +#define s1 $17 +#define s2 $18 +#define s3 $19 +#define s4 $20 +#define s5 $21 +#define s6 $22 +#define s7 $23 +#define t8 $24 +#define t9 $25 +#define k0 $26 +#define k1 $27 +#define gp $28 +#define sp $29 +#define fp $30 +#define s8 $30 +#define ra $31 + +#define f0 $f0 +#define f1 $f1 +#define f2 $f2 +#define f3 $f3 +#define f4 $f4 +#define f5 $f5 +#define f6 $f6 +#define f7 $f7 +#define f8 $f8 +#define f9 $f9 +#define f10 $f10 +#define f11 $f11 +#define f12 $f12 +#define f13 $f13 +#define f14 $f14 +#define f15 $f15 +#define f16 $f16 +#define f17 $f17 +#define f18 $f18 +#define f19 $f19 +#define f20 $f20 +#define f21 $f21 +#define f22 $f22 +#define f23 $f23 +#define f24 $f24 +#define f25 $f25 +#define f26 $f26 +#define f27 $f27 +#define f28 $f28 +#define f29 $f29 +#define f30 $f30 +#define f31 $f31 + +/* + * LEAF_MIPS32R2 - declare leaf routine for MIPS32r2 + */ +#define LEAF_MIPS32R2(symbol) \ + .globl symbol; \ + .align 2; \ + .type symbol, @function; \ + .ent symbol, 0; \ +symbol: .frame sp, 0, ra; \ + .set push; \ + .set arch=mips32r2; \ + .set noreorder; \ + .set noat; + +/* + * LEAF_MIPS_DSPR2 - declare leaf routine for MIPS DSPr2 + */ +#define LEAF_MIPS_DSPR2(symbol) \ +LEAF_MIPS32R2(symbol) \ + .set dspr2; + +/* + * END - mark end of function + */ +#define END(function) \ + .set pop; \ + .end function; \ + .size function,.-function + +/* + * Checks if stack offset is big enough for storing/restoring regs_num + * number of register to/from stack. Stack offset must be greater than + * or equal to the number of bytes needed for storing registers (regs_num*4). + * Since MIPS ABI allows usage of first 16 bytes of stack frame (this is + * preserved for input arguments of the functions, already stored in a0-a3), + * stack size can be further optimized by utilizing this space. + */ +.macro CHECK_STACK_OFFSET regs_num, stack_offset +.if \stack_offset < \regs_num * 4 - 16 +.error "Stack offset too small." +.endif +.endm + +/* + * Saves set of registers on stack. Maximum number of registers that + * can be saved on stack is limitted to 14 (a0-a3, v0-v1 and s0-s7). + * Stack offset is number of bytes that are added to stack pointer (sp) + * before registers are pushed in order to provide enough space on stack + * (offset must be multiple of 4, and must be big enough, as described by + * CHECK_STACK_OFFSET macro). This macro is intended to be used in + * combination with RESTORE_REGS_FROM_STACK macro. Example: + * SAVE_REGS_ON_STACK 4, v0, v1, s0, s1 + * RESTORE_REGS_FROM_STACK 4, v0, v1, s0, s1 + */ +.macro SAVE_REGS_ON_STACK stack_offset = 0, r1, \ + r2 = 0, r3 = 0, r4 = 0, \ + r5 = 0, r6 = 0, r7 = 0, \ + r8 = 0, r9 = 0, r10 = 0, \ + r11 = 0, r12 = 0, r13 = 0, \ + r14 = 0 + .if (\stack_offset < 0) || (\stack_offset - (\stack_offset / 4) * 4) + .error "Stack offset must be pozitive and multiple of 4." + .endif + .if \stack_offset != 0 + addiu sp, sp, -\stack_offset + .endif + sw \r1, 0(sp) + .if \r2 != 0 + sw \r2, 4(sp) + .endif + .if \r3 != 0 + sw \r3, 8(sp) + .endif + .if \r4 != 0 + sw \r4, 12(sp) + .endif + .if \r5 != 0 + CHECK_STACK_OFFSET 5, \stack_offset + sw \r5, 16(sp) + .endif + .if \r6 != 0 + CHECK_STACK_OFFSET 6, \stack_offset + sw \r6, 20(sp) + .endif + .if \r7 != 0 + CHECK_STACK_OFFSET 7, \stack_offset + sw \r7, 24(sp) + .endif + .if \r8 != 0 + CHECK_STACK_OFFSET 8, \stack_offset + sw \r8, 28(sp) + .endif + .if \r9 != 0 + CHECK_STACK_OFFSET 9, \stack_offset + sw \r9, 32(sp) + .endif + .if \r10 != 0 + CHECK_STACK_OFFSET 10, \stack_offset + sw \r10, 36(sp) + .endif + .if \r11 != 0 + CHECK_STACK_OFFSET 11, \stack_offset + sw \r11, 40(sp) + .endif + .if \r12 != 0 + CHECK_STACK_OFFSET 12, \stack_offset + sw \r12, 44(sp) + .endif + .if \r13 != 0 + CHECK_STACK_OFFSET 13, \stack_offset + sw \r13, 48(sp) + .endif + .if \r14 != 0 + CHECK_STACK_OFFSET 14, \stack_offset + sw \r14, 52(sp) + .endif +.endm + +/* + * Restores set of registers from stack. Maximum number of registers that + * can be restored from stack is limitted to 14 (a0-a3, v0-v1 and s0-s7). + * Stack offset is number of bytes that are added to stack pointer (sp) + * after registers are restored (offset must be multiple of 4, and must + * be big enough, as described by CHECK_STACK_OFFSET macro). This macro is + * intended to be used in combination with RESTORE_REGS_FROM_STACK macro. + * Example: + * SAVE_REGS_ON_STACK 4, v0, v1, s0, s1 + * RESTORE_REGS_FROM_STACK 4, v0, v1, s0, s1 + */ +.macro RESTORE_REGS_FROM_STACK stack_offset = 0, r1, \ + r2 = 0, r3 = 0, r4 = 0, \ + r5 = 0, r6 = 0, r7 = 0, \ + r8 = 0, r9 = 0, r10 = 0, \ + r11 = 0, r12 = 0, r13 = 0, \ + r14 = 0 + .if (\stack_offset < 0) || (\stack_offset - (\stack_offset/4)*4) + .error "Stack offset must be pozitive and multiple of 4." + .endif + lw \r1, 0(sp) + .if \r2 != 0 + lw \r2, 4(sp) + .endif + .if \r3 != 0 + lw \r3, 8(sp) + .endif + .if \r4 != 0 + lw \r4, 12(sp) + .endif + .if \r5 != 0 + CHECK_STACK_OFFSET 5, \stack_offset + lw \r5, 16(sp) + .endif + .if \r6 != 0 + CHECK_STACK_OFFSET 6, \stack_offset + lw \r6, 20(sp) + .endif + .if \r7 != 0 + CHECK_STACK_OFFSET 7, \stack_offset + lw \r7, 24(sp) + .endif + .if \r8 != 0 + CHECK_STACK_OFFSET 8, \stack_offset + lw \r8, 28(sp) + .endif + .if \r9 != 0 + CHECK_STACK_OFFSET 9, \stack_offset + lw \r9, 32(sp) + .endif + .if \r10 != 0 + CHECK_STACK_OFFSET 10, \stack_offset + lw \r10, 36(sp) + .endif + .if \r11 != 0 + CHECK_STACK_OFFSET 11, \stack_offset + lw \r11, 40(sp) + .endif + .if \r12 != 0 + CHECK_STACK_OFFSET 12, \stack_offset + lw \r12, 44(sp) + .endif + .if \r13 != 0 + CHECK_STACK_OFFSET 13, \stack_offset + lw \r13, 48(sp) + .endif + .if \r14 != 0 + CHECK_STACK_OFFSET 14, \stack_offset + lw \r14, 52(sp) + .endif + .if \stack_offset != 0 + addiu sp, sp, \stack_offset + .endif +.endm + + diff --git a/app/src/main/jni/libjpeg-turbo/simd/jsimd_x86_64.c b/app/src/main/jni/libjpeg-turbo/simd/jsimd_x86_64.c new file mode 100644 index 000000000..2c47a7f36 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jsimd_x86_64.c @@ -0,0 +1,757 @@ +/* + * jsimd_x86_64.c + * + * Copyright 2009 Pierre Ossman for Cendio AB + * Copyright 2009-2011, 2014 D. R. Commander + * + * Based on the x86 SIMD extension for IJG JPEG library, + * Copyright (C) 1999-2006, MIYASAKA Masaru. + * For conditions of distribution and use, see copyright notice in jsimdext.inc + * + * This file contains the interface between the "normal" portions + * of the library and the SIMD implementations when running on a + * 64-bit x86 architecture. + */ + +#define JPEG_INTERNALS +#include "../jinclude.h" +#include "../jpeglib.h" +#include "../jsimd.h" +#include "../jdct.h" +#include "../jsimddct.h" +#include "jsimd.h" + +/* + * In the PIC cases, we have no guarantee that constants will keep + * their alignment. This macro allows us to verify it at runtime. + */ +#define IS_ALIGNED(ptr, order) (((size_t)ptr & ((1 << order) - 1)) == 0) + +#define IS_ALIGNED_SSE(ptr) (IS_ALIGNED(ptr, 4)) /* 16 byte alignment */ + +GLOBAL(int) +jsimd_can_rgb_ycc (void) +{ + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) + return 0; + + if (!IS_ALIGNED_SSE(jconst_rgb_ycc_convert_sse2)) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_rgb_gray (void) +{ + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) + return 0; + + if (!IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2)) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_ycc_rgb (void) +{ + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4)) + return 0; + + if (!IS_ALIGNED_SSE(jconst_ycc_rgb_convert_sse2)) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_ycc_rgb565 (void) +{ + return 0; +} + +GLOBAL(void) +jsimd_rgb_ycc_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ + void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int); + + switch(cinfo->in_color_space) { + case JCS_EXT_RGB: + sse2fct=jsimd_extrgb_ycc_convert_sse2; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + sse2fct=jsimd_extrgbx_ycc_convert_sse2; + break; + case JCS_EXT_BGR: + sse2fct=jsimd_extbgr_ycc_convert_sse2; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + sse2fct=jsimd_extbgrx_ycc_convert_sse2; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + sse2fct=jsimd_extxbgr_ycc_convert_sse2; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + sse2fct=jsimd_extxrgb_ycc_convert_sse2; + break; + default: + sse2fct=jsimd_rgb_ycc_convert_sse2; + break; + } + + sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows); +} + +GLOBAL(void) +jsimd_rgb_gray_convert (j_compress_ptr cinfo, + JSAMPARRAY input_buf, JSAMPIMAGE output_buf, + JDIMENSION output_row, int num_rows) +{ + void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int); + + switch(cinfo->in_color_space) { + case JCS_EXT_RGB: + sse2fct=jsimd_extrgb_gray_convert_sse2; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + sse2fct=jsimd_extrgbx_gray_convert_sse2; + break; + case JCS_EXT_BGR: + sse2fct=jsimd_extbgr_gray_convert_sse2; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + sse2fct=jsimd_extbgrx_gray_convert_sse2; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + sse2fct=jsimd_extxbgr_gray_convert_sse2; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + sse2fct=jsimd_extxrgb_gray_convert_sse2; + break; + default: + sse2fct=jsimd_rgb_gray_convert_sse2; + break; + } + + sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows); +} + +GLOBAL(void) +jsimd_ycc_rgb_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ + void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int); + + switch(cinfo->out_color_space) { + case JCS_EXT_RGB: + sse2fct=jsimd_ycc_extrgb_convert_sse2; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + sse2fct=jsimd_ycc_extrgbx_convert_sse2; + break; + case JCS_EXT_BGR: + sse2fct=jsimd_ycc_extbgr_convert_sse2; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + sse2fct=jsimd_ycc_extbgrx_convert_sse2; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + sse2fct=jsimd_ycc_extxbgr_convert_sse2; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + sse2fct=jsimd_ycc_extxrgb_convert_sse2; + break; + default: + sse2fct=jsimd_ycc_rgb_convert_sse2; + break; + } + + sse2fct(cinfo->output_width, input_buf, input_row, output_buf, num_rows); +} + +GLOBAL(void) +jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, JDIMENSION input_row, + JSAMPARRAY output_buf, int num_rows) +{ +} + +GLOBAL(int) +jsimd_can_h2v2_downsample (void) +{ + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_h2v1_downsample (void) +{ + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + return 1; +} + +GLOBAL(void) +jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ + jsimd_h2v2_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor, + compptr->v_samp_factor, compptr->width_in_blocks, + input_data, output_data); +} + +GLOBAL(void) +jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr, + JSAMPARRAY input_data, JSAMPARRAY output_data) +{ + jsimd_h2v1_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor, + compptr->v_samp_factor, compptr->width_in_blocks, + input_data, output_data); +} + +GLOBAL(int) +jsimd_can_h2v2_upsample (void) +{ + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_h2v1_upsample (void) +{ + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + return 1; +} + +GLOBAL(void) +jsimd_h2v2_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ + jsimd_h2v2_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width, + input_data, output_data_ptr); +} + +GLOBAL(void) +jsimd_h2v1_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ + jsimd_h2v1_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width, + input_data, output_data_ptr); +} + +GLOBAL(int) +jsimd_can_h2v2_fancy_upsample (void) +{ + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (!IS_ALIGNED_SSE(jconst_fancy_upsample_sse2)) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_h2v1_fancy_upsample (void) +{ + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (!IS_ALIGNED_SSE(jconst_fancy_upsample_sse2)) + return 0; + + return 1; +} + +GLOBAL(void) +jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ + jsimd_h2v2_fancy_upsample_sse2(cinfo->max_v_samp_factor, + compptr->downsampled_width, input_data, + output_data_ptr); +} + +GLOBAL(void) +jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo, + jpeg_component_info * compptr, + JSAMPARRAY input_data, + JSAMPARRAY * output_data_ptr) +{ + jsimd_h2v1_fancy_upsample_sse2(cinfo->max_v_samp_factor, + compptr->downsampled_width, input_data, + output_data_ptr); +} + +GLOBAL(int) +jsimd_can_h2v2_merged_upsample (void) +{ + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (!IS_ALIGNED_SSE(jconst_merged_upsample_sse2)) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_h2v1_merged_upsample (void) +{ + /* The code is optimised for these values only */ + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + + if (!IS_ALIGNED_SSE(jconst_merged_upsample_sse2)) + return 0; + + return 1; +} + +GLOBAL(void) +jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY); + + switch(cinfo->out_color_space) { + case JCS_EXT_RGB: + sse2fct=jsimd_h2v2_extrgb_merged_upsample_sse2; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + sse2fct=jsimd_h2v2_extrgbx_merged_upsample_sse2; + break; + case JCS_EXT_BGR: + sse2fct=jsimd_h2v2_extbgr_merged_upsample_sse2; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + sse2fct=jsimd_h2v2_extbgrx_merged_upsample_sse2; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + sse2fct=jsimd_h2v2_extxbgr_merged_upsample_sse2; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + sse2fct=jsimd_h2v2_extxrgb_merged_upsample_sse2; + break; + default: + sse2fct=jsimd_h2v2_merged_upsample_sse2; + break; + } + + sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf); +} + +GLOBAL(void) +jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo, + JSAMPIMAGE input_buf, + JDIMENSION in_row_group_ctr, + JSAMPARRAY output_buf) +{ + void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY); + + switch(cinfo->out_color_space) { + case JCS_EXT_RGB: + sse2fct=jsimd_h2v1_extrgb_merged_upsample_sse2; + break; + case JCS_EXT_RGBX: + case JCS_EXT_RGBA: + sse2fct=jsimd_h2v1_extrgbx_merged_upsample_sse2; + break; + case JCS_EXT_BGR: + sse2fct=jsimd_h2v1_extbgr_merged_upsample_sse2; + break; + case JCS_EXT_BGRX: + case JCS_EXT_BGRA: + sse2fct=jsimd_h2v1_extbgrx_merged_upsample_sse2; + break; + case JCS_EXT_XBGR: + case JCS_EXT_ABGR: + sse2fct=jsimd_h2v1_extxbgr_merged_upsample_sse2; + break; + case JCS_EXT_XRGB: + case JCS_EXT_ARGB: + sse2fct=jsimd_h2v1_extxrgb_merged_upsample_sse2; + break; + default: + sse2fct=jsimd_h2v1_merged_upsample_sse2; + break; + } + + sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf); +} + +GLOBAL(int) +jsimd_can_convsamp (void) +{ + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_convsamp_float (void) +{ + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(FAST_FLOAT) != 4) + return 0; + + return 1; +} + +GLOBAL(void) +jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col, + DCTELEM * workspace) +{ + jsimd_convsamp_sse2(sample_data, start_col, workspace); +} + +GLOBAL(void) +jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col, + FAST_FLOAT * workspace) +{ + jsimd_convsamp_float_sse2(sample_data, start_col, workspace); +} + +GLOBAL(int) +jsimd_can_fdct_islow (void) +{ + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + if (!IS_ALIGNED_SSE(jconst_fdct_islow_sse2)) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_fdct_ifast (void) +{ + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + if (!IS_ALIGNED_SSE(jconst_fdct_ifast_sse2)) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_fdct_float (void) +{ + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(FAST_FLOAT) != 4) + return 0; + + if (!IS_ALIGNED_SSE(jconst_fdct_float_sse)) + return 0; + + return 1; +} + +GLOBAL(void) +jsimd_fdct_islow (DCTELEM * data) +{ + jsimd_fdct_islow_sse2(data); +} + +GLOBAL(void) +jsimd_fdct_ifast (DCTELEM * data) +{ + jsimd_fdct_ifast_sse2(data); +} + +GLOBAL(void) +jsimd_fdct_float (FAST_FLOAT * data) +{ + jsimd_fdct_float_sse(data); +} + +GLOBAL(int) +jsimd_can_quantize (void) +{ + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (sizeof(DCTELEM) != 2) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_quantize_float (void) +{ + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (sizeof(FAST_FLOAT) != 4) + return 0; + + return 1; +} + +GLOBAL(void) +jsimd_quantize (JCOEFPTR coef_block, DCTELEM * divisors, + DCTELEM * workspace) +{ + jsimd_quantize_sse2(coef_block, divisors, workspace); +} + +GLOBAL(void) +jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT * divisors, + FAST_FLOAT * workspace) +{ + jsimd_quantize_float_sse2(coef_block, divisors, workspace); +} + +GLOBAL(int) +jsimd_can_idct_2x2 (void) +{ + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (!IS_ALIGNED_SSE(jconst_idct_red_sse2)) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_idct_4x4 (void) +{ + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (!IS_ALIGNED_SSE(jconst_idct_red_sse2)) + return 0; + + return 1; +} + +GLOBAL(void) +jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + jsimd_idct_2x2_sse2(compptr->dct_table, coef_block, output_buf, output_col); +} + +GLOBAL(void) +jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + jsimd_idct_4x4_sse2(compptr->dct_table, coef_block, output_buf, output_col); +} + +GLOBAL(int) +jsimd_can_idct_islow (void) +{ + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(ISLOW_MULT_TYPE) != 2) + return 0; + + if (!IS_ALIGNED_SSE(jconst_idct_islow_sse2)) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_idct_ifast (void) +{ + /* The code is optimised for these values only */ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(IFAST_MULT_TYPE) != 2) + return 0; + if (IFAST_SCALE_BITS != 2) + return 0; + + if (!IS_ALIGNED_SSE(jconst_idct_ifast_sse2)) + return 0; + + return 1; +} + +GLOBAL(int) +jsimd_can_idct_float (void) +{ + if (DCTSIZE != 8) + return 0; + if (sizeof(JCOEF) != 2) + return 0; + if (BITS_IN_JSAMPLE != 8) + return 0; + if (sizeof(JDIMENSION) != 4) + return 0; + if (sizeof(FAST_FLOAT) != 4) + return 0; + if (sizeof(FLOAT_MULT_TYPE) != 4) + return 0; + + if (!IS_ALIGNED_SSE(jconst_idct_float_sse2)) + return 0; + + return 1; +} + +GLOBAL(void) +jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + jsimd_idct_islow_sse2(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(void) +jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + jsimd_idct_ifast_sse2(compptr->dct_table, coef_block, output_buf, + output_col); +} + +GLOBAL(void) +jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr, + JCOEFPTR coef_block, JSAMPARRAY output_buf, + JDIMENSION output_col) +{ + jsimd_idct_float_sse2(compptr->dct_table, coef_block, output_buf, + output_col); +} + diff --git a/app/src/main/jni/libjpeg-turbo/simd/jsimdcfg.inc b/app/src/main/jni/libjpeg-turbo/simd/jsimdcfg.inc new file mode 100644 index 000000000..8e71b5d8e --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jsimdcfg.inc @@ -0,0 +1,92 @@ +; +; Automatically generated include file from jsimdcfg.inc.h +; +; +; -- jpeglib.h +; +%define DCTSIZE 8 +%define DCTSIZE2 64 +; +; -- jmorecfg.h +; +%define RGB_RED 0 +%define RGB_GREEN 1 +%define RGB_BLUE 2 +%define RGB_PIXELSIZE 3 +%define EXT_RGB_RED 0 +%define EXT_RGB_GREEN 1 +%define EXT_RGB_BLUE 2 +%define EXT_RGB_PIXELSIZE 3 +%define EXT_RGBX_RED 0 +%define EXT_RGBX_GREEN 1 +%define EXT_RGBX_BLUE 2 +%define EXT_RGBX_PIXELSIZE 4 +%define EXT_BGR_RED 2 +%define EXT_BGR_GREEN 1 +%define EXT_BGR_BLUE 0 +%define EXT_BGR_PIXELSIZE 3 +%define EXT_BGRX_RED 2 +%define EXT_BGRX_GREEN 1 +%define EXT_BGRX_BLUE 0 +%define EXT_BGRX_PIXELSIZE 4 +%define EXT_XBGR_RED 3 +%define EXT_XBGR_GREEN 2 +%define EXT_XBGR_BLUE 1 +%define EXT_XBGR_PIXELSIZE 4 +%define EXT_XRGB_RED 1 +%define EXT_XRGB_GREEN 2 +%define EXT_XRGB_BLUE 3 +%define EXT_XRGB_PIXELSIZE 4 +%define RGBX_FILLER_0XFF 1 +; Representation of a single sample (pixel element value). +; On this SIMD implementation, this must be 'unsigned char'. +; +%define JSAMPLE byte ; unsigned char +%define SIZEOF_JSAMPLE SIZEOF_BYTE ; sizeof(JSAMPLE) +%define CENTERJSAMPLE 128 +; Representation of a DCT frequency coefficient. +; On this SIMD implementation, this must be 'short'. +; +%define JCOEF word ; short +%define SIZEOF_JCOEF SIZEOF_WORD ; sizeof(JCOEF) +; Datatype used for image dimensions. +; On this SIMD implementation, this must be 'unsigned int'. +; +%define JDIMENSION dword ; unsigned int +%define SIZEOF_JDIMENSION SIZEOF_DWORD ; sizeof(JDIMENSION) +%define JSAMPROW POINTER ; JSAMPLE * (jpeglib.h) +%define JSAMPARRAY POINTER ; JSAMPROW * (jpeglib.h) +%define JSAMPIMAGE POINTER ; JSAMPARRAY * (jpeglib.h) +%define JCOEFPTR POINTER ; JCOEF * (jpeglib.h) +%define SIZEOF_JSAMPROW SIZEOF_POINTER ; sizeof(JSAMPROW) +%define SIZEOF_JSAMPARRAY SIZEOF_POINTER ; sizeof(JSAMPARRAY) +%define SIZEOF_JSAMPIMAGE SIZEOF_POINTER ; sizeof(JSAMPIMAGE) +%define SIZEOF_JCOEFPTR SIZEOF_POINTER ; sizeof(JCOEFPTR) +; +; -- jdct.h +; +; A forward DCT routine is given a pointer to a work area of type DCTELEM[]; +; the DCT is to be performed in-place in that buffer. +; To maximize parallelism, Type DCTELEM is changed to short (originally, int). +; +%define DCTELEM word ; short +%define SIZEOF_DCTELEM SIZEOF_WORD ; sizeof(DCTELEM) +%define float FP32 ; float +%define SIZEOF_FAST_FLOAT SIZEOF_FP32 ; sizeof(float) +; To maximize parallelism, Type short is changed to short. +; +%define ISLOW_MULT_TYPE word ; must be short +%define SIZEOF_ISLOW_MULT_TYPE SIZEOF_WORD ; sizeof(ISLOW_MULT_TYPE) +%define IFAST_MULT_TYPE word ; must be short +%define SIZEOF_IFAST_MULT_TYPE SIZEOF_WORD ; sizeof(IFAST_MULT_TYPE) +%define IFAST_SCALE_BITS 2 ; fractional bits in scale factors +%define FLOAT_MULT_TYPE FP32 ; must be float +%define SIZEOF_FLOAT_MULT_TYPE SIZEOF_FP32 ; sizeof(FLOAT_MULT_TYPE) +; +; -- jsimd.h +; +%define JSIMD_NONE 0x00 +%define JSIMD_MMX 0x01 +%define JSIMD_3DNOW 0x02 +%define JSIMD_SSE 0x04 +%define JSIMD_SSE2 0x08 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jsimdcpu.asm b/app/src/main/jni/libjpeg-turbo/simd/jsimdcpu.asm new file mode 100644 index 000000000..c42c4ad62 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jsimdcpu.asm @@ -0,0 +1,105 @@ +; +; jsimdcpu.asm - SIMD instruction support check +; +; Copyright 2009 Pierre Ossman for Cendio AB +; +; Based on +; x86 SIMD extension for IJG JPEG library +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; For conditions of distribution and use, see copyright notice in jsimdext.inc +; +; This file should be assembled with NASM (Netwide Assembler), +; can *not* be assembled with Microsoft's MASM or any compatible +; assembler (including Borland's Turbo Assembler). +; NASM is available from http://nasm.sourceforge.net/ or +; http://sourceforge.net/project/showfiles.php?group_id=6208 +; +; [TAB8] + +%include "jsimdext.inc" + +; -------------------------------------------------------------------------- + SECTION SEG_TEXT + BITS 32 +; +; Check if the CPU supports SIMD instructions +; +; GLOBAL(unsigned int) +; jpeg_simd_cpu_support (void) +; + + align 16 + global EXTN(jpeg_simd_cpu_support) + +EXTN(jpeg_simd_cpu_support): + push ebx +; push ecx ; need not be preserved +; push edx ; need not be preserved +; push esi ; unused + push edi + + xor edi,edi ; simd support flag + + pushfd + pop eax + mov edx,eax + xor eax, 1<<21 ; flip ID bit in EFLAGS + push eax + popfd + pushfd + pop eax + xor eax,edx + jz short .return ; CPUID is not supported + + ; Check for MMX instruction support + xor eax,eax + cpuid + test eax,eax + jz short .return + + xor eax,eax + inc eax + cpuid + mov eax,edx ; eax = Standard feature flags + + test eax, 1<<23 ; bit23:MMX + jz short .no_mmx + or edi, byte JSIMD_MMX +.no_mmx: + test eax, 1<<25 ; bit25:SSE + jz short .no_sse + or edi, byte JSIMD_SSE +.no_sse: + test eax, 1<<26 ; bit26:SSE2 + jz short .no_sse2 + or edi, byte JSIMD_SSE2 +.no_sse2: + + ; Check for 3DNow! instruction support + mov eax, 0x80000000 + cpuid + cmp eax, 0x80000000 + jbe short .return + + mov eax, 0x80000001 + cpuid + mov eax,edx ; eax = Extended feature flags + + test eax, 1<<31 ; bit31:3DNow!(vendor independent) + jz short .no_3dnow + or edi, byte JSIMD_3DNOW +.no_3dnow: + +.return: + mov eax,edi + + pop edi +; pop esi ; unused +; pop edx ; need not be preserved +; pop ecx ; need not be preserved + pop ebx + ret + +; For some reason, the OS X linker does not honor the request to align the +; segment unless we do this. + align 16 diff --git a/app/src/main/jni/libjpeg-turbo/simd/jsimdext.inc b/app/src/main/jni/libjpeg-turbo/simd/jsimdext.inc new file mode 100644 index 000000000..5983f5e31 --- /dev/null +++ b/app/src/main/jni/libjpeg-turbo/simd/jsimdext.inc @@ -0,0 +1,379 @@ +; +; jsimdext.inc - common declarations +; +; Copyright 2009 Pierre Ossman for Cendio AB +; Copyright 2010 D. R. Commander +; +; Based on +; x86 SIMD extension for IJG JPEG library - version 1.02 +; +; Copyright (C) 1999-2006, MIYASAKA Masaru. +; +; This software is provided 'as-is', without any express or implied +; warranty. In no event will the authors be held liable for any damages +; arising from the use of this software. +; +; Permission is granted to anyone to use this software for any purpose, +; including commercial applications, and to alter it and redistribute it +; freely, subject to the following restrictions: +; +; 1. The origin of this software must not be misrepresented; you must not +; claim that you wrote the original software. If you use this software +; in a product, an acknowledgment in the product documentation would be +; appreciated but is not required. +; 2. Altered source versions must be plainly marked as such, and must not be +; misrepresented as being the original software. +; 3. This notice may not be removed or altered from any source distribution. +; +; [TAB8] + +; ========================================================================== +; System-dependent configurations + +%ifdef WIN32 ; ----(nasm -fwin32 -DWIN32 ...)-------- +; * Microsoft Visual C++ +; * MinGW (Minimalist GNU for Windows) +; * CygWin +; * LCC-Win32 + +; -- segment definition -- +; +%ifdef __YASM_VER__ +%define SEG_TEXT .text align=16 +%define SEG_CONST .rdata align=16 +%else +%define SEG_TEXT .text align=16 public use32 class=CODE +%define SEG_CONST .rdata align=16 public use32 class=CONST +%endif + +%elifdef WIN64 ; ----(nasm -fwin64 -DWIN64 ...)-------- +; * Microsoft Visual C++ + +; -- segment definition -- +; +%ifdef __YASM_VER__ +%define SEG_TEXT .text align=16 +%define SEG_CONST .rdata align=16 +%else +%define SEG_TEXT .text align=16 public use64 class=CODE +%define SEG_CONST .rdata align=16 public use64 class=CONST +%endif +%define EXTN(name) name ; foo() -> foo + +%elifdef OBJ32 ; ----(nasm -fobj -DOBJ32 ...)---------- +; * Borland C++ (Win32) + +; -- segment definition -- +; +%define SEG_TEXT _text align=16 public use32 class=CODE +%define SEG_CONST _data align=16 public use32 class=DATA + +%elifdef ELF ; ----(nasm -felf[64] -DELF ...)------------ +; * Linux +; * *BSD family Unix using elf format +; * Unix System V, including Solaris x86, UnixWare and SCO Unix + +; mark stack as non-executable +section .note.GNU-stack noalloc noexec nowrite progbits + +; -- segment definition -- +; +%ifdef __x86_64__ +%define SEG_TEXT .text progbits align=16 +%define SEG_CONST .rodata progbits align=16 +%else +%define SEG_TEXT .text progbits alloc exec nowrite align=16 +%define SEG_CONST .rodata progbits alloc noexec nowrite align=16 +%endif + +; To make the code position-independent, append -DPIC to the commandline +; +%define GOT_SYMBOL _GLOBAL_OFFSET_TABLE_ ; ELF supports PIC +%define EXTN(name) name ; foo() -> foo + +%elifdef AOUT ; ----(nasm -faoutb/aout -DAOUT ...)---- +; * Older Linux using a.out format (nasm -f aout -DAOUT ...) +; * *BSD family Unix using a.out format (nasm -f aoutb -DAOUT ...) + +; -- segment definition -- +; +%define SEG_TEXT .text +%define SEG_CONST .data + +; To make the code position-independent, append -DPIC to the commandline +; +%define GOT_SYMBOL __GLOBAL_OFFSET_TABLE_ ; BSD-style a.out supports PIC + +%elifdef MACHO ; ----(nasm -fmacho -DMACHO ...)-------- +; * NeXTstep/OpenStep/Rhapsody/Darwin/MacOS X (Mach-O format) + +; -- segment definition -- +; +%define SEG_TEXT .text ;align=16 ; nasm doesn't accept align=16. why? +%define SEG_CONST .rodata align=16 + +; The generation of position-independent code (PIC) is the default on Darwin. +; +%define PIC +%define GOT_SYMBOL _MACHO_PIC_ ; Mach-O style code-relative addressing + +%else ; ----(Other case)---------------------- + +; -- segment definition -- +; +%define SEG_TEXT .text +%define SEG_CONST .data + +%endif ; ---------------------------------------------- + +; ========================================================================== + +; -------------------------------------------------------------------------- +; Common types +; +%ifdef __x86_64__ +%define POINTER qword ; general pointer type +%define SIZEOF_POINTER SIZEOF_QWORD ; sizeof(POINTER) +%define POINTER_BIT QWORD_BIT ; sizeof(POINTER)*BYTE_BIT +%else +%define POINTER dword ; general pointer type +%define SIZEOF_POINTER SIZEOF_DWORD ; sizeof(POINTER) +%define POINTER_BIT DWORD_BIT ; sizeof(POINTER)*BYTE_BIT +%endif + +%define INT dword ; signed integer type +%define SIZEOF_INT SIZEOF_DWORD ; sizeof(INT) +%define INT_BIT DWORD_BIT ; sizeof(INT)*BYTE_BIT + +%define FP32 dword ; IEEE754 single +%define SIZEOF_FP32 SIZEOF_DWORD ; sizeof(FP32) +%define FP32_BIT DWORD_BIT ; sizeof(FP32)*BYTE_BIT + +%define MMWORD qword ; int64 (MMX register) +%define SIZEOF_MMWORD SIZEOF_QWORD ; sizeof(MMWORD) +%define MMWORD_BIT QWORD_BIT ; sizeof(MMWORD)*BYTE_BIT + +; NASM is buggy and doesn't properly handle operand sizes for SSE +; instructions, so for now we have to define XMMWORD as blank. +%define XMMWORD ; int128 (SSE register) +%define SIZEOF_XMMWORD SIZEOF_OWORD ; sizeof(XMMWORD) +%define XMMWORD_BIT OWORD_BIT ; sizeof(XMMWORD)*BYTE_BIT + +; Similar hacks for when we load a dword or MMWORD into an xmm# register +%define XMM_DWORD +%define XMM_MMWORD + +%define SIZEOF_BYTE 1 ; sizeof(BYTE) +%define SIZEOF_WORD 2 ; sizeof(WORD) +%define SIZEOF_DWORD 4 ; sizeof(DWORD) +%define SIZEOF_QWORD 8 ; sizeof(QWORD) +%define SIZEOF_OWORD 16 ; sizeof(OWORD) + +%define BYTE_BIT 8 ; CHAR_BIT in C +%define WORD_BIT 16 ; sizeof(WORD)*BYTE_BIT +%define DWORD_BIT 32 ; sizeof(DWORD)*BYTE_BIT +%define QWORD_BIT 64 ; sizeof(QWORD)*BYTE_BIT +%define OWORD_BIT 128 ; sizeof(OWORD)*BYTE_BIT + +; -------------------------------------------------------------------------- +; External Symbol Name +; wocao +; %ifndef EXTN +; %define EXTN(name) _ %+ name ; foo() -> _foo +; %endif +%ifndef EXTN +%define EXTN(name) name +%endif + +; -------------------------------------------------------------------------- +; Macros for position-independent code (PIC) support +; +%ifndef GOT_SYMBOL +%undef PIC +%endif + +%ifdef PIC ; ------------------------------------------- + +%ifidn GOT_SYMBOL,_MACHO_PIC_ ; -------------------- + +; At present, nasm doesn't seem to support PIC generation for Mach-O. +; The PIC support code below is a little tricky. + + SECTION SEG_CONST +const_base: + +%define GOTOFF(got,sym) (got) + (sym) - const_base + +%imacro get_GOT 1 + ; NOTE: this macro destroys ecx resister. + call %%geteip + add ecx, byte (%%ref - $) + jmp short %%adjust +%%geteip: + mov ecx, POINTER [esp] + ret +%%adjust: + push ebp + xor ebp,ebp ; ebp = 0 +%ifidni %1,ebx ; (%1 == ebx) + ; db 0x8D,0x9C + jmp near const_base = + ; lea ebx, [ecx+ebp*8+(const_base-%%ref)] ; 8D,9C,E9,(offset32) + db 0x8D,0x9C ; 8D,9C + jmp near const_base ; E9,(const_base-%%ref) +%%ref: +%else ; (%1 != ebx) + ; db 0x8D,0x8C + jmp near const_base = + ; lea ecx, [ecx+ebp*8+(const_base-%%ref)] ; 8D,8C,E9,(offset32) + db 0x8D,0x8C ; 8D,8C + jmp near const_base ; E9,(const_base-%%ref) +%%ref: mov %1, ecx +%endif ; (%1 == ebx) + pop ebp +%endmacro + +%else ; GOT_SYMBOL != _MACHO_PIC_ ---------------- + +%define GOTOFF(got,sym) (got) + (sym) wrt ..gotoff + +%imacro get_GOT 1 + extern GOT_SYMBOL + call %%geteip + add %1, GOT_SYMBOL + $$ - $ wrt ..gotpc + jmp short %%done +%%geteip: + mov %1, POINTER [esp] + ret +%%done: +%endmacro + +%endif ; GOT_SYMBOL == _MACHO_PIC_ ---------------- + +%imacro pushpic 1.nolist + push %1 +%endmacro +%imacro poppic 1.nolist + pop %1 +%endmacro +%imacro movpic 2.nolist + mov %1,%2 +%endmacro + +%else ; !PIC ----------------------------------------- + +%define GOTOFF(got,sym) (sym) + +%imacro get_GOT 1.nolist +%endmacro +%imacro pushpic 1.nolist +%endmacro +%imacro poppic 1.nolist +%endmacro +%imacro movpic 2.nolist +%endmacro + +%endif ; PIC ----------------------------------------- + +; -------------------------------------------------------------------------- +; Align the next instruction on {2,4,8,16,..}-byte boundary. +; ".balign n,,m" in GNU as +; +%define MSKLE(x,y) (~(((y) & 0xFFFF) - ((x) & 0xFFFF)) >> 16) +%define FILLB(b,n) (($$-(b)) & ((n)-1)) + +%imacro alignx 1-2.nolist 0xFFFF +%%bs: times MSKLE(FILLB(%%bs,%1),%2) & MSKLE(16,FILLB($,%1)) & FILLB($,%1) \ + db 0x90 ; nop + times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/9 \ + db 0x8D,0x9C,0x23,0x00,0x00,0x00,0x00 ; lea ebx,[ebx+0x00000000] + times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/7 \ + db 0x8D,0xAC,0x25,0x00,0x00,0x00,0x00 ; lea ebp,[ebp+0x00000000] + times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/6 \ + db 0x8D,0xAD,0x00,0x00,0x00,0x00 ; lea ebp,[ebp+0x00000000] + times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/4 \ + db 0x8D,0x6C,0x25,0x00 ; lea ebp,[ebp+0x00] + times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/3 \ + db 0x8D,0x6D,0x00 ; lea ebp,[ebp+0x00] + times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/2 \ + db 0x8B,0xED ; mov ebp,ebp + times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/1 \ + db 0x90 ; nop +%endmacro + +; Align the next data on {2,4,8,16,..}-byte boundary. +; +%imacro alignz 1.nolist + align %1, db 0 ; filling zeros +%endmacro + +%ifdef __x86_64__ + +%ifdef WIN64 + +%imacro collect_args 0 + push r12 + push r13 + push r14 + push r15 + mov r10, rcx + mov r11, rdx + mov r12, r8 + mov r13, r9 + mov r14, [rax+48] + mov r15, [rax+56] + push rsi + push rdi + sub rsp, SIZEOF_XMMWORD + movaps XMMWORD [rsp], xmm6 + sub rsp, SIZEOF_XMMWORD + movaps XMMWORD [rsp], xmm7 +%endmacro + +%imacro uncollect_args 0 + movaps xmm7, XMMWORD [rsp] + add rsp, SIZEOF_XMMWORD + movaps xmm6, XMMWORD [rsp] + add rsp, SIZEOF_XMMWORD + pop rdi + pop rsi + pop r15 + pop r14 + pop r13 + pop r12 +%endmacro + +%else + +%imacro collect_args 0 + push r10 + push r11 + push r12 + push r13 + push r14 + push r15 + mov r10, rdi + mov r11, rsi + mov r12, rdx + mov r13, rcx + mov r14, r8 + mov r15, r9 +%endmacro + +%imacro uncollect_args 0 + pop r15 + pop r14 + pop r13 + pop r12 + pop r11 + pop r10 +%endmacro + +%endif + +%endif + +; -------------------------------------------------------------------------- +; Defines picked up from the C headers +; +%include "jsimdcfg.inc" + +; -------------------------------------------------------------------------- diff --git a/app/src/main/jni/libpng/arm/arm_init.c b/app/src/main/jni/libpng/arm/arm_init.c new file mode 100644 index 000000000..712607a0b --- /dev/null +++ b/app/src/main/jni/libpng/arm/arm_init.c @@ -0,0 +1,134 @@ + +/* arm_init.c - NEON optimised filter functions + * + * Copyright (c) 2014 Glenn Randers-Pehrson + * Written by Mans Rullgard, 2011. + * Last changed in libpng 1.6.16 [December 22, 2014] + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + */ +/* Below, after checking __linux__, various non-C90 POSIX 1003.1 functions are + * called. + */ +#define _POSIX_SOURCE 1 + +#include "../pngpriv.h" + +#ifdef PNG_READ_SUPPORTED + +#if PNG_ARM_NEON_OPT > 0 +#ifdef PNG_ARM_NEON_CHECK_SUPPORTED /* Do run-time checks */ +/* WARNING: it is strongly recommended that you do not build libpng with + * run-time checks for CPU features if at all possible. In the case of the ARM + * NEON instructions there is no processor-specific way of detecting the + * presence of the required support, therefore run-time detection is extremely + * OS specific. + * + * You may set the macro PNG_ARM_NEON_FILE to the file name of file containing + * a fragment of C source code which defines the png_have_neon function. There + * are a number of implementations in contrib/arm-neon, but the only one that + * has partial support is contrib/arm-neon/linux.c - a generic Linux + * implementation which reads /proc/cpufino. + */ +#ifndef PNG_ARM_NEON_FILE +# ifdef __linux__ +# define PNG_ARM_NEON_FILE "../contrib/arm-neon/linux.c" +# endif +#endif + +#ifdef PNG_ARM_NEON_FILE + +#include /* for sig_atomic_t */ +static int png_have_neon(png_structp png_ptr); +#include PNG_ARM_NEON_FILE + +#else /* PNG_ARM_NEON_FILE */ +# error "PNG_ARM_NEON_FILE undefined: no support for run-time ARM NEON checks" +#endif /* PNG_ARM_NEON_FILE */ +#endif /* PNG_ARM_NEON_CHECK_SUPPORTED */ + +#ifndef PNG_ALIGNED_MEMORY_SUPPORTED +# error "ALIGNED_MEMORY is required; set: -DPNG_ALIGNED_MEMORY_SUPPORTED" +#endif + +void +png_init_filter_functions_neon(png_structp pp, unsigned int bpp) +{ + /* The switch statement is compiled in for ARM_NEON_API, the call to + * png_have_neon is compiled in for ARM_NEON_CHECK. If both are defined + * the check is only performed if the API has not set the NEON option on + * or off explicitly. In this case the check controls what happens. + * + * If the CHECK is not compiled in and the option is UNSET the behavior prior + * to 1.6.7 was to use the NEON code - this was a bug caused by having the + * wrong order of the 'ON' and 'default' cases. UNSET now defaults to OFF, + * as documented in png.h + */ +#ifdef PNG_ARM_NEON_API_SUPPORTED + switch ((pp->options >> PNG_ARM_NEON) & 3) + { + case PNG_OPTION_UNSET: + /* Allow the run-time check to execute if it has been enabled - + * thus both API and CHECK can be turned on. If it isn't supported + * this case will fall through to the 'default' below, which just + * returns. + */ +#endif /* PNG_ARM_NEON_API_SUPPORTED */ +#ifdef PNG_ARM_NEON_CHECK_SUPPORTED + { + static volatile sig_atomic_t no_neon = -1; /* not checked */ + + if (no_neon < 0) + no_neon = !png_have_neon(pp); + + if (no_neon) + return; + } +#ifdef PNG_ARM_NEON_API_SUPPORTED + break; +#endif +#endif /* PNG_ARM_NEON_CHECK_SUPPORTED */ + +#ifdef PNG_ARM_NEON_API_SUPPORTED + default: /* OFF or INVALID */ + return; + + case PNG_OPTION_ON: + /* Option turned on */ + break; + } +#endif + + /* IMPORTANT: any new external functions used here must be declared using + * PNG_INTERNAL_FUNCTION in ../pngpriv.h. This is required so that the + * 'prefix' option to configure works: + * + * ./configure --with-libpng-prefix=foobar_ + * + * Verify you have got this right by running the above command, doing a build + * and examining pngprefix.h; it must contain a #define for every external + * function you add. (Notice that this happens automatically for the + * initialization function.) + */ + pp->read_filter[PNG_FILTER_VALUE_UP-1] = png_read_filter_row_up_neon; + + if (bpp == 3) + { + pp->read_filter[PNG_FILTER_VALUE_SUB-1] = png_read_filter_row_sub3_neon; + pp->read_filter[PNG_FILTER_VALUE_AVG-1] = png_read_filter_row_avg3_neon; + pp->read_filter[PNG_FILTER_VALUE_PAETH-1] = + png_read_filter_row_paeth3_neon; + } + + else if (bpp == 4) + { + pp->read_filter[PNG_FILTER_VALUE_SUB-1] = png_read_filter_row_sub4_neon; + pp->read_filter[PNG_FILTER_VALUE_AVG-1] = png_read_filter_row_avg4_neon; + pp->read_filter[PNG_FILTER_VALUE_PAETH-1] = + png_read_filter_row_paeth4_neon; + } +} +#endif /* PNG_ARM_NEON_OPT > 0 */ +#endif /* READ */ diff --git a/app/src/main/jni/libpng/arm/filter_neon.S b/app/src/main/jni/libpng/arm/filter_neon.S new file mode 100644 index 000000000..3b061d6bb --- /dev/null +++ b/app/src/main/jni/libpng/arm/filter_neon.S @@ -0,0 +1,253 @@ + +/* filter_neon.S - NEON optimised filter functions + * + * Copyright (c) 2014 Glenn Randers-Pehrson + * Written by Mans Rullgard, 2011. + * Last changed in libpng 1.6.16 [December 22, 2014] + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + */ + +/* This is required to get the symbol renames, which are #defines, and the + * definitions (or not) of PNG_ARM_NEON_OPT and PNG_ARM_NEON_IMPLEMENTATION. + */ +#define PNG_VERSION_INFO_ONLY +#include "../pngpriv.h" + +#if defined(__linux__) && defined(__ELF__) +.section .note.GNU-stack,"",%progbits /* mark stack as non-executable */ +#endif + +#ifdef PNG_READ_SUPPORTED + +/* Assembler NEON support - only works for 32-bit ARM (i.e. it does not work for + * ARM64). The code in arm/filter_neon_intrinsics.c supports ARM64, however it + * only works if -mfpu=neon is specified on the GCC command line. See pngpriv.h + * for the logic which sets PNG_USE_ARM_NEON_ASM: + */ +#if PNG_ARM_NEON_IMPLEMENTATION == 2 /* hand-coded assembler */ + +#if PNG_ARM_NEON_OPT > 0 + +#ifdef __ELF__ +# define ELF +#else +# define ELF @ +#endif + + .arch armv7-a + .fpu neon + +.macro func name, export=0 + .macro endfunc +ELF .size \name, . - \name + .endfunc + .purgem endfunc + .endm + .text + + /* Explicitly specifying alignment here because some versions of + * GAS don't align code correctly. This is harmless in correctly + * written versions of GAS. + */ + .align 2 + + .if \export + .global \name + .endif +ELF .type \name, STT_FUNC + .func \name +\name: +.endm + +func png_read_filter_row_sub4_neon, export=1 + ldr r3, [r0, #4] @ rowbytes + vmov.i8 d3, #0 +1: + vld4.32 {d4[],d5[],d6[],d7[]}, [r1,:128] + vadd.u8 d0, d3, d4 + vadd.u8 d1, d0, d5 + vadd.u8 d2, d1, d6 + vadd.u8 d3, d2, d7 + vst4.32 {d0[0],d1[0],d2[0],d3[0]},[r1,:128]! + subs r3, r3, #16 + bgt 1b + + bx lr +endfunc + +func png_read_filter_row_sub3_neon, export=1 + ldr r3, [r0, #4] @ rowbytes + vmov.i8 d3, #0 + mov r0, r1 + mov r2, #3 + mov r12, #12 + vld1.8 {q11}, [r0], r12 +1: + vext.8 d5, d22, d23, #3 + vadd.u8 d0, d3, d22 + vext.8 d6, d22, d23, #6 + vadd.u8 d1, d0, d5 + vext.8 d7, d23, d23, #1 + vld1.8 {q11}, [r0], r12 + vst1.32 {d0[0]}, [r1,:32], r2 + vadd.u8 d2, d1, d6 + vst1.32 {d1[0]}, [r1], r2 + vadd.u8 d3, d2, d7 + vst1.32 {d2[0]}, [r1], r2 + vst1.32 {d3[0]}, [r1], r2 + subs r3, r3, #12 + bgt 1b + + bx lr +endfunc + +func png_read_filter_row_up_neon, export=1 + ldr r3, [r0, #4] @ rowbytes +1: + vld1.8 {q0}, [r1,:128] + vld1.8 {q1}, [r2,:128]! + vadd.u8 q0, q0, q1 + vst1.8 {q0}, [r1,:128]! + subs r3, r3, #16 + bgt 1b + + bx lr +endfunc + +func png_read_filter_row_avg4_neon, export=1 + ldr r12, [r0, #4] @ rowbytes + vmov.i8 d3, #0 +1: + vld4.32 {d4[],d5[],d6[],d7[]}, [r1,:128] + vld4.32 {d16[],d17[],d18[],d19[]},[r2,:128]! + vhadd.u8 d0, d3, d16 + vadd.u8 d0, d0, d4 + vhadd.u8 d1, d0, d17 + vadd.u8 d1, d1, d5 + vhadd.u8 d2, d1, d18 + vadd.u8 d2, d2, d6 + vhadd.u8 d3, d2, d19 + vadd.u8 d3, d3, d7 + vst4.32 {d0[0],d1[0],d2[0],d3[0]},[r1,:128]! + subs r12, r12, #16 + bgt 1b + + bx lr +endfunc + +func png_read_filter_row_avg3_neon, export=1 + push {r4,lr} + ldr r12, [r0, #4] @ rowbytes + vmov.i8 d3, #0 + mov r0, r1 + mov r4, #3 + mov lr, #12 + vld1.8 {q11}, [r0], lr +1: + vld1.8 {q10}, [r2], lr + vext.8 d5, d22, d23, #3 + vhadd.u8 d0, d3, d20 + vext.8 d17, d20, d21, #3 + vadd.u8 d0, d0, d22 + vext.8 d6, d22, d23, #6 + vhadd.u8 d1, d0, d17 + vext.8 d18, d20, d21, #6 + vadd.u8 d1, d1, d5 + vext.8 d7, d23, d23, #1 + vld1.8 {q11}, [r0], lr + vst1.32 {d0[0]}, [r1,:32], r4 + vhadd.u8 d2, d1, d18 + vst1.32 {d1[0]}, [r1], r4 + vext.8 d19, d21, d21, #1 + vadd.u8 d2, d2, d6 + vhadd.u8 d3, d2, d19 + vst1.32 {d2[0]}, [r1], r4 + vadd.u8 d3, d3, d7 + vst1.32 {d3[0]}, [r1], r4 + subs r12, r12, #12 + bgt 1b + + pop {r4,pc} +endfunc + +.macro paeth rx, ra, rb, rc + vaddl.u8 q12, \ra, \rb @ a + b + vaddl.u8 q15, \rc, \rc @ 2*c + vabdl.u8 q13, \rb, \rc @ pa + vabdl.u8 q14, \ra, \rc @ pb + vabd.u16 q15, q12, q15 @ pc + vcle.u16 q12, q13, q14 @ pa <= pb + vcle.u16 q13, q13, q15 @ pa <= pc + vcle.u16 q14, q14, q15 @ pb <= pc + vand q12, q12, q13 @ pa <= pb && pa <= pc + vmovn.u16 d28, q14 + vmovn.u16 \rx, q12 + vbsl d28, \rb, \rc + vbsl \rx, \ra, d28 +.endm + +func png_read_filter_row_paeth4_neon, export=1 + ldr r12, [r0, #4] @ rowbytes + vmov.i8 d3, #0 + vmov.i8 d20, #0 +1: + vld4.32 {d4[],d5[],d6[],d7[]}, [r1,:128] + vld4.32 {d16[],d17[],d18[],d19[]},[r2,:128]! + paeth d0, d3, d16, d20 + vadd.u8 d0, d0, d4 + paeth d1, d0, d17, d16 + vadd.u8 d1, d1, d5 + paeth d2, d1, d18, d17 + vadd.u8 d2, d2, d6 + paeth d3, d2, d19, d18 + vmov d20, d19 + vadd.u8 d3, d3, d7 + vst4.32 {d0[0],d1[0],d2[0],d3[0]},[r1,:128]! + subs r12, r12, #16 + bgt 1b + + bx lr +endfunc + +func png_read_filter_row_paeth3_neon, export=1 + push {r4,lr} + ldr r12, [r0, #4] @ rowbytes + vmov.i8 d3, #0 + vmov.i8 d4, #0 + mov r0, r1 + mov r4, #3 + mov lr, #12 + vld1.8 {q11}, [r0], lr +1: + vld1.8 {q10}, [r2], lr + paeth d0, d3, d20, d4 + vext.8 d5, d22, d23, #3 + vadd.u8 d0, d0, d22 + vext.8 d17, d20, d21, #3 + paeth d1, d0, d17, d20 + vst1.32 {d0[0]}, [r1,:32], r4 + vext.8 d6, d22, d23, #6 + vadd.u8 d1, d1, d5 + vext.8 d18, d20, d21, #6 + paeth d2, d1, d18, d17 + vext.8 d7, d23, d23, #1 + vld1.8 {q11}, [r0], lr + vst1.32 {d1[0]}, [r1], r4 + vadd.u8 d2, d2, d6 + vext.8 d19, d21, d21, #1 + paeth d3, d2, d19, d18 + vst1.32 {d2[0]}, [r1], r4 + vmov d4, d19 + vadd.u8 d3, d3, d7 + vst1.32 {d3[0]}, [r1], r4 + subs r12, r12, #12 + bgt 1b + + pop {r4,pc} +endfunc +#endif /* PNG_ARM_NEON_OPT > 0 */ +#endif /* PNG_ARM_NEON_IMPLEMENTATION == 2 (assembler) */ +#endif /* READ */ diff --git a/app/src/main/jni/libpng/arm/filter_neon_intrinsics.c b/app/src/main/jni/libpng/arm/filter_neon_intrinsics.c new file mode 100644 index 000000000..d42c78890 --- /dev/null +++ b/app/src/main/jni/libpng/arm/filter_neon_intrinsics.c @@ -0,0 +1,373 @@ + +/* filter_neon_intrinsics.c - NEON optimised filter functions + * + * Copyright (c) 2014 Glenn Randers-Pehrson + * Written by James Yu , October 2013. + * Based on filter_neon.S, written by Mans Rullgard, 2011. + * + * Last changed in libpng 1.6.16 [December 22, 2014] + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + */ + +#include "../pngpriv.h" + +#ifdef PNG_READ_SUPPORTED + +/* This code requires -mfpu=neon on the command line: */ +#if PNG_ARM_NEON_IMPLEMENTATION == 1 /* intrinsics code from pngpriv.h */ + +#include + +/* libpng row pointers are not necessarily aligned to any particular boundary, + * however this code will only work with appropriate alignment. arm/arm_init.c + * checks for this (and will not compile unless it is done). This code uses + * variants of png_aligncast to avoid compiler warnings. + */ +#define png_ptr(type,pointer) png_aligncast(type *,pointer) +#define png_ptrc(type,pointer) png_aligncastconst(const type *,pointer) + +/* The following relies on a variable 'temp_pointer' being declared with type + * 'type'. This is written this way just to hide the GCC strict aliasing + * warning; note that the code is safe because there never is an alias between + * the input and output pointers. + */ +#define png_ldr(type,pointer)\ + (temp_pointer = png_ptr(type,pointer), *temp_pointer) + +#if PNG_ARM_NEON_OPT > 0 + +void +png_read_filter_row_up_neon(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + png_bytep rp = row; + png_bytep rp_stop = row + row_info->rowbytes; + png_const_bytep pp = prev_row; + + for (; rp < rp_stop; rp += 16, pp += 16) + { + uint8x16_t qrp, qpp; + + qrp = vld1q_u8(rp); + qpp = vld1q_u8(pp); + qrp = vaddq_u8(qrp, qpp); + vst1q_u8(rp, qrp); + } +} + +void +png_read_filter_row_sub3_neon(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + png_bytep rp = row; + png_bytep rp_stop = row + row_info->rowbytes; + + uint8x16_t vtmp = vld1q_u8(rp); + uint8x8x2_t *vrpt = png_ptr(uint8x8x2_t, &vtmp); + uint8x8x2_t vrp = *vrpt; + + uint8x8x4_t vdest; + vdest.val[3] = vdup_n_u8(0); + + for (; rp < rp_stop;) + { + uint8x8_t vtmp1, vtmp2; + uint32x2_t *temp_pointer; + + vtmp1 = vext_u8(vrp.val[0], vrp.val[1], 3); + vdest.val[0] = vadd_u8(vdest.val[3], vrp.val[0]); + vtmp2 = vext_u8(vrp.val[0], vrp.val[1], 6); + vdest.val[1] = vadd_u8(vdest.val[0], vtmp1); + + vtmp1 = vext_u8(vrp.val[1], vrp.val[1], 1); + vdest.val[2] = vadd_u8(vdest.val[1], vtmp2); + vdest.val[3] = vadd_u8(vdest.val[2], vtmp1); + + vtmp = vld1q_u8(rp + 12); + vrpt = png_ptr(uint8x8x2_t, &vtmp); + vrp = *vrpt; + + vst1_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2_t,&vdest.val[0]), 0); + rp += 3; + vst1_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2_t,&vdest.val[1]), 0); + rp += 3; + vst1_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2_t,&vdest.val[2]), 0); + rp += 3; + vst1_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2_t,&vdest.val[3]), 0); + rp += 3; + } + + PNG_UNUSED(prev_row) +} + +void +png_read_filter_row_sub4_neon(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + png_bytep rp = row; + png_bytep rp_stop = row + row_info->rowbytes; + + uint8x8x4_t vdest; + vdest.val[3] = vdup_n_u8(0); + + for (; rp < rp_stop; rp += 16) + { + uint32x2x4_t vtmp = vld4_u32(png_ptr(uint32_t,rp)); + uint8x8x4_t *vrpt = png_ptr(uint8x8x4_t,&vtmp); + uint8x8x4_t vrp = *vrpt; + uint32x2x4_t *temp_pointer; + + vdest.val[0] = vadd_u8(vdest.val[3], vrp.val[0]); + vdest.val[1] = vadd_u8(vdest.val[0], vrp.val[1]); + vdest.val[2] = vadd_u8(vdest.val[1], vrp.val[2]); + vdest.val[3] = vadd_u8(vdest.val[2], vrp.val[3]); + vst4_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2x4_t,&vdest), 0); + } + + PNG_UNUSED(prev_row) +} + +void +png_read_filter_row_avg3_neon(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + png_bytep rp = row; + png_const_bytep pp = prev_row; + png_bytep rp_stop = row + row_info->rowbytes; + + uint8x16_t vtmp; + uint8x8x2_t *vrpt; + uint8x8x2_t vrp; + uint8x8x4_t vdest; + vdest.val[3] = vdup_n_u8(0); + + vtmp = vld1q_u8(rp); + vrpt = png_ptr(uint8x8x2_t,&vtmp); + vrp = *vrpt; + + for (; rp < rp_stop; pp += 12) + { + uint8x8_t vtmp1, vtmp2, vtmp3; + + uint8x8x2_t *vppt; + uint8x8x2_t vpp; + + uint32x2_t *temp_pointer; + + vtmp = vld1q_u8(pp); + vppt = png_ptr(uint8x8x2_t,&vtmp); + vpp = *vppt; + + vtmp1 = vext_u8(vrp.val[0], vrp.val[1], 3); + vdest.val[0] = vhadd_u8(vdest.val[3], vpp.val[0]); + vdest.val[0] = vadd_u8(vdest.val[0], vrp.val[0]); + + vtmp2 = vext_u8(vpp.val[0], vpp.val[1], 3); + vtmp3 = vext_u8(vrp.val[0], vrp.val[1], 6); + vdest.val[1] = vhadd_u8(vdest.val[0], vtmp2); + vdest.val[1] = vadd_u8(vdest.val[1], vtmp1); + + vtmp2 = vext_u8(vpp.val[0], vpp.val[1], 6); + vtmp1 = vext_u8(vrp.val[1], vrp.val[1], 1); + + vtmp = vld1q_u8(rp + 12); + vrpt = png_ptr(uint8x8x2_t,&vtmp); + vrp = *vrpt; + + vdest.val[2] = vhadd_u8(vdest.val[1], vtmp2); + vdest.val[2] = vadd_u8(vdest.val[2], vtmp3); + + vtmp2 = vext_u8(vpp.val[1], vpp.val[1], 1); + + vdest.val[3] = vhadd_u8(vdest.val[2], vtmp2); + vdest.val[3] = vadd_u8(vdest.val[3], vtmp1); + + vst1_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2_t,&vdest.val[0]), 0); + rp += 3; + vst1_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2_t,&vdest.val[1]), 0); + rp += 3; + vst1_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2_t,&vdest.val[2]), 0); + rp += 3; + vst1_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2_t,&vdest.val[3]), 0); + rp += 3; + } +} + +void +png_read_filter_row_avg4_neon(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + png_bytep rp = row; + png_bytep rp_stop = row + row_info->rowbytes; + png_const_bytep pp = prev_row; + + uint8x8x4_t vdest; + vdest.val[3] = vdup_n_u8(0); + + for (; rp < rp_stop; rp += 16, pp += 16) + { + uint32x2x4_t vtmp; + uint8x8x4_t *vrpt, *vppt; + uint8x8x4_t vrp, vpp; + uint32x2x4_t *temp_pointer; + + vtmp = vld4_u32(png_ptr(uint32_t,rp)); + vrpt = png_ptr(uint8x8x4_t,&vtmp); + vrp = *vrpt; + vtmp = vld4_u32(png_ptrc(uint32_t,pp)); + vppt = png_ptr(uint8x8x4_t,&vtmp); + vpp = *vppt; + + vdest.val[0] = vhadd_u8(vdest.val[3], vpp.val[0]); + vdest.val[0] = vadd_u8(vdest.val[0], vrp.val[0]); + vdest.val[1] = vhadd_u8(vdest.val[0], vpp.val[1]); + vdest.val[1] = vadd_u8(vdest.val[1], vrp.val[1]); + vdest.val[2] = vhadd_u8(vdest.val[1], vpp.val[2]); + vdest.val[2] = vadd_u8(vdest.val[2], vrp.val[2]); + vdest.val[3] = vhadd_u8(vdest.val[2], vpp.val[3]); + vdest.val[3] = vadd_u8(vdest.val[3], vrp.val[3]); + + vst4_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2x4_t,&vdest), 0); + } +} + +static uint8x8_t +paeth(uint8x8_t a, uint8x8_t b, uint8x8_t c) +{ + uint8x8_t d, e; + uint16x8_t p1, pa, pb, pc; + + p1 = vaddl_u8(a, b); /* a + b */ + pc = vaddl_u8(c, c); /* c * 2 */ + pa = vabdl_u8(b, c); /* pa */ + pb = vabdl_u8(a, c); /* pb */ + pc = vabdq_u16(p1, pc); /* pc */ + + p1 = vcleq_u16(pa, pb); /* pa <= pb */ + pa = vcleq_u16(pa, pc); /* pa <= pc */ + pb = vcleq_u16(pb, pc); /* pb <= pc */ + + p1 = vandq_u16(p1, pa); /* pa <= pb && pa <= pc */ + + d = vmovn_u16(pb); + e = vmovn_u16(p1); + + d = vbsl_u8(d, b, c); + e = vbsl_u8(e, a, d); + + return e; +} + +void +png_read_filter_row_paeth3_neon(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + png_bytep rp = row; + png_const_bytep pp = prev_row; + png_bytep rp_stop = row + row_info->rowbytes; + + uint8x16_t vtmp; + uint8x8x2_t *vrpt; + uint8x8x2_t vrp; + uint8x8_t vlast = vdup_n_u8(0); + uint8x8x4_t vdest; + vdest.val[3] = vdup_n_u8(0); + + vtmp = vld1q_u8(rp); + vrpt = png_ptr(uint8x8x2_t,&vtmp); + vrp = *vrpt; + + for (; rp < rp_stop; pp += 12) + { + uint8x8x2_t *vppt; + uint8x8x2_t vpp; + uint8x8_t vtmp1, vtmp2, vtmp3; + uint32x2_t *temp_pointer; + + vtmp = vld1q_u8(pp); + vppt = png_ptr(uint8x8x2_t,&vtmp); + vpp = *vppt; + + vdest.val[0] = paeth(vdest.val[3], vpp.val[0], vlast); + vdest.val[0] = vadd_u8(vdest.val[0], vrp.val[0]); + + vtmp1 = vext_u8(vrp.val[0], vrp.val[1], 3); + vtmp2 = vext_u8(vpp.val[0], vpp.val[1], 3); + vdest.val[1] = paeth(vdest.val[0], vtmp2, vpp.val[0]); + vdest.val[1] = vadd_u8(vdest.val[1], vtmp1); + + vtmp1 = vext_u8(vrp.val[0], vrp.val[1], 6); + vtmp3 = vext_u8(vpp.val[0], vpp.val[1], 6); + vdest.val[2] = paeth(vdest.val[1], vtmp3, vtmp2); + vdest.val[2] = vadd_u8(vdest.val[2], vtmp1); + + vtmp1 = vext_u8(vrp.val[1], vrp.val[1], 1); + vtmp2 = vext_u8(vpp.val[1], vpp.val[1], 1); + + vtmp = vld1q_u8(rp + 12); + vrpt = png_ptr(uint8x8x2_t,&vtmp); + vrp = *vrpt; + + vdest.val[3] = paeth(vdest.val[2], vtmp2, vtmp3); + vdest.val[3] = vadd_u8(vdest.val[3], vtmp1); + + vlast = vtmp2; + + vst1_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2_t,&vdest.val[0]), 0); + rp += 3; + vst1_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2_t,&vdest.val[1]), 0); + rp += 3; + vst1_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2_t,&vdest.val[2]), 0); + rp += 3; + vst1_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2_t,&vdest.val[3]), 0); + rp += 3; + } +} + +void +png_read_filter_row_paeth4_neon(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + png_bytep rp = row; + png_bytep rp_stop = row + row_info->rowbytes; + png_const_bytep pp = prev_row; + + uint8x8_t vlast = vdup_n_u8(0); + uint8x8x4_t vdest; + vdest.val[3] = vdup_n_u8(0); + + for (; rp < rp_stop; rp += 16, pp += 16) + { + uint32x2x4_t vtmp; + uint8x8x4_t *vrpt, *vppt; + uint8x8x4_t vrp, vpp; + uint32x2x4_t *temp_pointer; + + vtmp = vld4_u32(png_ptr(uint32_t,rp)); + vrpt = png_ptr(uint8x8x4_t,&vtmp); + vrp = *vrpt; + vtmp = vld4_u32(png_ptrc(uint32_t,pp)); + vppt = png_ptr(uint8x8x4_t,&vtmp); + vpp = *vppt; + + vdest.val[0] = paeth(vdest.val[3], vpp.val[0], vlast); + vdest.val[0] = vadd_u8(vdest.val[0], vrp.val[0]); + vdest.val[1] = paeth(vdest.val[0], vpp.val[1], vpp.val[0]); + vdest.val[1] = vadd_u8(vdest.val[1], vrp.val[1]); + vdest.val[2] = paeth(vdest.val[1], vpp.val[2], vpp.val[1]); + vdest.val[2] = vadd_u8(vdest.val[2], vrp.val[2]); + vdest.val[3] = paeth(vdest.val[2], vpp.val[3], vpp.val[2]); + vdest.val[3] = vadd_u8(vdest.val[3], vrp.val[3]); + + vlast = vpp.val[3]; + + vst4_lane_u32(png_ptr(uint32_t,rp), png_ldr(uint32x2x4_t,&vdest), 0); + } +} + +#endif /* PNG_ARM_NEON_OPT > 0 */ +#endif /* PNG_ARM_NEON_IMPLEMENTATION == 1 (intrinsics) */ +#endif /* READ */ diff --git a/app/src/main/jni/libpng/config.h b/app/src/main/jni/libpng/config.h new file mode 100644 index 000000000..86d9933b6 --- /dev/null +++ b/app/src/main/jni/libpng/config.h @@ -0,0 +1,110 @@ +/* config.h. Generated from config.h.in by configure. */ +/* config.h.in. Generated from configure.ac by autoheader. */ + +/* Define to 1 if you have the header file. */ +#define HAVE_DLFCN_H 1 + +/* Define to 1 if you have the `feenableexcept' function. */ +#define HAVE_FEENABLEEXCEPT 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_INTTYPES_H 1 + +/* Define to 1 if you have the `m' library (-lm). */ +#define HAVE_LIBM 1 + +/* Define to 1 if you have the `z' library (-lz). */ +#define HAVE_LIBZ 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_MEMORY_H 1 + +/* Define to 1 if you have the `memset' function. */ +#define HAVE_MEMSET 1 + +/* Define to 1 if you have the `pow' function. */ +#define HAVE_POW 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STDINT_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STDLIB_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STRINGS_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_STRING_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_SYS_STAT_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_SYS_TYPES_H 1 + +/* Define to 1 if you have the header file. */ +#define HAVE_UNISTD_H 1 + +/* Define to the sub-directory in which libtool stores uninstalled libraries. + */ +#define LT_OBJDIR ".libs/" + +/* Name of package */ +#define PACKAGE "libpng" + +/* Define to the address where bug reports for this package should be sent. */ +#define PACKAGE_BUGREPORT "png-mng-implement@lists.sourceforge.net" + +/* Define to the full name of this package. */ +#define PACKAGE_NAME "libpng" + +/* Define to the full name and version of this package. */ +#define PACKAGE_STRING "libpng 1.6.16" + +/* Define to the one symbol short name of this package. */ +#define PACKAGE_TARNAME "libpng" + +/* Define to the home page for this package. */ +#define PACKAGE_URL "" + +/* Define to the version of this package. */ +#define PACKAGE_VERSION "1.6.16" + +/* Turn on ARM Neon optimizations at run-time */ +/* #undef PNG_ARM_NEON_API_SUPPORTED */ + +/* Check for ARM Neon support at run-time */ +/* #undef PNG_ARM_NEON_CHECK_SUPPORTED */ + +/* Enable ARM Neon optimizations */ +/* #undef PNG_ARM_NEON_OPT */ + +/* Define to 1 if you have the ANSI C header files. */ +#define STDC_HEADERS 1 + +/* Define to 1 if your declares `struct tm'. */ +/* #undef TM_IN_SYS_TIME */ + +/* Version number of package */ +#define VERSION "1.6.16" + +/* Define to empty if `const' does not conform to ANSI C. */ +/* #undef const */ + +/* Define to the equivalent of the C99 'restrict' keyword, or to + nothing if this is not supported. Do not define if restrict is + supported directly. */ +#define restrict __restrict +/* Work around a bug in Sun C++: it does not support _Restrict or + __restrict__, even though the corresponding Sun C compiler ends up with + "#define restrict _Restrict" or "#define restrict __restrict__" in the + previous line. Perhaps some future version of Sun C++ will work with + restrict; if so, hopefully it defines __RESTRICT like Sun C does. */ +#if defined __SUNPRO_CC && !defined __RESTRICT +# define _Restrict +# define __restrict__ +#endif + +/* Define to `unsigned int' if does not define. */ +/* #undef size_t */ diff --git a/app/src/main/jni/libpng/contrib/arm-neon/linux.c b/app/src/main/jni/libpng/contrib/arm-neon/linux.c new file mode 100644 index 000000000..abae9e33b --- /dev/null +++ b/app/src/main/jni/libpng/contrib/arm-neon/linux.c @@ -0,0 +1,161 @@ +/* contrib/arm-neon/linux.c + * + * Copyright (c) 2014 Glenn Randers-Pehrson + * Written by John Bowler, 2014. + * Last changed in libpng 1.6.16 [December 22, 2014] + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + * + * SEE contrib/arm-neon/README before reporting bugs + * + * STATUS: SUPPORTED + * BUG REPORTS: png-mng-implement@sourceforge.net + * + * png_have_neon implemented for Linux by reading the widely available + * pseudo-file /proc/cpuinfo. + * + * This code is strict ANSI-C and is probably moderately portable; it does + * however use and it assumes that /proc/cpuinfo is never localized. + */ +#include + +static int +png_have_neon(png_structp png_ptr) +{ + FILE *f = fopen("/proc/cpuinfo", "rb"); + + if (f != NULL) + { + /* This is a simple state machine which reads the input byte-by-byte until + * it gets a match on the 'neon' feature or reaches the end of the stream. + */ + static const char ch_feature[] = { 70, 69, 65, 84, 85, 82, 69, 83 }; + static const char ch_neon[] = { 78, 69, 79, 78 }; + + enum + { + StartLine, Feature, Colon, StartTag, Neon, HaveNeon, SkipTag, SkipLine + } state; + int counter; + + for (state=StartLine, counter=0;;) + { + int ch = fgetc(f); + + if (ch == EOF) + { + /* EOF means error or end-of-file, return false; neon at EOF is + * assumed to be a mistake. + */ + fclose(f); + return 0; + } + + switch (state) + { + case StartLine: + /* Match spaces at the start of line */ + if (ch <= 32) /* skip control characters and space */ + break; + + counter=0; + state = Feature; + /* FALL THROUGH */ + + case Feature: + /* Match 'FEATURE', ASCII case insensitive. */ + if ((ch & ~0x20) == ch_feature[counter]) + { + if (++counter == (sizeof ch_feature)) + state = Colon; + break; + } + + /* did not match 'feature' */ + state = SkipLine; + /* FALL THROUGH */ + + case SkipLine: + skipLine: + /* Skip everything until we see linefeed or carriage return */ + if (ch != 10 && ch != 13) + break; + + state = StartLine; + break; + + case Colon: + /* Match any number of space or tab followed by ':' */ + if (ch == 32 || ch == 9) + break; + + if (ch == 58) /* i.e. ':' */ + { + state = StartTag; + break; + } + + /* Either a bad line format or a 'feature' prefix followed by + * other characters. + */ + state = SkipLine; + goto skipLine; + + case StartTag: + /* Skip space characters before a tag */ + if (ch == 32 || ch == 9) + break; + + state = Neon; + counter = 0; + /* FALL THROUGH */ + + case Neon: + /* Look for 'neon' tag */ + if ((ch & ~0x20) == ch_neon[counter]) + { + if (++counter == (sizeof ch_neon)) + state = HaveNeon; + break; + } + + state = SkipTag; + /* FALL THROUGH */ + + case SkipTag: + /* Skip non-space characters */ + if (ch == 10 || ch == 13) + state = StartLine; + + else if (ch == 32 || ch == 9) + state = StartTag; + break; + + case HaveNeon: + /* Have seen a 'neon' prefix, but there must be a space or new + * line character to terminate it. + */ + if (ch == 10 || ch == 13 || ch == 32 || ch == 9) + { + fclose(f); + return 1; + } + + state = SkipTag; + break; + + default: + png_error(png_ptr, "png_have_neon: internal error (bug)"); + } + } + } + +#ifdef PNG_WARNINGS_SUPPORTED + else + png_warning(png_ptr, "/proc/cpuinfo open failed"); +#endif + + return 0; +} diff --git a/app/src/main/jni/libpng/png.c b/app/src/main/jni/libpng/png.c new file mode 100644 index 000000000..4d93dfa91 --- /dev/null +++ b/app/src/main/jni/libpng/png.c @@ -0,0 +1,4493 @@ + +/* png.c - location for general purpose libpng functions + * + * Last changed in libpng 1.6.16 [December 22, 2014] + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + */ + +#include "pngpriv.h" + +/* Generate a compiler error if there is an old png.h in the search path. */ +typedef png_libpng_version_1_6_16 Your_png_h_is_not_version_1_6_16; + +/* Tells libpng that we have already handled the first "num_bytes" bytes + * of the PNG file signature. If the PNG data is embedded into another + * stream we can set num_bytes = 8 so that libpng will not attempt to read + * or write any of the magic bytes before it starts on the IHDR. + */ + +#ifdef PNG_READ_SUPPORTED +void PNGAPI +png_set_sig_bytes(png_structrp png_ptr, int num_bytes) +{ + png_debug(1, "in png_set_sig_bytes"); + + if (png_ptr == NULL) + return; + + if (num_bytes > 8) + png_error(png_ptr, "Too many bytes for PNG signature"); + + png_ptr->sig_bytes = (png_byte)(num_bytes < 0 ? 0 : num_bytes); +} + +/* Checks whether the supplied bytes match the PNG signature. We allow + * checking less than the full 8-byte signature so that those apps that + * already read the first few bytes of a file to determine the file type + * can simply check the remaining bytes for extra assurance. Returns + * an integer less than, equal to, or greater than zero if sig is found, + * respectively, to be less than, to match, or be greater than the correct + * PNG signature (this is the same behavior as strcmp, memcmp, etc). + */ +int PNGAPI +png_sig_cmp(png_const_bytep sig, png_size_t start, png_size_t num_to_check) +{ + png_byte png_signature[8] = {137, 80, 78, 71, 13, 10, 26, 10}; + + if (num_to_check > 8) + num_to_check = 8; + + else if (num_to_check < 1) + return (-1); + + if (start > 7) + return (-1); + + if (start + num_to_check > 8) + num_to_check = 8 - start; + + return ((int)(memcmp(&sig[start], &png_signature[start], num_to_check))); +} + +#endif /* READ */ + +#if defined(PNG_READ_SUPPORTED) || defined(PNG_WRITE_SUPPORTED) +/* Function to allocate memory for zlib */ +PNG_FUNCTION(voidpf /* PRIVATE */, +png_zalloc,(voidpf png_ptr, uInt items, uInt size),PNG_ALLOCATED) +{ + png_alloc_size_t num_bytes = size; + + if (png_ptr == NULL) + return NULL; + + if (items >= (~(png_alloc_size_t)0)/size) + { + png_warning (png_voidcast(png_structrp, png_ptr), + "Potential overflow in png_zalloc()"); + return NULL; + } + + num_bytes *= items; + return png_malloc_warn(png_voidcast(png_structrp, png_ptr), num_bytes); +} + +/* Function to free memory for zlib */ +void /* PRIVATE */ +png_zfree(voidpf png_ptr, voidpf ptr) +{ + png_free(png_voidcast(png_const_structrp,png_ptr), ptr); +} + +/* Reset the CRC variable to 32 bits of 1's. Care must be taken + * in case CRC is > 32 bits to leave the top bits 0. + */ +void /* PRIVATE */ +png_reset_crc(png_structrp png_ptr) +{ + /* The cast is safe because the crc is a 32 bit value. */ + png_ptr->crc = (png_uint_32)crc32(0, Z_NULL, 0); +} + +/* Calculate the CRC over a section of data. We can only pass as + * much data to this routine as the largest single buffer size. We + * also check that this data will actually be used before going to the + * trouble of calculating it. + */ +void /* PRIVATE */ +png_calculate_crc(png_structrp png_ptr, png_const_bytep ptr, png_size_t length) +{ + int need_crc = 1; + + if (PNG_CHUNK_ANCILLARY(png_ptr->chunk_name) != 0) + { + if ((png_ptr->flags & PNG_FLAG_CRC_ANCILLARY_MASK) == + (PNG_FLAG_CRC_ANCILLARY_USE | PNG_FLAG_CRC_ANCILLARY_NOWARN)) + need_crc = 0; + } + + else /* critical */ + { + if ((png_ptr->flags & PNG_FLAG_CRC_CRITICAL_IGNORE) != 0) + need_crc = 0; + } + + /* 'uLong' is defined in zlib.h as unsigned long; this means that on some + * systems it is a 64 bit value. crc32, however, returns 32 bits so the + * following cast is safe. 'uInt' may be no more than 16 bits, so it is + * necessary to perform a loop here. + */ + if (need_crc != 0 && length > 0) + { + uLong crc = png_ptr->crc; /* Should never issue a warning */ + + do + { + uInt safe_length = (uInt)length; + if (safe_length == 0) + safe_length = (uInt)-1; /* evil, but safe */ + + crc = crc32(crc, ptr, safe_length); + + /* The following should never issue compiler warnings; if they do the + * target system has characteristics that will probably violate other + * assumptions within the libpng code. + */ + ptr += safe_length; + length -= safe_length; + } + while (length > 0); + + /* And the following is always safe because the crc is only 32 bits. */ + png_ptr->crc = (png_uint_32)crc; + } +} + +/* Check a user supplied version number, called from both read and write + * functions that create a png_struct. + */ +int +png_user_version_check(png_structrp png_ptr, png_const_charp user_png_ver) +{ + /* Libpng versions 1.0.0 and later are binary compatible if the version + * string matches through the second '.'; we must recompile any + * applications that use any older library version. + */ + + if (user_png_ver != NULL) + { + int i = -1; + int found_dots = 0; + + do + { + i++; + if (user_png_ver[i] != PNG_LIBPNG_VER_STRING[i]) + png_ptr->flags |= PNG_FLAG_LIBRARY_MISMATCH; + if (user_png_ver[i] == '.') + found_dots++; + } while (found_dots < 2 && user_png_ver[i] != 0 && + PNG_LIBPNG_VER_STRING[i] != 0); + } + + else + png_ptr->flags |= PNG_FLAG_LIBRARY_MISMATCH; + + if ((png_ptr->flags & PNG_FLAG_LIBRARY_MISMATCH) != 0) + { +#ifdef PNG_WARNINGS_SUPPORTED + size_t pos = 0; + char m[128]; + + pos = png_safecat(m, (sizeof m), pos, + "Application built with libpng-"); + pos = png_safecat(m, (sizeof m), pos, user_png_ver); + pos = png_safecat(m, (sizeof m), pos, " but running with "); + pos = png_safecat(m, (sizeof m), pos, PNG_LIBPNG_VER_STRING); + PNG_UNUSED(pos) + + png_warning(png_ptr, m); +#endif + +#ifdef PNG_ERROR_NUMBERS_SUPPORTED + png_ptr->flags = 0; +#endif + + return 0; + } + + /* Success return. */ + return 1; +} + +/* Generic function to create a png_struct for either read or write - this + * contains the common initialization. + */ +PNG_FUNCTION(png_structp /* PRIVATE */, +png_create_png_struct,(png_const_charp user_png_ver, png_voidp error_ptr, + png_error_ptr error_fn, png_error_ptr warn_fn, png_voidp mem_ptr, + png_malloc_ptr malloc_fn, png_free_ptr free_fn),PNG_ALLOCATED) +{ + png_struct create_struct; +# ifdef PNG_SETJMP_SUPPORTED + jmp_buf create_jmp_buf; +# endif + + /* This temporary stack-allocated structure is used to provide a place to + * build enough context to allow the user provided memory allocator (if any) + * to be called. + */ + memset(&create_struct, 0, (sizeof create_struct)); + + /* Added at libpng-1.2.6 */ +# ifdef PNG_USER_LIMITS_SUPPORTED + create_struct.user_width_max = PNG_USER_WIDTH_MAX; + create_struct.user_height_max = PNG_USER_HEIGHT_MAX; + +# ifdef PNG_USER_CHUNK_CACHE_MAX + /* Added at libpng-1.2.43 and 1.4.0 */ + create_struct.user_chunk_cache_max = PNG_USER_CHUNK_CACHE_MAX; +# endif + +# ifdef PNG_USER_CHUNK_MALLOC_MAX + /* Added at libpng-1.2.43 and 1.4.1, required only for read but exists + * in png_struct regardless. + */ + create_struct.user_chunk_malloc_max = PNG_USER_CHUNK_MALLOC_MAX; +# endif +# endif + + /* The following two API calls simply set fields in png_struct, so it is safe + * to do them now even though error handling is not yet set up. + */ +# ifdef PNG_USER_MEM_SUPPORTED + png_set_mem_fn(&create_struct, mem_ptr, malloc_fn, free_fn); +# else + PNG_UNUSED(mem_ptr) + PNG_UNUSED(malloc_fn) + PNG_UNUSED(free_fn) +# endif + + /* (*error_fn) can return control to the caller after the error_ptr is set, + * this will result in a memory leak unless the error_fn does something + * extremely sophisticated. The design lacks merit but is implicit in the + * API. + */ + png_set_error_fn(&create_struct, error_ptr, error_fn, warn_fn); + +# ifdef PNG_SETJMP_SUPPORTED + if (!setjmp(create_jmp_buf)) + { + /* Temporarily fake out the longjmp information until we have + * successfully completed this function. This only works if we have + * setjmp() support compiled in, but it is safe - this stuff should + * never happen. + */ + create_struct.jmp_buf_ptr = &create_jmp_buf; + create_struct.jmp_buf_size = 0; /*stack allocation*/ + create_struct.longjmp_fn = longjmp; +# else + { +# endif + /* Call the general version checker (shared with read and write code): + */ + if (png_user_version_check(&create_struct, user_png_ver) != 0) + { + png_structrp png_ptr = png_voidcast(png_structrp, + png_malloc_warn(&create_struct, (sizeof *png_ptr))); + + if (png_ptr != NULL) + { + /* png_ptr->zstream holds a back-pointer to the png_struct, so + * this can only be done now: + */ + create_struct.zstream.zalloc = png_zalloc; + create_struct.zstream.zfree = png_zfree; + create_struct.zstream.opaque = png_ptr; + +# ifdef PNG_SETJMP_SUPPORTED + /* Eliminate the local error handling: */ + create_struct.jmp_buf_ptr = NULL; + create_struct.jmp_buf_size = 0; + create_struct.longjmp_fn = 0; +# endif + + *png_ptr = create_struct; + + /* This is the successful return point */ + return png_ptr; + } + } + } + + /* A longjmp because of a bug in the application storage allocator or a + * simple failure to allocate the png_struct. + */ + return NULL; +} + +/* Allocate the memory for an info_struct for the application. */ +PNG_FUNCTION(png_infop,PNGAPI +png_create_info_struct,(png_const_structrp png_ptr),PNG_ALLOCATED) +{ + png_inforp info_ptr; + + png_debug(1, "in png_create_info_struct"); + + if (png_ptr == NULL) + return NULL; + + /* Use the internal API that does not (or at least should not) error out, so + * that this call always returns ok. The application typically sets up the + * error handling *after* creating the info_struct because this is the way it + * has always been done in 'example.c'. + */ + info_ptr = png_voidcast(png_inforp, png_malloc_base(png_ptr, + (sizeof *info_ptr))); + + if (info_ptr != NULL) + memset(info_ptr, 0, (sizeof *info_ptr)); + + return info_ptr; +} + +/* This function frees the memory associated with a single info struct. + * Normally, one would use either png_destroy_read_struct() or + * png_destroy_write_struct() to free an info struct, but this may be + * useful for some applications. From libpng 1.6.0 this function is also used + * internally to implement the png_info release part of the 'struct' destroy + * APIs. This ensures that all possible approaches free the same data (all of + * it). + */ +void PNGAPI +png_destroy_info_struct(png_const_structrp png_ptr, png_infopp info_ptr_ptr) +{ + png_inforp info_ptr = NULL; + + png_debug(1, "in png_destroy_info_struct"); + + if (png_ptr == NULL) + return; + + if (info_ptr_ptr != NULL) + info_ptr = *info_ptr_ptr; + + if (info_ptr != NULL) + { + /* Do this first in case of an error below; if the app implements its own + * memory management this can lead to png_free calling png_error, which + * will abort this routine and return control to the app error handler. + * An infinite loop may result if it then tries to free the same info + * ptr. + */ + *info_ptr_ptr = NULL; + + png_free_data(png_ptr, info_ptr, PNG_FREE_ALL, -1); + memset(info_ptr, 0, (sizeof *info_ptr)); + png_free(png_ptr, info_ptr); + } +} + +/* Initialize the info structure. This is now an internal function (0.89) + * and applications using it are urged to use png_create_info_struct() + * instead. Use deprecated in 1.6.0, internal use removed (used internally it + * is just a memset). + * + * NOTE: it is almost inconceivable that this API is used because it bypasses + * the user-memory mechanism and the user error handling/warning mechanisms in + * those cases where it does anything other than a memset. + */ +PNG_FUNCTION(void,PNGAPI +png_info_init_3,(png_infopp ptr_ptr, png_size_t png_info_struct_size), + PNG_DEPRECATED) +{ + png_inforp info_ptr = *ptr_ptr; + + png_debug(1, "in png_info_init_3"); + + if (info_ptr == NULL) + return; + + if ((sizeof (png_info)) > png_info_struct_size) + { + *ptr_ptr = NULL; + /* The following line is why this API should not be used: */ + free(info_ptr); + info_ptr = png_voidcast(png_inforp, png_malloc_base(NULL, + (sizeof *info_ptr))); + *ptr_ptr = info_ptr; + } + + /* Set everything to 0 */ + memset(info_ptr, 0, (sizeof *info_ptr)); +} + +/* The following API is not called internally */ +void PNGAPI +png_data_freer(png_const_structrp png_ptr, png_inforp info_ptr, + int freer, png_uint_32 mask) +{ + png_debug(1, "in png_data_freer"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + if (freer == PNG_DESTROY_WILL_FREE_DATA) + info_ptr->free_me |= mask; + + else if (freer == PNG_USER_WILL_FREE_DATA) + info_ptr->free_me &= ~mask; + + else + png_error(png_ptr, "Unknown freer parameter in png_data_freer"); +} + +void PNGAPI +png_free_data(png_const_structrp png_ptr, png_inforp info_ptr, png_uint_32 mask, + int num) +{ + png_debug(1, "in png_free_data"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + +#ifdef PNG_TEXT_SUPPORTED + /* Free text item num or (if num == -1) all text items */ + if (info_ptr->text != 0 && + ((mask & PNG_FREE_TEXT) & info_ptr->free_me) != 0) + { + if (num != -1) + { + png_free(png_ptr, info_ptr->text[num].key); + info_ptr->text[num].key = NULL; + } + + else + { + int i; + + for (i = 0; i < info_ptr->num_text; i++) + png_free(png_ptr, info_ptr->text[i].key); + + png_free(png_ptr, info_ptr->text); + info_ptr->text = NULL; + info_ptr->num_text = 0; + } + } +#endif + +#ifdef PNG_tRNS_SUPPORTED + /* Free any tRNS entry */ + if (((mask & PNG_FREE_TRNS) & info_ptr->free_me) != 0) + { + png_free(png_ptr, info_ptr->trans_alpha); + info_ptr->trans_alpha = NULL; + info_ptr->valid &= ~PNG_INFO_tRNS; + } +#endif + +#ifdef PNG_sCAL_SUPPORTED + /* Free any sCAL entry */ + if (((mask & PNG_FREE_SCAL) & info_ptr->free_me) != 0) + { + png_free(png_ptr, info_ptr->scal_s_width); + png_free(png_ptr, info_ptr->scal_s_height); + info_ptr->scal_s_width = NULL; + info_ptr->scal_s_height = NULL; + info_ptr->valid &= ~PNG_INFO_sCAL; + } +#endif + +#ifdef PNG_pCAL_SUPPORTED + /* Free any pCAL entry */ + if (((mask & PNG_FREE_PCAL) & info_ptr->free_me) != 0) + { + png_free(png_ptr, info_ptr->pcal_purpose); + png_free(png_ptr, info_ptr->pcal_units); + info_ptr->pcal_purpose = NULL; + info_ptr->pcal_units = NULL; + + if (info_ptr->pcal_params != NULL) + { + int i; + + for (i = 0; i < info_ptr->pcal_nparams; i++) + png_free(png_ptr, info_ptr->pcal_params[i]); + + png_free(png_ptr, info_ptr->pcal_params); + info_ptr->pcal_params = NULL; + } + info_ptr->valid &= ~PNG_INFO_pCAL; + } +#endif + +#ifdef PNG_iCCP_SUPPORTED + /* Free any profile entry */ + if (((mask & PNG_FREE_ICCP) & info_ptr->free_me) != 0) + { + png_free(png_ptr, info_ptr->iccp_name); + png_free(png_ptr, info_ptr->iccp_profile); + info_ptr->iccp_name = NULL; + info_ptr->iccp_profile = NULL; + info_ptr->valid &= ~PNG_INFO_iCCP; + } +#endif + +#ifdef PNG_sPLT_SUPPORTED + /* Free a given sPLT entry, or (if num == -1) all sPLT entries */ + if (info_ptr->splt_palettes != 0 && + ((mask & PNG_FREE_SPLT) & info_ptr->free_me) != 0) + { + if (num != -1) + { + png_free(png_ptr, info_ptr->splt_palettes[num].name); + png_free(png_ptr, info_ptr->splt_palettes[num].entries); + info_ptr->splt_palettes[num].name = NULL; + info_ptr->splt_palettes[num].entries = NULL; + } + + else + { + if (info_ptr->splt_palettes_num != 0) + { + int i; + + for (i = 0; i < info_ptr->splt_palettes_num; i++) + { + png_free(png_ptr, info_ptr->splt_palettes[i].name); + png_free(png_ptr, info_ptr->splt_palettes[i].entries); + } + + png_free(png_ptr, info_ptr->splt_palettes); + info_ptr->splt_palettes = NULL; + info_ptr->splt_palettes_num = 0; + } + info_ptr->valid &= ~PNG_INFO_sPLT; + } + } +#endif + +#ifdef PNG_STORE_UNKNOWN_CHUNKS_SUPPORTED + if (info_ptr->unknown_chunks != 0 && + ((mask & PNG_FREE_UNKN) & info_ptr->free_me) != 0) + { + if (num != -1) + { + png_free(png_ptr, info_ptr->unknown_chunks[num].data); + info_ptr->unknown_chunks[num].data = NULL; + } + + else + { + int i; + + if (info_ptr->unknown_chunks_num != 0) + { + for (i = 0; i < info_ptr->unknown_chunks_num; i++) + png_free(png_ptr, info_ptr->unknown_chunks[i].data); + + png_free(png_ptr, info_ptr->unknown_chunks); + info_ptr->unknown_chunks = NULL; + info_ptr->unknown_chunks_num = 0; + } + } + } +#endif + +#ifdef PNG_hIST_SUPPORTED + /* Free any hIST entry */ + if (((mask & PNG_FREE_HIST) & info_ptr->free_me) != 0) + { + png_free(png_ptr, info_ptr->hist); + info_ptr->hist = NULL; + info_ptr->valid &= ~PNG_INFO_hIST; + } +#endif + + /* Free any PLTE entry that was internally allocated */ + if (((mask & PNG_FREE_PLTE) & info_ptr->free_me) != 0) + { + png_free(png_ptr, info_ptr->palette); + info_ptr->palette = NULL; + info_ptr->valid &= ~PNG_INFO_PLTE; + info_ptr->num_palette = 0; + } + +#ifdef PNG_INFO_IMAGE_SUPPORTED + /* Free any image bits attached to the info structure */ + if (((mask & PNG_FREE_ROWS) & info_ptr->free_me) != 0) + { + if (info_ptr->row_pointers != 0) + { + png_uint_32 row; + for (row = 0; row < info_ptr->height; row++) + png_free(png_ptr, info_ptr->row_pointers[row]); + + png_free(png_ptr, info_ptr->row_pointers); + info_ptr->row_pointers = NULL; + } + info_ptr->valid &= ~PNG_INFO_IDAT; + } +#endif + + if (num != -1) + mask &= ~PNG_FREE_MUL; + + info_ptr->free_me &= ~mask; +} +#endif /* READ || WRITE */ + +/* This function returns a pointer to the io_ptr associated with the user + * functions. The application should free any memory associated with this + * pointer before png_write_destroy() or png_read_destroy() are called. + */ +png_voidp PNGAPI +png_get_io_ptr(png_const_structrp png_ptr) +{ + if (png_ptr == NULL) + return (NULL); + + return (png_ptr->io_ptr); +} + +#if defined(PNG_READ_SUPPORTED) || defined(PNG_WRITE_SUPPORTED) +# ifdef PNG_STDIO_SUPPORTED +/* Initialize the default input/output functions for the PNG file. If you + * use your own read or write routines, you can call either png_set_read_fn() + * or png_set_write_fn() instead of png_init_io(). If you have defined + * PNG_NO_STDIO or otherwise disabled PNG_STDIO_SUPPORTED, you must use a + * function of your own because "FILE *" isn't necessarily available. + */ +void PNGAPI +png_init_io(png_structrp png_ptr, png_FILE_p fp) +{ + png_debug(1, "in png_init_io"); + + if (png_ptr == NULL) + return; + + png_ptr->io_ptr = (png_voidp)fp; +} +# endif + +#ifdef PNG_SAVE_INT_32_SUPPORTED +/* The png_save_int_32 function assumes integers are stored in two's + * complement format. If this isn't the case, then this routine needs to + * be modified to write data in two's complement format. Note that, + * the following works correctly even if png_int_32 has more than 32 bits + * (compare the more complex code required on read for sign extension.) + */ +void PNGAPI +png_save_int_32(png_bytep buf, png_int_32 i) +{ + buf[0] = (png_byte)((i >> 24) & 0xff); + buf[1] = (png_byte)((i >> 16) & 0xff); + buf[2] = (png_byte)((i >> 8) & 0xff); + buf[3] = (png_byte)(i & 0xff); +} +#endif + +# ifdef PNG_TIME_RFC1123_SUPPORTED +/* Convert the supplied time into an RFC 1123 string suitable for use in + * a "Creation Time" or other text-based time string. + */ +int PNGAPI +png_convert_to_rfc1123_buffer(char out[29], png_const_timep ptime) +{ + static PNG_CONST char short_months[12][4] = + {"Jan", "Feb", "Mar", "Apr", "May", "Jun", + "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"}; + + if (out == NULL) + return 0; + + if (ptime->year > 9999 /* RFC1123 limitation */ || + ptime->month == 0 || ptime->month > 12 || + ptime->day == 0 || ptime->day > 31 || + ptime->hour > 23 || ptime->minute > 59 || + ptime->second > 60) + return 0; + + { + size_t pos = 0; + char number_buf[5]; /* enough for a four-digit year */ + +# define APPEND_STRING(string) pos = png_safecat(out, 29, pos, (string)) +# define APPEND_NUMBER(format, value)\ + APPEND_STRING(PNG_FORMAT_NUMBER(number_buf, format, (value))) +# define APPEND(ch) if (pos < 28) out[pos++] = (ch) + + APPEND_NUMBER(PNG_NUMBER_FORMAT_u, (unsigned)ptime->day); + APPEND(' '); + APPEND_STRING(short_months[(ptime->month - 1)]); + APPEND(' '); + APPEND_NUMBER(PNG_NUMBER_FORMAT_u, ptime->year); + APPEND(' '); + APPEND_NUMBER(PNG_NUMBER_FORMAT_02u, (unsigned)ptime->hour); + APPEND(':'); + APPEND_NUMBER(PNG_NUMBER_FORMAT_02u, (unsigned)ptime->minute); + APPEND(':'); + APPEND_NUMBER(PNG_NUMBER_FORMAT_02u, (unsigned)ptime->second); + APPEND_STRING(" +0000"); /* This reliably terminates the buffer */ + +# undef APPEND +# undef APPEND_NUMBER +# undef APPEND_STRING + } + + return 1; +} + +# if PNG_LIBPNG_VER < 10700 +/* To do: remove the following from libpng-1.7 */ +/* Original API that uses a private buffer in png_struct. + * Deprecated because it causes png_struct to carry a spurious temporary + * buffer (png_struct::time_buffer), better to have the caller pass this in. + */ +png_const_charp PNGAPI +png_convert_to_rfc1123(png_structrp png_ptr, png_const_timep ptime) +{ + if (png_ptr != NULL) + { + /* The only failure above if png_ptr != NULL is from an invalid ptime */ + if (png_convert_to_rfc1123_buffer(png_ptr->time_buffer, ptime) == 0) + png_warning(png_ptr, "Ignoring invalid time value"); + + else + return png_ptr->time_buffer; + } + + return NULL; +} +# endif +# endif /* TIME_RFC1123 */ + +#endif /* READ || WRITE */ + +png_const_charp PNGAPI +png_get_copyright(png_const_structrp png_ptr) +{ + PNG_UNUSED(png_ptr) /* Silence compiler warning about unused png_ptr */ +#ifdef PNG_STRING_COPYRIGHT + return PNG_STRING_COPYRIGHT +#else +# ifdef __STDC__ + return PNG_STRING_NEWLINE \ + "libpng version 1.6.16 - December 22, 2014" PNG_STRING_NEWLINE \ + "Copyright (c) 1998-2014 Glenn Randers-Pehrson" PNG_STRING_NEWLINE \ + "Copyright (c) 1996-1997 Andreas Dilger" PNG_STRING_NEWLINE \ + "Copyright (c) 1995-1996 Guy Eric Schalnat, Group 42, Inc." \ + PNG_STRING_NEWLINE; +# else + return "libpng version 1.6.16 - December 22, 2014\ + Copyright (c) 1998-2014 Glenn Randers-Pehrson\ + Copyright (c) 1996-1997 Andreas Dilger\ + Copyright (c) 1995-1996 Guy Eric Schalnat, Group 42, Inc."; +# endif +#endif +} + +/* The following return the library version as a short string in the + * format 1.0.0 through 99.99.99zz. To get the version of *.h files + * used with your application, print out PNG_LIBPNG_VER_STRING, which + * is defined in png.h. + * Note: now there is no difference between png_get_libpng_ver() and + * png_get_header_ver(). Due to the version_nn_nn_nn typedef guard, + * it is guaranteed that png.c uses the correct version of png.h. + */ +png_const_charp PNGAPI +png_get_libpng_ver(png_const_structrp png_ptr) +{ + /* Version of *.c files used when building libpng */ + return png_get_header_ver(png_ptr); +} + +png_const_charp PNGAPI +png_get_header_ver(png_const_structrp png_ptr) +{ + /* Version of *.h files used when building libpng */ + PNG_UNUSED(png_ptr) /* Silence compiler warning about unused png_ptr */ + return PNG_LIBPNG_VER_STRING; +} + +png_const_charp PNGAPI +png_get_header_version(png_const_structrp png_ptr) +{ + /* Returns longer string containing both version and date */ + PNG_UNUSED(png_ptr) /* Silence compiler warning about unused png_ptr */ +#ifdef __STDC__ + return PNG_HEADER_VERSION_STRING +# ifndef PNG_READ_SUPPORTED + " (NO READ SUPPORT)" +# endif + PNG_STRING_NEWLINE; +#else + return PNG_HEADER_VERSION_STRING; +#endif +} + +#ifdef PNG_BUILD_GRAYSCALE_PALETTE_SUPPORTED +/* NOTE: this routine is not used internally! */ +/* Build a grayscale palette. Palette is assumed to be 1 << bit_depth + * large of png_color. This lets grayscale images be treated as + * paletted. Most useful for gamma correction and simplification + * of code. This API is not used internally. + */ +void PNGAPI +png_build_grayscale_palette(int bit_depth, png_colorp palette) +{ + int num_palette; + int color_inc; + int i; + int v; + + png_debug(1, "in png_do_build_grayscale_palette"); + + if (palette == NULL) + return; + + switch (bit_depth) + { + case 1: + num_palette = 2; + color_inc = 0xff; + break; + + case 2: + num_palette = 4; + color_inc = 0x55; + break; + + case 4: + num_palette = 16; + color_inc = 0x11; + break; + + case 8: + num_palette = 256; + color_inc = 1; + break; + + default: + num_palette = 0; + color_inc = 0; + break; + } + + for (i = 0, v = 0; i < num_palette; i++, v += color_inc) + { + palette[i].red = (png_byte)v; + palette[i].green = (png_byte)v; + palette[i].blue = (png_byte)v; + } +} +#endif + +#ifdef PNG_SET_UNKNOWN_CHUNKS_SUPPORTED +int PNGAPI +png_handle_as_unknown(png_const_structrp png_ptr, png_const_bytep chunk_name) +{ + /* Check chunk_name and return "keep" value if it's on the list, else 0 */ + png_const_bytep p, p_end; + + if (png_ptr == NULL || chunk_name == NULL || png_ptr->num_chunk_list == 0) + return PNG_HANDLE_CHUNK_AS_DEFAULT; + + p_end = png_ptr->chunk_list; + p = p_end + png_ptr->num_chunk_list*5; /* beyond end */ + + /* The code is the fifth byte after each four byte string. Historically this + * code was always searched from the end of the list, this is no longer + * necessary because the 'set' routine handles duplicate entries correcty. + */ + do /* num_chunk_list > 0, so at least one */ + { + p -= 5; + + if (memcmp(chunk_name, p, 4) == 0) + return p[4]; + } + while (p > p_end); + + /* This means that known chunks should be processed and unknown chunks should + * be handled according to the value of png_ptr->unknown_default; this can be + * confusing because, as a result, there are two levels of defaulting for + * unknown chunks. + */ + return PNG_HANDLE_CHUNK_AS_DEFAULT; +} + +#if defined(PNG_READ_UNKNOWN_CHUNKS_SUPPORTED) ||\ + defined(PNG_HANDLE_AS_UNKNOWN_SUPPORTED) +int /* PRIVATE */ +png_chunk_unknown_handling(png_const_structrp png_ptr, png_uint_32 chunk_name) +{ + png_byte chunk_string[5]; + + PNG_CSTRING_FROM_CHUNK(chunk_string, chunk_name); + return png_handle_as_unknown(png_ptr, chunk_string); +} +#endif /* READ_UNKNOWN_CHUNKS || HANDLE_AS_UNKNOWN */ +#endif /* SET_UNKNOWN_CHUNKS */ + +#ifdef PNG_READ_SUPPORTED +/* This function, added to libpng-1.0.6g, is untested. */ +int PNGAPI +png_reset_zstream(png_structrp png_ptr) +{ + if (png_ptr == NULL) + return Z_STREAM_ERROR; + + /* WARNING: this resets the window bits to the maximum! */ + return (inflateReset(&png_ptr->zstream)); +} +#endif /* READ */ + +/* This function was added to libpng-1.0.7 */ +png_uint_32 PNGAPI +png_access_version_number(void) +{ + /* Version of *.c files used when building libpng */ + return((png_uint_32)PNG_LIBPNG_VER); +} + + + +#if defined(PNG_READ_SUPPORTED) || defined(PNG_WRITE_SUPPORTED) +/* Ensure that png_ptr->zstream.msg holds some appropriate error message string. + * If it doesn't 'ret' is used to set it to something appropriate, even in cases + * like Z_OK or Z_STREAM_END where the error code is apparently a success code. + */ +void /* PRIVATE */ +png_zstream_error(png_structrp png_ptr, int ret) +{ + /* Translate 'ret' into an appropriate error string, priority is given to the + * one in zstream if set. This always returns a string, even in cases like + * Z_OK or Z_STREAM_END where the error code is a success code. + */ + if (png_ptr->zstream.msg == NULL) switch (ret) + { + default: + case Z_OK: + png_ptr->zstream.msg = PNGZ_MSG_CAST("unexpected zlib return code"); + break; + + case Z_STREAM_END: + /* Normal exit */ + png_ptr->zstream.msg = PNGZ_MSG_CAST("unexpected end of LZ stream"); + break; + + case Z_NEED_DICT: + /* This means the deflate stream did not have a dictionary; this + * indicates a bogus PNG. + */ + png_ptr->zstream.msg = PNGZ_MSG_CAST("missing LZ dictionary"); + break; + + case Z_ERRNO: + /* gz APIs only: should not happen */ + png_ptr->zstream.msg = PNGZ_MSG_CAST("zlib IO error"); + break; + + case Z_STREAM_ERROR: + /* internal libpng error */ + png_ptr->zstream.msg = PNGZ_MSG_CAST("bad parameters to zlib"); + break; + + case Z_DATA_ERROR: + png_ptr->zstream.msg = PNGZ_MSG_CAST("damaged LZ stream"); + break; + + case Z_MEM_ERROR: + png_ptr->zstream.msg = PNGZ_MSG_CAST("insufficient memory"); + break; + + case Z_BUF_ERROR: + /* End of input or output; not a problem if the caller is doing + * incremental read or write. + */ + png_ptr->zstream.msg = PNGZ_MSG_CAST("truncated"); + break; + + case Z_VERSION_ERROR: + png_ptr->zstream.msg = PNGZ_MSG_CAST("unsupported zlib version"); + break; + + case PNG_UNEXPECTED_ZLIB_RETURN: + /* Compile errors here mean that zlib now uses the value co-opted in + * pngpriv.h for PNG_UNEXPECTED_ZLIB_RETURN; update the switch above + * and change pngpriv.h. Note that this message is "... return", + * whereas the default/Z_OK one is "... return code". + */ + png_ptr->zstream.msg = PNGZ_MSG_CAST("unexpected zlib return"); + break; + } +} + +/* png_convert_size: a PNGAPI but no longer in png.h, so deleted + * at libpng 1.5.5! + */ + +/* Added at libpng version 1.2.34 and 1.4.0 (moved from pngset.c) */ +#ifdef PNG_GAMMA_SUPPORTED /* always set if COLORSPACE */ +static int +png_colorspace_check_gamma(png_const_structrp png_ptr, + png_colorspacerp colorspace, png_fixed_point gAMA, int from) + /* This is called to check a new gamma value against an existing one. The + * routine returns false if the new gamma value should not be written. + * + * 'from' says where the new gamma value comes from: + * + * 0: the new gamma value is the libpng estimate for an ICC profile + * 1: the new gamma value comes from a gAMA chunk + * 2: the new gamma value comes from an sRGB chunk + */ +{ + png_fixed_point gtest; + + if ((colorspace->flags & PNG_COLORSPACE_HAVE_GAMMA) != 0 && + (png_muldiv(>est, colorspace->gamma, PNG_FP_1, gAMA) == 0 || + png_gamma_significant(gtest) != 0)) + { + /* Either this is an sRGB image, in which case the calculated gamma + * approximation should match, or this is an image with a profile and the + * value libpng calculates for the gamma of the profile does not match the + * value recorded in the file. The former, sRGB, case is an error, the + * latter is just a warning. + */ + if ((colorspace->flags & PNG_COLORSPACE_FROM_sRGB) != 0 || from == 2) + { + png_chunk_report(png_ptr, "gamma value does not match sRGB", + PNG_CHUNK_ERROR); + /* Do not overwrite an sRGB value */ + return from == 2; + } + + else /* sRGB tag not involved */ + { + png_chunk_report(png_ptr, "gamma value does not match libpng estimate", + PNG_CHUNK_WARNING); + return from == 1; + } + } + + return 1; +} + +void /* PRIVATE */ +png_colorspace_set_gamma(png_const_structrp png_ptr, + png_colorspacerp colorspace, png_fixed_point gAMA) +{ + /* Changed in libpng-1.5.4 to limit the values to ensure overflow can't + * occur. Since the fixed point representation is asymetrical it is + * possible for 1/gamma to overflow the limit of 21474 and this means the + * gamma value must be at least 5/100000 and hence at most 20000.0. For + * safety the limits here are a little narrower. The values are 0.00016 to + * 6250.0, which are truly ridiculous gamma values (and will produce + * displays that are all black or all white.) + * + * In 1.6.0 this test replaces the ones in pngrutil.c, in the gAMA chunk + * handling code, which only required the value to be >0. + */ + png_const_charp errmsg; + + if (gAMA < 16 || gAMA > 625000000) + errmsg = "gamma value out of range"; + +# ifdef PNG_READ_gAMA_SUPPORTED + /* Allow the application to set the gamma value more than once */ + else if ((png_ptr->mode & PNG_IS_READ_STRUCT) != 0 && + (colorspace->flags & PNG_COLORSPACE_FROM_gAMA) != 0) + errmsg = "duplicate"; +# endif + + /* Do nothing if the colorspace is already invalid */ + else if ((colorspace->flags & PNG_COLORSPACE_INVALID) != 0) + return; + + else + { + if (png_colorspace_check_gamma(png_ptr, colorspace, gAMA, + 1/*from gAMA*/) != 0) + { + /* Store this gamma value. */ + colorspace->gamma = gAMA; + colorspace->flags |= + (PNG_COLORSPACE_HAVE_GAMMA | PNG_COLORSPACE_FROM_gAMA); + } + + /* At present if the check_gamma test fails the gamma of the colorspace is + * not updated however the colorspace is not invalidated. This + * corresponds to the case where the existing gamma comes from an sRGB + * chunk or profile. An error message has already been output. + */ + return; + } + + /* Error exit - errmsg has been set. */ + colorspace->flags |= PNG_COLORSPACE_INVALID; + png_chunk_report(png_ptr, errmsg, PNG_CHUNK_WRITE_ERROR); +} + +void /* PRIVATE */ +png_colorspace_sync_info(png_const_structrp png_ptr, png_inforp info_ptr) +{ + if ((info_ptr->colorspace.flags & PNG_COLORSPACE_INVALID) != 0) + { + /* Everything is invalid */ + info_ptr->valid &= ~(PNG_INFO_gAMA|PNG_INFO_cHRM|PNG_INFO_sRGB| + PNG_INFO_iCCP); + +# ifdef PNG_COLORSPACE_SUPPORTED + /* Clean up the iCCP profile now if it won't be used. */ + png_free_data(png_ptr, info_ptr, PNG_FREE_ICCP, -1/*not used*/); +# else + PNG_UNUSED(png_ptr) +# endif + } + + else + { +# ifdef PNG_COLORSPACE_SUPPORTED + /* Leave the INFO_iCCP flag set if the pngset.c code has already set + * it; this allows a PNG to contain a profile which matches sRGB and + * yet still have that profile retrievable by the application. + */ + if ((info_ptr->colorspace.flags & PNG_COLORSPACE_MATCHES_sRGB) != 0) + info_ptr->valid |= PNG_INFO_sRGB; + + else + info_ptr->valid &= ~PNG_INFO_sRGB; + + if ((info_ptr->colorspace.flags & PNG_COLORSPACE_HAVE_ENDPOINTS) != 0) + info_ptr->valid |= PNG_INFO_cHRM; + + else + info_ptr->valid &= ~PNG_INFO_cHRM; +# endif + + if ((info_ptr->colorspace.flags & PNG_COLORSPACE_HAVE_GAMMA) != 0) + info_ptr->valid |= PNG_INFO_gAMA; + + else + info_ptr->valid &= ~PNG_INFO_gAMA; + } +} + +#ifdef PNG_READ_SUPPORTED +void /* PRIVATE */ +png_colorspace_sync(png_const_structrp png_ptr, png_inforp info_ptr) +{ + if (info_ptr == NULL) /* reduce code size; check here not in the caller */ + return; + + info_ptr->colorspace = png_ptr->colorspace; + png_colorspace_sync_info(png_ptr, info_ptr); +} +#endif +#endif + +#ifdef PNG_COLORSPACE_SUPPORTED +/* Added at libpng-1.5.5 to support read and write of true CIEXYZ values for + * cHRM, as opposed to using chromaticities. These internal APIs return + * non-zero on a parameter error. The X, Y and Z values are required to be + * positive and less than 1.0. + */ +static int +png_xy_from_XYZ(png_xy *xy, const png_XYZ *XYZ) +{ + png_int_32 d, dwhite, whiteX, whiteY; + + d = XYZ->red_X + XYZ->red_Y + XYZ->red_Z; + if (png_muldiv(&xy->redx, XYZ->red_X, PNG_FP_1, d) == 0) + return 1; + if (png_muldiv(&xy->redy, XYZ->red_Y, PNG_FP_1, d) == 0) + return 1; + dwhite = d; + whiteX = XYZ->red_X; + whiteY = XYZ->red_Y; + + d = XYZ->green_X + XYZ->green_Y + XYZ->green_Z; + if (png_muldiv(&xy->greenx, XYZ->green_X, PNG_FP_1, d) == 0) + return 1; + if (png_muldiv(&xy->greeny, XYZ->green_Y, PNG_FP_1, d) == 0) + return 1; + dwhite += d; + whiteX += XYZ->green_X; + whiteY += XYZ->green_Y; + + d = XYZ->blue_X + XYZ->blue_Y + XYZ->blue_Z; + if (png_muldiv(&xy->bluex, XYZ->blue_X, PNG_FP_1, d) == 0) + return 1; + if (png_muldiv(&xy->bluey, XYZ->blue_Y, PNG_FP_1, d) == 0) + return 1; + dwhite += d; + whiteX += XYZ->blue_X; + whiteY += XYZ->blue_Y; + + /* The reference white is simply the sum of the end-point (X,Y,Z) vectors, + * thus: + */ + if (png_muldiv(&xy->whitex, whiteX, PNG_FP_1, dwhite) == 0) + return 1; + if (png_muldiv(&xy->whitey, whiteY, PNG_FP_1, dwhite) == 0) + return 1; + + return 0; +} + +static int +png_XYZ_from_xy(png_XYZ *XYZ, const png_xy *xy) +{ + png_fixed_point red_inverse, green_inverse, blue_scale; + png_fixed_point left, right, denominator; + + /* Check xy and, implicitly, z. Note that wide gamut color spaces typically + * have end points with 0 tristimulus values (these are impossible end + * points, but they are used to cover the possible colors.) + */ + if (xy->redx < 0 || xy->redx > PNG_FP_1) return 1; + if (xy->redy < 0 || xy->redy > PNG_FP_1-xy->redx) return 1; + if (xy->greenx < 0 || xy->greenx > PNG_FP_1) return 1; + if (xy->greeny < 0 || xy->greeny > PNG_FP_1-xy->greenx) return 1; + if (xy->bluex < 0 || xy->bluex > PNG_FP_1) return 1; + if (xy->bluey < 0 || xy->bluey > PNG_FP_1-xy->bluex) return 1; + if (xy->whitex < 0 || xy->whitex > PNG_FP_1) return 1; + if (xy->whitey < 0 || xy->whitey > PNG_FP_1-xy->whitex) return 1; + + /* The reverse calculation is more difficult because the original tristimulus + * value had 9 independent values (red,green,blue)x(X,Y,Z) however only 8 + * derived values were recorded in the cHRM chunk; + * (red,green,blue,white)x(x,y). This loses one degree of freedom and + * therefore an arbitrary ninth value has to be introduced to undo the + * original transformations. + * + * Think of the original end-points as points in (X,Y,Z) space. The + * chromaticity values (c) have the property: + * + * C + * c = --------- + * X + Y + Z + * + * For each c (x,y,z) from the corresponding original C (X,Y,Z). Thus the + * three chromaticity values (x,y,z) for each end-point obey the + * relationship: + * + * x + y + z = 1 + * + * This describes the plane in (X,Y,Z) space that intersects each axis at the + * value 1.0; call this the chromaticity plane. Thus the chromaticity + * calculation has scaled each end-point so that it is on the x+y+z=1 plane + * and chromaticity is the intersection of the vector from the origin to the + * (X,Y,Z) value with the chromaticity plane. + * + * To fully invert the chromaticity calculation we would need the three + * end-point scale factors, (red-scale, green-scale, blue-scale), but these + * were not recorded. Instead we calculated the reference white (X,Y,Z) and + * recorded the chromaticity of this. The reference white (X,Y,Z) would have + * given all three of the scale factors since: + * + * color-C = color-c * color-scale + * white-C = red-C + green-C + blue-C + * = red-c*red-scale + green-c*green-scale + blue-c*blue-scale + * + * But cHRM records only white-x and white-y, so we have lost the white scale + * factor: + * + * white-C = white-c*white-scale + * + * To handle this the inverse transformation makes an arbitrary assumption + * about white-scale: + * + * Assume: white-Y = 1.0 + * Hence: white-scale = 1/white-y + * Or: red-Y + green-Y + blue-Y = 1.0 + * + * Notice the last statement of the assumption gives an equation in three of + * the nine values we want to calculate. 8 more equations come from the + * above routine as summarised at the top above (the chromaticity + * calculation): + * + * Given: color-x = color-X / (color-X + color-Y + color-Z) + * Hence: (color-x - 1)*color-X + color.x*color-Y + color.x*color-Z = 0 + * + * This is 9 simultaneous equations in the 9 variables "color-C" and can be + * solved by Cramer's rule. Cramer's rule requires calculating 10 9x9 matrix + * determinants, however this is not as bad as it seems because only 28 of + * the total of 90 terms in the various matrices are non-zero. Nevertheless + * Cramer's rule is notoriously numerically unstable because the determinant + * calculation involves the difference of large, but similar, numbers. It is + * difficult to be sure that the calculation is stable for real world values + * and it is certain that it becomes unstable where the end points are close + * together. + * + * So this code uses the perhaps slightly less optimal but more + * understandable and totally obvious approach of calculating color-scale. + * + * This algorithm depends on the precision in white-scale and that is + * (1/white-y), so we can immediately see that as white-y approaches 0 the + * accuracy inherent in the cHRM chunk drops off substantially. + * + * libpng arithmetic: a simple inversion of the above equations + * ------------------------------------------------------------ + * + * white_scale = 1/white-y + * white-X = white-x * white-scale + * white-Y = 1.0 + * white-Z = (1 - white-x - white-y) * white_scale + * + * white-C = red-C + green-C + blue-C + * = red-c*red-scale + green-c*green-scale + blue-c*blue-scale + * + * This gives us three equations in (red-scale,green-scale,blue-scale) where + * all the coefficients are now known: + * + * red-x*red-scale + green-x*green-scale + blue-x*blue-scale + * = white-x/white-y + * red-y*red-scale + green-y*green-scale + blue-y*blue-scale = 1 + * red-z*red-scale + green-z*green-scale + blue-z*blue-scale + * = (1 - white-x - white-y)/white-y + * + * In the last equation color-z is (1 - color-x - color-y) so we can add all + * three equations together to get an alternative third: + * + * red-scale + green-scale + blue-scale = 1/white-y = white-scale + * + * So now we have a Cramer's rule solution where the determinants are just + * 3x3 - far more tractible. Unfortunately 3x3 determinants still involve + * multiplication of three coefficients so we can't guarantee to avoid + * overflow in the libpng fixed point representation. Using Cramer's rule in + * floating point is probably a good choice here, but it's not an option for + * fixed point. Instead proceed to simplify the first two equations by + * eliminating what is likely to be the largest value, blue-scale: + * + * blue-scale = white-scale - red-scale - green-scale + * + * Hence: + * + * (red-x - blue-x)*red-scale + (green-x - blue-x)*green-scale = + * (white-x - blue-x)*white-scale + * + * (red-y - blue-y)*red-scale + (green-y - blue-y)*green-scale = + * 1 - blue-y*white-scale + * + * And now we can trivially solve for (red-scale,green-scale): + * + * green-scale = + * (white-x - blue-x)*white-scale - (red-x - blue-x)*red-scale + * ----------------------------------------------------------- + * green-x - blue-x + * + * red-scale = + * 1 - blue-y*white-scale - (green-y - blue-y) * green-scale + * --------------------------------------------------------- + * red-y - blue-y + * + * Hence: + * + * red-scale = + * ( (green-x - blue-x) * (white-y - blue-y) - + * (green-y - blue-y) * (white-x - blue-x) ) / white-y + * ------------------------------------------------------------------------- + * (green-x - blue-x)*(red-y - blue-y)-(green-y - blue-y)*(red-x - blue-x) + * + * green-scale = + * ( (red-y - blue-y) * (white-x - blue-x) - + * (red-x - blue-x) * (white-y - blue-y) ) / white-y + * ------------------------------------------------------------------------- + * (green-x - blue-x)*(red-y - blue-y)-(green-y - blue-y)*(red-x - blue-x) + * + * Accuracy: + * The input values have 5 decimal digits of accuracy. The values are all in + * the range 0 < value < 1, so simple products are in the same range but may + * need up to 10 decimal digits to preserve the original precision and avoid + * underflow. Because we are using a 32-bit signed representation we cannot + * match this; the best is a little over 9 decimal digits, less than 10. + * + * The approach used here is to preserve the maximum precision within the + * signed representation. Because the red-scale calculation above uses the + * difference between two products of values that must be in the range -1..+1 + * it is sufficient to divide the product by 7; ceil(100,000/32767*2). The + * factor is irrelevant in the calculation because it is applied to both + * numerator and denominator. + * + * Note that the values of the differences of the products of the + * chromaticities in the above equations tend to be small, for example for + * the sRGB chromaticities they are: + * + * red numerator: -0.04751 + * green numerator: -0.08788 + * denominator: -0.2241 (without white-y multiplication) + * + * The resultant Y coefficients from the chromaticities of some widely used + * color space definitions are (to 15 decimal places): + * + * sRGB + * 0.212639005871510 0.715168678767756 0.072192315360734 + * Kodak ProPhoto + * 0.288071128229293 0.711843217810102 0.000085653960605 + * Adobe RGB + * 0.297344975250536 0.627363566255466 0.075291458493998 + * Adobe Wide Gamut RGB + * 0.258728243040113 0.724682314948566 0.016589442011321 + */ + /* By the argument, above overflow should be impossible here. The return + * value of 2 indicates an internal error to the caller. + */ + if (png_muldiv(&left, xy->greenx-xy->bluex, xy->redy - xy->bluey, 7) == 0) + return 2; + if (png_muldiv(&right, xy->greeny-xy->bluey, xy->redx - xy->bluex, 7) == 0) + return 2; + denominator = left - right; + + /* Now find the red numerator. */ + if (png_muldiv(&left, xy->greenx-xy->bluex, xy->whitey-xy->bluey, 7) == 0) + return 2; + if (png_muldiv(&right, xy->greeny-xy->bluey, xy->whitex-xy->bluex, 7) == 0) + return 2; + + /* Overflow is possible here and it indicates an extreme set of PNG cHRM + * chunk values. This calculation actually returns the reciprocal of the + * scale value because this allows us to delay the multiplication of white-y + * into the denominator, which tends to produce a small number. + */ + if (png_muldiv(&red_inverse, xy->whitey, denominator, left-right) == 0 || + red_inverse <= xy->whitey /* r+g+b scales = white scale */) + return 1; + + /* Similarly for green_inverse: */ + if (png_muldiv(&left, xy->redy-xy->bluey, xy->whitex-xy->bluex, 7) == 0) + return 2; + if (png_muldiv(&right, xy->redx-xy->bluex, xy->whitey-xy->bluey, 7) == 0) + return 2; + if (png_muldiv(&green_inverse, xy->whitey, denominator, left-right) == 0 || + green_inverse <= xy->whitey) + return 1; + + /* And the blue scale, the checks above guarantee this can't overflow but it + * can still produce 0 for extreme cHRM values. + */ + blue_scale = png_reciprocal(xy->whitey) - png_reciprocal(red_inverse) - + png_reciprocal(green_inverse); + if (blue_scale <= 0) + return 1; + + + /* And fill in the png_XYZ: */ + if (png_muldiv(&XYZ->red_X, xy->redx, PNG_FP_1, red_inverse) == 0) + return 1; + if (png_muldiv(&XYZ->red_Y, xy->redy, PNG_FP_1, red_inverse) == 0) + return 1; + if (png_muldiv(&XYZ->red_Z, PNG_FP_1 - xy->redx - xy->redy, PNG_FP_1, + red_inverse) == 0) + return 1; + + if (png_muldiv(&XYZ->green_X, xy->greenx, PNG_FP_1, green_inverse) == 0) + return 1; + if (png_muldiv(&XYZ->green_Y, xy->greeny, PNG_FP_1, green_inverse) == 0) + return 1; + if (png_muldiv(&XYZ->green_Z, PNG_FP_1 - xy->greenx - xy->greeny, PNG_FP_1, + green_inverse) == 0) + return 1; + + if (png_muldiv(&XYZ->blue_X, xy->bluex, blue_scale, PNG_FP_1) == 0) + return 1; + if (png_muldiv(&XYZ->blue_Y, xy->bluey, blue_scale, PNG_FP_1) == 0) + return 1; + if (png_muldiv(&XYZ->blue_Z, PNG_FP_1 - xy->bluex - xy->bluey, blue_scale, + PNG_FP_1) == 0) + return 1; + + return 0; /*success*/ +} + +static int +png_XYZ_normalize(png_XYZ *XYZ) +{ + png_int_32 Y; + + if (XYZ->red_Y < 0 || XYZ->green_Y < 0 || XYZ->blue_Y < 0 || + XYZ->red_X < 0 || XYZ->green_X < 0 || XYZ->blue_X < 0 || + XYZ->red_Z < 0 || XYZ->green_Z < 0 || XYZ->blue_Z < 0) + return 1; + + /* Normalize by scaling so the sum of the end-point Y values is PNG_FP_1. + * IMPLEMENTATION NOTE: ANSI requires signed overflow not to occur, therefore + * relying on addition of two positive values producing a negative one is not + * safe. + */ + Y = XYZ->red_Y; + if (0x7fffffff - Y < XYZ->green_X) + return 1; + Y += XYZ->green_Y; + if (0x7fffffff - Y < XYZ->blue_X) + return 1; + Y += XYZ->blue_Y; + + if (Y != PNG_FP_1) + { + if (png_muldiv(&XYZ->red_X, XYZ->red_X, PNG_FP_1, Y) == 0) + return 1; + if (png_muldiv(&XYZ->red_Y, XYZ->red_Y, PNG_FP_1, Y) == 0) + return 1; + if (png_muldiv(&XYZ->red_Z, XYZ->red_Z, PNG_FP_1, Y) == 0) + return 1; + + if (png_muldiv(&XYZ->green_X, XYZ->green_X, PNG_FP_1, Y) == 0) + return 1; + if (png_muldiv(&XYZ->green_Y, XYZ->green_Y, PNG_FP_1, Y) == 0) + return 1; + if (png_muldiv(&XYZ->green_Z, XYZ->green_Z, PNG_FP_1, Y) == 0) + return 1; + + if (png_muldiv(&XYZ->blue_X, XYZ->blue_X, PNG_FP_1, Y) == 0) + return 1; + if (png_muldiv(&XYZ->blue_Y, XYZ->blue_Y, PNG_FP_1, Y) == 0) + return 1; + if (png_muldiv(&XYZ->blue_Z, XYZ->blue_Z, PNG_FP_1, Y) == 0) + return 1; + } + + return 0; +} + +static int +png_colorspace_endpoints_match(const png_xy *xy1, const png_xy *xy2, int delta) +{ + /* Allow an error of +/-0.01 (absolute value) on each chromaticity */ + if (PNG_OUT_OF_RANGE(xy1->whitex, xy2->whitex,delta) || + PNG_OUT_OF_RANGE(xy1->whitey, xy2->whitey,delta) || + PNG_OUT_OF_RANGE(xy1->redx, xy2->redx, delta) || + PNG_OUT_OF_RANGE(xy1->redy, xy2->redy, delta) || + PNG_OUT_OF_RANGE(xy1->greenx, xy2->greenx,delta) || + PNG_OUT_OF_RANGE(xy1->greeny, xy2->greeny,delta) || + PNG_OUT_OF_RANGE(xy1->bluex, xy2->bluex, delta) || + PNG_OUT_OF_RANGE(xy1->bluey, xy2->bluey, delta)) + return 0; + return 1; +} + +/* Added in libpng-1.6.0, a different check for the validity of a set of cHRM + * chunk chromaticities. Earlier checks used to simply look for the overflow + * condition (where the determinant of the matrix to solve for XYZ ends up zero + * because the chromaticity values are not all distinct.) Despite this it is + * theoretically possible to produce chromaticities that are apparently valid + * but that rapidly degrade to invalid, potentially crashing, sets because of + * arithmetic inaccuracies when calculations are performed on them. The new + * check is to round-trip xy -> XYZ -> xy and then check that the result is + * within a small percentage of the original. + */ +static int +png_colorspace_check_xy(png_XYZ *XYZ, const png_xy *xy) +{ + int result; + png_xy xy_test; + + /* As a side-effect this routine also returns the XYZ endpoints. */ + result = png_XYZ_from_xy(XYZ, xy); + if (result != 0) + return result; + + result = png_xy_from_XYZ(&xy_test, XYZ); + if (result != 0) + return result; + + if (png_colorspace_endpoints_match(xy, &xy_test, + 5/*actually, the math is pretty accurate*/) != 0) + return 0; + + /* Too much slip */ + return 1; +} + +/* This is the check going the other way. The XYZ is modified to normalize it + * (another side-effect) and the xy chromaticities are returned. + */ +static int +png_colorspace_check_XYZ(png_xy *xy, png_XYZ *XYZ) +{ + int result; + png_XYZ XYZtemp; + + result = png_XYZ_normalize(XYZ); + if (result != 0) + return result; + + result = png_xy_from_XYZ(xy, XYZ); + if (result != 0) + return result; + + XYZtemp = *XYZ; + return png_colorspace_check_xy(&XYZtemp, xy); +} + +/* Used to check for an endpoint match against sRGB */ +static const png_xy sRGB_xy = /* From ITU-R BT.709-3 */ +{ + /* color x y */ + /* red */ 64000, 33000, + /* green */ 30000, 60000, + /* blue */ 15000, 6000, + /* white */ 31270, 32900 +}; + +static int +png_colorspace_set_xy_and_XYZ(png_const_structrp png_ptr, + png_colorspacerp colorspace, const png_xy *xy, const png_XYZ *XYZ, + int preferred) +{ + if ((colorspace->flags & PNG_COLORSPACE_INVALID) != 0) + return 0; + + /* The consistency check is performed on the chromaticities; this factors out + * variations because of the normalization (or not) of the end point Y + * values. + */ + if (preferred < 2 && + (colorspace->flags & PNG_COLORSPACE_HAVE_ENDPOINTS) != 0) + { + /* The end points must be reasonably close to any we already have. The + * following allows an error of up to +/-.001 + */ + if (png_colorspace_endpoints_match(xy, &colorspace->end_points_xy, + 100) == 0) + { + colorspace->flags |= PNG_COLORSPACE_INVALID; + png_benign_error(png_ptr, "inconsistent chromaticities"); + return 0; /* failed */ + } + + /* Only overwrite with preferred values */ + if (preferred == 0) + return 1; /* ok, but no change */ + } + + colorspace->end_points_xy = *xy; + colorspace->end_points_XYZ = *XYZ; + colorspace->flags |= PNG_COLORSPACE_HAVE_ENDPOINTS; + + /* The end points are normally quoted to two decimal digits, so allow +/-0.01 + * on this test. + */ + if (png_colorspace_endpoints_match(xy, &sRGB_xy, 1000) != 0) + colorspace->flags |= PNG_COLORSPACE_ENDPOINTS_MATCH_sRGB; + + else + colorspace->flags &= PNG_COLORSPACE_CANCEL( + PNG_COLORSPACE_ENDPOINTS_MATCH_sRGB); + + return 2; /* ok and changed */ +} + +int /* PRIVATE */ +png_colorspace_set_chromaticities(png_const_structrp png_ptr, + png_colorspacerp colorspace, const png_xy *xy, int preferred) +{ + /* We must check the end points to ensure they are reasonable - in the past + * color management systems have crashed as a result of getting bogus + * colorant values, while this isn't the fault of libpng it is the + * responsibility of libpng because PNG carries the bomb and libpng is in a + * position to protect against it. + */ + png_XYZ XYZ; + + switch (png_colorspace_check_xy(&XYZ, xy)) + { + case 0: /* success */ + return png_colorspace_set_xy_and_XYZ(png_ptr, colorspace, xy, &XYZ, + preferred); + + case 1: + /* We can't invert the chromaticities so we can't produce value XYZ + * values. Likely as not a color management system will fail too. + */ + colorspace->flags |= PNG_COLORSPACE_INVALID; + png_benign_error(png_ptr, "invalid chromaticities"); + break; + + default: + /* libpng is broken; this should be a warning but if it happens we + * want error reports so for the moment it is an error. + */ + colorspace->flags |= PNG_COLORSPACE_INVALID; + png_error(png_ptr, "internal error checking chromaticities"); + break; + } + + return 0; /* failed */ +} + +int /* PRIVATE */ +png_colorspace_set_endpoints(png_const_structrp png_ptr, + png_colorspacerp colorspace, const png_XYZ *XYZ_in, int preferred) +{ + png_XYZ XYZ = *XYZ_in; + png_xy xy; + + switch (png_colorspace_check_XYZ(&xy, &XYZ)) + { + case 0: + return png_colorspace_set_xy_and_XYZ(png_ptr, colorspace, &xy, &XYZ, + preferred); + + case 1: + /* End points are invalid. */ + colorspace->flags |= PNG_COLORSPACE_INVALID; + png_benign_error(png_ptr, "invalid end points"); + break; + + default: + colorspace->flags |= PNG_COLORSPACE_INVALID; + png_error(png_ptr, "internal error checking chromaticities"); + break; + } + + return 0; /* failed */ +} + +#if defined(PNG_sRGB_SUPPORTED) || defined(PNG_iCCP_SUPPORTED) +/* Error message generation */ +static char +png_icc_tag_char(png_uint_32 byte) +{ + byte &= 0xff; + if (byte >= 32 && byte <= 126) + return (char)byte; + else + return '?'; +} + +static void +png_icc_tag_name(char *name, png_uint_32 tag) +{ + name[0] = '\''; + name[1] = png_icc_tag_char(tag >> 24); + name[2] = png_icc_tag_char(tag >> 16); + name[3] = png_icc_tag_char(tag >> 8); + name[4] = png_icc_tag_char(tag ); + name[5] = '\''; +} + +static int +is_ICC_signature_char(png_alloc_size_t it) +{ + return it == 32 || (it >= 48 && it <= 57) || (it >= 65 && it <= 90) || + (it >= 97 && it <= 122); +} + +static int +is_ICC_signature(png_alloc_size_t it) +{ + return is_ICC_signature_char(it >> 24) /* checks all the top bits */ && + is_ICC_signature_char((it >> 16) & 0xff) && + is_ICC_signature_char((it >> 8) & 0xff) && + is_ICC_signature_char(it & 0xff); +} + +static int +png_icc_profile_error(png_const_structrp png_ptr, png_colorspacerp colorspace, + png_const_charp name, png_alloc_size_t value, png_const_charp reason) +{ + size_t pos; + char message[196]; /* see below for calculation */ + + if (colorspace != NULL) + colorspace->flags |= PNG_COLORSPACE_INVALID; + + pos = png_safecat(message, (sizeof message), 0, "profile '"); /* 9 chars */ + pos = png_safecat(message, pos+79, pos, name); /* Truncate to 79 chars */ + pos = png_safecat(message, (sizeof message), pos, "': "); /* +2 = 90 */ + if (is_ICC_signature(value) != 0) + { + /* So 'value' is at most 4 bytes and the following cast is safe */ + png_icc_tag_name(message+pos, (png_uint_32)value); + pos += 6; /* total +8; less than the else clause */ + message[pos++] = ':'; + message[pos++] = ' '; + } +# ifdef PNG_WARNINGS_SUPPORTED + else + { + char number[PNG_NUMBER_BUFFER_SIZE]; /* +24 = 114*/ + + pos = png_safecat(message, (sizeof message), pos, + png_format_number(number, number+(sizeof number), + PNG_NUMBER_FORMAT_x, value)); + pos = png_safecat(message, (sizeof message), pos, "h: "); /*+2 = 116*/ + } +# endif + /* The 'reason' is an arbitrary message, allow +79 maximum 195 */ + pos = png_safecat(message, (sizeof message), pos, reason); + PNG_UNUSED(pos) + + /* This is recoverable, but make it unconditionally an app_error on write to + * avoid writing invalid ICC profiles into PNG files (i.e., we handle them + * on read, with a warning, but on write unless the app turns off + * application errors the PNG won't be written.) + */ + png_chunk_report(png_ptr, message, + (colorspace != NULL) ? PNG_CHUNK_ERROR : PNG_CHUNK_WRITE_ERROR); + + return 0; +} +#endif /* sRGB || iCCP */ + +#ifdef PNG_sRGB_SUPPORTED +int /* PRIVATE */ +png_colorspace_set_sRGB(png_const_structrp png_ptr, png_colorspacerp colorspace, + int intent) +{ + /* sRGB sets known gamma, end points and (from the chunk) intent. */ + /* IMPORTANT: these are not necessarily the values found in an ICC profile + * because ICC profiles store values adapted to a D50 environment; it is + * expected that the ICC profile mediaWhitePointTag will be D50; see the + * checks and code elsewhere to understand this better. + * + * These XYZ values, which are accurate to 5dp, produce rgb to gray + * coefficients of (6968,23435,2366), which are reduced (because they add up + * to 32769 not 32768) to (6968,23434,2366). These are the values that + * libpng has traditionally used (and are the best values given the 15bit + * algorithm used by the rgb to gray code.) + */ + static const png_XYZ sRGB_XYZ = /* D65 XYZ (*not* the D50 adapted values!) */ + { + /* color X Y Z */ + /* red */ 41239, 21264, 1933, + /* green */ 35758, 71517, 11919, + /* blue */ 18048, 7219, 95053 + }; + + /* Do nothing if the colorspace is already invalidated. */ + if ((colorspace->flags & PNG_COLORSPACE_INVALID) != 0) + return 0; + + /* Check the intent, then check for existing settings. It is valid for the + * PNG file to have cHRM or gAMA chunks along with sRGB, but the values must + * be consistent with the correct values. If, however, this function is + * called below because an iCCP chunk matches sRGB then it is quite + * conceivable that an older app recorded incorrect gAMA and cHRM because of + * an incorrect calculation based on the values in the profile - this does + * *not* invalidate the profile (though it still produces an error, which can + * be ignored.) + */ + if (intent < 0 || intent >= PNG_sRGB_INTENT_LAST) + return png_icc_profile_error(png_ptr, colorspace, "sRGB", + (unsigned)intent, "invalid sRGB rendering intent"); + + if ((colorspace->flags & PNG_COLORSPACE_HAVE_INTENT) != 0 && + colorspace->rendering_intent != intent) + return png_icc_profile_error(png_ptr, colorspace, "sRGB", + (unsigned)intent, "inconsistent rendering intents"); + + if ((colorspace->flags & PNG_COLORSPACE_FROM_sRGB) != 0) + { + png_benign_error(png_ptr, "duplicate sRGB information ignored"); + return 0; + } + + /* If the standard sRGB cHRM chunk does not match the one from the PNG file + * warn but overwrite the value with the correct one. + */ + if ((colorspace->flags & PNG_COLORSPACE_HAVE_ENDPOINTS) != 0 && + !png_colorspace_endpoints_match(&sRGB_xy, &colorspace->end_points_xy, + 100)) + png_chunk_report(png_ptr, "cHRM chunk does not match sRGB", + PNG_CHUNK_ERROR); + + /* This check is just done for the error reporting - the routine always + * returns true when the 'from' argument corresponds to sRGB (2). + */ + (void)png_colorspace_check_gamma(png_ptr, colorspace, PNG_GAMMA_sRGB_INVERSE, + 2/*from sRGB*/); + + /* intent: bugs in GCC force 'int' to be used as the parameter type. */ + colorspace->rendering_intent = (png_uint_16)intent; + colorspace->flags |= PNG_COLORSPACE_HAVE_INTENT; + + /* endpoints */ + colorspace->end_points_xy = sRGB_xy; + colorspace->end_points_XYZ = sRGB_XYZ; + colorspace->flags |= + (PNG_COLORSPACE_HAVE_ENDPOINTS|PNG_COLORSPACE_ENDPOINTS_MATCH_sRGB); + + /* gamma */ + colorspace->gamma = PNG_GAMMA_sRGB_INVERSE; + colorspace->flags |= PNG_COLORSPACE_HAVE_GAMMA; + + /* Finally record that we have an sRGB profile */ + colorspace->flags |= + (PNG_COLORSPACE_MATCHES_sRGB|PNG_COLORSPACE_FROM_sRGB); + + return 1; /* set */ +} +#endif /* sRGB */ + +#ifdef PNG_iCCP_SUPPORTED +/* Encoded value of D50 as an ICC XYZNumber. From the ICC 2010 spec the value + * is XYZ(0.9642,1.0,0.8249), which scales to: + * + * (63189.8112, 65536, 54060.6464) + */ +static const png_byte D50_nCIEXYZ[12] = + { 0x00, 0x00, 0xf6, 0xd6, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0xd3, 0x2d }; + +int /* PRIVATE */ +png_icc_check_length(png_const_structrp png_ptr, png_colorspacerp colorspace, + png_const_charp name, png_uint_32 profile_length) +{ + if (profile_length < 132) + return png_icc_profile_error(png_ptr, colorspace, name, profile_length, + "too short"); + + return 1; +} + +int /* PRIVATE */ +png_icc_check_header(png_const_structrp png_ptr, png_colorspacerp colorspace, + png_const_charp name, png_uint_32 profile_length, + png_const_bytep profile/* first 132 bytes only */, int color_type) +{ + png_uint_32 temp; + + /* Length check; this cannot be ignored in this code because profile_length + * is used later to check the tag table, so even if the profile seems over + * long profile_length from the caller must be correct. The caller can fix + * this up on read or write by just passing in the profile header length. + */ + temp = png_get_uint_32(profile); + if (temp != profile_length) + return png_icc_profile_error(png_ptr, colorspace, name, temp, + "length does not match profile"); + + temp = (png_uint_32) (*(profile+8)); + if (temp > 3 && (profile_length & 3)) + return png_icc_profile_error(png_ptr, colorspace, name, profile_length, + "invalid length"); + + temp = png_get_uint_32(profile+128); /* tag count: 12 bytes/tag */ + if (temp > 357913930 || /* (2^32-4-132)/12: maximum possible tag count */ + profile_length < 132+12*temp) /* truncated tag table */ + return png_icc_profile_error(png_ptr, colorspace, name, temp, + "tag count too large"); + + /* The 'intent' must be valid or we can't store it, ICC limits the intent to + * 16 bits. + */ + temp = png_get_uint_32(profile+64); + if (temp >= 0xffff) /* The ICC limit */ + return png_icc_profile_error(png_ptr, colorspace, name, temp, + "invalid rendering intent"); + + /* This is just a warning because the profile may be valid in future + * versions. + */ + if (temp >= PNG_sRGB_INTENT_LAST) + (void)png_icc_profile_error(png_ptr, NULL, name, temp, + "intent outside defined range"); + + /* At this point the tag table can't be checked because it hasn't necessarily + * been loaded; however, various header fields can be checked. These checks + * are for values permitted by the PNG spec in an ICC profile; the PNG spec + * restricts the profiles that can be passed in an iCCP chunk (they must be + * appropriate to processing PNG data!) + */ + + /* Data checks (could be skipped). These checks must be independent of the + * version number; however, the version number doesn't accomodate changes in + * the header fields (just the known tags and the interpretation of the + * data.) + */ + temp = png_get_uint_32(profile+36); /* signature 'ascp' */ + if (temp != 0x61637370) + return png_icc_profile_error(png_ptr, colorspace, name, temp, + "invalid signature"); + + /* Currently the PCS illuminant/adopted white point (the computational + * white point) are required to be D50, + * however the profile contains a record of the illuminant so perhaps ICC + * expects to be able to change this in the future (despite the rationale in + * the introduction for using a fixed PCS adopted white.) Consequently the + * following is just a warning. + */ + if (memcmp(profile+68, D50_nCIEXYZ, 12) != 0) + (void)png_icc_profile_error(png_ptr, NULL, name, 0/*no tag value*/, + "PCS illuminant is not D50"); + + /* The PNG spec requires this: + * "If the iCCP chunk is present, the image samples conform to the colour + * space represented by the embedded ICC profile as defined by the + * International Color Consortium [ICC]. The colour space of the ICC profile + * shall be an RGB colour space for colour images (PNG colour types 2, 3, and + * 6), or a greyscale colour space for greyscale images (PNG colour types 0 + * and 4)." + * + * This checking code ensures the embedded profile (on either read or write) + * conforms to the specification requirements. Notice that an ICC 'gray' + * color-space profile contains the information to transform the monochrome + * data to XYZ or L*a*b (according to which PCS the profile uses) and this + * should be used in preference to the standard libpng K channel replication + * into R, G and B channels. + * + * Previously it was suggested that an RGB profile on grayscale data could be + * handled. However it it is clear that using an RGB profile in this context + * must be an error - there is no specification of what it means. Thus it is + * almost certainly more correct to ignore the profile. + */ + temp = png_get_uint_32(profile+16); /* data colour space field */ + switch (temp) + { + case 0x52474220: /* 'RGB ' */ + if ((color_type & PNG_COLOR_MASK_COLOR) == 0) + return png_icc_profile_error(png_ptr, colorspace, name, temp, + "RGB color space not permitted on grayscale PNG"); + break; + + case 0x47524159: /* 'GRAY' */ + if ((color_type & PNG_COLOR_MASK_COLOR) != 0) + return png_icc_profile_error(png_ptr, colorspace, name, temp, + "Gray color space not permitted on RGB PNG"); + break; + + default: + return png_icc_profile_error(png_ptr, colorspace, name, temp, + "invalid ICC profile color space"); + } + + /* It is up to the application to check that the profile class matches the + * application requirements; the spec provides no guidance, but it's pretty + * weird if the profile is not scanner ('scnr'), monitor ('mntr'), printer + * ('prtr') or 'spac' (for generic color spaces). Issue a warning in these + * cases. Issue an error for device link or abstract profiles - these don't + * contain the records necessary to transform the color-space to anything + * other than the target device (and not even that for an abstract profile). + * Profiles of these classes may not be embedded in images. + */ + temp = png_get_uint_32(profile+12); /* profile/device class */ + switch (temp) + { + case 0x73636E72: /* 'scnr' */ + case 0x6D6E7472: /* 'mntr' */ + case 0x70727472: /* 'prtr' */ + case 0x73706163: /* 'spac' */ + /* All supported */ + break; + + case 0x61627374: /* 'abst' */ + /* May not be embedded in an image */ + return png_icc_profile_error(png_ptr, colorspace, name, temp, + "invalid embedded Abstract ICC profile"); + + case 0x6C696E6B: /* 'link' */ + /* DeviceLink profiles cannot be interpreted in a non-device specific + * fashion, if an app uses the AToB0Tag in the profile the results are + * undefined unless the result is sent to the intended device, + * therefore a DeviceLink profile should not be found embedded in a + * PNG. + */ + return png_icc_profile_error(png_ptr, colorspace, name, temp, + "unexpected DeviceLink ICC profile class"); + + case 0x6E6D636C: /* 'nmcl' */ + /* A NamedColor profile is also device specific, however it doesn't + * contain an AToB0 tag that is open to misinterpretation. Almost + * certainly it will fail the tests below. + */ + (void)png_icc_profile_error(png_ptr, NULL, name, temp, + "unexpected NamedColor ICC profile class"); + break; + + default: + /* To allow for future enhancements to the profile accept unrecognized + * profile classes with a warning, these then hit the test below on the + * tag content to ensure they are backward compatible with one of the + * understood profiles. + */ + (void)png_icc_profile_error(png_ptr, NULL, name, temp, + "unrecognized ICC profile class"); + break; + } + + /* For any profile other than a device link one the PCS must be encoded + * either in XYZ or Lab. + */ + temp = png_get_uint_32(profile+20); + switch (temp) + { + case 0x58595A20: /* 'XYZ ' */ + case 0x4C616220: /* 'Lab ' */ + break; + + default: + return png_icc_profile_error(png_ptr, colorspace, name, temp, + "unexpected ICC PCS encoding"); + } + + return 1; +} + +int /* PRIVATE */ +png_icc_check_tag_table(png_const_structrp png_ptr, png_colorspacerp colorspace, + png_const_charp name, png_uint_32 profile_length, + png_const_bytep profile /* header plus whole tag table */) +{ + png_uint_32 tag_count = png_get_uint_32(profile+128); + png_uint_32 itag; + png_const_bytep tag = profile+132; /* The first tag */ + + /* First scan all the tags in the table and add bits to the icc_info value + * (temporarily in 'tags'). + */ + for (itag=0; itag < tag_count; ++itag, tag += 12) + { + png_uint_32 tag_id = png_get_uint_32(tag+0); + png_uint_32 tag_start = png_get_uint_32(tag+4); /* must be aligned */ + png_uint_32 tag_length = png_get_uint_32(tag+8);/* not padded */ + + /* The ICC specification does not exclude zero length tags, therefore the + * start might actually be anywhere if there is no data, but this would be + * a clear abuse of the intent of the standard so the start is checked for + * being in range. All defined tag types have an 8 byte header - a 4 byte + * type signature then 0. + */ + if ((tag_start & 3) != 0) + { + /* CNHP730S.icc shipped with Microsoft Windows 64 violates this, it is + * only a warning here because libpng does not care about the + * alignment. + */ + (void)png_icc_profile_error(png_ptr, NULL, name, tag_id, + "ICC profile tag start not a multiple of 4"); + } + + /* This is a hard error; potentially it can cause read outside the + * profile. + */ + if (tag_start > profile_length || tag_length > profile_length - tag_start) + return png_icc_profile_error(png_ptr, colorspace, name, tag_id, + "ICC profile tag outside profile"); + } + + return 1; /* success, maybe with warnings */ +} + +#if defined(PNG_sRGB_SUPPORTED) && PNG_sRGB_PROFILE_CHECKS >= 0 +/* Information about the known ICC sRGB profiles */ +static const struct +{ + png_uint_32 adler, crc, length; + png_uint_32 md5[4]; + png_byte have_md5; + png_byte is_broken; + png_uint_16 intent; + +# define PNG_MD5(a,b,c,d) { a, b, c, d }, (a!=0)||(b!=0)||(c!=0)||(d!=0) +# define PNG_ICC_CHECKSUM(adler, crc, md5, intent, broke, date, length, fname)\ + { adler, crc, length, md5, broke, intent }, + +} png_sRGB_checks[] = +{ + /* This data comes from contrib/tools/checksum-icc run on downloads of + * all four ICC sRGB profiles from www.color.org. + */ + /* adler32, crc32, MD5[4], intent, date, length, file-name */ + PNG_ICC_CHECKSUM(0x0a3fd9f6, 0x3b8772b9, + PNG_MD5(0x29f83dde, 0xaff255ae, 0x7842fae4, 0xca83390d), 0, 0, + "2009/03/27 21:36:31", 3048, "sRGB_IEC61966-2-1_black_scaled.icc") + + /* ICC sRGB v2 perceptual no black-compensation: */ + PNG_ICC_CHECKSUM(0x4909e5e1, 0x427ebb21, + PNG_MD5(0xc95bd637, 0xe95d8a3b, 0x0df38f99, 0xc1320389), 1, 0, + "2009/03/27 21:37:45", 3052, "sRGB_IEC61966-2-1_no_black_scaling.icc") + + PNG_ICC_CHECKSUM(0xfd2144a1, 0x306fd8ae, + PNG_MD5(0xfc663378, 0x37e2886b, 0xfd72e983, 0x8228f1b8), 0, 0, + "2009/08/10 17:28:01", 60988, "sRGB_v4_ICC_preference_displayclass.icc") + + /* ICC sRGB v4 perceptual */ + PNG_ICC_CHECKSUM(0x209c35d2, 0xbbef7812, + PNG_MD5(0x34562abf, 0x994ccd06, 0x6d2c5721, 0xd0d68c5d), 0, 0, + "2007/07/25 00:05:37", 60960, "sRGB_v4_ICC_preference.icc") + + /* The following profiles have no known MD5 checksum. If there is a match + * on the (empty) MD5 the other fields are used to attempt a match and + * a warning is produced. The first two of these profiles have a 'cprt' tag + * which suggests that they were also made by Hewlett Packard. + */ + PNG_ICC_CHECKSUM(0xa054d762, 0x5d5129ce, + PNG_MD5(0x00000000, 0x00000000, 0x00000000, 0x00000000), 1, 0, + "2004/07/21 18:57:42", 3024, "sRGB_IEC61966-2-1_noBPC.icc") + + /* This is a 'mntr' (display) profile with a mediaWhitePointTag that does not + * match the D50 PCS illuminant in the header (it is in fact the D65 values, + * so the white point is recorded as the un-adapted value.) The profiles + * below only differ in one byte - the intent - and are basically the same as + * the previous profile except for the mediaWhitePointTag error and a missing + * chromaticAdaptationTag. + */ + PNG_ICC_CHECKSUM(0xf784f3fb, 0x182ea552, + PNG_MD5(0x00000000, 0x00000000, 0x00000000, 0x00000000), 0, 1/*broken*/, + "1998/02/09 06:49:00", 3144, "HP-Microsoft sRGB v2 perceptual") + + PNG_ICC_CHECKSUM(0x0398f3fc, 0xf29e526d, + PNG_MD5(0x00000000, 0x00000000, 0x00000000, 0x00000000), 1, 1/*broken*/, + "1998/02/09 06:49:00", 3144, "HP-Microsoft sRGB v2 media-relative") +}; + +static int +png_compare_ICC_profile_with_sRGB(png_const_structrp png_ptr, + png_const_bytep profile, uLong adler) +{ + /* The quick check is to verify just the MD5 signature and trust the + * rest of the data. Because the profile has already been verified for + * correctness this is safe. png_colorspace_set_sRGB will check the 'intent' + * field too, so if the profile has been edited with an intent not defined + * by sRGB (but maybe defined by a later ICC specification) the read of + * the profile will fail at that point. + */ + + png_uint_32 length = 0; + png_uint_32 intent = 0x10000; /* invalid */ +#if PNG_sRGB_PROFILE_CHECKS > 1 + uLong crc = 0; /* the value for 0 length data */ +#endif + unsigned int i; + +#ifdef PNG_SET_OPTION_SUPPORTED + /* First see if PNG_SKIP_sRGB_CHECK_PROFILE has been set to "on" */ + if (((png_ptr->options >> PNG_SKIP_sRGB_CHECK_PROFILE) & 3) == + PNG_OPTION_ON) + return 0; +#endif + + for (i=0; i < (sizeof png_sRGB_checks) / (sizeof png_sRGB_checks[0]); ++i) + { + if (png_get_uint_32(profile+84) == png_sRGB_checks[i].md5[0] && + png_get_uint_32(profile+88) == png_sRGB_checks[i].md5[1] && + png_get_uint_32(profile+92) == png_sRGB_checks[i].md5[2] && + png_get_uint_32(profile+96) == png_sRGB_checks[i].md5[3]) + { + /* This may be one of the old HP profiles without an MD5, in that + * case we can only use the length and Adler32 (note that these + * are not used by default if there is an MD5!) + */ +# if PNG_sRGB_PROFILE_CHECKS == 0 + if (png_sRGB_checks[i].have_md5 != 0) + return 1+png_sRGB_checks[i].is_broken; +# endif + + /* Profile is unsigned or more checks have been configured in. */ + if (length == 0) + { + length = png_get_uint_32(profile); + intent = png_get_uint_32(profile+64); + } + + /* Length *and* intent must match */ + if (length == png_sRGB_checks[i].length && + intent == png_sRGB_checks[i].intent) + { + /* Now calculate the adler32 if not done already. */ + if (adler == 0) + { + adler = adler32(0, NULL, 0); + adler = adler32(adler, profile, length); + } + + if (adler == png_sRGB_checks[i].adler) + { + /* These basic checks suggest that the data has not been + * modified, but if the check level is more than 1 perform + * our own crc32 checksum on the data. + */ +# if PNG_sRGB_PROFILE_CHECKS > 1 + if (crc == 0) + { + crc = crc32(0, NULL, 0); + crc = crc32(crc, profile, length); + } + + /* So this check must pass for the 'return' below to happen. + */ + if (crc == png_sRGB_checks[i].crc) +# endif + { + if (png_sRGB_checks[i].is_broken != 0) + { + /* These profiles are known to have bad data that may cause + * problems if they are used, therefore attempt to + * discourage their use, skip the 'have_md5' warning below, + * which is made irrelevant by this error. + */ + png_chunk_report(png_ptr, "known incorrect sRGB profile", + PNG_CHUNK_ERROR); + } + + /* Warn that this being done; this isn't even an error since + * the profile is perfectly valid, but it would be nice if + * people used the up-to-date ones. + */ + else if (png_sRGB_checks[i].have_md5 == 0) + { + png_chunk_report(png_ptr, "out-of-date sRGB profile with" + " no signature", + PNG_CHUNK_WARNING); + } + + return 1+png_sRGB_checks[i].is_broken; + } + } + +# if PNG_sRGB_PROFILE_CHECKS > 0 + /* The signature matched, but the profile had been changed in some + * way. This probably indicates a data error or uninformed hacking. + * Fall through to "no match". + */ + png_chunk_report(png_ptr, "Not recognizing known sRGB profile that" + " has been edited", + PNG_CHUNK_WARNING); + break; +# endif + } + } + } + + return 0; /* no match */ +} +#endif + +#ifdef PNG_sRGB_SUPPORTED +void /* PRIVATE */ +png_icc_set_sRGB(png_const_structrp png_ptr, + png_colorspacerp colorspace, png_const_bytep profile, uLong adler) +{ + /* Is this profile one of the known ICC sRGB profiles? If it is, just set + * the sRGB information. + */ +#if PNG_sRGB_PROFILE_CHECKS >= 0 + if (png_compare_ICC_profile_with_sRGB(png_ptr, profile, adler) != 0) +#endif + (void)png_colorspace_set_sRGB(png_ptr, colorspace, + (int)/*already checked*/png_get_uint_32(profile+64)); +} +#endif /* READ_sRGB */ + +int /* PRIVATE */ +png_colorspace_set_ICC(png_const_structrp png_ptr, png_colorspacerp colorspace, + png_const_charp name, png_uint_32 profile_length, png_const_bytep profile, + int color_type) +{ + if ((colorspace->flags & PNG_COLORSPACE_INVALID) != 0) + return 0; + + if (png_icc_check_length(png_ptr, colorspace, name, profile_length) != 0 && + png_icc_check_header(png_ptr, colorspace, name, profile_length, profile, + color_type) != 0 && + png_icc_check_tag_table(png_ptr, colorspace, name, profile_length, + profile) != 0) + { +# ifdef PNG_sRGB_SUPPORTED + /* If no sRGB support, don't try storing sRGB information */ + png_icc_set_sRGB(png_ptr, colorspace, profile, 0); +# endif + return 1; + } + + /* Failure case */ + return 0; +} +#endif /* iCCP */ + +#ifdef PNG_READ_RGB_TO_GRAY_SUPPORTED +void /* PRIVATE */ +png_colorspace_set_rgb_coefficients(png_structrp png_ptr) +{ + /* Set the rgb_to_gray coefficients from the colorspace. */ + if (png_ptr->rgb_to_gray_coefficients_set == 0 && + (png_ptr->colorspace.flags & PNG_COLORSPACE_HAVE_ENDPOINTS) != 0) + { + /* png_set_background has not been called, get the coefficients from the Y + * values of the colorspace colorants. + */ + png_fixed_point r = png_ptr->colorspace.end_points_XYZ.red_Y; + png_fixed_point g = png_ptr->colorspace.end_points_XYZ.green_Y; + png_fixed_point b = png_ptr->colorspace.end_points_XYZ.blue_Y; + png_fixed_point total = r+g+b; + + if (total > 0 && + r >= 0 && png_muldiv(&r, r, 32768, total) && r >= 0 && r <= 32768 && + g >= 0 && png_muldiv(&g, g, 32768, total) && g >= 0 && g <= 32768 && + b >= 0 && png_muldiv(&b, b, 32768, total) && b >= 0 && b <= 32768 && + r+g+b <= 32769) + { + /* We allow 0 coefficients here. r+g+b may be 32769 if two or + * all of the coefficients were rounded up. Handle this by + * reducing the *largest* coefficient by 1; this matches the + * approach used for the default coefficients in pngrtran.c + */ + int add = 0; + + if (r+g+b > 32768) + add = -1; + else if (r+g+b < 32768) + add = 1; + + if (add != 0) + { + if (g >= r && g >= b) + g += add; + else if (r >= g && r >= b) + r += add; + else + b += add; + } + + /* Check for an internal error. */ + if (r+g+b != 32768) + png_error(png_ptr, + "internal error handling cHRM coefficients"); + + else + { + png_ptr->rgb_to_gray_red_coeff = (png_uint_16)r; + png_ptr->rgb_to_gray_green_coeff = (png_uint_16)g; + } + } + + /* This is a png_error at present even though it could be ignored - + * it should never happen, but it is important that if it does, the + * bug is fixed. + */ + else + png_error(png_ptr, "internal error handling cHRM->XYZ"); + } +} +#endif + +#endif /* COLORSPACE */ + +#ifdef __GNUC__ +/* This exists solely to work round a warning from GNU C. */ +static int /* PRIVATE */ +png_gt(size_t a, size_t b) +{ + return a > b; +} +#else +# define png_gt(a,b) ((a) > (b)) +#endif + +void /* PRIVATE */ +png_check_IHDR(png_const_structrp png_ptr, + png_uint_32 width, png_uint_32 height, int bit_depth, + int color_type, int interlace_type, int compression_type, + int filter_type) +{ + int error = 0; + + /* Check for width and height valid values */ + if (width == 0) + { + png_warning(png_ptr, "Image width is zero in IHDR"); + error = 1; + } + else if (width > PNG_UINT_31_MAX) + { + png_warning(png_ptr, "Invalid image width in IHDR"); + error = 1; + } + + else if (png_gt(width, + (PNG_SIZE_MAX >> 3) /* 8-byte RGBA pixels */ + - 48 /* big_row_buf hack */ + - 1 /* filter byte */ + - 7*8 /* rounding width to multiple of 8 pix */ + - 8)) /* extra max_pixel_depth pad */ + { + /* The size of the row must be within the limits of this architecture. + * Because the read code can perform arbitrary transformations the + * maximum size is checked here. Because the code in png_read_start_row + * adds extra space "for safety's sake" in several places a conservative + * limit is used here. + * + * NOTE: it would be far better to check the size that is actually used, + * but the effect in the real world is minor and the changes are more + * extensive, therefore much more dangerous and much more difficult to + * write in a way that avoids compiler warnings. + */ + png_warning(png_ptr, "Image width is too large for this architecture"); + error = 1; + } + else + { +# ifdef PNG_SET_USER_LIMITS_SUPPORTED + if (width > png_ptr->user_width_max) +# else + if (width > PNG_USER_WIDTH_MAX) +# endif + { + png_warning(png_ptr, "Image width exceeds user limit in IHDR"); + error = 1; + } + } + + if (height == 0) + { + png_warning(png_ptr, "Image height is zero in IHDR"); + error = 1; + } + else if (height > PNG_UINT_31_MAX) + { + png_warning(png_ptr, "Invalid image height in IHDR"); + error = 1; + } + else + { +# ifdef PNG_SET_USER_LIMITS_SUPPORTED + if (height > png_ptr->user_height_max) +# else + if (height > PNG_USER_HEIGHT_MAX) +# endif + { + png_warning(png_ptr, "Image height exceeds user limit in IHDR"); + error = 1; + } + } + + /* Check other values */ + if (bit_depth != 1 && bit_depth != 2 && bit_depth != 4 && + bit_depth != 8 && bit_depth != 16) + { + png_warning(png_ptr, "Invalid bit depth in IHDR"); + error = 1; + } + + if (color_type < 0 || color_type == 1 || + color_type == 5 || color_type > 6) + { + png_warning(png_ptr, "Invalid color type in IHDR"); + error = 1; + } + + if (((color_type == PNG_COLOR_TYPE_PALETTE) && bit_depth > 8) || + ((color_type == PNG_COLOR_TYPE_RGB || + color_type == PNG_COLOR_TYPE_GRAY_ALPHA || + color_type == PNG_COLOR_TYPE_RGB_ALPHA) && bit_depth < 8)) + { + png_warning(png_ptr, "Invalid color type/bit depth combination in IHDR"); + error = 1; + } + + if (interlace_type >= PNG_INTERLACE_LAST) + { + png_warning(png_ptr, "Unknown interlace method in IHDR"); + error = 1; + } + + if (compression_type != PNG_COMPRESSION_TYPE_BASE) + { + png_warning(png_ptr, "Unknown compression method in IHDR"); + error = 1; + } + +# ifdef PNG_MNG_FEATURES_SUPPORTED + /* Accept filter_method 64 (intrapixel differencing) only if + * 1. Libpng was compiled with PNG_MNG_FEATURES_SUPPORTED and + * 2. Libpng did not read a PNG signature (this filter_method is only + * used in PNG datastreams that are embedded in MNG datastreams) and + * 3. The application called png_permit_mng_features with a mask that + * included PNG_FLAG_MNG_FILTER_64 and + * 4. The filter_method is 64 and + * 5. The color_type is RGB or RGBA + */ + if ((png_ptr->mode & PNG_HAVE_PNG_SIGNATURE) != 0 && + png_ptr->mng_features_permitted != 0) + png_warning(png_ptr, "MNG features are not allowed in a PNG datastream"); + + if (filter_type != PNG_FILTER_TYPE_BASE) + { + if (!((png_ptr->mng_features_permitted & PNG_FLAG_MNG_FILTER_64) != 0 && + (filter_type == PNG_INTRAPIXEL_DIFFERENCING) && + ((png_ptr->mode & PNG_HAVE_PNG_SIGNATURE) == 0) && + (color_type == PNG_COLOR_TYPE_RGB || + color_type == PNG_COLOR_TYPE_RGB_ALPHA))) + { + png_warning(png_ptr, "Unknown filter method in IHDR"); + error = 1; + } + + if ((png_ptr->mode & PNG_HAVE_PNG_SIGNATURE) != 0) + { + png_warning(png_ptr, "Invalid filter method in IHDR"); + error = 1; + } + } + +# else + if (filter_type != PNG_FILTER_TYPE_BASE) + { + png_warning(png_ptr, "Unknown filter method in IHDR"); + error = 1; + } +# endif + + if (error == 1) + png_error(png_ptr, "Invalid IHDR data"); +} + +#if defined(PNG_sCAL_SUPPORTED) || defined(PNG_pCAL_SUPPORTED) +/* ASCII to fp functions */ +/* Check an ASCII formated floating point value, see the more detailed + * comments in pngpriv.h + */ +/* The following is used internally to preserve the sticky flags */ +#define png_fp_add(state, flags) ((state) |= (flags)) +#define png_fp_set(state, value) ((state) = (value) | ((state) & PNG_FP_STICKY)) + +int /* PRIVATE */ +png_check_fp_number(png_const_charp string, png_size_t size, int *statep, + png_size_tp whereami) +{ + int state = *statep; + png_size_t i = *whereami; + + while (i < size) + { + int type; + /* First find the type of the next character */ + switch (string[i]) + { + case 43: type = PNG_FP_SAW_SIGN; break; + case 45: type = PNG_FP_SAW_SIGN + PNG_FP_NEGATIVE; break; + case 46: type = PNG_FP_SAW_DOT; break; + case 48: type = PNG_FP_SAW_DIGIT; break; + case 49: case 50: case 51: case 52: + case 53: case 54: case 55: case 56: + case 57: type = PNG_FP_SAW_DIGIT + PNG_FP_NONZERO; break; + case 69: + case 101: type = PNG_FP_SAW_E; break; + default: goto PNG_FP_End; + } + + /* Now deal with this type according to the current + * state, the type is arranged to not overlap the + * bits of the PNG_FP_STATE. + */ + switch ((state & PNG_FP_STATE) + (type & PNG_FP_SAW_ANY)) + { + case PNG_FP_INTEGER + PNG_FP_SAW_SIGN: + if ((state & PNG_FP_SAW_ANY) != 0) + goto PNG_FP_End; /* not a part of the number */ + + png_fp_add(state, type); + break; + + case PNG_FP_INTEGER + PNG_FP_SAW_DOT: + /* Ok as trailer, ok as lead of fraction. */ + if ((state & PNG_FP_SAW_DOT) != 0) /* two dots */ + goto PNG_FP_End; + + else if ((state & PNG_FP_SAW_DIGIT) != 0) /* trailing dot? */ + png_fp_add(state, type); + + else + png_fp_set(state, PNG_FP_FRACTION | type); + + break; + + case PNG_FP_INTEGER + PNG_FP_SAW_DIGIT: + if ((state & PNG_FP_SAW_DOT) != 0) /* delayed fraction */ + png_fp_set(state, PNG_FP_FRACTION | PNG_FP_SAW_DOT); + + png_fp_add(state, type | PNG_FP_WAS_VALID); + + break; + + case PNG_FP_INTEGER + PNG_FP_SAW_E: + if ((state & PNG_FP_SAW_DIGIT) == 0) + goto PNG_FP_End; + + png_fp_set(state, PNG_FP_EXPONENT); + + break; + + /* case PNG_FP_FRACTION + PNG_FP_SAW_SIGN: + goto PNG_FP_End; ** no sign in fraction */ + + /* case PNG_FP_FRACTION + PNG_FP_SAW_DOT: + goto PNG_FP_End; ** Because SAW_DOT is always set */ + + case PNG_FP_FRACTION + PNG_FP_SAW_DIGIT: + png_fp_add(state, type | PNG_FP_WAS_VALID); + break; + + case PNG_FP_FRACTION + PNG_FP_SAW_E: + /* This is correct because the trailing '.' on an + * integer is handled above - so we can only get here + * with the sequence ".E" (with no preceding digits). + */ + if ((state & PNG_FP_SAW_DIGIT) == 0) + goto PNG_FP_End; + + png_fp_set(state, PNG_FP_EXPONENT); + + break; + + case PNG_FP_EXPONENT + PNG_FP_SAW_SIGN: + if ((state & PNG_FP_SAW_ANY) != 0) + goto PNG_FP_End; /* not a part of the number */ + + png_fp_add(state, PNG_FP_SAW_SIGN); + + break; + + /* case PNG_FP_EXPONENT + PNG_FP_SAW_DOT: + goto PNG_FP_End; */ + + case PNG_FP_EXPONENT + PNG_FP_SAW_DIGIT: + png_fp_add(state, PNG_FP_SAW_DIGIT | PNG_FP_WAS_VALID); + + break; + + /* case PNG_FP_EXPONEXT + PNG_FP_SAW_E: + goto PNG_FP_End; */ + + default: goto PNG_FP_End; /* I.e. break 2 */ + } + + /* The character seems ok, continue. */ + ++i; + } + +PNG_FP_End: + /* Here at the end, update the state and return the correct + * return code. + */ + *statep = state; + *whereami = i; + + return (state & PNG_FP_SAW_DIGIT) != 0; +} + + +/* The same but for a complete string. */ +int +png_check_fp_string(png_const_charp string, png_size_t size) +{ + int state=0; + png_size_t char_index=0; + + if (png_check_fp_number(string, size, &state, &char_index) != 0 && + (char_index == size || string[char_index] == 0)) + return state /* must be non-zero - see above */; + + return 0; /* i.e. fail */ +} +#endif /* pCAL || sCAL */ + +#ifdef PNG_sCAL_SUPPORTED +# ifdef PNG_FLOATING_POINT_SUPPORTED +/* Utility used below - a simple accurate power of ten from an integral + * exponent. + */ +static double +png_pow10(int power) +{ + int recip = 0; + double d = 1; + + /* Handle negative exponent with a reciprocal at the end because + * 10 is exact whereas .1 is inexact in base 2 + */ + if (power < 0) + { + if (power < DBL_MIN_10_EXP) return 0; + recip = 1, power = -power; + } + + if (power > 0) + { + /* Decompose power bitwise. */ + double mult = 10; + do + { + if (power & 1) d *= mult; + mult *= mult; + power >>= 1; + } + while (power > 0); + + if (recip != 0) d = 1/d; + } + /* else power is 0 and d is 1 */ + + return d; +} + +/* Function to format a floating point value in ASCII with a given + * precision. + */ +void /* PRIVATE */ +png_ascii_from_fp(png_const_structrp png_ptr, png_charp ascii, png_size_t size, + double fp, unsigned int precision) +{ + /* We use standard functions from math.h, but not printf because + * that would require stdio. The caller must supply a buffer of + * sufficient size or we will png_error. The tests on size and + * the space in ascii[] consumed are indicated below. + */ + if (precision < 1) + precision = DBL_DIG; + + /* Enforce the limit of the implementation precision too. */ + if (precision > DBL_DIG+1) + precision = DBL_DIG+1; + + /* Basic sanity checks */ + if (size >= precision+5) /* See the requirements below. */ + { + if (fp < 0) + { + fp = -fp; + *ascii++ = 45; /* '-' PLUS 1 TOTAL 1 */ + --size; + } + + if (fp >= DBL_MIN && fp <= DBL_MAX) + { + int exp_b10; /* A base 10 exponent */ + double base; /* 10^exp_b10 */ + + /* First extract a base 10 exponent of the number, + * the calculation below rounds down when converting + * from base 2 to base 10 (multiply by log10(2) - + * 0.3010, but 77/256 is 0.3008, so exp_b10 needs to + * be increased. Note that the arithmetic shift + * performs a floor() unlike C arithmetic - using a + * C multiply would break the following for negative + * exponents. + */ + (void)frexp(fp, &exp_b10); /* exponent to base 2 */ + + exp_b10 = (exp_b10 * 77) >> 8; /* <= exponent to base 10 */ + + /* Avoid underflow here. */ + base = png_pow10(exp_b10); /* May underflow */ + + while (base < DBL_MIN || base < fp) + { + /* And this may overflow. */ + double test = png_pow10(exp_b10+1); + + if (test <= DBL_MAX) + ++exp_b10, base = test; + + else + break; + } + + /* Normalize fp and correct exp_b10, after this fp is in the + * range [.1,1) and exp_b10 is both the exponent and the digit + * *before* which the decimal point should be inserted + * (starting with 0 for the first digit). Note that this + * works even if 10^exp_b10 is out of range because of the + * test on DBL_MAX above. + */ + fp /= base; + while (fp >= 1) fp /= 10, ++exp_b10; + + /* Because of the code above fp may, at this point, be + * less than .1, this is ok because the code below can + * handle the leading zeros this generates, so no attempt + * is made to correct that here. + */ + + { + int czero, clead, cdigits; + char exponent[10]; + + /* Allow up to two leading zeros - this will not lengthen + * the number compared to using E-n. + */ + if (exp_b10 < 0 && exp_b10 > -3) /* PLUS 3 TOTAL 4 */ + { + czero = -exp_b10; /* PLUS 2 digits: TOTAL 3 */ + exp_b10 = 0; /* Dot added below before first output. */ + } + else + czero = 0; /* No zeros to add */ + + /* Generate the digit list, stripping trailing zeros and + * inserting a '.' before a digit if the exponent is 0. + */ + clead = czero; /* Count of leading zeros */ + cdigits = 0; /* Count of digits in list. */ + + do + { + double d; + + fp *= 10; + /* Use modf here, not floor and subtract, so that + * the separation is done in one step. At the end + * of the loop don't break the number into parts so + * that the final digit is rounded. + */ + if (cdigits+czero-clead+1 < (int)precision) + fp = modf(fp, &d); + + else + { + d = floor(fp + .5); + + if (d > 9) + { + /* Rounding up to 10, handle that here. */ + if (czero > 0) + { + --czero, d = 1; + if (cdigits == 0) --clead; + } + else + { + while (cdigits > 0 && d > 9) + { + int ch = *--ascii; + + if (exp_b10 != (-1)) + ++exp_b10; + + else if (ch == 46) + { + ch = *--ascii, ++size; + /* Advance exp_b10 to '1', so that the + * decimal point happens after the + * previous digit. + */ + exp_b10 = 1; + } + + --cdigits; + d = ch - 47; /* I.e. 1+(ch-48) */ + } + + /* Did we reach the beginning? If so adjust the + * exponent but take into account the leading + * decimal point. + */ + if (d > 9) /* cdigits == 0 */ + { + if (exp_b10 == (-1)) + { + /* Leading decimal point (plus zeros?), if + * we lose the decimal point here it must + * be reentered below. + */ + int ch = *--ascii; + + if (ch == 46) + ++size, exp_b10 = 1; + + /* Else lost a leading zero, so 'exp_b10' is + * still ok at (-1) + */ + } + else + ++exp_b10; + + /* In all cases we output a '1' */ + d = 1; + } + } + } + fp = 0; /* Guarantees termination below. */ + } + + if (d == 0) + { + ++czero; + if (cdigits == 0) ++clead; + } + else + { + /* Included embedded zeros in the digit count. */ + cdigits += czero - clead; + clead = 0; + + while (czero > 0) + { + /* exp_b10 == (-1) means we just output the decimal + * place - after the DP don't adjust 'exp_b10' any + * more! + */ + if (exp_b10 != (-1)) + { + if (exp_b10 == 0) *ascii++ = 46, --size; + /* PLUS 1: TOTAL 4 */ + --exp_b10; + } + *ascii++ = 48, --czero; + } + + if (exp_b10 != (-1)) + { + if (exp_b10 == 0) + *ascii++ = 46, --size; /* counted above */ + + --exp_b10; + } + *ascii++ = (char)(48 + (int)d), ++cdigits; + } + } + while (cdigits+czero-clead < (int)precision && fp > DBL_MIN); + + /* The total output count (max) is now 4+precision */ + + /* Check for an exponent, if we don't need one we are + * done and just need to terminate the string. At + * this point exp_b10==(-1) is effectively if flag - it got + * to '-1' because of the decrement after outputing + * the decimal point above (the exponent required is + * *not* -1!) + */ + if (exp_b10 >= (-1) && exp_b10 <= 2) + { + /* The following only happens if we didn't output the + * leading zeros above for negative exponent, so this + * doest add to the digit requirement. Note that the + * two zeros here can only be output if the two leading + * zeros were *not* output, so this doesn't increase + * the output count. + */ + while (--exp_b10 >= 0) *ascii++ = 48; + + *ascii = 0; + + /* Total buffer requirement (including the '\0') is + * 5+precision - see check at the start. + */ + return; + } + + /* Here if an exponent is required, adjust size for + * the digits we output but did not count. The total + * digit output here so far is at most 1+precision - no + * decimal point and no leading or trailing zeros have + * been output. + */ + size -= cdigits; + + *ascii++ = 69, --size; /* 'E': PLUS 1 TOTAL 2+precision */ + + /* The following use of an unsigned temporary avoids ambiguities in + * the signed arithmetic on exp_b10 and permits GCC at least to do + * better optimization. + */ + { + unsigned int uexp_b10; + + if (exp_b10 < 0) + { + *ascii++ = 45, --size; /* '-': PLUS 1 TOTAL 3+precision */ + uexp_b10 = -exp_b10; + } + + else + uexp_b10 = exp_b10; + + cdigits = 0; + + while (uexp_b10 > 0) + { + exponent[cdigits++] = (char)(48 + uexp_b10 % 10); + uexp_b10 /= 10; + } + } + + /* Need another size check here for the exponent digits, so + * this need not be considered above. + */ + if ((int)size > cdigits) + { + while (cdigits > 0) *ascii++ = exponent[--cdigits]; + + *ascii = 0; + + return; + } + } + } + else if (!(fp >= DBL_MIN)) + { + *ascii++ = 48; /* '0' */ + *ascii = 0; + return; + } + else + { + *ascii++ = 105; /* 'i' */ + *ascii++ = 110; /* 'n' */ + *ascii++ = 102; /* 'f' */ + *ascii = 0; + return; + } + } + + /* Here on buffer too small. */ + png_error(png_ptr, "ASCII conversion buffer too small"); +} + +# endif /* FLOATING_POINT */ + +# ifdef PNG_FIXED_POINT_SUPPORTED +/* Function to format a fixed point value in ASCII. + */ +void /* PRIVATE */ +png_ascii_from_fixed(png_const_structrp png_ptr, png_charp ascii, + png_size_t size, png_fixed_point fp) +{ + /* Require space for 10 decimal digits, a decimal point, a minus sign and a + * trailing \0, 13 characters: + */ + if (size > 12) + { + png_uint_32 num; + + /* Avoid overflow here on the minimum integer. */ + if (fp < 0) + *ascii++ = 45, --size, num = -fp; + else + num = fp; + + if (num <= 0x80000000) /* else overflowed */ + { + unsigned int ndigits = 0, first = 16 /* flag value */; + char digits[10]; + + while (num) + { + /* Split the low digit off num: */ + unsigned int tmp = num/10; + num -= tmp*10; + digits[ndigits++] = (char)(48 + num); + /* Record the first non-zero digit, note that this is a number + * starting at 1, it's not actually the array index. + */ + if (first == 16 && num > 0) + first = ndigits; + num = tmp; + } + + if (ndigits > 0) + { + while (ndigits > 5) *ascii++ = digits[--ndigits]; + /* The remaining digits are fractional digits, ndigits is '5' or + * smaller at this point. It is certainly not zero. Check for a + * non-zero fractional digit: + */ + if (first <= 5) + { + unsigned int i; + *ascii++ = 46; /* decimal point */ + /* ndigits may be <5 for small numbers, output leading zeros + * then ndigits digits to first: + */ + i = 5; + while (ndigits < i) *ascii++ = 48, --i; + while (ndigits >= first) *ascii++ = digits[--ndigits]; + /* Don't output the trailing zeros! */ + } + } + else + *ascii++ = 48; + + /* And null terminate the string: */ + *ascii = 0; + return; + } + } + + /* Here on buffer too small. */ + png_error(png_ptr, "ASCII conversion buffer too small"); +} +# endif /* FIXED_POINT */ +#endif /* READ_SCAL */ + +#if defined(PNG_FLOATING_POINT_SUPPORTED) && \ + !defined(PNG_FIXED_POINT_MACRO_SUPPORTED) && \ + (defined(PNG_gAMA_SUPPORTED) || defined(PNG_cHRM_SUPPORTED) || \ + defined(PNG_sCAL_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED) || \ + defined(PNG_READ_RGB_TO_GRAY_SUPPORTED)) || \ + (defined(PNG_sCAL_SUPPORTED) && \ + defined(PNG_FLOATING_ARITHMETIC_SUPPORTED)) +png_fixed_point +png_fixed(png_const_structrp png_ptr, double fp, png_const_charp text) +{ + double r = floor(100000 * fp + .5); + + if (r > 2147483647. || r < -2147483648.) + png_fixed_error(png_ptr, text); + +# ifndef PNG_ERROR_TEXT_SUPPORTED + PNG_UNUSED(text) +# endif + + return (png_fixed_point)r; +} +#endif + +#if defined(PNG_GAMMA_SUPPORTED) || defined(PNG_COLORSPACE_SUPPORTED) ||\ + defined(PNG_INCH_CONVERSIONS_SUPPORTED) || defined(PNG_READ_pHYs_SUPPORTED) +/* muldiv functions */ +/* This API takes signed arguments and rounds the result to the nearest + * integer (or, for a fixed point number - the standard argument - to + * the nearest .00001). Overflow and divide by zero are signalled in + * the result, a boolean - true on success, false on overflow. + */ +int +png_muldiv(png_fixed_point_p res, png_fixed_point a, png_int_32 times, + png_int_32 divisor) +{ + /* Return a * times / divisor, rounded. */ + if (divisor != 0) + { + if (a == 0 || times == 0) + { + *res = 0; + return 1; + } + else + { +#ifdef PNG_FLOATING_ARITHMETIC_SUPPORTED + double r = a; + r *= times; + r /= divisor; + r = floor(r+.5); + + /* A png_fixed_point is a 32-bit integer. */ + if (r <= 2147483647. && r >= -2147483648.) + { + *res = (png_fixed_point)r; + return 1; + } +#else + int negative = 0; + png_uint_32 A, T, D; + png_uint_32 s16, s32, s00; + + if (a < 0) + negative = 1, A = -a; + else + A = a; + + if (times < 0) + negative = !negative, T = -times; + else + T = times; + + if (divisor < 0) + negative = !negative, D = -divisor; + else + D = divisor; + + /* Following can't overflow because the arguments only + * have 31 bits each, however the result may be 32 bits. + */ + s16 = (A >> 16) * (T & 0xffff) + + (A & 0xffff) * (T >> 16); + /* Can't overflow because the a*times bit is only 30 + * bits at most. + */ + s32 = (A >> 16) * (T >> 16) + (s16 >> 16); + s00 = (A & 0xffff) * (T & 0xffff); + + s16 = (s16 & 0xffff) << 16; + s00 += s16; + + if (s00 < s16) + ++s32; /* carry */ + + if (s32 < D) /* else overflow */ + { + /* s32.s00 is now the 64-bit product, do a standard + * division, we know that s32 < D, so the maximum + * required shift is 31. + */ + int bitshift = 32; + png_fixed_point result = 0; /* NOTE: signed */ + + while (--bitshift >= 0) + { + png_uint_32 d32, d00; + + if (bitshift > 0) + d32 = D >> (32-bitshift), d00 = D << bitshift; + + else + d32 = 0, d00 = D; + + if (s32 > d32) + { + if (s00 < d00) --s32; /* carry */ + s32 -= d32, s00 -= d00, result += 1<= d00) + s32 = 0, s00 -= d00, result += 1<= (D >> 1)) + ++result; + + if (negative != 0) + result = -result; + + /* Check for overflow. */ + if ((negative != 0 && result <= 0) || + (negative == 0 && result >= 0)) + { + *res = result; + return 1; + } + } +#endif + } + } + + return 0; +} +#endif /* READ_GAMMA || INCH_CONVERSIONS */ + +#if defined(PNG_READ_GAMMA_SUPPORTED) || defined(PNG_INCH_CONVERSIONS_SUPPORTED) +/* The following is for when the caller doesn't much care about the + * result. + */ +png_fixed_point +png_muldiv_warn(png_const_structrp png_ptr, png_fixed_point a, png_int_32 times, + png_int_32 divisor) +{ + png_fixed_point result; + + if (png_muldiv(&result, a, times, divisor) != 0) + return result; + + png_warning(png_ptr, "fixed point overflow ignored"); + return 0; +} +#endif + +#ifdef PNG_GAMMA_SUPPORTED /* more fixed point functions for gamma */ +/* Calculate a reciprocal, return 0 on div-by-zero or overflow. */ +png_fixed_point +png_reciprocal(png_fixed_point a) +{ +#ifdef PNG_FLOATING_ARITHMETIC_SUPPORTED + double r = floor(1E10/a+.5); + + if (r <= 2147483647. && r >= -2147483648.) + return (png_fixed_point)r; +#else + png_fixed_point res; + + if (png_muldiv(&res, 100000, 100000, a) != 0) + return res; +#endif + + return 0; /* error/overflow */ +} + +/* This is the shared test on whether a gamma value is 'significant' - whether + * it is worth doing gamma correction. + */ +int /* PRIVATE */ +png_gamma_significant(png_fixed_point gamma_val) +{ + return gamma_val < PNG_FP_1 - PNG_GAMMA_THRESHOLD_FIXED || + gamma_val > PNG_FP_1 + PNG_GAMMA_THRESHOLD_FIXED; +} +#endif + +#ifdef PNG_READ_GAMMA_SUPPORTED +#if defined(PNG_16BIT_SUPPORTED) || !defined(PNG_FLOATING_ARITHMETIC_SUPPORTED) +/* A local convenience routine. */ +static png_fixed_point +png_product2(png_fixed_point a, png_fixed_point b) +{ + /* The required result is 1/a * 1/b; the following preserves accuracy. */ +# ifdef PNG_FLOATING_ARITHMETIC_SUPPORTED + double r = a * 1E-5; + r *= b; + r = floor(r+.5); + + if (r <= 2147483647. && r >= -2147483648.) + return (png_fixed_point)r; +# else + png_fixed_point res; + + if (png_muldiv(&res, a, b, 100000) != 0) + return res; +# endif + + return 0; /* overflow */ +} +#endif /* 16BIT || !FLOATING_ARITHMETIC */ + +/* The inverse of the above. */ +png_fixed_point +png_reciprocal2(png_fixed_point a, png_fixed_point b) +{ + /* The required result is 1/a * 1/b; the following preserves accuracy. */ +#ifdef PNG_FLOATING_ARITHMETIC_SUPPORTED + double r = 1E15/a; + r /= b; + r = floor(r+.5); + + if (r <= 2147483647. && r >= -2147483648.) + return (png_fixed_point)r; +#else + /* This may overflow because the range of png_fixed_point isn't symmetric, + * but this API is only used for the product of file and screen gamma so it + * doesn't matter that the smallest number it can produce is 1/21474, not + * 1/100000 + */ + png_fixed_point res = png_product2(a, b); + + if (res != 0) + return png_reciprocal(res); +#endif + + return 0; /* overflow */ +} +#endif /* READ_GAMMA */ + +#ifdef PNG_READ_GAMMA_SUPPORTED /* gamma table code */ +#ifndef PNG_FLOATING_ARITHMETIC_SUPPORTED +/* Fixed point gamma. + * + * The code to calculate the tables used below can be found in the shell script + * contrib/tools/intgamma.sh + * + * To calculate gamma this code implements fast log() and exp() calls using only + * fixed point arithmetic. This code has sufficient precision for either 8-bit + * or 16-bit sample values. + * + * The tables used here were calculated using simple 'bc' programs, but C double + * precision floating point arithmetic would work fine. + * + * 8-bit log table + * This is a table of -log(value/255)/log(2) for 'value' in the range 128 to + * 255, so it's the base 2 logarithm of a normalized 8-bit floating point + * mantissa. The numbers are 32-bit fractions. + */ +static const png_uint_32 +png_8bit_l2[128] = +{ + 4270715492U, 4222494797U, 4174646467U, 4127164793U, 4080044201U, 4033279239U, + 3986864580U, 3940795015U, 3895065449U, 3849670902U, 3804606499U, 3759867474U, + 3715449162U, 3671346997U, 3627556511U, 3584073329U, 3540893168U, 3498011834U, + 3455425220U, 3413129301U, 3371120137U, 3329393864U, 3287946700U, 3246774933U, + 3205874930U, 3165243125U, 3124876025U, 3084770202U, 3044922296U, 3005329011U, + 2965987113U, 2926893432U, 2888044853U, 2849438323U, 2811070844U, 2772939474U, + 2735041326U, 2697373562U, 2659933400U, 2622718104U, 2585724991U, 2548951424U, + 2512394810U, 2476052606U, 2439922311U, 2404001468U, 2368287663U, 2332778523U, + 2297471715U, 2262364947U, 2227455964U, 2192742551U, 2158222529U, 2123893754U, + 2089754119U, 2055801552U, 2022034013U, 1988449497U, 1955046031U, 1921821672U, + 1888774511U, 1855902668U, 1823204291U, 1790677560U, 1758320682U, 1726131893U, + 1694109454U, 1662251657U, 1630556815U, 1599023271U, 1567649391U, 1536433567U, + 1505374214U, 1474469770U, 1443718700U, 1413119487U, 1382670639U, 1352370686U, + 1322218179U, 1292211689U, 1262349810U, 1232631153U, 1203054352U, 1173618059U, + 1144320946U, 1115161701U, 1086139034U, 1057251672U, 1028498358U, 999877854U, + 971388940U, 943030410U, 914801076U, 886699767U, 858725327U, 830876614U, + 803152505U, 775551890U, 748073672U, 720716771U, 693480120U, 666362667U, + 639363374U, 612481215U, 585715177U, 559064263U, 532527486U, 506103872U, + 479792461U, 453592303U, 427502463U, 401522014U, 375650043U, 349885648U, + 324227938U, 298676034U, 273229066U, 247886176U, 222646516U, 197509248U, + 172473545U, 147538590U, 122703574U, 97967701U, 73330182U, 48790236U, + 24347096U, 0U + +#if 0 + /* The following are the values for 16-bit tables - these work fine for the + * 8-bit conversions but produce very slightly larger errors in the 16-bit + * log (about 1.2 as opposed to 0.7 absolute error in the final value). To + * use these all the shifts below must be adjusted appropriately. + */ + 65166, 64430, 63700, 62976, 62257, 61543, 60835, 60132, 59434, 58741, 58054, + 57371, 56693, 56020, 55352, 54689, 54030, 53375, 52726, 52080, 51439, 50803, + 50170, 49542, 48918, 48298, 47682, 47070, 46462, 45858, 45257, 44661, 44068, + 43479, 42894, 42312, 41733, 41159, 40587, 40020, 39455, 38894, 38336, 37782, + 37230, 36682, 36137, 35595, 35057, 34521, 33988, 33459, 32932, 32408, 31887, + 31369, 30854, 30341, 29832, 29325, 28820, 28319, 27820, 27324, 26830, 26339, + 25850, 25364, 24880, 24399, 23920, 23444, 22970, 22499, 22029, 21562, 21098, + 20636, 20175, 19718, 19262, 18808, 18357, 17908, 17461, 17016, 16573, 16132, + 15694, 15257, 14822, 14390, 13959, 13530, 13103, 12678, 12255, 11834, 11415, + 10997, 10582, 10168, 9756, 9346, 8937, 8531, 8126, 7723, 7321, 6921, 6523, + 6127, 5732, 5339, 4947, 4557, 4169, 3782, 3397, 3014, 2632, 2251, 1872, 1495, + 1119, 744, 372 +#endif +}; + +static png_int_32 +png_log8bit(unsigned int x) +{ + unsigned int lg2 = 0; + /* Each time 'x' is multiplied by 2, 1 must be subtracted off the final log, + * because the log is actually negate that means adding 1. The final + * returned value thus has the range 0 (for 255 input) to 7.994 (for 1 + * input), return -1 for the overflow (log 0) case, - so the result is + * always at most 19 bits. + */ + if ((x &= 0xff) == 0) + return -1; + + if ((x & 0xf0) == 0) + lg2 = 4, x <<= 4; + + if ((x & 0xc0) == 0) + lg2 += 2, x <<= 2; + + if ((x & 0x80) == 0) + lg2 += 1, x <<= 1; + + /* result is at most 19 bits, so this cast is safe: */ + return (png_int_32)((lg2 << 16) + ((png_8bit_l2[x-128]+32768)>>16)); +} + +/* The above gives exact (to 16 binary places) log2 values for 8-bit images, + * for 16-bit images we use the most significant 8 bits of the 16-bit value to + * get an approximation then multiply the approximation by a correction factor + * determined by the remaining up to 8 bits. This requires an additional step + * in the 16-bit case. + * + * We want log2(value/65535), we have log2(v'/255), where: + * + * value = v' * 256 + v'' + * = v' * f + * + * So f is value/v', which is equal to (256+v''/v') since v' is in the range 128 + * to 255 and v'' is in the range 0 to 255 f will be in the range 256 to less + * than 258. The final factor also needs to correct for the fact that our 8-bit + * value is scaled by 255, whereas the 16-bit values must be scaled by 65535. + * + * This gives a final formula using a calculated value 'x' which is value/v' and + * scaling by 65536 to match the above table: + * + * log2(x/257) * 65536 + * + * Since these numbers are so close to '1' we can use simple linear + * interpolation between the two end values 256/257 (result -368.61) and 258/257 + * (result 367.179). The values used below are scaled by a further 64 to give + * 16-bit precision in the interpolation: + * + * Start (256): -23591 + * Zero (257): 0 + * End (258): 23499 + */ +#ifdef PNG_16BIT_SUPPORTED +static png_int_32 +png_log16bit(png_uint_32 x) +{ + unsigned int lg2 = 0; + + /* As above, but now the input has 16 bits. */ + if ((x &= 0xffff) == 0) + return -1; + + if ((x & 0xff00) == 0) + lg2 = 8, x <<= 8; + + if ((x & 0xf000) == 0) + lg2 += 4, x <<= 4; + + if ((x & 0xc000) == 0) + lg2 += 2, x <<= 2; + + if ((x & 0x8000) == 0) + lg2 += 1, x <<= 1; + + /* Calculate the base logarithm from the top 8 bits as a 28-bit fractional + * value. + */ + lg2 <<= 28; + lg2 += (png_8bit_l2[(x>>8)-128]+8) >> 4; + + /* Now we need to interpolate the factor, this requires a division by the top + * 8 bits. Do this with maximum precision. + */ + x = ((x << 16) + (x >> 9)) / (x >> 8); + + /* Since we divided by the top 8 bits of 'x' there will be a '1' at 1<<24, + * the value at 1<<16 (ignoring this) will be 0 or 1; this gives us exactly + * 16 bits to interpolate to get the low bits of the result. Round the + * answer. Note that the end point values are scaled by 64 to retain overall + * precision and that 'lg2' is current scaled by an extra 12 bits, so adjust + * the overall scaling by 6-12. Round at every step. + */ + x -= 1U << 24; + + if (x <= 65536U) /* <= '257' */ + lg2 += ((23591U * (65536U-x)) + (1U << (16+6-12-1))) >> (16+6-12); + + else + lg2 -= ((23499U * (x-65536U)) + (1U << (16+6-12-1))) >> (16+6-12); + + /* Safe, because the result can't have more than 20 bits: */ + return (png_int_32)((lg2 + 2048) >> 12); +} +#endif /* 16BIT */ + +/* The 'exp()' case must invert the above, taking a 20-bit fixed point + * logarithmic value and returning a 16 or 8-bit number as appropriate. In + * each case only the low 16 bits are relevant - the fraction - since the + * integer bits (the top 4) simply determine a shift. + * + * The worst case is the 16-bit distinction between 65535 and 65534. This + * requires perhaps spurious accuracy in the decoding of the logarithm to + * distinguish log2(65535/65534.5) - 10^-5 or 17 bits. There is little chance + * of getting this accuracy in practice. + * + * To deal with this the following exp() function works out the exponent of the + * frational part of the logarithm by using an accurate 32-bit value from the + * top four fractional bits then multiplying in the remaining bits. + */ +static const png_uint_32 +png_32bit_exp[16] = +{ + /* NOTE: the first entry is deliberately set to the maximum 32-bit value. */ + 4294967295U, 4112874773U, 3938502376U, 3771522796U, 3611622603U, 3458501653U, + 3311872529U, 3171459999U, 3037000500U, 2908241642U, 2784941738U, 2666869345U, + 2553802834U, 2445529972U, 2341847524U, 2242560872U +}; + +/* Adjustment table; provided to explain the numbers in the code below. */ +#if 0 +for (i=11;i>=0;--i){ print i, " ", (1 - e(-(2^i)/65536*l(2))) * 2^(32-i), "\n"} + 11 44937.64284865548751208448 + 10 45180.98734845585101160448 + 9 45303.31936980687359311872 + 8 45364.65110595323018870784 + 7 45395.35850361789624614912 + 6 45410.72259715102037508096 + 5 45418.40724413220722311168 + 4 45422.25021786898173001728 + 3 45424.17186732298419044352 + 2 45425.13273269940811464704 + 1 45425.61317555035558641664 + 0 45425.85339951654943850496 +#endif + +static png_uint_32 +png_exp(png_fixed_point x) +{ + if (x > 0 && x <= 0xfffff) /* Else overflow or zero (underflow) */ + { + /* Obtain a 4-bit approximation */ + png_uint_32 e = png_32bit_exp[(x >> 12) & 0xf]; + + /* Incorporate the low 12 bits - these decrease the returned value by + * multiplying by a number less than 1 if the bit is set. The multiplier + * is determined by the above table and the shift. Notice that the values + * converge on 45426 and this is used to allow linear interpolation of the + * low bits. + */ + if (x & 0x800) + e -= (((e >> 16) * 44938U) + 16U) >> 5; + + if (x & 0x400) + e -= (((e >> 16) * 45181U) + 32U) >> 6; + + if (x & 0x200) + e -= (((e >> 16) * 45303U) + 64U) >> 7; + + if (x & 0x100) + e -= (((e >> 16) * 45365U) + 128U) >> 8; + + if (x & 0x080) + e -= (((e >> 16) * 45395U) + 256U) >> 9; + + if (x & 0x040) + e -= (((e >> 16) * 45410U) + 512U) >> 10; + + /* And handle the low 6 bits in a single block. */ + e -= (((e >> 16) * 355U * (x & 0x3fU)) + 256U) >> 9; + + /* Handle the upper bits of x. */ + e >>= x >> 16; + return e; + } + + /* Check for overflow */ + if (x <= 0) + return png_32bit_exp[0]; + + /* Else underflow */ + return 0; +} + +static png_byte +png_exp8bit(png_fixed_point lg2) +{ + /* Get a 32-bit value: */ + png_uint_32 x = png_exp(lg2); + + /* Convert the 32-bit value to 0..255 by multiplying by 256-1. Note that the + * second, rounding, step can't overflow because of the first, subtraction, + * step. + */ + x -= x >> 8; + return (png_byte)((x + 0x7fffffU) >> 24); +} + +#ifdef PNG_16BIT_SUPPORTED +static png_uint_16 +png_exp16bit(png_fixed_point lg2) +{ + /* Get a 32-bit value: */ + png_uint_32 x = png_exp(lg2); + + /* Convert the 32-bit value to 0..65535 by multiplying by 65536-1: */ + x -= x >> 16; + return (png_uint_16)((x + 32767U) >> 16); +} +#endif /* 16BIT */ +#endif /* FLOATING_ARITHMETIC */ + +png_byte +png_gamma_8bit_correct(unsigned int value, png_fixed_point gamma_val) +{ + if (value > 0 && value < 255) + { +# ifdef PNG_FLOATING_ARITHMETIC_SUPPORTED + /* 'value' is unsigned, ANSI-C90 requires the compiler to correctly + * convert this to a floating point value. This includes values that + * would overflow if 'value' were to be converted to 'int'. + * + * Apparently GCC, however, does an intermediate conversion to (int) + * on some (ARM) but not all (x86) platforms, possibly because of + * hardware FP limitations. (E.g. if the hardware conversion always + * assumes the integer register contains a signed value.) This results + * in ANSI-C undefined behavior for large values. + * + * Other implementations on the same machine might actually be ANSI-C90 + * conformant and therefore compile spurious extra code for the large + * values. + * + * We can be reasonably sure that an unsigned to float conversion + * won't be faster than an int to float one. Therefore this code + * assumes responsibility for the undefined behavior, which it knows + * can't happen because of the check above. + * + * Note the argument to this routine is an (unsigned int) because, on + * 16-bit platforms, it is assigned a value which might be out of + * range for an (int); that would result in undefined behavior in the + * caller if the *argument* ('value') were to be declared (int). + */ + double r = floor(255*pow((int)/*SAFE*/value/255.,gamma_val*.00001)+.5); + return (png_byte)r; +# else + png_int_32 lg2 = png_log8bit(value); + png_fixed_point res; + + if (png_muldiv(&res, gamma_val, lg2, PNG_FP_1) != 0) + return png_exp8bit(res); + + /* Overflow. */ + value = 0; +# endif + } + + return (png_byte)value; +} + +#ifdef PNG_16BIT_SUPPORTED +png_uint_16 +png_gamma_16bit_correct(unsigned int value, png_fixed_point gamma_val) +{ + if (value > 0 && value < 65535) + { +# ifdef PNG_FLOATING_ARITHMETIC_SUPPORTED + /* The same (unsigned int)->(double) constraints apply here as above, + * however in this case the (unsigned int) to (int) conversion can + * overflow on an ANSI-C90 compliant system so the cast needs to ensure + * that this is not possible. + */ + double r = floor(65535*pow((png_int_32)value/65535., + gamma_val*.00001)+.5); + return (png_uint_16)r; +# else + png_int_32 lg2 = png_log16bit(value); + png_fixed_point res; + + if (png_muldiv(&res, gamma_val, lg2, PNG_FP_1) != 0) + return png_exp16bit(res); + + /* Overflow. */ + value = 0; +# endif + } + + return (png_uint_16)value; +} +#endif /* 16BIT */ + +/* This does the right thing based on the bit_depth field of the + * png_struct, interpreting values as 8-bit or 16-bit. While the result + * is nominally a 16-bit value if bit depth is 8 then the result is + * 8-bit (as are the arguments.) + */ +png_uint_16 /* PRIVATE */ +png_gamma_correct(png_structrp png_ptr, unsigned int value, + png_fixed_point gamma_val) +{ + if (png_ptr->bit_depth == 8) + return png_gamma_8bit_correct(value, gamma_val); + +#ifdef PNG_16BIT_SUPPORTED + else + return png_gamma_16bit_correct(value, gamma_val); +#else + /* should not reach this */ + return 0; +#endif /* 16BIT */ +} + +#ifdef PNG_16BIT_SUPPORTED +/* Internal function to build a single 16-bit table - the table consists of + * 'num' 256 entry subtables, where 'num' is determined by 'shift' - the amount + * to shift the input values right (or 16-number_of_signifiant_bits). + * + * The caller is responsible for ensuring that the table gets cleaned up on + * png_error (i.e. if one of the mallocs below fails) - i.e. the *table argument + * should be somewhere that will be cleaned. + */ +static void +png_build_16bit_table(png_structrp png_ptr, png_uint_16pp *ptable, + PNG_CONST unsigned int shift, PNG_CONST png_fixed_point gamma_val) +{ + /* Various values derived from 'shift': */ + PNG_CONST unsigned int num = 1U << (8U - shift); +#ifdef PNG_FLOATING_ARITHMETIC_SUPPORTED + /* CSE the division and work round wacky GCC warnings (see the comments + * in png_gamma_8bit_correct for where these come from.) + */ + PNG_CONST double fmax = 1./(((png_int_32)1 << (16U - shift))-1); +#endif + PNG_CONST unsigned int max = (1U << (16U - shift))-1U; + PNG_CONST unsigned int max_by_2 = 1U << (15U-shift); + unsigned int i; + + png_uint_16pp table = *ptable = + (png_uint_16pp)png_calloc(png_ptr, num * (sizeof (png_uint_16p))); + + for (i = 0; i < num; i++) + { + png_uint_16p sub_table = table[i] = + (png_uint_16p)png_malloc(png_ptr, 256 * (sizeof (png_uint_16))); + + /* The 'threshold' test is repeated here because it can arise for one of + * the 16-bit tables even if the others don't hit it. + */ + if (png_gamma_significant(gamma_val) != 0) + { + /* The old code would overflow at the end and this would cause the + * 'pow' function to return a result >1, resulting in an + * arithmetic error. This code follows the spec exactly; ig is + * the recovered input sample, it always has 8-16 bits. + * + * We want input * 65535/max, rounded, the arithmetic fits in 32 + * bits (unsigned) so long as max <= 32767. + */ + unsigned int j; + for (j = 0; j < 256; j++) + { + png_uint_32 ig = (j << (8-shift)) + i; +# ifdef PNG_FLOATING_ARITHMETIC_SUPPORTED + /* Inline the 'max' scaling operation: */ + /* See png_gamma_8bit_correct for why the cast to (int) is + * required here. + */ + double d = floor(65535.*pow(ig*fmax, gamma_val*.00001)+.5); + sub_table[j] = (png_uint_16)d; +# else + if (shift != 0) + ig = (ig * 65535U + max_by_2)/max; + + sub_table[j] = png_gamma_16bit_correct(ig, gamma_val); +# endif + } + } + else + { + /* We must still build a table, but do it the fast way. */ + unsigned int j; + + for (j = 0; j < 256; j++) + { + png_uint_32 ig = (j << (8-shift)) + i; + + if (shift != 0) + ig = (ig * 65535U + max_by_2)/max; + + sub_table[j] = (png_uint_16)ig; + } + } + } +} + +/* NOTE: this function expects the *inverse* of the overall gamma transformation + * required. + */ +static void +png_build_16to8_table(png_structrp png_ptr, png_uint_16pp *ptable, + PNG_CONST unsigned int shift, PNG_CONST png_fixed_point gamma_val) +{ + PNG_CONST unsigned int num = 1U << (8U - shift); + PNG_CONST unsigned int max = (1U << (16U - shift))-1U; + unsigned int i; + png_uint_32 last; + + png_uint_16pp table = *ptable = + (png_uint_16pp)png_calloc(png_ptr, num * (sizeof (png_uint_16p))); + + /* 'num' is the number of tables and also the number of low bits of low + * bits of the input 16-bit value used to select a table. Each table is + * itself indexed by the high 8 bits of the value. + */ + for (i = 0; i < num; i++) + table[i] = (png_uint_16p)png_malloc(png_ptr, + 256 * (sizeof (png_uint_16))); + + /* 'gamma_val' is set to the reciprocal of the value calculated above, so + * pow(out,g) is an *input* value. 'last' is the last input value set. + * + * In the loop 'i' is used to find output values. Since the output is + * 8-bit there are only 256 possible values. The tables are set up to + * select the closest possible output value for each input by finding + * the input value at the boundary between each pair of output values + * and filling the table up to that boundary with the lower output + * value. + * + * The boundary values are 0.5,1.5..253.5,254.5. Since these are 9-bit + * values the code below uses a 16-bit value in i; the values start at + * 128.5 (for 0.5) and step by 257, for a total of 254 values (the last + * entries are filled with 255). Start i at 128 and fill all 'last' + * table entries <= 'max' + */ + last = 0; + for (i = 0; i < 255; ++i) /* 8-bit output value */ + { + /* Find the corresponding maximum input value */ + png_uint_16 out = (png_uint_16)(i * 257U); /* 16-bit output value */ + + /* Find the boundary value in 16 bits: */ + png_uint_32 bound = png_gamma_16bit_correct(out+128U, gamma_val); + + /* Adjust (round) to (16-shift) bits: */ + bound = (bound * max + 32768U)/65535U + 1U; + + while (last < bound) + { + table[last & (0xffU >> shift)][last >> (8U - shift)] = out; + last++; + } + } + + /* And fill in the final entries. */ + while (last < (num << 8)) + { + table[last & (0xff >> shift)][last >> (8U - shift)] = 65535U; + last++; + } +} +#endif /* 16BIT */ + +/* Build a single 8-bit table: same as the 16-bit case but much simpler (and + * typically much faster). Note that libpng currently does no sBIT processing + * (apparently contrary to the spec) so a 256-entry table is always generated. + */ +static void +png_build_8bit_table(png_structrp png_ptr, png_bytepp ptable, + PNG_CONST png_fixed_point gamma_val) +{ + unsigned int i; + png_bytep table = *ptable = (png_bytep)png_malloc(png_ptr, 256); + + if (png_gamma_significant(gamma_val) != 0) + for (i=0; i<256; i++) + table[i] = png_gamma_8bit_correct(i, gamma_val); + + else + for (i=0; i<256; ++i) + table[i] = (png_byte)i; +} + +/* Used from png_read_destroy and below to release the memory used by the gamma + * tables. + */ +void /* PRIVATE */ +png_destroy_gamma_table(png_structrp png_ptr) +{ + png_free(png_ptr, png_ptr->gamma_table); + png_ptr->gamma_table = NULL; + +#ifdef PNG_16BIT_SUPPORTED + if (png_ptr->gamma_16_table != NULL) + { + int i; + int istop = (1 << (8 - png_ptr->gamma_shift)); + for (i = 0; i < istop; i++) + { + png_free(png_ptr, png_ptr->gamma_16_table[i]); + } + png_free(png_ptr, png_ptr->gamma_16_table); + png_ptr->gamma_16_table = NULL; + } +#endif /* 16BIT */ + +#if defined(PNG_READ_BACKGROUND_SUPPORTED) || \ + defined(PNG_READ_ALPHA_MODE_SUPPORTED) || \ + defined(PNG_READ_RGB_TO_GRAY_SUPPORTED) + png_free(png_ptr, png_ptr->gamma_from_1); + png_ptr->gamma_from_1 = NULL; + png_free(png_ptr, png_ptr->gamma_to_1); + png_ptr->gamma_to_1 = NULL; + +#ifdef PNG_16BIT_SUPPORTED + if (png_ptr->gamma_16_from_1 != NULL) + { + int i; + int istop = (1 << (8 - png_ptr->gamma_shift)); + for (i = 0; i < istop; i++) + { + png_free(png_ptr, png_ptr->gamma_16_from_1[i]); + } + png_free(png_ptr, png_ptr->gamma_16_from_1); + png_ptr->gamma_16_from_1 = NULL; + } + if (png_ptr->gamma_16_to_1 != NULL) + { + int i; + int istop = (1 << (8 - png_ptr->gamma_shift)); + for (i = 0; i < istop; i++) + { + png_free(png_ptr, png_ptr->gamma_16_to_1[i]); + } + png_free(png_ptr, png_ptr->gamma_16_to_1); + png_ptr->gamma_16_to_1 = NULL; + } +#endif /* 16BIT */ +#endif /* READ_BACKGROUND || READ_ALPHA_MODE || RGB_TO_GRAY */ +} + +/* We build the 8- or 16-bit gamma tables here. Note that for 16-bit + * tables, we don't make a full table if we are reducing to 8-bit in + * the future. Note also how the gamma_16 tables are segmented so that + * we don't need to allocate > 64K chunks for a full 16-bit table. + */ +void /* PRIVATE */ +png_build_gamma_table(png_structrp png_ptr, int bit_depth) +{ + png_debug(1, "in png_build_gamma_table"); + + /* Remove any existing table; this copes with multiple calls to + * png_read_update_info. The warning is because building the gamma tables + * multiple times is a performance hit - it's harmless but the ability to call + * png_read_update_info() multiple times is new in 1.5.6 so it seems sensible + * to warn if the app introduces such a hit. + */ + if (png_ptr->gamma_table != NULL || png_ptr->gamma_16_table != NULL) + { + png_warning(png_ptr, "gamma table being rebuilt"); + png_destroy_gamma_table(png_ptr); + } + + if (bit_depth <= 8) + { + png_build_8bit_table(png_ptr, &png_ptr->gamma_table, + png_ptr->screen_gamma > 0 ? png_reciprocal2(png_ptr->colorspace.gamma, + png_ptr->screen_gamma) : PNG_FP_1); + +#if defined(PNG_READ_BACKGROUND_SUPPORTED) || \ + defined(PNG_READ_ALPHA_MODE_SUPPORTED) || \ + defined(PNG_READ_RGB_TO_GRAY_SUPPORTED) + if ((png_ptr->transformations & (PNG_COMPOSE | PNG_RGB_TO_GRAY)) != 0) + { + png_build_8bit_table(png_ptr, &png_ptr->gamma_to_1, + png_reciprocal(png_ptr->colorspace.gamma)); + + png_build_8bit_table(png_ptr, &png_ptr->gamma_from_1, + png_ptr->screen_gamma > 0 ? png_reciprocal(png_ptr->screen_gamma) : + png_ptr->colorspace.gamma/* Probably doing rgb_to_gray */); + } +#endif /* READ_BACKGROUND || READ_ALPHA_MODE || RGB_TO_GRAY */ + } +#ifdef PNG_16BIT_SUPPORTED + else + { + png_byte shift, sig_bit; + + if ((png_ptr->color_type & PNG_COLOR_MASK_COLOR) != 0) + { + sig_bit = png_ptr->sig_bit.red; + + if (png_ptr->sig_bit.green > sig_bit) + sig_bit = png_ptr->sig_bit.green; + + if (png_ptr->sig_bit.blue > sig_bit) + sig_bit = png_ptr->sig_bit.blue; + } + else + sig_bit = png_ptr->sig_bit.gray; + + /* 16-bit gamma code uses this equation: + * + * ov = table[(iv & 0xff) >> gamma_shift][iv >> 8] + * + * Where 'iv' is the input color value and 'ov' is the output value - + * pow(iv, gamma). + * + * Thus the gamma table consists of up to 256 256-entry tables. The table + * is selected by the (8-gamma_shift) most significant of the low 8 bits of + * the color value then indexed by the upper 8 bits: + * + * table[low bits][high 8 bits] + * + * So the table 'n' corresponds to all those 'iv' of: + * + * ..<(n+1 << gamma_shift)-1> + * + */ + if (sig_bit > 0 && sig_bit < 16U) + shift = (png_byte)(16U - sig_bit); /* shift == insignificant bits */ + + else + shift = 0; /* keep all 16 bits */ + + if ((png_ptr->transformations & (PNG_16_TO_8 | PNG_SCALE_16_TO_8)) != 0) + { + /* PNG_MAX_GAMMA_8 is the number of bits to keep - effectively + * the significant bits in the *input* when the output will + * eventually be 8 bits. By default it is 11. + */ + if (shift < (16U - PNG_MAX_GAMMA_8)) + shift = (16U - PNG_MAX_GAMMA_8); + } + + if (shift > 8U) + shift = 8U; /* Guarantees at least one table! */ + + png_ptr->gamma_shift = shift; + + /* NOTE: prior to 1.5.4 this test used to include PNG_BACKGROUND (now + * PNG_COMPOSE). This effectively smashed the background calculation for + * 16-bit output because the 8-bit table assumes the result will be reduced + * to 8 bits. + */ + if ((png_ptr->transformations & (PNG_16_TO_8 | PNG_SCALE_16_TO_8)) != 0) + png_build_16to8_table(png_ptr, &png_ptr->gamma_16_table, shift, + png_ptr->screen_gamma > 0 ? png_product2(png_ptr->colorspace.gamma, + png_ptr->screen_gamma) : PNG_FP_1); + + else + png_build_16bit_table(png_ptr, &png_ptr->gamma_16_table, shift, + png_ptr->screen_gamma > 0 ? png_reciprocal2(png_ptr->colorspace.gamma, + png_ptr->screen_gamma) : PNG_FP_1); + +#if defined(PNG_READ_BACKGROUND_SUPPORTED) || \ + defined(PNG_READ_ALPHA_MODE_SUPPORTED) || \ + defined(PNG_READ_RGB_TO_GRAY_SUPPORTED) + if ((png_ptr->transformations & (PNG_COMPOSE | PNG_RGB_TO_GRAY)) != 0) + { + png_build_16bit_table(png_ptr, &png_ptr->gamma_16_to_1, shift, + png_reciprocal(png_ptr->colorspace.gamma)); + + /* Notice that the '16 from 1' table should be full precision, however + * the lookup on this table still uses gamma_shift, so it can't be. + * TODO: fix this. + */ + png_build_16bit_table(png_ptr, &png_ptr->gamma_16_from_1, shift, + png_ptr->screen_gamma > 0 ? png_reciprocal(png_ptr->screen_gamma) : + png_ptr->colorspace.gamma/* Probably doing rgb_to_gray */); + } +#endif /* READ_BACKGROUND || READ_ALPHA_MODE || RGB_TO_GRAY */ + } +#endif /* 16BIT */ +} +#endif /* READ_GAMMA */ + +/* HARDWARE OR SOFTWARE OPTION SUPPORT */ +#ifdef PNG_SET_OPTION_SUPPORTED +int PNGAPI +png_set_option(png_structrp png_ptr, int option, int onoff) +{ + if (png_ptr != NULL && option >= 0 && option < PNG_OPTION_NEXT && + (option & 1) == 0) + { + int mask = 3 << option; + int setting = (2 + (onoff != 0)) << option; + int current = png_ptr->options; + + png_ptr->options = (png_byte)((current & ~mask) | setting); + + return (current & mask) >> option; + } + + return PNG_OPTION_INVALID; +} +#endif + +/* sRGB support */ +#if defined(PNG_SIMPLIFIED_READ_SUPPORTED) ||\ + defined(PNG_SIMPLIFIED_WRITE_SUPPORTED) +/* sRGB conversion tables; these are machine generated with the code in + * contrib/tools/makesRGB.c. The actual sRGB transfer curve defined in the + * specification (see the article at http://en.wikipedia.org/wiki/SRGB) + * is used, not the gamma=1/2.2 approximation use elsewhere in libpng. + * The sRGB to linear table is exact (to the nearest 16 bit linear fraction). + * The inverse (linear to sRGB) table has accuracies as follows: + * + * For all possible (255*65535+1) input values: + * + * error: -0.515566 - 0.625971, 79441 (0.475369%) of readings inexact + * + * For the input values corresponding to the 65536 16-bit values: + * + * error: -0.513727 - 0.607759, 308 (0.469978%) of readings inexact + * + * In all cases the inexact readings are only off by one. + */ + +#ifdef PNG_SIMPLIFIED_READ_SUPPORTED +/* The convert-to-sRGB table is only currently required for read. */ +const png_uint_16 png_sRGB_table[256] = +{ + 0,20,40,60,80,99,119,139, + 159,179,199,219,241,264,288,313, + 340,367,396,427,458,491,526,562, + 599,637,677,718,761,805,851,898, + 947,997,1048,1101,1156,1212,1270,1330, + 1391,1453,1517,1583,1651,1720,1790,1863, + 1937,2013,2090,2170,2250,2333,2418,2504, + 2592,2681,2773,2866,2961,3058,3157,3258, + 3360,3464,3570,3678,3788,3900,4014,4129, + 4247,4366,4488,4611,4736,4864,4993,5124, + 5257,5392,5530,5669,5810,5953,6099,6246, + 6395,6547,6700,6856,7014,7174,7335,7500, + 7666,7834,8004,8177,8352,8528,8708,8889, + 9072,9258,9445,9635,9828,10022,10219,10417, + 10619,10822,11028,11235,11446,11658,11873,12090, + 12309,12530,12754,12980,13209,13440,13673,13909, + 14146,14387,14629,14874,15122,15371,15623,15878, + 16135,16394,16656,16920,17187,17456,17727,18001, + 18277,18556,18837,19121,19407,19696,19987,20281, + 20577,20876,21177,21481,21787,22096,22407,22721, + 23038,23357,23678,24002,24329,24658,24990,25325, + 25662,26001,26344,26688,27036,27386,27739,28094, + 28452,28813,29176,29542,29911,30282,30656,31033, + 31412,31794,32179,32567,32957,33350,33745,34143, + 34544,34948,35355,35764,36176,36591,37008,37429, + 37852,38278,38706,39138,39572,40009,40449,40891, + 41337,41785,42236,42690,43147,43606,44069,44534, + 45002,45473,45947,46423,46903,47385,47871,48359, + 48850,49344,49841,50341,50844,51349,51858,52369, + 52884,53401,53921,54445,54971,55500,56032,56567, + 57105,57646,58190,58737,59287,59840,60396,60955, + 61517,62082,62650,63221,63795,64372,64952,65535 +}; +#endif /* SIMPLIFIED_READ */ + +/* The base/delta tables are required for both read and write (but currently + * only the simplified versions.) + */ +const png_uint_16 png_sRGB_base[512] = +{ + 128,1782,3383,4644,5675,6564,7357,8074, + 8732,9346,9921,10463,10977,11466,11935,12384, + 12816,13233,13634,14024,14402,14769,15125,15473, + 15812,16142,16466,16781,17090,17393,17690,17981, + 18266,18546,18822,19093,19359,19621,19879,20133, + 20383,20630,20873,21113,21349,21583,21813,22041, + 22265,22487,22707,22923,23138,23350,23559,23767, + 23972,24175,24376,24575,24772,24967,25160,25352, + 25542,25730,25916,26101,26284,26465,26645,26823, + 27000,27176,27350,27523,27695,27865,28034,28201, + 28368,28533,28697,28860,29021,29182,29341,29500, + 29657,29813,29969,30123,30276,30429,30580,30730, + 30880,31028,31176,31323,31469,31614,31758,31902, + 32045,32186,32327,32468,32607,32746,32884,33021, + 33158,33294,33429,33564,33697,33831,33963,34095, + 34226,34357,34486,34616,34744,34873,35000,35127, + 35253,35379,35504,35629,35753,35876,35999,36122, + 36244,36365,36486,36606,36726,36845,36964,37083, + 37201,37318,37435,37551,37668,37783,37898,38013, + 38127,38241,38354,38467,38580,38692,38803,38915, + 39026,39136,39246,39356,39465,39574,39682,39790, + 39898,40005,40112,40219,40325,40431,40537,40642, + 40747,40851,40955,41059,41163,41266,41369,41471, + 41573,41675,41777,41878,41979,42079,42179,42279, + 42379,42478,42577,42676,42775,42873,42971,43068, + 43165,43262,43359,43456,43552,43648,43743,43839, + 43934,44028,44123,44217,44311,44405,44499,44592, + 44685,44778,44870,44962,45054,45146,45238,45329, + 45420,45511,45601,45692,45782,45872,45961,46051, + 46140,46229,46318,46406,46494,46583,46670,46758, + 46846,46933,47020,47107,47193,47280,47366,47452, + 47538,47623,47709,47794,47879,47964,48048,48133, + 48217,48301,48385,48468,48552,48635,48718,48801, + 48884,48966,49048,49131,49213,49294,49376,49458, + 49539,49620,49701,49782,49862,49943,50023,50103, + 50183,50263,50342,50422,50501,50580,50659,50738, + 50816,50895,50973,51051,51129,51207,51285,51362, + 51439,51517,51594,51671,51747,51824,51900,51977, + 52053,52129,52205,52280,52356,52432,52507,52582, + 52657,52732,52807,52881,52956,53030,53104,53178, + 53252,53326,53400,53473,53546,53620,53693,53766, + 53839,53911,53984,54056,54129,54201,54273,54345, + 54417,54489,54560,54632,54703,54774,54845,54916, + 54987,55058,55129,55199,55269,55340,55410,55480, + 55550,55620,55689,55759,55828,55898,55967,56036, + 56105,56174,56243,56311,56380,56448,56517,56585, + 56653,56721,56789,56857,56924,56992,57059,57127, + 57194,57261,57328,57395,57462,57529,57595,57662, + 57728,57795,57861,57927,57993,58059,58125,58191, + 58256,58322,58387,58453,58518,58583,58648,58713, + 58778,58843,58908,58972,59037,59101,59165,59230, + 59294,59358,59422,59486,59549,59613,59677,59740, + 59804,59867,59930,59993,60056,60119,60182,60245, + 60308,60370,60433,60495,60558,60620,60682,60744, + 60806,60868,60930,60992,61054,61115,61177,61238, + 61300,61361,61422,61483,61544,61605,61666,61727, + 61788,61848,61909,61969,62030,62090,62150,62211, + 62271,62331,62391,62450,62510,62570,62630,62689, + 62749,62808,62867,62927,62986,63045,63104,63163, + 63222,63281,63340,63398,63457,63515,63574,63632, + 63691,63749,63807,63865,63923,63981,64039,64097, + 64155,64212,64270,64328,64385,64443,64500,64557, + 64614,64672,64729,64786,64843,64900,64956,65013, + 65070,65126,65183,65239,65296,65352,65409,65465 +}; + +const png_byte png_sRGB_delta[512] = +{ + 207,201,158,129,113,100,90,82,77,72,68,64,61,59,56,54, + 52,50,49,47,46,45,43,42,41,40,39,39,38,37,36,36, + 35,34,34,33,33,32,32,31,31,30,30,30,29,29,28,28, + 28,27,27,27,27,26,26,26,25,25,25,25,24,24,24,24, + 23,23,23,23,23,22,22,22,22,22,22,21,21,21,21,21, + 21,20,20,20,20,20,20,20,20,19,19,19,19,19,19,19, + 19,18,18,18,18,18,18,18,18,18,18,17,17,17,17,17, + 17,17,17,17,17,17,16,16,16,16,16,16,16,16,16,16, + 16,16,16,16,15,15,15,15,15,15,15,15,15,15,15,15, + 15,15,15,15,14,14,14,14,14,14,14,14,14,14,14,14, + 14,14,14,14,14,14,14,13,13,13,13,13,13,13,13,13, + 13,13,13,13,13,13,13,13,13,13,13,13,13,13,12,12, + 12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12, + 12,12,12,12,12,12,12,12,12,12,12,12,11,11,11,11, + 11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11, + 11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11, + 11,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10, + 10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10, + 10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10, + 10,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,7,7,7,7,7,7,7, + 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, + 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, + 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7 +}; +#endif /* SIMPLIFIED READ/WRITE sRGB support */ + +/* SIMPLIFIED READ/WRITE SUPPORT */ +#if defined(PNG_SIMPLIFIED_READ_SUPPORTED) ||\ + defined(PNG_SIMPLIFIED_WRITE_SUPPORTED) +static int +png_image_free_function(png_voidp argument) +{ + png_imagep image = png_voidcast(png_imagep, argument); + png_controlp cp = image->opaque; + png_control c; + + /* Double check that we have a png_ptr - it should be impossible to get here + * without one. + */ + if (cp->png_ptr == NULL) + return 0; + + /* First free any data held in the control structure. */ +# ifdef PNG_STDIO_SUPPORTED + if (cp->owned_file != 0) + { + FILE *fp = png_voidcast(FILE*, cp->png_ptr->io_ptr); + cp->owned_file = 0; + + /* Ignore errors here. */ + if (fp != NULL) + { + cp->png_ptr->io_ptr = NULL; + (void)fclose(fp); + } + } +# endif + + /* Copy the control structure so that the original, allocated, version can be + * safely freed. Notice that a png_error here stops the remainder of the + * cleanup, but this is probably fine because that would indicate bad memory + * problems anyway. + */ + c = *cp; + image->opaque = &c; + png_free(c.png_ptr, cp); + + /* Then the structures, calling the correct API. */ + if (c.for_write != 0) + { +# ifdef PNG_SIMPLIFIED_WRITE_SUPPORTED + png_destroy_write_struct(&c.png_ptr, &c.info_ptr); +# else + png_error(c.png_ptr, "simplified write not supported"); +# endif + } + else + { +# ifdef PNG_SIMPLIFIED_READ_SUPPORTED + png_destroy_read_struct(&c.png_ptr, &c.info_ptr, NULL); +# else + png_error(c.png_ptr, "simplified read not supported"); +# endif + } + + /* Success. */ + return 1; +} + +void PNGAPI +png_image_free(png_imagep image) +{ + /* Safely call the real function, but only if doing so is safe at this point + * (if not inside an error handling context). Otherwise assume + * png_safe_execute will call this API after the return. + */ + if (image != NULL && image->opaque != NULL && + image->opaque->error_buf == NULL) + { + /* Ignore errors here: */ + (void)png_safe_execute(image, png_image_free_function, image); + image->opaque = NULL; + } +} + +int /* PRIVATE */ +png_image_error(png_imagep image, png_const_charp error_message) +{ + /* Utility to log an error. */ + png_safecat(image->message, (sizeof image->message), 0, error_message); + image->warning_or_error |= PNG_IMAGE_ERROR; + png_image_free(image); + return 0; +} + +#endif /* SIMPLIFIED READ/WRITE */ +#endif /* READ || WRITE */ diff --git a/app/src/main/jni/libpng/png.h b/app/src/main/jni/libpng/png.h new file mode 100644 index 000000000..997130d38 --- /dev/null +++ b/app/src/main/jni/libpng/png.h @@ -0,0 +1,3282 @@ + +/* png.h - header file for PNG reference library + * + * libpng version 1.6.16, December 22, 2014 + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * This code is released under the libpng license (See LICENSE, below) + * + * Authors and maintainers: + * libpng versions 0.71, May 1995, through 0.88, January 1996: Guy Schalnat + * libpng versions 0.89c, June 1996, through 0.96, May 1997: Andreas Dilger + * libpng versions 0.97, January 1998, through 1.6.16, December 22, 2014: Glenn + * See also "Contributing Authors", below. + * + * Note about libpng version numbers: + * + * Due to various miscommunications, unforeseen code incompatibilities + * and occasional factors outside the authors' control, version numbering + * on the library has not always been consistent and straightforward. + * The following table summarizes matters since version 0.89c, which was + * the first widely used release: + * + * source png.h png.h shared-lib + * version string int version + * ------- ------ ----- ---------- + * 0.89c "1.0 beta 3" 0.89 89 1.0.89 + * 0.90 "1.0 beta 4" 0.90 90 0.90 [should have been 2.0.90] + * 0.95 "1.0 beta 5" 0.95 95 0.95 [should have been 2.0.95] + * 0.96 "1.0 beta 6" 0.96 96 0.96 [should have been 2.0.96] + * 0.97b "1.00.97 beta 7" 1.00.97 97 1.0.1 [should have been 2.0.97] + * 0.97c 0.97 97 2.0.97 + * 0.98 0.98 98 2.0.98 + * 0.99 0.99 98 2.0.99 + * 0.99a-m 0.99 99 2.0.99 + * 1.00 1.00 100 2.1.0 [100 should be 10000] + * 1.0.0 (from here on, the 100 2.1.0 [100 should be 10000] + * 1.0.1 png.h string is 10001 2.1.0 + * 1.0.1a-e identical to the 10002 from here on, the shared library + * 1.0.2 source version) 10002 is 2.V where V is the source code + * 1.0.2a-b 10003 version, except as noted. + * 1.0.3 10003 + * 1.0.3a-d 10004 + * 1.0.4 10004 + * 1.0.4a-f 10005 + * 1.0.5 (+ 2 patches) 10005 + * 1.0.5a-d 10006 + * 1.0.5e-r 10100 (not source compatible) + * 1.0.5s-v 10006 (not binary compatible) + * 1.0.6 (+ 3 patches) 10006 (still binary incompatible) + * 1.0.6d-f 10007 (still binary incompatible) + * 1.0.6g 10007 + * 1.0.6h 10007 10.6h (testing xy.z so-numbering) + * 1.0.6i 10007 10.6i + * 1.0.6j 10007 2.1.0.6j (incompatible with 1.0.0) + * 1.0.7beta11-14 DLLNUM 10007 2.1.0.7beta11-14 (binary compatible) + * 1.0.7beta15-18 1 10007 2.1.0.7beta15-18 (binary compatible) + * 1.0.7rc1-2 1 10007 2.1.0.7rc1-2 (binary compatible) + * 1.0.7 1 10007 (still compatible) + * 1.0.8beta1-4 1 10008 2.1.0.8beta1-4 + * 1.0.8rc1 1 10008 2.1.0.8rc1 + * 1.0.8 1 10008 2.1.0.8 + * 1.0.9beta1-6 1 10009 2.1.0.9beta1-6 + * 1.0.9rc1 1 10009 2.1.0.9rc1 + * 1.0.9beta7-10 1 10009 2.1.0.9beta7-10 + * 1.0.9rc2 1 10009 2.1.0.9rc2 + * 1.0.9 1 10009 2.1.0.9 + * 1.0.10beta1 1 10010 2.1.0.10beta1 + * 1.0.10rc1 1 10010 2.1.0.10rc1 + * 1.0.10 1 10010 2.1.0.10 + * 1.0.11beta1-3 1 10011 2.1.0.11beta1-3 + * 1.0.11rc1 1 10011 2.1.0.11rc1 + * 1.0.11 1 10011 2.1.0.11 + * 1.0.12beta1-2 2 10012 2.1.0.12beta1-2 + * 1.0.12rc1 2 10012 2.1.0.12rc1 + * 1.0.12 2 10012 2.1.0.12 + * 1.1.0a-f - 10100 2.1.1.0a-f (branch abandoned) + * 1.2.0beta1-2 2 10200 2.1.2.0beta1-2 + * 1.2.0beta3-5 3 10200 3.1.2.0beta3-5 + * 1.2.0rc1 3 10200 3.1.2.0rc1 + * 1.2.0 3 10200 3.1.2.0 + * 1.2.1beta1-4 3 10201 3.1.2.1beta1-4 + * 1.2.1rc1-2 3 10201 3.1.2.1rc1-2 + * 1.2.1 3 10201 3.1.2.1 + * 1.2.2beta1-6 12 10202 12.so.0.1.2.2beta1-6 + * 1.0.13beta1 10 10013 10.so.0.1.0.13beta1 + * 1.0.13rc1 10 10013 10.so.0.1.0.13rc1 + * 1.2.2rc1 12 10202 12.so.0.1.2.2rc1 + * 1.0.13 10 10013 10.so.0.1.0.13 + * 1.2.2 12 10202 12.so.0.1.2.2 + * 1.2.3rc1-6 12 10203 12.so.0.1.2.3rc1-6 + * 1.2.3 12 10203 12.so.0.1.2.3 + * 1.2.4beta1-3 13 10204 12.so.0.1.2.4beta1-3 + * 1.0.14rc1 13 10014 10.so.0.1.0.14rc1 + * 1.2.4rc1 13 10204 12.so.0.1.2.4rc1 + * 1.0.14 10 10014 10.so.0.1.0.14 + * 1.2.4 13 10204 12.so.0.1.2.4 + * 1.2.5beta1-2 13 10205 12.so.0.1.2.5beta1-2 + * 1.0.15rc1-3 10 10015 10.so.0.1.0.15rc1-3 + * 1.2.5rc1-3 13 10205 12.so.0.1.2.5rc1-3 + * 1.0.15 10 10015 10.so.0.1.0.15 + * 1.2.5 13 10205 12.so.0.1.2.5 + * 1.2.6beta1-4 13 10206 12.so.0.1.2.6beta1-4 + * 1.0.16 10 10016 10.so.0.1.0.16 + * 1.2.6 13 10206 12.so.0.1.2.6 + * 1.2.7beta1-2 13 10207 12.so.0.1.2.7beta1-2 + * 1.0.17rc1 10 10017 12.so.0.1.0.17rc1 + * 1.2.7rc1 13 10207 12.so.0.1.2.7rc1 + * 1.0.17 10 10017 12.so.0.1.0.17 + * 1.2.7 13 10207 12.so.0.1.2.7 + * 1.2.8beta1-5 13 10208 12.so.0.1.2.8beta1-5 + * 1.0.18rc1-5 10 10018 12.so.0.1.0.18rc1-5 + * 1.2.8rc1-5 13 10208 12.so.0.1.2.8rc1-5 + * 1.0.18 10 10018 12.so.0.1.0.18 + * 1.2.8 13 10208 12.so.0.1.2.8 + * 1.2.9beta1-3 13 10209 12.so.0.1.2.9beta1-3 + * 1.2.9beta4-11 13 10209 12.so.0.9[.0] + * 1.2.9rc1 13 10209 12.so.0.9[.0] + * 1.2.9 13 10209 12.so.0.9[.0] + * 1.2.10beta1-7 13 10210 12.so.0.10[.0] + * 1.2.10rc1-2 13 10210 12.so.0.10[.0] + * 1.2.10 13 10210 12.so.0.10[.0] + * 1.4.0beta1-5 14 10400 14.so.0.0[.0] + * 1.2.11beta1-4 13 10211 12.so.0.11[.0] + * 1.4.0beta7-8 14 10400 14.so.0.0[.0] + * 1.2.11 13 10211 12.so.0.11[.0] + * 1.2.12 13 10212 12.so.0.12[.0] + * 1.4.0beta9-14 14 10400 14.so.0.0[.0] + * 1.2.13 13 10213 12.so.0.13[.0] + * 1.4.0beta15-36 14 10400 14.so.0.0[.0] + * 1.4.0beta37-87 14 10400 14.so.14.0[.0] + * 1.4.0rc01 14 10400 14.so.14.0[.0] + * 1.4.0beta88-109 14 10400 14.so.14.0[.0] + * 1.4.0rc02-08 14 10400 14.so.14.0[.0] + * 1.4.0 14 10400 14.so.14.0[.0] + * 1.4.1beta01-03 14 10401 14.so.14.1[.0] + * 1.4.1rc01 14 10401 14.so.14.1[.0] + * 1.4.1beta04-12 14 10401 14.so.14.1[.0] + * 1.4.1 14 10401 14.so.14.1[.0] + * 1.4.2 14 10402 14.so.14.2[.0] + * 1.4.3 14 10403 14.so.14.3[.0] + * 1.4.4 14 10404 14.so.14.4[.0] + * 1.5.0beta01-58 15 10500 15.so.15.0[.0] + * 1.5.0rc01-07 15 10500 15.so.15.0[.0] + * 1.5.0 15 10500 15.so.15.0[.0] + * 1.5.1beta01-11 15 10501 15.so.15.1[.0] + * 1.5.1rc01-02 15 10501 15.so.15.1[.0] + * 1.5.1 15 10501 15.so.15.1[.0] + * 1.5.2beta01-03 15 10502 15.so.15.2[.0] + * 1.5.2rc01-03 15 10502 15.so.15.2[.0] + * 1.5.2 15 10502 15.so.15.2[.0] + * 1.5.3beta01-10 15 10503 15.so.15.3[.0] + * 1.5.3rc01-02 15 10503 15.so.15.3[.0] + * 1.5.3beta11 15 10503 15.so.15.3[.0] + * 1.5.3 [omitted] + * 1.5.4beta01-08 15 10504 15.so.15.4[.0] + * 1.5.4rc01 15 10504 15.so.15.4[.0] + * 1.5.4 15 10504 15.so.15.4[.0] + * 1.5.5beta01-08 15 10505 15.so.15.5[.0] + * 1.5.5rc01 15 10505 15.so.15.5[.0] + * 1.5.5 15 10505 15.so.15.5[.0] + * 1.5.6beta01-07 15 10506 15.so.15.6[.0] + * 1.5.6rc01-03 15 10506 15.so.15.6[.0] + * 1.5.6 15 10506 15.so.15.6[.0] + * 1.5.7beta01-05 15 10507 15.so.15.7[.0] + * 1.5.7rc01-03 15 10507 15.so.15.7[.0] + * 1.5.7 15 10507 15.so.15.7[.0] + * 1.6.0beta01-40 16 10600 16.so.16.0[.0] + * 1.6.0rc01-08 16 10600 16.so.16.0[.0] + * 1.6.0 16 10600 16.so.16.0[.0] + * 1.6.1beta01-09 16 10601 16.so.16.1[.0] + * 1.6.1rc01 16 10601 16.so.16.1[.0] + * 1.6.1 16 10601 16.so.16.1[.0] + * 1.6.2beta01 16 10602 16.so.16.2[.0] + * 1.6.2rc01-06 16 10602 16.so.16.2[.0] + * 1.6.2 16 10602 16.so.16.2[.0] + * 1.6.3beta01-11 16 10603 16.so.16.3[.0] + * 1.6.3rc01 16 10603 16.so.16.3[.0] + * 1.6.3 16 10603 16.so.16.3[.0] + * 1.6.4beta01-02 16 10604 16.so.16.4[.0] + * 1.6.4rc01 16 10604 16.so.16.4[.0] + * 1.6.4 16 10604 16.so.16.4[.0] + * 1.6.5 16 10605 16.so.16.5[.0] + * 1.6.6 16 10606 16.so.16.6[.0] + * 1.6.7beta01-04 16 10607 16.so.16.7[.0] + * 1.6.7rc01-03 16 10607 16.so.16.7[.0] + * 1.6.7 16 10607 16.so.16.7[.0] + * 1.6.8beta01-02 16 10608 16.so.16.8[.0] + * 1.6.8rc01-02 16 10608 16.so.16.8[.0] + * 1.6.8 16 10608 16.so.16.8[.0] + * 1.6.9beta01-04 16 10609 16.so.16.9[.0] + * 1.6.9rc01-02 16 10609 16.so.16.9[.0] + * 1.6.9 16 10609 16.so.16.9[.0] + * 1.6.10beta01-03 16 10610 16.so.16.10[.0] + * 1.6.10rc01-03 16 10610 16.so.16.10[.0] + * 1.6.10 16 10610 16.so.16.10[.0] + * 1.6.11beta01-06 16 10611 16.so.16.11[.0] + * 1.6.11rc01-02 16 10611 16.so.16.11[.0] + * 1.6.11 16 10611 16.so.16.11[.0] + * 1.6.12rc01-03 16 10612 16.so.16.12[.0] + * 1.6.12 16 10612 16.so.16.12[.0] + * 1.6.13beta01-04 16 10613 16.so.16.13[.0] + * 1.6.13rc01-02 16 10613 16.so.16.13[.0] + * 1.6.13 16 10613 16.so.16.13[.0] + * 1.6.14beta01-07 16 10614 16.so.16.14[.0] + * 1.6.14rc01-02 16 10614 16.so.16.14[.0] + * 1.6.14 16 10614 16.so.16.14[.0] + * 1.6.15beta01-08 16 10615 16.so.16.15[.0] + * 1.6.15rc01-03 16 10615 16.so.16.15[.0] + * 1.6.15 16 10615 16.so.16.15[.0] + * 1.6.16beta01-03 16 10616 16.so.16.16[.0] + * 1.6.16rc01-02 16 10616 16.so.16.16[.0] + * 1.6.16 16 10616 16.so.16.16[.0] + * + * Henceforth the source version will match the shared-library major + * and minor numbers; the shared-library major version number will be + * used for changes in backward compatibility, as it is intended. The + * PNG_LIBPNG_VER macro, which is not used within libpng but is available + * for applications, is an unsigned integer of the form xyyzz corresponding + * to the source version x.y.z (leading zeros in y and z). Beta versions + * were given the previous public release number plus a letter, until + * version 1.0.6j; from then on they were given the upcoming public + * release number plus "betaNN" or "rcNN". + * + * Binary incompatibility exists only when applications make direct access + * to the info_ptr or png_ptr members through png.h, and the compiled + * application is loaded with a different version of the library. + * + * DLLNUM will change each time there are forward or backward changes + * in binary compatibility (e.g., when a new feature is added). + * + * See libpng-manual.txt or libpng.3 for more information. The PNG + * specification is available as a W3C Recommendation and as an ISO + * Specification, + * + * If you just need to read a PNG file and don't want to read the documentation + * skip to the end of this file and read the section entitled 'simplified API'. + */ + +/* Version information for png.h - this should match the version in png.c */ +#define PNG_LIBPNG_VER_STRING "1.6.16" +#define PNG_HEADER_VERSION_STRING \ + " libpng version 1.6.16 - December 22, 2014\n" + +#define PNG_LIBPNG_VER_SONUM 16 +#define PNG_LIBPNG_VER_DLLNUM 16 + +/* These should match the first 3 components of PNG_LIBPNG_VER_STRING: */ +#define PNG_LIBPNG_VER_MAJOR 1 +#define PNG_LIBPNG_VER_MINOR 6 +#define PNG_LIBPNG_VER_RELEASE 16 + +/* This should match the numeric part of the final component of + * PNG_LIBPNG_VER_STRING, omitting any leading zero: + */ + +#define PNG_LIBPNG_VER_BUILD 0 + +/* Release Status */ +#define PNG_LIBPNG_BUILD_ALPHA 1 +#define PNG_LIBPNG_BUILD_BETA 2 +#define PNG_LIBPNG_BUILD_RC 3 +#define PNG_LIBPNG_BUILD_STABLE 4 +#define PNG_LIBPNG_BUILD_RELEASE_STATUS_MASK 7 + +/* Release-Specific Flags */ +#define PNG_LIBPNG_BUILD_PATCH 8 /* Can be OR'ed with + PNG_LIBPNG_BUILD_STABLE only */ +#define PNG_LIBPNG_BUILD_PRIVATE 16 /* Cannot be OR'ed with + PNG_LIBPNG_BUILD_SPECIAL */ +#define PNG_LIBPNG_BUILD_SPECIAL 32 /* Cannot be OR'ed with + PNG_LIBPNG_BUILD_PRIVATE */ + +#define PNG_LIBPNG_BUILD_BASE_TYPE PNG_LIBPNG_BUILD_STABLE + +/* Careful here. At one time, Guy wanted to use 082, but that would be octal. + * We must not include leading zeros. + * Versions 0.7 through 1.0.0 were in the range 0 to 100 here (only + * version 1.0.0 was mis-numbered 100 instead of 10000). From + * version 1.0.1 it's xxyyzz, where x=major, y=minor, z=release + */ +#define PNG_LIBPNG_VER 10616 /* 1.6.16 */ + +/* Library configuration: these options cannot be changed after + * the library has been built. + */ +#ifndef PNGLCONF_H + /* If pnglibconf.h is missing, you can + * copy scripts/pnglibconf.h.prebuilt to pnglibconf.h + */ +# include "pnglibconf.h" +#endif + +#ifndef PNG_VERSION_INFO_ONLY + /* Machine specific configuration. */ +# include "pngconf.h" +#endif + +/* + * Added at libpng-1.2.8 + * + * Ref MSDN: Private as priority over Special + * VS_FF_PRIVATEBUILD File *was not* built using standard release + * procedures. If this value is given, the StringFileInfo block must + * contain a PrivateBuild string. + * + * VS_FF_SPECIALBUILD File *was* built by the original company using + * standard release procedures but is a variation of the standard + * file of the same version number. If this value is given, the + * StringFileInfo block must contain a SpecialBuild string. + */ + +#ifdef PNG_USER_PRIVATEBUILD /* From pnglibconf.h */ +# define PNG_LIBPNG_BUILD_TYPE \ + (PNG_LIBPNG_BUILD_BASE_TYPE | PNG_LIBPNG_BUILD_PRIVATE) +#else +# ifdef PNG_LIBPNG_SPECIALBUILD +# define PNG_LIBPNG_BUILD_TYPE \ + (PNG_LIBPNG_BUILD_BASE_TYPE | PNG_LIBPNG_BUILD_SPECIAL) +# else +# define PNG_LIBPNG_BUILD_TYPE (PNG_LIBPNG_BUILD_BASE_TYPE) +# endif +#endif + +#ifndef PNG_VERSION_INFO_ONLY + +/* Inhibit C++ name-mangling for libpng functions but not for system calls. */ +#ifdef __cplusplus +extern "C" { +#endif /* __cplusplus */ + +/* Version information for C files, stored in png.c. This had better match + * the version above. + */ +#define png_libpng_ver png_get_header_ver(NULL) + +/* This file is arranged in several sections: + * + * 1. Any configuration options that can be specified by for the application + * code when it is built. (Build time configuration is in pnglibconf.h) + * 2. Type definitions (base types are defined in pngconf.h), structure + * definitions. + * 3. Exported library functions. + * 4. Simplified API. + * + * The library source code has additional files (principally pngpriv.h) that + * allow configuration of the library. + */ +/* Section 1: run time configuration + * See pnglibconf.h for build time configuration + * + * Run time configuration allows the application to choose between + * implementations of certain arithmetic APIs. The default is set + * at build time and recorded in pnglibconf.h, but it is safe to + * override these (and only these) settings. Note that this won't + * change what the library does, only application code, and the + * settings can (and probably should) be made on a per-file basis + * by setting the #defines before including png.h + * + * Use macros to read integers from PNG data or use the exported + * functions? + * PNG_USE_READ_MACROS: use the macros (see below) Note that + * the macros evaluate their argument multiple times. + * PNG_NO_USE_READ_MACROS: call the relevant library function. + * + * Use the alternative algorithm for compositing alpha samples that + * does not use division? + * PNG_READ_COMPOSITE_NODIV_SUPPORTED: use the 'no division' + * algorithm. + * PNG_NO_READ_COMPOSITE_NODIV: use the 'division' algorithm. + * + * How to handle benign errors if PNG_ALLOW_BENIGN_ERRORS is + * false? + * PNG_ALLOW_BENIGN_ERRORS: map calls to the benign error + * APIs to png_warning. + * Otherwise the calls are mapped to png_error. + */ + +/* Section 2: type definitions, including structures and compile time + * constants. + * See pngconf.h for base types that vary by machine/system + */ + +/* This triggers a compiler error in png.c, if png.c and png.h + * do not agree upon the version number. + */ +typedef char* png_libpng_version_1_6_16; + +/* Basic control structions. Read libpng-manual.txt or libpng.3 for more info. + * + * png_struct is the cache of information used while reading or writing a single + * PNG file. One of these is always required, although the simplified API + * (below) hides the creation and destruction of it. + */ +typedef struct png_struct_def png_struct; +typedef const png_struct * png_const_structp; +typedef png_struct * png_structp; +typedef png_struct * * png_structpp; + +/* png_info contains information read from or to be written to a PNG file. One + * or more of these must exist while reading or creating a PNG file. The + * information is not used by libpng during read but is used to control what + * gets written when a PNG file is created. "png_get_" function calls read + * information during read and "png_set_" functions calls write information + * when creating a PNG. + * been moved into a separate header file that is not accessible to + * applications. Read libpng-manual.txt or libpng.3 for more info. + */ +typedef struct png_info_def png_info; +typedef png_info * png_infop; +typedef const png_info * png_const_infop; +typedef png_info * * png_infopp; + +/* Types with names ending 'p' are pointer types. The corresponding types with + * names ending 'rp' are identical pointer types except that the pointer is + * marked 'restrict', which means that it is the only pointer to the object + * passed to the function. Applications should not use the 'restrict' types; + * it is always valid to pass 'p' to a pointer with a function argument of the + * corresponding 'rp' type. Different compilers have different rules with + * regard to type matching in the presence of 'restrict'. For backward + * compatibility libpng callbacks never have 'restrict' in their parameters and, + * consequentially, writing portable application code is extremely difficult if + * an attempt is made to use 'restrict'. + */ +typedef png_struct * PNG_RESTRICT png_structrp; +typedef const png_struct * PNG_RESTRICT png_const_structrp; +typedef png_info * PNG_RESTRICT png_inforp; +typedef const png_info * PNG_RESTRICT png_const_inforp; + +/* Three color definitions. The order of the red, green, and blue, (and the + * exact size) is not important, although the size of the fields need to + * be png_byte or png_uint_16 (as defined below). + */ +typedef struct png_color_struct +{ + png_byte red; + png_byte green; + png_byte blue; +} png_color; +typedef png_color * png_colorp; +typedef const png_color * png_const_colorp; +typedef png_color * * png_colorpp; + +typedef struct png_color_16_struct +{ + png_byte index; /* used for palette files */ + png_uint_16 red; /* for use in red green blue files */ + png_uint_16 green; + png_uint_16 blue; + png_uint_16 gray; /* for use in grayscale files */ +} png_color_16; +typedef png_color_16 * png_color_16p; +typedef const png_color_16 * png_const_color_16p; +typedef png_color_16 * * png_color_16pp; + +typedef struct png_color_8_struct +{ + png_byte red; /* for use in red green blue files */ + png_byte green; + png_byte blue; + png_byte gray; /* for use in grayscale files */ + png_byte alpha; /* for alpha channel files */ +} png_color_8; +typedef png_color_8 * png_color_8p; +typedef const png_color_8 * png_const_color_8p; +typedef png_color_8 * * png_color_8pp; + +/* + * The following two structures are used for the in-core representation + * of sPLT chunks. + */ +typedef struct png_sPLT_entry_struct +{ + png_uint_16 red; + png_uint_16 green; + png_uint_16 blue; + png_uint_16 alpha; + png_uint_16 frequency; +} png_sPLT_entry; +typedef png_sPLT_entry * png_sPLT_entryp; +typedef const png_sPLT_entry * png_const_sPLT_entryp; +typedef png_sPLT_entry * * png_sPLT_entrypp; + +/* When the depth of the sPLT palette is 8 bits, the color and alpha samples + * occupy the LSB of their respective members, and the MSB of each member + * is zero-filled. The frequency member always occupies the full 16 bits. + */ + +typedef struct png_sPLT_struct +{ + png_charp name; /* palette name */ + png_byte depth; /* depth of palette samples */ + png_sPLT_entryp entries; /* palette entries */ + png_int_32 nentries; /* number of palette entries */ +} png_sPLT_t; +typedef png_sPLT_t * png_sPLT_tp; +typedef const png_sPLT_t * png_const_sPLT_tp; +typedef png_sPLT_t * * png_sPLT_tpp; + +#ifdef PNG_TEXT_SUPPORTED +/* png_text holds the contents of a text/ztxt/itxt chunk in a PNG file, + * and whether that contents is compressed or not. The "key" field + * points to a regular zero-terminated C string. The "text" fields can be a + * regular C string, an empty string, or a NULL pointer. + * However, the structure returned by png_get_text() will always contain + * the "text" field as a regular zero-terminated C string (possibly + * empty), never a NULL pointer, so it can be safely used in printf() and + * other string-handling functions. Note that the "itxt_length", "lang", and + * "lang_key" members of the structure only exist when the library is built + * with iTXt chunk support. Prior to libpng-1.4.0 the library was built by + * default without iTXt support. Also note that when iTXt *is* supported, + * the "lang" and "lang_key" fields contain NULL pointers when the + * "compression" field contains * PNG_TEXT_COMPRESSION_NONE or + * PNG_TEXT_COMPRESSION_zTXt. Note that the "compression value" is not the + * same as what appears in the PNG tEXt/zTXt/iTXt chunk's "compression flag" + * which is always 0 or 1, or its "compression method" which is always 0. + */ +typedef struct png_text_struct +{ + int compression; /* compression value: + -1: tEXt, none + 0: zTXt, deflate + 1: iTXt, none + 2: iTXt, deflate */ + png_charp key; /* keyword, 1-79 character description of "text" */ + png_charp text; /* comment, may be an empty string (ie "") + or a NULL pointer */ + png_size_t text_length; /* length of the text string */ + png_size_t itxt_length; /* length of the itxt string */ + png_charp lang; /* language code, 0-79 characters + or a NULL pointer */ + png_charp lang_key; /* keyword translated UTF-8 string, 0 or more + chars or a NULL pointer */ +} png_text; +typedef png_text * png_textp; +typedef const png_text * png_const_textp; +typedef png_text * * png_textpp; +#endif + +/* Supported compression types for text in PNG files (tEXt, and zTXt). + * The values of the PNG_TEXT_COMPRESSION_ defines should NOT be changed. */ +#define PNG_TEXT_COMPRESSION_NONE_WR -3 +#define PNG_TEXT_COMPRESSION_zTXt_WR -2 +#define PNG_TEXT_COMPRESSION_NONE -1 +#define PNG_TEXT_COMPRESSION_zTXt 0 +#define PNG_ITXT_COMPRESSION_NONE 1 +#define PNG_ITXT_COMPRESSION_zTXt 2 +#define PNG_TEXT_COMPRESSION_LAST 3 /* Not a valid value */ + +/* png_time is a way to hold the time in an machine independent way. + * Two conversions are provided, both from time_t and struct tm. There + * is no portable way to convert to either of these structures, as far + * as I know. If you know of a portable way, send it to me. As a side + * note - PNG has always been Year 2000 compliant! + */ +typedef struct png_time_struct +{ + png_uint_16 year; /* full year, as in, 1995 */ + png_byte month; /* month of year, 1 - 12 */ + png_byte day; /* day of month, 1 - 31 */ + png_byte hour; /* hour of day, 0 - 23 */ + png_byte minute; /* minute of hour, 0 - 59 */ + png_byte second; /* second of minute, 0 - 60 (for leap seconds) */ +} png_time; +typedef png_time * png_timep; +typedef const png_time * png_const_timep; +typedef png_time * * png_timepp; + +#if defined(PNG_STORE_UNKNOWN_CHUNKS_SUPPORTED) ||\ + defined(PNG_USER_CHUNKS_SUPPORTED) +/* png_unknown_chunk is a structure to hold queued chunks for which there is + * no specific support. The idea is that we can use this to queue + * up private chunks for output even though the library doesn't actually + * know about their semantics. + * + * The data in the structure is set by libpng on read and used on write. + */ +typedef struct png_unknown_chunk_t +{ + png_byte name[5]; /* Textual chunk name with '\0' terminator */ + png_byte *data; /* Data, should not be modified on read! */ + png_size_t size; + + /* On write 'location' must be set using the flag values listed below. + * Notice that on read it is set by libpng however the values stored have + * more bits set than are listed below. Always treat the value as a + * bitmask. On write set only one bit - setting multiple bits may cause the + * chunk to be written in multiple places. + */ + png_byte location; /* mode of operation at read time */ +} +png_unknown_chunk; + +typedef png_unknown_chunk * png_unknown_chunkp; +typedef const png_unknown_chunk * png_const_unknown_chunkp; +typedef png_unknown_chunk * * png_unknown_chunkpp; +#endif + +/* Flag values for the unknown chunk location byte. */ +#define PNG_HAVE_IHDR 0x01 +#define PNG_HAVE_PLTE 0x02 +#define PNG_AFTER_IDAT 0x08 + +/* Maximum positive integer used in PNG is (2^31)-1 */ +#define PNG_UINT_31_MAX ((png_uint_32)0x7fffffffL) +#define PNG_UINT_32_MAX ((png_uint_32)(-1)) +#define PNG_SIZE_MAX ((png_size_t)(-1)) + +/* These are constants for fixed point values encoded in the + * PNG specification manner (x100000) + */ +#define PNG_FP_1 100000 +#define PNG_FP_HALF 50000 +#define PNG_FP_MAX ((png_fixed_point)0x7fffffffL) +#define PNG_FP_MIN (-PNG_FP_MAX) + +/* These describe the color_type field in png_info. */ +/* color type masks */ +#define PNG_COLOR_MASK_PALETTE 1 +#define PNG_COLOR_MASK_COLOR 2 +#define PNG_COLOR_MASK_ALPHA 4 + +/* color types. Note that not all combinations are legal */ +#define PNG_COLOR_TYPE_GRAY 0 +#define PNG_COLOR_TYPE_PALETTE (PNG_COLOR_MASK_COLOR | PNG_COLOR_MASK_PALETTE) +#define PNG_COLOR_TYPE_RGB (PNG_COLOR_MASK_COLOR) +#define PNG_COLOR_TYPE_RGB_ALPHA (PNG_COLOR_MASK_COLOR | PNG_COLOR_MASK_ALPHA) +#define PNG_COLOR_TYPE_GRAY_ALPHA (PNG_COLOR_MASK_ALPHA) +/* aliases */ +#define PNG_COLOR_TYPE_RGBA PNG_COLOR_TYPE_RGB_ALPHA +#define PNG_COLOR_TYPE_GA PNG_COLOR_TYPE_GRAY_ALPHA + +/* This is for compression type. PNG 1.0-1.2 only define the single type. */ +#define PNG_COMPRESSION_TYPE_BASE 0 /* Deflate method 8, 32K window */ +#define PNG_COMPRESSION_TYPE_DEFAULT PNG_COMPRESSION_TYPE_BASE + +/* This is for filter type. PNG 1.0-1.2 only define the single type. */ +#define PNG_FILTER_TYPE_BASE 0 /* Single row per-byte filtering */ +#define PNG_INTRAPIXEL_DIFFERENCING 64 /* Used only in MNG datastreams */ +#define PNG_FILTER_TYPE_DEFAULT PNG_FILTER_TYPE_BASE + +/* These are for the interlacing type. These values should NOT be changed. */ +#define PNG_INTERLACE_NONE 0 /* Non-interlaced image */ +#define PNG_INTERLACE_ADAM7 1 /* Adam7 interlacing */ +#define PNG_INTERLACE_LAST 2 /* Not a valid value */ + +/* These are for the oFFs chunk. These values should NOT be changed. */ +#define PNG_OFFSET_PIXEL 0 /* Offset in pixels */ +#define PNG_OFFSET_MICROMETER 1 /* Offset in micrometers (1/10^6 meter) */ +#define PNG_OFFSET_LAST 2 /* Not a valid value */ + +/* These are for the pCAL chunk. These values should NOT be changed. */ +#define PNG_EQUATION_LINEAR 0 /* Linear transformation */ +#define PNG_EQUATION_BASE_E 1 /* Exponential base e transform */ +#define PNG_EQUATION_ARBITRARY 2 /* Arbitrary base exponential transform */ +#define PNG_EQUATION_HYPERBOLIC 3 /* Hyperbolic sine transformation */ +#define PNG_EQUATION_LAST 4 /* Not a valid value */ + +/* These are for the sCAL chunk. These values should NOT be changed. */ +#define PNG_SCALE_UNKNOWN 0 /* unknown unit (image scale) */ +#define PNG_SCALE_METER 1 /* meters per pixel */ +#define PNG_SCALE_RADIAN 2 /* radians per pixel */ +#define PNG_SCALE_LAST 3 /* Not a valid value */ + +/* These are for the pHYs chunk. These values should NOT be changed. */ +#define PNG_RESOLUTION_UNKNOWN 0 /* pixels/unknown unit (aspect ratio) */ +#define PNG_RESOLUTION_METER 1 /* pixels/meter */ +#define PNG_RESOLUTION_LAST 2 /* Not a valid value */ + +/* These are for the sRGB chunk. These values should NOT be changed. */ +#define PNG_sRGB_INTENT_PERCEPTUAL 0 +#define PNG_sRGB_INTENT_RELATIVE 1 +#define PNG_sRGB_INTENT_SATURATION 2 +#define PNG_sRGB_INTENT_ABSOLUTE 3 +#define PNG_sRGB_INTENT_LAST 4 /* Not a valid value */ + +/* This is for text chunks */ +#define PNG_KEYWORD_MAX_LENGTH 79 + +/* Maximum number of entries in PLTE/sPLT/tRNS arrays */ +#define PNG_MAX_PALETTE_LENGTH 256 + +/* These determine if an ancillary chunk's data has been successfully read + * from the PNG header, or if the application has filled in the corresponding + * data in the info_struct to be written into the output file. The values + * of the PNG_INFO_ defines should NOT be changed. + */ +#define PNG_INFO_gAMA 0x0001 +#define PNG_INFO_sBIT 0x0002 +#define PNG_INFO_cHRM 0x0004 +#define PNG_INFO_PLTE 0x0008 +#define PNG_INFO_tRNS 0x0010 +#define PNG_INFO_bKGD 0x0020 +#define PNG_INFO_hIST 0x0040 +#define PNG_INFO_pHYs 0x0080 +#define PNG_INFO_oFFs 0x0100 +#define PNG_INFO_tIME 0x0200 +#define PNG_INFO_pCAL 0x0400 +#define PNG_INFO_sRGB 0x0800 /* GR-P, 0.96a */ +#define PNG_INFO_iCCP 0x1000 /* ESR, 1.0.6 */ +#define PNG_INFO_sPLT 0x2000 /* ESR, 1.0.6 */ +#define PNG_INFO_sCAL 0x4000 /* ESR, 1.0.6 */ +#define PNG_INFO_IDAT 0x8000 /* ESR, 1.0.6 */ + +/* This is used for the transformation routines, as some of them + * change these values for the row. It also should enable using + * the routines for other purposes. + */ +typedef struct png_row_info_struct +{ + png_uint_32 width; /* width of row */ + png_size_t rowbytes; /* number of bytes in row */ + png_byte color_type; /* color type of row */ + png_byte bit_depth; /* bit depth of row */ + png_byte channels; /* number of channels (1, 2, 3, or 4) */ + png_byte pixel_depth; /* bits per pixel (depth * channels) */ +} png_row_info; + +typedef png_row_info * png_row_infop; +typedef png_row_info * * png_row_infopp; + +/* These are the function types for the I/O functions and for the functions + * that allow the user to override the default I/O functions with his or her + * own. The png_error_ptr type should match that of user-supplied warning + * and error functions, while the png_rw_ptr type should match that of the + * user read/write data functions. Note that the 'write' function must not + * modify the buffer it is passed. The 'read' function, on the other hand, is + * expected to return the read data in the buffer. + */ +typedef PNG_CALLBACK(void, *png_error_ptr, (png_structp, png_const_charp)); +typedef PNG_CALLBACK(void, *png_rw_ptr, (png_structp, png_bytep, png_size_t)); +typedef PNG_CALLBACK(void, *png_flush_ptr, (png_structp)); +typedef PNG_CALLBACK(void, *png_read_status_ptr, (png_structp, png_uint_32, + int)); +typedef PNG_CALLBACK(void, *png_write_status_ptr, (png_structp, png_uint_32, + int)); + +#ifdef PNG_PROGRESSIVE_READ_SUPPORTED +typedef PNG_CALLBACK(void, *png_progressive_info_ptr, (png_structp, png_infop)); +typedef PNG_CALLBACK(void, *png_progressive_end_ptr, (png_structp, png_infop)); + +/* The following callback receives png_uint_32 row_number, int pass for the + * png_bytep data of the row. When transforming an interlaced image the + * row number is the row number within the sub-image of the interlace pass, so + * the value will increase to the height of the sub-image (not the full image) + * then reset to 0 for the next pass. + * + * Use PNG_ROW_FROM_PASS_ROW(row, pass) and PNG_COL_FROM_PASS_COL(col, pass) to + * find the output pixel (x,y) given an interlaced sub-image pixel + * (row,col,pass). (See below for these macros.) + */ +typedef PNG_CALLBACK(void, *png_progressive_row_ptr, (png_structp, png_bytep, + png_uint_32, int)); +#endif + +#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED) || \ + defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED) +typedef PNG_CALLBACK(void, *png_user_transform_ptr, (png_structp, png_row_infop, + png_bytep)); +#endif + +#ifdef PNG_USER_CHUNKS_SUPPORTED +typedef PNG_CALLBACK(int, *png_user_chunk_ptr, (png_structp, + png_unknown_chunkp)); +#endif +#ifdef PNG_UNKNOWN_CHUNKS_SUPPORTED +/* not used anywhere */ +/* typedef PNG_CALLBACK(void, *png_unknown_chunk_ptr, (png_structp)); */ +#endif + +#ifdef PNG_SETJMP_SUPPORTED +/* This must match the function definition in , and the application + * must include this before png.h to obtain the definition of jmp_buf. The + * function is required to be PNG_NORETURN, but this is not checked. If the + * function does return the application will crash via an abort() or similar + * system level call. + * + * If you get a warning here while building the library you may need to make + * changes to ensure that pnglibconf.h records the calling convention used by + * your compiler. This may be very difficult - try using a different compiler + * to build the library! + */ +PNG_FUNCTION(void, (PNGCAPI *png_longjmp_ptr), PNGARG((jmp_buf, int)), typedef); +#endif + +/* Transform masks for the high-level interface */ +#define PNG_TRANSFORM_IDENTITY 0x0000 /* read and write */ +#define PNG_TRANSFORM_STRIP_16 0x0001 /* read only */ +#define PNG_TRANSFORM_STRIP_ALPHA 0x0002 /* read only */ +#define PNG_TRANSFORM_PACKING 0x0004 /* read and write */ +#define PNG_TRANSFORM_PACKSWAP 0x0008 /* read and write */ +#define PNG_TRANSFORM_EXPAND 0x0010 /* read only */ +#define PNG_TRANSFORM_INVERT_MONO 0x0020 /* read and write */ +#define PNG_TRANSFORM_SHIFT 0x0040 /* read and write */ +#define PNG_TRANSFORM_BGR 0x0080 /* read and write */ +#define PNG_TRANSFORM_SWAP_ALPHA 0x0100 /* read and write */ +#define PNG_TRANSFORM_SWAP_ENDIAN 0x0200 /* read and write */ +#define PNG_TRANSFORM_INVERT_ALPHA 0x0400 /* read and write */ +#define PNG_TRANSFORM_STRIP_FILLER 0x0800 /* write only */ +/* Added to libpng-1.2.34 */ +#define PNG_TRANSFORM_STRIP_FILLER_BEFORE PNG_TRANSFORM_STRIP_FILLER +#define PNG_TRANSFORM_STRIP_FILLER_AFTER 0x1000 /* write only */ +/* Added to libpng-1.4.0 */ +#define PNG_TRANSFORM_GRAY_TO_RGB 0x2000 /* read only */ +/* Added to libpng-1.5.4 */ +#define PNG_TRANSFORM_EXPAND_16 0x4000 /* read only */ +#define PNG_TRANSFORM_SCALE_16 0x8000 /* read only */ + +/* Flags for MNG supported features */ +#define PNG_FLAG_MNG_EMPTY_PLTE 0x01 +#define PNG_FLAG_MNG_FILTER_64 0x04 +#define PNG_ALL_MNG_FEATURES 0x05 + +/* NOTE: prior to 1.5 these functions had no 'API' style declaration, + * this allowed the zlib default functions to be used on Windows + * platforms. In 1.5 the zlib default malloc (which just calls malloc and + * ignores the first argument) should be completely compatible with the + * following. + */ +typedef PNG_CALLBACK(png_voidp, *png_malloc_ptr, (png_structp, + png_alloc_size_t)); +typedef PNG_CALLBACK(void, *png_free_ptr, (png_structp, png_voidp)); + +/* Section 3: exported functions + * Here are the function definitions most commonly used. This is not + * the place to find out how to use libpng. See libpng-manual.txt for the + * full explanation, see example.c for the summary. This just provides + * a simple one line description of the use of each function. + * + * The PNG_EXPORT() and PNG_EXPORTA() macros used below are defined in + * pngconf.h and in the *.dfn files in the scripts directory. + * + * PNG_EXPORT(ordinal, type, name, (args)); + * + * ordinal: ordinal that is used while building + * *.def files. The ordinal value is only + * relevant when preprocessing png.h with + * the *.dfn files for building symbol table + * entries, and are removed by pngconf.h. + * type: return type of the function + * name: function name + * args: function arguments, with types + * + * When we wish to append attributes to a function prototype we use + * the PNG_EXPORTA() macro instead. + * + * PNG_EXPORTA(ordinal, type, name, (args), attributes); + * + * ordinal, type, name, and args: same as in PNG_EXPORT(). + * attributes: function attributes + */ + +/* Returns the version number of the library */ +PNG_EXPORT(1, png_uint_32, png_access_version_number, (void)); + +/* Tell lib we have already handled the first magic bytes. + * Handling more than 8 bytes from the beginning of the file is an error. + */ +PNG_EXPORT(2, void, png_set_sig_bytes, (png_structrp png_ptr, int num_bytes)); + +/* Check sig[start] through sig[start + num_to_check - 1] to see if it's a + * PNG file. Returns zero if the supplied bytes match the 8-byte PNG + * signature, and non-zero otherwise. Having num_to_check == 0 or + * start > 7 will always fail (ie return non-zero). + */ +PNG_EXPORT(3, int, png_sig_cmp, (png_const_bytep sig, png_size_t start, + png_size_t num_to_check)); + +/* Simple signature checking function. This is the same as calling + * png_check_sig(sig, n) := !png_sig_cmp(sig, 0, n). + */ +#define png_check_sig(sig, n) !png_sig_cmp((sig), 0, (n)) + +/* Allocate and initialize png_ptr struct for reading, and any other memory. */ +PNG_EXPORTA(4, png_structp, png_create_read_struct, + (png_const_charp user_png_ver, png_voidp error_ptr, + png_error_ptr error_fn, png_error_ptr warn_fn), + PNG_ALLOCATED); + +/* Allocate and initialize png_ptr struct for writing, and any other memory */ +PNG_EXPORTA(5, png_structp, png_create_write_struct, + (png_const_charp user_png_ver, png_voidp error_ptr, png_error_ptr error_fn, + png_error_ptr warn_fn), + PNG_ALLOCATED); + +PNG_EXPORT(6, png_size_t, png_get_compression_buffer_size, + (png_const_structrp png_ptr)); + +PNG_EXPORT(7, void, png_set_compression_buffer_size, (png_structrp png_ptr, + png_size_t size)); + +/* Moved from pngconf.h in 1.4.0 and modified to ensure setjmp/longjmp + * match up. + */ +#ifdef PNG_SETJMP_SUPPORTED +/* This function returns the jmp_buf built in to *png_ptr. It must be + * supplied with an appropriate 'longjmp' function to use on that jmp_buf + * unless the default error function is overridden in which case NULL is + * acceptable. The size of the jmp_buf is checked against the actual size + * allocated by the library - the call will return NULL on a mismatch + * indicating an ABI mismatch. + */ +PNG_EXPORT(8, jmp_buf*, png_set_longjmp_fn, (png_structrp png_ptr, + png_longjmp_ptr longjmp_fn, size_t jmp_buf_size)); +# define png_jmpbuf(png_ptr) \ + (*png_set_longjmp_fn((png_ptr), longjmp, (sizeof (jmp_buf)))) +#else +# define png_jmpbuf(png_ptr) \ + (LIBPNG_WAS_COMPILED_WITH__PNG_NO_SETJMP) +#endif +/* This function should be used by libpng applications in place of + * longjmp(png_ptr->jmpbuf, val). If longjmp_fn() has been set, it + * will use it; otherwise it will call PNG_ABORT(). This function was + * added in libpng-1.5.0. + */ +PNG_EXPORTA(9, void, png_longjmp, (png_const_structrp png_ptr, int val), + PNG_NORETURN); + +#ifdef PNG_READ_SUPPORTED +/* Reset the compression stream */ +PNG_EXPORTA(10, int, png_reset_zstream, (png_structrp png_ptr), PNG_DEPRECATED); +#endif + +/* New functions added in libpng-1.0.2 (not enabled by default until 1.2.0) */ +#ifdef PNG_USER_MEM_SUPPORTED +PNG_EXPORTA(11, png_structp, png_create_read_struct_2, + (png_const_charp user_png_ver, png_voidp error_ptr, png_error_ptr error_fn, + png_error_ptr warn_fn, + png_voidp mem_ptr, png_malloc_ptr malloc_fn, png_free_ptr free_fn), + PNG_ALLOCATED); +PNG_EXPORTA(12, png_structp, png_create_write_struct_2, + (png_const_charp user_png_ver, png_voidp error_ptr, png_error_ptr error_fn, + png_error_ptr warn_fn, + png_voidp mem_ptr, png_malloc_ptr malloc_fn, png_free_ptr free_fn), + PNG_ALLOCATED); +#endif + +/* Write the PNG file signature. */ +PNG_EXPORT(13, void, png_write_sig, (png_structrp png_ptr)); + +/* Write a PNG chunk - size, type, (optional) data, CRC. */ +PNG_EXPORT(14, void, png_write_chunk, (png_structrp png_ptr, png_const_bytep + chunk_name, png_const_bytep data, png_size_t length)); + +/* Write the start of a PNG chunk - length and chunk name. */ +PNG_EXPORT(15, void, png_write_chunk_start, (png_structrp png_ptr, + png_const_bytep chunk_name, png_uint_32 length)); + +/* Write the data of a PNG chunk started with png_write_chunk_start(). */ +PNG_EXPORT(16, void, png_write_chunk_data, (png_structrp png_ptr, + png_const_bytep data, png_size_t length)); + +/* Finish a chunk started with png_write_chunk_start() (includes CRC). */ +PNG_EXPORT(17, void, png_write_chunk_end, (png_structrp png_ptr)); + +/* Allocate and initialize the info structure */ +PNG_EXPORTA(18, png_infop, png_create_info_struct, (png_const_structrp png_ptr), + PNG_ALLOCATED); + +/* DEPRECATED: this function allowed init structures to be created using the + * default allocation method (typically malloc). Use is deprecated in 1.6.0 and + * the API will be removed in the future. + */ +PNG_EXPORTA(19, void, png_info_init_3, (png_infopp info_ptr, + png_size_t png_info_struct_size), PNG_DEPRECATED); + +/* Writes all the PNG information before the image. */ +PNG_EXPORT(20, void, png_write_info_before_PLTE, + (png_structrp png_ptr, png_const_inforp info_ptr)); +PNG_EXPORT(21, void, png_write_info, + (png_structrp png_ptr, png_const_inforp info_ptr)); + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +/* Read the information before the actual image data. */ +PNG_EXPORT(22, void, png_read_info, + (png_structrp png_ptr, png_inforp info_ptr)); +#endif + +#ifdef PNG_TIME_RFC1123_SUPPORTED + /* Convert to a US string format: there is no localization support in this + * routine. The original implementation used a 29 character buffer in + * png_struct, this will be removed in future versions. + */ +#if PNG_LIBPNG_VER < 10700 +/* To do: remove this from libpng17 (and from libpng17/png.c and pngstruct.h) */ +PNG_EXPORTA(23, png_const_charp, png_convert_to_rfc1123, (png_structrp png_ptr, + png_const_timep ptime),PNG_DEPRECATED); +#endif +PNG_EXPORT(241, int, png_convert_to_rfc1123_buffer, (char out[29], + png_const_timep ptime)); +#endif + +#ifdef PNG_CONVERT_tIME_SUPPORTED +/* Convert from a struct tm to png_time */ +PNG_EXPORT(24, void, png_convert_from_struct_tm, (png_timep ptime, + const struct tm * ttime)); + +/* Convert from time_t to png_time. Uses gmtime() */ +PNG_EXPORT(25, void, png_convert_from_time_t, (png_timep ptime, time_t ttime)); +#endif /* CONVERT_tIME */ + +#ifdef PNG_READ_EXPAND_SUPPORTED +/* Expand data to 24-bit RGB, or 8-bit grayscale, with alpha if available. */ +PNG_EXPORT(26, void, png_set_expand, (png_structrp png_ptr)); +PNG_EXPORT(27, void, png_set_expand_gray_1_2_4_to_8, (png_structrp png_ptr)); +PNG_EXPORT(28, void, png_set_palette_to_rgb, (png_structrp png_ptr)); +PNG_EXPORT(29, void, png_set_tRNS_to_alpha, (png_structrp png_ptr)); +#endif + +#ifdef PNG_READ_EXPAND_16_SUPPORTED +/* Expand to 16-bit channels, forces conversion of palette to RGB and expansion + * of a tRNS chunk if present. + */ +PNG_EXPORT(221, void, png_set_expand_16, (png_structrp png_ptr)); +#endif + +#if defined(PNG_READ_BGR_SUPPORTED) || defined(PNG_WRITE_BGR_SUPPORTED) +/* Use blue, green, red order for pixels. */ +PNG_EXPORT(30, void, png_set_bgr, (png_structrp png_ptr)); +#endif + +#ifdef PNG_READ_GRAY_TO_RGB_SUPPORTED +/* Expand the grayscale to 24-bit RGB if necessary. */ +PNG_EXPORT(31, void, png_set_gray_to_rgb, (png_structrp png_ptr)); +#endif + +#ifdef PNG_READ_RGB_TO_GRAY_SUPPORTED +/* Reduce RGB to grayscale. */ +#define PNG_ERROR_ACTION_NONE 1 +#define PNG_ERROR_ACTION_WARN 2 +#define PNG_ERROR_ACTION_ERROR 3 +#define PNG_RGB_TO_GRAY_DEFAULT (-1)/*for red/green coefficients*/ + +PNG_FP_EXPORT(32, void, png_set_rgb_to_gray, (png_structrp png_ptr, + int error_action, double red, double green)) +PNG_FIXED_EXPORT(33, void, png_set_rgb_to_gray_fixed, (png_structrp png_ptr, + int error_action, png_fixed_point red, png_fixed_point green)) + +PNG_EXPORT(34, png_byte, png_get_rgb_to_gray_status, (png_const_structrp + png_ptr)); +#endif + +#ifdef PNG_BUILD_GRAYSCALE_PALETTE_SUPPORTED +PNG_EXPORT(35, void, png_build_grayscale_palette, (int bit_depth, + png_colorp palette)); +#endif + +#ifdef PNG_READ_ALPHA_MODE_SUPPORTED +/* How the alpha channel is interpreted - this affects how the color channels + * of a PNG file are returned to the calling application when an alpha channel, + * or a tRNS chunk in a palette file, is present. + * + * This has no effect on the way pixels are written into a PNG output + * datastream. The color samples in a PNG datastream are never premultiplied + * with the alpha samples. + * + * The default is to return data according to the PNG specification: the alpha + * channel is a linear measure of the contribution of the pixel to the + * corresponding composited pixel, and the color channels are unassociated + * (not premultiplied). The gamma encoded color channels must be scaled + * according to the contribution and to do this it is necessary to undo + * the encoding, scale the color values, perform the composition and reencode + * the values. This is the 'PNG' mode. + * + * The alternative is to 'associate' the alpha with the color information by + * storing color channel values that have been scaled by the alpha. + * image. These are the 'STANDARD', 'ASSOCIATED' or 'PREMULTIPLIED' modes + * (the latter being the two common names for associated alpha color channels). + * + * For the 'OPTIMIZED' mode, a pixel is treated as opaque only if the alpha + * value is equal to the maximum value. + * + * The final choice is to gamma encode the alpha channel as well. This is + * broken because, in practice, no implementation that uses this choice + * correctly undoes the encoding before handling alpha composition. Use this + * choice only if other serious errors in the software or hardware you use + * mandate it; the typical serious error is for dark halos to appear around + * opaque areas of the composited PNG image because of arithmetic overflow. + * + * The API function png_set_alpha_mode specifies which of these choices to use + * with an enumerated 'mode' value and the gamma of the required output: + */ +#define PNG_ALPHA_PNG 0 /* according to the PNG standard */ +#define PNG_ALPHA_STANDARD 1 /* according to Porter/Duff */ +#define PNG_ALPHA_ASSOCIATED 1 /* as above; this is the normal practice */ +#define PNG_ALPHA_PREMULTIPLIED 1 /* as above */ +#define PNG_ALPHA_OPTIMIZED 2 /* 'PNG' for opaque pixels, else 'STANDARD' */ +#define PNG_ALPHA_BROKEN 3 /* the alpha channel is gamma encoded */ + +PNG_FP_EXPORT(227, void, png_set_alpha_mode, (png_structrp png_ptr, int mode, + double output_gamma)) +PNG_FIXED_EXPORT(228, void, png_set_alpha_mode_fixed, (png_structrp png_ptr, + int mode, png_fixed_point output_gamma)) +#endif + +#if defined(PNG_GAMMA_SUPPORTED) || defined(PNG_READ_ALPHA_MODE_SUPPORTED) +/* The output_gamma value is a screen gamma in libpng terminology: it expresses + * how to decode the output values, not how they are encoded. + */ +#define PNG_DEFAULT_sRGB -1 /* sRGB gamma and color space */ +#define PNG_GAMMA_MAC_18 -2 /* Old Mac '1.8' gamma and color space */ +#define PNG_GAMMA_sRGB 220000 /* Television standards--matches sRGB gamma */ +#define PNG_GAMMA_LINEAR PNG_FP_1 /* Linear */ +#endif + +/* The following are examples of calls to png_set_alpha_mode to achieve the + * required overall gamma correction and, where necessary, alpha + * premultiplication. + * + * png_set_alpha_mode(pp, PNG_ALPHA_PNG, PNG_DEFAULT_sRGB); + * This is the default libpng handling of the alpha channel - it is not + * pre-multiplied into the color components. In addition the call states + * that the output is for a sRGB system and causes all PNG files without gAMA + * chunks to be assumed to be encoded using sRGB. + * + * png_set_alpha_mode(pp, PNG_ALPHA_PNG, PNG_GAMMA_MAC); + * In this case the output is assumed to be something like an sRGB conformant + * display preceeded by a power-law lookup table of power 1.45. This is how + * early Mac systems behaved. + * + * png_set_alpha_mode(pp, PNG_ALPHA_STANDARD, PNG_GAMMA_LINEAR); + * This is the classic Jim Blinn approach and will work in academic + * environments where everything is done by the book. It has the shortcoming + * of assuming that input PNG data with no gamma information is linear - this + * is unlikely to be correct unless the PNG files where generated locally. + * Most of the time the output precision will be so low as to show + * significant banding in dark areas of the image. + * + * png_set_expand_16(pp); + * png_set_alpha_mode(pp, PNG_ALPHA_STANDARD, PNG_DEFAULT_sRGB); + * This is a somewhat more realistic Jim Blinn inspired approach. PNG files + * are assumed to have the sRGB encoding if not marked with a gamma value and + * the output is always 16 bits per component. This permits accurate scaling + * and processing of the data. If you know that your input PNG files were + * generated locally you might need to replace PNG_DEFAULT_sRGB with the + * correct value for your system. + * + * png_set_alpha_mode(pp, PNG_ALPHA_OPTIMIZED, PNG_DEFAULT_sRGB); + * If you just need to composite the PNG image onto an existing background + * and if you control the code that does this you can use the optimization + * setting. In this case you just copy completely opaque pixels to the + * output. For pixels that are not completely transparent (you just skip + * those) you do the composition math using png_composite or png_composite_16 + * below then encode the resultant 8-bit or 16-bit values to match the output + * encoding. + * + * Other cases + * If neither the PNG nor the standard linear encoding work for you because + * of the software or hardware you use then you have a big problem. The PNG + * case will probably result in halos around the image. The linear encoding + * will probably result in a washed out, too bright, image (it's actually too + * contrasty.) Try the ALPHA_OPTIMIZED mode above - this will probably + * substantially reduce the halos. Alternatively try: + * + * png_set_alpha_mode(pp, PNG_ALPHA_BROKEN, PNG_DEFAULT_sRGB); + * This option will also reduce the halos, but there will be slight dark + * halos round the opaque parts of the image where the background is light. + * In the OPTIMIZED mode the halos will be light halos where the background + * is dark. Take your pick - the halos are unavoidable unless you can get + * your hardware/software fixed! (The OPTIMIZED approach is slightly + * faster.) + * + * When the default gamma of PNG files doesn't match the output gamma. + * If you have PNG files with no gamma information png_set_alpha_mode allows + * you to provide a default gamma, but it also sets the ouput gamma to the + * matching value. If you know your PNG files have a gamma that doesn't + * match the output you can take advantage of the fact that + * png_set_alpha_mode always sets the output gamma but only sets the PNG + * default if it is not already set: + * + * png_set_alpha_mode(pp, PNG_ALPHA_PNG, PNG_DEFAULT_sRGB); + * png_set_alpha_mode(pp, PNG_ALPHA_PNG, PNG_GAMMA_MAC); + * The first call sets both the default and the output gamma values, the + * second call overrides the output gamma without changing the default. This + * is easier than achieving the same effect with png_set_gamma. You must use + * PNG_ALPHA_PNG for the first call - internal checking in png_set_alpha will + * fire if more than one call to png_set_alpha_mode and png_set_background is + * made in the same read operation, however multiple calls with PNG_ALPHA_PNG + * are ignored. + */ + +#ifdef PNG_READ_STRIP_ALPHA_SUPPORTED +PNG_EXPORT(36, void, png_set_strip_alpha, (png_structrp png_ptr)); +#endif + +#if defined(PNG_READ_SWAP_ALPHA_SUPPORTED) || \ + defined(PNG_WRITE_SWAP_ALPHA_SUPPORTED) +PNG_EXPORT(37, void, png_set_swap_alpha, (png_structrp png_ptr)); +#endif + +#if defined(PNG_READ_INVERT_ALPHA_SUPPORTED) || \ + defined(PNG_WRITE_INVERT_ALPHA_SUPPORTED) +PNG_EXPORT(38, void, png_set_invert_alpha, (png_structrp png_ptr)); +#endif + +#if defined(PNG_READ_FILLER_SUPPORTED) || defined(PNG_WRITE_FILLER_SUPPORTED) +/* Add a filler byte to 8-bit Gray or 24-bit RGB images. */ +PNG_EXPORT(39, void, png_set_filler, (png_structrp png_ptr, png_uint_32 filler, + int flags)); +/* The values of the PNG_FILLER_ defines should NOT be changed */ +# define PNG_FILLER_BEFORE 0 +# define PNG_FILLER_AFTER 1 +/* Add an alpha byte to 8-bit Gray or 24-bit RGB images. */ +PNG_EXPORT(40, void, png_set_add_alpha, (png_structrp png_ptr, + png_uint_32 filler, int flags)); +#endif /* READ_FILLER || WRITE_FILLER */ + +#if defined(PNG_READ_SWAP_SUPPORTED) || defined(PNG_WRITE_SWAP_SUPPORTED) +/* Swap bytes in 16-bit depth files. */ +PNG_EXPORT(41, void, png_set_swap, (png_structrp png_ptr)); +#endif + +#if defined(PNG_READ_PACK_SUPPORTED) || defined(PNG_WRITE_PACK_SUPPORTED) +/* Use 1 byte per pixel in 1, 2, or 4-bit depth files. */ +PNG_EXPORT(42, void, png_set_packing, (png_structrp png_ptr)); +#endif + +#if defined(PNG_READ_PACKSWAP_SUPPORTED) || \ + defined(PNG_WRITE_PACKSWAP_SUPPORTED) +/* Swap packing order of pixels in bytes. */ +PNG_EXPORT(43, void, png_set_packswap, (png_structrp png_ptr)); +#endif + +#if defined(PNG_READ_SHIFT_SUPPORTED) || defined(PNG_WRITE_SHIFT_SUPPORTED) +/* Converts files to legal bit depths. */ +PNG_EXPORT(44, void, png_set_shift, (png_structrp png_ptr, png_const_color_8p + true_bits)); +#endif + +#if defined(PNG_READ_INTERLACING_SUPPORTED) || \ + defined(PNG_WRITE_INTERLACING_SUPPORTED) +/* Have the code handle the interlacing. Returns the number of passes. + * MUST be called before png_read_update_info or png_start_read_image, + * otherwise it will not have the desired effect. Note that it is still + * necessary to call png_read_row or png_read_rows png_get_image_height + * times for each pass. +*/ +PNG_EXPORT(45, int, png_set_interlace_handling, (png_structrp png_ptr)); +#endif + +#if defined(PNG_READ_INVERT_SUPPORTED) || defined(PNG_WRITE_INVERT_SUPPORTED) +/* Invert monochrome files */ +PNG_EXPORT(46, void, png_set_invert_mono, (png_structrp png_ptr)); +#endif + +#ifdef PNG_READ_BACKGROUND_SUPPORTED +/* Handle alpha and tRNS by replacing with a background color. Prior to + * libpng-1.5.4 this API must not be called before the PNG file header has been + * read. Doing so will result in unexpected behavior and possible warnings or + * errors if the PNG file contains a bKGD chunk. + */ +PNG_FP_EXPORT(47, void, png_set_background, (png_structrp png_ptr, + png_const_color_16p background_color, int background_gamma_code, + int need_expand, double background_gamma)) +PNG_FIXED_EXPORT(215, void, png_set_background_fixed, (png_structrp png_ptr, + png_const_color_16p background_color, int background_gamma_code, + int need_expand, png_fixed_point background_gamma)) +#endif +#ifdef PNG_READ_BACKGROUND_SUPPORTED +# define PNG_BACKGROUND_GAMMA_UNKNOWN 0 +# define PNG_BACKGROUND_GAMMA_SCREEN 1 +# define PNG_BACKGROUND_GAMMA_FILE 2 +# define PNG_BACKGROUND_GAMMA_UNIQUE 3 +#endif + +#ifdef PNG_READ_SCALE_16_TO_8_SUPPORTED +/* Scale a 16-bit depth file down to 8-bit, accurately. */ +PNG_EXPORT(229, void, png_set_scale_16, (png_structrp png_ptr)); +#endif + +#ifdef PNG_READ_STRIP_16_TO_8_SUPPORTED +#define PNG_READ_16_TO_8 SUPPORTED /* Name prior to 1.5.4 */ +/* Strip the second byte of information from a 16-bit depth file. */ +PNG_EXPORT(48, void, png_set_strip_16, (png_structrp png_ptr)); +#endif + +#ifdef PNG_READ_QUANTIZE_SUPPORTED +/* Turn on quantizing, and reduce the palette to the number of colors + * available. + */ +PNG_EXPORT(49, void, png_set_quantize, (png_structrp png_ptr, + png_colorp palette, int num_palette, int maximum_colors, + png_const_uint_16p histogram, int full_quantize)); +#endif + +#ifdef PNG_READ_GAMMA_SUPPORTED +/* The threshold on gamma processing is configurable but hard-wired into the + * library. The following is the floating point variant. + */ +#define PNG_GAMMA_THRESHOLD (PNG_GAMMA_THRESHOLD_FIXED*.00001) + +/* Handle gamma correction. Screen_gamma=(display_exponent). + * NOTE: this API simply sets the screen and file gamma values. It will + * therefore override the value for gamma in a PNG file if it is called after + * the file header has been read - use with care - call before reading the PNG + * file for best results! + * + * These routines accept the same gamma values as png_set_alpha_mode (described + * above). The PNG_GAMMA_ defines and PNG_DEFAULT_sRGB can be passed to either + * API (floating point or fixed.) Notice, however, that the 'file_gamma' value + * is the inverse of a 'screen gamma' value. + */ +PNG_FP_EXPORT(50, void, png_set_gamma, (png_structrp png_ptr, + double screen_gamma, double override_file_gamma)) +PNG_FIXED_EXPORT(208, void, png_set_gamma_fixed, (png_structrp png_ptr, + png_fixed_point screen_gamma, png_fixed_point override_file_gamma)) +#endif + +#ifdef PNG_WRITE_FLUSH_SUPPORTED +/* Set how many lines between output flushes - 0 for no flushing */ +PNG_EXPORT(51, void, png_set_flush, (png_structrp png_ptr, int nrows)); +/* Flush the current PNG output buffer */ +PNG_EXPORT(52, void, png_write_flush, (png_structrp png_ptr)); +#endif + +/* Optional update palette with requested transformations */ +PNG_EXPORT(53, void, png_start_read_image, (png_structrp png_ptr)); + +/* Optional call to update the users info structure */ +PNG_EXPORT(54, void, png_read_update_info, (png_structrp png_ptr, + png_inforp info_ptr)); + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +/* Read one or more rows of image data. */ +PNG_EXPORT(55, void, png_read_rows, (png_structrp png_ptr, png_bytepp row, + png_bytepp display_row, png_uint_32 num_rows)); +#endif + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +/* Read a row of data. */ +PNG_EXPORT(56, void, png_read_row, (png_structrp png_ptr, png_bytep row, + png_bytep display_row)); +#endif + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +/* Read the whole image into memory at once. */ +PNG_EXPORT(57, void, png_read_image, (png_structrp png_ptr, png_bytepp image)); +#endif + +/* Write a row of image data */ +PNG_EXPORT(58, void, png_write_row, (png_structrp png_ptr, + png_const_bytep row)); + +/* Write a few rows of image data: (*row) is not written; however, the type + * is declared as writeable to maintain compatibility with previous versions + * of libpng and to allow the 'display_row' array from read_rows to be passed + * unchanged to write_rows. + */ +PNG_EXPORT(59, void, png_write_rows, (png_structrp png_ptr, png_bytepp row, + png_uint_32 num_rows)); + +/* Write the image data */ +PNG_EXPORT(60, void, png_write_image, (png_structrp png_ptr, png_bytepp image)); + +/* Write the end of the PNG file. */ +PNG_EXPORT(61, void, png_write_end, (png_structrp png_ptr, + png_inforp info_ptr)); + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +/* Read the end of the PNG file. */ +PNG_EXPORT(62, void, png_read_end, (png_structrp png_ptr, png_inforp info_ptr)); +#endif + +/* Free any memory associated with the png_info_struct */ +PNG_EXPORT(63, void, png_destroy_info_struct, (png_const_structrp png_ptr, + png_infopp info_ptr_ptr)); + +/* Free any memory associated with the png_struct and the png_info_structs */ +PNG_EXPORT(64, void, png_destroy_read_struct, (png_structpp png_ptr_ptr, + png_infopp info_ptr_ptr, png_infopp end_info_ptr_ptr)); + +/* Free any memory associated with the png_struct and the png_info_structs */ +PNG_EXPORT(65, void, png_destroy_write_struct, (png_structpp png_ptr_ptr, + png_infopp info_ptr_ptr)); + +/* Set the libpng method of handling chunk CRC errors */ +PNG_EXPORT(66, void, png_set_crc_action, (png_structrp png_ptr, int crit_action, + int ancil_action)); + +/* Values for png_set_crc_action() say how to handle CRC errors in + * ancillary and critical chunks, and whether to use the data contained + * therein. Note that it is impossible to "discard" data in a critical + * chunk. For versions prior to 0.90, the action was always error/quit, + * whereas in version 0.90 and later, the action for CRC errors in ancillary + * chunks is warn/discard. These values should NOT be changed. + * + * value action:critical action:ancillary + */ +#define PNG_CRC_DEFAULT 0 /* error/quit warn/discard data */ +#define PNG_CRC_ERROR_QUIT 1 /* error/quit error/quit */ +#define PNG_CRC_WARN_DISCARD 2 /* (INVALID) warn/discard data */ +#define PNG_CRC_WARN_USE 3 /* warn/use data warn/use data */ +#define PNG_CRC_QUIET_USE 4 /* quiet/use data quiet/use data */ +#define PNG_CRC_NO_CHANGE 5 /* use current value use current value */ + +/* These functions give the user control over the scan-line filtering in + * libpng and the compression methods used by zlib. These functions are + * mainly useful for testing, as the defaults should work with most users. + * Those users who are tight on memory or want faster performance at the + * expense of compression can modify them. See the compression library + * header file (zlib.h) for an explination of the compression functions. + */ + +/* Set the filtering method(s) used by libpng. Currently, the only valid + * value for "method" is 0. + */ +PNG_EXPORT(67, void, png_set_filter, (png_structrp png_ptr, int method, + int filters)); + +/* Flags for png_set_filter() to say which filters to use. The flags + * are chosen so that they don't conflict with real filter types + * below, in case they are supplied instead of the #defined constants. + * These values should NOT be changed. + */ +#define PNG_NO_FILTERS 0x00 +#define PNG_FILTER_NONE 0x08 +#define PNG_FILTER_SUB 0x10 +#define PNG_FILTER_UP 0x20 +#define PNG_FILTER_AVG 0x40 +#define PNG_FILTER_PAETH 0x80 +#define PNG_ALL_FILTERS (PNG_FILTER_NONE | PNG_FILTER_SUB | PNG_FILTER_UP | \ + PNG_FILTER_AVG | PNG_FILTER_PAETH) + +/* Filter values (not flags) - used in pngwrite.c, pngwutil.c for now. + * These defines should NOT be changed. + */ +#define PNG_FILTER_VALUE_NONE 0 +#define PNG_FILTER_VALUE_SUB 1 +#define PNG_FILTER_VALUE_UP 2 +#define PNG_FILTER_VALUE_AVG 3 +#define PNG_FILTER_VALUE_PAETH 4 +#define PNG_FILTER_VALUE_LAST 5 + +#ifdef PNG_WRITE_WEIGHTED_FILTER_SUPPORTED /* EXPERIMENTAL */ +/* The "heuristic_method" is given by one of the PNG_FILTER_HEURISTIC_ + * defines, either the default (minimum-sum-of-absolute-differences), or + * the experimental method (weighted-minimum-sum-of-absolute-differences). + * + * Weights are factors >= 1.0, indicating how important it is to keep the + * filter type consistent between rows. Larger numbers mean the current + * filter is that many times as likely to be the same as the "num_weights" + * previous filters. This is cumulative for each previous row with a weight. + * There needs to be "num_weights" values in "filter_weights", or it can be + * NULL if the weights aren't being specified. Weights have no influence on + * the selection of the first row filter. Well chosen weights can (in theory) + * improve the compression for a given image. + * + * Costs are factors >= 1.0 indicating the relative decoding costs of a + * filter type. Higher costs indicate more decoding expense, and are + * therefore less likely to be selected over a filter with lower computational + * costs. There needs to be a value in "filter_costs" for each valid filter + * type (given by PNG_FILTER_VALUE_LAST), or it can be NULL if you aren't + * setting the costs. Costs try to improve the speed of decompression without + * unduly increasing the compressed image size. + * + * A negative weight or cost indicates the default value is to be used, and + * values in the range [0.0, 1.0) indicate the value is to remain unchanged. + * The default values for both weights and costs are currently 1.0, but may + * change if good general weighting/cost heuristics can be found. If both + * the weights and costs are set to 1.0, this degenerates the WEIGHTED method + * to the UNWEIGHTED method, but with added encoding time/computation. + */ +PNG_FP_EXPORT(68, void, png_set_filter_heuristics, (png_structrp png_ptr, + int heuristic_method, int num_weights, png_const_doublep filter_weights, + png_const_doublep filter_costs)) +PNG_FIXED_EXPORT(209, void, png_set_filter_heuristics_fixed, + (png_structrp png_ptr, int heuristic_method, int num_weights, + png_const_fixed_point_p filter_weights, + png_const_fixed_point_p filter_costs)) +#endif /* WRITE_WEIGHTED_FILTER */ + +/* Heuristic used for row filter selection. These defines should NOT be + * changed. + */ +#define PNG_FILTER_HEURISTIC_DEFAULT 0 /* Currently "UNWEIGHTED" */ +#define PNG_FILTER_HEURISTIC_UNWEIGHTED 1 /* Used by libpng < 0.95 */ +#define PNG_FILTER_HEURISTIC_WEIGHTED 2 /* Experimental feature */ +#define PNG_FILTER_HEURISTIC_LAST 3 /* Not a valid value */ + +#ifdef PNG_WRITE_SUPPORTED +/* Set the library compression level. Currently, valid values range from + * 0 - 9, corresponding directly to the zlib compression levels 0 - 9 + * (0 - no compression, 9 - "maximal" compression). Note that tests have + * shown that zlib compression levels 3-6 usually perform as well as level 9 + * for PNG images, and do considerably fewer caclulations. In the future, + * these values may not correspond directly to the zlib compression levels. + */ +PNG_EXPORT(69, void, png_set_compression_level, (png_structrp png_ptr, + int level)); + +PNG_EXPORT(70, void, png_set_compression_mem_level, (png_structrp png_ptr, + int mem_level)); + +PNG_EXPORT(71, void, png_set_compression_strategy, (png_structrp png_ptr, + int strategy)); + +/* If PNG_WRITE_OPTIMIZE_CMF_SUPPORTED is defined, libpng will use a + * smaller value of window_bits if it can do so safely. + */ +PNG_EXPORT(72, void, png_set_compression_window_bits, (png_structrp png_ptr, + int window_bits)); + +PNG_EXPORT(73, void, png_set_compression_method, (png_structrp png_ptr, + int method)); +#endif + +#ifdef PNG_WRITE_CUSTOMIZE_ZTXT_COMPRESSION_SUPPORTED +/* Also set zlib parameters for compressing non-IDAT chunks */ +PNG_EXPORT(222, void, png_set_text_compression_level, (png_structrp png_ptr, + int level)); + +PNG_EXPORT(223, void, png_set_text_compression_mem_level, (png_structrp png_ptr, + int mem_level)); + +PNG_EXPORT(224, void, png_set_text_compression_strategy, (png_structrp png_ptr, + int strategy)); + +/* If PNG_WRITE_OPTIMIZE_CMF_SUPPORTED is defined, libpng will use a + * smaller value of window_bits if it can do so safely. + */ +PNG_EXPORT(225, void, png_set_text_compression_window_bits, + (png_structrp png_ptr, int window_bits)); + +PNG_EXPORT(226, void, png_set_text_compression_method, (png_structrp png_ptr, + int method)); +#endif /* WRITE_CUSTOMIZE_ZTXT_COMPRESSION */ + +/* These next functions are called for input/output, memory, and error + * handling. They are in the file pngrio.c, pngwio.c, and pngerror.c, + * and call standard C I/O routines such as fread(), fwrite(), and + * fprintf(). These functions can be made to use other I/O routines + * at run time for those applications that need to handle I/O in a + * different manner by calling png_set_???_fn(). See libpng-manual.txt for + * more information. + */ + +#ifdef PNG_STDIO_SUPPORTED +/* Initialize the input/output for the PNG file to the default functions. */ +PNG_EXPORT(74, void, png_init_io, (png_structrp png_ptr, png_FILE_p fp)); +#endif + +/* Replace the (error and abort), and warning functions with user + * supplied functions. If no messages are to be printed you must still + * write and use replacement functions. The replacement error_fn should + * still do a longjmp to the last setjmp location if you are using this + * method of error handling. If error_fn or warning_fn is NULL, the + * default function will be used. + */ + +PNG_EXPORT(75, void, png_set_error_fn, (png_structrp png_ptr, + png_voidp error_ptr, png_error_ptr error_fn, png_error_ptr warning_fn)); + +/* Return the user pointer associated with the error functions */ +PNG_EXPORT(76, png_voidp, png_get_error_ptr, (png_const_structrp png_ptr)); + +/* Replace the default data output functions with a user supplied one(s). + * If buffered output is not used, then output_flush_fn can be set to NULL. + * If PNG_WRITE_FLUSH_SUPPORTED is not defined at libpng compile time + * output_flush_fn will be ignored (and thus can be NULL). + * It is probably a mistake to use NULL for output_flush_fn if + * write_data_fn is not also NULL unless you have built libpng with + * PNG_WRITE_FLUSH_SUPPORTED undefined, because in this case libpng's + * default flush function, which uses the standard *FILE structure, will + * be used. + */ +PNG_EXPORT(77, void, png_set_write_fn, (png_structrp png_ptr, png_voidp io_ptr, + png_rw_ptr write_data_fn, png_flush_ptr output_flush_fn)); + +/* Replace the default data input function with a user supplied one. */ +PNG_EXPORT(78, void, png_set_read_fn, (png_structrp png_ptr, png_voidp io_ptr, + png_rw_ptr read_data_fn)); + +/* Return the user pointer associated with the I/O functions */ +PNG_EXPORT(79, png_voidp, png_get_io_ptr, (png_const_structrp png_ptr)); + +PNG_EXPORT(80, void, png_set_read_status_fn, (png_structrp png_ptr, + png_read_status_ptr read_row_fn)); + +PNG_EXPORT(81, void, png_set_write_status_fn, (png_structrp png_ptr, + png_write_status_ptr write_row_fn)); + +#ifdef PNG_USER_MEM_SUPPORTED +/* Replace the default memory allocation functions with user supplied one(s). */ +PNG_EXPORT(82, void, png_set_mem_fn, (png_structrp png_ptr, png_voidp mem_ptr, + png_malloc_ptr malloc_fn, png_free_ptr free_fn)); +/* Return the user pointer associated with the memory functions */ +PNG_EXPORT(83, png_voidp, png_get_mem_ptr, (png_const_structrp png_ptr)); +#endif + +#ifdef PNG_READ_USER_TRANSFORM_SUPPORTED +PNG_EXPORT(84, void, png_set_read_user_transform_fn, (png_structrp png_ptr, + png_user_transform_ptr read_user_transform_fn)); +#endif + +#ifdef PNG_WRITE_USER_TRANSFORM_SUPPORTED +PNG_EXPORT(85, void, png_set_write_user_transform_fn, (png_structrp png_ptr, + png_user_transform_ptr write_user_transform_fn)); +#endif + +#ifdef PNG_USER_TRANSFORM_PTR_SUPPORTED +PNG_EXPORT(86, void, png_set_user_transform_info, (png_structrp png_ptr, + png_voidp user_transform_ptr, int user_transform_depth, + int user_transform_channels)); +/* Return the user pointer associated with the user transform functions */ +PNG_EXPORT(87, png_voidp, png_get_user_transform_ptr, + (png_const_structrp png_ptr)); +#endif + +#ifdef PNG_USER_TRANSFORM_INFO_SUPPORTED +/* Return information about the row currently being processed. Note that these + * APIs do not fail but will return unexpected results if called outside a user + * transform callback. Also note that when transforming an interlaced image the + * row number is the row number within the sub-image of the interlace pass, so + * the value will increase to the height of the sub-image (not the full image) + * then reset to 0 for the next pass. + * + * Use PNG_ROW_FROM_PASS_ROW(row, pass) and PNG_COL_FROM_PASS_COL(col, pass) to + * find the output pixel (x,y) given an interlaced sub-image pixel + * (row,col,pass). (See below for these macros.) + */ +PNG_EXPORT(217, png_uint_32, png_get_current_row_number, (png_const_structrp)); +PNG_EXPORT(218, png_byte, png_get_current_pass_number, (png_const_structrp)); +#endif + +#ifdef PNG_READ_USER_CHUNKS_SUPPORTED +/* This callback is called only for *unknown* chunks. If + * PNG_HANDLE_AS_UNKNOWN_SUPPORTED is set then it is possible to set known + * chunks to be treated as unknown, however in this case the callback must do + * any processing required by the chunk (e.g. by calling the appropriate + * png_set_ APIs.) + * + * There is no write support - on write, by default, all the chunks in the + * 'unknown' list are written in the specified position. + * + * The integer return from the callback function is interpreted thus: + * + * negative: An error occured, png_chunk_error will be called. + * zero: The chunk was not handled, the chunk will be saved. A critical + * chunk will cause an error at this point unless it is to be saved. + * positive: The chunk was handled, libpng will ignore/discard it. + * + * See "INTERACTION WTIH USER CHUNK CALLBACKS" below for important notes about + * how this behavior will change in libpng 1.7 + */ +PNG_EXPORT(88, void, png_set_read_user_chunk_fn, (png_structrp png_ptr, + png_voidp user_chunk_ptr, png_user_chunk_ptr read_user_chunk_fn)); +#endif + +#ifdef PNG_USER_CHUNKS_SUPPORTED +PNG_EXPORT(89, png_voidp, png_get_user_chunk_ptr, (png_const_structrp png_ptr)); +#endif + +#ifdef PNG_PROGRESSIVE_READ_SUPPORTED +/* Sets the function callbacks for the push reader, and a pointer to a + * user-defined structure available to the callback functions. + */ +PNG_EXPORT(90, void, png_set_progressive_read_fn, (png_structrp png_ptr, + png_voidp progressive_ptr, png_progressive_info_ptr info_fn, + png_progressive_row_ptr row_fn, png_progressive_end_ptr end_fn)); + +/* Returns the user pointer associated with the push read functions */ +PNG_EXPORT(91, png_voidp, png_get_progressive_ptr, + (png_const_structrp png_ptr)); + +/* Function to be called when data becomes available */ +PNG_EXPORT(92, void, png_process_data, (png_structrp png_ptr, + png_inforp info_ptr, png_bytep buffer, png_size_t buffer_size)); + +/* A function which may be called *only* within png_process_data to stop the + * processing of any more data. The function returns the number of bytes + * remaining, excluding any that libpng has cached internally. A subsequent + * call to png_process_data must supply these bytes again. If the argument + * 'save' is set to true the routine will first save all the pending data and + * will always return 0. + */ +PNG_EXPORT(219, png_size_t, png_process_data_pause, (png_structrp, int save)); + +/* A function which may be called *only* outside (after) a call to + * png_process_data. It returns the number of bytes of data to skip in the + * input. Normally it will return 0, but if it returns a non-zero value the + * application must skip than number of bytes of input data and pass the + * following data to the next call to png_process_data. + */ +PNG_EXPORT(220, png_uint_32, png_process_data_skip, (png_structrp)); + +/* Function that combines rows. 'new_row' is a flag that should come from + * the callback and be non-NULL if anything needs to be done; the library + * stores its own version of the new data internally and ignores the passed + * in value. + */ +PNG_EXPORT(93, void, png_progressive_combine_row, (png_const_structrp png_ptr, + png_bytep old_row, png_const_bytep new_row)); +#endif /* PROGRESSIVE_READ */ + +PNG_EXPORTA(94, png_voidp, png_malloc, (png_const_structrp png_ptr, + png_alloc_size_t size), PNG_ALLOCATED); +/* Added at libpng version 1.4.0 */ +PNG_EXPORTA(95, png_voidp, png_calloc, (png_const_structrp png_ptr, + png_alloc_size_t size), PNG_ALLOCATED); + +/* Added at libpng version 1.2.4 */ +PNG_EXPORTA(96, png_voidp, png_malloc_warn, (png_const_structrp png_ptr, + png_alloc_size_t size), PNG_ALLOCATED); + +/* Frees a pointer allocated by png_malloc() */ +PNG_EXPORT(97, void, png_free, (png_const_structrp png_ptr, png_voidp ptr)); + +/* Free data that was allocated internally */ +PNG_EXPORT(98, void, png_free_data, (png_const_structrp png_ptr, + png_inforp info_ptr, png_uint_32 free_me, int num)); + +/* Reassign responsibility for freeing existing data, whether allocated + * by libpng or by the application; this works on the png_info structure passed + * in, it does not change the state for other png_info structures. + * + * It is unlikely that this function works correctly as of 1.6.0 and using it + * may result either in memory leaks or double free of allocated data. + */ +PNG_EXPORT(99, void, png_data_freer, (png_const_structrp png_ptr, + png_inforp info_ptr, int freer, png_uint_32 mask)); + +/* Assignments for png_data_freer */ +#define PNG_DESTROY_WILL_FREE_DATA 1 +#define PNG_SET_WILL_FREE_DATA 1 +#define PNG_USER_WILL_FREE_DATA 2 +/* Flags for png_ptr->free_me and info_ptr->free_me */ +#define PNG_FREE_HIST 0x0008 +#define PNG_FREE_ICCP 0x0010 +#define PNG_FREE_SPLT 0x0020 +#define PNG_FREE_ROWS 0x0040 +#define PNG_FREE_PCAL 0x0080 +#define PNG_FREE_SCAL 0x0100 +#ifdef PNG_STORE_UNKNOWN_CHUNKS_SUPPORTED +# define PNG_FREE_UNKN 0x0200 +#endif +/* PNG_FREE_LIST 0x0400 removed in 1.6.0 because it is ignored */ +#define PNG_FREE_PLTE 0x1000 +#define PNG_FREE_TRNS 0x2000 +#define PNG_FREE_TEXT 0x4000 +#define PNG_FREE_ALL 0x7fff +#define PNG_FREE_MUL 0x4220 /* PNG_FREE_SPLT|PNG_FREE_TEXT|PNG_FREE_UNKN */ + +#ifdef PNG_USER_MEM_SUPPORTED +PNG_EXPORTA(100, png_voidp, png_malloc_default, (png_const_structrp png_ptr, + png_alloc_size_t size), PNG_ALLOCATED PNG_DEPRECATED); +PNG_EXPORTA(101, void, png_free_default, (png_const_structrp png_ptr, + png_voidp ptr), PNG_DEPRECATED); +#endif + +#ifdef PNG_ERROR_TEXT_SUPPORTED +/* Fatal error in PNG image of libpng - can't continue */ +PNG_EXPORTA(102, void, png_error, (png_const_structrp png_ptr, + png_const_charp error_message), PNG_NORETURN); + +/* The same, but the chunk name is prepended to the error string. */ +PNG_EXPORTA(103, void, png_chunk_error, (png_const_structrp png_ptr, + png_const_charp error_message), PNG_NORETURN); + +#else +/* Fatal error in PNG image of libpng - can't continue */ +PNG_EXPORTA(104, void, png_err, (png_const_structrp png_ptr), PNG_NORETURN); +# define png_error(s1,s2) png_err(s1) +# define png_chunk_error(s1,s2) png_err(s1) +#endif + +#ifdef PNG_WARNINGS_SUPPORTED +/* Non-fatal error in libpng. Can continue, but may have a problem. */ +PNG_EXPORT(105, void, png_warning, (png_const_structrp png_ptr, + png_const_charp warning_message)); + +/* Non-fatal error in libpng, chunk name is prepended to message. */ +PNG_EXPORT(106, void, png_chunk_warning, (png_const_structrp png_ptr, + png_const_charp warning_message)); +#else +# define png_warning(s1,s2) ((void)(s1)) +# define png_chunk_warning(s1,s2) ((void)(s1)) +#endif + +#ifdef PNG_BENIGN_ERRORS_SUPPORTED +/* Benign error in libpng. Can continue, but may have a problem. + * User can choose whether to handle as a fatal error or as a warning. */ +PNG_EXPORT(107, void, png_benign_error, (png_const_structrp png_ptr, + png_const_charp warning_message)); + +#ifdef PNG_READ_SUPPORTED +/* Same, chunk name is prepended to message (only during read) */ +PNG_EXPORT(108, void, png_chunk_benign_error, (png_const_structrp png_ptr, + png_const_charp warning_message)); +#endif + +PNG_EXPORT(109, void, png_set_benign_errors, + (png_structrp png_ptr, int allowed)); +#else +# ifdef PNG_ALLOW_BENIGN_ERRORS +# define png_benign_error png_warning +# define png_chunk_benign_error png_chunk_warning +# else +# define png_benign_error png_error +# define png_chunk_benign_error png_chunk_error +# endif +#endif + +/* The png_set_ functions are for storing values in the png_info_struct. + * Similarly, the png_get_ calls are used to read values from the + * png_info_struct, either storing the parameters in the passed variables, or + * setting pointers into the png_info_struct where the data is stored. The + * png_get_ functions return a non-zero value if the data was available + * in info_ptr, or return zero and do not change any of the parameters if the + * data was not available. + * + * These functions should be used instead of directly accessing png_info + * to avoid problems with future changes in the size and internal layout of + * png_info_struct. + */ +/* Returns "flag" if chunk data is valid in info_ptr. */ +PNG_EXPORT(110, png_uint_32, png_get_valid, (png_const_structrp png_ptr, + png_const_inforp info_ptr, png_uint_32 flag)); + +/* Returns number of bytes needed to hold a transformed row. */ +PNG_EXPORT(111, png_size_t, png_get_rowbytes, (png_const_structrp png_ptr, + png_const_inforp info_ptr)); + +#ifdef PNG_INFO_IMAGE_SUPPORTED +/* Returns row_pointers, which is an array of pointers to scanlines that was + * returned from png_read_png(). + */ +PNG_EXPORT(112, png_bytepp, png_get_rows, (png_const_structrp png_ptr, + png_const_inforp info_ptr)); + +/* Set row_pointers, which is an array of pointers to scanlines for use + * by png_write_png(). + */ +PNG_EXPORT(113, void, png_set_rows, (png_const_structrp png_ptr, + png_inforp info_ptr, png_bytepp row_pointers)); +#endif + +/* Returns number of color channels in image. */ +PNG_EXPORT(114, png_byte, png_get_channels, (png_const_structrp png_ptr, + png_const_inforp info_ptr)); + +#ifdef PNG_EASY_ACCESS_SUPPORTED +/* Returns image width in pixels. */ +PNG_EXPORT(115, png_uint_32, png_get_image_width, (png_const_structrp png_ptr, + png_const_inforp info_ptr)); + +/* Returns image height in pixels. */ +PNG_EXPORT(116, png_uint_32, png_get_image_height, (png_const_structrp png_ptr, + png_const_inforp info_ptr)); + +/* Returns image bit_depth. */ +PNG_EXPORT(117, png_byte, png_get_bit_depth, (png_const_structrp png_ptr, + png_const_inforp info_ptr)); + +/* Returns image color_type. */ +PNG_EXPORT(118, png_byte, png_get_color_type, (png_const_structrp png_ptr, + png_const_inforp info_ptr)); + +/* Returns image filter_type. */ +PNG_EXPORT(119, png_byte, png_get_filter_type, (png_const_structrp png_ptr, + png_const_inforp info_ptr)); + +/* Returns image interlace_type. */ +PNG_EXPORT(120, png_byte, png_get_interlace_type, (png_const_structrp png_ptr, + png_const_inforp info_ptr)); + +/* Returns image compression_type. */ +PNG_EXPORT(121, png_byte, png_get_compression_type, (png_const_structrp png_ptr, + png_const_inforp info_ptr)); + +/* Returns image resolution in pixels per meter, from pHYs chunk data. */ +PNG_EXPORT(122, png_uint_32, png_get_pixels_per_meter, + (png_const_structrp png_ptr, png_const_inforp info_ptr)); +PNG_EXPORT(123, png_uint_32, png_get_x_pixels_per_meter, + (png_const_structrp png_ptr, png_const_inforp info_ptr)); +PNG_EXPORT(124, png_uint_32, png_get_y_pixels_per_meter, + (png_const_structrp png_ptr, png_const_inforp info_ptr)); + +/* Returns pixel aspect ratio, computed from pHYs chunk data. */ +PNG_FP_EXPORT(125, float, png_get_pixel_aspect_ratio, + (png_const_structrp png_ptr, png_const_inforp info_ptr)) +PNG_FIXED_EXPORT(210, png_fixed_point, png_get_pixel_aspect_ratio_fixed, + (png_const_structrp png_ptr, png_const_inforp info_ptr)) + +/* Returns image x, y offset in pixels or microns, from oFFs chunk data. */ +PNG_EXPORT(126, png_int_32, png_get_x_offset_pixels, + (png_const_structrp png_ptr, png_const_inforp info_ptr)); +PNG_EXPORT(127, png_int_32, png_get_y_offset_pixels, + (png_const_structrp png_ptr, png_const_inforp info_ptr)); +PNG_EXPORT(128, png_int_32, png_get_x_offset_microns, + (png_const_structrp png_ptr, png_const_inforp info_ptr)); +PNG_EXPORT(129, png_int_32, png_get_y_offset_microns, + (png_const_structrp png_ptr, png_const_inforp info_ptr)); + +#endif /* EASY_ACCESS */ + +#ifdef PNG_READ_SUPPORTED +/* Returns pointer to signature string read from PNG header */ +PNG_EXPORT(130, png_const_bytep, png_get_signature, (png_const_structrp png_ptr, + png_const_inforp info_ptr)); +#endif + +#ifdef PNG_bKGD_SUPPORTED +PNG_EXPORT(131, png_uint_32, png_get_bKGD, (png_const_structrp png_ptr, + png_inforp info_ptr, png_color_16p *background)); +#endif + +#ifdef PNG_bKGD_SUPPORTED +PNG_EXPORT(132, void, png_set_bKGD, (png_const_structrp png_ptr, + png_inforp info_ptr, png_const_color_16p background)); +#endif + +#ifdef PNG_cHRM_SUPPORTED +PNG_FP_EXPORT(133, png_uint_32, png_get_cHRM, (png_const_structrp png_ptr, + png_const_inforp info_ptr, double *white_x, double *white_y, double *red_x, + double *red_y, double *green_x, double *green_y, double *blue_x, + double *blue_y)) +PNG_FP_EXPORT(230, png_uint_32, png_get_cHRM_XYZ, (png_const_structrp png_ptr, + png_const_inforp info_ptr, double *red_X, double *red_Y, double *red_Z, + double *green_X, double *green_Y, double *green_Z, double *blue_X, + double *blue_Y, double *blue_Z)) +PNG_FIXED_EXPORT(134, png_uint_32, png_get_cHRM_fixed, + (png_const_structrp png_ptr, png_const_inforp info_ptr, + png_fixed_point *int_white_x, png_fixed_point *int_white_y, + png_fixed_point *int_red_x, png_fixed_point *int_red_y, + png_fixed_point *int_green_x, png_fixed_point *int_green_y, + png_fixed_point *int_blue_x, png_fixed_point *int_blue_y)) +PNG_FIXED_EXPORT(231, png_uint_32, png_get_cHRM_XYZ_fixed, + (png_const_structrp png_ptr, png_const_inforp info_ptr, + png_fixed_point *int_red_X, png_fixed_point *int_red_Y, + png_fixed_point *int_red_Z, png_fixed_point *int_green_X, + png_fixed_point *int_green_Y, png_fixed_point *int_green_Z, + png_fixed_point *int_blue_X, png_fixed_point *int_blue_Y, + png_fixed_point *int_blue_Z)) +#endif + +#ifdef PNG_cHRM_SUPPORTED +PNG_FP_EXPORT(135, void, png_set_cHRM, (png_const_structrp png_ptr, + png_inforp info_ptr, + double white_x, double white_y, double red_x, double red_y, double green_x, + double green_y, double blue_x, double blue_y)) +PNG_FP_EXPORT(232, void, png_set_cHRM_XYZ, (png_const_structrp png_ptr, + png_inforp info_ptr, double red_X, double red_Y, double red_Z, + double green_X, double green_Y, double green_Z, double blue_X, + double blue_Y, double blue_Z)) +PNG_FIXED_EXPORT(136, void, png_set_cHRM_fixed, (png_const_structrp png_ptr, + png_inforp info_ptr, png_fixed_point int_white_x, + png_fixed_point int_white_y, png_fixed_point int_red_x, + png_fixed_point int_red_y, png_fixed_point int_green_x, + png_fixed_point int_green_y, png_fixed_point int_blue_x, + png_fixed_point int_blue_y)) +PNG_FIXED_EXPORT(233, void, png_set_cHRM_XYZ_fixed, (png_const_structrp png_ptr, + png_inforp info_ptr, png_fixed_point int_red_X, png_fixed_point int_red_Y, + png_fixed_point int_red_Z, png_fixed_point int_green_X, + png_fixed_point int_green_Y, png_fixed_point int_green_Z, + png_fixed_point int_blue_X, png_fixed_point int_blue_Y, + png_fixed_point int_blue_Z)) +#endif + +#ifdef PNG_gAMA_SUPPORTED +PNG_FP_EXPORT(137, png_uint_32, png_get_gAMA, (png_const_structrp png_ptr, + png_const_inforp info_ptr, double *file_gamma)) +PNG_FIXED_EXPORT(138, png_uint_32, png_get_gAMA_fixed, + (png_const_structrp png_ptr, png_const_inforp info_ptr, + png_fixed_point *int_file_gamma)) +#endif + +#ifdef PNG_gAMA_SUPPORTED +PNG_FP_EXPORT(139, void, png_set_gAMA, (png_const_structrp png_ptr, + png_inforp info_ptr, double file_gamma)) +PNG_FIXED_EXPORT(140, void, png_set_gAMA_fixed, (png_const_structrp png_ptr, + png_inforp info_ptr, png_fixed_point int_file_gamma)) +#endif + +#ifdef PNG_hIST_SUPPORTED +PNG_EXPORT(141, png_uint_32, png_get_hIST, (png_const_structrp png_ptr, + png_inforp info_ptr, png_uint_16p *hist)); +#endif + +#ifdef PNG_hIST_SUPPORTED +PNG_EXPORT(142, void, png_set_hIST, (png_const_structrp png_ptr, + png_inforp info_ptr, png_const_uint_16p hist)); +#endif + +PNG_EXPORT(143, png_uint_32, png_get_IHDR, (png_const_structrp png_ptr, + png_const_inforp info_ptr, png_uint_32 *width, png_uint_32 *height, + int *bit_depth, int *color_type, int *interlace_method, + int *compression_method, int *filter_method)); + +PNG_EXPORT(144, void, png_set_IHDR, (png_const_structrp png_ptr, + png_inforp info_ptr, png_uint_32 width, png_uint_32 height, int bit_depth, + int color_type, int interlace_method, int compression_method, + int filter_method)); + +#ifdef PNG_oFFs_SUPPORTED +PNG_EXPORT(145, png_uint_32, png_get_oFFs, (png_const_structrp png_ptr, + png_const_inforp info_ptr, png_int_32 *offset_x, png_int_32 *offset_y, + int *unit_type)); +#endif + +#ifdef PNG_oFFs_SUPPORTED +PNG_EXPORT(146, void, png_set_oFFs, (png_const_structrp png_ptr, + png_inforp info_ptr, png_int_32 offset_x, png_int_32 offset_y, + int unit_type)); +#endif + +#ifdef PNG_pCAL_SUPPORTED +PNG_EXPORT(147, png_uint_32, png_get_pCAL, (png_const_structrp png_ptr, + png_inforp info_ptr, png_charp *purpose, png_int_32 *X0, + png_int_32 *X1, int *type, int *nparams, png_charp *units, + png_charpp *params)); +#endif + +#ifdef PNG_pCAL_SUPPORTED +PNG_EXPORT(148, void, png_set_pCAL, (png_const_structrp png_ptr, + png_inforp info_ptr, png_const_charp purpose, png_int_32 X0, png_int_32 X1, + int type, int nparams, png_const_charp units, png_charpp params)); +#endif + +#ifdef PNG_pHYs_SUPPORTED +PNG_EXPORT(149, png_uint_32, png_get_pHYs, (png_const_structrp png_ptr, + png_const_inforp info_ptr, png_uint_32 *res_x, png_uint_32 *res_y, + int *unit_type)); +#endif + +#ifdef PNG_pHYs_SUPPORTED +PNG_EXPORT(150, void, png_set_pHYs, (png_const_structrp png_ptr, + png_inforp info_ptr, png_uint_32 res_x, png_uint_32 res_y, int unit_type)); +#endif + +PNG_EXPORT(151, png_uint_32, png_get_PLTE, (png_const_structrp png_ptr, + png_inforp info_ptr, png_colorp *palette, int *num_palette)); + +PNG_EXPORT(152, void, png_set_PLTE, (png_structrp png_ptr, + png_inforp info_ptr, png_const_colorp palette, int num_palette)); + +#ifdef PNG_sBIT_SUPPORTED +PNG_EXPORT(153, png_uint_32, png_get_sBIT, (png_const_structrp png_ptr, + png_inforp info_ptr, png_color_8p *sig_bit)); +#endif + +#ifdef PNG_sBIT_SUPPORTED +PNG_EXPORT(154, void, png_set_sBIT, (png_const_structrp png_ptr, + png_inforp info_ptr, png_const_color_8p sig_bit)); +#endif + +#ifdef PNG_sRGB_SUPPORTED +PNG_EXPORT(155, png_uint_32, png_get_sRGB, (png_const_structrp png_ptr, + png_const_inforp info_ptr, int *file_srgb_intent)); +#endif + +#ifdef PNG_sRGB_SUPPORTED +PNG_EXPORT(156, void, png_set_sRGB, (png_const_structrp png_ptr, + png_inforp info_ptr, int srgb_intent)); +PNG_EXPORT(157, void, png_set_sRGB_gAMA_and_cHRM, (png_const_structrp png_ptr, + png_inforp info_ptr, int srgb_intent)); +#endif + +#ifdef PNG_iCCP_SUPPORTED +PNG_EXPORT(158, png_uint_32, png_get_iCCP, (png_const_structrp png_ptr, + png_inforp info_ptr, png_charpp name, int *compression_type, + png_bytepp profile, png_uint_32 *proflen)); +#endif + +#ifdef PNG_iCCP_SUPPORTED +PNG_EXPORT(159, void, png_set_iCCP, (png_const_structrp png_ptr, + png_inforp info_ptr, png_const_charp name, int compression_type, + png_const_bytep profile, png_uint_32 proflen)); +#endif + +#ifdef PNG_sPLT_SUPPORTED +PNG_EXPORT(160, int, png_get_sPLT, (png_const_structrp png_ptr, + png_inforp info_ptr, png_sPLT_tpp entries)); +#endif + +#ifdef PNG_sPLT_SUPPORTED +PNG_EXPORT(161, void, png_set_sPLT, (png_const_structrp png_ptr, + png_inforp info_ptr, png_const_sPLT_tp entries, int nentries)); +#endif + +#ifdef PNG_TEXT_SUPPORTED +/* png_get_text also returns the number of text chunks in *num_text */ +PNG_EXPORT(162, int, png_get_text, (png_const_structrp png_ptr, + png_inforp info_ptr, png_textp *text_ptr, int *num_text)); +#endif + +/* Note while png_set_text() will accept a structure whose text, + * language, and translated keywords are NULL pointers, the structure + * returned by png_get_text will always contain regular + * zero-terminated C strings. They might be empty strings but + * they will never be NULL pointers. + */ + +#ifdef PNG_TEXT_SUPPORTED +PNG_EXPORT(163, void, png_set_text, (png_const_structrp png_ptr, + png_inforp info_ptr, png_const_textp text_ptr, int num_text)); +#endif + +#ifdef PNG_tIME_SUPPORTED +PNG_EXPORT(164, png_uint_32, png_get_tIME, (png_const_structrp png_ptr, + png_inforp info_ptr, png_timep *mod_time)); +#endif + +#ifdef PNG_tIME_SUPPORTED +PNG_EXPORT(165, void, png_set_tIME, (png_const_structrp png_ptr, + png_inforp info_ptr, png_const_timep mod_time)); +#endif + +#ifdef PNG_tRNS_SUPPORTED +PNG_EXPORT(166, png_uint_32, png_get_tRNS, (png_const_structrp png_ptr, + png_inforp info_ptr, png_bytep *trans_alpha, int *num_trans, + png_color_16p *trans_color)); +#endif + +#ifdef PNG_tRNS_SUPPORTED +PNG_EXPORT(167, void, png_set_tRNS, (png_structrp png_ptr, + png_inforp info_ptr, png_const_bytep trans_alpha, int num_trans, + png_const_color_16p trans_color)); +#endif + +#ifdef PNG_sCAL_SUPPORTED +PNG_FP_EXPORT(168, png_uint_32, png_get_sCAL, (png_const_structrp png_ptr, + png_const_inforp info_ptr, int *unit, double *width, double *height)) +#if defined(PNG_FLOATING_ARITHMETIC_SUPPORTED) || \ + defined(PNG_FLOATING_POINT_SUPPORTED) +/* NOTE: this API is currently implemented using floating point arithmetic, + * consequently it can only be used on systems with floating point support. + * In any case the range of values supported by png_fixed_point is small and it + * is highly recommended that png_get_sCAL_s be used instead. + */ +PNG_FIXED_EXPORT(214, png_uint_32, png_get_sCAL_fixed, + (png_const_structrp png_ptr, png_const_inforp info_ptr, int *unit, + png_fixed_point *width, png_fixed_point *height)) +#endif +PNG_EXPORT(169, png_uint_32, png_get_sCAL_s, + (png_const_structrp png_ptr, png_const_inforp info_ptr, int *unit, + png_charpp swidth, png_charpp sheight)); + +PNG_FP_EXPORT(170, void, png_set_sCAL, (png_const_structrp png_ptr, + png_inforp info_ptr, int unit, double width, double height)) +PNG_FIXED_EXPORT(213, void, png_set_sCAL_fixed, (png_const_structrp png_ptr, + png_inforp info_ptr, int unit, png_fixed_point width, + png_fixed_point height)) +PNG_EXPORT(171, void, png_set_sCAL_s, (png_const_structrp png_ptr, + png_inforp info_ptr, int unit, + png_const_charp swidth, png_const_charp sheight)); +#endif /* sCAL */ + +#ifdef PNG_SET_UNKNOWN_CHUNKS_SUPPORTED +/* Provide the default handling for all unknown chunks or, optionally, for + * specific unknown chunks. + * + * NOTE: prior to 1.6.0 the handling specified for particular chunks on read was + * ignored and the default was used, the per-chunk setting only had an effect on + * write. If you wish to have chunk-specific handling on read in code that must + * work on earlier versions you must use a user chunk callback to specify the + * desired handling (keep or discard.) + * + * The 'keep' parameter is a PNG_HANDLE_CHUNK_ value as listed below. The + * parameter is interpreted as follows: + * + * READ: + * PNG_HANDLE_CHUNK_AS_DEFAULT: + * Known chunks: do normal libpng processing, do not keep the chunk (but + * see the comments below about PNG_HANDLE_AS_UNKNOWN_SUPPORTED) + * Unknown chunks: for a specific chunk use the global default, when used + * as the default discard the chunk data. + * PNG_HANDLE_CHUNK_NEVER: + * Discard the chunk data. + * PNG_HANDLE_CHUNK_IF_SAFE: + * Keep the chunk data if the chunk is not critical else raise a chunk + * error. + * PNG_HANDLE_CHUNK_ALWAYS: + * Keep the chunk data. + * + * If the chunk data is saved it can be retrieved using png_get_unknown_chunks, + * below. Notice that specifying "AS_DEFAULT" as a global default is equivalent + * to specifying "NEVER", however when "AS_DEFAULT" is used for specific chunks + * it simply resets the behavior to the libpng default. + * + * INTERACTION WTIH USER CHUNK CALLBACKS: + * The per-chunk handling is always used when there is a png_user_chunk_ptr + * callback and the callback returns 0; the chunk is then always stored *unless* + * it is critical and the per-chunk setting is other than ALWAYS. Notice that + * the global default is *not* used in this case. (In effect the per-chunk + * value is incremented to at least IF_SAFE.) + * + * IMPORTANT NOTE: this behavior will change in libpng 1.7 - the global and + * per-chunk defaults will be honored. If you want to preserve the current + * behavior when your callback returns 0 you must set PNG_HANDLE_CHUNK_IF_SAFE + * as the default - if you don't do this libpng 1.6 will issue a warning. + * + * If you want unhandled unknown chunks to be discarded in libpng 1.6 and + * earlier simply return '1' (handled). + * + * PNG_HANDLE_AS_UNKNOWN_SUPPORTED: + * If this is *not* set known chunks will always be handled by libpng and + * will never be stored in the unknown chunk list. Known chunks listed to + * png_set_keep_unknown_chunks will have no effect. If it is set then known + * chunks listed with a keep other than AS_DEFAULT will *never* be processed + * by libpng, in addition critical chunks must either be processed by the + * callback or saved. + * + * The IHDR and IEND chunks must not be listed. Because this turns off the + * default handling for chunks that would otherwise be recognized the + * behavior of libpng transformations may well become incorrect! + * + * WRITE: + * When writing chunks the options only apply to the chunks specified by + * png_set_unknown_chunks (below), libpng will *always* write known chunks + * required by png_set_ calls and will always write the core critical chunks + * (as required for PLTE). + * + * Each chunk in the png_set_unknown_chunks list is looked up in the + * png_set_keep_unknown_chunks list to find the keep setting, this is then + * interpreted as follows: + * + * PNG_HANDLE_CHUNK_AS_DEFAULT: + * Write safe-to-copy chunks and write other chunks if the global + * default is set to _ALWAYS, otherwise don't write this chunk. + * PNG_HANDLE_CHUNK_NEVER: + * Do not write the chunk. + * PNG_HANDLE_CHUNK_IF_SAFE: + * Write the chunk if it is safe-to-copy, otherwise do not write it. + * PNG_HANDLE_CHUNK_ALWAYS: + * Write the chunk. + * + * Note that the default behavior is effectively the opposite of the read case - + * in read unknown chunks are not stored by default, in write they are written + * by default. Also the behavior of PNG_HANDLE_CHUNK_IF_SAFE is very different + * - on write the safe-to-copy bit is checked, on read the critical bit is + * checked and on read if the chunk is critical an error will be raised. + * + * num_chunks: + * =========== + * If num_chunks is positive, then the "keep" parameter specifies the manner + * for handling only those chunks appearing in the chunk_list array, + * otherwise the chunk list array is ignored. + * + * If num_chunks is 0 the "keep" parameter specifies the default behavior for + * unknown chunks, as described above. + * + * If num_chunks is negative, then the "keep" parameter specifies the manner + * for handling all unknown chunks plus all chunks recognized by libpng + * except for the IHDR, PLTE, tRNS, IDAT, and IEND chunks (which continue to + * be processed by libpng. + */ +PNG_EXPORT(172, void, png_set_keep_unknown_chunks, (png_structrp png_ptr, + int keep, png_const_bytep chunk_list, int num_chunks)); + +/* The "keep" PNG_HANDLE_CHUNK_ parameter for the specified chunk is returned; + * the result is therefore true (non-zero) if special handling is required, + * false for the default handling. + */ +PNG_EXPORT(173, int, png_handle_as_unknown, (png_const_structrp png_ptr, + png_const_bytep chunk_name)); +#endif + +#ifdef PNG_STORE_UNKNOWN_CHUNKS_SUPPORTED +PNG_EXPORT(174, void, png_set_unknown_chunks, (png_const_structrp png_ptr, + png_inforp info_ptr, png_const_unknown_chunkp unknowns, + int num_unknowns)); + /* NOTE: prior to 1.6.0 this routine set the 'location' field of the added + * unknowns to the location currently stored in the png_struct. This is + * invariably the wrong value on write. To fix this call the following API + * for each chunk in the list with the correct location. If you know your + * code won't be compiled on earlier versions you can rely on + * png_set_unknown_chunks(write-ptr, png_get_unknown_chunks(read-ptr)) doing + * the correct thing. + */ + +PNG_EXPORT(175, void, png_set_unknown_chunk_location, + (png_const_structrp png_ptr, png_inforp info_ptr, int chunk, int location)); + +PNG_EXPORT(176, int, png_get_unknown_chunks, (png_const_structrp png_ptr, + png_inforp info_ptr, png_unknown_chunkpp entries)); +#endif + +/* Png_free_data() will turn off the "valid" flag for anything it frees. + * If you need to turn it off for a chunk that your application has freed, + * you can use png_set_invalid(png_ptr, info_ptr, PNG_INFO_CHNK); + */ +PNG_EXPORT(177, void, png_set_invalid, (png_const_structrp png_ptr, + png_inforp info_ptr, int mask)); + +#ifdef PNG_INFO_IMAGE_SUPPORTED +/* The "params" pointer is currently not used and is for future expansion. */ +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +PNG_EXPORT(178, void, png_read_png, (png_structrp png_ptr, png_inforp info_ptr, + int transforms, png_voidp params)); +#endif +#ifdef PNG_WRITE_SUPPORTED +PNG_EXPORT(179, void, png_write_png, (png_structrp png_ptr, png_inforp info_ptr, + int transforms, png_voidp params)); +#endif +#endif + +PNG_EXPORT(180, png_const_charp, png_get_copyright, + (png_const_structrp png_ptr)); +PNG_EXPORT(181, png_const_charp, png_get_header_ver, + (png_const_structrp png_ptr)); +PNG_EXPORT(182, png_const_charp, png_get_header_version, + (png_const_structrp png_ptr)); +PNG_EXPORT(183, png_const_charp, png_get_libpng_ver, + (png_const_structrp png_ptr)); + +#ifdef PNG_MNG_FEATURES_SUPPORTED +PNG_EXPORT(184, png_uint_32, png_permit_mng_features, (png_structrp png_ptr, + png_uint_32 mng_features_permitted)); +#endif + +/* For use in png_set_keep_unknown, added to version 1.2.6 */ +#define PNG_HANDLE_CHUNK_AS_DEFAULT 0 +#define PNG_HANDLE_CHUNK_NEVER 1 +#define PNG_HANDLE_CHUNK_IF_SAFE 2 +#define PNG_HANDLE_CHUNK_ALWAYS 3 +#define PNG_HANDLE_CHUNK_LAST 4 + +/* Strip the prepended error numbers ("#nnn ") from error and warning + * messages before passing them to the error or warning handler. + */ +#ifdef PNG_ERROR_NUMBERS_SUPPORTED +PNG_EXPORT(185, void, png_set_strip_error_numbers, (png_structrp png_ptr, + png_uint_32 strip_mode)); +#endif + +/* Added in libpng-1.2.6 */ +#ifdef PNG_SET_USER_LIMITS_SUPPORTED +PNG_EXPORT(186, void, png_set_user_limits, (png_structrp png_ptr, + png_uint_32 user_width_max, png_uint_32 user_height_max)); +PNG_EXPORT(187, png_uint_32, png_get_user_width_max, + (png_const_structrp png_ptr)); +PNG_EXPORT(188, png_uint_32, png_get_user_height_max, + (png_const_structrp png_ptr)); +/* Added in libpng-1.4.0 */ +PNG_EXPORT(189, void, png_set_chunk_cache_max, (png_structrp png_ptr, + png_uint_32 user_chunk_cache_max)); +PNG_EXPORT(190, png_uint_32, png_get_chunk_cache_max, + (png_const_structrp png_ptr)); +/* Added in libpng-1.4.1 */ +PNG_EXPORT(191, void, png_set_chunk_malloc_max, (png_structrp png_ptr, + png_alloc_size_t user_chunk_cache_max)); +PNG_EXPORT(192, png_alloc_size_t, png_get_chunk_malloc_max, + (png_const_structrp png_ptr)); +#endif + +#if defined(PNG_INCH_CONVERSIONS_SUPPORTED) +PNG_EXPORT(193, png_uint_32, png_get_pixels_per_inch, + (png_const_structrp png_ptr, png_const_inforp info_ptr)); + +PNG_EXPORT(194, png_uint_32, png_get_x_pixels_per_inch, + (png_const_structrp png_ptr, png_const_inforp info_ptr)); + +PNG_EXPORT(195, png_uint_32, png_get_y_pixels_per_inch, + (png_const_structrp png_ptr, png_const_inforp info_ptr)); + +PNG_FP_EXPORT(196, float, png_get_x_offset_inches, + (png_const_structrp png_ptr, png_const_inforp info_ptr)) +#ifdef PNG_FIXED_POINT_SUPPORTED /* otherwise not implemented. */ +PNG_FIXED_EXPORT(211, png_fixed_point, png_get_x_offset_inches_fixed, + (png_const_structrp png_ptr, png_const_inforp info_ptr)) +#endif + +PNG_FP_EXPORT(197, float, png_get_y_offset_inches, (png_const_structrp png_ptr, + png_const_inforp info_ptr)) +#ifdef PNG_FIXED_POINT_SUPPORTED /* otherwise not implemented. */ +PNG_FIXED_EXPORT(212, png_fixed_point, png_get_y_offset_inches_fixed, + (png_const_structrp png_ptr, png_const_inforp info_ptr)) +#endif + +# ifdef PNG_pHYs_SUPPORTED +PNG_EXPORT(198, png_uint_32, png_get_pHYs_dpi, (png_const_structrp png_ptr, + png_const_inforp info_ptr, png_uint_32 *res_x, png_uint_32 *res_y, + int *unit_type)); +# endif /* pHYs */ +#endif /* INCH_CONVERSIONS */ + +/* Added in libpng-1.4.0 */ +#ifdef PNG_IO_STATE_SUPPORTED +PNG_EXPORT(199, png_uint_32, png_get_io_state, (png_const_structrp png_ptr)); + +/* Removed from libpng 1.6; use png_get_io_chunk_type. */ +PNG_REMOVED(200, png_const_bytep, png_get_io_chunk_name, (png_structrp png_ptr), + PNG_DEPRECATED) + +PNG_EXPORT(216, png_uint_32, png_get_io_chunk_type, + (png_const_structrp png_ptr)); + +/* The flags returned by png_get_io_state() are the following: */ +# define PNG_IO_NONE 0x0000 /* no I/O at this moment */ +# define PNG_IO_READING 0x0001 /* currently reading */ +# define PNG_IO_WRITING 0x0002 /* currently writing */ +# define PNG_IO_SIGNATURE 0x0010 /* currently at the file signature */ +# define PNG_IO_CHUNK_HDR 0x0020 /* currently at the chunk header */ +# define PNG_IO_CHUNK_DATA 0x0040 /* currently at the chunk data */ +# define PNG_IO_CHUNK_CRC 0x0080 /* currently at the chunk crc */ +# define PNG_IO_MASK_OP 0x000f /* current operation: reading/writing */ +# define PNG_IO_MASK_LOC 0x00f0 /* current location: sig/hdr/data/crc */ +#endif /* IO_STATE */ + +/* Interlace support. The following macros are always defined so that if + * libpng interlace handling is turned off the macros may be used to handle + * interlaced images within the application. + */ +#define PNG_INTERLACE_ADAM7_PASSES 7 + +/* Two macros to return the first row and first column of the original, + * full, image which appears in a given pass. 'pass' is in the range 0 + * to 6 and the result is in the range 0 to 7. + */ +#define PNG_PASS_START_ROW(pass) (((1&~(pass))<<(3-((pass)>>1)))&7) +#define PNG_PASS_START_COL(pass) (((1& (pass))<<(3-(((pass)+1)>>1)))&7) + +/* A macro to return the offset between pixels in the output row for a pair of + * pixels in the input - effectively the inverse of the 'COL_SHIFT' macro that + * follows. Note that ROW_OFFSET is the offset from one row to the next whereas + * COL_OFFSET is from one column to the next, within a row. + */ +#define PNG_PASS_ROW_OFFSET(pass) ((pass)>2?(8>>(((pass)-1)>>1)):8) +#define PNG_PASS_COL_OFFSET(pass) (1<<((7-(pass))>>1)) + +/* Two macros to help evaluate the number of rows or columns in each + * pass. This is expressed as a shift - effectively log2 of the number or + * rows or columns in each 8x8 tile of the original image. + */ +#define PNG_PASS_ROW_SHIFT(pass) ((pass)>2?(8-(pass))>>1:3) +#define PNG_PASS_COL_SHIFT(pass) ((pass)>1?(7-(pass))>>1:3) + +/* Hence two macros to determine the number of rows or columns in a given + * pass of an image given its height or width. In fact these macros may + * return non-zero even though the sub-image is empty, because the other + * dimension may be empty for a small image. + */ +#define PNG_PASS_ROWS(height, pass) (((height)+(((1<>PNG_PASS_ROW_SHIFT(pass)) +#define PNG_PASS_COLS(width, pass) (((width)+(((1<>PNG_PASS_COL_SHIFT(pass)) + +/* For the reader row callbacks (both progressive and sequential) it is + * necessary to find the row in the output image given a row in an interlaced + * image, so two more macros: + */ +#define PNG_ROW_FROM_PASS_ROW(y_in, pass) \ + (((y_in)<>(((7-(off))-(pass))<<2)) & 0xF) | \ + ((0x01145AF0>>(((7-(off))-(pass))<<2)) & 0xF0)) + +#define PNG_ROW_IN_INTERLACE_PASS(y, pass) \ + ((PNG_PASS_MASK(pass,0) >> ((y)&7)) & 1) +#define PNG_COL_IN_INTERLACE_PASS(x, pass) \ + ((PNG_PASS_MASK(pass,1) >> ((x)&7)) & 1) + +#ifdef PNG_READ_COMPOSITE_NODIV_SUPPORTED +/* With these routines we avoid an integer divide, which will be slower on + * most machines. However, it does take more operations than the corresponding + * divide method, so it may be slower on a few RISC systems. There are two + * shifts (by 8 or 16 bits) and an addition, versus a single integer divide. + * + * Note that the rounding factors are NOT supposed to be the same! 128 and + * 32768 are correct for the NODIV code; 127 and 32767 are correct for the + * standard method. + * + * [Optimized code by Greg Roelofs and Mark Adler...blame us for bugs. :-) ] + */ + + /* fg and bg should be in `gamma 1.0' space; alpha is the opacity */ + +# define png_composite(composite, fg, alpha, bg) \ + { png_uint_16 temp = (png_uint_16)((png_uint_16)(fg) \ + * (png_uint_16)(alpha) \ + + (png_uint_16)(bg)*(png_uint_16)(255 \ + - (png_uint_16)(alpha)) + 128); \ + (composite) = (png_byte)((temp + (temp >> 8)) >> 8); } + +# define png_composite_16(composite, fg, alpha, bg) \ + { png_uint_32 temp = (png_uint_32)((png_uint_32)(fg) \ + * (png_uint_32)(alpha) \ + + (png_uint_32)(bg)*(65535 \ + - (png_uint_32)(alpha)) + 32768); \ + (composite) = (png_uint_16)((temp + (temp >> 16)) >> 16); } + +#else /* Standard method using integer division */ + +# define png_composite(composite, fg, alpha, bg) \ + (composite) = (png_byte)(((png_uint_16)(fg) * (png_uint_16)(alpha) + \ + (png_uint_16)(bg) * (png_uint_16)(255 - (png_uint_16)(alpha)) + \ + 127) / 255) + +# define png_composite_16(composite, fg, alpha, bg) \ + (composite) = (png_uint_16)(((png_uint_32)(fg) * (png_uint_32)(alpha) + \ + (png_uint_32)(bg)*(png_uint_32)(65535 - (png_uint_32)(alpha)) + \ + 32767) / 65535) +#endif /* READ_COMPOSITE_NODIV */ + +#ifdef PNG_READ_INT_FUNCTIONS_SUPPORTED +PNG_EXPORT(201, png_uint_32, png_get_uint_32, (png_const_bytep buf)); +PNG_EXPORT(202, png_uint_16, png_get_uint_16, (png_const_bytep buf)); +PNG_EXPORT(203, png_int_32, png_get_int_32, (png_const_bytep buf)); +#endif + +PNG_EXPORT(204, png_uint_32, png_get_uint_31, (png_const_structrp png_ptr, + png_const_bytep buf)); +/* No png_get_int_16 -- may be added if there's a real need for it. */ + +/* Place a 32-bit number into a buffer in PNG byte order (big-endian). */ +#ifdef PNG_WRITE_INT_FUNCTIONS_SUPPORTED +PNG_EXPORT(205, void, png_save_uint_32, (png_bytep buf, png_uint_32 i)); +#endif +#ifdef PNG_SAVE_INT_32_SUPPORTED +PNG_EXPORT(206, void, png_save_int_32, (png_bytep buf, png_int_32 i)); +#endif + +/* Place a 16-bit number into a buffer in PNG byte order. + * The parameter is declared unsigned int, not png_uint_16, + * just to avoid potential problems on pre-ANSI C compilers. + */ +#ifdef PNG_WRITE_INT_FUNCTIONS_SUPPORTED +PNG_EXPORT(207, void, png_save_uint_16, (png_bytep buf, unsigned int i)); +/* No png_save_int_16 -- may be added if there's a real need for it. */ +#endif + +#ifdef PNG_USE_READ_MACROS +/* Inline macros to do direct reads of bytes from the input buffer. + * The png_get_int_32() routine assumes we are using two's complement + * format for negative values, which is almost certainly true. + */ +# define PNG_get_uint_32(buf) \ + (((png_uint_32)(*(buf)) << 24) + \ + ((png_uint_32)(*((buf) + 1)) << 16) + \ + ((png_uint_32)(*((buf) + 2)) << 8) + \ + ((png_uint_32)(*((buf) + 3)))) + + /* From libpng-1.4.0 until 1.4.4, the png_get_uint_16 macro (but not the + * function) incorrectly returned a value of type png_uint_32. + */ +# define PNG_get_uint_16(buf) \ + ((png_uint_16) \ + (((unsigned int)(*(buf)) << 8) + \ + ((unsigned int)(*((buf) + 1))))) + +# define PNG_get_int_32(buf) \ + ((png_int_32)((*(buf) & 0x80) \ + ? -((png_int_32)((png_get_uint_32(buf) ^ 0xffffffffL) + 1)) \ + : (png_int_32)png_get_uint_32(buf))) + + /* If PNG_PREFIX is defined the same thing as below happens in pnglibconf.h, + * but defining a macro name prefixed with PNG_PREFIX. + */ +# ifndef PNG_PREFIX +# define png_get_uint_32(buf) PNG_get_uint_32(buf) +# define png_get_uint_16(buf) PNG_get_uint_16(buf) +# define png_get_int_32(buf) PNG_get_int_32(buf) +# endif +#else +# ifdef PNG_PREFIX + /* No macros; revert to the (redefined) function */ +# define PNG_get_uint_32 (png_get_uint_32) +# define PNG_get_uint_16 (png_get_uint_16) +# define PNG_get_int_32 (png_get_int_32) +# endif +#endif + +#if defined(PNG_SIMPLIFIED_READ_SUPPORTED) || \ + defined(PNG_SIMPLIFIED_WRITE_SUPPORTED) +/******************************************************************************* + * SIMPLIFIED API + ******************************************************************************* + * + * Please read the documentation in libpng-manual.txt (TODO: write said + * documentation) if you don't understand what follows. + * + * The simplified API hides the details of both libpng and the PNG file format + * itself. It allows PNG files to be read into a very limited number of + * in-memory bitmap formats or to be written from the same formats. If these + * formats do not accomodate your needs then you can, and should, use the more + * sophisticated APIs above - these support a wide variety of in-memory formats + * and a wide variety of sophisticated transformations to those formats as well + * as a wide variety of APIs to manipulate ancillary information. + * + * To read a PNG file using the simplified API: + * + * 1) Declare a 'png_image' structure (see below) on the stack and set the + * version field to PNG_IMAGE_VERSION. + * 2) Call the appropriate png_image_begin_read... function. + * 3) Set the png_image 'format' member to the required sample format. + * 4) Allocate a buffer for the image and, if required, the color-map. + * 5) Call png_image_finish_read to read the image and, if required, the + * color-map into your buffers. + * + * There are no restrictions on the format of the PNG input itself; all valid + * color types, bit depths, and interlace methods are acceptable, and the + * input image is transformed as necessary to the requested in-memory format + * during the png_image_finish_read() step. The only caveat is that if you + * request a color-mapped image from a PNG that is full-color or makes + * complex use of an alpha channel the transformation is extremely lossy and the + * result may look terrible. + * + * To write a PNG file using the simplified API: + * + * 1) Declare a 'png_image' structure on the stack and memset() it to all zero. + * 2) Initialize the members of the structure that describe the image, setting + * the 'format' member to the format of the image samples. + * 3) Call the appropriate png_image_write... function with a pointer to the + * image and, if necessary, the color-map to write the PNG data. + * + * png_image is a structure that describes the in-memory format of an image + * when it is being read or defines the in-memory format of an image that you + * need to write: + */ +#define PNG_IMAGE_VERSION 1 + +typedef struct png_control *png_controlp; +typedef struct +{ + png_controlp opaque; /* Initialize to NULL, free with png_image_free */ + png_uint_32 version; /* Set to PNG_IMAGE_VERSION */ + png_uint_32 width; /* Image width in pixels (columns) */ + png_uint_32 height; /* Image height in pixels (rows) */ + png_uint_32 format; /* Image format as defined below */ + png_uint_32 flags; /* A bit mask containing informational flags */ + png_uint_32 colormap_entries; + /* Number of entries in the color-map */ + + /* In the event of an error or warning the following field will be set to a + * non-zero value and the 'message' field will contain a '\0' terminated + * string with the libpng error or warning message. If both warnings and + * an error were encountered, only the error is recorded. If there + * are multiple warnings, only the first one is recorded. + * + * The upper 30 bits of this value are reserved, the low two bits contain + * a value as follows: + */ +# define PNG_IMAGE_WARNING 1 +# define PNG_IMAGE_ERROR 2 + /* + * The result is a two-bit code such that a value more than 1 indicates + * a failure in the API just called: + * + * 0 - no warning or error + * 1 - warning + * 2 - error + * 3 - error preceded by warning + */ +# define PNG_IMAGE_FAILED(png_cntrl) ((((png_cntrl).warning_or_error)&0x03)>1) + + png_uint_32 warning_or_error; + + char message[64]; +} png_image, *png_imagep; + +/* The samples of the image have one to four channels whose components have + * original values in the range 0 to 1.0: + * + * 1: A single gray or luminance channel (G). + * 2: A gray/luminance channel and an alpha channel (GA). + * 3: Three red, green, blue color channels (RGB). + * 4: Three color channels and an alpha channel (RGBA). + * + * The components are encoded in one of two ways: + * + * a) As a small integer, value 0..255, contained in a single byte. For the + * alpha channel the original value is simply value/255. For the color or + * luminance channels the value is encoded according to the sRGB specification + * and matches the 8-bit format expected by typical display devices. + * + * The color/gray channels are not scaled (pre-multiplied) by the alpha + * channel and are suitable for passing to color management software. + * + * b) As a value in the range 0..65535, contained in a 2-byte integer. All + * channels can be converted to the original value by dividing by 65535; all + * channels are linear. Color channels use the RGB encoding (RGB end-points) of + * the sRGB specification. This encoding is identified by the + * PNG_FORMAT_FLAG_LINEAR flag below. + * + * When the simplified API needs to convert between sRGB and linear colorspaces, + * the actual sRGB transfer curve defined in the sRGB specification (see the + * article at http://en.wikipedia.org/wiki/SRGB) is used, not the gamma=1/2.2 + * approximation used elsewhere in libpng. + * + * When an alpha channel is present it is expected to denote pixel coverage + * of the color or luminance channels and is returned as an associated alpha + * channel: the color/gray channels are scaled (pre-multiplied) by the alpha + * value. + * + * The samples are either contained directly in the image data, between 1 and 8 + * bytes per pixel according to the encoding, or are held in a color-map indexed + * by bytes in the image data. In the case of a color-map the color-map entries + * are individual samples, encoded as above, and the image data has one byte per + * pixel to select the relevant sample from the color-map. + */ + +/* PNG_FORMAT_* + * + * #defines to be used in png_image::format. Each #define identifies a + * particular layout of sample data and, if present, alpha values. There are + * separate defines for each of the two component encodings. + * + * A format is built up using single bit flag values. All combinations are + * valid. Formats can be built up from the flag values or you can use one of + * the predefined values below. When testing formats always use the FORMAT_FLAG + * macros to test for individual features - future versions of the library may + * add new flags. + * + * When reading or writing color-mapped images the format should be set to the + * format of the entries in the color-map then png_image_{read,write}_colormap + * called to read or write the color-map and set the format correctly for the + * image data. Do not set the PNG_FORMAT_FLAG_COLORMAP bit directly! + * + * NOTE: libpng can be built with particular features disabled, if you see + * compiler errors because the definition of one of the following flags has been + * compiled out it is because libpng does not have the required support. It is + * possible, however, for the libpng configuration to enable the format on just + * read or just write; in that case you may see an error at run time. You can + * guard against this by checking for the definition of the appropriate + * "_SUPPORTED" macro, one of: + * + * PNG_SIMPLIFIED_{READ,WRITE}_{BGR,AFIRST}_SUPPORTED + */ +#define PNG_FORMAT_FLAG_ALPHA 0x01U /* format with an alpha channel */ +#define PNG_FORMAT_FLAG_COLOR 0x02U /* color format: otherwise grayscale */ +#define PNG_FORMAT_FLAG_LINEAR 0x04U /* 2 byte channels else 1 byte */ +#define PNG_FORMAT_FLAG_COLORMAP 0x08U /* image data is color-mapped */ + +#ifdef PNG_FORMAT_BGR_SUPPORTED +# define PNG_FORMAT_FLAG_BGR 0x10U /* BGR colors, else order is RGB */ +#endif + +#ifdef PNG_FORMAT_AFIRST_SUPPORTED +# define PNG_FORMAT_FLAG_AFIRST 0x20U /* alpha channel comes first */ +#endif + +/* Commonly used formats have predefined macros. + * + * First the single byte (sRGB) formats: + */ +#define PNG_FORMAT_GRAY 0 +#define PNG_FORMAT_GA PNG_FORMAT_FLAG_ALPHA +#define PNG_FORMAT_AG (PNG_FORMAT_GA|PNG_FORMAT_FLAG_AFIRST) +#define PNG_FORMAT_RGB PNG_FORMAT_FLAG_COLOR +#define PNG_FORMAT_BGR (PNG_FORMAT_FLAG_COLOR|PNG_FORMAT_FLAG_BGR) +#define PNG_FORMAT_RGBA (PNG_FORMAT_RGB|PNG_FORMAT_FLAG_ALPHA) +#define PNG_FORMAT_ARGB (PNG_FORMAT_RGBA|PNG_FORMAT_FLAG_AFIRST) +#define PNG_FORMAT_BGRA (PNG_FORMAT_BGR|PNG_FORMAT_FLAG_ALPHA) +#define PNG_FORMAT_ABGR (PNG_FORMAT_BGRA|PNG_FORMAT_FLAG_AFIRST) + +/* Then the linear 2-byte formats. When naming these "Y" is used to + * indicate a luminance (gray) channel. + */ +#define PNG_FORMAT_LINEAR_Y PNG_FORMAT_FLAG_LINEAR +#define PNG_FORMAT_LINEAR_Y_ALPHA (PNG_FORMAT_FLAG_LINEAR|PNG_FORMAT_FLAG_ALPHA) +#define PNG_FORMAT_LINEAR_RGB (PNG_FORMAT_FLAG_LINEAR|PNG_FORMAT_FLAG_COLOR) +#define PNG_FORMAT_LINEAR_RGB_ALPHA \ + (PNG_FORMAT_FLAG_LINEAR|PNG_FORMAT_FLAG_COLOR|PNG_FORMAT_FLAG_ALPHA) + +/* With color-mapped formats the image data is one byte for each pixel, the byte + * is an index into the color-map which is formatted as above. To obtain a + * color-mapped format it is sufficient just to add the PNG_FOMAT_FLAG_COLORMAP + * to one of the above definitions, or you can use one of the definitions below. + */ +#define PNG_FORMAT_RGB_COLORMAP (PNG_FORMAT_RGB|PNG_FORMAT_FLAG_COLORMAP) +#define PNG_FORMAT_BGR_COLORMAP (PNG_FORMAT_BGR|PNG_FORMAT_FLAG_COLORMAP) +#define PNG_FORMAT_RGBA_COLORMAP (PNG_FORMAT_RGBA|PNG_FORMAT_FLAG_COLORMAP) +#define PNG_FORMAT_ARGB_COLORMAP (PNG_FORMAT_ARGB|PNG_FORMAT_FLAG_COLORMAP) +#define PNG_FORMAT_BGRA_COLORMAP (PNG_FORMAT_BGRA|PNG_FORMAT_FLAG_COLORMAP) +#define PNG_FORMAT_ABGR_COLORMAP (PNG_FORMAT_ABGR|PNG_FORMAT_FLAG_COLORMAP) + +/* PNG_IMAGE macros + * + * These are convenience macros to derive information from a png_image + * structure. The PNG_IMAGE_SAMPLE_ macros return values appropriate to the + * actual image sample values - either the entries in the color-map or the + * pixels in the image. The PNG_IMAGE_PIXEL_ macros return corresponding values + * for the pixels and will always return 1 for color-mapped formats. The + * remaining macros return information about the rows in the image and the + * complete image. + * + * NOTE: All the macros that take a png_image::format parameter are compile time + * constants if the format parameter is, itself, a constant. Therefore these + * macros can be used in array declarations and case labels where required. + * Similarly the macros are also pre-processor constants (sizeof is not used) so + * they can be used in #if tests. + * + * First the information about the samples. + */ +#define PNG_IMAGE_SAMPLE_CHANNELS(fmt)\ + (((fmt)&(PNG_FORMAT_FLAG_COLOR|PNG_FORMAT_FLAG_ALPHA))+1) + /* Return the total number of channels in a given format: 1..4 */ + +#define PNG_IMAGE_SAMPLE_COMPONENT_SIZE(fmt)\ + ((((fmt) & PNG_FORMAT_FLAG_LINEAR) >> 2)+1) + /* Return the size in bytes of a single component of a pixel or color-map + * entry (as appropriate) in the image: 1 or 2. + */ + +#define PNG_IMAGE_SAMPLE_SIZE(fmt)\ + (PNG_IMAGE_SAMPLE_CHANNELS(fmt) * PNG_IMAGE_SAMPLE_COMPONENT_SIZE(fmt)) + /* This is the size of the sample data for one sample. If the image is + * color-mapped it is the size of one color-map entry (and image pixels are + * one byte in size), otherwise it is the size of one image pixel. + */ + +#define PNG_IMAGE_MAXIMUM_COLORMAP_COMPONENTS(fmt)\ + (PNG_IMAGE_SAMPLE_CHANNELS(fmt) * 256) + /* The maximum size of the color-map required by the format expressed in a + * count of components. This can be used to compile-time allocate a + * color-map: + * + * png_uint_16 colormap[PNG_IMAGE_MAXIMUM_COLORMAP_COMPONENTS(linear_fmt)]; + * + * png_byte colormap[PNG_IMAGE_MAXIMUM_COLORMAP_COMPONENTS(sRGB_fmt)]; + * + * Alternatively use the PNG_IMAGE_COLORMAP_SIZE macro below to use the + * information from one of the png_image_begin_read_ APIs and dynamically + * allocate the required memory. + */ + +/* Corresponding information about the pixels */ +#define PNG_IMAGE_PIXEL_(test,fmt)\ + (((fmt)&PNG_FORMAT_FLAG_COLORMAP)?1:test(fmt)) + +#define PNG_IMAGE_PIXEL_CHANNELS(fmt)\ + PNG_IMAGE_PIXEL_(PNG_IMAGE_SAMPLE_CHANNELS,fmt) + /* The number of separate channels (components) in a pixel; 1 for a + * color-mapped image. + */ + +#define PNG_IMAGE_PIXEL_COMPONENT_SIZE(fmt)\ + PNG_IMAGE_PIXEL_(PNG_IMAGE_SAMPLE_COMPONENT_SIZE,fmt) + /* The size, in bytes, of each component in a pixel; 1 for a color-mapped + * image. + */ + +#define PNG_IMAGE_PIXEL_SIZE(fmt) PNG_IMAGE_PIXEL_(PNG_IMAGE_SAMPLE_SIZE,fmt) + /* The size, in bytes, of a complete pixel; 1 for a color-mapped image. */ + +/* Information about the whole row, or whole image */ +#define PNG_IMAGE_ROW_STRIDE(image)\ + (PNG_IMAGE_PIXEL_CHANNELS((image).format) * (image).width) + /* Return the total number of components in a single row of the image; this + * is the minimum 'row stride', the minimum count of components between each + * row. For a color-mapped image this is the minimum number of bytes in a + * row. + */ + +#define PNG_IMAGE_BUFFER_SIZE(image, row_stride)\ + (PNG_IMAGE_PIXEL_COMPONENT_SIZE((image).format)*(image).height*(row_stride)) + /* Return the size, in bytes, of an image buffer given a png_image and a row + * stride - the number of components to leave space for in each row. + */ + +#define PNG_IMAGE_SIZE(image)\ + PNG_IMAGE_BUFFER_SIZE(image, PNG_IMAGE_ROW_STRIDE(image)) + /* Return the size, in bytes, of the image in memory given just a png_image; + * the row stride is the minimum stride required for the image. + */ + +#define PNG_IMAGE_COLORMAP_SIZE(image)\ + (PNG_IMAGE_SAMPLE_SIZE((image).format) * (image).colormap_entries) + /* Return the size, in bytes, of the color-map of this image. If the image + * format is not a color-map format this will return a size sufficient for + * 256 entries in the given format; check PNG_FORMAT_FLAG_COLORMAP if + * you don't want to allocate a color-map in this case. + */ + +/* PNG_IMAGE_FLAG_* + * + * Flags containing additional information about the image are held in the + * 'flags' field of png_image. + */ +#define PNG_IMAGE_FLAG_COLORSPACE_NOT_sRGB 0x01 + /* This indicates the the RGB values of the in-memory bitmap do not + * correspond to the red, green and blue end-points defined by sRGB. + */ + +#define PNG_IMAGE_FLAG_FAST 0x02 + /* On write emphasise speed over compression; the resultant PNG file will be + * larger but will be produced significantly faster, particular for large + * images. Do not use this option for images which will be distributed, only + * used it when producing intermediate files that will be read back in + * repeatedly. For a typical 24-bit image the option will double the read + * speed at the cost of increasing the image size by 25%, however for many + * more compressible images the PNG file can be 10 times larger with only a + * slight speed gain. + */ + +#define PNG_IMAGE_FLAG_16BIT_sRGB 0x04 + /* On read if the image is a 16-bit per component image and there is no gAMA + * or sRGB chunk assume that the components are sRGB encoded. Notice that + * images output by the simplified API always have gamma information; setting + * this flag only affects the interpretation of 16-bit images from an + * external source. It is recommended that the application expose this flag + * to the user; the user can normally easily recognize the difference between + * linear and sRGB encoding. This flag has no effect on write - the data + * passed to the write APIs must have the correct encoding (as defined + * above.) + * + * If the flag is not set (the default) input 16-bit per component data is + * assumed to be linear. + * + * NOTE: the flag can only be set after the png_image_begin_read_ call, + * because that call initializes the 'flags' field. + */ + +#ifdef PNG_SIMPLIFIED_READ_SUPPORTED +/* READ APIs + * --------- + * + * The png_image passed to the read APIs must have been initialized by setting + * the png_controlp field 'opaque' to NULL (or, safer, memset the whole thing.) + */ +#ifdef PNG_STDIO_SUPPORTED +PNG_EXPORT(234, int, png_image_begin_read_from_file, (png_imagep image, + const char *file_name)); + /* The named file is opened for read and the image header is filled in + * from the PNG header in the file. + */ + +PNG_EXPORT(235, int, png_image_begin_read_from_stdio, (png_imagep image, + FILE* file)); + /* The PNG header is read from the stdio FILE object. */ +#endif /* STDIO */ + +PNG_EXPORT(236, int, png_image_begin_read_from_memory, (png_imagep image, + png_const_voidp memory, png_size_t size)); + /* The PNG header is read from the given memory buffer. */ + +PNG_EXPORT(237, int, png_image_finish_read, (png_imagep image, + png_const_colorp background, void *buffer, png_int_32 row_stride, + void *colormap)); + /* Finish reading the image into the supplied buffer and clean up the + * png_image structure. + * + * row_stride is the step, in byte or 2-byte units as appropriate, + * between adjacent rows. A positive stride indicates that the top-most row + * is first in the buffer - the normal top-down arrangement. A negative + * stride indicates that the bottom-most row is first in the buffer. + * + * background need only be supplied if an alpha channel must be removed from + * a png_byte format and the removal is to be done by compositing on a solid + * color; otherwise it may be NULL and any composition will be done directly + * onto the buffer. The value is an sRGB color to use for the background, + * for grayscale output the green channel is used. + * + * background must be supplied when an alpha channel must be removed from a + * single byte color-mapped output format, in other words if: + * + * 1) The original format from png_image_begin_read_from_* had + * PNG_FORMAT_FLAG_ALPHA set. + * 2) The format set by the application does not. + * 3) The format set by the application has PNG_FORMAT_FLAG_COLORMAP set and + * PNG_FORMAT_FLAG_LINEAR *not* set. + * + * For linear output removing the alpha channel is always done by compositing + * on black and background is ignored. + * + * colormap must be supplied when PNG_FORMAT_FLAG_COLORMAP is set. It must + * be at least the size (in bytes) returned by PNG_IMAGE_COLORMAP_SIZE. + * image->colormap_entries will be updated to the actual number of entries + * written to the colormap; this may be less than the original value. + */ + +PNG_EXPORT(238, void, png_image_free, (png_imagep image)); + /* Free any data allocated by libpng in image->opaque, setting the pointer to + * NULL. May be called at any time after the structure is initialized. + */ +#endif /* SIMPLIFIED_READ */ + +#ifdef PNG_SIMPLIFIED_WRITE_SUPPORTED +#ifdef PNG_STDIO_SUPPORTED +/* WRITE APIS + * ---------- + * For write you must initialize a png_image structure to describe the image to + * be written. To do this use memset to set the whole structure to 0 then + * initialize fields describing your image. + * + * version: must be set to PNG_IMAGE_VERSION + * opaque: must be initialized to NULL + * width: image width in pixels + * height: image height in rows + * format: the format of the data (image and color-map) you wish to write + * flags: set to 0 unless one of the defined flags applies; set + * PNG_IMAGE_FLAG_COLORSPACE_NOT_sRGB for color format images where the RGB + * values do not correspond to the colors in sRGB. + * colormap_entries: set to the number of entries in the color-map (0 to 256) + */ +PNG_EXPORT(239, int, png_image_write_to_file, (png_imagep image, + const char *file, int convert_to_8bit, const void *buffer, + png_int_32 row_stride, const void *colormap)); + /* Write the image to the named file. */ + +PNG_EXPORT(240, int, png_image_write_to_stdio, (png_imagep image, FILE *file, + int convert_to_8_bit, const void *buffer, png_int_32 row_stride, + const void *colormap)); + /* Write the image to the given (FILE*). */ + +/* With both write APIs if image is in one of the linear formats with 16-bit + * data then setting convert_to_8_bit will cause the output to be an 8-bit PNG + * gamma encoded according to the sRGB specification, otherwise a 16-bit linear + * encoded PNG file is written. + * + * With color-mapped data formats the colormap parameter point to a color-map + * with at least image->colormap_entries encoded in the specified format. If + * the format is linear the written PNG color-map will be converted to sRGB + * regardless of the convert_to_8_bit flag. + * + * With all APIs row_stride is handled as in the read APIs - it is the spacing + * from one row to the next in component sized units (1 or 2 bytes) and if + * negative indicates a bottom-up row layout in the buffer. + * + * Note that the write API does not support interlacing or sub-8-bit pixels. + */ +#endif /* STDIO */ +#endif /* SIMPLIFIED_WRITE */ +/******************************************************************************* + * END OF SIMPLIFIED API + ******************************************************************************/ +#endif /* SIMPLIFIED_{READ|WRITE} */ + +#ifdef PNG_CHECK_FOR_INVALID_INDEX_SUPPORTED +PNG_EXPORT(242, void, png_set_check_for_invalid_index, + (png_structrp png_ptr, int allowed)); +# ifdef PNG_GET_PALETTE_MAX_SUPPORTED +PNG_EXPORT(243, int, png_get_palette_max, (png_const_structp png_ptr, + png_const_infop info_ptr)); +# endif +#endif /* CHECK_FOR_INVALID_INDEX */ + +/******************************************************************************* + * IMPLEMENTATION OPTIONS + ******************************************************************************* + * + * Support for arbitrary implementation-specific optimizations. The API allows + * particular options to be turned on or off. 'Option' is the number of the + * option and 'onoff' is 0 (off) or non-0 (on). The value returned is given + * by the PNG_OPTION_ defines below. + * + * HARDWARE: normally hardware capabilites, such as the Intel SSE instructions, + * are detected at run time, however sometimes it may be impossible + * to do this in user mode, in which case it is necessary to discover + * the capabilities in an OS specific way. Such capabilities are + * listed here when libpng has support for them and must be turned + * ON by the application if present. + * + * SOFTWARE: sometimes software optimizations actually result in performance + * decrease on some architectures or systems, or with some sets of + * PNG images. 'Software' options allow such optimizations to be + * selected at run time. + */ +#ifdef PNG_SET_OPTION_SUPPORTED +#ifdef PNG_ARM_NEON_API_SUPPORTED +# define PNG_ARM_NEON 0 /* HARDWARE: ARM Neon SIMD instructions supported */ +#endif +#define PNG_MAXIMUM_INFLATE_WINDOW 2 /* SOFTWARE: force maximum window */ +#define PNG_SKIP_sRGB_CHECK_PROFILE 4 /* SOFTWARE: Check ICC profile for sRGB */ +#define PNG_OPTION_NEXT 6 /* Next option - numbers must be even */ + +/* Return values: NOTE: there are four values and 'off' is *not* zero */ +#define PNG_OPTION_UNSET 0 /* Unset - defaults to off */ +#define PNG_OPTION_INVALID 1 /* Option number out of range */ +#define PNG_OPTION_OFF 2 +#define PNG_OPTION_ON 3 + +PNG_EXPORT(244, int, png_set_option, (png_structrp png_ptr, int option, + int onoff)); +#endif /* SET_OPTION */ + +/******************************************************************************* + * END OF HARDWARE AND SOFTWARE OPTIONS + ******************************************************************************/ + +/* Maintainer: Put new public prototypes here ^, in libpng.3, in project + * defs, and in scripts/symbols.def. + */ + +/* The last ordinal number (this is the *last* one already used; the next + * one to use is one more than this.) + */ +#ifdef PNG_EXPORT_LAST_ORDINAL + PNG_EXPORT_LAST_ORDINAL(244); +#endif + +#ifdef __cplusplus +} +#endif + +#endif /* PNG_VERSION_INFO_ONLY */ +/* Do not put anything past this line */ +#endif /* PNG_H */ diff --git a/app/src/main/jni/libpng/pngconf.h b/app/src/main/jni/libpng/pngconf.h new file mode 100644 index 000000000..03615f0e7 --- /dev/null +++ b/app/src/main/jni/libpng/pngconf.h @@ -0,0 +1,644 @@ + +/* pngconf.h - machine configurable file for libpng + * + * libpng version 1.6.16,December 22, 2014 + * + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + * + */ + +/* Any machine specific code is near the front of this file, so if you + * are configuring libpng for a machine, you may want to read the section + * starting here down to where it starts to typedef png_color, png_text, + * and png_info. + */ + +#ifndef PNGCONF_H +#define PNGCONF_H + +/* To do: Do all of this in scripts/pnglibconf.dfa */ +#ifdef PNG_SAFE_LIMITS_SUPPORTED +# ifdef PNG_USER_WIDTH_MAX +# undef PNG_USER_WIDTH_MAX +# define PNG_USER_WIDTH_MAX 1000000L +# endif +# ifdef PNG_USER_HEIGHT_MAX +# undef PNG_USER_HEIGHT_MAX +# define PNG_USER_HEIGHT_MAX 1000000L +# endif +# ifdef PNG_USER_CHUNK_MALLOC_MAX +# undef PNG_USER_CHUNK_MALLOC_MAX +# define PNG_USER_CHUNK_MALLOC_MAX 4000000L +# endif +# ifdef PNG_USER_CHUNK_CACHE_MAX +# undef PNG_USER_CHUNK_CACHE_MAX +# define PNG_USER_CHUNK_CACHE_MAX 128 +# endif +#endif + +#ifndef PNG_BUILDING_SYMBOL_TABLE /* else includes may cause problems */ + +/* From libpng 1.6.0 libpng requires an ANSI X3.159-1989 ("ISOC90") compliant C + * compiler for correct compilation. The following header files are required by + * the standard. If your compiler doesn't provide these header files, or they + * do not match the standard, you will need to provide/improve them. + */ +#include +#include + +/* Library header files. These header files are all defined by ISOC90; libpng + * expects conformant implementations, however, an ISOC90 conformant system need + * not provide these header files if the functionality cannot be implemented. + * In this case it will be necessary to disable the relevant parts of libpng in + * the build of pnglibconf.h. + * + * Prior to 1.6.0 string.h was included here; the API changes in 1.6.0 to not + * include this unnecessary header file. + */ + +#ifdef PNG_STDIO_SUPPORTED + /* Required for the definition of FILE: */ +# include +#endif + +#ifdef PNG_SETJMP_SUPPORTED + /* Required for the definition of jmp_buf and the declaration of longjmp: */ +# include +#endif + +#ifdef PNG_CONVERT_tIME_SUPPORTED + /* Required for struct tm: */ +# include +#endif + +#endif /* PNG_BUILDING_SYMBOL_TABLE */ + +/* Prior to 1.6.0 it was possible to turn off 'const' in declarations using + * PNG_NO_CONST; this is no longer supported except for data declarations which + * apparently still cause problems in 2011 on some compilers. + */ +#define PNG_CONST const /* backward compatibility only */ + +/* This controls optimization of the reading of 16 and 32 bit values + * from PNG files. It can be set on a per-app-file basis - it + * just changes whether a macro is used when the function is called. + * The library builder sets the default; if read functions are not + * built into the library the macro implementation is forced on. + */ +#ifndef PNG_READ_INT_FUNCTIONS_SUPPORTED +# define PNG_USE_READ_MACROS +#endif +#if !defined(PNG_NO_USE_READ_MACROS) && !defined(PNG_USE_READ_MACROS) +# if PNG_DEFAULT_READ_MACROS +# define PNG_USE_READ_MACROS +# endif +#endif + +/* COMPILER SPECIFIC OPTIONS. + * + * These options are provided so that a variety of difficult compilers + * can be used. Some are fixed at build time (e.g. PNG_API_RULE + * below) but still have compiler specific implementations, others + * may be changed on a per-file basis when compiling against libpng. + */ + +/* The PNGARG macro was used in versions of libpng prior to 1.6.0 to protect + * against legacy (pre ISOC90) compilers that did not understand function + * prototypes. It is not required for modern C compilers. + */ +#ifndef PNGARG +# define PNGARG(arglist) arglist +#endif + +/* Function calling conventions. + * ============================= + * Normally it is not necessary to specify to the compiler how to call + * a function - it just does it - however on x86 systems derived from + * Microsoft and Borland C compilers ('IBM PC', 'DOS', 'Windows' systems + * and some others) there are multiple ways to call a function and the + * default can be changed on the compiler command line. For this reason + * libpng specifies the calling convention of every exported function and + * every function called via a user supplied function pointer. This is + * done in this file by defining the following macros: + * + * PNGAPI Calling convention for exported functions. + * PNGCBAPI Calling convention for user provided (callback) functions. + * PNGCAPI Calling convention used by the ANSI-C library (required + * for longjmp callbacks and sometimes used internally to + * specify the calling convention for zlib). + * + * These macros should never be overridden. If it is necessary to + * change calling convention in a private build this can be done + * by setting PNG_API_RULE (which defaults to 0) to one of the values + * below to select the correct 'API' variants. + * + * PNG_API_RULE=0 Use PNGCAPI - the 'C' calling convention - throughout. + * This is correct in every known environment. + * PNG_API_RULE=1 Use the operating system convention for PNGAPI and + * the 'C' calling convention (from PNGCAPI) for + * callbacks (PNGCBAPI). This is no longer required + * in any known environment - if it has to be used + * please post an explanation of the problem to the + * libpng mailing list. + * + * These cases only differ if the operating system does not use the C + * calling convention, at present this just means the above cases + * (x86 DOS/Windows sytems) and, even then, this does not apply to + * Cygwin running on those systems. + * + * Note that the value must be defined in pnglibconf.h so that what + * the application uses to call the library matches the conventions + * set when building the library. + */ + +/* Symbol export + * ============= + * When building a shared library it is almost always necessary to tell + * the compiler which symbols to export. The png.h macro 'PNG_EXPORT' + * is used to mark the symbols. On some systems these symbols can be + * extracted at link time and need no special processing by the compiler, + * on other systems the symbols are flagged by the compiler and just + * the declaration requires a special tag applied (unfortunately) in a + * compiler dependent way. Some systems can do either. + * + * A small number of older systems also require a symbol from a DLL to + * be flagged to the program that calls it. This is a problem because + * we do not know in the header file included by application code that + * the symbol will come from a shared library, as opposed to a statically + * linked one. For this reason the application must tell us by setting + * the magic flag PNG_USE_DLL to turn on the special processing before + * it includes png.h. + * + * Four additional macros are used to make this happen: + * + * PNG_IMPEXP The magic (if any) to cause a symbol to be exported from + * the build or imported if PNG_USE_DLL is set - compiler + * and system specific. + * + * PNG_EXPORT_TYPE(type) A macro that pre or appends PNG_IMPEXP to + * 'type', compiler specific. + * + * PNG_DLL_EXPORT Set to the magic to use during a libpng build to + * make a symbol exported from the DLL. Not used in the + * public header files; see pngpriv.h for how it is used + * in the libpng build. + * + * PNG_DLL_IMPORT Set to the magic to force the libpng symbols to come + * from a DLL - used to define PNG_IMPEXP when + * PNG_USE_DLL is set. + */ + +/* System specific discovery. + * ========================== + * This code is used at build time to find PNG_IMPEXP, the API settings + * and PNG_EXPORT_TYPE(), it may also set a macro to indicate the DLL + * import processing is possible. On Windows systems it also sets + * compiler-specific macros to the values required to change the calling + * conventions of the various functions. + */ +#if defined(_Windows) || defined(_WINDOWS) || defined(WIN32) ||\ + defined(_WIN32) || defined(__WIN32__) || defined(__CYGWIN__) + /* Windows system (DOS doesn't support DLLs). Includes builds under Cygwin or + * MinGW on any architecture currently supported by Windows. Also includes + * Watcom builds but these need special treatment because they are not + * compatible with GCC or Visual C because of different calling conventions. + */ +# if PNG_API_RULE == 2 + /* If this line results in an error, either because __watcall is not + * understood or because of a redefine just below you cannot use *this* + * build of the library with the compiler you are using. *This* build was + * build using Watcom and applications must also be built using Watcom! + */ +# define PNGCAPI __watcall +# endif + +# if defined(__GNUC__) || (defined(_MSC_VER) && (_MSC_VER >= 800)) +# define PNGCAPI __cdecl +# if PNG_API_RULE == 1 + /* If this line results in an error __stdcall is not understood and + * PNG_API_RULE should not have been set to '1'. + */ +# define PNGAPI __stdcall +# endif +# else + /* An older compiler, or one not detected (erroneously) above, + * if necessary override on the command line to get the correct + * variants for the compiler. + */ +# ifndef PNGCAPI +# define PNGCAPI _cdecl +# endif +# if PNG_API_RULE == 1 && !defined(PNGAPI) +# define PNGAPI _stdcall +# endif +# endif /* compiler/api */ + + /* NOTE: PNGCBAPI always defaults to PNGCAPI. */ + +# if defined(PNGAPI) && !defined(PNG_USER_PRIVATEBUILD) +# error "PNG_USER_PRIVATEBUILD must be defined if PNGAPI is changed" +# endif + +# if (defined(_MSC_VER) && _MSC_VER < 800) ||\ + (defined(__BORLANDC__) && __BORLANDC__ < 0x500) + /* older Borland and MSC + * compilers used '__export' and required this to be after + * the type. + */ +# ifndef PNG_EXPORT_TYPE +# define PNG_EXPORT_TYPE(type) type PNG_IMPEXP +# endif +# define PNG_DLL_EXPORT __export +# else /* newer compiler */ +# define PNG_DLL_EXPORT __declspec(dllexport) +# ifndef PNG_DLL_IMPORT +# define PNG_DLL_IMPORT __declspec(dllimport) +# endif +# endif /* compiler */ + +#else /* !Windows */ +# if (defined(__IBMC__) || defined(__IBMCPP__)) && defined(__OS2__) +# define PNGAPI _System +# else /* !Windows/x86 && !OS/2 */ + /* Use the defaults, or define PNG*API on the command line (but + * this will have to be done for every compile!) + */ +# endif /* other system, !OS/2 */ +#endif /* !Windows/x86 */ + +/* Now do all the defaulting . */ +#ifndef PNGCAPI +# define PNGCAPI +#endif +#ifndef PNGCBAPI +# define PNGCBAPI PNGCAPI +#endif +#ifndef PNGAPI +# define PNGAPI PNGCAPI +#endif + +/* PNG_IMPEXP may be set on the compilation system command line or (if not set) + * then in an internal header file when building the library, otherwise (when + * using the library) it is set here. + */ +#ifndef PNG_IMPEXP +# if defined(PNG_USE_DLL) && defined(PNG_DLL_IMPORT) + /* This forces use of a DLL, disallowing static linking */ +# define PNG_IMPEXP PNG_DLL_IMPORT +# endif + +# ifndef PNG_IMPEXP +# define PNG_IMPEXP +# endif +#endif + +/* In 1.5.2 the definition of PNG_FUNCTION has been changed to always treat + * 'attributes' as a storage class - the attributes go at the start of the + * function definition, and attributes are always appended regardless of the + * compiler. This considerably simplifies these macros but may cause problems + * if any compilers both need function attributes and fail to handle them as + * a storage class (this is unlikely.) + */ +#ifndef PNG_FUNCTION +# define PNG_FUNCTION(type, name, args, attributes) attributes type name args +#endif + +#ifndef PNG_EXPORT_TYPE +# define PNG_EXPORT_TYPE(type) PNG_IMPEXP type +#endif + + /* The ordinal value is only relevant when preprocessing png.h for symbol + * table entries, so we discard it here. See the .dfn files in the + * scripts directory. + */ +#ifndef PNG_EXPORTA + +# define PNG_EXPORTA(ordinal, type, name, args, attributes)\ + PNG_FUNCTION(PNG_EXPORT_TYPE(type),(PNGAPI name),PNGARG(args), \ + extern attributes) +#endif + +/* ANSI-C (C90) does not permit a macro to be invoked with an empty argument, + * so make something non-empty to satisfy the requirement: + */ +#define PNG_EMPTY /*empty list*/ + +#define PNG_EXPORT(ordinal, type, name, args)\ + PNG_EXPORTA(ordinal, type, name, args, PNG_EMPTY) + +/* Use PNG_REMOVED to comment out a removed interface. */ +#ifndef PNG_REMOVED +# define PNG_REMOVED(ordinal, type, name, args, attributes) +#endif + +#ifndef PNG_CALLBACK +# define PNG_CALLBACK(type, name, args) type (PNGCBAPI name) PNGARG(args) +#endif + +/* Support for compiler specific function attributes. These are used + * so that where compiler support is available incorrect use of API + * functions in png.h will generate compiler warnings. + * + * Added at libpng-1.2.41. + */ + +#ifndef PNG_NO_PEDANTIC_WARNINGS +# ifndef PNG_PEDANTIC_WARNINGS_SUPPORTED +# define PNG_PEDANTIC_WARNINGS_SUPPORTED +# endif +#endif + +#ifdef PNG_PEDANTIC_WARNINGS_SUPPORTED + /* Support for compiler specific function attributes. These are used + * so that where compiler support is available, incorrect use of API + * functions in png.h will generate compiler warnings. Added at libpng + * version 1.2.41. Disabling these removes the warnings but may also produce + * less efficient code. + */ +# if defined(__clang__) && defined(__has_attribute) + /* Clang defines both __clang__ and __GNUC__. Check __clang__ first. */ +# if !defined(PNG_USE_RESULT) && __has_attribute(__warn_unused_result__) +# define PNG_USE_RESULT __attribute__((__warn_unused_result__)) +# endif +# if !defined(PNG_NORETURN) && __has_attribute(__noreturn__) +# define PNG_NORETURN __attribute__((__noreturn__)) +# endif +# if !defined(PNG_ALLOCATED) && __has_attribute(__malloc__) +# define PNG_ALLOCATED __attribute__((__malloc__)) +# endif +# if !defined(PNG_DEPRECATED) && __has_attribute(__deprecated__) +# define PNG_DEPRECATED __attribute__((__deprecated__)) +# endif +# if !defined(PNG_PRIVATE) +# ifdef __has_extension +# if __has_extension(attribute_unavailable_with_message) +# define PNG_PRIVATE __attribute__((__unavailable__(\ + "This function is not exported by libpng."))) +# endif +# endif +# endif +# ifndef PNG_RESTRICT +# define PNG_RESTRICT __restrict +# endif + +# elif defined(__GNUC__) +# ifndef PNG_USE_RESULT +# define PNG_USE_RESULT __attribute__((__warn_unused_result__)) +# endif +# ifndef PNG_NORETURN +# define PNG_NORETURN __attribute__((__noreturn__)) +# endif +# if __GNUC__ >= 3 +# ifndef PNG_ALLOCATED +# define PNG_ALLOCATED __attribute__((__malloc__)) +# endif +# ifndef PNG_DEPRECATED +# define PNG_DEPRECATED __attribute__((__deprecated__)) +# endif +# ifndef PNG_PRIVATE +# if 0 /* Doesn't work so we use deprecated instead*/ +# define PNG_PRIVATE \ + __attribute__((warning("This function is not exported by libpng."))) +# else +# define PNG_PRIVATE \ + __attribute__((__deprecated__)) +# endif +# endif +# if ((__GNUC__ > 3) || !defined(__GNUC_MINOR__) || (__GNUC_MINOR__ >= 1)) +# ifndef PNG_RESTRICT +# define PNG_RESTRICT __restrict +# endif +# endif /* __GNUC__.__GNUC_MINOR__ > 3.0 */ +# endif /* __GNUC__ >= 3 */ + +# elif defined(_MSC_VER) && (_MSC_VER >= 1300) +# ifndef PNG_USE_RESULT +# define PNG_USE_RESULT /* not supported */ +# endif +# ifndef PNG_NORETURN +# define PNG_NORETURN __declspec(noreturn) +# endif +# ifndef PNG_ALLOCATED +# if (_MSC_VER >= 1400) +# define PNG_ALLOCATED __declspec(restrict) +# endif +# endif +# ifndef PNG_DEPRECATED +# define PNG_DEPRECATED __declspec(deprecated) +# endif +# ifndef PNG_PRIVATE +# define PNG_PRIVATE __declspec(deprecated) +# endif +# ifndef PNG_RESTRICT +# if (_MSC_VER >= 1400) +# define PNG_RESTRICT __restrict +# endif +# endif + +# elif defined(__WATCOMC__) +# ifndef PNG_RESTRICT +# define PNG_RESTRICT __restrict +# endif +# endif +#endif /* PNG_PEDANTIC_WARNINGS */ + +#ifndef PNG_DEPRECATED +# define PNG_DEPRECATED /* Use of this function is deprecated */ +#endif +#ifndef PNG_USE_RESULT +# define PNG_USE_RESULT /* The result of this function must be checked */ +#endif +#ifndef PNG_NORETURN +# define PNG_NORETURN /* This function does not return */ +#endif +#ifndef PNG_ALLOCATED +# define PNG_ALLOCATED /* The result of the function is new memory */ +#endif +#ifndef PNG_PRIVATE +# define PNG_PRIVATE /* This is a private libpng function */ +#endif +#ifndef PNG_RESTRICT +# define PNG_RESTRICT /* The C99 "restrict" feature */ +#endif + +#ifndef PNG_FP_EXPORT /* A floating point API. */ +# ifdef PNG_FLOATING_POINT_SUPPORTED +# define PNG_FP_EXPORT(ordinal, type, name, args)\ + PNG_EXPORT(ordinal, type, name, args); +# else /* No floating point APIs */ +# define PNG_FP_EXPORT(ordinal, type, name, args) +# endif +#endif +#ifndef PNG_FIXED_EXPORT /* A fixed point API. */ +# ifdef PNG_FIXED_POINT_SUPPORTED +# define PNG_FIXED_EXPORT(ordinal, type, name, args)\ + PNG_EXPORT(ordinal, type, name, args); +# else /* No fixed point APIs */ +# define PNG_FIXED_EXPORT(ordinal, type, name, args) +# endif +#endif + +#ifndef PNG_BUILDING_SYMBOL_TABLE +/* Some typedefs to get us started. These should be safe on most of the common + * platforms. + * + * png_uint_32 and png_int_32 may, currently, be larger than required to hold a + * 32-bit value however this is not normally advisable. + * + * png_uint_16 and png_int_16 should always be two bytes in size - this is + * verified at library build time. + * + * png_byte must always be one byte in size. + * + * The checks below use constants from limits.h, as defined by the ISOC90 + * standard. + */ +#if CHAR_BIT == 8 && UCHAR_MAX == 255 + typedef unsigned char png_byte; +#else +# error "libpng requires 8 bit bytes" +#endif + +#if INT_MIN == -32768 && INT_MAX == 32767 + typedef int png_int_16; +#elif SHRT_MIN == -32768 && SHRT_MAX == 32767 + typedef short png_int_16; +#else +# error "libpng requires a signed 16 bit type" +#endif + +#if UINT_MAX == 65535 + typedef unsigned int png_uint_16; +#elif USHRT_MAX == 65535 + typedef unsigned short png_uint_16; +#else +# error "libpng requires an unsigned 16 bit type" +#endif + +#if INT_MIN < -2147483646 && INT_MAX > 2147483646 + typedef int png_int_32; +#elif LONG_MIN < -2147483646 && LONG_MAX > 2147483646 + typedef long int png_int_32; +#else +# error "libpng requires a signed 32 bit (or more) type" +#endif + +#if UINT_MAX > 4294967294 + typedef unsigned int png_uint_32; +#elif ULONG_MAX > 4294967294 + typedef unsigned long int png_uint_32; +#else +# error "libpng requires an unsigned 32 bit (or more) type" +#endif + +/* Prior to 1.6.0 it was possible to disable the use of size_t, 1.6.0, however, + * requires an ISOC90 compiler and relies on consistent behavior of sizeof. + */ +typedef size_t png_size_t; +typedef ptrdiff_t png_ptrdiff_t; + +/* libpng needs to know the maximum value of 'size_t' and this controls the + * definition of png_alloc_size_t, below. This maximum value of size_t limits + * but does not control the maximum allocations the library makes - there is + * direct application control of this through png_set_user_limits(). + */ +#ifndef PNG_SMALL_SIZE_T + /* Compiler specific tests for systems where size_t is known to be less than + * 32 bits (some of these systems may no longer work because of the lack of + * 'far' support; see above.) + */ +# if (defined(__TURBOC__) && !defined(__FLAT__)) ||\ + (defined(_MSC_VER) && defined(MAXSEG_64K)) +# define PNG_SMALL_SIZE_T +# endif +#endif + +/* png_alloc_size_t is guaranteed to be no smaller than png_size_t, and no + * smaller than png_uint_32. Casts from png_size_t or png_uint_32 to + * png_alloc_size_t are not necessary; in fact, it is recommended not to use + * them at all so that the compiler can complain when something turns out to be + * problematic. + * + * Casts in the other direction (from png_alloc_size_t to png_size_t or + * png_uint_32) should be explicitly applied; however, we do not expect to + * encounter practical situations that require such conversions. + * + * PNG_SMALL_SIZE_T must be defined if the maximum value of size_t is less than + * 4294967295 - i.e. less than the maximum value of png_uint_32. + */ +#ifdef PNG_SMALL_SIZE_T + typedef png_uint_32 png_alloc_size_t; +#else + typedef png_size_t png_alloc_size_t; +#endif + +/* Prior to 1.6.0 libpng offered limited support for Microsoft C compiler + * implementations of Intel CPU specific support of user-mode segmented address + * spaces, where 16-bit pointers address more than 65536 bytes of memory using + * separate 'segment' registers. The implementation requires two different + * types of pointer (only one of which includes the segment value.) + * + * If required this support is available in version 1.2 of libpng and may be + * available in versions through 1.5, although the correctness of the code has + * not been verified recently. + */ + +/* Typedef for floating-point numbers that are converted to fixed-point with a + * multiple of 100,000, e.g., gamma + */ +typedef png_int_32 png_fixed_point; + +/* Add typedefs for pointers */ +typedef void * png_voidp; +typedef const void * png_const_voidp; +typedef png_byte * png_bytep; +typedef const png_byte * png_const_bytep; +typedef png_uint_32 * png_uint_32p; +typedef const png_uint_32 * png_const_uint_32p; +typedef png_int_32 * png_int_32p; +typedef const png_int_32 * png_const_int_32p; +typedef png_uint_16 * png_uint_16p; +typedef const png_uint_16 * png_const_uint_16p; +typedef png_int_16 * png_int_16p; +typedef const png_int_16 * png_const_int_16p; +typedef char * png_charp; +typedef const char * png_const_charp; +typedef png_fixed_point * png_fixed_point_p; +typedef const png_fixed_point * png_const_fixed_point_p; +typedef png_size_t * png_size_tp; +typedef const png_size_t * png_const_size_tp; + +#ifdef PNG_STDIO_SUPPORTED +typedef FILE * png_FILE_p; +#endif + +#ifdef PNG_FLOATING_POINT_SUPPORTED +typedef double * png_doublep; +typedef const double * png_const_doublep; +#endif + +/* Pointers to pointers; i.e. arrays */ +typedef png_byte * * png_bytepp; +typedef png_uint_32 * * png_uint_32pp; +typedef png_int_32 * * png_int_32pp; +typedef png_uint_16 * * png_uint_16pp; +typedef png_int_16 * * png_int_16pp; +typedef const char * * png_const_charpp; +typedef char * * png_charpp; +typedef png_fixed_point * * png_fixed_point_pp; +#ifdef PNG_FLOATING_POINT_SUPPORTED +typedef double * * png_doublepp; +#endif + +/* Pointers to pointers to pointers; i.e., pointer to array */ +typedef char * * * png_charppp; + +#endif /* PNG_BUILDING_SYMBOL_TABLE */ + +#endif /* PNGCONF_H */ diff --git a/app/src/main/jni/libpng/pngdebug.h b/app/src/main/jni/libpng/pngdebug.h new file mode 100644 index 000000000..b43c59cde --- /dev/null +++ b/app/src/main/jni/libpng/pngdebug.h @@ -0,0 +1,154 @@ + +/* pngdebug.h - Debugging macros for libpng, also used in pngtest.c + * + * Copyright (c) 1998-2013 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * Last changed in libpng 1.6.8 [December 19, 2013] + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + */ + +/* Define PNG_DEBUG at compile time for debugging information. Higher + * numbers for PNG_DEBUG mean more debugging information. This has + * only been added since version 0.95 so it is not implemented throughout + * libpng yet, but more support will be added as needed. + * + * png_debug[1-2]?(level, message ,arg{0-2}) + * Expands to a statement (either a simple expression or a compound + * do..while(0) statement) that outputs a message with parameter + * substitution if PNG_DEBUG is defined to 2 or more. If PNG_DEBUG + * is undefined, 0 or 1 every png_debug expands to a simple expression + * (actually ((void)0)). + * + * level: level of detail of message, starting at 0. A level 'n' + * message is preceded by 'n' 3-space indentations (not implemented + * on Microsoft compilers unless PNG_DEBUG_FILE is also + * defined, to allow debug DLL compilation with no standard IO). + * message: a printf(3) style text string. A trailing '\n' is added + * to the message. + * arg: 0 to 2 arguments for printf(3) style substitution in message. + */ +#ifndef PNGDEBUG_H +#define PNGDEBUG_H +/* These settings control the formatting of messages in png.c and pngerror.c */ +/* Moved to pngdebug.h at 1.5.0 */ +# ifndef PNG_LITERAL_SHARP +# define PNG_LITERAL_SHARP 0x23 +# endif +# ifndef PNG_LITERAL_LEFT_SQUARE_BRACKET +# define PNG_LITERAL_LEFT_SQUARE_BRACKET 0x5b +# endif +# ifndef PNG_LITERAL_RIGHT_SQUARE_BRACKET +# define PNG_LITERAL_RIGHT_SQUARE_BRACKET 0x5d +# endif +# ifndef PNG_STRING_NEWLINE +# define PNG_STRING_NEWLINE "\n" +# endif + +#ifdef PNG_DEBUG +# if (PNG_DEBUG > 0) +# if !defined(PNG_DEBUG_FILE) && defined(_MSC_VER) +# include +# if (PNG_DEBUG > 1) +# ifndef _DEBUG +# define _DEBUG +# endif +# ifndef png_debug +# define png_debug(l,m) _RPT0(_CRT_WARN,m PNG_STRING_NEWLINE) +# endif +# ifndef png_debug1 +# define png_debug1(l,m,p1) _RPT1(_CRT_WARN,m PNG_STRING_NEWLINE,p1) +# endif +# ifndef png_debug2 +# define png_debug2(l,m,p1,p2) \ + _RPT2(_CRT_WARN,m PNG_STRING_NEWLINE,p1,p2) +# endif +# endif +# else /* PNG_DEBUG_FILE || !_MSC_VER */ +# ifndef PNG_STDIO_SUPPORTED +# include /* not included yet */ +# endif +# ifndef PNG_DEBUG_FILE +# define PNG_DEBUG_FILE stderr +# endif /* PNG_DEBUG_FILE */ + +# if (PNG_DEBUG > 1) +# ifdef __STDC__ +# ifndef png_debug +# define png_debug(l,m) \ + do { \ + int num_tabs=l; \ + fprintf(PNG_DEBUG_FILE,"%s" m PNG_STRING_NEWLINE,(num_tabs==1 ? " " : \ + (num_tabs==2 ? " " : (num_tabs>2 ? " " : "")))); \ + } while (0) +# endif +# ifndef png_debug1 +# define png_debug1(l,m,p1) \ + do { \ + int num_tabs=l; \ + fprintf(PNG_DEBUG_FILE,"%s" m PNG_STRING_NEWLINE,(num_tabs==1 ? " " : \ + (num_tabs==2 ? " " : (num_tabs>2 ? " " : ""))),p1); \ + } while (0) +# endif +# ifndef png_debug2 +# define png_debug2(l,m,p1,p2) \ + do { \ + int num_tabs=l; \ + fprintf(PNG_DEBUG_FILE,"%s" m PNG_STRING_NEWLINE,(num_tabs==1 ? " " : \ + (num_tabs==2 ? " " : (num_tabs>2 ? " " : ""))),p1,p2);\ + } while (0) +# endif +# else /* __STDC __ */ +# ifndef png_debug +# define png_debug(l,m) \ + do { \ + int num_tabs=l; \ + char format[256]; \ + snprintf(format,256,"%s%s%s",(num_tabs==1 ? "\t" : \ + (num_tabs==2 ? "\t\t":(num_tabs>2 ? "\t\t\t":""))), \ + m,PNG_STRING_NEWLINE); \ + fprintf(PNG_DEBUG_FILE,format); \ + } while (0) +# endif +# ifndef png_debug1 +# define png_debug1(l,m,p1) \ + do { \ + int num_tabs=l; \ + char format[256]; \ + snprintf(format,256,"%s%s%s",(num_tabs==1 ? "\t" : \ + (num_tabs==2 ? "\t\t":(num_tabs>2 ? "\t\t\t":""))), \ + m,PNG_STRING_NEWLINE); \ + fprintf(PNG_DEBUG_FILE,format,p1); \ + } while (0) +# endif +# ifndef png_debug2 +# define png_debug2(l,m,p1,p2) \ + do { \ + int num_tabs=l; \ + char format[256]; \ + snprintf(format,256,"%s%s%s",(num_tabs==1 ? "\t" : \ + (num_tabs==2 ? "\t\t":(num_tabs>2 ? "\t\t\t":""))), \ + m,PNG_STRING_NEWLINE); \ + fprintf(PNG_DEBUG_FILE,format,p1,p2); \ + } while (0) +# endif +# endif /* __STDC __ */ +# endif /* (PNG_DEBUG > 1) */ + +# endif /* _MSC_VER */ +# endif /* (PNG_DEBUG > 0) */ +#endif /* PNG_DEBUG */ +#ifndef png_debug +# define png_debug(l, m) ((void)0) +#endif +#ifndef png_debug1 +# define png_debug1(l, m, p1) ((void)0) +#endif +#ifndef png_debug2 +# define png_debug2(l, m, p1, p2) ((void)0) +#endif +#endif /* PNGDEBUG_H */ diff --git a/app/src/main/jni/libpng/pngerror.c b/app/src/main/jni/libpng/pngerror.c new file mode 100644 index 000000000..0781866a8 --- /dev/null +++ b/app/src/main/jni/libpng/pngerror.c @@ -0,0 +1,963 @@ + +/* pngerror.c - stub functions for i/o and memory allocation + * + * Last changed in libpng 1.6.15 [November 20, 2014] + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + * + * This file provides a location for all error handling. Users who + * need special error handling are expected to write replacement functions + * and use png_set_error_fn() to use those functions. See the instructions + * at each function. + */ + +#include "pngpriv.h" + +#if defined(PNG_READ_SUPPORTED) || defined(PNG_WRITE_SUPPORTED) + +static PNG_FUNCTION(void, png_default_error,PNGARG((png_const_structrp png_ptr, + png_const_charp error_message)),PNG_NORETURN); + +#ifdef PNG_WARNINGS_SUPPORTED +static void /* PRIVATE */ +png_default_warning PNGARG((png_const_structrp png_ptr, + png_const_charp warning_message)); +#endif /* WARNINGS */ + +/* This function is called whenever there is a fatal error. This function + * should not be changed. If there is a need to handle errors differently, + * you should supply a replacement error function and use png_set_error_fn() + * to replace the error function at run-time. + */ +#ifdef PNG_ERROR_TEXT_SUPPORTED +PNG_FUNCTION(void,PNGAPI +png_error,(png_const_structrp png_ptr, png_const_charp error_message), + PNG_NORETURN) +{ +#ifdef PNG_ERROR_NUMBERS_SUPPORTED + char msg[16]; + if (png_ptr != NULL) + { + if ((png_ptr->flags & + (PNG_FLAG_STRIP_ERROR_NUMBERS|PNG_FLAG_STRIP_ERROR_TEXT)) != 0 + { + if (*error_message == PNG_LITERAL_SHARP) + { + /* Strip "#nnnn " from beginning of error message. */ + int offset; + for (offset = 1; offset<15; offset++) + if (error_message[offset] == ' ') + break; + + if ((png_ptr->flags & PNG_FLAG_STRIP_ERROR_TEXT) != 0) + { + int i; + for (i = 0; i < offset - 1; i++) + msg[i] = error_message[i + 1]; + msg[i - 1] = '\0'; + error_message = msg; + } + + else + error_message += offset; + } + + else + { + if ((png_ptr->flags & PNG_FLAG_STRIP_ERROR_TEXT) != 0) + { + msg[0] = '0'; + msg[1] = '\0'; + error_message = msg; + } + } + } + } +#endif + if (png_ptr != NULL && png_ptr->error_fn != NULL) + (*(png_ptr->error_fn))(png_constcast(png_structrp,png_ptr), + error_message); + + /* If the custom handler doesn't exist, or if it returns, + use the default handler, which will not return. */ + png_default_error(png_ptr, error_message); +} +#else +PNG_FUNCTION(void,PNGAPI +png_err,(png_const_structrp png_ptr),PNG_NORETURN) +{ + /* Prior to 1.5.2 the error_fn received a NULL pointer, expressed + * erroneously as '\0', instead of the empty string "". This was + * apparently an error, introduced in libpng-1.2.20, and png_default_error + * will crash in this case. + */ + if (png_ptr != NULL && png_ptr->error_fn != NULL) + (*(png_ptr->error_fn))(png_constcast(png_structrp,png_ptr), ""); + + /* If the custom handler doesn't exist, or if it returns, + use the default handler, which will not return. */ + png_default_error(png_ptr, ""); +} +#endif /* ERROR_TEXT */ + +/* Utility to safely appends strings to a buffer. This never errors out so + * error checking is not required in the caller. + */ +size_t +png_safecat(png_charp buffer, size_t bufsize, size_t pos, + png_const_charp string) +{ + if (buffer != NULL && pos < bufsize) + { + if (string != NULL) + while (*string != '\0' && pos < bufsize-1) + buffer[pos++] = *string++; + + buffer[pos] = '\0'; + } + + return pos; +} + +#if defined(PNG_WARNINGS_SUPPORTED) || defined(PNG_TIME_RFC1123_SUPPORTED) +/* Utility to dump an unsigned value into a buffer, given a start pointer and + * and end pointer (which should point just *beyond* the end of the buffer!) + * Returns the pointer to the start of the formatted string. + */ +png_charp +png_format_number(png_const_charp start, png_charp end, int format, + png_alloc_size_t number) +{ + int count = 0; /* number of digits output */ + int mincount = 1; /* minimum number required */ + int output = 0; /* digit output (for the fixed point format) */ + + *--end = '\0'; + + /* This is written so that the loop always runs at least once, even with + * number zero. + */ + while (end > start && (number != 0 || count < mincount)) + { + + static const char digits[] = "0123456789ABCDEF"; + + switch (format) + { + case PNG_NUMBER_FORMAT_fixed: + /* Needs five digits (the fraction) */ + mincount = 5; + if (output != 0 || number % 10 != 0) + { + *--end = digits[number % 10]; + output = 1; + } + number /= 10; + break; + + case PNG_NUMBER_FORMAT_02u: + /* Expects at least 2 digits. */ + mincount = 2; + /* FALL THROUGH */ + + case PNG_NUMBER_FORMAT_u: + *--end = digits[number % 10]; + number /= 10; + break; + + case PNG_NUMBER_FORMAT_02x: + /* This format expects at least two digits */ + mincount = 2; + /* FALL THROUGH */ + + case PNG_NUMBER_FORMAT_x: + *--end = digits[number & 0xf]; + number >>= 4; + break; + + default: /* an error */ + number = 0; + break; + } + + /* Keep track of the number of digits added */ + ++count; + + /* Float a fixed number here: */ + if ((format == PNG_NUMBER_FORMAT_fixed) && (count == 5) && (end > start)) + { + /* End of the fraction, but maybe nothing was output? In that case + * drop the decimal point. If the number is a true zero handle that + * here. + */ + if (output != 0) + *--end = '.'; + else if (number == 0) /* and !output */ + *--end = '0'; + } + } + + return end; +} +#endif + +#ifdef PNG_WARNINGS_SUPPORTED +/* This function is called whenever there is a non-fatal error. This function + * should not be changed. If there is a need to handle warnings differently, + * you should supply a replacement warning function and use + * png_set_error_fn() to replace the warning function at run-time. + */ +void PNGAPI +png_warning(png_const_structrp png_ptr, png_const_charp warning_message) +{ + int offset = 0; + if (png_ptr != NULL) + { +#ifdef PNG_ERROR_NUMBERS_SUPPORTED + if ((png_ptr->flags & + (PNG_FLAG_STRIP_ERROR_NUMBERS|PNG_FLAG_STRIP_ERROR_TEXT)) != 0) +#endif + { + if (*warning_message == PNG_LITERAL_SHARP) + { + for (offset = 1; offset < 15; offset++) + if (warning_message[offset] == ' ') + break; + } + } + } + if (png_ptr != NULL && png_ptr->warning_fn != NULL) + (*(png_ptr->warning_fn))(png_constcast(png_structrp,png_ptr), + warning_message + offset); + else + png_default_warning(png_ptr, warning_message + offset); +} + +/* These functions support 'formatted' warning messages with up to + * PNG_WARNING_PARAMETER_COUNT parameters. In the format string the parameter + * is introduced by @, where 'number' starts at 1. This follows the + * standard established by X/Open for internationalizable error messages. + */ +void +png_warning_parameter(png_warning_parameters p, int number, + png_const_charp string) +{ + if (number > 0 && number <= PNG_WARNING_PARAMETER_COUNT) + (void)png_safecat(p[number-1], (sizeof p[number-1]), 0, string); +} + +void +png_warning_parameter_unsigned(png_warning_parameters p, int number, int format, + png_alloc_size_t value) +{ + char buffer[PNG_NUMBER_BUFFER_SIZE]; + png_warning_parameter(p, number, PNG_FORMAT_NUMBER(buffer, format, value)); +} + +void +png_warning_parameter_signed(png_warning_parameters p, int number, int format, + png_int_32 value) +{ + png_alloc_size_t u; + png_charp str; + char buffer[PNG_NUMBER_BUFFER_SIZE]; + + /* Avoid overflow by doing the negate in a png_alloc_size_t: */ + u = (png_alloc_size_t)value; + if (value < 0) + u = ~u + 1; + + str = PNG_FORMAT_NUMBER(buffer, format, u); + + if (value < 0 && str > buffer) + *--str = '-'; + + png_warning_parameter(p, number, str); +} + +void +png_formatted_warning(png_const_structrp png_ptr, png_warning_parameters p, + png_const_charp message) +{ + /* The internal buffer is just 192 bytes - enough for all our messages, + * overflow doesn't happen because this code checks! If someone figures + * out how to send us a message longer than 192 bytes, all that will + * happen is that the message will be truncated appropriately. + */ + size_t i = 0; /* Index in the msg[] buffer: */ + char msg[192]; + + /* Each iteration through the following loop writes at most one character + * to msg[i++] then returns here to validate that there is still space for + * the trailing '\0'. It may (in the case of a parameter) read more than + * one character from message[]; it must check for '\0' and continue to the + * test if it finds the end of string. + */ + while (i<(sizeof msg)-1 && *message != '\0') + { + /* '@' at end of string is now just printed (previously it was skipped); + * it is an error in the calling code to terminate the string with @. + */ + if (p != NULL && *message == '@' && message[1] != '\0') + { + int parameter_char = *++message; /* Consume the '@' */ + static const char valid_parameters[] = "123456789"; + int parameter = 0; + + /* Search for the parameter digit, the index in the string is the + * parameter to use. + */ + while (valid_parameters[parameter] != parameter_char && + valid_parameters[parameter] != '\0') + ++parameter; + + /* If the parameter digit is out of range it will just get printed. */ + if (parameter < PNG_WARNING_PARAMETER_COUNT) + { + /* Append this parameter */ + png_const_charp parm = p[parameter]; + png_const_charp pend = p[parameter] + (sizeof p[parameter]); + + /* No need to copy the trailing '\0' here, but there is no guarantee + * that parm[] has been initialized, so there is no guarantee of a + * trailing '\0': + */ + while (i<(sizeof msg)-1 && *parm != '\0' && parm < pend) + msg[i++] = *parm++; + + /* Consume the parameter digit too: */ + ++message; + continue; + } + + /* else not a parameter and there is a character after the @ sign; just + * copy that. This is known not to be '\0' because of the test above. + */ + } + + /* At this point *message can't be '\0', even in the bad parameter case + * above where there is a lone '@' at the end of the message string. + */ + msg[i++] = *message++; + } + + /* i is always less than (sizeof msg), so: */ + msg[i] = '\0'; + + /* And this is the formatted message. It may be larger than + * PNG_MAX_ERROR_TEXT, but that is only used for 'chunk' errors and these + * are not (currently) formatted. + */ + png_warning(png_ptr, msg); +} +#endif /* WARNINGS */ + +#ifdef PNG_BENIGN_ERRORS_SUPPORTED +void PNGAPI +png_benign_error(png_const_structrp png_ptr, png_const_charp error_message) +{ + if ((png_ptr->flags & PNG_FLAG_BENIGN_ERRORS_WARN) != 0) + { +# ifdef PNG_READ_SUPPORTED + if ((png_ptr->mode & PNG_IS_READ_STRUCT) != 0 && + png_ptr->chunk_name != 0) + png_chunk_warning(png_ptr, error_message); + else +# endif + png_warning(png_ptr, error_message); + } + + else + { +# ifdef PNG_READ_SUPPORTED + if ((png_ptr->mode & PNG_IS_READ_STRUCT) != 0 && + png_ptr->chunk_name != 0) + png_chunk_error(png_ptr, error_message); + else +# endif + png_error(png_ptr, error_message); + } + +# ifndef PNG_ERROR_TEXT_SUPPORTED + PNG_UNUSED(error_message) +# endif +} + +void /* PRIVATE */ +png_app_warning(png_const_structrp png_ptr, png_const_charp error_message) +{ + if ((png_ptr->flags & PNG_FLAG_APP_WARNINGS_WARN) != 0) + png_warning(png_ptr, error_message); + else + png_error(png_ptr, error_message); + +# ifndef PNG_ERROR_TEXT_SUPPORTED + PNG_UNUSED(error_message) +# endif +} + +void /* PRIVATE */ +png_app_error(png_const_structrp png_ptr, png_const_charp error_message) +{ + if ((png_ptr->flags & PNG_FLAG_APP_ERRORS_WARN) != 0) + png_warning(png_ptr, error_message); + else + png_error(png_ptr, error_message); + +# ifndef PNG_ERROR_TEXT_SUPPORTED + PNG_UNUSED(error_message) +# endif +} +#endif /* BENIGN_ERRORS */ + +#define PNG_MAX_ERROR_TEXT 196 /* Currently limited by profile_error in png.c */ +#if defined(PNG_WARNINGS_SUPPORTED) || \ + (defined(PNG_READ_SUPPORTED) && defined(PNG_ERROR_TEXT_SUPPORTED)) +/* These utilities are used internally to build an error message that relates + * to the current chunk. The chunk name comes from png_ptr->chunk_name, + * which is used to prefix the message. The message is limited in length + * to 63 bytes. The name characters are output as hex digits wrapped in [] + * if the character is invalid. + */ +#define isnonalpha(c) ((c) < 65 || (c) > 122 || ((c) > 90 && (c) < 97)) +static PNG_CONST char png_digit[16] = { + '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', + 'A', 'B', 'C', 'D', 'E', 'F' +}; + +static void /* PRIVATE */ +png_format_buffer(png_const_structrp png_ptr, png_charp buffer, png_const_charp + error_message) +{ + png_uint_32 chunk_name = png_ptr->chunk_name; + int iout = 0, ishift = 24; + + while (ishift >= 0) + { + int c = (int)(chunk_name >> ishift) & 0xff; + + ishift -= 8; + if (isnonalpha(c) != 0) + { + buffer[iout++] = PNG_LITERAL_LEFT_SQUARE_BRACKET; + buffer[iout++] = png_digit[(c & 0xf0) >> 4]; + buffer[iout++] = png_digit[c & 0x0f]; + buffer[iout++] = PNG_LITERAL_RIGHT_SQUARE_BRACKET; + } + + else + { + buffer[iout++] = (char)c; + } + } + + if (error_message == NULL) + buffer[iout] = '\0'; + + else + { + int iin = 0; + + buffer[iout++] = ':'; + buffer[iout++] = ' '; + + while (iin < PNG_MAX_ERROR_TEXT-1 && error_message[iin] != '\0') + buffer[iout++] = error_message[iin++]; + + /* iin < PNG_MAX_ERROR_TEXT, so the following is safe: */ + buffer[iout] = '\0'; + } +} +#endif /* WARNINGS || ERROR_TEXT */ + +#if defined(PNG_READ_SUPPORTED) && defined(PNG_ERROR_TEXT_SUPPORTED) +PNG_FUNCTION(void,PNGAPI +png_chunk_error,(png_const_structrp png_ptr, png_const_charp error_message), + PNG_NORETURN) +{ + char msg[18+PNG_MAX_ERROR_TEXT]; + if (png_ptr == NULL) + png_error(png_ptr, error_message); + + else + { + png_format_buffer(png_ptr, msg, error_message); + png_error(png_ptr, msg); + } +} +#endif /* READ && ERROR_TEXT */ + +#ifdef PNG_WARNINGS_SUPPORTED +void PNGAPI +png_chunk_warning(png_const_structrp png_ptr, png_const_charp warning_message) +{ + char msg[18+PNG_MAX_ERROR_TEXT]; + if (png_ptr == NULL) + png_warning(png_ptr, warning_message); + + else + { + png_format_buffer(png_ptr, msg, warning_message); + png_warning(png_ptr, msg); + } +} +#endif /* WARNINGS */ + +#ifdef PNG_READ_SUPPORTED +#ifdef PNG_BENIGN_ERRORS_SUPPORTED +void PNGAPI +png_chunk_benign_error(png_const_structrp png_ptr, png_const_charp + error_message) +{ + if ((png_ptr->flags & PNG_FLAG_BENIGN_ERRORS_WARN) != 0) + png_chunk_warning(png_ptr, error_message); + + else + png_chunk_error(png_ptr, error_message); + +# ifndef PNG_ERROR_TEXT_SUPPORTED + PNG_UNUSED(error_message) +# endif +} +#endif +#endif /* READ */ + +void /* PRIVATE */ +png_chunk_report(png_const_structrp png_ptr, png_const_charp message, int error) +{ +# ifndef PNG_WARNINGS_SUPPORTED + PNG_UNUSED(message) +# endif + + /* This is always supported, but for just read or just write it + * unconditionally does the right thing. + */ +# if defined(PNG_READ_SUPPORTED) && defined(PNG_WRITE_SUPPORTED) + if ((png_ptr->mode & PNG_IS_READ_STRUCT) != 0) +# endif + +# ifdef PNG_READ_SUPPORTED + { + if (error < PNG_CHUNK_ERROR) + png_chunk_warning(png_ptr, message); + + else + png_chunk_benign_error(png_ptr, message); + } +# endif + +# if defined(PNG_READ_SUPPORTED) && defined(PNG_WRITE_SUPPORTED) + else if ((png_ptr->mode & PNG_IS_READ_STRUCT) == 0) +# endif + +# ifdef PNG_WRITE_SUPPORTED + { + if (error < PNG_CHUNK_WRITE_ERROR) + png_app_warning(png_ptr, message); + + else + png_app_error(png_ptr, message); + } +# endif +} + +#ifdef PNG_ERROR_TEXT_SUPPORTED +#ifdef PNG_FLOATING_POINT_SUPPORTED +PNG_FUNCTION(void, +png_fixed_error,(png_const_structrp png_ptr, png_const_charp name),PNG_NORETURN) +{ +# define fixed_message "fixed point overflow in " +# define fixed_message_ln ((sizeof fixed_message)-1) + int iin; + char msg[fixed_message_ln+PNG_MAX_ERROR_TEXT]; + memcpy(msg, fixed_message, fixed_message_ln); + iin = 0; + if (name != NULL) + while (iin < (PNG_MAX_ERROR_TEXT-1) && name[iin] != 0) + { + msg[fixed_message_ln + iin] = name[iin]; + ++iin; + } + msg[fixed_message_ln + iin] = 0; + png_error(png_ptr, msg); +} +#endif +#endif + +#ifdef PNG_SETJMP_SUPPORTED +/* This API only exists if ANSI-C style error handling is used, + * otherwise it is necessary for png_default_error to be overridden. + */ +jmp_buf* PNGAPI +png_set_longjmp_fn(png_structrp png_ptr, png_longjmp_ptr longjmp_fn, + size_t jmp_buf_size) +{ + /* From libpng 1.6.0 the app gets one chance to set a 'jmpbuf_size' value + * and it must not change after that. Libpng doesn't care how big the + * buffer is, just that it doesn't change. + * + * If the buffer size is no *larger* than the size of jmp_buf when libpng is + * compiled a built in jmp_buf is returned; this preserves the pre-1.6.0 + * semantics that this call will not fail. If the size is larger, however, + * the buffer is allocated and this may fail, causing the function to return + * NULL. + */ + if (png_ptr == NULL) + return NULL; + + if (png_ptr->jmp_buf_ptr == NULL) + { + png_ptr->jmp_buf_size = 0; /* not allocated */ + + if (jmp_buf_size <= (sizeof png_ptr->jmp_buf_local)) + png_ptr->jmp_buf_ptr = &png_ptr->jmp_buf_local; + + else + { + png_ptr->jmp_buf_ptr = png_voidcast(jmp_buf *, + png_malloc_warn(png_ptr, jmp_buf_size)); + + if (png_ptr->jmp_buf_ptr == NULL) + return NULL; /* new NULL return on OOM */ + + png_ptr->jmp_buf_size = jmp_buf_size; + } + } + + else /* Already allocated: check the size */ + { + size_t size = png_ptr->jmp_buf_size; + + if (size == 0) + { + size = (sizeof png_ptr->jmp_buf_local); + if (png_ptr->jmp_buf_ptr != &png_ptr->jmp_buf_local) + { + /* This is an internal error in libpng: somehow we have been left + * with a stack allocated jmp_buf when the application regained + * control. It's always possible to fix this up, but for the moment + * this is a png_error because that makes it easy to detect. + */ + png_error(png_ptr, "Libpng jmp_buf still allocated"); + /* png_ptr->jmp_buf_ptr = &png_ptr->jmp_buf_local; */ + } + } + + if (size != jmp_buf_size) + { + png_warning(png_ptr, "Application jmp_buf size changed"); + return NULL; /* caller will probably crash: no choice here */ + } + } + + /* Finally fill in the function, now we have a satisfactory buffer. It is + * valid to change the function on every call. + */ + png_ptr->longjmp_fn = longjmp_fn; + return png_ptr->jmp_buf_ptr; +} + +void /* PRIVATE */ +png_free_jmpbuf(png_structrp png_ptr) +{ + if (png_ptr != NULL) + { + jmp_buf *jb = png_ptr->jmp_buf_ptr; + + /* A size of 0 is used to indicate a local, stack, allocation of the + * pointer; used here and in png.c + */ + if (jb != NULL && png_ptr->jmp_buf_size > 0) + { + + /* This stuff is so that a failure to free the error control structure + * does not leave libpng in a state with no valid error handling: the + * free always succeeds, if there is an error it gets ignored. + */ + if (jb != &png_ptr->jmp_buf_local) + { + /* Make an internal, libpng, jmp_buf to return here */ + jmp_buf free_jmp_buf; + + if (!setjmp(free_jmp_buf)) + { + png_ptr->jmp_buf_ptr = &free_jmp_buf; /* come back here */ + png_ptr->jmp_buf_size = 0; /* stack allocation */ + png_ptr->longjmp_fn = longjmp; + png_free(png_ptr, jb); /* Return to setjmp on error */ + } + } + } + + /* *Always* cancel everything out: */ + png_ptr->jmp_buf_size = 0; + png_ptr->jmp_buf_ptr = NULL; + png_ptr->longjmp_fn = 0; + } +} +#endif + +/* This is the default error handling function. Note that replacements for + * this function MUST NOT RETURN, or the program will likely crash. This + * function is used by default, or if the program supplies NULL for the + * error function pointer in png_set_error_fn(). + */ +static PNG_FUNCTION(void /* PRIVATE */, +png_default_error,(png_const_structrp png_ptr, png_const_charp error_message), + PNG_NORETURN) +{ +#ifdef PNG_CONSOLE_IO_SUPPORTED +#ifdef PNG_ERROR_NUMBERS_SUPPORTED + /* Check on NULL only added in 1.5.4 */ + if (error_message != NULL && *error_message == PNG_LITERAL_SHARP) + { + /* Strip "#nnnn " from beginning of error message. */ + int offset; + char error_number[16]; + for (offset = 0; offset<15; offset++) + { + error_number[offset] = error_message[offset + 1]; + if (error_message[offset] == ' ') + break; + } + + if ((offset > 1) && (offset < 15)) + { + error_number[offset - 1] = '\0'; + fprintf(stderr, "libpng error no. %s: %s", + error_number, error_message + offset + 1); + fprintf(stderr, PNG_STRING_NEWLINE); + } + + else + { + fprintf(stderr, "libpng error: %s, offset=%d", + error_message, offset); + fprintf(stderr, PNG_STRING_NEWLINE); + } + } + else +#endif + { + fprintf(stderr, "libpng error: %s", error_message ? error_message : + "undefined"); + fprintf(stderr, PNG_STRING_NEWLINE); + } +#else + PNG_UNUSED(error_message) /* Make compiler happy */ +#endif + png_longjmp(png_ptr, 1); +} + +PNG_FUNCTION(void,PNGAPI +png_longjmp,(png_const_structrp png_ptr, int val),PNG_NORETURN) +{ +#ifdef PNG_SETJMP_SUPPORTED + if (png_ptr != NULL && png_ptr->longjmp_fn != NULL && + png_ptr->jmp_buf_ptr != NULL) + png_ptr->longjmp_fn(*png_ptr->jmp_buf_ptr, val); +#else + PNG_UNUSED(png_ptr) + PNG_UNUSED(val) +#endif + + /* If control reaches this point, png_longjmp() must not return. The only + * choice is to terminate the whole process (or maybe the thread); to do + * this the ANSI-C abort() function is used unless a different method is + * implemented by overriding the default configuration setting for + * PNG_ABORT(). + */ + PNG_ABORT(); +} + +#ifdef PNG_WARNINGS_SUPPORTED +/* This function is called when there is a warning, but the library thinks + * it can continue anyway. Replacement functions don't have to do anything + * here if you don't want them to. In the default configuration, png_ptr is + * not used, but it is passed in case it may be useful. + */ +static void /* PRIVATE */ +png_default_warning(png_const_structrp png_ptr, png_const_charp warning_message) +{ +#ifdef PNG_CONSOLE_IO_SUPPORTED +# ifdef PNG_ERROR_NUMBERS_SUPPORTED + if (*warning_message == PNG_LITERAL_SHARP) + { + int offset; + char warning_number[16]; + for (offset = 0; offset < 15; offset++) + { + warning_number[offset] = warning_message[offset + 1]; + if (warning_message[offset] == ' ') + break; + } + + if ((offset > 1) && (offset < 15)) + { + warning_number[offset + 1] = '\0'; + fprintf(stderr, "libpng warning no. %s: %s", + warning_number, warning_message + offset); + fprintf(stderr, PNG_STRING_NEWLINE); + } + + else + { + fprintf(stderr, "libpng warning: %s", + warning_message); + fprintf(stderr, PNG_STRING_NEWLINE); + } + } + else +# endif + + { + fprintf(stderr, "libpng warning: %s", warning_message); + fprintf(stderr, PNG_STRING_NEWLINE); + } +#else + PNG_UNUSED(warning_message) /* Make compiler happy */ +#endif + PNG_UNUSED(png_ptr) /* Make compiler happy */ +} +#endif /* WARNINGS */ + +/* This function is called when the application wants to use another method + * of handling errors and warnings. Note that the error function MUST NOT + * return to the calling routine or serious problems will occur. The return + * method used in the default routine calls longjmp(png_ptr->jmp_buf_ptr, 1) + */ +void PNGAPI +png_set_error_fn(png_structrp png_ptr, png_voidp error_ptr, + png_error_ptr error_fn, png_error_ptr warning_fn) +{ + if (png_ptr == NULL) + return; + + png_ptr->error_ptr = error_ptr; + png_ptr->error_fn = error_fn; +#ifdef PNG_WARNINGS_SUPPORTED + png_ptr->warning_fn = warning_fn; +#else + PNG_UNUSED(warning_fn) +#endif +} + + +/* This function returns a pointer to the error_ptr associated with the user + * functions. The application should free any memory associated with this + * pointer before png_write_destroy and png_read_destroy are called. + */ +png_voidp PNGAPI +png_get_error_ptr(png_const_structrp png_ptr) +{ + if (png_ptr == NULL) + return NULL; + + return ((png_voidp)png_ptr->error_ptr); +} + + +#ifdef PNG_ERROR_NUMBERS_SUPPORTED +void PNGAPI +png_set_strip_error_numbers(png_structrp png_ptr, png_uint_32 strip_mode) +{ + if (png_ptr != NULL) + { + png_ptr->flags &= + ((~(PNG_FLAG_STRIP_ERROR_NUMBERS | + PNG_FLAG_STRIP_ERROR_TEXT))&strip_mode); + } +} +#endif + +#if defined(PNG_SIMPLIFIED_READ_SUPPORTED) ||\ + defined(PNG_SIMPLIFIED_WRITE_SUPPORTED) + /* Currently the above both depend on SETJMP_SUPPORTED, however it would be + * possible to implement without setjmp support just so long as there is some + * way to handle the error return here: + */ +PNG_FUNCTION(void /* PRIVATE */, (PNGCBAPI +png_safe_error),(png_structp png_nonconst_ptr, png_const_charp error_message), + PNG_NORETURN) +{ + const png_const_structrp png_ptr = png_nonconst_ptr; + png_imagep image = png_voidcast(png_imagep, png_ptr->error_ptr); + + /* An error is always logged here, overwriting anything (typically a warning) + * that is already there: + */ + if (image != NULL) + { + png_safecat(image->message, (sizeof image->message), 0, error_message); + image->warning_or_error |= PNG_IMAGE_ERROR; + + /* Retrieve the jmp_buf from within the png_control, making this work for + * C++ compilation too is pretty tricky: C++ wants a pointer to the first + * element of a jmp_buf, but C doesn't tell us the type of that. + */ + if (image->opaque != NULL && image->opaque->error_buf != NULL) + longjmp(png_control_jmp_buf(image->opaque), 1); + + /* Missing longjmp buffer, the following is to help debugging: */ + { + size_t pos = png_safecat(image->message, (sizeof image->message), 0, + "bad longjmp: "); + png_safecat(image->message, (sizeof image->message), pos, + error_message); + } + } + + /* Here on an internal programming error. */ + abort(); +} + +#ifdef PNG_WARNINGS_SUPPORTED +void /* PRIVATE */ PNGCBAPI +png_safe_warning(png_structp png_nonconst_ptr, png_const_charp warning_message) +{ + const png_const_structrp png_ptr = png_nonconst_ptr; + png_imagep image = png_voidcast(png_imagep, png_ptr->error_ptr); + + /* A warning is only logged if there is no prior warning or error. */ + if (image->warning_or_error == 0) + { + png_safecat(image->message, (sizeof image->message), 0, warning_message); + image->warning_or_error |= PNG_IMAGE_WARNING; + } +} +#endif + +int /* PRIVATE */ +png_safe_execute(png_imagep image_in, int (*function)(png_voidp), png_voidp arg) +{ + volatile png_imagep image = image_in; + volatile int result; + volatile png_voidp saved_error_buf; + jmp_buf safe_jmpbuf; + + /* Safely execute function(arg) with png_error returning to this function. */ + saved_error_buf = image->opaque->error_buf; + result = setjmp(safe_jmpbuf) == 0; + + if (result != 0) + { + + image->opaque->error_buf = safe_jmpbuf; + result = function(arg); + } + + image->opaque->error_buf = saved_error_buf; + + /* And do the cleanup prior to any failure return. */ + if (result == 0) + png_image_free(image); + + return result; +} +#endif /* SIMPLIFIED READ || SIMPLIFIED_WRITE */ +#endif /* READ || WRITE */ diff --git a/app/src/main/jni/libpng/pngget.c b/app/src/main/jni/libpng/pngget.c new file mode 100644 index 000000000..fce126cd0 --- /dev/null +++ b/app/src/main/jni/libpng/pngget.c @@ -0,0 +1,1213 @@ + +/* pngget.c - retrieval of values from info struct + * + * Last changed in libpng 1.6.15 [November 20, 2014] + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + * + */ + +#include "pngpriv.h" + +#if defined(PNG_READ_SUPPORTED) || defined(PNG_WRITE_SUPPORTED) + +png_uint_32 PNGAPI +png_get_valid(png_const_structrp png_ptr, png_const_inforp info_ptr, + png_uint_32 flag) +{ + if (png_ptr != NULL && info_ptr != NULL) + return(info_ptr->valid & flag); + + return(0); +} + +png_size_t PNGAPI +png_get_rowbytes(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + if (png_ptr != NULL && info_ptr != NULL) + return(info_ptr->rowbytes); + + return(0); +} + +#ifdef PNG_INFO_IMAGE_SUPPORTED +png_bytepp PNGAPI +png_get_rows(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + if (png_ptr != NULL && info_ptr != NULL) + return(info_ptr->row_pointers); + + return(0); +} +#endif + +#ifdef PNG_EASY_ACCESS_SUPPORTED +/* Easy access to info, added in libpng-0.99 */ +png_uint_32 PNGAPI +png_get_image_width(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + if (png_ptr != NULL && info_ptr != NULL) + return info_ptr->width; + + return (0); +} + +png_uint_32 PNGAPI +png_get_image_height(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + if (png_ptr != NULL && info_ptr != NULL) + return info_ptr->height; + + return (0); +} + +png_byte PNGAPI +png_get_bit_depth(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + if (png_ptr != NULL && info_ptr != NULL) + return info_ptr->bit_depth; + + return (0); +} + +png_byte PNGAPI +png_get_color_type(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + if (png_ptr != NULL && info_ptr != NULL) + return info_ptr->color_type; + + return (0); +} + +png_byte PNGAPI +png_get_filter_type(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + if (png_ptr != NULL && info_ptr != NULL) + return info_ptr->filter_type; + + return (0); +} + +png_byte PNGAPI +png_get_interlace_type(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + if (png_ptr != NULL && info_ptr != NULL) + return info_ptr->interlace_type; + + return (0); +} + +png_byte PNGAPI +png_get_compression_type(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + if (png_ptr != NULL && info_ptr != NULL) + return info_ptr->compression_type; + + return (0); +} + +png_uint_32 PNGAPI +png_get_x_pixels_per_meter(png_const_structrp png_ptr, png_const_inforp + info_ptr) +{ +#ifdef PNG_pHYs_SUPPORTED + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_pHYs) != 0) + { + png_debug1(1, "in %s retrieval function", + "png_get_x_pixels_per_meter"); + + if (info_ptr->phys_unit_type == PNG_RESOLUTION_METER) + return (info_ptr->x_pixels_per_unit); + } +#else + PNG_UNUSED(png_ptr) + PNG_UNUSED(info_ptr) +#endif + + return (0); +} + +png_uint_32 PNGAPI +png_get_y_pixels_per_meter(png_const_structrp png_ptr, png_const_inforp + info_ptr) +{ +#ifdef PNG_pHYs_SUPPORTED + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_pHYs) != 0) + { + png_debug1(1, "in %s retrieval function", + "png_get_y_pixels_per_meter"); + + if (info_ptr->phys_unit_type == PNG_RESOLUTION_METER) + return (info_ptr->y_pixels_per_unit); + } +#else + PNG_UNUSED(png_ptr) + PNG_UNUSED(info_ptr) +#endif + + return (0); +} + +png_uint_32 PNGAPI +png_get_pixels_per_meter(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ +#ifdef PNG_pHYs_SUPPORTED + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_pHYs) != 0) + { + png_debug1(1, "in %s retrieval function", "png_get_pixels_per_meter"); + + if (info_ptr->phys_unit_type == PNG_RESOLUTION_METER && + info_ptr->x_pixels_per_unit == info_ptr->y_pixels_per_unit) + return (info_ptr->x_pixels_per_unit); + } +#else + PNG_UNUSED(png_ptr) + PNG_UNUSED(info_ptr) +#endif + + return (0); +} + +#ifdef PNG_FLOATING_POINT_SUPPORTED +float PNGAPI +png_get_pixel_aspect_ratio(png_const_structrp png_ptr, png_const_inforp + info_ptr) +{ +#ifdef PNG_READ_pHYs_SUPPORTED + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_pHYs) != 0) + { + png_debug1(1, "in %s retrieval function", "png_get_aspect_ratio"); + + if (info_ptr->x_pixels_per_unit != 0) + return ((float)((float)info_ptr->y_pixels_per_unit + /(float)info_ptr->x_pixels_per_unit)); + } +#else + PNG_UNUSED(png_ptr) + PNG_UNUSED(info_ptr) +#endif + + return ((float)0.0); +} +#endif + +#ifdef PNG_FIXED_POINT_SUPPORTED +png_fixed_point PNGAPI +png_get_pixel_aspect_ratio_fixed(png_const_structrp png_ptr, + png_const_inforp info_ptr) +{ +#ifdef PNG_READ_pHYs_SUPPORTED + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_pHYs) != 0 && + info_ptr->x_pixels_per_unit > 0 && info_ptr->y_pixels_per_unit > 0 && + info_ptr->x_pixels_per_unit <= PNG_UINT_31_MAX && + info_ptr->y_pixels_per_unit <= PNG_UINT_31_MAX) + { + png_fixed_point res; + + png_debug1(1, "in %s retrieval function", "png_get_aspect_ratio_fixed"); + + /* The following casts work because a PNG 4 byte integer only has a valid + * range of 0..2^31-1; otherwise the cast might overflow. + */ + if (png_muldiv(&res, (png_int_32)info_ptr->y_pixels_per_unit, PNG_FP_1, + (png_int_32)info_ptr->x_pixels_per_unit) != 0) + return res; + } +#else + PNG_UNUSED(png_ptr) + PNG_UNUSED(info_ptr) +#endif + + return 0; +} +#endif + +png_int_32 PNGAPI +png_get_x_offset_microns(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ +#ifdef PNG_oFFs_SUPPORTED + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_oFFs) != 0) + { + png_debug1(1, "in %s retrieval function", "png_get_x_offset_microns"); + + if (info_ptr->offset_unit_type == PNG_OFFSET_MICROMETER) + return (info_ptr->x_offset); + } +#else + PNG_UNUSED(png_ptr) + PNG_UNUSED(info_ptr) +#endif + + return (0); +} + +png_int_32 PNGAPI +png_get_y_offset_microns(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ +#ifdef PNG_oFFs_SUPPORTED + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_oFFs) != 0) + { + png_debug1(1, "in %s retrieval function", "png_get_y_offset_microns"); + + if (info_ptr->offset_unit_type == PNG_OFFSET_MICROMETER) + return (info_ptr->y_offset); + } +#else + PNG_UNUSED(png_ptr) + PNG_UNUSED(info_ptr) +#endif + + return (0); +} + +png_int_32 PNGAPI +png_get_x_offset_pixels(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ +#ifdef PNG_oFFs_SUPPORTED + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_oFFs) != 0) + { + png_debug1(1, "in %s retrieval function", "png_get_x_offset_pixels"); + + if (info_ptr->offset_unit_type == PNG_OFFSET_PIXEL) + return (info_ptr->x_offset); + } +#else + PNG_UNUSED(png_ptr) + PNG_UNUSED(info_ptr) +#endif + + return (0); +} + +png_int_32 PNGAPI +png_get_y_offset_pixels(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ +#ifdef PNG_oFFs_SUPPORTED + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_oFFs) != 0) + { + png_debug1(1, "in %s retrieval function", "png_get_y_offset_pixels"); + + if (info_ptr->offset_unit_type == PNG_OFFSET_PIXEL) + return (info_ptr->y_offset); + } +#else + PNG_UNUSED(png_ptr) + PNG_UNUSED(info_ptr) +#endif + + return (0); +} + +#ifdef PNG_INCH_CONVERSIONS_SUPPORTED +static png_uint_32 +ppi_from_ppm(png_uint_32 ppm) +{ +#if 0 + /* The conversion is *(2.54/100), in binary (32 digits): + * .00000110100000001001110101001001 + */ + png_uint_32 t1001, t1101; + ppm >>= 1; /* .1 */ + t1001 = ppm + (ppm >> 3); /* .1001 */ + t1101 = t1001 + (ppm >> 1); /* .1101 */ + ppm >>= 20; /* .000000000000000000001 */ + t1101 += t1101 >> 15; /* .1101000000000001101 */ + t1001 >>= 11; /* .000000000001001 */ + t1001 += t1001 >> 12; /* .000000000001001000000001001 */ + ppm += t1001; /* .000000000001001000001001001 */ + ppm += t1101; /* .110100000001001110101001001 */ + return (ppm + 16) >> 5;/* .00000110100000001001110101001001 */ +#else + /* The argument is a PNG unsigned integer, so it is not permitted + * to be bigger than 2^31. + */ + png_fixed_point result; + if (ppm <= PNG_UINT_31_MAX && png_muldiv(&result, (png_int_32)ppm, 127, + 5000) != 0) + return result; + + /* Overflow. */ + return 0; +#endif +} + +png_uint_32 PNGAPI +png_get_pixels_per_inch(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + return ppi_from_ppm(png_get_pixels_per_meter(png_ptr, info_ptr)); +} + +png_uint_32 PNGAPI +png_get_x_pixels_per_inch(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + return ppi_from_ppm(png_get_x_pixels_per_meter(png_ptr, info_ptr)); +} + +png_uint_32 PNGAPI +png_get_y_pixels_per_inch(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + return ppi_from_ppm(png_get_y_pixels_per_meter(png_ptr, info_ptr)); +} + +#ifdef PNG_FIXED_POINT_SUPPORTED +static png_fixed_point +png_fixed_inches_from_microns(png_const_structrp png_ptr, png_int_32 microns) +{ + /* Convert from metres * 1,000,000 to inches * 100,000, meters to + * inches is simply *(100/2.54), so we want *(10/2.54) == 500/127. + * Notice that this can overflow - a warning is output and 0 is + * returned. + */ + return png_muldiv_warn(png_ptr, microns, 500, 127); +} + +png_fixed_point PNGAPI +png_get_x_offset_inches_fixed(png_const_structrp png_ptr, + png_const_inforp info_ptr) +{ + return png_fixed_inches_from_microns(png_ptr, + png_get_x_offset_microns(png_ptr, info_ptr)); +} +#endif + +#ifdef PNG_FIXED_POINT_SUPPORTED +png_fixed_point PNGAPI +png_get_y_offset_inches_fixed(png_const_structrp png_ptr, + png_const_inforp info_ptr) +{ + return png_fixed_inches_from_microns(png_ptr, + png_get_y_offset_microns(png_ptr, info_ptr)); +} +#endif + +#ifdef PNG_FLOATING_POINT_SUPPORTED +float PNGAPI +png_get_x_offset_inches(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + /* To avoid the overflow do the conversion directly in floating + * point. + */ + return (float)(png_get_x_offset_microns(png_ptr, info_ptr) * .00003937); +} +#endif + +#ifdef PNG_FLOATING_POINT_SUPPORTED +float PNGAPI +png_get_y_offset_inches(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + /* To avoid the overflow do the conversion directly in floating + * point. + */ + return (float)(png_get_y_offset_microns(png_ptr, info_ptr) * .00003937); +} +#endif + +#ifdef PNG_pHYs_SUPPORTED +png_uint_32 PNGAPI +png_get_pHYs_dpi(png_const_structrp png_ptr, png_const_inforp info_ptr, + png_uint_32 *res_x, png_uint_32 *res_y, int *unit_type) +{ + png_uint_32 retval = 0; + + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_pHYs) != 0) + { + png_debug1(1, "in %s retrieval function", "pHYs"); + + if (res_x != NULL) + { + *res_x = info_ptr->x_pixels_per_unit; + retval |= PNG_INFO_pHYs; + } + + if (res_y != NULL) + { + *res_y = info_ptr->y_pixels_per_unit; + retval |= PNG_INFO_pHYs; + } + + if (unit_type != NULL) + { + *unit_type = (int)info_ptr->phys_unit_type; + retval |= PNG_INFO_pHYs; + + if (*unit_type == 1) + { + if (res_x != NULL) *res_x = (png_uint_32)(*res_x * .0254 + .50); + if (res_y != NULL) *res_y = (png_uint_32)(*res_y * .0254 + .50); + } + } + } + + return (retval); +} +#endif /* pHYs */ +#endif /* INCH_CONVERSIONS */ + +/* png_get_channels really belongs in here, too, but it's been around longer */ + +#endif /* EASY_ACCESS */ + + +png_byte PNGAPI +png_get_channels(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + if (png_ptr != NULL && info_ptr != NULL) + return(info_ptr->channels); + + return (0); +} + +#ifdef PNG_READ_SUPPORTED +png_const_bytep PNGAPI +png_get_signature(png_const_structrp png_ptr, png_const_inforp info_ptr) +{ + if (png_ptr != NULL && info_ptr != NULL) + return(info_ptr->signature); + + return (NULL); +} +#endif + +#ifdef PNG_bKGD_SUPPORTED +png_uint_32 PNGAPI +png_get_bKGD(png_const_structrp png_ptr, png_inforp info_ptr, + png_color_16p *background) +{ + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_bKGD) != 0 && + background != NULL) + { + png_debug1(1, "in %s retrieval function", "bKGD"); + + *background = &(info_ptr->background); + return (PNG_INFO_bKGD); + } + + return (0); +} +#endif + +#ifdef PNG_cHRM_SUPPORTED +/* The XYZ APIs were added in 1.5.5 to take advantage of the code added at the + * same time to correct the rgb grayscale coefficient defaults obtained from the + * cHRM chunk in 1.5.4 + */ +# ifdef PNG_FLOATING_POINT_SUPPORTED +png_uint_32 PNGAPI +png_get_cHRM(png_const_structrp png_ptr, png_const_inforp info_ptr, + double *white_x, double *white_y, double *red_x, double *red_y, + double *green_x, double *green_y, double *blue_x, double *blue_y) +{ + /* Quiet API change: this code used to only return the end points if a cHRM + * chunk was present, but the end points can also come from iCCP or sRGB + * chunks, so in 1.6.0 the png_get_ APIs return the end points regardless and + * the png_set_ APIs merely check that set end points are mutually + * consistent. + */ + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->colorspace.flags & PNG_COLORSPACE_HAVE_ENDPOINTS) != 0) + { + png_debug1(1, "in %s retrieval function", "cHRM"); + + if (white_x != NULL) + *white_x = png_float(png_ptr, + info_ptr->colorspace.end_points_xy.whitex, "cHRM white X"); + if (white_y != NULL) + *white_y = png_float(png_ptr, + info_ptr->colorspace.end_points_xy.whitey, "cHRM white Y"); + if (red_x != NULL) + *red_x = png_float(png_ptr, info_ptr->colorspace.end_points_xy.redx, + "cHRM red X"); + if (red_y != NULL) + *red_y = png_float(png_ptr, info_ptr->colorspace.end_points_xy.redy, + "cHRM red Y"); + if (green_x != NULL) + *green_x = png_float(png_ptr, + info_ptr->colorspace.end_points_xy.greenx, "cHRM green X"); + if (green_y != NULL) + *green_y = png_float(png_ptr, + info_ptr->colorspace.end_points_xy.greeny, "cHRM green Y"); + if (blue_x != NULL) + *blue_x = png_float(png_ptr, info_ptr->colorspace.end_points_xy.bluex, + "cHRM blue X"); + if (blue_y != NULL) + *blue_y = png_float(png_ptr, info_ptr->colorspace.end_points_xy.bluey, + "cHRM blue Y"); + return (PNG_INFO_cHRM); + } + + return (0); +} + +png_uint_32 PNGAPI +png_get_cHRM_XYZ(png_const_structrp png_ptr, png_const_inforp info_ptr, + double *red_X, double *red_Y, double *red_Z, double *green_X, + double *green_Y, double *green_Z, double *blue_X, double *blue_Y, + double *blue_Z) +{ + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->colorspace.flags & PNG_COLORSPACE_HAVE_ENDPOINTS) != 0) + { + png_debug1(1, "in %s retrieval function", "cHRM_XYZ(float)"); + + if (red_X != NULL) + *red_X = png_float(png_ptr, info_ptr->colorspace.end_points_XYZ.red_X, + "cHRM red X"); + if (red_Y != NULL) + *red_Y = png_float(png_ptr, info_ptr->colorspace.end_points_XYZ.red_Y, + "cHRM red Y"); + if (red_Z != NULL) + *red_Z = png_float(png_ptr, info_ptr->colorspace.end_points_XYZ.red_Z, + "cHRM red Z"); + if (green_X != NULL) + *green_X = png_float(png_ptr, + info_ptr->colorspace.end_points_XYZ.green_X, "cHRM green X"); + if (green_Y != NULL) + *green_Y = png_float(png_ptr, + info_ptr->colorspace.end_points_XYZ.green_Y, "cHRM green Y"); + if (green_Z != NULL) + *green_Z = png_float(png_ptr, + info_ptr->colorspace.end_points_XYZ.green_Z, "cHRM green Z"); + if (blue_X != NULL) + *blue_X = png_float(png_ptr, + info_ptr->colorspace.end_points_XYZ.blue_X, "cHRM blue X"); + if (blue_Y != NULL) + *blue_Y = png_float(png_ptr, + info_ptr->colorspace.end_points_XYZ.blue_Y, "cHRM blue Y"); + if (blue_Z != NULL) + *blue_Z = png_float(png_ptr, + info_ptr->colorspace.end_points_XYZ.blue_Z, "cHRM blue Z"); + return (PNG_INFO_cHRM); + } + + return (0); +} +# endif + +# ifdef PNG_FIXED_POINT_SUPPORTED +png_uint_32 PNGAPI +png_get_cHRM_XYZ_fixed(png_const_structrp png_ptr, png_const_inforp info_ptr, + png_fixed_point *int_red_X, png_fixed_point *int_red_Y, + png_fixed_point *int_red_Z, png_fixed_point *int_green_X, + png_fixed_point *int_green_Y, png_fixed_point *int_green_Z, + png_fixed_point *int_blue_X, png_fixed_point *int_blue_Y, + png_fixed_point *int_blue_Z) +{ + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->colorspace.flags & PNG_COLORSPACE_HAVE_ENDPOINTS) != 0) + { + png_debug1(1, "in %s retrieval function", "cHRM_XYZ"); + + if (int_red_X != NULL) + *int_red_X = info_ptr->colorspace.end_points_XYZ.red_X; + if (int_red_Y != NULL) + *int_red_Y = info_ptr->colorspace.end_points_XYZ.red_Y; + if (int_red_Z != NULL) + *int_red_Z = info_ptr->colorspace.end_points_XYZ.red_Z; + if (int_green_X != NULL) + *int_green_X = info_ptr->colorspace.end_points_XYZ.green_X; + if (int_green_Y != NULL) + *int_green_Y = info_ptr->colorspace.end_points_XYZ.green_Y; + if (int_green_Z != NULL) + *int_green_Z = info_ptr->colorspace.end_points_XYZ.green_Z; + if (int_blue_X != NULL) + *int_blue_X = info_ptr->colorspace.end_points_XYZ.blue_X; + if (int_blue_Y != NULL) + *int_blue_Y = info_ptr->colorspace.end_points_XYZ.blue_Y; + if (int_blue_Z != NULL) + *int_blue_Z = info_ptr->colorspace.end_points_XYZ.blue_Z; + return (PNG_INFO_cHRM); + } + + return (0); +} + +png_uint_32 PNGAPI +png_get_cHRM_fixed(png_const_structrp png_ptr, png_const_inforp info_ptr, + png_fixed_point *white_x, png_fixed_point *white_y, png_fixed_point *red_x, + png_fixed_point *red_y, png_fixed_point *green_x, png_fixed_point *green_y, + png_fixed_point *blue_x, png_fixed_point *blue_y) +{ + png_debug1(1, "in %s retrieval function", "cHRM"); + + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->colorspace.flags & PNG_COLORSPACE_HAVE_ENDPOINTS) != 0) + { + if (white_x != NULL) + *white_x = info_ptr->colorspace.end_points_xy.whitex; + if (white_y != NULL) + *white_y = info_ptr->colorspace.end_points_xy.whitey; + if (red_x != NULL) + *red_x = info_ptr->colorspace.end_points_xy.redx; + if (red_y != NULL) + *red_y = info_ptr->colorspace.end_points_xy.redy; + if (green_x != NULL) + *green_x = info_ptr->colorspace.end_points_xy.greenx; + if (green_y != NULL) + *green_y = info_ptr->colorspace.end_points_xy.greeny; + if (blue_x != NULL) + *blue_x = info_ptr->colorspace.end_points_xy.bluex; + if (blue_y != NULL) + *blue_y = info_ptr->colorspace.end_points_xy.bluey; + return (PNG_INFO_cHRM); + } + + return (0); +} +# endif +#endif + +#ifdef PNG_gAMA_SUPPORTED +# ifdef PNG_FIXED_POINT_SUPPORTED +png_uint_32 PNGAPI +png_get_gAMA_fixed(png_const_structrp png_ptr, png_const_inforp info_ptr, + png_fixed_point *file_gamma) +{ + png_debug1(1, "in %s retrieval function", "gAMA"); + + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->colorspace.flags & PNG_COLORSPACE_HAVE_GAMMA) != 0 && + file_gamma != NULL) + { + *file_gamma = info_ptr->colorspace.gamma; + return (PNG_INFO_gAMA); + } + + return (0); +} +# endif + +# ifdef PNG_FLOATING_POINT_SUPPORTED +png_uint_32 PNGAPI +png_get_gAMA(png_const_structrp png_ptr, png_const_inforp info_ptr, + double *file_gamma) +{ + png_debug1(1, "in %s retrieval function", "gAMA(float)"); + + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->colorspace.flags & PNG_COLORSPACE_HAVE_GAMMA) != 0 && + file_gamma != NULL) + { + *file_gamma = png_float(png_ptr, info_ptr->colorspace.gamma, + "png_get_gAMA"); + return (PNG_INFO_gAMA); + } + + return (0); +} +# endif +#endif + +#ifdef PNG_sRGB_SUPPORTED +png_uint_32 PNGAPI +png_get_sRGB(png_const_structrp png_ptr, png_const_inforp info_ptr, + int *file_srgb_intent) +{ + png_debug1(1, "in %s retrieval function", "sRGB"); + + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_sRGB) != 0 && file_srgb_intent != NULL) + { + *file_srgb_intent = info_ptr->colorspace.rendering_intent; + return (PNG_INFO_sRGB); + } + + return (0); +} +#endif + +#ifdef PNG_iCCP_SUPPORTED +png_uint_32 PNGAPI +png_get_iCCP(png_const_structrp png_ptr, png_inforp info_ptr, + png_charpp name, int *compression_type, + png_bytepp profile, png_uint_32 *proflen) +{ + png_debug1(1, "in %s retrieval function", "iCCP"); + + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_iCCP) != 0 && + name != NULL && compression_type != NULL && profile != NULL && + proflen != NULL) + { + *name = info_ptr->iccp_name; + *profile = info_ptr->iccp_profile; + *proflen = png_get_uint_32(info_ptr->iccp_profile); + /* This is somewhat irrelevant since the profile data returned has + * actually been uncompressed. + */ + *compression_type = PNG_COMPRESSION_TYPE_BASE; + return (PNG_INFO_iCCP); + } + + return (0); +} +#endif + +#ifdef PNG_sPLT_SUPPORTED +int PNGAPI +png_get_sPLT(png_const_structrp png_ptr, png_inforp info_ptr, + png_sPLT_tpp spalettes) +{ + if (png_ptr != NULL && info_ptr != NULL && spalettes != NULL) + { + *spalettes = info_ptr->splt_palettes; + return info_ptr->splt_palettes_num; + } + + return (0); +} +#endif + +#ifdef PNG_hIST_SUPPORTED +png_uint_32 PNGAPI +png_get_hIST(png_const_structrp png_ptr, png_inforp info_ptr, + png_uint_16p *hist) +{ + png_debug1(1, "in %s retrieval function", "hIST"); + + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_hIST) != 0 && hist != NULL) + { + *hist = info_ptr->hist; + return (PNG_INFO_hIST); + } + + return (0); +} +#endif + +png_uint_32 PNGAPI +png_get_IHDR(png_const_structrp png_ptr, png_const_inforp info_ptr, + png_uint_32 *width, png_uint_32 *height, int *bit_depth, + int *color_type, int *interlace_type, int *compression_type, + int *filter_type) +{ + png_debug1(1, "in %s retrieval function", "IHDR"); + + if (png_ptr == NULL || info_ptr == NULL || width == NULL || + height == NULL || bit_depth == NULL || color_type == NULL) + return (0); + + *width = info_ptr->width; + *height = info_ptr->height; + *bit_depth = info_ptr->bit_depth; + *color_type = info_ptr->color_type; + + if (compression_type != NULL) + *compression_type = info_ptr->compression_type; + + if (filter_type != NULL) + *filter_type = info_ptr->filter_type; + + if (interlace_type != NULL) + *interlace_type = info_ptr->interlace_type; + + /* This is redundant if we can be sure that the info_ptr values were all + * assigned in png_set_IHDR(). We do the check anyhow in case an + * application has ignored our advice not to mess with the members + * of info_ptr directly. + */ + png_check_IHDR(png_ptr, info_ptr->width, info_ptr->height, + info_ptr->bit_depth, info_ptr->color_type, info_ptr->interlace_type, + info_ptr->compression_type, info_ptr->filter_type); + + return (1); +} + +#ifdef PNG_oFFs_SUPPORTED +png_uint_32 PNGAPI +png_get_oFFs(png_const_structrp png_ptr, png_const_inforp info_ptr, + png_int_32 *offset_x, png_int_32 *offset_y, int *unit_type) +{ + png_debug1(1, "in %s retrieval function", "oFFs"); + + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_oFFs) != 0 && + offset_x != NULL && offset_y != NULL && unit_type != NULL) + { + *offset_x = info_ptr->x_offset; + *offset_y = info_ptr->y_offset; + *unit_type = (int)info_ptr->offset_unit_type; + return (PNG_INFO_oFFs); + } + + return (0); +} +#endif + +#ifdef PNG_pCAL_SUPPORTED +png_uint_32 PNGAPI +png_get_pCAL(png_const_structrp png_ptr, png_inforp info_ptr, + png_charp *purpose, png_int_32 *X0, png_int_32 *X1, int *type, int *nparams, + png_charp *units, png_charpp *params) +{ + png_debug1(1, "in %s retrieval function", "pCAL"); + + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_pCAL) != 0 && + purpose != NULL && X0 != NULL && X1 != NULL && type != NULL && + nparams != NULL && units != NULL && params != NULL) + { + *purpose = info_ptr->pcal_purpose; + *X0 = info_ptr->pcal_X0; + *X1 = info_ptr->pcal_X1; + *type = (int)info_ptr->pcal_type; + *nparams = (int)info_ptr->pcal_nparams; + *units = info_ptr->pcal_units; + *params = info_ptr->pcal_params; + return (PNG_INFO_pCAL); + } + + return (0); +} +#endif + +#ifdef PNG_sCAL_SUPPORTED +# ifdef PNG_FIXED_POINT_SUPPORTED +# if defined(PNG_FLOATING_ARITHMETIC_SUPPORTED) || \ + defined(PNG_FLOATING_POINT_SUPPORTED) +png_uint_32 PNGAPI +png_get_sCAL_fixed(png_const_structrp png_ptr, png_const_inforp info_ptr, + int *unit, png_fixed_point *width, png_fixed_point *height) +{ + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_sCAL) != 0) + { + *unit = info_ptr->scal_unit; + /*TODO: make this work without FP support; the API is currently eliminated + * if neither floating point APIs nor internal floating point arithmetic + * are enabled. + */ + *width = png_fixed(png_ptr, atof(info_ptr->scal_s_width), "sCAL width"); + *height = png_fixed(png_ptr, atof(info_ptr->scal_s_height), + "sCAL height"); + return (PNG_INFO_sCAL); + } + + return(0); +} +# endif /* FLOATING_ARITHMETIC */ +# endif /* FIXED_POINT */ +# ifdef PNG_FLOATING_POINT_SUPPORTED +png_uint_32 PNGAPI +png_get_sCAL(png_const_structrp png_ptr, png_const_inforp info_ptr, + int *unit, double *width, double *height) +{ + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_sCAL) != 0) + { + *unit = info_ptr->scal_unit; + *width = atof(info_ptr->scal_s_width); + *height = atof(info_ptr->scal_s_height); + return (PNG_INFO_sCAL); + } + + return(0); +} +# endif /* FLOATING POINT */ +png_uint_32 PNGAPI +png_get_sCAL_s(png_const_structrp png_ptr, png_const_inforp info_ptr, + int *unit, png_charpp width, png_charpp height) +{ + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_sCAL) != 0) + { + *unit = info_ptr->scal_unit; + *width = info_ptr->scal_s_width; + *height = info_ptr->scal_s_height; + return (PNG_INFO_sCAL); + } + + return(0); +} +#endif /* sCAL */ + +#ifdef PNG_pHYs_SUPPORTED +png_uint_32 PNGAPI +png_get_pHYs(png_const_structrp png_ptr, png_const_inforp info_ptr, + png_uint_32 *res_x, png_uint_32 *res_y, int *unit_type) +{ + png_uint_32 retval = 0; + + png_debug1(1, "in %s retrieval function", "pHYs"); + + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_pHYs) != 0) + { + if (res_x != NULL) + { + *res_x = info_ptr->x_pixels_per_unit; + retval |= PNG_INFO_pHYs; + } + + if (res_y != NULL) + { + *res_y = info_ptr->y_pixels_per_unit; + retval |= PNG_INFO_pHYs; + } + + if (unit_type != NULL) + { + *unit_type = (int)info_ptr->phys_unit_type; + retval |= PNG_INFO_pHYs; + } + } + + return (retval); +} +#endif /* pHYs */ + +png_uint_32 PNGAPI +png_get_PLTE(png_const_structrp png_ptr, png_inforp info_ptr, + png_colorp *palette, int *num_palette) +{ + png_debug1(1, "in %s retrieval function", "PLTE"); + + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_PLTE) != 0 && palette != NULL) + { + *palette = info_ptr->palette; + *num_palette = info_ptr->num_palette; + png_debug1(3, "num_palette = %d", *num_palette); + return (PNG_INFO_PLTE); + } + + return (0); +} + +#ifdef PNG_sBIT_SUPPORTED +png_uint_32 PNGAPI +png_get_sBIT(png_const_structrp png_ptr, png_inforp info_ptr, + png_color_8p *sig_bit) +{ + png_debug1(1, "in %s retrieval function", "sBIT"); + + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_sBIT) != 0 && sig_bit != NULL) + { + *sig_bit = &(info_ptr->sig_bit); + return (PNG_INFO_sBIT); + } + + return (0); +} +#endif + +#ifdef PNG_TEXT_SUPPORTED +int PNGAPI +png_get_text(png_const_structrp png_ptr, png_inforp info_ptr, + png_textp *text_ptr, int *num_text) +{ + if (png_ptr != NULL && info_ptr != NULL && info_ptr->num_text > 0) + { + png_debug1(1, "in 0x%lx retrieval function", + (unsigned long)png_ptr->chunk_name); + + if (text_ptr != NULL) + *text_ptr = info_ptr->text; + + if (num_text != NULL) + *num_text = info_ptr->num_text; + + return info_ptr->num_text; + } + + if (num_text != NULL) + *num_text = 0; + + return(0); +} +#endif + +#ifdef PNG_tIME_SUPPORTED +png_uint_32 PNGAPI +png_get_tIME(png_const_structrp png_ptr, png_inforp info_ptr, + png_timep *mod_time) +{ + png_debug1(1, "in %s retrieval function", "tIME"); + + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_tIME) != 0 && mod_time != NULL) + { + *mod_time = &(info_ptr->mod_time); + return (PNG_INFO_tIME); + } + + return (0); +} +#endif + +#ifdef PNG_tRNS_SUPPORTED +png_uint_32 PNGAPI +png_get_tRNS(png_const_structrp png_ptr, png_inforp info_ptr, + png_bytep *trans_alpha, int *num_trans, png_color_16p *trans_color) +{ + png_uint_32 retval = 0; + if (png_ptr != NULL && info_ptr != NULL && + (info_ptr->valid & PNG_INFO_tRNS) != 0) + { + png_debug1(1, "in %s retrieval function", "tRNS"); + + if (info_ptr->color_type == PNG_COLOR_TYPE_PALETTE) + { + if (trans_alpha != NULL) + { + *trans_alpha = info_ptr->trans_alpha; + retval |= PNG_INFO_tRNS; + } + + if (trans_color != NULL) + *trans_color = &(info_ptr->trans_color); + } + + else /* if (info_ptr->color_type != PNG_COLOR_TYPE_PALETTE) */ + { + if (trans_color != NULL) + { + *trans_color = &(info_ptr->trans_color); + retval |= PNG_INFO_tRNS; + } + + if (trans_alpha != NULL) + *trans_alpha = NULL; + } + + if (num_trans != NULL) + { + *num_trans = info_ptr->num_trans; + retval |= PNG_INFO_tRNS; + } + } + + return (retval); +} +#endif + +#ifdef PNG_STORE_UNKNOWN_CHUNKS_SUPPORTED +int PNGAPI +png_get_unknown_chunks(png_const_structrp png_ptr, png_inforp info_ptr, + png_unknown_chunkpp unknowns) +{ + if (png_ptr != NULL && info_ptr != NULL && unknowns != NULL) + { + *unknowns = info_ptr->unknown_chunks; + return info_ptr->unknown_chunks_num; + } + + return (0); +} +#endif + +#ifdef PNG_READ_RGB_TO_GRAY_SUPPORTED +png_byte PNGAPI +png_get_rgb_to_gray_status (png_const_structrp png_ptr) +{ + return (png_byte)(png_ptr ? png_ptr->rgb_to_gray_status : 0); +} +#endif + +#ifdef PNG_USER_CHUNKS_SUPPORTED +png_voidp PNGAPI +png_get_user_chunk_ptr(png_const_structrp png_ptr) +{ + return (png_ptr ? png_ptr->user_chunk_ptr : NULL); +} +#endif + +png_size_t PNGAPI +png_get_compression_buffer_size(png_const_structrp png_ptr) +{ + if (png_ptr == NULL) + return 0; + +# ifdef PNG_WRITE_SUPPORTED + if ((png_ptr->mode & PNG_IS_READ_STRUCT) != 0) +# endif + { +# ifdef PNG_SEQUENTIAL_READ_SUPPORTED + return png_ptr->IDAT_read_size; +# else + return PNG_IDAT_READ_SIZE; +# endif + } + +# ifdef PNG_WRITE_SUPPORTED + else + return png_ptr->zbuffer_size; +# endif +} + +#ifdef PNG_SET_USER_LIMITS_SUPPORTED +/* These functions were added to libpng 1.2.6 and were enabled + * by default in libpng-1.4.0 */ +png_uint_32 PNGAPI +png_get_user_width_max (png_const_structrp png_ptr) +{ + return (png_ptr ? png_ptr->user_width_max : 0); +} + +png_uint_32 PNGAPI +png_get_user_height_max (png_const_structrp png_ptr) +{ + return (png_ptr ? png_ptr->user_height_max : 0); +} + +/* This function was added to libpng 1.4.0 */ +png_uint_32 PNGAPI +png_get_chunk_cache_max (png_const_structrp png_ptr) +{ + return (png_ptr ? png_ptr->user_chunk_cache_max : 0); +} + +/* This function was added to libpng 1.4.1 */ +png_alloc_size_t PNGAPI +png_get_chunk_malloc_max (png_const_structrp png_ptr) +{ + return (png_ptr ? png_ptr->user_chunk_malloc_max : 0); +} +#endif /* SET_USER_LIMITS */ + +/* These functions were added to libpng 1.4.0 */ +#ifdef PNG_IO_STATE_SUPPORTED +png_uint_32 PNGAPI +png_get_io_state (png_const_structrp png_ptr) +{ + return png_ptr->io_state; +} + +png_uint_32 PNGAPI +png_get_io_chunk_type (png_const_structrp png_ptr) +{ + return png_ptr->chunk_name; +} +#endif /* IO_STATE */ + +#ifdef PNG_CHECK_FOR_INVALID_INDEX_SUPPORTED +# ifdef PNG_GET_PALETTE_MAX_SUPPORTED +int PNGAPI +png_get_palette_max(png_const_structp png_ptr, png_const_infop info_ptr) +{ + if (png_ptr != NULL && info_ptr != NULL) + return png_ptr->num_palette_max; + + return (-1); +} +# endif +#endif + +#endif /* READ || WRITE */ diff --git a/app/src/main/jni/libpng/pnginfo.h b/app/src/main/jni/libpng/pnginfo.h new file mode 100644 index 000000000..122a90a41 --- /dev/null +++ b/app/src/main/jni/libpng/pnginfo.h @@ -0,0 +1,260 @@ + +/* pnginfo.h - header file for PNG reference library + * + * Copyright (c) 1998-2013 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * Last changed in libpng 1.6.1 [March 28, 2013] + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + */ + + /* png_info is a structure that holds the information in a PNG file so + * that the application can find out the characteristics of the image. + * If you are reading the file, this structure will tell you what is + * in the PNG file. If you are writing the file, fill in the information + * you want to put into the PNG file, using png_set_*() functions, then + * call png_write_info(). + * + * The names chosen should be very close to the PNG specification, so + * consult that document for information about the meaning of each field. + * + * With libpng < 0.95, it was only possible to directly set and read the + * the values in the png_info_struct, which meant that the contents and + * order of the values had to remain fixed. With libpng 0.95 and later, + * however, there are now functions that abstract the contents of + * png_info_struct from the application, so this makes it easier to use + * libpng with dynamic libraries, and even makes it possible to use + * libraries that don't have all of the libpng ancillary chunk-handing + * functionality. In libpng-1.5.0 this was moved into a separate private + * file that is not visible to applications. + * + * The following members may have allocated storage attached that should be + * cleaned up before the structure is discarded: palette, trans, text, + * pcal_purpose, pcal_units, pcal_params, hist, iccp_name, iccp_profile, + * splt_palettes, scal_unit, row_pointers, and unknowns. By default, these + * are automatically freed when the info structure is deallocated, if they were + * allocated internally by libpng. This behavior can be changed by means + * of the png_data_freer() function. + * + * More allocation details: all the chunk-reading functions that + * change these members go through the corresponding png_set_* + * functions. A function to clear these members is available: see + * png_free_data(). The png_set_* functions do not depend on being + * able to point info structure members to any of the storage they are + * passed (they make their own copies), EXCEPT that the png_set_text + * functions use the same storage passed to them in the text_ptr or + * itxt_ptr structure argument, and the png_set_rows and png_set_unknowns + * functions do not make their own copies. + */ +#ifndef PNGINFO_H +#define PNGINFO_H + +struct png_info_def +{ + /* The following are necessary for every PNG file */ + png_uint_32 width; /* width of image in pixels (from IHDR) */ + png_uint_32 height; /* height of image in pixels (from IHDR) */ + png_uint_32 valid; /* valid chunk data (see PNG_INFO_ below) */ + png_size_t rowbytes; /* bytes needed to hold an untransformed row */ + png_colorp palette; /* array of color values (valid & PNG_INFO_PLTE) */ + png_uint_16 num_palette; /* number of color entries in "palette" (PLTE) */ + png_uint_16 num_trans; /* number of transparent palette color (tRNS) */ + png_byte bit_depth; /* 1, 2, 4, 8, or 16 bits/channel (from IHDR) */ + png_byte color_type; /* see PNG_COLOR_TYPE_ below (from IHDR) */ + /* The following three should have been named *_method not *_type */ + png_byte compression_type; /* must be PNG_COMPRESSION_TYPE_BASE (IHDR) */ + png_byte filter_type; /* must be PNG_FILTER_TYPE_BASE (from IHDR) */ + png_byte interlace_type; /* One of PNG_INTERLACE_NONE, PNG_INTERLACE_ADAM7 */ + + /* The following are set by png_set_IHDR, called from the application on + * write, but the are never actually used by the write code. + */ + png_byte channels; /* number of data channels per pixel (1, 2, 3, 4) */ + png_byte pixel_depth; /* number of bits per pixel */ + png_byte spare_byte; /* to align the data, and for future use */ + +#ifdef PNG_READ_SUPPORTED + /* This is never set during write */ + png_byte signature[8]; /* magic bytes read by libpng from start of file */ +#endif + + /* The rest of the data is optional. If you are reading, check the + * valid field to see if the information in these are valid. If you + * are writing, set the valid field to those chunks you want written, + * and initialize the appropriate fields below. + */ + +#if defined(PNG_COLORSPACE_SUPPORTED) || defined(PNG_GAMMA_SUPPORTED) + /* png_colorspace only contains 'flags' if neither GAMMA or COLORSPACE are + * defined. When COLORSPACE is switched on all the colorspace-defining + * chunks should be enabled, when GAMMA is switched on all the gamma-defining + * chunks should be enabled. If this is not done it becomes possible to read + * inconsistent PNG files and assign a probably incorrect interpretation to + * the information. (In other words, by carefully choosing which chunks to + * recognize the system configuration can select an interpretation for PNG + * files containing ambiguous data and this will result in inconsistent + * behavior between different libpng builds!) + */ + png_colorspace colorspace; +#endif + +#ifdef PNG_iCCP_SUPPORTED + /* iCCP chunk data. */ + png_charp iccp_name; /* profile name */ + png_bytep iccp_profile; /* International Color Consortium profile data */ + png_uint_32 iccp_proflen; /* ICC profile data length */ +#endif + +#ifdef PNG_TEXT_SUPPORTED + /* The tEXt, and zTXt chunks contain human-readable textual data in + * uncompressed, compressed, and optionally compressed forms, respectively. + * The data in "text" is an array of pointers to uncompressed, + * null-terminated C strings. Each chunk has a keyword that describes the + * textual data contained in that chunk. Keywords are not required to be + * unique, and the text string may be empty. Any number of text chunks may + * be in an image. + */ + int num_text; /* number of comments read or comments to write */ + int max_text; /* current size of text array */ + png_textp text; /* array of comments read or comments to write */ +#endif /* TEXT */ + +#ifdef PNG_tIME_SUPPORTED + /* The tIME chunk holds the last time the displayed image data was + * modified. See the png_time struct for the contents of this struct. + */ + png_time mod_time; +#endif + +#ifdef PNG_sBIT_SUPPORTED + /* The sBIT chunk specifies the number of significant high-order bits + * in the pixel data. Values are in the range [1, bit_depth], and are + * only specified for the channels in the pixel data. The contents of + * the low-order bits is not specified. Data is valid if + * (valid & PNG_INFO_sBIT) is non-zero. + */ + png_color_8 sig_bit; /* significant bits in color channels */ +#endif + +#if defined(PNG_tRNS_SUPPORTED) || defined(PNG_READ_EXPAND_SUPPORTED) || \ +defined(PNG_READ_BACKGROUND_SUPPORTED) + /* The tRNS chunk supplies transparency data for paletted images and + * other image types that don't need a full alpha channel. There are + * "num_trans" transparency values for a paletted image, stored in the + * same order as the palette colors, starting from index 0. Values + * for the data are in the range [0, 255], ranging from fully transparent + * to fully opaque, respectively. For non-paletted images, there is a + * single color specified that should be treated as fully transparent. + * Data is valid if (valid & PNG_INFO_tRNS) is non-zero. + */ + png_bytep trans_alpha; /* alpha values for paletted image */ + png_color_16 trans_color; /* transparent color for non-palette image */ +#endif + +#if defined(PNG_bKGD_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED) + /* The bKGD chunk gives the suggested image background color if the + * display program does not have its own background color and the image + * is needs to composited onto a background before display. The colors + * in "background" are normally in the same color space/depth as the + * pixel data. Data is valid if (valid & PNG_INFO_bKGD) is non-zero. + */ + png_color_16 background; +#endif + +#ifdef PNG_oFFs_SUPPORTED + /* The oFFs chunk gives the offset in "offset_unit_type" units rightwards + * and downwards from the top-left corner of the display, page, or other + * application-specific co-ordinate space. See the PNG_OFFSET_ defines + * below for the unit types. Valid if (valid & PNG_INFO_oFFs) non-zero. + */ + png_int_32 x_offset; /* x offset on page */ + png_int_32 y_offset; /* y offset on page */ + png_byte offset_unit_type; /* offset units type */ +#endif + +#ifdef PNG_pHYs_SUPPORTED + /* The pHYs chunk gives the physical pixel density of the image for + * display or printing in "phys_unit_type" units (see PNG_RESOLUTION_ + * defines below). Data is valid if (valid & PNG_INFO_pHYs) is non-zero. + */ + png_uint_32 x_pixels_per_unit; /* horizontal pixel density */ + png_uint_32 y_pixels_per_unit; /* vertical pixel density */ + png_byte phys_unit_type; /* resolution type (see PNG_RESOLUTION_ below) */ +#endif + +#ifdef PNG_hIST_SUPPORTED + /* The hIST chunk contains the relative frequency or importance of the + * various palette entries, so that a viewer can intelligently select a + * reduced-color palette, if required. Data is an array of "num_palette" + * values in the range [0,65535]. Data valid if (valid & PNG_INFO_hIST) + * is non-zero. + */ + png_uint_16p hist; +#endif + +#ifdef PNG_pCAL_SUPPORTED + /* The pCAL chunk describes a transformation between the stored pixel + * values and original physical data values used to create the image. + * The integer range [0, 2^bit_depth - 1] maps to the floating-point + * range given by [pcal_X0, pcal_X1], and are further transformed by a + * (possibly non-linear) transformation function given by "pcal_type" + * and "pcal_params" into "pcal_units". Please see the PNG_EQUATION_ + * defines below, and the PNG-Group's PNG extensions document for a + * complete description of the transformations and how they should be + * implemented, and for a description of the ASCII parameter strings. + * Data values are valid if (valid & PNG_INFO_pCAL) non-zero. + */ + png_charp pcal_purpose; /* pCAL chunk description string */ + png_int_32 pcal_X0; /* minimum value */ + png_int_32 pcal_X1; /* maximum value */ + png_charp pcal_units; /* Latin-1 string giving physical units */ + png_charpp pcal_params; /* ASCII strings containing parameter values */ + png_byte pcal_type; /* equation type (see PNG_EQUATION_ below) */ + png_byte pcal_nparams; /* number of parameters given in pcal_params */ +#endif + +/* New members added in libpng-1.0.6 */ + png_uint_32 free_me; /* flags items libpng is responsible for freeing */ + +#ifdef PNG_STORE_UNKNOWN_CHUNKS_SUPPORTED + /* Storage for unknown chunks that the library doesn't recognize. */ + png_unknown_chunkp unknown_chunks; + + /* The type of this field is limited by the type of + * png_struct::user_chunk_cache_max, else overflow can occur. + */ + int unknown_chunks_num; +#endif + +#ifdef PNG_sPLT_SUPPORTED + /* Data on sPLT chunks (there may be more than one). */ + png_sPLT_tp splt_palettes; + int splt_palettes_num; /* Match type returned by png_get API */ +#endif + +#ifdef PNG_sCAL_SUPPORTED + /* The sCAL chunk describes the actual physical dimensions of the + * subject matter of the graphic. The chunk contains a unit specification + * a byte value, and two ASCII strings representing floating-point + * values. The values are width and height corresponsing to one pixel + * in the image. Data values are valid if (valid & PNG_INFO_sCAL) is + * non-zero. + */ + png_byte scal_unit; /* unit of physical scale */ + png_charp scal_s_width; /* string containing height */ + png_charp scal_s_height; /* string containing width */ +#endif + +#ifdef PNG_INFO_IMAGE_SUPPORTED + /* Memory has been allocated if (valid & PNG_ALLOCATED_INFO_ROWS) + non-zero */ + /* Data valid if (valid & PNG_INFO_IDAT) non-zero */ + png_bytepp row_pointers; /* the image bits */ +#endif + +}; +#endif /* PNGINFO_H */ diff --git a/app/src/main/jni/libpng/pnglibconf.h b/app/src/main/jni/libpng/pnglibconf.h new file mode 100644 index 000000000..621a1890f --- /dev/null +++ b/app/src/main/jni/libpng/pnglibconf.h @@ -0,0 +1,210 @@ +/* pnglibconf.h - library build configuration */ + +/* libpng version 1.6.16,December 22, 2014 */ + +/* Copyright (c) 1998-2014 Glenn Randers-Pehrson */ + +/* This code is released under the libpng license. */ +/* For conditions of distribution and use, see the disclaimer */ +/* and license in png.h */ + +/* pnglibconf.h */ +/* Machine generated file: DO NOT EDIT */ +/* Derived from: scripts/pnglibconf.dfa */ +#ifndef PNGLCONF_H +#define PNGLCONF_H +/* options */ +#define PNG_16BIT_SUPPORTED +#define PNG_ALIGNED_MEMORY_SUPPORTED +#define PNG_ARM_NEON_API_SUPPORTED +#define PNG_ARM_NEON_CHECK_SUPPORTED +#define PNG_BENIGN_ERRORS_SUPPORTED +#define PNG_BENIGN_READ_ERRORS_SUPPORTED +/*#undef PNG_BENIGN_WRITE_ERRORS_SUPPORTED*/ +#define PNG_BUILD_GRAYSCALE_PALETTE_SUPPORTED +#define PNG_CHECK_FOR_INVALID_INDEX_SUPPORTED +#define PNG_COLORSPACE_SUPPORTED +#define PNG_CONSOLE_IO_SUPPORTED +#define PNG_CONVERT_tIME_SUPPORTED +#define PNG_EASY_ACCESS_SUPPORTED +/*#undef PNG_ERROR_NUMBERS_SUPPORTED*/ +#define PNG_ERROR_TEXT_SUPPORTED +#define PNG_FIXED_POINT_SUPPORTED +#define PNG_FLOATING_ARITHMETIC_SUPPORTED +#define PNG_FLOATING_POINT_SUPPORTED +#define PNG_FORMAT_AFIRST_SUPPORTED +#define PNG_FORMAT_BGR_SUPPORTED +#define PNG_GAMMA_SUPPORTED +#define PNG_GET_PALETTE_MAX_SUPPORTED +#define PNG_HANDLE_AS_UNKNOWN_SUPPORTED +#define PNG_INCH_CONVERSIONS_SUPPORTED +#define PNG_INFO_IMAGE_SUPPORTED +#define PNG_IO_STATE_SUPPORTED +#define PNG_MNG_FEATURES_SUPPORTED +#define PNG_POINTER_INDEXING_SUPPORTED +#define PNG_PROGRESSIVE_READ_SUPPORTED +#define PNG_READ_16BIT_SUPPORTED +#define PNG_READ_ALPHA_MODE_SUPPORTED +#define PNG_READ_ANCILLARY_CHUNKS_SUPPORTED +#define PNG_READ_BACKGROUND_SUPPORTED +#define PNG_READ_BGR_SUPPORTED +#define PNG_READ_CHECK_FOR_INVALID_INDEX_SUPPORTED +#define PNG_READ_COMPOSITE_NODIV_SUPPORTED +#define PNG_READ_COMPRESSED_TEXT_SUPPORTED +#define PNG_READ_EXPAND_16_SUPPORTED +#define PNG_READ_EXPAND_SUPPORTED +#define PNG_READ_FILLER_SUPPORTED +#define PNG_READ_GAMMA_SUPPORTED +#define PNG_READ_GET_PALETTE_MAX_SUPPORTED +#define PNG_READ_GRAY_TO_RGB_SUPPORTED +#define PNG_READ_INTERLACING_SUPPORTED +#define PNG_READ_INT_FUNCTIONS_SUPPORTED +#define PNG_READ_INVERT_ALPHA_SUPPORTED +#define PNG_READ_INVERT_SUPPORTED +#define PNG_READ_OPT_PLTE_SUPPORTED +#define PNG_READ_PACKSWAP_SUPPORTED +#define PNG_READ_PACK_SUPPORTED +#define PNG_READ_QUANTIZE_SUPPORTED +#define PNG_READ_RGB_TO_GRAY_SUPPORTED +#define PNG_READ_SCALE_16_TO_8_SUPPORTED +#define PNG_READ_SHIFT_SUPPORTED +#define PNG_READ_STRIP_16_TO_8_SUPPORTED +#define PNG_READ_STRIP_ALPHA_SUPPORTED +#define PNG_READ_SUPPORTED +#define PNG_READ_SWAP_ALPHA_SUPPORTED +#define PNG_READ_SWAP_SUPPORTED +#define PNG_READ_TEXT_SUPPORTED +#define PNG_READ_TRANSFORMS_SUPPORTED +#define PNG_READ_UNKNOWN_CHUNKS_SUPPORTED +#define PNG_READ_USER_CHUNKS_SUPPORTED +#define PNG_READ_USER_TRANSFORM_SUPPORTED +#define PNG_READ_bKGD_SUPPORTED +#define PNG_READ_cHRM_SUPPORTED +#define PNG_READ_gAMA_SUPPORTED +#define PNG_READ_hIST_SUPPORTED +#define PNG_READ_iCCP_SUPPORTED +#define PNG_READ_iTXt_SUPPORTED +#define PNG_READ_oFFs_SUPPORTED +#define PNG_READ_pCAL_SUPPORTED +#define PNG_READ_pHYs_SUPPORTED +#define PNG_READ_sBIT_SUPPORTED +#define PNG_READ_sCAL_SUPPORTED +#define PNG_READ_sPLT_SUPPORTED +#define PNG_READ_sRGB_SUPPORTED +#define PNG_READ_tEXt_SUPPORTED +#define PNG_READ_tIME_SUPPORTED +#define PNG_READ_tRNS_SUPPORTED +#define PNG_READ_zTXt_SUPPORTED +/*#undef PNG_SAFE_LIMITS_SUPPORTED*/ +#define PNG_SAVE_INT_32_SUPPORTED +#define PNG_SAVE_UNKNOWN_CHUNKS_SUPPORTED +#define PNG_SEQUENTIAL_READ_SUPPORTED +#define PNG_SETJMP_SUPPORTED +#define PNG_SET_CHUNK_CACHE_LIMIT_SUPPORTED +#define PNG_SET_CHUNK_MALLOC_LIMIT_SUPPORTED +#define PNG_SET_OPTION_SUPPORTED +#define PNG_SET_UNKNOWN_CHUNKS_SUPPORTED +#define PNG_SET_USER_LIMITS_SUPPORTED +#define PNG_SIMPLIFIED_READ_AFIRST_SUPPORTED +#define PNG_SIMPLIFIED_READ_BGR_SUPPORTED +#define PNG_SIMPLIFIED_READ_SUPPORTED +/*#undef PNG_SIMPLIFIED_WRITE_AFIRST_SUPPORTED*/ +/*#undef PNG_SIMPLIFIED_WRITE_BGR_SUPPORTED*/ +/*#undef PNG_SIMPLIFIED_WRITE_SUPPORTED*/ +#define PNG_STDIO_SUPPORTED +#define PNG_STORE_UNKNOWN_CHUNKS_SUPPORTED +#define PNG_TEXT_SUPPORTED +#define PNG_TIME_RFC1123_SUPPORTED +#define PNG_UNKNOWN_CHUNKS_SUPPORTED +#define PNG_USER_CHUNKS_SUPPORTED +#define PNG_USER_LIMITS_SUPPORTED +#define PNG_USER_MEM_SUPPORTED +#define PNG_USER_TRANSFORM_INFO_SUPPORTED +#define PNG_USER_TRANSFORM_PTR_SUPPORTED +#define PNG_WARNINGS_SUPPORTED +/* +#define PNG_WRITE_16BIT_SUPPORTED +#define PNG_WRITE_ANCILLARY_CHUNKS_SUPPORTED +#define PNG_WRITE_BGR_SUPPORTED +#define PNG_WRITE_CHECK_FOR_INVALID_INDEX_SUPPORTED +#define PNG_WRITE_COMPRESSED_TEXT_SUPPORTED +#define PNG_WRITE_CUSTOMIZE_ZTXT_COMPRESSION_SUPPORTED +#define PNG_WRITE_FILLER_SUPPORTED +#define PNG_WRITE_FILTER_SUPPORTED +#define PNG_WRITE_FLUSH_SUPPORTED +#define PNG_WRITE_GET_PALETTE_MAX_SUPPORTED +#define PNG_WRITE_INTERLACING_SUPPORTED +#define PNG_WRITE_INT_FUNCTIONS_SUPPORTED +#define PNG_WRITE_INVERT_ALPHA_SUPPORTED +#define PNG_WRITE_INVERT_SUPPORTED +#define PNG_WRITE_OPTIMIZE_CMF_SUPPORTED +#define PNG_WRITE_PACKSWAP_SUPPORTED +#define PNG_WRITE_PACK_SUPPORTED +#define PNG_WRITE_SHIFT_SUPPORTED +#define PNG_WRITE_SUPPORTED +#define PNG_WRITE_SWAP_ALPHA_SUPPORTED +#define PNG_WRITE_SWAP_SUPPORTED +#define PNG_WRITE_TEXT_SUPPORTED +#define PNG_WRITE_TRANSFORMS_SUPPORTED +#define PNG_WRITE_UNKNOWN_CHUNKS_SUPPORTED +#define PNG_WRITE_USER_TRANSFORM_SUPPORTED +#define PNG_WRITE_WEIGHTED_FILTER_SUPPORTED +#define PNG_WRITE_bKGD_SUPPORTED +#define PNG_WRITE_cHRM_SUPPORTED +#define PNG_WRITE_gAMA_SUPPORTED +#define PNG_WRITE_hIST_SUPPORTED +#define PNG_WRITE_iCCP_SUPPORTED +#define PNG_WRITE_iTXt_SUPPORTED +#define PNG_WRITE_oFFs_SUPPORTED +#define PNG_WRITE_pCAL_SUPPORTED +#define PNG_WRITE_pHYs_SUPPORTED +#define PNG_WRITE_sBIT_SUPPORTED +#define PNG_WRITE_sCAL_SUPPORTED +#define PNG_WRITE_sPLT_SUPPORTED +#define PNG_WRITE_sRGB_SUPPORTED +#define PNG_WRITE_tEXt_SUPPORTED +#define PNG_WRITE_tIME_SUPPORTED +#define PNG_WRITE_tRNS_SUPPORTED +#define PNG_WRITE_zTXt_SUPPORTED +*/ +#define PNG_bKGD_SUPPORTED +#define PNG_cHRM_SUPPORTED +#define PNG_gAMA_SUPPORTED +#define PNG_hIST_SUPPORTED +#define PNG_iCCP_SUPPORTED +#define PNG_iTXt_SUPPORTED +#define PNG_oFFs_SUPPORTED +#define PNG_pCAL_SUPPORTED +#define PNG_pHYs_SUPPORTED +#define PNG_sBIT_SUPPORTED +#define PNG_sCAL_SUPPORTED +#define PNG_sPLT_SUPPORTED +#define PNG_sRGB_SUPPORTED +#define PNG_tEXt_SUPPORTED +#define PNG_tIME_SUPPORTED +#define PNG_tRNS_SUPPORTED +#define PNG_zTXt_SUPPORTED +/* end of options */ +/* settings */ +#define PNG_API_RULE 0 +#define PNG_COST_SHIFT 3 +#define PNG_DEFAULT_READ_MACROS 1 +#define PNG_GAMMA_THRESHOLD_FIXED 5000 +#define PNG_IDAT_READ_SIZE PNG_ZBUF_SIZE +#define PNG_INFLATE_BUF_SIZE 1024 +#define PNG_MAX_GAMMA_8 11 +#define PNG_QUANTIZE_BLUE_BITS 5 +#define PNG_QUANTIZE_GREEN_BITS 5 +#define PNG_QUANTIZE_RED_BITS 5 +#define PNG_TEXT_Z_DEFAULT_COMPRESSION (-1) +#define PNG_TEXT_Z_DEFAULT_STRATEGY 0 +#define PNG_WEIGHT_SHIFT 8 +#define PNG_ZBUF_SIZE 8192 +#define PNG_ZLIB_VERNUM 0x1280 +#define PNG_Z_DEFAULT_COMPRESSION (-1) +#define PNG_Z_DEFAULT_NOFILTER_STRATEGY 0 +#define PNG_Z_DEFAULT_STRATEGY 1 +#define PNG_sCAL_PRECISION 5 +#define PNG_sRGB_PROFILE_CHECKS 2 +/* end of settings */ +#endif /* PNGLCONF_H */ diff --git a/app/src/main/jni/libpng/pngmem.c b/app/src/main/jni/libpng/pngmem.c new file mode 100644 index 000000000..d6caf273f --- /dev/null +++ b/app/src/main/jni/libpng/pngmem.c @@ -0,0 +1,281 @@ + +/* pngmem.c - stub functions for memory allocation + * + * Last changed in libpng 1.6.15 [November 20, 2014] + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + * + * This file provides a location for all memory allocation. Users who + * need special memory handling are expected to supply replacement + * functions for png_malloc() and png_free(), and to use + * png_create_read_struct_2() and png_create_write_struct_2() to + * identify the replacement functions. + */ + +#include "pngpriv.h" + +#if defined(PNG_READ_SUPPORTED) || defined(PNG_WRITE_SUPPORTED) +/* Free a png_struct */ +void /* PRIVATE */ +png_destroy_png_struct(png_structrp png_ptr) +{ + if (png_ptr != NULL) + { + /* png_free might call png_error and may certainly call + * png_get_mem_ptr, so fake a temporary png_struct to support this. + */ + png_struct dummy_struct = *png_ptr; + memset(png_ptr, 0, (sizeof *png_ptr)); + png_free(&dummy_struct, png_ptr); + +# ifdef PNG_SETJMP_SUPPORTED + /* We may have a jmp_buf left to deallocate. */ + png_free_jmpbuf(&dummy_struct); +# endif + } +} + +/* Allocate memory. For reasonable files, size should never exceed + * 64K. However, zlib may allocate more then 64K if you don't tell + * it not to. See zconf.h and png.h for more information. zlib does + * need to allocate exactly 64K, so whatever you call here must + * have the ability to do that. + */ +PNG_FUNCTION(png_voidp,PNGAPI +png_calloc,(png_const_structrp png_ptr, png_alloc_size_t size),PNG_ALLOCATED) +{ + png_voidp ret; + + ret = png_malloc(png_ptr, size); + + if (ret != NULL) + memset(ret, 0, size); + + return ret; +} + +/* png_malloc_base, an internal function added at libpng 1.6.0, does the work of + * allocating memory, taking into account limits and PNG_USER_MEM_SUPPORTED. + * Checking and error handling must happen outside this routine; it returns NULL + * if the allocation cannot be done (for any reason.) + */ +PNG_FUNCTION(png_voidp /* PRIVATE */, +png_malloc_base,(png_const_structrp png_ptr, png_alloc_size_t size), + PNG_ALLOCATED) +{ + /* Moved to png_malloc_base from png_malloc_default in 1.6.0; the DOS + * allocators have also been removed in 1.6.0, so any 16-bit system now has + * to implement a user memory handler. This checks to be sure it isn't + * called with big numbers. + */ +#ifndef PNG_USER_MEM_SUPPORTED + PNG_UNUSED(png_ptr) +#endif + + if (size > 0 && size <= PNG_SIZE_MAX +# ifdef PNG_MAX_MALLOC_64K + && size <= 65536U +# endif + ) + { +#ifdef PNG_USER_MEM_SUPPORTED + if (png_ptr != NULL && png_ptr->malloc_fn != NULL) + return png_ptr->malloc_fn(png_constcast(png_structrp,png_ptr), size); + + else +#endif + return malloc((size_t)size); /* checked for truncation above */ + } + + else + return NULL; +} + +#if defined(PNG_TEXT_SUPPORTED) || defined(PNG_sPLT_SUPPORTED) ||\ + defined(PNG_STORE_UNKNOWN_CHUNKS_SUPPORTED) +/* This is really here only to work round a spurious warning in GCC 4.6 and 4.7 + * that arises because of the checks in png_realloc_array that are repeated in + * png_malloc_array. + */ +static png_voidp +png_malloc_array_checked(png_const_structrp png_ptr, int nelements, + size_t element_size) +{ + png_alloc_size_t req = nelements; /* known to be > 0 */ + + if (req <= PNG_SIZE_MAX/element_size) + return png_malloc_base(png_ptr, req * element_size); + + /* The failure case when the request is too large */ + return NULL; +} + +PNG_FUNCTION(png_voidp /* PRIVATE */, +png_malloc_array,(png_const_structrp png_ptr, int nelements, + size_t element_size),PNG_ALLOCATED) +{ + if (nelements <= 0 || element_size == 0) + png_error(png_ptr, "internal error: array alloc"); + + return png_malloc_array_checked(png_ptr, nelements, element_size); +} + +PNG_FUNCTION(png_voidp /* PRIVATE */, +png_realloc_array,(png_const_structrp png_ptr, png_const_voidp old_array, + int old_elements, int add_elements, size_t element_size),PNG_ALLOCATED) +{ + /* These are internal errors: */ + if (add_elements <= 0 || element_size == 0 || old_elements < 0 || + (old_array == NULL && old_elements > 0)) + png_error(png_ptr, "internal error: array realloc"); + + /* Check for overflow on the elements count (so the caller does not have to + * check.) + */ + if (add_elements <= INT_MAX - old_elements) + { + png_voidp new_array = png_malloc_array_checked(png_ptr, + old_elements+add_elements, element_size); + + if (new_array != NULL) + { + /* Because png_malloc_array worked the size calculations below cannot + * overflow. + */ + if (old_elements > 0) + memcpy(new_array, old_array, element_size*(unsigned)old_elements); + + memset((char*)new_array + element_size*(unsigned)old_elements, 0, + element_size*(unsigned)add_elements); + + return new_array; + } + } + + return NULL; /* error */ +} +#endif /* TEXT || sPLT || STORE_UNKNOWN_CHUNKS */ + +/* Various functions that have different error handling are derived from this. + * png_malloc always exists, but if PNG_USER_MEM_SUPPORTED is defined a separate + * function png_malloc_default is also provided. + */ +PNG_FUNCTION(png_voidp,PNGAPI +png_malloc,(png_const_structrp png_ptr, png_alloc_size_t size),PNG_ALLOCATED) +{ + png_voidp ret; + + if (png_ptr == NULL) + return NULL; + + ret = png_malloc_base(png_ptr, size); + + if (ret == NULL) + png_error(png_ptr, "Out of memory"); /* 'm' means png_malloc */ + + return ret; +} + +#ifdef PNG_USER_MEM_SUPPORTED +PNG_FUNCTION(png_voidp,PNGAPI +png_malloc_default,(png_const_structrp png_ptr, png_alloc_size_t size), + PNG_ALLOCATED PNG_DEPRECATED) +{ + png_voidp ret; + + if (png_ptr == NULL) + return NULL; + + /* Passing 'NULL' here bypasses the application provided memory handler. */ + ret = png_malloc_base(NULL/*use malloc*/, size); + + if (ret == NULL) + png_error(png_ptr, "Out of Memory"); /* 'M' means png_malloc_default */ + + return ret; +} +#endif /* USER_MEM */ + +/* This function was added at libpng version 1.2.3. The png_malloc_warn() + * function will issue a png_warning and return NULL instead of issuing a + * png_error, if it fails to allocate the requested memory. + */ +PNG_FUNCTION(png_voidp,PNGAPI +png_malloc_warn,(png_const_structrp png_ptr, png_alloc_size_t size), + PNG_ALLOCATED) +{ + if (png_ptr != NULL) + { + png_voidp ret = png_malloc_base(png_ptr, size); + + if (ret != NULL) + return ret; + + png_warning(png_ptr, "Out of memory"); + } + + return NULL; +} + +/* Free a pointer allocated by png_malloc(). If ptr is NULL, return + * without taking any action. + */ +void PNGAPI +png_free(png_const_structrp png_ptr, png_voidp ptr) +{ + if (png_ptr == NULL || ptr == NULL) + return; + +#ifdef PNG_USER_MEM_SUPPORTED + if (png_ptr->free_fn != NULL) + png_ptr->free_fn(png_constcast(png_structrp,png_ptr), ptr); + + else + png_free_default(png_ptr, ptr); +} + +PNG_FUNCTION(void,PNGAPI +png_free_default,(png_const_structrp png_ptr, png_voidp ptr),PNG_DEPRECATED) +{ + if (png_ptr == NULL || ptr == NULL) + return; +#endif /* USER_MEM */ + + free(ptr); +} + +#ifdef PNG_USER_MEM_SUPPORTED +/* This function is called when the application wants to use another method + * of allocating and freeing memory. + */ +void PNGAPI +png_set_mem_fn(png_structrp png_ptr, png_voidp mem_ptr, png_malloc_ptr + malloc_fn, png_free_ptr free_fn) +{ + if (png_ptr != NULL) + { + png_ptr->mem_ptr = mem_ptr; + png_ptr->malloc_fn = malloc_fn; + png_ptr->free_fn = free_fn; + } +} + +/* This function returns a pointer to the mem_ptr associated with the user + * functions. The application should free any memory associated with this + * pointer before png_write_destroy and png_read_destroy are called. + */ +png_voidp PNGAPI +png_get_mem_ptr(png_const_structrp png_ptr) +{ + if (png_ptr == NULL) + return NULL; + + return png_ptr->mem_ptr; +} +#endif /* USER_MEM */ +#endif /* READ || WRITE */ diff --git a/app/src/main/jni/libpng/pngpread.c b/app/src/main/jni/libpng/pngpread.c new file mode 100644 index 000000000..fb40e7db3 --- /dev/null +++ b/app/src/main/jni/libpng/pngpread.c @@ -0,0 +1,1168 @@ + +/* pngpread.c - read a png file in push mode + * + * Last changed in libpng 1.6.15 [November 20, 2014] + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + */ + +#include "pngpriv.h" + +#ifdef PNG_PROGRESSIVE_READ_SUPPORTED + +/* Push model modes */ +#define PNG_READ_SIG_MODE 0 +#define PNG_READ_CHUNK_MODE 1 +#define PNG_READ_IDAT_MODE 2 +#define PNG_SKIP_MODE 3 +#define PNG_READ_tEXt_MODE 4 +#define PNG_READ_zTXt_MODE 5 +#define PNG_READ_DONE_MODE 6 +#define PNG_READ_iTXt_MODE 7 +#define PNG_ERROR_MODE 8 + +#define PNG_PUSH_SAVE_BUFFER_IF_FULL \ +if (png_ptr->push_length + 4 > png_ptr->buffer_size) \ + { png_push_save_buffer(png_ptr); return; } +#define PNG_PUSH_SAVE_BUFFER_IF_LT(N) \ +if (png_ptr->buffer_size < N) \ + { png_push_save_buffer(png_ptr); return; } + +void PNGAPI +png_process_data(png_structrp png_ptr, png_inforp info_ptr, + png_bytep buffer, png_size_t buffer_size) +{ + if (png_ptr == NULL || info_ptr == NULL) + return; + + png_push_restore_buffer(png_ptr, buffer, buffer_size); + + while (png_ptr->buffer_size) + { + png_process_some_data(png_ptr, info_ptr); + } +} + +png_size_t PNGAPI +png_process_data_pause(png_structrp png_ptr, int save) +{ + if (png_ptr != NULL) + { + /* It's easiest for the caller if we do the save; then the caller doesn't + * have to supply the same data again: + */ + if (save != 0) + png_push_save_buffer(png_ptr); + else + { + /* This includes any pending saved bytes: */ + png_size_t remaining = png_ptr->buffer_size; + png_ptr->buffer_size = 0; + + /* So subtract the saved buffer size, unless all the data + * is actually 'saved', in which case we just return 0 + */ + if (png_ptr->save_buffer_size < remaining) + return remaining - png_ptr->save_buffer_size; + } + } + + return 0; +} + +png_uint_32 PNGAPI +png_process_data_skip(png_structrp png_ptr) +{ + png_uint_32 remaining = 0; + + if (png_ptr != NULL && png_ptr->process_mode == PNG_SKIP_MODE && + png_ptr->skip_length > 0) + { + /* At the end of png_process_data the buffer size must be 0 (see the loop + * above) so we can detect a broken call here: + */ + if (png_ptr->buffer_size != 0) + png_error(png_ptr, + "png_process_data_skip called inside png_process_data"); + + /* If is impossible for there to be a saved buffer at this point - + * otherwise we could not be in SKIP mode. This will also happen if + * png_process_skip is called inside png_process_data (but only very + * rarely.) + */ + if (png_ptr->save_buffer_size != 0) + png_error(png_ptr, "png_process_data_skip called with saved data"); + + remaining = png_ptr->skip_length; + png_ptr->skip_length = 0; + png_ptr->process_mode = PNG_READ_CHUNK_MODE; + } + + return remaining; +} + +/* What we do with the incoming data depends on what we were previously + * doing before we ran out of data... + */ +void /* PRIVATE */ +png_process_some_data(png_structrp png_ptr, png_inforp info_ptr) +{ + if (png_ptr == NULL) + return; + + switch (png_ptr->process_mode) + { + case PNG_READ_SIG_MODE: + { + png_push_read_sig(png_ptr, info_ptr); + break; + } + + case PNG_READ_CHUNK_MODE: + { + png_push_read_chunk(png_ptr, info_ptr); + break; + } + + case PNG_READ_IDAT_MODE: + { + png_push_read_IDAT(png_ptr); + break; + } + + case PNG_SKIP_MODE: + { + png_push_crc_finish(png_ptr); + break; + } + + default: + { + png_ptr->buffer_size = 0; + break; + } + } +} + +/* Read any remaining signature bytes from the stream and compare them with + * the correct PNG signature. It is possible that this routine is called + * with bytes already read from the signature, either because they have been + * checked by the calling application, or because of multiple calls to this + * routine. + */ +void /* PRIVATE */ +png_push_read_sig(png_structrp png_ptr, png_inforp info_ptr) +{ + png_size_t num_checked = png_ptr->sig_bytes, /* SAFE, does not exceed 8 */ + num_to_check = 8 - num_checked; + + if (png_ptr->buffer_size < num_to_check) + { + num_to_check = png_ptr->buffer_size; + } + + png_push_fill_buffer(png_ptr, &(info_ptr->signature[num_checked]), + num_to_check); + png_ptr->sig_bytes = (png_byte)(png_ptr->sig_bytes + num_to_check); + + if (png_sig_cmp(info_ptr->signature, num_checked, num_to_check)) + { + if (num_checked < 4 && + png_sig_cmp(info_ptr->signature, num_checked, num_to_check - 4)) + png_error(png_ptr, "Not a PNG file"); + + else + png_error(png_ptr, "PNG file corrupted by ASCII conversion"); + } + else + { + if (png_ptr->sig_bytes >= 8) + { + png_ptr->process_mode = PNG_READ_CHUNK_MODE; + } + } +} + +void /* PRIVATE */ +png_push_read_chunk(png_structrp png_ptr, png_inforp info_ptr) +{ + png_uint_32 chunk_name; +#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED + int keep; /* unknown handling method */ +#endif + + /* First we make sure we have enough data for the 4-byte chunk name + * and the 4-byte chunk length before proceeding with decoding the + * chunk data. To fully decode each of these chunks, we also make + * sure we have enough data in the buffer for the 4-byte CRC at the + * end of every chunk (except IDAT, which is handled separately). + */ + if ((png_ptr->mode & PNG_HAVE_CHUNK_HEADER) == 0) + { + png_byte chunk_length[4]; + png_byte chunk_tag[4]; + + PNG_PUSH_SAVE_BUFFER_IF_LT(8) + png_push_fill_buffer(png_ptr, chunk_length, 4); + png_ptr->push_length = png_get_uint_31(png_ptr, chunk_length); + png_reset_crc(png_ptr); + png_crc_read(png_ptr, chunk_tag, 4); + png_ptr->chunk_name = PNG_CHUNK_FROM_STRING(chunk_tag); + png_check_chunk_name(png_ptr, png_ptr->chunk_name); + png_ptr->mode |= PNG_HAVE_CHUNK_HEADER; + } + + chunk_name = png_ptr->chunk_name; + + if (chunk_name == png_IDAT) + { + if ((png_ptr->mode & PNG_AFTER_IDAT) != 0) + png_ptr->mode |= PNG_HAVE_CHUNK_AFTER_IDAT; + + /* If we reach an IDAT chunk, this means we have read all of the + * header chunks, and we can start reading the image (or if this + * is called after the image has been read - we have an error). + */ + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_error(png_ptr, "Missing IHDR before IDAT"); + + else if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE && + (png_ptr->mode & PNG_HAVE_PLTE) == 0) + png_error(png_ptr, "Missing PLTE before IDAT"); + + png_ptr->mode |= PNG_HAVE_IDAT; + png_ptr->process_mode = PNG_READ_IDAT_MODE; + + if ((png_ptr->mode & PNG_HAVE_CHUNK_AFTER_IDAT) == 0) + if (png_ptr->push_length == 0) + return; + + if ((png_ptr->mode & PNG_AFTER_IDAT) != 0) + png_benign_error(png_ptr, "Too many IDATs found"); + } + + if (chunk_name == png_IHDR) + { + if (png_ptr->push_length != 13) + png_error(png_ptr, "Invalid IHDR length"); + + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_IHDR(png_ptr, info_ptr, png_ptr->push_length); + } + + else if (chunk_name == png_IEND) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_IEND(png_ptr, info_ptr, png_ptr->push_length); + + png_ptr->process_mode = PNG_READ_DONE_MODE; + png_push_have_end(png_ptr, info_ptr); + } + +#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED + else if ((keep = png_chunk_unknown_handling(png_ptr, chunk_name)) != 0) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_unknown(png_ptr, info_ptr, png_ptr->push_length, keep); + + if (chunk_name == png_PLTE) + png_ptr->mode |= PNG_HAVE_PLTE; + } +#endif + + else if (chunk_name == png_PLTE) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_PLTE(png_ptr, info_ptr, png_ptr->push_length); + } + + else if (chunk_name == png_IDAT) + { + png_ptr->idat_size = png_ptr->push_length; + png_ptr->process_mode = PNG_READ_IDAT_MODE; + png_push_have_info(png_ptr, info_ptr); + png_ptr->zstream.avail_out = + (uInt) PNG_ROWBYTES(png_ptr->pixel_depth, + png_ptr->iwidth) + 1; + png_ptr->zstream.next_out = png_ptr->row_buf; + return; + } + +#ifdef PNG_READ_gAMA_SUPPORTED + else if (png_ptr->chunk_name == png_gAMA) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_gAMA(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_sBIT_SUPPORTED + else if (png_ptr->chunk_name == png_sBIT) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_sBIT(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_cHRM_SUPPORTED + else if (png_ptr->chunk_name == png_cHRM) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_cHRM(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_sRGB_SUPPORTED + else if (chunk_name == png_sRGB) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_sRGB(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_iCCP_SUPPORTED + else if (png_ptr->chunk_name == png_iCCP) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_iCCP(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_sPLT_SUPPORTED + else if (chunk_name == png_sPLT) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_sPLT(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_tRNS_SUPPORTED + else if (chunk_name == png_tRNS) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_tRNS(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_bKGD_SUPPORTED + else if (chunk_name == png_bKGD) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_bKGD(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_hIST_SUPPORTED + else if (chunk_name == png_hIST) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_hIST(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_pHYs_SUPPORTED + else if (chunk_name == png_pHYs) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_pHYs(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_oFFs_SUPPORTED + else if (chunk_name == png_oFFs) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_oFFs(png_ptr, info_ptr, png_ptr->push_length); + } +#endif + +#ifdef PNG_READ_pCAL_SUPPORTED + else if (chunk_name == png_pCAL) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_pCAL(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_sCAL_SUPPORTED + else if (chunk_name == png_sCAL) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_sCAL(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_tIME_SUPPORTED + else if (chunk_name == png_tIME) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_tIME(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_tEXt_SUPPORTED + else if (chunk_name == png_tEXt) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_tEXt(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_zTXt_SUPPORTED + else if (chunk_name == png_zTXt) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_zTXt(png_ptr, info_ptr, png_ptr->push_length); + } + +#endif +#ifdef PNG_READ_iTXt_SUPPORTED + else if (chunk_name == png_iTXt) + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_iTXt(png_ptr, info_ptr, png_ptr->push_length); + } +#endif + + else + { + PNG_PUSH_SAVE_BUFFER_IF_FULL + png_handle_unknown(png_ptr, info_ptr, png_ptr->push_length, + PNG_HANDLE_CHUNK_AS_DEFAULT); + } + + png_ptr->mode &= ~PNG_HAVE_CHUNK_HEADER; +} + +void /* PRIVATE */ +png_push_crc_skip(png_structrp png_ptr, png_uint_32 skip) +{ + png_ptr->process_mode = PNG_SKIP_MODE; + png_ptr->skip_length = skip; +} + +void /* PRIVATE */ +png_push_crc_finish(png_structrp png_ptr) +{ + if (png_ptr->skip_length != 0 && png_ptr->save_buffer_size != 0) + { + png_size_t save_size = png_ptr->save_buffer_size; + png_uint_32 skip_length = png_ptr->skip_length; + + /* We want the smaller of 'skip_length' and 'save_buffer_size', but + * they are of different types and we don't know which variable has the + * fewest bits. Carefully select the smaller and cast it to the type of + * the larger - this cannot overflow. Do not cast in the following test + * - it will break on either 16 or 64 bit platforms. + */ + if (skip_length < save_size) + save_size = (png_size_t)skip_length; + + else + skip_length = (png_uint_32)save_size; + + png_calculate_crc(png_ptr, png_ptr->save_buffer_ptr, save_size); + + png_ptr->skip_length -= skip_length; + png_ptr->buffer_size -= save_size; + png_ptr->save_buffer_size -= save_size; + png_ptr->save_buffer_ptr += save_size; + } + if (png_ptr->skip_length != 0 && png_ptr->current_buffer_size != 0) + { + png_size_t save_size = png_ptr->current_buffer_size; + png_uint_32 skip_length = png_ptr->skip_length; + + /* We want the smaller of 'skip_length' and 'current_buffer_size', here, + * the same problem exists as above and the same solution. + */ + if (skip_length < save_size) + save_size = (png_size_t)skip_length; + + else + skip_length = (png_uint_32)save_size; + + png_calculate_crc(png_ptr, png_ptr->current_buffer_ptr, save_size); + + png_ptr->skip_length -= skip_length; + png_ptr->buffer_size -= save_size; + png_ptr->current_buffer_size -= save_size; + png_ptr->current_buffer_ptr += save_size; + } + if (png_ptr->skip_length == 0) + { + PNG_PUSH_SAVE_BUFFER_IF_LT(4) + png_crc_finish(png_ptr, 0); + png_ptr->process_mode = PNG_READ_CHUNK_MODE; + } +} + +void PNGCBAPI +png_push_fill_buffer(png_structp png_ptr, png_bytep buffer, png_size_t length) +{ + png_bytep ptr; + + if (png_ptr == NULL) + return; + + ptr = buffer; + if (png_ptr->save_buffer_size != 0) + { + png_size_t save_size; + + if (length < png_ptr->save_buffer_size) + save_size = length; + + else + save_size = png_ptr->save_buffer_size; + + memcpy(ptr, png_ptr->save_buffer_ptr, save_size); + length -= save_size; + ptr += save_size; + png_ptr->buffer_size -= save_size; + png_ptr->save_buffer_size -= save_size; + png_ptr->save_buffer_ptr += save_size; + } + if (length != 0 && png_ptr->current_buffer_size != 0) + { + png_size_t save_size; + + if (length < png_ptr->current_buffer_size) + save_size = length; + + else + save_size = png_ptr->current_buffer_size; + + memcpy(ptr, png_ptr->current_buffer_ptr, save_size); + png_ptr->buffer_size -= save_size; + png_ptr->current_buffer_size -= save_size; + png_ptr->current_buffer_ptr += save_size; + } +} + +void /* PRIVATE */ +png_push_save_buffer(png_structrp png_ptr) +{ + if (png_ptr->save_buffer_size != 0) + { + if (png_ptr->save_buffer_ptr != png_ptr->save_buffer) + { + png_size_t i, istop; + png_bytep sp; + png_bytep dp; + + istop = png_ptr->save_buffer_size; + for (i = 0, sp = png_ptr->save_buffer_ptr, dp = png_ptr->save_buffer; + i < istop; i++, sp++, dp++) + { + *dp = *sp; + } + } + } + if (png_ptr->save_buffer_size + png_ptr->current_buffer_size > + png_ptr->save_buffer_max) + { + png_size_t new_max; + png_bytep old_buffer; + + if (png_ptr->save_buffer_size > PNG_SIZE_MAX - + (png_ptr->current_buffer_size + 256)) + { + png_error(png_ptr, "Potential overflow of save_buffer"); + } + + new_max = png_ptr->save_buffer_size + png_ptr->current_buffer_size + 256; + old_buffer = png_ptr->save_buffer; + png_ptr->save_buffer = (png_bytep)png_malloc_warn(png_ptr, + (png_size_t)new_max); + + if (png_ptr->save_buffer == NULL) + { + png_free(png_ptr, old_buffer); + old_buffer = NULL; + png_error(png_ptr, "Insufficient memory for save_buffer"); + } + + memcpy(png_ptr->save_buffer, old_buffer, png_ptr->save_buffer_size); + png_free(png_ptr, old_buffer); + old_buffer = NULL; + png_ptr->save_buffer_max = new_max; + } + if (png_ptr->current_buffer_size) + { + memcpy(png_ptr->save_buffer + png_ptr->save_buffer_size, + png_ptr->current_buffer_ptr, png_ptr->current_buffer_size); + png_ptr->save_buffer_size += png_ptr->current_buffer_size; + png_ptr->current_buffer_size = 0; + } + png_ptr->save_buffer_ptr = png_ptr->save_buffer; + png_ptr->buffer_size = 0; +} + +void /* PRIVATE */ +png_push_restore_buffer(png_structrp png_ptr, png_bytep buffer, + png_size_t buffer_length) +{ + png_ptr->current_buffer = buffer; + png_ptr->current_buffer_size = buffer_length; + png_ptr->buffer_size = buffer_length + png_ptr->save_buffer_size; + png_ptr->current_buffer_ptr = png_ptr->current_buffer; +} + +void /* PRIVATE */ +png_push_read_IDAT(png_structrp png_ptr) +{ + if ((png_ptr->mode & PNG_HAVE_CHUNK_HEADER) == 0) + { + png_byte chunk_length[4]; + png_byte chunk_tag[4]; + + /* TODO: this code can be commoned up with the same code in push_read */ + PNG_PUSH_SAVE_BUFFER_IF_LT(8) + png_push_fill_buffer(png_ptr, chunk_length, 4); + png_ptr->push_length = png_get_uint_31(png_ptr, chunk_length); + png_reset_crc(png_ptr); + png_crc_read(png_ptr, chunk_tag, 4); + png_ptr->chunk_name = PNG_CHUNK_FROM_STRING(chunk_tag); + png_ptr->mode |= PNG_HAVE_CHUNK_HEADER; + + if (png_ptr->chunk_name != png_IDAT) + { + png_ptr->process_mode = PNG_READ_CHUNK_MODE; + + if ((png_ptr->flags & PNG_FLAG_ZSTREAM_ENDED) == 0) + png_error(png_ptr, "Not enough compressed data"); + + return; + } + + png_ptr->idat_size = png_ptr->push_length; + } + + if (png_ptr->idat_size != 0 && png_ptr->save_buffer_size != 0) + { + png_size_t save_size = png_ptr->save_buffer_size; + png_uint_32 idat_size = png_ptr->idat_size; + + /* We want the smaller of 'idat_size' and 'current_buffer_size', but they + * are of different types and we don't know which variable has the fewest + * bits. Carefully select the smaller and cast it to the type of the + * larger - this cannot overflow. Do not cast in the following test - it + * will break on either 16 or 64 bit platforms. + */ + if (idat_size < save_size) + save_size = (png_size_t)idat_size; + + else + idat_size = (png_uint_32)save_size; + + png_calculate_crc(png_ptr, png_ptr->save_buffer_ptr, save_size); + + png_process_IDAT_data(png_ptr, png_ptr->save_buffer_ptr, save_size); + + png_ptr->idat_size -= idat_size; + png_ptr->buffer_size -= save_size; + png_ptr->save_buffer_size -= save_size; + png_ptr->save_buffer_ptr += save_size; + } + + if (png_ptr->idat_size != 0 && png_ptr->current_buffer_size != 0) + { + png_size_t save_size = png_ptr->current_buffer_size; + png_uint_32 idat_size = png_ptr->idat_size; + + /* We want the smaller of 'idat_size' and 'current_buffer_size', but they + * are of different types and we don't know which variable has the fewest + * bits. Carefully select the smaller and cast it to the type of the + * larger - this cannot overflow. + */ + if (idat_size < save_size) + save_size = (png_size_t)idat_size; + + else + idat_size = (png_uint_32)save_size; + + png_calculate_crc(png_ptr, png_ptr->current_buffer_ptr, save_size); + + png_process_IDAT_data(png_ptr, png_ptr->current_buffer_ptr, save_size); + + png_ptr->idat_size -= idat_size; + png_ptr->buffer_size -= save_size; + png_ptr->current_buffer_size -= save_size; + png_ptr->current_buffer_ptr += save_size; + } + if (png_ptr->idat_size == 0) + { + PNG_PUSH_SAVE_BUFFER_IF_LT(4) + png_crc_finish(png_ptr, 0); + png_ptr->mode &= ~PNG_HAVE_CHUNK_HEADER; + png_ptr->mode |= PNG_AFTER_IDAT; + png_ptr->zowner = 0; + } +} + +void /* PRIVATE */ +png_process_IDAT_data(png_structrp png_ptr, png_bytep buffer, + png_size_t buffer_length) +{ + /* The caller checks for a non-zero buffer length. */ + if (!(buffer_length > 0) || buffer == NULL) + png_error(png_ptr, "No IDAT data (internal error)"); + + /* This routine must process all the data it has been given + * before returning, calling the row callback as required to + * handle the uncompressed results. + */ + png_ptr->zstream.next_in = buffer; + /* TODO: WARNING: TRUNCATION ERROR: DANGER WILL ROBINSON: */ + png_ptr->zstream.avail_in = (uInt)buffer_length; + + /* Keep going until the decompressed data is all processed + * or the stream marked as finished. + */ + while (png_ptr->zstream.avail_in > 0 && + !(png_ptr->flags & PNG_FLAG_ZSTREAM_ENDED)) + { + int ret; + + /* We have data for zlib, but we must check that zlib + * has someplace to put the results. It doesn't matter + * if we don't expect any results -- it may be the input + * data is just the LZ end code. + */ + if (!(png_ptr->zstream.avail_out > 0)) + { + /* TODO: WARNING: TRUNCATION ERROR: DANGER WILL ROBINSON: */ + png_ptr->zstream.avail_out = (uInt)(PNG_ROWBYTES(png_ptr->pixel_depth, + png_ptr->iwidth) + 1); + + png_ptr->zstream.next_out = png_ptr->row_buf; + } + + /* Using Z_SYNC_FLUSH here means that an unterminated + * LZ stream (a stream with a missing end code) can still + * be handled, otherwise (Z_NO_FLUSH) a future zlib + * implementation might defer output and therefore + * change the current behavior (see comments in inflate.c + * for why this doesn't happen at present with zlib 1.2.5). + */ + ret = inflate(&png_ptr->zstream, Z_SYNC_FLUSH); + + /* Check for any failure before proceeding. */ + if (ret != Z_OK && ret != Z_STREAM_END) + { + /* Terminate the decompression. */ + png_ptr->flags |= PNG_FLAG_ZSTREAM_ENDED; + png_ptr->zowner = 0; + + /* This may be a truncated stream (missing or + * damaged end code). Treat that as a warning. + */ + if (png_ptr->row_number >= png_ptr->num_rows || + png_ptr->pass > 6) + png_warning(png_ptr, "Truncated compressed data in IDAT"); + + else + png_error(png_ptr, "Decompression error in IDAT"); + + /* Skip the check on unprocessed input */ + return; + } + + /* Did inflate output any data? */ + if (png_ptr->zstream.next_out != png_ptr->row_buf) + { + /* Is this unexpected data after the last row? + * If it is, artificially terminate the LZ output + * here. + */ + if (png_ptr->row_number >= png_ptr->num_rows || + png_ptr->pass > 6) + { + /* Extra data. */ + png_warning(png_ptr, "Extra compressed data in IDAT"); + png_ptr->flags |= PNG_FLAG_ZSTREAM_ENDED; + png_ptr->zowner = 0; + + /* Do no more processing; skip the unprocessed + * input check below. + */ + return; + } + + /* Do we have a complete row? */ + if (png_ptr->zstream.avail_out == 0) + png_push_process_row(png_ptr); + } + + /* And check for the end of the stream. */ + if (ret == Z_STREAM_END) + png_ptr->flags |= PNG_FLAG_ZSTREAM_ENDED; + } + + /* All the data should have been processed, if anything + * is left at this point we have bytes of IDAT data + * after the zlib end code. + */ + if (png_ptr->zstream.avail_in > 0) + png_warning(png_ptr, "Extra compression data in IDAT"); +} + +void /* PRIVATE */ +png_push_process_row(png_structrp png_ptr) +{ + /* 1.5.6: row_info moved out of png_struct to a local here. */ + png_row_info row_info; + + row_info.width = png_ptr->iwidth; /* NOTE: width of current interlaced row */ + row_info.color_type = png_ptr->color_type; + row_info.bit_depth = png_ptr->bit_depth; + row_info.channels = png_ptr->channels; + row_info.pixel_depth = png_ptr->pixel_depth; + row_info.rowbytes = PNG_ROWBYTES(row_info.pixel_depth, row_info.width); + + if (png_ptr->row_buf[0] > PNG_FILTER_VALUE_NONE) + { + if (png_ptr->row_buf[0] < PNG_FILTER_VALUE_LAST) + png_read_filter_row(png_ptr, &row_info, png_ptr->row_buf + 1, + png_ptr->prev_row + 1, png_ptr->row_buf[0]); + else + png_error(png_ptr, "bad adaptive filter value"); + } + + /* libpng 1.5.6: the following line was copying png_ptr->rowbytes before + * 1.5.6, while the buffer really is this big in current versions of libpng + * it may not be in the future, so this was changed just to copy the + * interlaced row count: + */ + memcpy(png_ptr->prev_row, png_ptr->row_buf, row_info.rowbytes + 1); + +#ifdef PNG_READ_TRANSFORMS_SUPPORTED + if (png_ptr->transformations != 0) + png_do_read_transformations(png_ptr, &row_info); +#endif + + /* The transformed pixel depth should match the depth now in row_info. */ + if (png_ptr->transformed_pixel_depth == 0) + { + png_ptr->transformed_pixel_depth = row_info.pixel_depth; + if (row_info.pixel_depth > png_ptr->maximum_pixel_depth) + png_error(png_ptr, "progressive row overflow"); + } + + else if (png_ptr->transformed_pixel_depth != row_info.pixel_depth) + png_error(png_ptr, "internal progressive row size calculation error"); + + +#ifdef PNG_READ_INTERLACING_SUPPORTED + /* Expand interlaced rows to full size */ + if (png_ptr->interlaced != 0 && + (png_ptr->transformations & PNG_INTERLACE) != 0) + { + if (png_ptr->pass < 6) + png_do_read_interlace(&row_info, png_ptr->row_buf + 1, png_ptr->pass, + png_ptr->transformations); + + switch (png_ptr->pass) + { + case 0: + { + int i; + for (i = 0; i < 8 && png_ptr->pass == 0; i++) + { + png_push_have_row(png_ptr, png_ptr->row_buf + 1); + png_read_push_finish_row(png_ptr); /* Updates png_ptr->pass */ + } + + if (png_ptr->pass == 2) /* Pass 1 might be empty */ + { + for (i = 0; i < 4 && png_ptr->pass == 2; i++) + { + png_push_have_row(png_ptr, NULL); + png_read_push_finish_row(png_ptr); + } + } + + if (png_ptr->pass == 4 && png_ptr->height <= 4) + { + for (i = 0; i < 2 && png_ptr->pass == 4; i++) + { + png_push_have_row(png_ptr, NULL); + png_read_push_finish_row(png_ptr); + } + } + + if (png_ptr->pass == 6 && png_ptr->height <= 4) + { + png_push_have_row(png_ptr, NULL); + png_read_push_finish_row(png_ptr); + } + + break; + } + + case 1: + { + int i; + for (i = 0; i < 8 && png_ptr->pass == 1; i++) + { + png_push_have_row(png_ptr, png_ptr->row_buf + 1); + png_read_push_finish_row(png_ptr); + } + + if (png_ptr->pass == 2) /* Skip top 4 generated rows */ + { + for (i = 0; i < 4 && png_ptr->pass == 2; i++) + { + png_push_have_row(png_ptr, NULL); + png_read_push_finish_row(png_ptr); + } + } + + break; + } + + case 2: + { + int i; + + for (i = 0; i < 4 && png_ptr->pass == 2; i++) + { + png_push_have_row(png_ptr, png_ptr->row_buf + 1); + png_read_push_finish_row(png_ptr); + } + + for (i = 0; i < 4 && png_ptr->pass == 2; i++) + { + png_push_have_row(png_ptr, NULL); + png_read_push_finish_row(png_ptr); + } + + if (png_ptr->pass == 4) /* Pass 3 might be empty */ + { + for (i = 0; i < 2 && png_ptr->pass == 4; i++) + { + png_push_have_row(png_ptr, NULL); + png_read_push_finish_row(png_ptr); + } + } + + break; + } + + case 3: + { + int i; + + for (i = 0; i < 4 && png_ptr->pass == 3; i++) + { + png_push_have_row(png_ptr, png_ptr->row_buf + 1); + png_read_push_finish_row(png_ptr); + } + + if (png_ptr->pass == 4) /* Skip top two generated rows */ + { + for (i = 0; i < 2 && png_ptr->pass == 4; i++) + { + png_push_have_row(png_ptr, NULL); + png_read_push_finish_row(png_ptr); + } + } + + break; + } + + case 4: + { + int i; + + for (i = 0; i < 2 && png_ptr->pass == 4; i++) + { + png_push_have_row(png_ptr, png_ptr->row_buf + 1); + png_read_push_finish_row(png_ptr); + } + + for (i = 0; i < 2 && png_ptr->pass == 4; i++) + { + png_push_have_row(png_ptr, NULL); + png_read_push_finish_row(png_ptr); + } + + if (png_ptr->pass == 6) /* Pass 5 might be empty */ + { + png_push_have_row(png_ptr, NULL); + png_read_push_finish_row(png_ptr); + } + + break; + } + + case 5: + { + int i; + + for (i = 0; i < 2 && png_ptr->pass == 5; i++) + { + png_push_have_row(png_ptr, png_ptr->row_buf + 1); + png_read_push_finish_row(png_ptr); + } + + if (png_ptr->pass == 6) /* Skip top generated row */ + { + png_push_have_row(png_ptr, NULL); + png_read_push_finish_row(png_ptr); + } + + break; + } + + default: + case 6: + { + png_push_have_row(png_ptr, png_ptr->row_buf + 1); + png_read_push_finish_row(png_ptr); + + if (png_ptr->pass != 6) + break; + + png_push_have_row(png_ptr, NULL); + png_read_push_finish_row(png_ptr); + } + } + } + else + { + png_push_have_row(png_ptr, png_ptr->row_buf + 1); + png_read_push_finish_row(png_ptr); + } +} + +void /* PRIVATE */ +png_read_push_finish_row(png_structrp png_ptr) +{ + /* Arrays to facilitate easy interlacing - use pass (0 - 6) as index */ + + /* Start of interlace block */ + static PNG_CONST png_byte png_pass_start[] = {0, 4, 0, 2, 0, 1, 0}; + + /* Offset to next interlace block */ + static PNG_CONST png_byte png_pass_inc[] = {8, 8, 4, 4, 2, 2, 1}; + + /* Start of interlace block in the y direction */ + static PNG_CONST png_byte png_pass_ystart[] = {0, 0, 4, 0, 2, 0, 1}; + + /* Offset to next interlace block in the y direction */ + static PNG_CONST png_byte png_pass_yinc[] = {8, 8, 8, 4, 4, 2, 2}; + + /* Height of interlace block. This is not currently used - if you need + * it, uncomment it here and in png.h + static PNG_CONST png_byte png_pass_height[] = {8, 8, 4, 4, 2, 2, 1}; + */ +#endif + + png_ptr->row_number++; + if (png_ptr->row_number < png_ptr->num_rows) + return; + + if (png_ptr->interlaced != 0) + { + png_ptr->row_number = 0; + memset(png_ptr->prev_row, 0, png_ptr->rowbytes + 1); + + do + { + png_ptr->pass++; + if ((png_ptr->pass == 1 && png_ptr->width < 5) || + (png_ptr->pass == 3 && png_ptr->width < 3) || + (png_ptr->pass == 5 && png_ptr->width < 2)) + png_ptr->pass++; + + if (png_ptr->pass > 7) + png_ptr->pass--; + + if (png_ptr->pass >= 7) + break; + + png_ptr->iwidth = (png_ptr->width + + png_pass_inc[png_ptr->pass] - 1 - + png_pass_start[png_ptr->pass]) / + png_pass_inc[png_ptr->pass]; + + if ((png_ptr->transformations & PNG_INTERLACE) != 0) + break; + + png_ptr->num_rows = (png_ptr->height + + png_pass_yinc[png_ptr->pass] - 1 - + png_pass_ystart[png_ptr->pass]) / + png_pass_yinc[png_ptr->pass]; + + } while (png_ptr->iwidth == 0 || png_ptr->num_rows == 0); + } +} + +void /* PRIVATE */ +png_push_have_info(png_structrp png_ptr, png_inforp info_ptr) +{ + if (png_ptr->info_fn != NULL) + (*(png_ptr->info_fn))(png_ptr, info_ptr); +} + +void /* PRIVATE */ +png_push_have_end(png_structrp png_ptr, png_inforp info_ptr) +{ + if (png_ptr->end_fn != NULL) + (*(png_ptr->end_fn))(png_ptr, info_ptr); +} + +void /* PRIVATE */ +png_push_have_row(png_structrp png_ptr, png_bytep row) +{ + if (png_ptr->row_fn != NULL) + (*(png_ptr->row_fn))(png_ptr, row, png_ptr->row_number, + (int)png_ptr->pass); +} + +void PNGAPI +png_progressive_combine_row(png_const_structrp png_ptr, png_bytep old_row, + png_const_bytep new_row) +{ + if (png_ptr == NULL) + return; + + /* new_row is a flag here - if it is NULL then the app callback was called + * from an empty row (see the calls to png_struct::row_fn below), otherwise + * it must be png_ptr->row_buf+1 + */ + if (new_row != NULL) + png_combine_row(png_ptr, old_row, 1/*blocky display*/); +} + +void PNGAPI +png_set_progressive_read_fn(png_structrp png_ptr, png_voidp progressive_ptr, + png_progressive_info_ptr info_fn, png_progressive_row_ptr row_fn, + png_progressive_end_ptr end_fn) +{ + if (png_ptr == NULL) + return; + + png_ptr->info_fn = info_fn; + png_ptr->row_fn = row_fn; + png_ptr->end_fn = end_fn; + + png_set_read_fn(png_ptr, progressive_ptr, png_push_fill_buffer); +} + +png_voidp PNGAPI +png_get_progressive_ptr(png_const_structrp png_ptr) +{ + if (png_ptr == NULL) + return (NULL); + + return png_ptr->io_ptr; +} +#endif /* PROGRESSIVE_READ */ diff --git a/app/src/main/jni/libpng/pngpriv.h b/app/src/main/jni/libpng/pngpriv.h new file mode 100644 index 000000000..e0aec2034 --- /dev/null +++ b/app/src/main/jni/libpng/pngpriv.h @@ -0,0 +1,1944 @@ + +/* pngpriv.h - private declarations for use inside libpng + * + * For conditions of distribution and use, see copyright notice in png.h + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * Last changed in libpng 1.6.10 [March 6, 1014]] + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + */ + +/* The symbols declared in this file (including the functions declared + * as extern) are PRIVATE. They are not part of the libpng public + * interface, and are not recommended for use by regular applications. + * Some of them may become public in the future; others may stay private, + * change in an incompatible way, or even disappear. + * Although the libpng users are not forbidden to include this header, + * they should be well aware of the issues that may arise from doing so. + */ + +#ifndef PNGPRIV_H +#define PNGPRIV_H + +/* Feature Test Macros. The following are defined here to ensure that correctly + * implemented libraries reveal the APIs libpng needs to build and hide those + * that are not needed and potentially damaging to the compilation. + * + * Feature Test Macros must be defined before any system header is included (see + * POSIX 1003.1 2.8.2 "POSIX Symbols." + * + * These macros only have an effect if the operating system supports either + * POSIX 1003.1 or C99, or both. On other operating systems (particularly + * Windows/Visual Studio) there is no effect; the OS specific tests below are + * still required (as of 2011-05-02.) + */ +#define _POSIX_SOURCE 1 /* Just the POSIX 1003.1 and C89 APIs */ + +#ifndef PNG_VERSION_INFO_ONLY +/* Standard library headers not required by png.h: */ +# include +# include +#endif + +#define PNGLIB_BUILD /*libpng is being built, not used*/ + +/* If HAVE_CONFIG_H is defined during the build then the build system must + * provide an appropriate "config.h" file on the include path. The header file + * must provide definitions as required below (search for "HAVE_CONFIG_H"); + * see configure.ac for more details of the requirements. The macro + * "PNG_NO_CONFIG_H" is provided for maintainers to test for dependencies on + * 'configure'; define this macro to prevent the configure build including the + * configure generated config.h. Libpng is expected to compile without *any* + * special build system support on a reasonably ANSI-C compliant system. + */ +#if defined(HAVE_CONFIG_H) && !defined(PNG_NO_CONFIG_H) +# include + + /* Pick up the definition of 'restrict' from config.h if it was read: */ +# define PNG_RESTRICT restrict +#endif + +/* To support symbol prefixing it is necessary to know *before* including png.h + * whether the fixed point (and maybe other) APIs are exported, because if they + * are not internal definitions may be required. This is handled below just + * before png.h is included, but load the configuration now if it is available. + */ +#ifndef PNGLCONF_H +# include "pnglibconf.h" +#endif + +/* Local renames may change non-exported API functions from png.h */ +#if defined(PNG_PREFIX) && !defined(PNGPREFIX_H) +# include "pngprefix.h" +#endif + +#ifdef PNG_USER_CONFIG +# include "pngusr.h" + /* These should have been defined in pngusr.h */ +# ifndef PNG_USER_PRIVATEBUILD +# define PNG_USER_PRIVATEBUILD "Custom libpng build" +# endif +# ifndef PNG_USER_DLLFNAME_POSTFIX +# define PNG_USER_DLLFNAME_POSTFIX "Cb" +# endif +#endif + +/* Compile time options. + * ===================== + * In a multi-arch build the compiler may compile the code several times for the + * same object module, producing different binaries for different architectures. + * When this happens configure-time setting of the target host options cannot be + * done and this interferes with the handling of the ARM NEON optimizations, and + * possibly other similar optimizations. Put additional tests here; in general + * this is needed when the same option can be changed at both compile time and + * run time depending on the target OS (i.e. iOS vs Android.) + * + * NOTE: symbol prefixing does not pass $(CFLAGS) to the preprocessor, because + * this is not possible with certain compilers (Oracle SUN OS CC), as a result + * it is necessary to ensure that all extern functions that *might* be used + * regardless of $(CFLAGS) get declared in this file. The test on __ARM_NEON__ + * below is one example of this behavior because it is controlled by the + * presence or not of -mfpu=neon on the GCC command line, it is possible to do + * this in $(CC), e.g. "CC=gcc -mfpu=neon", but people who build libpng rarely + * do this. + */ +#ifndef PNG_ARM_NEON_OPT + /* ARM NEON optimizations are being controlled by the compiler settings, + * typically the target FPU. If the FPU has been set to NEON (-mfpu=neon + * with GCC) then the compiler will define __ARM_NEON__ and we can rely + * unconditionally on NEON instructions not crashing, otherwise we must + * disable use of NEON instructions. + * + * NOTE: at present these optimizations depend on 'ALIGNED_MEMORY', so they + * can only be turned on automatically if that is supported too. If + * PNG_ARM_NEON_OPT is set in CPPFLAGS (to >0) then arm/arm_init.c will fail + * to compile with an appropriate #error if ALIGNED_MEMORY has been turned + * off. + * + * Note that gcc-4.9 defines __ARM_NEON instead of __ARM_NEON__, so we + * check both variants. + */ +# if (defined(__ARM_NEON__) || defined(__ARM_NEON)) && \ + defined(PNG_ALIGNED_MEMORY_SUPPORTED) +# define PNG_ARM_NEON_OPT 2 +# else +# define PNG_ARM_NEON_OPT 0 +# endif +#endif + +#if PNG_ARM_NEON_OPT > 0 + /* NEON optimizations are to be at least considered by libpng, so enable the + * callbacks to do this. + */ +# define PNG_FILTER_OPTIMIZATIONS png_init_filter_functions_neon + + /* By default the 'intrinsics' code in arm/filter_neon_intrinsics.c is used + * if possible - if __ARM_NEON__ is set and the compiler version is not known + * to be broken. This is controlled by PNG_ARM_NEON_IMPLEMENTATION which can + * be: + * + * 1 The intrinsics code (the default with __ARM_NEON__) + * 2 The hand coded assembler (the default without __ARM_NEON__) + * + * It is possible to set PNG_ARM_NEON_IMPLEMENTATION in CPPFLAGS, however + * this is *NOT* supported and may cease to work even after a minor revision + * to libpng. It *is* valid to do this for testing purposes, e.g. speed + * testing or a new compiler, but the results should be communicated to the + * libpng implementation list for incorporation in the next minor release. + */ +# ifndef PNG_ARM_NEON_IMPLEMENTATION +# if defined(__ARM_NEON__) || defined(__ARM_NEON) +# if defined(__clang__) + /* At present it is unknown by the libpng developers which versions + * of clang support the intrinsics, however some or perhaps all + * versions do not work with the assembler so this may be + * irrelevant, so just use the default (do nothing here.) + */ +# elif defined(__GNUC__) + /* GCC 4.5.4 NEON support is known to be broken. 4.6.3 is known to + * work, so if this *is* GCC, or G++, look for a version >4.5 + */ +# if __GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 6) +# define PNG_ARM_NEON_IMPLEMENTATION 2 +# endif /* no GNUC support */ +# endif /* __GNUC__ */ +# else /* !defined __ARM_NEON__ */ + /* The 'intrinsics' code simply won't compile without this -mfpu=neon: + */ +# define PNG_ARM_NEON_IMPLEMENTATION 2 +# endif /* __ARM_NEON__ */ +# endif /* !PNG_ARM_NEON_IMPLEMENTATION */ + +# ifndef PNG_ARM_NEON_IMPLEMENTATION + /* Use the intrinsics code by default. */ +# define PNG_ARM_NEON_IMPLEMENTATION 1 +# endif +#endif /* PNG_ARM_NEON_OPT > 0 */ + +/* Is this a build of a DLL where compilation of the object modules requires + * different preprocessor settings to those required for a simple library? If + * so PNG_BUILD_DLL must be set. + * + * If libpng is used inside a DLL but that DLL does not export the libpng APIs + * PNG_BUILD_DLL must not be set. To avoid the code below kicking in build a + * static library of libpng then link the DLL against that. + */ +#ifndef PNG_BUILD_DLL +# ifdef DLL_EXPORT + /* This is set by libtool when files are compiled for a DLL; libtool + * always compiles twice, even on systems where it isn't necessary. Set + * PNG_BUILD_DLL in case it is necessary: + */ +# define PNG_BUILD_DLL +# else +# ifdef _WINDLL + /* This is set by the Microsoft Visual Studio IDE in projects that + * build a DLL. It can't easily be removed from those projects (it + * isn't visible in the Visual Studio UI) so it is a fairly reliable + * indication that PNG_IMPEXP needs to be set to the DLL export + * attributes. + */ +# define PNG_BUILD_DLL +# else +# ifdef __DLL__ + /* This is set by the Borland C system when compiling for a DLL + * (as above.) + */ +# define PNG_BUILD_DLL +# else + /* Add additional compiler cases here. */ +# endif +# endif +# endif +#endif /* Setting PNG_BUILD_DLL if required */ + +/* See pngconf.h for more details: the builder of the library may set this on + * the command line to the right thing for the specific compilation system or it + * may be automagically set above (at present we know of no system where it does + * need to be set on the command line.) + * + * PNG_IMPEXP must be set here when building the library to prevent pngconf.h + * setting it to the "import" setting for a DLL build. + */ +#ifndef PNG_IMPEXP +# ifdef PNG_BUILD_DLL +# define PNG_IMPEXP PNG_DLL_EXPORT +# else + /* Not building a DLL, or the DLL doesn't require specific export + * definitions. + */ +# define PNG_IMPEXP +# endif +#endif + +/* No warnings for private or deprecated functions in the build: */ +#ifndef PNG_DEPRECATED +# define PNG_DEPRECATED +#endif +#ifndef PNG_PRIVATE +# define PNG_PRIVATE +#endif + +/* Symbol preprocessing support. + * + * To enable listing global, but internal, symbols the following macros should + * always be used to declare an extern data or function object in this file. + */ +#ifndef PNG_INTERNAL_DATA +# define PNG_INTERNAL_DATA(type, name, array) extern type name array +#endif + +#ifndef PNG_INTERNAL_FUNCTION +# define PNG_INTERNAL_FUNCTION(type, name, args, attributes)\ + extern PNG_FUNCTION(type, name, args, PNG_EMPTY attributes) +#endif + +#ifndef PNG_INTERNAL_CALLBACK +# define PNG_INTERNAL_CALLBACK(type, name, args, attributes)\ + extern PNG_FUNCTION(type, (PNGCBAPI name), args, PNG_EMPTY attributes) +#endif + +/* If floating or fixed point APIs are disabled they may still be compiled + * internally. To handle this make sure they are declared as the appropriate + * internal extern function (otherwise the symbol prefixing stuff won't work and + * the functions will be used without definitions.) + * + * NOTE: although all the API functions are declared here they are not all + * actually built! Because the declarations are still made it is necessary to + * fake out types that they depend on. + */ +#ifndef PNG_FP_EXPORT +# ifndef PNG_FLOATING_POINT_SUPPORTED +# define PNG_FP_EXPORT(ordinal, type, name, args)\ + PNG_INTERNAL_FUNCTION(type, name, args, PNG_EMPTY); +# ifndef PNG_VERSION_INFO_ONLY + typedef struct png_incomplete png_double; + typedef png_double* png_doublep; + typedef const png_double* png_const_doublep; + typedef png_double** png_doublepp; +# endif +# endif +#endif +#ifndef PNG_FIXED_EXPORT +# ifndef PNG_FIXED_POINT_SUPPORTED +# define PNG_FIXED_EXPORT(ordinal, type, name, args)\ + PNG_INTERNAL_FUNCTION(type, name, args, PNG_EMPTY); +# endif +#endif + +#include "png.h" + +/* pngconf.h does not set PNG_DLL_EXPORT unless it is required, so: */ +#ifndef PNG_DLL_EXPORT +# define PNG_DLL_EXPORT +#endif + +/* SECURITY and SAFETY: + * + * By default libpng is built without any internal limits on image size, + * individual heap (png_malloc) allocations or the total amount of memory used. + * If PNG_SAFE_LIMITS_SUPPORTED is defined, however, the limits below are used + * (unless individually overridden). These limits are believed to be fairly + * safe, but builders of secure systems should verify the values against the + * real system capabilities. + */ +#ifdef PNG_SAFE_LIMITS_SUPPORTED + /* 'safe' limits */ +# ifndef PNG_USER_WIDTH_MAX +# define PNG_USER_WIDTH_MAX 1000000 +# endif +# ifndef PNG_USER_HEIGHT_MAX +# define PNG_USER_HEIGHT_MAX 1000000 +# endif +# ifndef PNG_USER_CHUNK_CACHE_MAX +# define PNG_USER_CHUNK_CACHE_MAX 128 +# endif +# ifndef PNG_USER_CHUNK_MALLOC_MAX +# define PNG_USER_CHUNK_MALLOC_MAX 8000000 +# endif +#else + /* values for no limits */ +# ifndef PNG_USER_WIDTH_MAX +# define PNG_USER_WIDTH_MAX 0x7fffffff +# endif +# ifndef PNG_USER_HEIGHT_MAX +# define PNG_USER_HEIGHT_MAX 0x7fffffff +# endif +# ifndef PNG_USER_CHUNK_CACHE_MAX +# define PNG_USER_CHUNK_CACHE_MAX 0 +# endif +# ifndef PNG_USER_CHUNK_MALLOC_MAX +# define PNG_USER_CHUNK_MALLOC_MAX 0 +# endif +#endif + +/* Moved to pngpriv.h at libpng-1.5.0 */ +/* NOTE: some of these may have been used in external applications as + * these definitions were exposed in pngconf.h prior to 1.5. + */ + +/* If you are running on a machine where you cannot allocate more + * than 64K of memory at once, uncomment this. While libpng will not + * normally need that much memory in a chunk (unless you load up a very + * large file), zlib needs to know how big of a chunk it can use, and + * libpng thus makes sure to check any memory allocation to verify it + * will fit into memory. + * + * zlib provides 'MAXSEG_64K' which, if defined, indicates the + * same limit and pngconf.h (already included) sets the limit + * if certain operating systems are detected. + */ +#if defined(MAXSEG_64K) && !defined(PNG_MAX_MALLOC_64K) +# define PNG_MAX_MALLOC_64K +#endif + +#ifndef PNG_UNUSED +/* Unused formal parameter warnings are silenced using the following macro + * which is expected to have no bad effects on performance (optimizing + * compilers will probably remove it entirely). Note that if you replace + * it with something other than whitespace, you must include the terminating + * semicolon. + */ +# define PNG_UNUSED(param) (void)param; +#endif + +/* Just a little check that someone hasn't tried to define something + * contradictory. + */ +#if (PNG_ZBUF_SIZE > 65536L) && defined(PNG_MAX_MALLOC_64K) +# undef PNG_ZBUF_SIZE +# define PNG_ZBUF_SIZE 65536L +#endif + +/* If warnings or errors are turned off the code is disabled or redirected here. + * From 1.5.4 functions have been added to allow very limited formatting of + * error and warning messages - this code will also be disabled here. + */ +#ifdef PNG_WARNINGS_SUPPORTED +# define PNG_WARNING_PARAMETERS(p) png_warning_parameters p; +#else +# define png_warning_parameter(p,number,string) ((void)0) +# define png_warning_parameter_unsigned(p,number,format,value) ((void)0) +# define png_warning_parameter_signed(p,number,format,value) ((void)0) +# define png_formatted_warning(pp,p,message) ((void)(pp)) +# define PNG_WARNING_PARAMETERS(p) +#endif +#ifndef PNG_ERROR_TEXT_SUPPORTED +# define png_fixed_error(s1,s2) png_err(s1) +#endif + +/* C allows up-casts from (void*) to any pointer and (const void*) to any + * pointer to a const object. C++ regards this as a type error and requires an + * explicit, static, cast and provides the static_cast<> rune to ensure that + * const is not cast away. + */ +#ifdef __cplusplus +# define png_voidcast(type, value) static_cast(value) +# define png_constcast(type, value) const_cast(value) +# define png_aligncast(type, value) \ + static_cast(static_cast(value)) +# define png_aligncastconst(type, value) \ + static_cast(static_cast(value)) +#else +# define png_voidcast(type, value) (value) +# define png_constcast(type, value) ((type)(value)) +# define png_aligncast(type, value) ((void*)(value)) +# define png_aligncastconst(type, value) ((const void*)(value)) +#endif /* __cplusplus */ + +/* Some fixed point APIs are still required even if not exported because + * they get used by the corresponding floating point APIs. This magic + * deals with this: + */ +#ifdef PNG_FIXED_POINT_SUPPORTED +# define PNGFAPI PNGAPI +#else +# define PNGFAPI /* PRIVATE */ +#endif + +#ifndef PNG_VERSION_INFO_ONLY +/* Other defines specific to compilers can go here. Try to keep + * them inside an appropriate ifdef/endif pair for portability. + */ +#if defined(PNG_FLOATING_POINT_SUPPORTED) ||\ + defined(PNG_FLOATING_ARITHMETIC_SUPPORTED) + /* png.c requires the following ANSI-C constants if the conversion of + * floating point to ASCII is implemented therein: + * + * DBL_DIG Maximum number of decimal digits (can be set to any constant) + * DBL_MIN Smallest normalized fp number (can be set to an arbitrary value) + * DBL_MAX Maximum floating point number (can be set to an arbitrary value) + */ +# include + +# if (defined(__MWERKS__) && defined(macintosh)) || defined(applec) || \ + defined(THINK_C) || defined(__SC__) || defined(TARGET_OS_MAC) + /* We need to check that hasn't already been included earlier + * as it seems it doesn't agree with , yet we should really use + * if possible. + */ +# if !defined(__MATH_H__) && !defined(__MATH_H) && !defined(__cmath__) +# include +# endif +# else +# include +# endif +# if defined(_AMIGA) && defined(__SASC) && defined(_M68881) + /* Amiga SAS/C: We must include builtin FPU functions when compiling using + * MATH=68881 + */ +# include +# endif +#endif + +/* This provides the non-ANSI (far) memory allocation routines. */ +#if defined(__TURBOC__) && defined(__MSDOS__) +# include +# include +#endif + +#if defined(WIN32) || defined(_Windows) || defined(_WINDOWS) || \ + defined(_WIN32) || defined(__WIN32__) +# include /* defines _WINDOWS_ macro */ +#endif +#endif /* PNG_VERSION_INFO_ONLY */ + +/* Moved here around 1.5.0beta36 from pngconf.h */ +/* Users may want to use these so they are not private. Any library + * functions that are passed far data must be model-independent. + */ + +/* Memory model/platform independent fns */ +#ifndef PNG_ABORT +# ifdef _WINDOWS_ +# define PNG_ABORT() ExitProcess(0) +# else +# define PNG_ABORT() abort() +# endif +#endif + +/* These macros may need to be architecture dependent. */ +#define PNG_ALIGN_NONE 0 /* do not use data alignment */ +#define PNG_ALIGN_ALWAYS 1 /* assume unaligned accesses are OK */ +#ifdef offsetof +# define PNG_ALIGN_OFFSET 2 /* use offsetof to determine alignment */ +#else +# define PNG_ALIGN_OFFSET -1 /* prevent the use of this */ +#endif +#define PNG_ALIGN_SIZE 3 /* use sizeof to determine alignment */ + +#ifndef PNG_ALIGN_TYPE + /* Default to using aligned access optimizations and requiring alignment to a + * multiple of the data type size. Override in a compiler specific fashion + * if necessary by inserting tests here: + */ +# define PNG_ALIGN_TYPE PNG_ALIGN_SIZE +#endif + +#if PNG_ALIGN_TYPE == PNG_ALIGN_SIZE + /* This is used because in some compiler implementations non-aligned + * structure members are supported, so the offsetof approach below fails. + * Set PNG_ALIGN_SIZE=0 for compiler combinations where unaligned access + * is good for performance. Do not do this unless you have tested the result + * and understand it. + */ +# define png_alignof(type) (sizeof (type)) +#else +# if PNG_ALIGN_TYPE == PNG_ALIGN_OFFSET +# define png_alignof(type) offsetof(struct{char c; type t;}, t) +# else +# if PNG_ALIGN_TYPE == PNG_ALIGN_ALWAYS +# define png_alignof(type) (1) +# endif + /* Else leave png_alignof undefined to prevent use thereof */ +# endif +#endif + +/* This implicitly assumes alignment is always to a power of 2. */ +#ifdef png_alignof +# define png_isaligned(ptr, type)\ + ((((const char*)ptr-(const char*)0) & (png_alignof(type)-1)) == 0) +#else +# define png_isaligned(ptr, type) 0 +#endif + +/* End of memory model/platform independent support */ +/* End of 1.5.0beta36 move from pngconf.h */ + +/* CONSTANTS and UTILITY MACROS + * These are used internally by libpng and not exposed in the API + */ + +/* Various modes of operation. Note that after an init, mode is set to + * zero automatically when the structure is created. Three of these + * are defined in png.h because they need to be visible to applications + * that call png_set_unknown_chunk(). + */ +/* #define PNG_HAVE_IHDR 0x01 (defined in png.h) */ +/* #define PNG_HAVE_PLTE 0x02 (defined in png.h) */ +#define PNG_HAVE_IDAT 0x04 +/* #define PNG_AFTER_IDAT 0x08 (defined in png.h) */ +#define PNG_HAVE_IEND 0x10 + /* 0x20 (unused) */ + /* 0x40 (unused) */ + /* 0x80 (unused) */ +#define PNG_HAVE_CHUNK_HEADER 0x100 +#define PNG_WROTE_tIME 0x200 +#define PNG_WROTE_INFO_BEFORE_PLTE 0x400 +#define PNG_BACKGROUND_IS_GRAY 0x800 +#define PNG_HAVE_PNG_SIGNATURE 0x1000 +#define PNG_HAVE_CHUNK_AFTER_IDAT 0x2000 /* Have another chunk after IDAT */ + /* 0x4000 (unused) */ +#define PNG_IS_READ_STRUCT 0x8000 /* Else is a write struct */ + +/* Flags for the transformations the PNG library does on the image data */ +#define PNG_BGR 0x0001 +#define PNG_INTERLACE 0x0002 +#define PNG_PACK 0x0004 +#define PNG_SHIFT 0x0008 +#define PNG_SWAP_BYTES 0x0010 +#define PNG_INVERT_MONO 0x0020 +#define PNG_QUANTIZE 0x0040 +#define PNG_COMPOSE 0x0080 /* Was PNG_BACKGROUND */ +#define PNG_BACKGROUND_EXPAND 0x0100 +#define PNG_EXPAND_16 0x0200 /* Added to libpng 1.5.2 */ +#define PNG_16_TO_8 0x0400 /* Becomes 'chop' in 1.5.4 */ +#define PNG_RGBA 0x0800 +#define PNG_EXPAND 0x1000 +#define PNG_GAMMA 0x2000 +#define PNG_GRAY_TO_RGB 0x4000 +#define PNG_FILLER 0x8000 +#define PNG_PACKSWAP 0x10000 +#define PNG_SWAP_ALPHA 0x20000 +#define PNG_STRIP_ALPHA 0x40000 +#define PNG_INVERT_ALPHA 0x80000 +#define PNG_USER_TRANSFORM 0x100000 +#define PNG_RGB_TO_GRAY_ERR 0x200000 +#define PNG_RGB_TO_GRAY_WARN 0x400000 +#define PNG_RGB_TO_GRAY 0x600000 /* two bits, RGB_TO_GRAY_ERR|WARN */ +#define PNG_ENCODE_ALPHA 0x800000 /* Added to libpng-1.5.4 */ +#define PNG_ADD_ALPHA 0x1000000 /* Added to libpng-1.2.7 */ +#define PNG_EXPAND_tRNS 0x2000000 /* Added to libpng-1.2.9 */ +#define PNG_SCALE_16_TO_8 0x4000000 /* Added to libpng-1.5.4 */ + /* 0x8000000 unused */ + /* 0x10000000 unused */ + /* 0x20000000 unused */ + /* 0x40000000 unused */ +/* Flags for png_create_struct */ +#define PNG_STRUCT_PNG 0x0001 +#define PNG_STRUCT_INFO 0x0002 + +/* Scaling factor for filter heuristic weighting calculations */ +#define PNG_WEIGHT_FACTOR (1<<(PNG_WEIGHT_SHIFT)) +#define PNG_COST_FACTOR (1<<(PNG_COST_SHIFT)) + +/* Flags for the png_ptr->flags rather than declaring a byte for each one */ +#define PNG_FLAG_ZLIB_CUSTOM_STRATEGY 0x0001 +#define PNG_FLAG_ZSTREAM_INITIALIZED 0x0002 /* Added to libpng-1.6.0 */ + /* 0x0004 unused */ +#define PNG_FLAG_ZSTREAM_ENDED 0x0008 /* Added to libpng-1.6.0 */ + /* 0x0010 unused */ + /* 0x0020 unused */ +#define PNG_FLAG_ROW_INIT 0x0040 +#define PNG_FLAG_FILLER_AFTER 0x0080 +#define PNG_FLAG_CRC_ANCILLARY_USE 0x0100 +#define PNG_FLAG_CRC_ANCILLARY_NOWARN 0x0200 +#define PNG_FLAG_CRC_CRITICAL_USE 0x0400 +#define PNG_FLAG_CRC_CRITICAL_IGNORE 0x0800 +#define PNG_FLAG_ASSUME_sRGB 0x1000 /* Added to libpng-1.5.4 */ +#define PNG_FLAG_OPTIMIZE_ALPHA 0x2000 /* Added to libpng-1.5.4 */ +#define PNG_FLAG_DETECT_UNINITIALIZED 0x4000 /* Added to libpng-1.5.4 */ +/* #define PNG_FLAG_KEEP_UNKNOWN_CHUNKS 0x8000 */ +/* #define PNG_FLAG_KEEP_UNSAFE_CHUNKS 0x10000 */ +#define PNG_FLAG_LIBRARY_MISMATCH 0x20000 +#define PNG_FLAG_STRIP_ERROR_NUMBERS 0x40000 +#define PNG_FLAG_STRIP_ERROR_TEXT 0x80000 +#define PNG_FLAG_BENIGN_ERRORS_WARN 0x100000 /* Added to libpng-1.4.0 */ +#define PNG_FLAG_APP_WARNINGS_WARN 0x200000 /* Added to libpng-1.6.0 */ +#define PNG_FLAG_APP_ERRORS_WARN 0x400000 /* Added to libpng-1.6.0 */ + /* 0x800000 unused */ + /* 0x1000000 unused */ + /* 0x2000000 unused */ + /* 0x4000000 unused */ + /* 0x8000000 unused */ + /* 0x10000000 unused */ + /* 0x20000000 unused */ + /* 0x40000000 unused */ + +#define PNG_FLAG_CRC_ANCILLARY_MASK (PNG_FLAG_CRC_ANCILLARY_USE | \ + PNG_FLAG_CRC_ANCILLARY_NOWARN) + +#define PNG_FLAG_CRC_CRITICAL_MASK (PNG_FLAG_CRC_CRITICAL_USE | \ + PNG_FLAG_CRC_CRITICAL_IGNORE) + +#define PNG_FLAG_CRC_MASK (PNG_FLAG_CRC_ANCILLARY_MASK | \ + PNG_FLAG_CRC_CRITICAL_MASK) + +/* Save typing and make code easier to understand */ + +#define PNG_COLOR_DIST(c1, c2) (abs((int)((c1).red) - (int)((c2).red)) + \ + abs((int)((c1).green) - (int)((c2).green)) + \ + abs((int)((c1).blue) - (int)((c2).blue))) + +/* Added to libpng-1.6.0: scale a 16-bit value in the range 0..65535 to 0..255 + * by dividing by 257 *with rounding*. This macro is exact for the given range. + * See the discourse in pngrtran.c png_do_scale_16_to_8. The values in the + * macro were established by experiment (modifying the added value). The macro + * has a second variant that takes a value already scaled by 255 and divides by + * 65535 - this has a maximum error of .502. Over the range 0..65535*65535 it + * only gives off-by-one errors and only for 0.5% (1 in 200) of the values. + */ +#define PNG_DIV65535(v24) (((v24) + 32895) >> 16) +#define PNG_DIV257(v16) PNG_DIV65535((png_uint_32)(v16) * 255) + +/* Added to libpng-1.2.6 JB */ +#define PNG_ROWBYTES(pixel_bits, width) \ + ((pixel_bits) >= 8 ? \ + ((png_size_t)(width) * (((png_size_t)(pixel_bits)) >> 3)) : \ + (( ((png_size_t)(width) * ((png_size_t)(pixel_bits))) + 7) >> 3) ) + +/* PNG_OUT_OF_RANGE returns true if value is outside the range + * ideal-delta..ideal+delta. Each argument is evaluated twice. + * "ideal" and "delta" should be constants, normally simple + * integers, "value" a variable. Added to libpng-1.2.6 JB + */ +#define PNG_OUT_OF_RANGE(value, ideal, delta) \ + ( (value) < (ideal)-(delta) || (value) > (ideal)+(delta) ) + +/* Conversions between fixed and floating point, only defined if + * required (to make sure the code doesn't accidentally use float + * when it is supposedly disabled.) + */ +#ifdef PNG_FLOATING_POINT_SUPPORTED +/* The floating point conversion can't overflow, though it can and + * does lose accuracy relative to the original fixed point value. + * In practice this doesn't matter because png_fixed_point only + * stores numbers with very low precision. The png_ptr and s + * arguments are unused by default but are there in case error + * checking becomes a requirement. + */ +#define png_float(png_ptr, fixed, s) (.00001 * (fixed)) + +/* The fixed point conversion performs range checking and evaluates + * its argument multiple times, so must be used with care. The + * range checking uses the PNG specification values for a signed + * 32 bit fixed point value except that the values are deliberately + * rounded-to-zero to an integral value - 21474 (21474.83 is roughly + * (2^31-1) * 100000). 's' is a string that describes the value being + * converted. + * + * NOTE: this macro will raise a png_error if the range check fails, + * therefore it is normally only appropriate to use this on values + * that come from API calls or other sources where an out of range + * error indicates a programming error, not a data error! + * + * NOTE: by default this is off - the macro is not used - because the + * function call saves a lot of code. + */ +#ifdef PNG_FIXED_POINT_MACRO_SUPPORTED +#define png_fixed(png_ptr, fp, s) ((fp) <= 21474 && (fp) >= -21474 ?\ + ((png_fixed_point)(100000 * (fp))) : (png_fixed_error(png_ptr, s),0)) +#endif +/* else the corresponding function is defined below, inside the scope of the + * cplusplus test. + */ +#endif + +/* Constants for known chunk types. If you need to add a chunk, define the name + * here. For historical reasons these constants have the form png_; i.e. + * the prefix is lower case. Please use decimal values as the parameters to + * match the ISO PNG specification and to avoid relying on the C locale + * interpretation of character values. + * + * Prior to 1.5.6 these constants were strings, as of 1.5.6 png_uint_32 values + * are computed and a new macro (PNG_STRING_FROM_CHUNK) added to allow a string + * to be generated if required. + * + * PNG_32b correctly produces a value shifted by up to 24 bits, even on + * architectures where (int) is only 16 bits. + */ +#define PNG_32b(b,s) ((png_uint_32)(b) << (s)) +#define PNG_U32(b1,b2,b3,b4) \ + (PNG_32b(b1,24) | PNG_32b(b2,16) | PNG_32b(b3,8) | PNG_32b(b4,0)) + +/* Constants for known chunk types. + * + * MAINTAINERS: If you need to add a chunk, define the name here. + * For historical reasons these constants have the form png_; i.e. + * the prefix is lower case. Please use decimal values as the parameters to + * match the ISO PNG specification and to avoid relying on the C locale + * interpretation of character values. Please keep the list sorted. + * + * Notice that PNG_U32 is used to define a 32-bit value for the 4 byte chunk + * type. In fact the specification does not express chunk types this way, + * however using a 32-bit value means that the chunk type can be read from the + * stream using exactly the same code as used for a 32-bit unsigned value and + * can be examined far more efficiently (using one arithmetic compare). + * + * Prior to 1.5.6 the chunk type constants were expressed as C strings. The + * libpng API still uses strings for 'unknown' chunks and a macro, + * PNG_STRING_FROM_CHUNK, allows a string to be generated if required. Notice + * that for portable code numeric values must still be used; the string "IHDR" + * is not portable and neither is PNG_U32('I', 'H', 'D', 'R'). + * + * In 1.7.0 the definitions will be made public in png.h to avoid having to + * duplicate the same definitions in application code. + */ +#define png_IDAT PNG_U32( 73, 68, 65, 84) +#define png_IEND PNG_U32( 73, 69, 78, 68) +#define png_IHDR PNG_U32( 73, 72, 68, 82) +#define png_PLTE PNG_U32( 80, 76, 84, 69) +#define png_bKGD PNG_U32( 98, 75, 71, 68) +#define png_cHRM PNG_U32( 99, 72, 82, 77) +#define png_fRAc PNG_U32(102, 82, 65, 99) /* registered, not defined */ +#define png_gAMA PNG_U32(103, 65, 77, 65) +#define png_gIFg PNG_U32(103, 73, 70, 103) +#define png_gIFt PNG_U32(103, 73, 70, 116) /* deprecated */ +#define png_gIFx PNG_U32(103, 73, 70, 120) +#define png_hIST PNG_U32(104, 73, 83, 84) +#define png_iCCP PNG_U32(105, 67, 67, 80) +#define png_iTXt PNG_U32(105, 84, 88, 116) +#define png_oFFs PNG_U32(111, 70, 70, 115) +#define png_pCAL PNG_U32(112, 67, 65, 76) +#define png_pHYs PNG_U32(112, 72, 89, 115) +#define png_sBIT PNG_U32(115, 66, 73, 84) +#define png_sCAL PNG_U32(115, 67, 65, 76) +#define png_sPLT PNG_U32(115, 80, 76, 84) +#define png_sRGB PNG_U32(115, 82, 71, 66) +#define png_sTER PNG_U32(115, 84, 69, 82) +#define png_tEXt PNG_U32(116, 69, 88, 116) +#define png_tIME PNG_U32(116, 73, 77, 69) +#define png_tRNS PNG_U32(116, 82, 78, 83) +#define png_zTXt PNG_U32(122, 84, 88, 116) + +/* The following will work on (signed char*) strings, whereas the get_uint_32 + * macro will fail on top-bit-set values because of the sign extension. + */ +#define PNG_CHUNK_FROM_STRING(s)\ + PNG_U32(0xff&(s)[0], 0xff&(s)[1], 0xff&(s)[2], 0xff&(s)[3]) + +/* This uses (char), not (png_byte) to avoid warnings on systems where (char) is + * signed and the argument is a (char[]) This macro will fail miserably on + * systems where (char) is more than 8 bits. + */ +#define PNG_STRING_FROM_CHUNK(s,c)\ + (void)(((char*)(s))[0]=(char)((c)>>24), ((char*)(s))[1]=(char)((c)>>16),\ + ((char*)(s))[2]=(char)((c)>>8), ((char*)(s))[3]=(char)((c))) + +/* Do the same but terminate with a null character. */ +#define PNG_CSTRING_FROM_CHUNK(s,c)\ + (void)(PNG_STRING_FROM_CHUNK(s,c), ((char*)(s))[4] = 0) + +/* Test on flag values as defined in the spec (section 5.4): */ +#define PNG_CHUNK_ANCILLARY(c) (1 & ((c) >> 29)) +#define PNG_CHUNK_CRITICAL(c) (!PNG_CHUNK_ANCILLARY(c)) +#define PNG_CHUNK_PRIVATE(c) (1 & ((c) >> 21)) +#define PNG_CHUNK_RESERVED(c) (1 & ((c) >> 13)) +#define PNG_CHUNK_SAFE_TO_COPY(c) (1 & ((c) >> 5)) + +/* Gamma values (new at libpng-1.5.4): */ +#define PNG_GAMMA_MAC_OLD 151724 /* Assume '1.8' is really 2.2/1.45! */ +#define PNG_GAMMA_MAC_INVERSE 65909 +#define PNG_GAMMA_sRGB_INVERSE 45455 + +/* Almost everything below is C specific; the #defines above can be used in + * non-C code (so long as it is C-preprocessed) the rest of this stuff cannot. + */ +#ifndef PNG_VERSION_INFO_ONLY + +#include "pngstruct.h" +#include "pnginfo.h" + +/* Validate the include paths - the include path used to generate pnglibconf.h + * must match that used in the build, or we must be using pnglibconf.h.prebuilt: + */ +//#if PNG_ZLIB_VERNUM != 0 && PNG_ZLIB_VERNUM != ZLIB_VERNUM +//# error ZLIB_VERNUM != PNG_ZLIB_VERNUM \ +// "-I (include path) error: see the notes in pngpriv.h" + /* This means that when pnglibconf.h was built the copy of zlib.h that it + * used is not the same as the one being used here. Because the build of + * libpng makes decisions to use inflateInit2 and inflateReset2 based on the + * zlib version number and because this affects handling of certain broken + * PNG files the -I directives must match. + * + * The most likely explanation is that you passed a -I in CFLAGS. This will + * not work; all the preprocessor directories and in particular all the -I + * directives must be in CPPFLAGS. + */ +//#endif + +/* This is used for 16 bit gamma tables -- only the top level pointers are + * const; this could be changed: + */ +typedef const png_uint_16p * png_const_uint_16pp; + +/* Added to libpng-1.5.7: sRGB conversion tables */ +#if defined(PNG_SIMPLIFIED_READ_SUPPORTED) ||\ + defined(PNG_SIMPLIFIED_WRITE_SUPPORTED) +#ifdef PNG_SIMPLIFIED_READ_SUPPORTED +PNG_INTERNAL_DATA(const png_uint_16, png_sRGB_table, [256]); + /* Convert from an sRGB encoded value 0..255 to a 16-bit linear value, + * 0..65535. This table gives the closest 16-bit answers (no errors). + */ +#endif + +PNG_INTERNAL_DATA(const png_uint_16, png_sRGB_base, [512]); +PNG_INTERNAL_DATA(const png_byte, png_sRGB_delta, [512]); + +#define PNG_sRGB_FROM_LINEAR(linear) ((png_byte)((png_sRGB_base[(linear)>>15] +\ + ((((linear)&0x7fff)*png_sRGB_delta[(linear)>>15])>>12)) >> 8)) + /* Given a value 'linear' in the range 0..255*65535 calculate the 8-bit sRGB + * encoded value with maximum error 0.646365. Note that the input is not a + * 16-bit value; it has been multiplied by 255! */ +#endif /* SIMPLIFIED_READ/WRITE */ + + +/* Inhibit C++ name-mangling for libpng functions but not for system calls. */ +#ifdef __cplusplus +extern "C" { +#endif /* __cplusplus */ + +/* Internal functions; these are not exported from a DLL however because they + * are used within several of the C source files they have to be C extern. + * + * All of these functions must be declared with PNG_INTERNAL_FUNCTION. + */ + +/* Zlib support */ +#define PNG_UNEXPECTED_ZLIB_RETURN (-7) +PNG_INTERNAL_FUNCTION(void, png_zstream_error,(png_structrp png_ptr, int ret), + PNG_EMPTY); + /* Used by the zlib handling functions to ensure that z_stream::msg is always + * set before they return. + */ + +#ifdef PNG_WRITE_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_free_buffer_list,(png_structrp png_ptr, + png_compression_bufferp *list),PNG_EMPTY); + /* Free the buffer list used by the compressed write code. */ +#endif + +#if defined(PNG_FLOATING_POINT_SUPPORTED) && \ + !defined(PNG_FIXED_POINT_MACRO_SUPPORTED) && \ + (defined(PNG_gAMA_SUPPORTED) || defined(PNG_cHRM_SUPPORTED) || \ + defined(PNG_sCAL_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED) || \ + defined(PNG_READ_RGB_TO_GRAY_SUPPORTED)) || \ + (defined(PNG_sCAL_SUPPORTED) && \ + defined(PNG_FLOATING_ARITHMETIC_SUPPORTED)) +PNG_INTERNAL_FUNCTION(png_fixed_point,png_fixed,(png_const_structrp png_ptr, + double fp, png_const_charp text),PNG_EMPTY); +#endif + +/* Check the user version string for compatibility, returns false if the version + * numbers aren't compatible. + */ +PNG_INTERNAL_FUNCTION(int,png_user_version_check,(png_structrp png_ptr, + png_const_charp user_png_ver),PNG_EMPTY); + +/* Internal base allocator - no messages, NULL on failure to allocate. This + * does, however, call the application provided allocator and that could call + * png_error (although that would be a bug in the application implementation.) + */ +PNG_INTERNAL_FUNCTION(png_voidp,png_malloc_base,(png_const_structrp png_ptr, + png_alloc_size_t size),PNG_ALLOCATED); + +#if defined(PNG_TEXT_SUPPORTED) || defined(PNG_sPLT_SUPPORTED) ||\ + defined(PNG_STORE_UNKNOWN_CHUNKS_SUPPORTED) +/* Internal array allocator, outputs no error or warning messages on failure, + * just returns NULL. + */ +PNG_INTERNAL_FUNCTION(png_voidp,png_malloc_array,(png_const_structrp png_ptr, + int nelements, size_t element_size),PNG_ALLOCATED); + +/* The same but an existing array is extended by add_elements. This function + * also memsets the new elements to 0 and copies the old elements. The old + * array is not freed or altered. + */ +PNG_INTERNAL_FUNCTION(png_voidp,png_realloc_array,(png_const_structrp png_ptr, + png_const_voidp array, int old_elements, int add_elements, + size_t element_size),PNG_ALLOCATED); +#endif /* text, sPLT or unknown chunks */ + +/* Magic to create a struct when there is no struct to call the user supplied + * memory allocators. Because error handling has not been set up the memory + * handlers can't safely call png_error, but this is an obscure and undocumented + * restriction so libpng has to assume that the 'free' handler, at least, might + * call png_error. + */ +PNG_INTERNAL_FUNCTION(png_structp,png_create_png_struct, + (png_const_charp user_png_ver, png_voidp error_ptr, png_error_ptr error_fn, + png_error_ptr warn_fn, png_voidp mem_ptr, png_malloc_ptr malloc_fn, + png_free_ptr free_fn),PNG_ALLOCATED); + +/* Free memory from internal libpng struct */ +PNG_INTERNAL_FUNCTION(void,png_destroy_png_struct,(png_structrp png_ptr), + PNG_EMPTY); + +/* Free an allocated jmp_buf (always succeeds) */ +PNG_INTERNAL_FUNCTION(void,png_free_jmpbuf,(png_structrp png_ptr),PNG_EMPTY); + +/* Function to allocate memory for zlib. PNGAPI is disallowed. */ +PNG_INTERNAL_FUNCTION(voidpf,png_zalloc,(voidpf png_ptr, uInt items, uInt size), + PNG_ALLOCATED); + +/* Function to free memory for zlib. PNGAPI is disallowed. */ +PNG_INTERNAL_FUNCTION(void,png_zfree,(voidpf png_ptr, voidpf ptr),PNG_EMPTY); + +/* Next four functions are used internally as callbacks. PNGCBAPI is required + * but not PNG_EXPORT. PNGAPI added at libpng version 1.2.3, changed to + * PNGCBAPI at 1.5.0 + */ + +PNG_INTERNAL_FUNCTION(void PNGCBAPI,png_default_read_data,(png_structp png_ptr, + png_bytep data, png_size_t length),PNG_EMPTY); + +#ifdef PNG_PROGRESSIVE_READ_SUPPORTED +PNG_INTERNAL_FUNCTION(void PNGCBAPI,png_push_fill_buffer,(png_structp png_ptr, + png_bytep buffer, png_size_t length),PNG_EMPTY); +#endif + +PNG_INTERNAL_FUNCTION(void PNGCBAPI,png_default_write_data,(png_structp png_ptr, + png_bytep data, png_size_t length),PNG_EMPTY); + +#ifdef PNG_WRITE_FLUSH_SUPPORTED +# ifdef PNG_STDIO_SUPPORTED +PNG_INTERNAL_FUNCTION(void PNGCBAPI,png_default_flush,(png_structp png_ptr), + PNG_EMPTY); +# endif +#endif + +/* Reset the CRC variable */ +PNG_INTERNAL_FUNCTION(void,png_reset_crc,(png_structrp png_ptr),PNG_EMPTY); + +/* Write the "data" buffer to whatever output you are using */ +PNG_INTERNAL_FUNCTION(void,png_write_data,(png_structrp png_ptr, + png_const_bytep data, png_size_t length),PNG_EMPTY); + +/* Read and check the PNG file signature */ +PNG_INTERNAL_FUNCTION(void,png_read_sig,(png_structrp png_ptr, + png_inforp info_ptr),PNG_EMPTY); + +/* Read the chunk header (length + type name) */ +PNG_INTERNAL_FUNCTION(png_uint_32,png_read_chunk_header,(png_structrp png_ptr), + PNG_EMPTY); + +/* Read data from whatever input you are using into the "data" buffer */ +PNG_INTERNAL_FUNCTION(void,png_read_data,(png_structrp png_ptr, png_bytep data, + png_size_t length),PNG_EMPTY); + +/* Read bytes into buf, and update png_ptr->crc */ +PNG_INTERNAL_FUNCTION(void,png_crc_read,(png_structrp png_ptr, png_bytep buf, + png_uint_32 length),PNG_EMPTY); + +/* Read "skip" bytes, read the file crc, and (optionally) verify png_ptr->crc */ +PNG_INTERNAL_FUNCTION(int,png_crc_finish,(png_structrp png_ptr, + png_uint_32 skip),PNG_EMPTY); + +/* Read the CRC from the file and compare it to the libpng calculated CRC */ +PNG_INTERNAL_FUNCTION(int,png_crc_error,(png_structrp png_ptr),PNG_EMPTY); + +/* Calculate the CRC over a section of data. Note that we are only + * passing a maximum of 64K on systems that have this as a memory limit, + * since this is the maximum buffer size we can specify. + */ +PNG_INTERNAL_FUNCTION(void,png_calculate_crc,(png_structrp png_ptr, + png_const_bytep ptr, png_size_t length),PNG_EMPTY); + +#ifdef PNG_WRITE_FLUSH_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_flush,(png_structrp png_ptr),PNG_EMPTY); +#endif + +/* Write various chunks */ + +/* Write the IHDR chunk, and update the png_struct with the necessary + * information. + */ +PNG_INTERNAL_FUNCTION(void,png_write_IHDR,(png_structrp png_ptr, + png_uint_32 width, png_uint_32 height, int bit_depth, int color_type, + int compression_method, int filter_method, int interlace_method),PNG_EMPTY); + +PNG_INTERNAL_FUNCTION(void,png_write_PLTE,(png_structrp png_ptr, + png_const_colorp palette, png_uint_32 num_pal),PNG_EMPTY); + +PNG_INTERNAL_FUNCTION(void,png_compress_IDAT,(png_structrp png_ptr, + png_const_bytep row_data, png_alloc_size_t row_data_length, int flush), + PNG_EMPTY); + +PNG_INTERNAL_FUNCTION(void,png_write_IEND,(png_structrp png_ptr),PNG_EMPTY); + +#ifdef PNG_WRITE_gAMA_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_gAMA_fixed,(png_structrp png_ptr, + png_fixed_point file_gamma),PNG_EMPTY); +#endif + +#ifdef PNG_WRITE_sBIT_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_sBIT,(png_structrp png_ptr, + png_const_color_8p sbit, int color_type),PNG_EMPTY); +#endif + +#ifdef PNG_WRITE_cHRM_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_cHRM_fixed,(png_structrp png_ptr, + const png_xy *xy), PNG_EMPTY); + /* The xy value must have been previously validated */ +#endif + +#ifdef PNG_WRITE_sRGB_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_sRGB,(png_structrp png_ptr, + int intent),PNG_EMPTY); +#endif + +#ifdef PNG_WRITE_iCCP_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_iCCP,(png_structrp png_ptr, + png_const_charp name, png_const_bytep profile), PNG_EMPTY); + /* The profile must have been previously validated for correctness, the + * length comes from the first four bytes. Only the base, deflate, + * compression is supported. + */ +#endif + +#ifdef PNG_WRITE_sPLT_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_sPLT,(png_structrp png_ptr, + png_const_sPLT_tp palette),PNG_EMPTY); +#endif + +#ifdef PNG_WRITE_tRNS_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_tRNS,(png_structrp png_ptr, + png_const_bytep trans, png_const_color_16p values, int number, + int color_type),PNG_EMPTY); +#endif + +#ifdef PNG_WRITE_bKGD_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_bKGD,(png_structrp png_ptr, + png_const_color_16p values, int color_type),PNG_EMPTY); +#endif + +#ifdef PNG_WRITE_hIST_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_hIST,(png_structrp png_ptr, + png_const_uint_16p hist, int num_hist),PNG_EMPTY); +#endif + +/* Chunks that have keywords */ +#ifdef PNG_WRITE_tEXt_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_tEXt,(png_structrp png_ptr, + png_const_charp key, png_const_charp text, png_size_t text_len),PNG_EMPTY); +#endif + +#ifdef PNG_WRITE_zTXt_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_zTXt,(png_structrp png_ptr, png_const_charp + key, png_const_charp text, int compression),PNG_EMPTY); +#endif + +#ifdef PNG_WRITE_iTXt_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_iTXt,(png_structrp png_ptr, + int compression, png_const_charp key, png_const_charp lang, + png_const_charp lang_key, png_const_charp text),PNG_EMPTY); +#endif + +#ifdef PNG_TEXT_SUPPORTED /* Added at version 1.0.14 and 1.2.4 */ +PNG_INTERNAL_FUNCTION(int,png_set_text_2,(png_const_structrp png_ptr, + png_inforp info_ptr, png_const_textp text_ptr, int num_text),PNG_EMPTY); +#endif + +#ifdef PNG_WRITE_oFFs_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_oFFs,(png_structrp png_ptr, + png_int_32 x_offset, png_int_32 y_offset, int unit_type),PNG_EMPTY); +#endif + +#ifdef PNG_WRITE_pCAL_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_pCAL,(png_structrp png_ptr, + png_charp purpose, png_int_32 X0, png_int_32 X1, int type, int nparams, + png_const_charp units, png_charpp params),PNG_EMPTY); +#endif + +#ifdef PNG_WRITE_pHYs_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_pHYs,(png_structrp png_ptr, + png_uint_32 x_pixels_per_unit, png_uint_32 y_pixels_per_unit, + int unit_type),PNG_EMPTY); +#endif + +#ifdef PNG_WRITE_tIME_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_tIME,(png_structrp png_ptr, + png_const_timep mod_time),PNG_EMPTY); +#endif + +#ifdef PNG_WRITE_sCAL_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_write_sCAL_s,(png_structrp png_ptr, + int unit, png_const_charp width, png_const_charp height),PNG_EMPTY); +#endif + +/* Called when finished processing a row of data */ +PNG_INTERNAL_FUNCTION(void,png_write_finish_row,(png_structrp png_ptr), + PNG_EMPTY); + +/* Internal use only. Called before first row of data */ +PNG_INTERNAL_FUNCTION(void,png_write_start_row,(png_structrp png_ptr), + PNG_EMPTY); + +/* Combine a row of data, dealing with alpha, etc. if requested. 'row' is an + * array of png_ptr->width pixels. If the image is not interlaced or this + * is the final pass this just does a memcpy, otherwise the "display" flag + * is used to determine whether to copy pixels that are not in the current pass. + * + * Because 'png_do_read_interlace' (below) replicates pixels this allows this + * function to achieve the documented 'blocky' appearance during interlaced read + * if display is 1 and the 'sparkle' appearance, where existing pixels in 'row' + * are not changed if they are not in the current pass, when display is 0. + * + * 'display' must be 0 or 1, otherwise the memcpy will be done regardless. + * + * The API always reads from the png_struct row buffer and always assumes that + * it is full width (png_do_read_interlace has already been called.) + * + * This function is only ever used to write to row buffers provided by the + * caller of the relevant libpng API and the row must have already been + * transformed by the read transformations. + * + * The PNG_USE_COMPILE_TIME_MASKS option causes generation of pre-computed + * bitmasks for use within the code, otherwise runtime generated masks are used. + * The default is compile time masks. + */ +#ifndef PNG_USE_COMPILE_TIME_MASKS +# define PNG_USE_COMPILE_TIME_MASKS 1 +#endif +PNG_INTERNAL_FUNCTION(void,png_combine_row,(png_const_structrp png_ptr, + png_bytep row, int display),PNG_EMPTY); + +#ifdef PNG_READ_INTERLACING_SUPPORTED +/* Expand an interlaced row: the 'row_info' describes the pass data that has + * been read in and must correspond to the pixels in 'row', the pixels are + * expanded (moved apart) in 'row' to match the final layout, when doing this + * the pixels are *replicated* to the intervening space. This is essential for + * the correct operation of png_combine_row, above. + */ +PNG_INTERNAL_FUNCTION(void,png_do_read_interlace,(png_row_infop row_info, + png_bytep row, int pass, png_uint_32 transformations),PNG_EMPTY); +#endif + +/* GRR TO DO (2.0 or whenever): simplify other internal calling interfaces */ + +#ifdef PNG_WRITE_INTERLACING_SUPPORTED +/* Grab pixels out of a row for an interlaced pass */ +PNG_INTERNAL_FUNCTION(void,png_do_write_interlace,(png_row_infop row_info, + png_bytep row, int pass),PNG_EMPTY); +#endif + +/* Unfilter a row: check the filter value before calling this, there is no point + * calling it for PNG_FILTER_VALUE_NONE. + */ +PNG_INTERNAL_FUNCTION(void,png_read_filter_row,(png_structrp pp, png_row_infop + row_info, png_bytep row, png_const_bytep prev_row, int filter),PNG_EMPTY); + +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_up_neon,(png_row_infop row_info, + png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub3_neon,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_sub4_neon,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg3_neon,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_avg4_neon,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth3_neon,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_filter_row_paeth4_neon,(png_row_infop + row_info, png_bytep row, png_const_bytep prev_row),PNG_EMPTY); + +/* Choose the best filter to use and filter the row data */ +PNG_INTERNAL_FUNCTION(void,png_write_find_filter,(png_structrp png_ptr, + png_row_infop row_info),PNG_EMPTY); + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_read_IDAT_data,(png_structrp png_ptr, + png_bytep output, png_alloc_size_t avail_out),PNG_EMPTY); + /* Read 'avail_out' bytes of data from the IDAT stream. If the output buffer + * is NULL the function checks, instead, for the end of the stream. In this + * case a benign error will be issued if the stream end is not found or if + * extra data has to be consumed. + */ +PNG_INTERNAL_FUNCTION(void,png_read_finish_IDAT,(png_structrp png_ptr), + PNG_EMPTY); + /* This cleans up when the IDAT LZ stream does not end when the last image + * byte is read; there is still some pending input. + */ + +PNG_INTERNAL_FUNCTION(void,png_read_finish_row,(png_structrp png_ptr), + PNG_EMPTY); + /* Finish a row while reading, dealing with interlacing passes, etc. */ +#endif /* SEQUENTIAL_READ */ + +/* Initialize the row buffers, etc. */ +PNG_INTERNAL_FUNCTION(void,png_read_start_row,(png_structrp png_ptr),PNG_EMPTY); + +#ifdef PNG_READ_TRANSFORMS_SUPPORTED +/* Optional call to update the users info structure */ +PNG_INTERNAL_FUNCTION(void,png_read_transform_info,(png_structrp png_ptr, + png_inforp info_ptr),PNG_EMPTY); +#endif + +/* Shared transform functions, defined in pngtran.c */ +#if defined(PNG_WRITE_FILLER_SUPPORTED) || \ + defined(PNG_READ_STRIP_ALPHA_SUPPORTED) +PNG_INTERNAL_FUNCTION(void,png_do_strip_channel,(png_row_infop row_info, + png_bytep row, int at_start),PNG_EMPTY); +#endif + +#ifdef PNG_16BIT_SUPPORTED +#if defined(PNG_READ_SWAP_SUPPORTED) || defined(PNG_WRITE_SWAP_SUPPORTED) +PNG_INTERNAL_FUNCTION(void,png_do_swap,(png_row_infop row_info, + png_bytep row),PNG_EMPTY); +#endif +#endif + +#if defined(PNG_READ_PACKSWAP_SUPPORTED) || \ + defined(PNG_WRITE_PACKSWAP_SUPPORTED) +PNG_INTERNAL_FUNCTION(void,png_do_packswap,(png_row_infop row_info, + png_bytep row),PNG_EMPTY); +#endif + +#if defined(PNG_READ_INVERT_SUPPORTED) || defined(PNG_WRITE_INVERT_SUPPORTED) +PNG_INTERNAL_FUNCTION(void,png_do_invert,(png_row_infop row_info, + png_bytep row),PNG_EMPTY); +#endif + +#if defined(PNG_READ_BGR_SUPPORTED) || defined(PNG_WRITE_BGR_SUPPORTED) +PNG_INTERNAL_FUNCTION(void,png_do_bgr,(png_row_infop row_info, + png_bytep row),PNG_EMPTY); +#endif + +/* The following decodes the appropriate chunks, and does error correction, + * then calls the appropriate callback for the chunk if it is valid. + */ + +/* Decode the IHDR chunk */ +PNG_INTERNAL_FUNCTION(void,png_handle_IHDR,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_handle_PLTE,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_handle_IEND,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); + +#ifdef PNG_READ_bKGD_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_bKGD,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_cHRM_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_cHRM,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_gAMA_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_gAMA,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_hIST_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_hIST,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_iCCP_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_iCCP,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif /* READ_iCCP */ + +#ifdef PNG_READ_iTXt_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_iTXt,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_oFFs_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_oFFs,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_pCAL_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_pCAL,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_pHYs_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_pHYs,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_sBIT_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_sBIT,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_sCAL_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_sCAL,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_sPLT_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_sPLT,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif /* READ_sPLT */ + +#ifdef PNG_READ_sRGB_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_sRGB,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_tEXt_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_tEXt,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_tIME_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_tIME,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_tRNS_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_tRNS,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +#ifdef PNG_READ_zTXt_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_handle_zTXt,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +#endif + +PNG_INTERNAL_FUNCTION(void,png_check_chunk_name,(png_structrp png_ptr, + png_uint_32 chunk_name),PNG_EMPTY); + +PNG_INTERNAL_FUNCTION(void,png_handle_unknown,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length, int keep),PNG_EMPTY); + /* This is the function that gets called for unknown chunks. The 'keep' + * argument is either non-zero for a known chunk that has been set to be + * handled as unknown or zero for an unknown chunk. By default the function + * just skips the chunk or errors out if it is critical. + */ + +#if defined(PNG_READ_UNKNOWN_CHUNKS_SUPPORTED) ||\ + defined(PNG_HANDLE_AS_UNKNOWN_SUPPORTED) +PNG_INTERNAL_FUNCTION(int,png_chunk_unknown_handling, + (png_const_structrp png_ptr, png_uint_32 chunk_name),PNG_EMPTY); + /* Exactly as the API png_handle_as_unknown() except that the argument is a + * 32-bit chunk name, not a string. + */ +#endif /* READ_UNKNOWN_CHUNKS || HANDLE_AS_UNKNOWN */ + +/* Handle the transformations for reading and writing */ +#ifdef PNG_READ_TRANSFORMS_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_do_read_transformations,(png_structrp png_ptr, + png_row_infop row_info),PNG_EMPTY); +#endif +#ifdef PNG_WRITE_TRANSFORMS_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_do_write_transformations,(png_structrp png_ptr, + png_row_infop row_info),PNG_EMPTY); +#endif + +#ifdef PNG_READ_TRANSFORMS_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_init_read_transformations,(png_structrp png_ptr), + PNG_EMPTY); +#endif + +#ifdef PNG_PROGRESSIVE_READ_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_push_read_chunk,(png_structrp png_ptr, + png_inforp info_ptr),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_read_sig,(png_structrp png_ptr, + png_inforp info_ptr),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_check_crc,(png_structrp png_ptr),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_crc_skip,(png_structrp png_ptr, + png_uint_32 length),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_crc_finish,(png_structrp png_ptr), + PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_save_buffer,(png_structrp png_ptr), + PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_restore_buffer,(png_structrp png_ptr, + png_bytep buffer, png_size_t buffer_length),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_read_IDAT,(png_structrp png_ptr),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_process_IDAT_data,(png_structrp png_ptr, + png_bytep buffer, png_size_t buffer_length),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_process_row,(png_structrp png_ptr), + PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_handle_unknown,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_have_info,(png_structrp png_ptr, + png_inforp info_ptr),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_have_end,(png_structrp png_ptr, + png_inforp info_ptr),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_have_row,(png_structrp png_ptr, + png_bytep row),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_read_end,(png_structrp png_ptr, + png_inforp info_ptr),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_process_some_data,(png_structrp png_ptr, + png_inforp info_ptr),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_read_push_finish_row,(png_structrp png_ptr), + PNG_EMPTY); +# ifdef PNG_READ_tEXt_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_push_handle_tEXt,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_read_tEXt,(png_structrp png_ptr, + png_inforp info_ptr),PNG_EMPTY); +# endif +# ifdef PNG_READ_zTXt_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_push_handle_zTXt,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_read_zTXt,(png_structrp png_ptr, + png_inforp info_ptr),PNG_EMPTY); +# endif +# ifdef PNG_READ_iTXt_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_push_handle_iTXt,(png_structrp png_ptr, + png_inforp info_ptr, png_uint_32 length),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_push_read_iTXt,(png_structrp png_ptr, + png_inforp info_ptr),PNG_EMPTY); +# endif + +#endif /* PROGRESSIVE_READ */ + +/* Added at libpng version 1.6.0 */ +#ifdef PNG_GAMMA_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_colorspace_set_gamma,(png_const_structrp png_ptr, + png_colorspacerp colorspace, png_fixed_point gAMA), PNG_EMPTY); + /* Set the colorspace gamma with a value provided by the application or by + * the gAMA chunk on read. The value will override anything set by an ICC + * profile. + */ + +PNG_INTERNAL_FUNCTION(void,png_colorspace_sync_info,(png_const_structrp png_ptr, + png_inforp info_ptr), PNG_EMPTY); + /* Synchronize the info 'valid' flags with the colorspace */ + +PNG_INTERNAL_FUNCTION(void,png_colorspace_sync,(png_const_structrp png_ptr, + png_inforp info_ptr), PNG_EMPTY); + /* Copy the png_struct colorspace to the info_struct and call the above to + * synchronize the flags. Checks for NULL info_ptr and does nothing. + */ +#endif + +/* Added at libpng version 1.4.0 */ +#ifdef PNG_COLORSPACE_SUPPORTED +/* These internal functions are for maintaining the colorspace structure within + * a png_info or png_struct (or, indeed, both). + */ +PNG_INTERNAL_FUNCTION(int,png_colorspace_set_chromaticities, + (png_const_structrp png_ptr, png_colorspacerp colorspace, const png_xy *xy, + int preferred), PNG_EMPTY); + +PNG_INTERNAL_FUNCTION(int,png_colorspace_set_endpoints, + (png_const_structrp png_ptr, png_colorspacerp colorspace, const png_XYZ *XYZ, + int preferred), PNG_EMPTY); + +#ifdef PNG_sRGB_SUPPORTED +PNG_INTERNAL_FUNCTION(int,png_colorspace_set_sRGB,(png_const_structrp png_ptr, + png_colorspacerp colorspace, int intent), PNG_EMPTY); + /* This does set the colorspace gAMA and cHRM values too, but doesn't set the + * flags to write them, if it returns false there was a problem and an error + * message has already been output (but the colorspace may still need to be + * synced to record the invalid flag). + */ +#endif /* sRGB */ + +#ifdef PNG_iCCP_SUPPORTED +PNG_INTERNAL_FUNCTION(int,png_colorspace_set_ICC,(png_const_structrp png_ptr, + png_colorspacerp colorspace, png_const_charp name, + png_uint_32 profile_length, png_const_bytep profile, int color_type), + PNG_EMPTY); + /* The 'name' is used for information only */ + +/* Routines for checking parts of an ICC profile. */ +PNG_INTERNAL_FUNCTION(int,png_icc_check_length,(png_const_structrp png_ptr, + png_colorspacerp colorspace, png_const_charp name, + png_uint_32 profile_length), PNG_EMPTY); +PNG_INTERNAL_FUNCTION(int,png_icc_check_header,(png_const_structrp png_ptr, + png_colorspacerp colorspace, png_const_charp name, + png_uint_32 profile_length, + png_const_bytep profile /* first 132 bytes only */, int color_type), + PNG_EMPTY); +PNG_INTERNAL_FUNCTION(int,png_icc_check_tag_table,(png_const_structrp png_ptr, + png_colorspacerp colorspace, png_const_charp name, + png_uint_32 profile_length, + png_const_bytep profile /* header plus whole tag table */), PNG_EMPTY); +#ifdef PNG_sRGB_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_icc_set_sRGB,( + png_const_structrp png_ptr, png_colorspacerp colorspace, + png_const_bytep profile, uLong adler), PNG_EMPTY); + /* 'adler' is the Adler32 checksum of the uncompressed profile data. It may + * be zero to indicate that it is not available. It is used, if provided, + * as a fast check on the profile when checking to see if it is sRGB. + */ +#endif +#endif /* iCCP */ + +#ifdef PNG_READ_RGB_TO_GRAY_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_colorspace_set_rgb_coefficients, + (png_structrp png_ptr), PNG_EMPTY); + /* Set the rgb_to_gray coefficients from the colorspace Y values */ +#endif /* READ_RGB_TO_GRAY */ +#endif /* COLORSPACE */ + +/* Added at libpng version 1.4.0 */ +PNG_INTERNAL_FUNCTION(void,png_check_IHDR,(png_const_structrp png_ptr, + png_uint_32 width, png_uint_32 height, int bit_depth, + int color_type, int interlace_type, int compression_type, + int filter_type),PNG_EMPTY); + +/* Added at libpng version 1.5.10 */ +#if defined(PNG_READ_CHECK_FOR_INVALID_INDEX_SUPPORTED) || \ + defined(PNG_WRITE_CHECK_FOR_INVALID_INDEX_SUPPORTED) +PNG_INTERNAL_FUNCTION(void,png_do_check_palette_indexes, + (png_structrp png_ptr, png_row_infop row_info),PNG_EMPTY); +#endif + +#if defined(PNG_FLOATING_POINT_SUPPORTED) && defined(PNG_ERROR_TEXT_SUPPORTED) +PNG_INTERNAL_FUNCTION(void,png_fixed_error,(png_const_structrp png_ptr, + png_const_charp name),PNG_NORETURN); +#endif + +/* Puts 'string' into 'buffer' at buffer[pos], taking care never to overwrite + * the end. Always leaves the buffer nul terminated. Never errors out (and + * there is no error code.) + */ +PNG_INTERNAL_FUNCTION(size_t,png_safecat,(png_charp buffer, size_t bufsize, + size_t pos, png_const_charp string),PNG_EMPTY); + +/* Various internal functions to handle formatted warning messages, currently + * only implemented for warnings. + */ +#if defined(PNG_WARNINGS_SUPPORTED) || defined(PNG_TIME_RFC1123_SUPPORTED) +/* Utility to dump an unsigned value into a buffer, given a start pointer and + * and end pointer (which should point just *beyond* the end of the buffer!) + * Returns the pointer to the start of the formatted string. This utility only + * does unsigned values. + */ +PNG_INTERNAL_FUNCTION(png_charp,png_format_number,(png_const_charp start, + png_charp end, int format, png_alloc_size_t number),PNG_EMPTY); + +/* Convenience macro that takes an array: */ +#define PNG_FORMAT_NUMBER(buffer,format,number) \ + png_format_number(buffer, buffer + (sizeof buffer), format, number) + +/* Suggested size for a number buffer (enough for 64 bits and a sign!) */ +#define PNG_NUMBER_BUFFER_SIZE 24 + +/* These are the integer formats currently supported, the name is formed from + * the standard printf(3) format string. + */ +#define PNG_NUMBER_FORMAT_u 1 /* chose unsigned API! */ +#define PNG_NUMBER_FORMAT_02u 2 +#define PNG_NUMBER_FORMAT_d 1 /* chose signed API! */ +#define PNG_NUMBER_FORMAT_02d 2 +#define PNG_NUMBER_FORMAT_x 3 +#define PNG_NUMBER_FORMAT_02x 4 +#define PNG_NUMBER_FORMAT_fixed 5 /* choose the signed API */ +#endif + +#ifdef PNG_WARNINGS_SUPPORTED +/* New defines and members adding in libpng-1.5.4 */ +# define PNG_WARNING_PARAMETER_SIZE 32 +# define PNG_WARNING_PARAMETER_COUNT 8 /* Maximum 9; see pngerror.c */ + +/* An l-value of this type has to be passed to the APIs below to cache the + * values of the parameters to a formatted warning message. + */ +typedef char png_warning_parameters[PNG_WARNING_PARAMETER_COUNT][ + PNG_WARNING_PARAMETER_SIZE]; + +PNG_INTERNAL_FUNCTION(void,png_warning_parameter,(png_warning_parameters p, + int number, png_const_charp string),PNG_EMPTY); + /* Parameters are limited in size to PNG_WARNING_PARAMETER_SIZE characters, + * including the trailing '\0'. + */ +PNG_INTERNAL_FUNCTION(void,png_warning_parameter_unsigned, + (png_warning_parameters p, int number, int format, png_alloc_size_t value), + PNG_EMPTY); + /* Use png_alloc_size_t because it is an unsigned type as big as any we + * need to output. Use the following for a signed value. + */ +PNG_INTERNAL_FUNCTION(void,png_warning_parameter_signed, + (png_warning_parameters p, int number, int format, png_int_32 value), + PNG_EMPTY); + +PNG_INTERNAL_FUNCTION(void,png_formatted_warning,(png_const_structrp png_ptr, + png_warning_parameters p, png_const_charp message),PNG_EMPTY); + /* 'message' follows the X/Open approach of using @1, @2 to insert + * parameters previously supplied using the above functions. Errors in + * specifying the parameters will simply result in garbage substitutions. + */ +#endif + +#ifdef PNG_BENIGN_ERRORS_SUPPORTED +/* Application errors (new in 1.6); use these functions (declared below) for + * errors in the parameters or order of API function calls on read. The + * 'warning' should be used for an error that can be handled completely; the + * 'error' for one which can be handled safely but which may lose application + * information or settings. + * + * By default these both result in a png_error call prior to release, while in a + * released version the 'warning' is just a warning. However if the application + * explicitly disables benign errors (explicitly permitting the code to lose + * information) they both turn into warnings. + * + * If benign errors aren't supported they end up as the corresponding base call + * (png_warning or png_error.) + */ +PNG_INTERNAL_FUNCTION(void,png_app_warning,(png_const_structrp png_ptr, + png_const_charp message),PNG_EMPTY); + /* The application provided invalid parameters to an API function or called + * an API function at the wrong time, libpng can completely recover. + */ + +PNG_INTERNAL_FUNCTION(void,png_app_error,(png_const_structrp png_ptr, + png_const_charp message),PNG_EMPTY); + /* As above but libpng will ignore the call, or attempt some other partial + * recovery from the error. + */ +#else +# define png_app_warning(pp,s) png_warning(pp,s) +# define png_app_error(pp,s) png_error(pp,s) +#endif + +PNG_INTERNAL_FUNCTION(void,png_chunk_report,(png_const_structrp png_ptr, + png_const_charp message, int error),PNG_EMPTY); + /* Report a recoverable issue in chunk data. On read this is used to report + * a problem found while reading a particular chunk and the + * png_chunk_benign_error or png_chunk_warning function is used as + * appropriate. On write this is used to report an error that comes from + * data set via an application call to a png_set_ API and png_app_error or + * png_app_warning is used as appropriate. + * + * The 'error' parameter must have one of the following values: + */ +#define PNG_CHUNK_WARNING 0 /* never an error */ +#define PNG_CHUNK_WRITE_ERROR 1 /* an error only on write */ +#define PNG_CHUNK_ERROR 2 /* always an error */ + +/* ASCII to FP interfaces, currently only implemented if sCAL + * support is required. + */ +#if defined(PNG_sCAL_SUPPORTED) +/* MAX_DIGITS is actually the maximum number of characters in an sCAL + * width or height, derived from the precision (number of significant + * digits - a build time settable option) and assumptions about the + * maximum ridiculous exponent. + */ +#define PNG_sCAL_MAX_DIGITS (PNG_sCAL_PRECISION+1/*.*/+1/*E*/+10/*exponent*/) + +#ifdef PNG_FLOATING_POINT_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_ascii_from_fp,(png_const_structrp png_ptr, + png_charp ascii, png_size_t size, double fp, unsigned int precision), + PNG_EMPTY); +#endif /* FLOATING_POINT */ + +#ifdef PNG_FIXED_POINT_SUPPORTED +PNG_INTERNAL_FUNCTION(void,png_ascii_from_fixed,(png_const_structrp png_ptr, + png_charp ascii, png_size_t size, png_fixed_point fp),PNG_EMPTY); +#endif /* FIXED_POINT */ +#endif /* sCAL */ + +#if defined(PNG_sCAL_SUPPORTED) || defined(PNG_pCAL_SUPPORTED) +/* An internal API to validate the format of a floating point number. + * The result is the index of the next character. If the number is + * not valid it will be the index of a character in the supposed number. + * + * The format of a number is defined in the PNG extensions specification + * and this API is strictly conformant to that spec, not anyone elses! + * + * The format as a regular expression is: + * + * [+-]?[0-9]+.?([Ee][+-]?[0-9]+)? + * + * or: + * + * [+-]?.[0-9]+(.[0-9]+)?([Ee][+-]?[0-9]+)? + * + * The complexity is that either integer or fraction must be present and the + * fraction is permitted to have no digits only if the integer is present. + * + * NOTE: The dangling E problem. + * There is a PNG valid floating point number in the following: + * + * PNG floating point numbers are not greedy. + * + * Working this out requires *TWO* character lookahead (because of the + * sign), the parser does not do this - it will fail at the 'r' - this + * doesn't matter for PNG sCAL chunk values, but it requires more care + * if the value were ever to be embedded in something more complex. Use + * ANSI-C strtod if you need the lookahead. + */ +/* State table for the parser. */ +#define PNG_FP_INTEGER 0 /* before or in integer */ +#define PNG_FP_FRACTION 1 /* before or in fraction */ +#define PNG_FP_EXPONENT 2 /* before or in exponent */ +#define PNG_FP_STATE 3 /* mask for the above */ +#define PNG_FP_SAW_SIGN 4 /* Saw +/- in current state */ +#define PNG_FP_SAW_DIGIT 8 /* Saw a digit in current state */ +#define PNG_FP_SAW_DOT 16 /* Saw a dot in current state */ +#define PNG_FP_SAW_E 32 /* Saw an E (or e) in current state */ +#define PNG_FP_SAW_ANY 60 /* Saw any of the above 4 */ + +/* These three values don't affect the parser. They are set but not used. + */ +#define PNG_FP_WAS_VALID 64 /* Preceding substring is a valid fp number */ +#define PNG_FP_NEGATIVE 128 /* A negative number, including "-0" */ +#define PNG_FP_NONZERO 256 /* A non-zero value */ +#define PNG_FP_STICKY 448 /* The above three flags */ + +/* This is available for the caller to store in 'state' if required. Do not + * call the parser after setting it (the parser sometimes clears it.) + */ +#define PNG_FP_INVALID 512 /* Available for callers as a distinct value */ + +/* Result codes for the parser (boolean - true meants ok, false means + * not ok yet.) + */ +#define PNG_FP_MAYBE 0 /* The number may be valid in the future */ +#define PNG_FP_OK 1 /* The number is valid */ + +/* Tests on the sticky non-zero and negative flags. To pass these checks + * the state must also indicate that the whole number is valid - this is + * achieved by testing PNG_FP_SAW_DIGIT (see the implementation for why this + * is equivalent to PNG_FP_OK above.) + */ +#define PNG_FP_NZ_MASK (PNG_FP_SAW_DIGIT | PNG_FP_NEGATIVE | PNG_FP_NONZERO) + /* NZ_MASK: the string is valid and a non-zero negative value */ +#define PNG_FP_Z_MASK (PNG_FP_SAW_DIGIT | PNG_FP_NONZERO) + /* Z MASK: the string is valid and a non-zero value. */ + /* PNG_FP_SAW_DIGIT: the string is valid. */ +#define PNG_FP_IS_ZERO(state) (((state) & PNG_FP_Z_MASK) == PNG_FP_SAW_DIGIT) +#define PNG_FP_IS_POSITIVE(state) (((state) & PNG_FP_NZ_MASK) == PNG_FP_Z_MASK) +#define PNG_FP_IS_NEGATIVE(state) (((state) & PNG_FP_NZ_MASK) == PNG_FP_NZ_MASK) + +/* The actual parser. This can be called repeatedly. It updates + * the index into the string and the state variable (which must + * be initialized to 0). It returns a result code, as above. There + * is no point calling the parser any more if it fails to advance to + * the end of the string - it is stuck on an invalid character (or + * terminated by '\0'). + * + * Note that the pointer will consume an E or even an E+ and then leave + * a 'maybe' state even though a preceding integer.fraction is valid. + * The PNG_FP_WAS_VALID flag indicates that a preceding substring was + * a valid number. It's possible to recover from this by calling + * the parser again (from the start, with state 0) but with a string + * that omits the last character (i.e. set the size to the index of + * the problem character.) This has not been tested within libpng. + */ +PNG_INTERNAL_FUNCTION(int,png_check_fp_number,(png_const_charp string, + png_size_t size, int *statep, png_size_tp whereami),PNG_EMPTY); + +/* This is the same but it checks a complete string and returns true + * only if it just contains a floating point number. As of 1.5.4 this + * function also returns the state at the end of parsing the number if + * it was valid (otherwise it returns 0.) This can be used for testing + * for negative or zero values using the sticky flag. + */ +PNG_INTERNAL_FUNCTION(int,png_check_fp_string,(png_const_charp string, + png_size_t size),PNG_EMPTY); +#endif /* pCAL || sCAL */ + +#if defined(PNG_GAMMA_SUPPORTED) ||\ + defined(PNG_INCH_CONVERSIONS_SUPPORTED) || defined(PNG_READ_pHYs_SUPPORTED) +/* Added at libpng version 1.5.0 */ +/* This is a utility to provide a*times/div (rounded) and indicate + * if there is an overflow. The result is a boolean - false (0) + * for overflow, true (1) if no overflow, in which case *res + * holds the result. + */ +PNG_INTERNAL_FUNCTION(int,png_muldiv,(png_fixed_point_p res, png_fixed_point a, + png_int_32 multiplied_by, png_int_32 divided_by),PNG_EMPTY); +#endif + +#if defined(PNG_READ_GAMMA_SUPPORTED) || defined(PNG_INCH_CONVERSIONS_SUPPORTED) +/* Same deal, but issue a warning on overflow and return 0. */ +PNG_INTERNAL_FUNCTION(png_fixed_point,png_muldiv_warn, + (png_const_structrp png_ptr, png_fixed_point a, png_int_32 multiplied_by, + png_int_32 divided_by),PNG_EMPTY); +#endif + +#ifdef PNG_GAMMA_SUPPORTED +/* Calculate a reciprocal - used for gamma values. This returns + * 0 if the argument is 0 in order to maintain an undefined value; + * there are no warnings. + */ +PNG_INTERNAL_FUNCTION(png_fixed_point,png_reciprocal,(png_fixed_point a), + PNG_EMPTY); + +#ifdef PNG_READ_GAMMA_SUPPORTED +/* The same but gives a reciprocal of the product of two fixed point + * values. Accuracy is suitable for gamma calculations but this is + * not exact - use png_muldiv for that. Only required at present on read. + */ +PNG_INTERNAL_FUNCTION(png_fixed_point,png_reciprocal2,(png_fixed_point a, + png_fixed_point b),PNG_EMPTY); +#endif + +/* Return true if the gamma value is significantly different from 1.0 */ +PNG_INTERNAL_FUNCTION(int,png_gamma_significant,(png_fixed_point gamma_value), + PNG_EMPTY); +#endif + +#ifdef PNG_READ_GAMMA_SUPPORTED +/* Internal fixed point gamma correction. These APIs are called as + * required to convert single values - they don't need to be fast, + * they are not used when processing image pixel values. + * + * While the input is an 'unsigned' value it must actually be the + * correct bit value - 0..255 or 0..65535 as required. + */ +PNG_INTERNAL_FUNCTION(png_uint_16,png_gamma_correct,(png_structrp png_ptr, + unsigned int value, png_fixed_point gamma_value),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(png_uint_16,png_gamma_16bit_correct,(unsigned int value, + png_fixed_point gamma_value),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(png_byte,png_gamma_8bit_correct,(unsigned int value, + png_fixed_point gamma_value),PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_destroy_gamma_table,(png_structrp png_ptr), + PNG_EMPTY); +PNG_INTERNAL_FUNCTION(void,png_build_gamma_table,(png_structrp png_ptr, + int bit_depth),PNG_EMPTY); +#endif + +/* SIMPLIFIED READ/WRITE SUPPORT */ +#if defined(PNG_SIMPLIFIED_READ_SUPPORTED) ||\ + defined(PNG_SIMPLIFIED_WRITE_SUPPORTED) +/* The internal structure that png_image::opaque points to. */ +typedef struct png_control +{ + png_structp png_ptr; + png_infop info_ptr; + png_voidp error_buf; /* Always a jmp_buf at present. */ + + png_const_bytep memory; /* Memory buffer. */ + png_size_t size; /* Size of the memory buffer. */ + + unsigned int for_write :1; /* Otherwise it is a read structure */ + unsigned int owned_file :1; /* We own the file in io_ptr */ +} png_control; + +/* Return the pointer to the jmp_buf from a png_control: necessary because C + * does not reveal the type of the elements of jmp_buf. + */ +#ifdef __cplusplus +# define png_control_jmp_buf(pc) (((jmp_buf*)((pc)->error_buf))[0]) +#else +# define png_control_jmp_buf(pc) ((pc)->error_buf) +#endif + +/* Utility to safely execute a piece of libpng code catching and logging any + * errors that might occur. Returns true on success, false on failure (either + * of the function or as a result of a png_error.) + */ +PNG_INTERNAL_CALLBACK(void,png_safe_error,(png_structp png_ptr, + png_const_charp error_message),PNG_NORETURN); + +#ifdef PNG_WARNINGS_SUPPORTED +PNG_INTERNAL_CALLBACK(void,png_safe_warning,(png_structp png_ptr, + png_const_charp warning_message),PNG_EMPTY); +#else +# define png_safe_warning 0/*dummy argument*/ +#endif + +PNG_INTERNAL_FUNCTION(int,png_safe_execute,(png_imagep image, + int (*function)(png_voidp), png_voidp arg),PNG_EMPTY); + +/* Utility to log an error; this also cleans up the png_image; the function + * always returns 0 (false). + */ +PNG_INTERNAL_FUNCTION(int,png_image_error,(png_imagep image, + png_const_charp error_message),PNG_EMPTY); + +#ifndef PNG_SIMPLIFIED_READ_SUPPORTED +/* png_image_free is used by the write code but not exported */ +PNG_INTERNAL_FUNCTION(void, png_image_free, (png_imagep image), PNG_EMPTY); +#endif /* !SIMPLIFIED_READ */ + +#endif /* SIMPLIFIED READ/WRITE */ + +/* These are initialization functions for hardware specific PNG filter + * optimizations; list these here then select the appropriate one at compile + * time using the macro PNG_FILTER_OPTIMIZATIONS. If the macro is not defined + * the generic code is used. + */ +#ifdef PNG_FILTER_OPTIMIZATIONS +PNG_INTERNAL_FUNCTION(void, PNG_FILTER_OPTIMIZATIONS, (png_structp png_ptr, + unsigned int bpp), PNG_EMPTY); + /* Just declare the optimization that will be used */ +#else + /* List *all* the possible optimizations here - this branch is required if + * the builder of libpng passes the definition of PNG_FILTER_OPTIMIZATIONS in + * CFLAGS in place of CPPFLAGS *and* uses symbol prefixing. + */ +PNG_INTERNAL_FUNCTION(void, png_init_filter_functions_neon, + (png_structp png_ptr, unsigned int bpp), PNG_EMPTY); +#endif + +/* Maintainer: Put new private prototypes here ^ */ + +#include "pngdebug.h" + +#ifdef __cplusplus +} +#endif + +#endif /* PNG_VERSION_INFO_ONLY */ +#endif /* PNGPRIV_H */ diff --git a/app/src/main/jni/libpng/pngread.c b/app/src/main/jni/libpng/pngread.c new file mode 100644 index 000000000..87a789157 --- /dev/null +++ b/app/src/main/jni/libpng/pngread.c @@ -0,0 +1,4121 @@ + +/* pngread.c - read a PNG file + * + * Last changed in libpng 1.6.15 [November 20, 2014] + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + * + * This file contains routines that an application calls directly to + * read a PNG file or stream. + */ + +#include "pngpriv.h" +#if defined(PNG_SIMPLIFIED_READ_SUPPORTED) && defined(PNG_STDIO_SUPPORTED) +# include +#endif + +#ifdef PNG_READ_SUPPORTED + +/* Create a PNG structure for reading, and allocate any memory needed. */ +PNG_FUNCTION(png_structp,PNGAPI +png_create_read_struct,(png_const_charp user_png_ver, png_voidp error_ptr, + png_error_ptr error_fn, png_error_ptr warn_fn),PNG_ALLOCATED) +{ +#ifndef PNG_USER_MEM_SUPPORTED + png_structp png_ptr = png_create_png_struct(user_png_ver, error_ptr, + error_fn, warn_fn, NULL, NULL, NULL); +#else + return png_create_read_struct_2(user_png_ver, error_ptr, error_fn, + warn_fn, NULL, NULL, NULL); +} + +/* Alternate create PNG structure for reading, and allocate any memory + * needed. + */ +PNG_FUNCTION(png_structp,PNGAPI +png_create_read_struct_2,(png_const_charp user_png_ver, png_voidp error_ptr, + png_error_ptr error_fn, png_error_ptr warn_fn, png_voidp mem_ptr, + png_malloc_ptr malloc_fn, png_free_ptr free_fn),PNG_ALLOCATED) +{ + png_structp png_ptr = png_create_png_struct(user_png_ver, error_ptr, + error_fn, warn_fn, mem_ptr, malloc_fn, free_fn); +#endif /* USER_MEM */ + + if (png_ptr != NULL) + { + png_ptr->mode = PNG_IS_READ_STRUCT; + + /* Added in libpng-1.6.0; this can be used to detect a read structure if + * required (it will be zero in a write structure.) + */ +# ifdef PNG_SEQUENTIAL_READ_SUPPORTED + png_ptr->IDAT_read_size = PNG_IDAT_READ_SIZE; +# endif + +# ifdef PNG_BENIGN_READ_ERRORS_SUPPORTED + png_ptr->flags |= PNG_FLAG_BENIGN_ERRORS_WARN; + + /* In stable builds only warn if an application error can be completely + * handled. + */ +# if PNG_LIBPNG_BUILD_BASE_TYPE >= PNG_LIBPNG_BUILD_RC + png_ptr->flags |= PNG_FLAG_APP_WARNINGS_WARN; +# endif +# endif + + /* TODO: delay this, it can be done in png_init_io (if the app doesn't + * do it itself) avoiding setting the default function if it is not + * required. + */ + png_set_read_fn(png_ptr, NULL, NULL); + } + + return png_ptr; +} + + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +/* Read the information before the actual image data. This has been + * changed in v0.90 to allow reading a file that already has the magic + * bytes read from the stream. You can tell libpng how many bytes have + * been read from the beginning of the stream (up to the maximum of 8) + * via png_set_sig_bytes(), and we will only check the remaining bytes + * here. The application can then have access to the signature bytes we + * read if it is determined that this isn't a valid PNG file. + */ +void PNGAPI +png_read_info(png_structrp png_ptr, png_inforp info_ptr) +{ +#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED + int keep; +#endif + + png_debug(1, "in png_read_info"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + /* Read and check the PNG file signature. */ + png_read_sig(png_ptr, info_ptr); + + for (;;) + { + png_uint_32 length = png_read_chunk_header(png_ptr); + png_uint_32 chunk_name = png_ptr->chunk_name; + + /* IDAT logic needs to happen here to simplify getting the two flags + * right. + */ + if (chunk_name == png_IDAT) + { + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "Missing IHDR before IDAT"); + + else if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE && + (png_ptr->mode & PNG_HAVE_PLTE) == 0) + png_chunk_error(png_ptr, "Missing PLTE before IDAT"); + + else if ((png_ptr->mode & PNG_AFTER_IDAT) != 0) + png_chunk_benign_error(png_ptr, "Too many IDATs found"); + + png_ptr->mode |= PNG_HAVE_IDAT; + } + + else if ((png_ptr->mode & PNG_HAVE_IDAT) != 0) + png_ptr->mode |= PNG_AFTER_IDAT; + + /* This should be a binary subdivision search or a hash for + * matching the chunk name rather than a linear search. + */ + if (chunk_name == png_IHDR) + png_handle_IHDR(png_ptr, info_ptr, length); + + else if (chunk_name == png_IEND) + png_handle_IEND(png_ptr, info_ptr, length); + +#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED + else if ((keep = png_chunk_unknown_handling(png_ptr, chunk_name)) != 0) + { + png_handle_unknown(png_ptr, info_ptr, length, keep); + + if (chunk_name == png_PLTE) + png_ptr->mode |= PNG_HAVE_PLTE; + + else if (chunk_name == png_IDAT) + { + png_ptr->idat_size = 0; /* It has been consumed */ + break; + } + } +#endif + else if (chunk_name == png_PLTE) + png_handle_PLTE(png_ptr, info_ptr, length); + + else if (chunk_name == png_IDAT) + { + png_ptr->idat_size = length; + break; + } + +#ifdef PNG_READ_bKGD_SUPPORTED + else if (chunk_name == png_bKGD) + png_handle_bKGD(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_cHRM_SUPPORTED + else if (chunk_name == png_cHRM) + png_handle_cHRM(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_gAMA_SUPPORTED + else if (chunk_name == png_gAMA) + png_handle_gAMA(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_hIST_SUPPORTED + else if (chunk_name == png_hIST) + png_handle_hIST(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_oFFs_SUPPORTED + else if (chunk_name == png_oFFs) + png_handle_oFFs(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_pCAL_SUPPORTED + else if (chunk_name == png_pCAL) + png_handle_pCAL(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_sCAL_SUPPORTED + else if (chunk_name == png_sCAL) + png_handle_sCAL(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_pHYs_SUPPORTED + else if (chunk_name == png_pHYs) + png_handle_pHYs(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_sBIT_SUPPORTED + else if (chunk_name == png_sBIT) + png_handle_sBIT(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_sRGB_SUPPORTED + else if (chunk_name == png_sRGB) + png_handle_sRGB(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_iCCP_SUPPORTED + else if (chunk_name == png_iCCP) + png_handle_iCCP(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_sPLT_SUPPORTED + else if (chunk_name == png_sPLT) + png_handle_sPLT(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_tEXt_SUPPORTED + else if (chunk_name == png_tEXt) + png_handle_tEXt(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_tIME_SUPPORTED + else if (chunk_name == png_tIME) + png_handle_tIME(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_tRNS_SUPPORTED + else if (chunk_name == png_tRNS) + png_handle_tRNS(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_zTXt_SUPPORTED + else if (chunk_name == png_zTXt) + png_handle_zTXt(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_iTXt_SUPPORTED + else if (chunk_name == png_iTXt) + png_handle_iTXt(png_ptr, info_ptr, length); +#endif + + else + png_handle_unknown(png_ptr, info_ptr, length, + PNG_HANDLE_CHUNK_AS_DEFAULT); + } +} +#endif /* SEQUENTIAL_READ */ + +/* Optional call to update the users info_ptr structure */ +void PNGAPI +png_read_update_info(png_structrp png_ptr, png_inforp info_ptr) +{ + png_debug(1, "in png_read_update_info"); + + if (png_ptr != NULL) + { + if ((png_ptr->flags & PNG_FLAG_ROW_INIT) == 0) + { + png_read_start_row(png_ptr); + +# ifdef PNG_READ_TRANSFORMS_SUPPORTED + png_read_transform_info(png_ptr, info_ptr); +# else + PNG_UNUSED(info_ptr) +# endif + } + + /* New in 1.6.0 this avoids the bug of doing the initializations twice */ + else + png_app_error(png_ptr, + "png_read_update_info/png_start_read_image: duplicate call"); + } +} + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +/* Initialize palette, background, etc, after transformations + * are set, but before any reading takes place. This allows + * the user to obtain a gamma-corrected palette, for example. + * If the user doesn't call this, we will do it ourselves. + */ +void PNGAPI +png_start_read_image(png_structrp png_ptr) +{ + png_debug(1, "in png_start_read_image"); + + if (png_ptr != NULL) + { + if ((png_ptr->flags & PNG_FLAG_ROW_INIT) == 0) + png_read_start_row(png_ptr); + + /* New in 1.6.0 this avoids the bug of doing the initializations twice */ + else + png_app_error(png_ptr, + "png_start_read_image/png_read_update_info: duplicate call"); + } +} +#endif /* SEQUENTIAL_READ */ + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +#ifdef PNG_MNG_FEATURES_SUPPORTED +/* Undoes intrapixel differencing, + * NOTE: this is apparently only supported in the 'sequential' reader. + */ +static void +png_do_read_intrapixel(png_row_infop row_info, png_bytep row) +{ + png_debug(1, "in png_do_read_intrapixel"); + + if ( + (row_info->color_type & PNG_COLOR_MASK_COLOR) != 0) + { + int bytes_per_pixel; + png_uint_32 row_width = row_info->width; + + if (row_info->bit_depth == 8) + { + png_bytep rp; + png_uint_32 i; + + if (row_info->color_type == PNG_COLOR_TYPE_RGB) + bytes_per_pixel = 3; + + else if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA) + bytes_per_pixel = 4; + + else + return; + + for (i = 0, rp = row; i < row_width; i++, rp += bytes_per_pixel) + { + *(rp) = (png_byte)((256 + *rp + *(rp + 1)) & 0xff); + *(rp+2) = (png_byte)((256 + *(rp + 2) + *(rp + 1)) & 0xff); + } + } + else if (row_info->bit_depth == 16) + { + png_bytep rp; + png_uint_32 i; + + if (row_info->color_type == PNG_COLOR_TYPE_RGB) + bytes_per_pixel = 6; + + else if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA) + bytes_per_pixel = 8; + + else + return; + + for (i = 0, rp = row; i < row_width; i++, rp += bytes_per_pixel) + { + png_uint_32 s0 = (*(rp ) << 8) | *(rp + 1); + png_uint_32 s1 = (*(rp + 2) << 8) | *(rp + 3); + png_uint_32 s2 = (*(rp + 4) << 8) | *(rp + 5); + png_uint_32 red = (s0 + s1 + 65536) & 0xffff; + png_uint_32 blue = (s2 + s1 + 65536) & 0xffff; + *(rp ) = (png_byte)((red >> 8) & 0xff); + *(rp + 1) = (png_byte)(red & 0xff); + *(rp + 4) = (png_byte)((blue >> 8) & 0xff); + *(rp + 5) = (png_byte)(blue & 0xff); + } + } + } +} +#endif /* MNG_FEATURES */ + +void PNGAPI +png_read_row(png_structrp png_ptr, png_bytep row, png_bytep dsp_row) +{ + png_row_info row_info; + + if (png_ptr == NULL) + return; + + png_debug2(1, "in png_read_row (row %lu, pass %d)", + (unsigned long)png_ptr->row_number, png_ptr->pass); + + /* png_read_start_row sets the information (in particular iwidth) for this + * interlace pass. + */ + if ((png_ptr->flags & PNG_FLAG_ROW_INIT) == 0) + png_read_start_row(png_ptr); + + /* 1.5.6: row_info moved out of png_struct to a local here. */ + row_info.width = png_ptr->iwidth; /* NOTE: width of current interlaced row */ + row_info.color_type = png_ptr->color_type; + row_info.bit_depth = png_ptr->bit_depth; + row_info.channels = png_ptr->channels; + row_info.pixel_depth = png_ptr->pixel_depth; + row_info.rowbytes = PNG_ROWBYTES(row_info.pixel_depth, row_info.width); + +#ifdef PNG_WARNINGS_SUPPORTED + if (png_ptr->row_number == 0 && png_ptr->pass == 0) + { + /* Check for transforms that have been set but were defined out */ +#if defined(PNG_WRITE_INVERT_SUPPORTED) && !defined(PNG_READ_INVERT_SUPPORTED) + if ((png_ptr->transformations & PNG_INVERT_MONO) != 0) + png_warning(png_ptr, "PNG_READ_INVERT_SUPPORTED is not defined"); +#endif + +#if defined(PNG_WRITE_FILLER_SUPPORTED) && !defined(PNG_READ_FILLER_SUPPORTED) + if ((png_ptr->transformations & PNG_FILLER) != 0) + png_warning(png_ptr, "PNG_READ_FILLER_SUPPORTED is not defined"); +#endif + +#if defined(PNG_WRITE_PACKSWAP_SUPPORTED) && \ + !defined(PNG_READ_PACKSWAP_SUPPORTED) + if ((png_ptr->transformations & PNG_PACKSWAP) != 0) + png_warning(png_ptr, "PNG_READ_PACKSWAP_SUPPORTED is not defined"); +#endif + +#if defined(PNG_WRITE_PACK_SUPPORTED) && !defined(PNG_READ_PACK_SUPPORTED) + if ((png_ptr->transformations & PNG_PACK) != 0) + png_warning(png_ptr, "PNG_READ_PACK_SUPPORTED is not defined"); +#endif + +#if defined(PNG_WRITE_SHIFT_SUPPORTED) && !defined(PNG_READ_SHIFT_SUPPORTED) + if ((png_ptr->transformations & PNG_SHIFT) != 0) + png_warning(png_ptr, "PNG_READ_SHIFT_SUPPORTED is not defined"); +#endif + +#if defined(PNG_WRITE_BGR_SUPPORTED) && !defined(PNG_READ_BGR_SUPPORTED) + if ((png_ptr->transformations & PNG_BGR) != 0) + png_warning(png_ptr, "PNG_READ_BGR_SUPPORTED is not defined"); +#endif + +#if defined(PNG_WRITE_SWAP_SUPPORTED) && !defined(PNG_READ_SWAP_SUPPORTED) + if ((png_ptr->transformations & PNG_SWAP_BYTES) != 0) + png_warning(png_ptr, "PNG_READ_SWAP_SUPPORTED is not defined"); +#endif + } +#endif /* WARNINGS */ + +#ifdef PNG_READ_INTERLACING_SUPPORTED + /* If interlaced and we do not need a new row, combine row and return. + * Notice that the pixels we have from previous rows have been transformed + * already; we can only combine like with like (transformed or + * untransformed) and, because of the libpng API for interlaced images, this + * means we must transform before de-interlacing. + */ + if (png_ptr->interlaced != 0 && + (png_ptr->transformations & PNG_INTERLACE) != 0) + { + switch (png_ptr->pass) + { + case 0: + if (png_ptr->row_number & 0x07) + { + if (dsp_row != NULL) + png_combine_row(png_ptr, dsp_row, 1/*display*/); + png_read_finish_row(png_ptr); + return; + } + break; + + case 1: + if ((png_ptr->row_number & 0x07) || png_ptr->width < 5) + { + if (dsp_row != NULL) + png_combine_row(png_ptr, dsp_row, 1/*display*/); + + png_read_finish_row(png_ptr); + return; + } + break; + + case 2: + if ((png_ptr->row_number & 0x07) != 4) + { + if (dsp_row != NULL && (png_ptr->row_number & 4)) + png_combine_row(png_ptr, dsp_row, 1/*display*/); + + png_read_finish_row(png_ptr); + return; + } + break; + + case 3: + if ((png_ptr->row_number & 3) || png_ptr->width < 3) + { + if (dsp_row != NULL) + png_combine_row(png_ptr, dsp_row, 1/*display*/); + + png_read_finish_row(png_ptr); + return; + } + break; + + case 4: + if ((png_ptr->row_number & 3) != 2) + { + if (dsp_row != NULL && (png_ptr->row_number & 2)) + png_combine_row(png_ptr, dsp_row, 1/*display*/); + + png_read_finish_row(png_ptr); + return; + } + break; + + case 5: + if ((png_ptr->row_number & 1) || png_ptr->width < 2) + { + if (dsp_row != NULL) + png_combine_row(png_ptr, dsp_row, 1/*display*/); + + png_read_finish_row(png_ptr); + return; + } + break; + + default: + case 6: + if ((png_ptr->row_number & 1) == 0) + { + png_read_finish_row(png_ptr); + return; + } + break; + } + } +#endif + + if ((png_ptr->mode & PNG_HAVE_IDAT) == 0) + png_error(png_ptr, "Invalid attempt to read row data"); + + /* Fill the row with IDAT data: */ + png_read_IDAT_data(png_ptr, png_ptr->row_buf, row_info.rowbytes + 1); + + if (png_ptr->row_buf[0] > PNG_FILTER_VALUE_NONE) + { + if (png_ptr->row_buf[0] < PNG_FILTER_VALUE_LAST) + png_read_filter_row(png_ptr, &row_info, png_ptr->row_buf + 1, + png_ptr->prev_row + 1, png_ptr->row_buf[0]); + else + png_error(png_ptr, "bad adaptive filter value"); + } + + /* libpng 1.5.6: the following line was copying png_ptr->rowbytes before + * 1.5.6, while the buffer really is this big in current versions of libpng + * it may not be in the future, so this was changed just to copy the + * interlaced count: + */ + memcpy(png_ptr->prev_row, png_ptr->row_buf, row_info.rowbytes + 1); + +#ifdef PNG_MNG_FEATURES_SUPPORTED + if ((png_ptr->mng_features_permitted & PNG_FLAG_MNG_FILTER_64) != 0 && + (png_ptr->filter_type == PNG_INTRAPIXEL_DIFFERENCING)) + { + /* Intrapixel differencing */ + png_do_read_intrapixel(&row_info, png_ptr->row_buf + 1); + } +#endif + +#ifdef PNG_READ_TRANSFORMS_SUPPORTED + if (png_ptr->transformations) + png_do_read_transformations(png_ptr, &row_info); +#endif + + /* The transformed pixel depth should match the depth now in row_info. */ + if (png_ptr->transformed_pixel_depth == 0) + { + png_ptr->transformed_pixel_depth = row_info.pixel_depth; + if (row_info.pixel_depth > png_ptr->maximum_pixel_depth) + png_error(png_ptr, "sequential row overflow"); + } + + else if (png_ptr->transformed_pixel_depth != row_info.pixel_depth) + png_error(png_ptr, "internal sequential row size calculation error"); + +#ifdef PNG_READ_INTERLACING_SUPPORTED + /* Expand interlaced rows to full size */ + if (png_ptr->interlaced != 0 && + (png_ptr->transformations & PNG_INTERLACE) != 0) + { + if (png_ptr->pass < 6) + png_do_read_interlace(&row_info, png_ptr->row_buf + 1, png_ptr->pass, + png_ptr->transformations); + + if (dsp_row != NULL) + png_combine_row(png_ptr, dsp_row, 1/*display*/); + + if (row != NULL) + png_combine_row(png_ptr, row, 0/*row*/); + } + + else +#endif + { + if (row != NULL) + png_combine_row(png_ptr, row, -1/*ignored*/); + + if (dsp_row != NULL) + png_combine_row(png_ptr, dsp_row, -1/*ignored*/); + } + png_read_finish_row(png_ptr); + + if (png_ptr->read_row_fn != NULL) + (*(png_ptr->read_row_fn))(png_ptr, png_ptr->row_number, png_ptr->pass); + +} +#endif /* SEQUENTIAL_READ */ + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +/* Read one or more rows of image data. If the image is interlaced, + * and png_set_interlace_handling() has been called, the rows need to + * contain the contents of the rows from the previous pass. If the + * image has alpha or transparency, and png_handle_alpha()[*] has been + * called, the rows contents must be initialized to the contents of the + * screen. + * + * "row" holds the actual image, and pixels are placed in it + * as they arrive. If the image is displayed after each pass, it will + * appear to "sparkle" in. "display_row" can be used to display a + * "chunky" progressive image, with finer detail added as it becomes + * available. If you do not want this "chunky" display, you may pass + * NULL for display_row. If you do not want the sparkle display, and + * you have not called png_handle_alpha(), you may pass NULL for rows. + * If you have called png_handle_alpha(), and the image has either an + * alpha channel or a transparency chunk, you must provide a buffer for + * rows. In this case, you do not have to provide a display_row buffer + * also, but you may. If the image is not interlaced, or if you have + * not called png_set_interlace_handling(), the display_row buffer will + * be ignored, so pass NULL to it. + * + * [*] png_handle_alpha() does not exist yet, as of this version of libpng + */ + +void PNGAPI +png_read_rows(png_structrp png_ptr, png_bytepp row, + png_bytepp display_row, png_uint_32 num_rows) +{ + png_uint_32 i; + png_bytepp rp; + png_bytepp dp; + + png_debug(1, "in png_read_rows"); + + if (png_ptr == NULL) + return; + + rp = row; + dp = display_row; + if (rp != NULL && dp != NULL) + for (i = 0; i < num_rows; i++) + { + png_bytep rptr = *rp++; + png_bytep dptr = *dp++; + + png_read_row(png_ptr, rptr, dptr); + } + + else if (rp != NULL) + for (i = 0; i < num_rows; i++) + { + png_bytep rptr = *rp; + png_read_row(png_ptr, rptr, NULL); + rp++; + } + + else if (dp != NULL) + for (i = 0; i < num_rows; i++) + { + png_bytep dptr = *dp; + png_read_row(png_ptr, NULL, dptr); + dp++; + } +} +#endif /* SEQUENTIAL_READ */ + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +/* Read the entire image. If the image has an alpha channel or a tRNS + * chunk, and you have called png_handle_alpha()[*], you will need to + * initialize the image to the current image that PNG will be overlaying. + * We set the num_rows again here, in case it was incorrectly set in + * png_read_start_row() by a call to png_read_update_info() or + * png_start_read_image() if png_set_interlace_handling() wasn't called + * prior to either of these functions like it should have been. You can + * only call this function once. If you desire to have an image for + * each pass of a interlaced image, use png_read_rows() instead. + * + * [*] png_handle_alpha() does not exist yet, as of this version of libpng + */ +void PNGAPI +png_read_image(png_structrp png_ptr, png_bytepp image) +{ + png_uint_32 i, image_height; + int pass, j; + png_bytepp rp; + + png_debug(1, "in png_read_image"); + + if (png_ptr == NULL) + return; + +#ifdef PNG_READ_INTERLACING_SUPPORTED + if ((png_ptr->flags & PNG_FLAG_ROW_INIT) == 0) + { + pass = png_set_interlace_handling(png_ptr); + /* And make sure transforms are initialized. */ + png_start_read_image(png_ptr); + } + else + { + if (png_ptr->interlaced != 0 && + (png_ptr->transformations & PNG_INTERLACE) == 0) + { + /* Caller called png_start_read_image or png_read_update_info without + * first turning on the PNG_INTERLACE transform. We can fix this here, + * but the caller should do it! + */ + png_warning(png_ptr, "Interlace handling should be turned on when " + "using png_read_image"); + /* Make sure this is set correctly */ + png_ptr->num_rows = png_ptr->height; + } + + /* Obtain the pass number, which also turns on the PNG_INTERLACE flag in + * the above error case. + */ + pass = png_set_interlace_handling(png_ptr); + } +#else + if (png_ptr->interlaced) + png_error(png_ptr, + "Cannot read interlaced image -- interlace handler disabled"); + + pass = 1; +#endif + + image_height=png_ptr->height; + + for (j = 0; j < pass; j++) + { + rp = image; + for (i = 0; i < image_height; i++) + { + png_read_row(png_ptr, *rp, NULL); + rp++; + } + } +} +#endif /* SEQUENTIAL_READ */ + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +/* Read the end of the PNG file. Will not read past the end of the + * file, will verify the end is accurate, and will read any comments + * or time information at the end of the file, if info is not NULL. + */ +void PNGAPI +png_read_end(png_structrp png_ptr, png_inforp info_ptr) +{ +#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED + int keep; +#endif + + png_debug(1, "in png_read_end"); + + if (png_ptr == NULL) + return; + + /* If png_read_end is called in the middle of reading the rows there may + * still be pending IDAT data and an owned zstream. Deal with this here. + */ +#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED + if (png_chunk_unknown_handling(png_ptr, png_IDAT) == 0) +#endif + png_read_finish_IDAT(png_ptr); + +#ifdef PNG_READ_CHECK_FOR_INVALID_INDEX_SUPPORTED + /* Report invalid palette index; added at libng-1.5.10 */ + if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE && + png_ptr->num_palette_max > png_ptr->num_palette) + png_benign_error(png_ptr, "Read palette index exceeding num_palette"); +#endif + + do + { + png_uint_32 length = png_read_chunk_header(png_ptr); + png_uint_32 chunk_name = png_ptr->chunk_name; + + if (chunk_name == png_IEND) + png_handle_IEND(png_ptr, info_ptr, length); + + else if (chunk_name == png_IHDR) + png_handle_IHDR(png_ptr, info_ptr, length); + + else if (info_ptr == NULL) + png_crc_finish(png_ptr, length); + +#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED + else if ((keep = png_chunk_unknown_handling(png_ptr, chunk_name)) != 0) + { + if (chunk_name == png_IDAT) + { + if ((length > 0) || + (png_ptr->mode & PNG_HAVE_CHUNK_AFTER_IDAT) != 0) + png_benign_error(png_ptr, "Too many IDATs found"); + } + png_handle_unknown(png_ptr, info_ptr, length, keep); + if (chunk_name == png_PLTE) + png_ptr->mode |= PNG_HAVE_PLTE; + } +#endif + + else if (chunk_name == png_IDAT) + { + /* Zero length IDATs are legal after the last IDAT has been + * read, but not after other chunks have been read. + */ + if ((length > 0) || + (png_ptr->mode & PNG_HAVE_CHUNK_AFTER_IDAT) != 0) + png_benign_error(png_ptr, "Too many IDATs found"); + + png_crc_finish(png_ptr, length); + } + else if (chunk_name == png_PLTE) + png_handle_PLTE(png_ptr, info_ptr, length); + +#ifdef PNG_READ_bKGD_SUPPORTED + else if (chunk_name == png_bKGD) + png_handle_bKGD(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_cHRM_SUPPORTED + else if (chunk_name == png_cHRM) + png_handle_cHRM(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_gAMA_SUPPORTED + else if (chunk_name == png_gAMA) + png_handle_gAMA(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_hIST_SUPPORTED + else if (chunk_name == png_hIST) + png_handle_hIST(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_oFFs_SUPPORTED + else if (chunk_name == png_oFFs) + png_handle_oFFs(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_pCAL_SUPPORTED + else if (chunk_name == png_pCAL) + png_handle_pCAL(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_sCAL_SUPPORTED + else if (chunk_name == png_sCAL) + png_handle_sCAL(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_pHYs_SUPPORTED + else if (chunk_name == png_pHYs) + png_handle_pHYs(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_sBIT_SUPPORTED + else if (chunk_name == png_sBIT) + png_handle_sBIT(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_sRGB_SUPPORTED + else if (chunk_name == png_sRGB) + png_handle_sRGB(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_iCCP_SUPPORTED + else if (chunk_name == png_iCCP) + png_handle_iCCP(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_sPLT_SUPPORTED + else if (chunk_name == png_sPLT) + png_handle_sPLT(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_tEXt_SUPPORTED + else if (chunk_name == png_tEXt) + png_handle_tEXt(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_tIME_SUPPORTED + else if (chunk_name == png_tIME) + png_handle_tIME(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_tRNS_SUPPORTED + else if (chunk_name == png_tRNS) + png_handle_tRNS(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_zTXt_SUPPORTED + else if (chunk_name == png_zTXt) + png_handle_zTXt(png_ptr, info_ptr, length); +#endif + +#ifdef PNG_READ_iTXt_SUPPORTED + else if (chunk_name == png_iTXt) + png_handle_iTXt(png_ptr, info_ptr, length); +#endif + + else + png_handle_unknown(png_ptr, info_ptr, length, + PNG_HANDLE_CHUNK_AS_DEFAULT); + } while ((png_ptr->mode & PNG_HAVE_IEND) == 0); +} +#endif /* SEQUENTIAL_READ */ + +/* Free all memory used in the read struct */ +static void +png_read_destroy(png_structrp png_ptr) +{ + png_debug(1, "in png_read_destroy"); + +#ifdef PNG_READ_GAMMA_SUPPORTED + png_destroy_gamma_table(png_ptr); +#endif + + png_free(png_ptr, png_ptr->big_row_buf); + png_ptr->big_row_buf = NULL; + png_free(png_ptr, png_ptr->big_prev_row); + png_ptr->big_prev_row = NULL; + png_free(png_ptr, png_ptr->read_buffer); + png_ptr->read_buffer = NULL; + +#ifdef PNG_READ_QUANTIZE_SUPPORTED + png_free(png_ptr, png_ptr->palette_lookup); + png_ptr->palette_lookup = NULL; + png_free(png_ptr, png_ptr->quantize_index); + png_ptr->quantize_index = NULL; +#endif + + if ((png_ptr->free_me & PNG_FREE_PLTE) != 0) + { + png_zfree(png_ptr, png_ptr->palette); + png_ptr->palette = NULL; + } + png_ptr->free_me &= ~PNG_FREE_PLTE; + +#if defined(PNG_tRNS_SUPPORTED) || \ + defined(PNG_READ_EXPAND_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED) + if ((png_ptr->free_me & PNG_FREE_TRNS) != 0) + { + png_free(png_ptr, png_ptr->trans_alpha); + png_ptr->trans_alpha = NULL; + } + png_ptr->free_me &= ~PNG_FREE_TRNS; +#endif + + inflateEnd(&png_ptr->zstream); + +#ifdef PNG_PROGRESSIVE_READ_SUPPORTED + png_free(png_ptr, png_ptr->save_buffer); + png_ptr->save_buffer = NULL; +#endif + +#if defined(PNG_STORE_UNKNOWN_CHUNKS_SUPPORTED) && \ + defined(PNG_READ_UNKNOWN_CHUNKS_SUPPORTED) + png_free(png_ptr, png_ptr->unknown_chunk.data); + png_ptr->unknown_chunk.data = NULL; +#endif + +#ifdef PNG_SET_UNKNOWN_CHUNKS_SUPPORTED + png_free(png_ptr, png_ptr->chunk_list); + png_ptr->chunk_list = NULL; +#endif + + /* NOTE: the 'setjmp' buffer may still be allocated and the memory and error + * callbacks are still set at this point. They are required to complete the + * destruction of the png_struct itself. + */ +} + +/* Free all memory used by the read */ +void PNGAPI +png_destroy_read_struct(png_structpp png_ptr_ptr, png_infopp info_ptr_ptr, + png_infopp end_info_ptr_ptr) +{ + png_structrp png_ptr = NULL; + + png_debug(1, "in png_destroy_read_struct"); + + if (png_ptr_ptr != NULL) + png_ptr = *png_ptr_ptr; + + if (png_ptr == NULL) + return; + + /* libpng 1.6.0: use the API to destroy info structs to ensure consistent + * behavior. Prior to 1.6.0 libpng did extra 'info' destruction in this API. + * The extra was, apparently, unnecessary yet this hides memory leak bugs. + */ + png_destroy_info_struct(png_ptr, end_info_ptr_ptr); + png_destroy_info_struct(png_ptr, info_ptr_ptr); + + *png_ptr_ptr = NULL; + png_read_destroy(png_ptr); + png_destroy_png_struct(png_ptr); +} + +void PNGAPI +png_set_read_status_fn(png_structrp png_ptr, png_read_status_ptr read_row_fn) +{ + if (png_ptr == NULL) + return; + + png_ptr->read_row_fn = read_row_fn; +} + + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +#ifdef PNG_INFO_IMAGE_SUPPORTED +void PNGAPI +png_read_png(png_structrp png_ptr, png_inforp info_ptr, + int transforms, + voidp params) +{ + if (png_ptr == NULL || info_ptr == NULL) + return; + + /* png_read_info() gives us all of the information from the + * PNG file before the first IDAT (image data chunk). + */ + png_read_info(png_ptr, info_ptr); + if (info_ptr->height > PNG_UINT_32_MAX/(sizeof (png_bytep))) + png_error(png_ptr, "Image is too high to process with png_read_png()"); + + /* -------------- image transformations start here ------------------- */ + /* libpng 1.6.10: add code to cause a png_app_error if a selected TRANSFORM + * is not implemented. This will only happen in de-configured (non-default) + * libpng builds. The results can be unexpected - png_read_png may return + * short or mal-formed rows because the transform is skipped. + */ + + /* Tell libpng to strip 16-bit/color files down to 8 bits per color. + */ + if ((transforms & PNG_TRANSFORM_SCALE_16) != 0) + /* Added at libpng-1.5.4. "strip_16" produces the same result that it + * did in earlier versions, while "scale_16" is now more accurate. + */ +#ifdef PNG_READ_SCALE_16_TO_8_SUPPORTED + png_set_scale_16(png_ptr); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_SCALE_16 not supported"); +#endif + + /* If both SCALE and STRIP are required pngrtran will effectively cancel the + * latter by doing SCALE first. This is ok and allows apps not to check for + * which is supported to get the right answer. + */ + if ((transforms & PNG_TRANSFORM_STRIP_16) != 0) +#ifdef PNG_READ_STRIP_16_TO_8_SUPPORTED + png_set_strip_16(png_ptr); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_STRIP_16 not supported"); +#endif + + /* Strip alpha bytes from the input data without combining with + * the background (not recommended). + */ + if ((transforms & PNG_TRANSFORM_STRIP_ALPHA) != 0) +#ifdef PNG_READ_STRIP_ALPHA_SUPPORTED + png_set_strip_alpha(png_ptr); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_STRIP_ALPHA not supported"); +#endif + + /* Extract multiple pixels with bit depths of 1, 2, or 4 from a single + * byte into separate bytes (useful for paletted and grayscale images). + */ + if ((transforms & PNG_TRANSFORM_PACKING) != 0) +#ifdef PNG_READ_PACK_SUPPORTED + png_set_packing(png_ptr); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_PACKING not supported"); +#endif + + /* Change the order of packed pixels to least significant bit first + * (not useful if you are using png_set_packing). + */ + if ((transforms & PNG_TRANSFORM_PACKSWAP) != 0) +#ifdef PNG_READ_PACKSWAP_SUPPORTED + png_set_packswap(png_ptr); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_PACKSWAP not supported"); +#endif + + /* Expand paletted colors into true RGB triplets + * Expand grayscale images to full 8 bits from 1, 2, or 4 bits/pixel + * Expand paletted or RGB images with transparency to full alpha + * channels so the data will be available as RGBA quartets. + */ + if ((transforms & PNG_TRANSFORM_EXPAND) != 0) +#ifdef PNG_READ_EXPAND_SUPPORTED + png_set_expand(png_ptr); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_EXPAND not supported"); +#endif + + /* We don't handle background color or gamma transformation or quantizing. + */ + + /* Invert monochrome files to have 0 as white and 1 as black + */ + if ((transforms & PNG_TRANSFORM_INVERT_MONO) != 0) +#ifdef PNG_READ_INVERT_SUPPORTED + png_set_invert_mono(png_ptr); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_INVERT_MONO not supported"); +#endif + + /* If you want to shift the pixel values from the range [0,255] or + * [0,65535] to the original [0,7] or [0,31], or whatever range the + * colors were originally in: + */ + if ((transforms & PNG_TRANSFORM_SHIFT) != 0) +#ifdef PNG_READ_SHIFT_SUPPORTED + if ((info_ptr->valid & PNG_INFO_sBIT) != 0) + png_set_shift(png_ptr, &info_ptr->sig_bit); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_SHIFT not supported"); +#endif + + /* Flip the RGB pixels to BGR (or RGBA to BGRA) */ + if ((transforms & PNG_TRANSFORM_BGR) != 0) +#ifdef PNG_READ_BGR_SUPPORTED + png_set_bgr(png_ptr); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_BGR not supported"); +#endif + + /* Swap the RGBA or GA data to ARGB or AG (or BGRA to ABGR) */ + if ((transforms & PNG_TRANSFORM_SWAP_ALPHA) != 0) +#ifdef PNG_READ_SWAP_ALPHA_SUPPORTED + png_set_swap_alpha(png_ptr); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_SWAP_ALPHA not supported"); +#endif + + /* Swap bytes of 16-bit files to least significant byte first */ + if ((transforms & PNG_TRANSFORM_SWAP_ENDIAN) != 0) +#ifdef PNG_READ_SWAP_SUPPORTED + png_set_swap(png_ptr); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_SWAP_ENDIAN not supported"); +#endif + +/* Added at libpng-1.2.41 */ + /* Invert the alpha channel from opacity to transparency */ + if ((transforms & PNG_TRANSFORM_INVERT_ALPHA) != 0) +#ifdef PNG_READ_INVERT_ALPHA_SUPPORTED + png_set_invert_alpha(png_ptr); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_INVERT_ALPHA not supported"); +#endif + +/* Added at libpng-1.2.41 */ + /* Expand grayscale image to RGB */ + if ((transforms & PNG_TRANSFORM_GRAY_TO_RGB) != 0) +#ifdef PNG_READ_GRAY_TO_RGB_SUPPORTED + png_set_gray_to_rgb(png_ptr); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_GRAY_TO_RGB not supported"); +#endif + +/* Added at libpng-1.5.4 */ + if ((transforms & PNG_TRANSFORM_EXPAND_16) != 0) +#ifdef PNG_READ_EXPAND_16_SUPPORTED + png_set_expand_16(png_ptr); +#else + png_app_error(png_ptr, "PNG_TRANSFORM_EXPAND_16 not supported"); +#endif + + /* We don't handle adding filler bytes */ + + /* We use png_read_image and rely on that for interlace handling, but we also + * call png_read_update_info therefore must turn on interlace handling now: + */ + (void)png_set_interlace_handling(png_ptr); + + /* Optional call to gamma correct and add the background to the palette + * and update info structure. REQUIRED if you are expecting libpng to + * update the palette for you (i.e., you selected such a transform above). + */ + png_read_update_info(png_ptr, info_ptr); + + /* -------------- image transformations end here ------------------- */ + + png_free_data(png_ptr, info_ptr, PNG_FREE_ROWS, 0); + if (info_ptr->row_pointers == NULL) + { + png_uint_32 iptr; + + info_ptr->row_pointers = png_voidcast(png_bytepp, png_malloc(png_ptr, + info_ptr->height * (sizeof (png_bytep)))); + + for (iptr=0; iptrheight; iptr++) + info_ptr->row_pointers[iptr] = NULL; + + info_ptr->free_me |= PNG_FREE_ROWS; + + for (iptr = 0; iptr < info_ptr->height; iptr++) + info_ptr->row_pointers[iptr] = png_voidcast(png_bytep, + png_malloc(png_ptr, info_ptr->rowbytes)); + } + + png_read_image(png_ptr, info_ptr->row_pointers); + info_ptr->valid |= PNG_INFO_IDAT; + + /* Read rest of file, and get additional chunks in info_ptr - REQUIRED */ + png_read_end(png_ptr, info_ptr); + + PNG_UNUSED(params) +} +#endif /* INFO_IMAGE */ +#endif /* SEQUENTIAL_READ */ + +#ifdef PNG_SIMPLIFIED_READ_SUPPORTED +/* SIMPLIFIED READ + * + * This code currently relies on the sequential reader, though it could easily + * be made to work with the progressive one. + */ +/* Arguments to png_image_finish_read: */ + +/* Encoding of PNG data (used by the color-map code) */ +# define P_NOTSET 0 /* File encoding not yet known */ +# define P_sRGB 1 /* 8-bit encoded to sRGB gamma */ +# define P_LINEAR 2 /* 16-bit linear: not encoded, NOT pre-multiplied! */ +# define P_FILE 3 /* 8-bit encoded to file gamma, not sRGB or linear */ +# define P_LINEAR8 4 /* 8-bit linear: only from a file value */ + +/* Color-map processing: after libpng has run on the PNG image further + * processing may be needed to convert the data to color-map indices. + */ +#define PNG_CMAP_NONE 0 +#define PNG_CMAP_GA 1 /* Process GA data to a color-map with alpha */ +#define PNG_CMAP_TRANS 2 /* Process GA data to a background index */ +#define PNG_CMAP_RGB 3 /* Process RGB data */ +#define PNG_CMAP_RGB_ALPHA 4 /* Process RGBA data */ + +/* The following document where the background is for each processing case. */ +#define PNG_CMAP_NONE_BACKGROUND 256 +#define PNG_CMAP_GA_BACKGROUND 231 +#define PNG_CMAP_TRANS_BACKGROUND 254 +#define PNG_CMAP_RGB_BACKGROUND 256 +#define PNG_CMAP_RGB_ALPHA_BACKGROUND 216 + +typedef struct +{ + /* Arguments: */ + png_imagep image; + png_voidp buffer; + png_int_32 row_stride; + png_voidp colormap; + png_const_colorp background; + /* Local variables: */ + png_voidp local_row; + png_voidp first_row; + ptrdiff_t row_bytes; /* step between rows */ + int file_encoding; /* E_ values above */ + png_fixed_point gamma_to_linear; /* For P_FILE, reciprocal of gamma */ + int colormap_processing; /* PNG_CMAP_ values above */ +} png_image_read_control; + +/* Do all the *safe* initialization - 'safe' means that png_error won't be + * called, so setting up the jmp_buf is not required. This means that anything + * called from here must *not* call png_malloc - it has to call png_malloc_warn + * instead so that control is returned safely back to this routine. + */ +static int +png_image_read_init(png_imagep image) +{ + if (image->opaque == NULL) + { + png_structp png_ptr = png_create_read_struct(PNG_LIBPNG_VER_STRING, image, + png_safe_error, png_safe_warning); + + /* And set the rest of the structure to NULL to ensure that the various + * fields are consistent. + */ + memset(image, 0, (sizeof *image)); + image->version = PNG_IMAGE_VERSION; + + if (png_ptr != NULL) + { + png_infop info_ptr = png_create_info_struct(png_ptr); + + if (info_ptr != NULL) + { + png_controlp control = png_voidcast(png_controlp, + png_malloc_warn(png_ptr, (sizeof *control))); + + if (control != NULL) + { + memset(control, 0, (sizeof *control)); + + control->png_ptr = png_ptr; + control->info_ptr = info_ptr; + control->for_write = 0; + + image->opaque = control; + return 1; + } + + /* Error clean up */ + png_destroy_info_struct(png_ptr, &info_ptr); + } + + png_destroy_read_struct(&png_ptr, NULL, NULL); + } + + return png_image_error(image, "png_image_read: out of memory"); + } + + return png_image_error(image, "png_image_read: opaque pointer not NULL"); +} + +/* Utility to find the base format of a PNG file from a png_struct. */ +static png_uint_32 +png_image_format(png_structrp png_ptr) +{ + png_uint_32 format = 0; + + if ((png_ptr->color_type & PNG_COLOR_MASK_COLOR) != 0) + format |= PNG_FORMAT_FLAG_COLOR; + + if ((png_ptr->color_type & PNG_COLOR_MASK_ALPHA) != 0) + format |= PNG_FORMAT_FLAG_ALPHA; + + /* Use png_ptr here, not info_ptr, because by examination png_handle_tRNS + * sets the png_struct fields; that's all we are interested in here. The + * precise interaction with an app call to png_set_tRNS and PNG file reading + * is unclear. + */ + else if (png_ptr->num_trans > 0) + format |= PNG_FORMAT_FLAG_ALPHA; + + if (png_ptr->bit_depth == 16) + format |= PNG_FORMAT_FLAG_LINEAR; + + if ((png_ptr->color_type & PNG_COLOR_MASK_PALETTE) != 0) + format |= PNG_FORMAT_FLAG_COLORMAP; + + return format; +} + +/* Is the given gamma significantly different from sRGB? The test is the same + * one used in pngrtran.c when deciding whether to do gamma correction. The + * arithmetic optimizes the division by using the fact that the inverse of the + * file sRGB gamma is 2.2 + */ +static int +png_gamma_not_sRGB(png_fixed_point g) +{ + if (g < PNG_FP_1) + { + /* An uninitialized gamma is assumed to be sRGB for the simplified API. */ + if (g == 0) + return 0; + + return png_gamma_significant((g * 11 + 2)/5 /* i.e. *2.2, rounded */); + } + + return 1; +} + +/* Do the main body of a 'png_image_begin_read' function; read the PNG file + * header and fill in all the information. This is executed in a safe context, + * unlike the init routine above. + */ +static int +png_image_read_header(png_voidp argument) +{ + png_imagep image = png_voidcast(png_imagep, argument); + png_structrp png_ptr = image->opaque->png_ptr; + png_inforp info_ptr = image->opaque->info_ptr; + + png_set_benign_errors(png_ptr, 1/*warn*/); + png_read_info(png_ptr, info_ptr); + + /* Do this the fast way; just read directly out of png_struct. */ + image->width = png_ptr->width; + image->height = png_ptr->height; + + { + png_uint_32 format = png_image_format(png_ptr); + + image->format = format; + +#ifdef PNG_COLORSPACE_SUPPORTED + /* Does the colorspace match sRGB? If there is no color endpoint + * (colorant) information assume yes, otherwise require the + * 'ENDPOINTS_MATCHP_sRGB' colorspace flag to have been set. If the + * colorspace has been determined to be invalid ignore it. + */ + if ((format & PNG_FORMAT_FLAG_COLOR) != 0 && ((png_ptr->colorspace.flags + & (PNG_COLORSPACE_HAVE_ENDPOINTS|PNG_COLORSPACE_ENDPOINTS_MATCH_sRGB| + PNG_COLORSPACE_INVALID)) == PNG_COLORSPACE_HAVE_ENDPOINTS)) + image->flags |= PNG_IMAGE_FLAG_COLORSPACE_NOT_sRGB; +#endif + } + + /* We need the maximum number of entries regardless of the format the + * application sets here. + */ + { + png_uint_32 cmap_entries; + + switch (png_ptr->color_type) + { + case PNG_COLOR_TYPE_GRAY: + cmap_entries = 1U << png_ptr->bit_depth; + break; + + case PNG_COLOR_TYPE_PALETTE: + cmap_entries = png_ptr->num_palette; + break; + + default: + cmap_entries = 256; + break; + } + + if (cmap_entries > 256) + cmap_entries = 256; + + image->colormap_entries = cmap_entries; + } + + return 1; +} + +#ifdef PNG_STDIO_SUPPORTED +int PNGAPI +png_image_begin_read_from_stdio(png_imagep image, FILE* file) +{ + if (image != NULL && image->version == PNG_IMAGE_VERSION) + { + if (file != NULL) + { + if (png_image_read_init(image) != 0) + { + /* This is slightly evil, but png_init_io doesn't do anything other + * than this and we haven't changed the standard IO functions so + * this saves a 'safe' function. + */ + image->opaque->png_ptr->io_ptr = file; + return png_safe_execute(image, png_image_read_header, image); + } + } + + else + return png_image_error(image, + "png_image_begin_read_from_stdio: invalid argument"); + } + + else if (image != NULL) + return png_image_error(image, + "png_image_begin_read_from_stdio: incorrect PNG_IMAGE_VERSION"); + + return 0; +} + +int PNGAPI +png_image_begin_read_from_file(png_imagep image, const char *file_name) +{ + if (image != NULL && image->version == PNG_IMAGE_VERSION) + { + if (file_name != NULL) + { + FILE *fp = fopen(file_name, "rb"); + + if (fp != NULL) + { + if (png_image_read_init(image) != 0) + { + image->opaque->png_ptr->io_ptr = fp; + image->opaque->owned_file = 1; + return png_safe_execute(image, png_image_read_header, image); + } + + /* Clean up: just the opened file. */ + (void)fclose(fp); + } + + else + return png_image_error(image, strerror(errno)); + } + + else + return png_image_error(image, + "png_image_begin_read_from_file: invalid argument"); + } + + else if (image != NULL) + return png_image_error(image, + "png_image_begin_read_from_file: incorrect PNG_IMAGE_VERSION"); + + return 0; +} +#endif /* STDIO */ + +static void PNGCBAPI +png_image_memory_read(png_structp png_ptr, png_bytep out, png_size_t need) +{ + if (png_ptr != NULL) + { + png_imagep image = png_voidcast(png_imagep, png_ptr->io_ptr); + if (image != NULL) + { + png_controlp cp = image->opaque; + if (cp != NULL) + { + png_const_bytep memory = cp->memory; + png_size_t size = cp->size; + + if (memory != NULL && size >= need) + { + memcpy(out, memory, need); + cp->memory = memory + need; + cp->size = size - need; + return; + } + + png_error(png_ptr, "read beyond end of data"); + } + } + + png_error(png_ptr, "invalid memory read"); + } +} + +int PNGAPI png_image_begin_read_from_memory(png_imagep image, + png_const_voidp memory, png_size_t size) +{ + if (image != NULL && image->version == PNG_IMAGE_VERSION) + { + if (memory != NULL && size > 0) + { + if (png_image_read_init(image) != 0) + { + /* Now set the IO functions to read from the memory buffer and + * store it into io_ptr. Again do this in-place to avoid calling a + * libpng function that requires error handling. + */ + image->opaque->memory = png_voidcast(png_const_bytep, memory); + image->opaque->size = size; + image->opaque->png_ptr->io_ptr = image; + image->opaque->png_ptr->read_data_fn = png_image_memory_read; + + return png_safe_execute(image, png_image_read_header, image); + } + } + + else + return png_image_error(image, + "png_image_begin_read_from_memory: invalid argument"); + } + + else if (image != NULL) + return png_image_error(image, + "png_image_begin_read_from_memory: incorrect PNG_IMAGE_VERSION"); + + return 0; +} + +/* Utility function to skip chunks that are not used by the simplified image + * read functions and an appropriate macro to call it. + */ +#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED +static void +png_image_skip_unused_chunks(png_structrp png_ptr) +{ + /* Prepare the reader to ignore all recognized chunks whose data will not + * be used, i.e., all chunks recognized by libpng except for those + * involved in basic image reading: + * + * IHDR, PLTE, IDAT, IEND + * + * Or image data handling: + * + * tRNS, bKGD, gAMA, cHRM, sRGB, [iCCP] and sBIT. + * + * This provides a small performance improvement and eliminates any + * potential vulnerability to security problems in the unused chunks. + * + * At present the iCCP chunk data isn't used, so iCCP chunk can be ignored + * too. This allows the simplified API to be compiled without iCCP support, + * however if the support is there the chunk is still checked to detect + * errors (which are unfortunately quite common.) + */ + { + static PNG_CONST png_byte chunks_to_process[] = { + 98, 75, 71, 68, '\0', /* bKGD */ + 99, 72, 82, 77, '\0', /* cHRM */ + 103, 65, 77, 65, '\0', /* gAMA */ +# ifdef PNG_READ_iCCP_SUPPORTED + 105, 67, 67, 80, '\0', /* iCCP */ +# endif + 115, 66, 73, 84, '\0', /* sBIT */ + 115, 82, 71, 66, '\0', /* sRGB */ + }; + + /* Ignore unknown chunks and all other chunks except for the + * IHDR, PLTE, tRNS, IDAT, and IEND chunks. + */ + png_set_keep_unknown_chunks(png_ptr, PNG_HANDLE_CHUNK_NEVER, + NULL, -1); + + /* But do not ignore image data handling chunks */ + png_set_keep_unknown_chunks(png_ptr, PNG_HANDLE_CHUNK_AS_DEFAULT, + chunks_to_process, (int)/*SAFE*/(sizeof chunks_to_process)/5); + } +} + +# define PNG_SKIP_CHUNKS(p) png_image_skip_unused_chunks(p) +#else +# define PNG_SKIP_CHUNKS(p) ((void)0) +#endif /* HANDLE_AS_UNKNOWN */ + +/* The following macro gives the exact rounded answer for all values in the + * range 0..255 (it actually divides by 51.2, but the rounding still generates + * the correct numbers 0..5 + */ +#define PNG_DIV51(v8) (((v8) * 5 + 130) >> 8) + +/* Utility functions to make particular color-maps */ +static void +set_file_encoding(png_image_read_control *display) +{ + png_fixed_point g = display->image->opaque->png_ptr->colorspace.gamma; + if (png_gamma_significant(g) != 0) + { + if (png_gamma_not_sRGB(g) != 0) + { + display->file_encoding = P_FILE; + display->gamma_to_linear = png_reciprocal(g); + } + + else + display->file_encoding = P_sRGB; + } + + else + display->file_encoding = P_LINEAR8; +} + +static unsigned int +decode_gamma(png_image_read_control *display, png_uint_32 value, int encoding) +{ + if (encoding == P_FILE) /* double check */ + encoding = display->file_encoding; + + if (encoding == P_NOTSET) /* must be the file encoding */ + { + set_file_encoding(display); + encoding = display->file_encoding; + } + + switch (encoding) + { + case P_FILE: + value = png_gamma_16bit_correct(value*257, display->gamma_to_linear); + break; + + case P_sRGB: + value = png_sRGB_table[value]; + break; + + case P_LINEAR: + break; + + case P_LINEAR8: + value *= 257; + break; + + default: + png_error(display->image->opaque->png_ptr, + "unexpected encoding (internal error)"); + break; + } + + return value; +} + +static png_uint_32 +png_colormap_compose(png_image_read_control *display, + png_uint_32 foreground, int foreground_encoding, png_uint_32 alpha, + png_uint_32 background, int encoding) +{ + /* The file value is composed on the background, the background has the given + * encoding and so does the result, the file is encoded with P_FILE and the + * file and alpha are 8-bit values. The (output) encoding will always be + * P_LINEAR or P_sRGB. + */ + png_uint_32 f = decode_gamma(display, foreground, foreground_encoding); + png_uint_32 b = decode_gamma(display, background, encoding); + + /* The alpha is always an 8-bit value (it comes from the palette), the value + * scaled by 255 is what PNG_sRGB_FROM_LINEAR requires. + */ + f = f * alpha + b * (255-alpha); + + if (encoding == P_LINEAR) + { + /* Scale to 65535; divide by 255, approximately (in fact this is extremely + * accurate, it divides by 255.00000005937181414556, with no overflow.) + */ + f *= 257; /* Now scaled by 65535 */ + f += f >> 16; + f = (f+32768) >> 16; + } + + else /* P_sRGB */ + f = PNG_sRGB_FROM_LINEAR(f); + + return f; +} + +/* NOTE: P_LINEAR values to this routine must be 16-bit, but P_FILE values must + * be 8-bit. + */ +static void +png_create_colormap_entry(png_image_read_control *display, + png_uint_32 ip, png_uint_32 red, png_uint_32 green, png_uint_32 blue, + png_uint_32 alpha, int encoding) +{ + png_imagep image = display->image; + const int output_encoding = (image->format & PNG_FORMAT_FLAG_LINEAR) != 0 ? + P_LINEAR : P_sRGB; + const int convert_to_Y = (image->format & PNG_FORMAT_FLAG_COLOR) == 0 && + (red != green || green != blue); + + if (ip > 255) + png_error(image->opaque->png_ptr, "color-map index out of range"); + + /* Update the cache with whether the file gamma is significantly different + * from sRGB. + */ + if (encoding == P_FILE) + { + if (display->file_encoding == P_NOTSET) + set_file_encoding(display); + + /* Note that the cached value may be P_FILE too, but if it is then the + * gamma_to_linear member has been set. + */ + encoding = display->file_encoding; + } + + if (encoding == P_FILE) + { + png_fixed_point g = display->gamma_to_linear; + + red = png_gamma_16bit_correct(red*257, g); + green = png_gamma_16bit_correct(green*257, g); + blue = png_gamma_16bit_correct(blue*257, g); + + if (convert_to_Y != 0 || output_encoding == P_LINEAR) + { + alpha *= 257; + encoding = P_LINEAR; + } + + else + { + red = PNG_sRGB_FROM_LINEAR(red * 255); + green = PNG_sRGB_FROM_LINEAR(green * 255); + blue = PNG_sRGB_FROM_LINEAR(blue * 255); + encoding = P_sRGB; + } + } + + else if (encoding == P_LINEAR8) + { + /* This encoding occurs quite frequently in test cases because PngSuite + * includes a gAMA 1.0 chunk with most images. + */ + red *= 257; + green *= 257; + blue *= 257; + alpha *= 257; + encoding = P_LINEAR; + } + + else if (encoding == P_sRGB && + (convert_to_Y != 0 || output_encoding == P_LINEAR)) + { + /* The values are 8-bit sRGB values, but must be converted to 16-bit + * linear. + */ + red = png_sRGB_table[red]; + green = png_sRGB_table[green]; + blue = png_sRGB_table[blue]; + alpha *= 257; + encoding = P_LINEAR; + } + + /* This is set if the color isn't gray but the output is. */ + if (encoding == P_LINEAR) + { + if (convert_to_Y != 0) + { + /* NOTE: these values are copied from png_do_rgb_to_gray */ + png_uint_32 y = (png_uint_32)6968 * red + (png_uint_32)23434 * green + + (png_uint_32)2366 * blue; + + if (output_encoding == P_LINEAR) + y = (y + 16384) >> 15; + + else + { + /* y is scaled by 32768, we need it scaled by 255: */ + y = (y + 128) >> 8; + y *= 255; + y = PNG_sRGB_FROM_LINEAR((y + 64) >> 7); + encoding = P_sRGB; + } + + blue = red = green = y; + } + + else if (output_encoding == P_sRGB) + { + red = PNG_sRGB_FROM_LINEAR(red * 255); + green = PNG_sRGB_FROM_LINEAR(green * 255); + blue = PNG_sRGB_FROM_LINEAR(blue * 255); + alpha = PNG_DIV257(alpha); + encoding = P_sRGB; + } + } + + if (encoding != output_encoding) + png_error(image->opaque->png_ptr, "bad encoding (internal error)"); + + /* Store the value. */ + { +# ifdef PNG_FORMAT_AFIRST_SUPPORTED + const int afirst = (image->format & PNG_FORMAT_FLAG_AFIRST) != 0 && + (image->format & PNG_FORMAT_FLAG_ALPHA) != 0; +# else +# define afirst 0 +# endif +# ifdef PNG_FORMAT_BGR_SUPPORTED + const int bgr = (image->format & PNG_FORMAT_FLAG_BGR) != 0 ? 2 : 0; +# else +# define bgr 0 +# endif + + if (output_encoding == P_LINEAR) + { + png_uint_16p entry = png_voidcast(png_uint_16p, display->colormap); + + entry += ip * PNG_IMAGE_SAMPLE_CHANNELS(image->format); + + /* The linear 16-bit values must be pre-multiplied by the alpha channel + * value, if less than 65535 (this is, effectively, composite on black + * if the alpha channel is removed.) + */ + switch (PNG_IMAGE_SAMPLE_CHANNELS(image->format)) + { + case 4: + entry[afirst ? 0 : 3] = (png_uint_16)alpha; + /* FALL THROUGH */ + + case 3: + if (alpha < 65535) + { + if (alpha > 0) + { + blue = (blue * alpha + 32767U)/65535U; + green = (green * alpha + 32767U)/65535U; + red = (red * alpha + 32767U)/65535U; + } + + else + red = green = blue = 0; + } + entry[afirst + (2 ^ bgr)] = (png_uint_16)blue; + entry[afirst + 1] = (png_uint_16)green; + entry[afirst + bgr] = (png_uint_16)red; + break; + + case 2: + entry[1 ^ afirst] = (png_uint_16)alpha; + /* FALL THROUGH */ + + case 1: + if (alpha < 65535) + { + if (alpha > 0) + green = (green * alpha + 32767U)/65535U; + + else + green = 0; + } + entry[afirst] = (png_uint_16)green; + break; + + default: + break; + } + } + + else /* output encoding is P_sRGB */ + { + png_bytep entry = png_voidcast(png_bytep, display->colormap); + + entry += ip * PNG_IMAGE_SAMPLE_CHANNELS(image->format); + + switch (PNG_IMAGE_SAMPLE_CHANNELS(image->format)) + { + case 4: + entry[afirst ? 0 : 3] = (png_byte)alpha; + case 3: + entry[afirst + (2 ^ bgr)] = (png_byte)blue; + entry[afirst + 1] = (png_byte)green; + entry[afirst + bgr] = (png_byte)red; + break; + + case 2: + entry[1 ^ afirst] = (png_byte)alpha; + case 1: + entry[afirst] = (png_byte)green; + break; + + default: + break; + } + } + +# ifdef afirst +# undef afirst +# endif +# ifdef bgr +# undef bgr +# endif + } +} + +static int +make_gray_file_colormap(png_image_read_control *display) +{ + unsigned int i; + + for (i=0; i<256; ++i) + png_create_colormap_entry(display, i, i, i, i, 255, P_FILE); + + return i; +} + +static int +make_gray_colormap(png_image_read_control *display) +{ + unsigned int i; + + for (i=0; i<256; ++i) + png_create_colormap_entry(display, i, i, i, i, 255, P_sRGB); + + return i; +} +#define PNG_GRAY_COLORMAP_ENTRIES 256 + +static int +make_ga_colormap(png_image_read_control *display) +{ + unsigned int i, a; + + /* Alpha is retained, the output will be a color-map with entries + * selected by six levels of alpha. One transparent entry, 6 gray + * levels for all the intermediate alpha values, leaving 230 entries + * for the opaque grays. The color-map entries are the six values + * [0..5]*51, the GA processing uses PNG_DIV51(value) to find the + * relevant entry. + * + * if (alpha > 229) // opaque + * { + * // The 231 entries are selected to make the math below work: + * base = 0; + * entry = (231 * gray + 128) >> 8; + * } + * else if (alpha < 26) // transparent + * { + * base = 231; + * entry = 0; + * } + * else // partially opaque + * { + * base = 226 + 6 * PNG_DIV51(alpha); + * entry = PNG_DIV51(gray); + * } + */ + i = 0; + while (i < 231) + { + unsigned int gray = (i * 256 + 115) / 231; + png_create_colormap_entry(display, i++, gray, gray, gray, 255, P_sRGB); + } + + /* 255 is used here for the component values for consistency with the code + * that undoes premultiplication in pngwrite.c. + */ + png_create_colormap_entry(display, i++, 255, 255, 255, 0, P_sRGB); + + for (a=1; a<5; ++a) + { + unsigned int g; + + for (g=0; g<6; ++g) + png_create_colormap_entry(display, i++, g*51, g*51, g*51, a*51, + P_sRGB); + } + + return i; +} + +#define PNG_GA_COLORMAP_ENTRIES 256 + +static int +make_rgb_colormap(png_image_read_control *display) +{ + unsigned int i, r; + + /* Build a 6x6x6 opaque RGB cube */ + for (i=r=0; r<6; ++r) + { + unsigned int g; + + for (g=0; g<6; ++g) + { + unsigned int b; + + for (b=0; b<6; ++b) + png_create_colormap_entry(display, i++, r*51, g*51, b*51, 255, + P_sRGB); + } + } + + return i; +} + +#define PNG_RGB_COLORMAP_ENTRIES 216 + +/* Return a palette index to the above palette given three 8-bit sRGB values. */ +#define PNG_RGB_INDEX(r,g,b) \ + ((png_byte)(6 * (6 * PNG_DIV51(r) + PNG_DIV51(g)) + PNG_DIV51(b))) + +static int +png_image_read_colormap(png_voidp argument) +{ + png_image_read_control *display = + png_voidcast(png_image_read_control*, argument); + const png_imagep image = display->image; + + const png_structrp png_ptr = image->opaque->png_ptr; + const png_uint_32 output_format = image->format; + const int output_encoding = (output_format & PNG_FORMAT_FLAG_LINEAR) != 0 ? + P_LINEAR : P_sRGB; + + unsigned int cmap_entries; + unsigned int output_processing; /* Output processing option */ + unsigned int data_encoding = P_NOTSET; /* Encoding libpng must produce */ + + /* Background information; the background color and the index of this color + * in the color-map if it exists (else 256). + */ + unsigned int background_index = 256; + png_uint_32 back_r, back_g, back_b; + + /* Flags to accumulate things that need to be done to the input. */ + int expand_tRNS = 0; + + /* Exclude the NYI feature of compositing onto a color-mapped buffer; it is + * very difficult to do, the results look awful, and it is difficult to see + * what possible use it is because the application can't control the + * color-map. + */ + if (((png_ptr->color_type & PNG_COLOR_MASK_ALPHA) != 0 || + png_ptr->num_trans > 0) /* alpha in input */ && + ((output_format & PNG_FORMAT_FLAG_ALPHA) == 0) /* no alpha in output */) + { + if (output_encoding == P_LINEAR) /* compose on black */ + back_b = back_g = back_r = 0; + + else if (display->background == NULL /* no way to remove it */) + png_error(png_ptr, + "a background color must be supplied to remove alpha/transparency"); + + /* Get a copy of the background color (this avoids repeating the checks + * below.) The encoding is 8-bit sRGB or 16-bit linear, depending on the + * output format. + */ + else + { + back_g = display->background->green; + if ((output_format & PNG_FORMAT_FLAG_COLOR) != 0) + { + back_r = display->background->red; + back_b = display->background->blue; + } + else + back_b = back_r = back_g; + } + } + + else if (output_encoding == P_LINEAR) + back_b = back_r = back_g = 65535; + + else + back_b = back_r = back_g = 255; + + /* Default the input file gamma if required - this is necessary because + * libpng assumes that if no gamma information is present the data is in the + * output format, but the simplified API deduces the gamma from the input + * format. + */ + if ((png_ptr->colorspace.flags & PNG_COLORSPACE_HAVE_GAMMA) == 0) + { + /* Do this directly, not using the png_colorspace functions, to ensure + * that it happens even if the colorspace is invalid (though probably if + * it is the setting will be ignored) Note that the same thing can be + * achieved at the application interface with png_set_gAMA. + */ + if (png_ptr->bit_depth == 16 && + (image->flags & PNG_IMAGE_FLAG_16BIT_sRGB) == 0) + png_ptr->colorspace.gamma = PNG_GAMMA_LINEAR; + + else + png_ptr->colorspace.gamma = PNG_GAMMA_sRGB_INVERSE; + + png_ptr->colorspace.flags |= PNG_COLORSPACE_HAVE_GAMMA; + } + + /* Decide what to do based on the PNG color type of the input data. The + * utility function png_create_colormap_entry deals with most aspects of the + * output transformations; this code works out how to produce bytes of + * color-map entries from the original format. + */ + switch (png_ptr->color_type) + { + case PNG_COLOR_TYPE_GRAY: + if (png_ptr->bit_depth <= 8) + { + /* There at most 256 colors in the output, regardless of + * transparency. + */ + unsigned int step, i, val, trans = 256/*ignore*/, back_alpha = 0; + + cmap_entries = 1U << png_ptr->bit_depth; + if (cmap_entries > image->colormap_entries) + png_error(png_ptr, "gray[8] color-map: too few entries"); + + step = 255 / (cmap_entries - 1); + output_processing = PNG_CMAP_NONE; + + /* If there is a tRNS chunk then this either selects a transparent + * value or, if the output has no alpha, the background color. + */ + if (png_ptr->num_trans > 0) + { + trans = png_ptr->trans_color.gray; + + if ((output_format & PNG_FORMAT_FLAG_ALPHA) == 0) + back_alpha = output_encoding == P_LINEAR ? 65535 : 255; + } + + /* png_create_colormap_entry just takes an RGBA and writes the + * corresponding color-map entry using the format from 'image', + * including the required conversion to sRGB or linear as + * appropriate. The input values are always either sRGB (if the + * gamma correction flag is 0) or 0..255 scaled file encoded values + * (if the function must gamma correct them). + */ + for (i=val=0; ibit_depth < 8) + png_set_packing(png_ptr); + } + + else /* bit depth is 16 */ + { + /* The 16-bit input values can be converted directly to 8-bit gamma + * encoded values; however, if a tRNS chunk is present 257 color-map + * entries are required. This means that the extra entry requires + * special processing; add an alpha channel, sacrifice gray level + * 254 and convert transparent (alpha==0) entries to that. + * + * Use libpng to chop the data to 8 bits. Convert it to sRGB at the + * same time to minimize quality loss. If a tRNS chunk is present + * this means libpng must handle it too; otherwise it is impossible + * to do the exact match on the 16-bit value. + * + * If the output has no alpha channel *and* the background color is + * gray then it is possible to let libpng handle the substitution by + * ensuring that the corresponding gray level matches the background + * color exactly. + */ + data_encoding = P_sRGB; + + if (PNG_GRAY_COLORMAP_ENTRIES > image->colormap_entries) + png_error(png_ptr, "gray[16] color-map: too few entries"); + + cmap_entries = make_gray_colormap(display); + + if (png_ptr->num_trans > 0) + { + unsigned int back_alpha; + + if ((output_format & PNG_FORMAT_FLAG_ALPHA) != 0) + back_alpha = 0; + + else + { + if (back_r == back_g && back_g == back_b) + { + /* Background is gray; no special processing will be + * required. + */ + png_color_16 c; + png_uint_32 gray = back_g; + + if (output_encoding == P_LINEAR) + { + gray = PNG_sRGB_FROM_LINEAR(gray * 255); + + /* And make sure the corresponding palette entry + * matches. + */ + png_create_colormap_entry(display, gray, back_g, back_g, + back_g, 65535, P_LINEAR); + } + + /* The background passed to libpng, however, must be the + * sRGB value. + */ + c.index = 0; /*unused*/ + c.gray = c.red = c.green = c.blue = (png_uint_16)gray; + + /* NOTE: does this work without expanding tRNS to alpha? + * It should be the color->gray case below apparently + * doesn't. + */ + png_set_background_fixed(png_ptr, &c, + PNG_BACKGROUND_GAMMA_SCREEN, 0/*need_expand*/, + 0/*gamma: not used*/); + + output_processing = PNG_CMAP_NONE; + break; + } + + back_alpha = output_encoding == P_LINEAR ? 65535 : 255; + } + + /* output_processing means that the libpng-processed row will be + * 8-bit GA and it has to be processing to single byte color-map + * values. Entry 254 is replaced by either a completely + * transparent entry or by the background color at full + * precision (and the background color is not a simple gray + * level in this case.) + */ + expand_tRNS = 1; + output_processing = PNG_CMAP_TRANS; + background_index = 254; + + /* And set (overwrite) color-map entry 254 to the actual + * background color at full precision. + */ + png_create_colormap_entry(display, 254, back_r, back_g, back_b, + back_alpha, output_encoding); + } + + else + output_processing = PNG_CMAP_NONE; + } + break; + + case PNG_COLOR_TYPE_GRAY_ALPHA: + /* 8-bit or 16-bit PNG with two channels - gray and alpha. A minimum + * of 65536 combinations. If, however, the alpha channel is to be + * removed there are only 256 possibilities if the background is gray. + * (Otherwise there is a subset of the 65536 possibilities defined by + * the triangle between black, white and the background color.) + * + * Reduce 16-bit files to 8-bit and sRGB encode the result. No need to + * worry about tRNS matching - tRNS is ignored if there is an alpha + * channel. + */ + data_encoding = P_sRGB; + + if ((output_format & PNG_FORMAT_FLAG_ALPHA) != 0) + { + if (PNG_GA_COLORMAP_ENTRIES > image->colormap_entries) + png_error(png_ptr, "gray+alpha color-map: too few entries"); + + cmap_entries = make_ga_colormap(display); + + background_index = PNG_CMAP_GA_BACKGROUND; + output_processing = PNG_CMAP_GA; + } + + else /* alpha is removed */ + { + /* Alpha must be removed as the PNG data is processed when the + * background is a color because the G and A channels are + * independent and the vector addition (non-parallel vectors) is a + * 2-D problem. + * + * This can be reduced to the same algorithm as above by making a + * colormap containing gray levels (for the opaque grays), a + * background entry (for a transparent pixel) and a set of four six + * level color values, one set for each intermediate alpha value. + * See the comments in make_ga_colormap for how this works in the + * per-pixel processing. + * + * If the background is gray, however, we only need a 256 entry gray + * level color map. It is sufficient to make the entry generated + * for the background color be exactly the color specified. + */ + if ((output_format & PNG_FORMAT_FLAG_COLOR) == 0 || + (back_r == back_g && back_g == back_b)) + { + /* Background is gray; no special processing will be required. */ + png_color_16 c; + png_uint_32 gray = back_g; + + if (PNG_GRAY_COLORMAP_ENTRIES > image->colormap_entries) + png_error(png_ptr, "gray-alpha color-map: too few entries"); + + cmap_entries = make_gray_colormap(display); + + if (output_encoding == P_LINEAR) + { + gray = PNG_sRGB_FROM_LINEAR(gray * 255); + + /* And make sure the corresponding palette entry matches. */ + png_create_colormap_entry(display, gray, back_g, back_g, + back_g, 65535, P_LINEAR); + } + + /* The background passed to libpng, however, must be the sRGB + * value. + */ + c.index = 0; /*unused*/ + c.gray = c.red = c.green = c.blue = (png_uint_16)gray; + + png_set_background_fixed(png_ptr, &c, + PNG_BACKGROUND_GAMMA_SCREEN, 0/*need_expand*/, + 0/*gamma: not used*/); + + output_processing = PNG_CMAP_NONE; + } + + else + { + png_uint_32 i, a; + + /* This is the same as png_make_ga_colormap, above, except that + * the entries are all opaque. + */ + if (PNG_GA_COLORMAP_ENTRIES > image->colormap_entries) + png_error(png_ptr, "ga-alpha color-map: too few entries"); + + i = 0; + while (i < 231) + { + png_uint_32 gray = (i * 256 + 115) / 231; + png_create_colormap_entry(display, i++, gray, gray, gray, + 255, P_sRGB); + } + + /* NOTE: this preserves the full precision of the application + * background color. + */ + background_index = i; + png_create_colormap_entry(display, i++, back_r, back_g, back_b, + output_encoding == P_LINEAR ? 65535U : 255U, output_encoding); + + /* For non-opaque input composite on the sRGB background - this + * requires inverting the encoding for each component. The input + * is still converted to the sRGB encoding because this is a + * reasonable approximate to the logarithmic curve of human + * visual sensitivity, at least over the narrow range which PNG + * represents. Consequently 'G' is always sRGB encoded, while + * 'A' is linear. We need the linear background colors. + */ + if (output_encoding == P_sRGB) /* else already linear */ + { + /* This may produce a value not exactly matching the + * background, but that's ok because these numbers are only + * used when alpha != 0 + */ + back_r = png_sRGB_table[back_r]; + back_g = png_sRGB_table[back_g]; + back_b = png_sRGB_table[back_b]; + } + + for (a=1; a<5; ++a) + { + unsigned int g; + + /* PNG_sRGB_FROM_LINEAR expects a 16-bit linear value scaled + * by an 8-bit alpha value (0..255). + */ + png_uint_32 alpha = 51 * a; + png_uint_32 back_rx = (255-alpha) * back_r; + png_uint_32 back_gx = (255-alpha) * back_g; + png_uint_32 back_bx = (255-alpha) * back_b; + + for (g=0; g<6; ++g) + { + png_uint_32 gray = png_sRGB_table[g*51] * alpha; + + png_create_colormap_entry(display, i++, + PNG_sRGB_FROM_LINEAR(gray + back_rx), + PNG_sRGB_FROM_LINEAR(gray + back_gx), + PNG_sRGB_FROM_LINEAR(gray + back_bx), 255, P_sRGB); + } + } + + cmap_entries = i; + output_processing = PNG_CMAP_GA; + } + } + break; + + case PNG_COLOR_TYPE_RGB: + case PNG_COLOR_TYPE_RGB_ALPHA: + /* Exclude the case where the output is gray; we can always handle this + * with the cases above. + */ + if ((output_format & PNG_FORMAT_FLAG_COLOR) == 0) + { + /* The color-map will be grayscale, so we may as well convert the + * input RGB values to a simple grayscale and use the grayscale + * code above. + * + * NOTE: calling this apparently damages the recognition of the + * transparent color in background color handling; call + * png_set_tRNS_to_alpha before png_set_background_fixed. + */ + png_set_rgb_to_gray_fixed(png_ptr, PNG_ERROR_ACTION_NONE, -1, + -1); + data_encoding = P_sRGB; + + /* The output will now be one or two 8-bit gray or gray+alpha + * channels. The more complex case arises when the input has alpha. + */ + if ((png_ptr->color_type == PNG_COLOR_TYPE_RGB_ALPHA || + png_ptr->num_trans > 0) && + (output_format & PNG_FORMAT_FLAG_ALPHA) != 0) + { + /* Both input and output have an alpha channel, so no background + * processing is required; just map the GA bytes to the right + * color-map entry. + */ + expand_tRNS = 1; + + if (PNG_GA_COLORMAP_ENTRIES > image->colormap_entries) + png_error(png_ptr, "rgb[ga] color-map: too few entries"); + + cmap_entries = make_ga_colormap(display); + background_index = PNG_CMAP_GA_BACKGROUND; + output_processing = PNG_CMAP_GA; + } + + else + { + /* Either the input or the output has no alpha channel, so there + * will be no non-opaque pixels in the color-map; it will just be + * grayscale. + */ + if (PNG_GRAY_COLORMAP_ENTRIES > image->colormap_entries) + png_error(png_ptr, "rgb[gray] color-map: too few entries"); + + /* Ideally this code would use libpng to do the gamma correction, + * but if an input alpha channel is to be removed we will hit the + * libpng bug in gamma+compose+rgb-to-gray (the double gamma + * correction bug). Fix this by dropping the gamma correction in + * this case and doing it in the palette; this will result in + * duplicate palette entries, but that's better than the + * alternative of double gamma correction. + */ + if ((png_ptr->color_type == PNG_COLOR_TYPE_RGB_ALPHA || + png_ptr->num_trans > 0) && + png_gamma_not_sRGB(png_ptr->colorspace.gamma) != 0) + { + cmap_entries = make_gray_file_colormap(display); + data_encoding = P_FILE; + } + + else + cmap_entries = make_gray_colormap(display); + + /* But if the input has alpha or transparency it must be removed + */ + if (png_ptr->color_type == PNG_COLOR_TYPE_RGB_ALPHA || + png_ptr->num_trans > 0) + { + png_color_16 c; + png_uint_32 gray = back_g; + + /* We need to ensure that the application background exists in + * the colormap and that completely transparent pixels map to + * it. Achieve this simply by ensuring that the entry + * selected for the background really is the background color. + */ + if (data_encoding == P_FILE) /* from the fixup above */ + { + /* The app supplied a gray which is in output_encoding, we + * need to convert it to a value of the input (P_FILE) + * encoding then set this palette entry to the required + * output encoding. + */ + if (output_encoding == P_sRGB) + gray = png_sRGB_table[gray]; /* now P_LINEAR */ + + gray = PNG_DIV257(png_gamma_16bit_correct(gray, + png_ptr->colorspace.gamma)); /* now P_FILE */ + + /* And make sure the corresponding palette entry contains + * exactly the required sRGB value. + */ + png_create_colormap_entry(display, gray, back_g, back_g, + back_g, 0/*unused*/, output_encoding); + } + + else if (output_encoding == P_LINEAR) + { + gray = PNG_sRGB_FROM_LINEAR(gray * 255); + + /* And make sure the corresponding palette entry matches. + */ + png_create_colormap_entry(display, gray, back_g, back_g, + back_g, 0/*unused*/, P_LINEAR); + } + + /* The background passed to libpng, however, must be the + * output (normally sRGB) value. + */ + c.index = 0; /*unused*/ + c.gray = c.red = c.green = c.blue = (png_uint_16)gray; + + /* NOTE: the following is apparently a bug in libpng. Without + * it the transparent color recognition in + * png_set_background_fixed seems to go wrong. + */ + expand_tRNS = 1; + png_set_background_fixed(png_ptr, &c, + PNG_BACKGROUND_GAMMA_SCREEN, 0/*need_expand*/, + 0/*gamma: not used*/); + } + + output_processing = PNG_CMAP_NONE; + } + } + + else /* output is color */ + { + /* We could use png_quantize here so long as there is no transparent + * color or alpha; png_quantize ignores alpha. Easier overall just + * to do it once and using PNG_DIV51 on the 6x6x6 reduced RGB cube. + * Consequently we always want libpng to produce sRGB data. + */ + data_encoding = P_sRGB; + + /* Is there any transparency or alpha? */ + if (png_ptr->color_type == PNG_COLOR_TYPE_RGB_ALPHA || + png_ptr->num_trans > 0) + { + /* Is there alpha in the output too? If so all four channels are + * processed into a special RGB cube with alpha support. + */ + if ((output_format & PNG_FORMAT_FLAG_ALPHA) != 0) + { + png_uint_32 r; + + if (PNG_RGB_COLORMAP_ENTRIES+1+27 > image->colormap_entries) + png_error(png_ptr, "rgb+alpha color-map: too few entries"); + + cmap_entries = make_rgb_colormap(display); + + /* Add a transparent entry. */ + png_create_colormap_entry(display, cmap_entries, 255, 255, + 255, 0, P_sRGB); + + /* This is stored as the background index for the processing + * algorithm. + */ + background_index = cmap_entries++; + + /* Add 27 r,g,b entries each with alpha 0.5. */ + for (r=0; r<256; r = (r << 1) | 0x7f) + { + png_uint_32 g; + + for (g=0; g<256; g = (g << 1) | 0x7f) + { + png_uint_32 b; + + /* This generates components with the values 0, 127 and + * 255 + */ + for (b=0; b<256; b = (b << 1) | 0x7f) + png_create_colormap_entry(display, cmap_entries++, + r, g, b, 128, P_sRGB); + } + } + + expand_tRNS = 1; + output_processing = PNG_CMAP_RGB_ALPHA; + } + + else + { + /* Alpha/transparency must be removed. The background must + * exist in the color map (achieved by setting adding it after + * the 666 color-map). If the standard processing code will + * pick up this entry automatically that's all that is + * required; libpng can be called to do the background + * processing. + */ + unsigned int sample_size = + PNG_IMAGE_SAMPLE_SIZE(output_format); + png_uint_32 r, g, b; /* sRGB background */ + + if (PNG_RGB_COLORMAP_ENTRIES+1+27 > image->colormap_entries) + png_error(png_ptr, "rgb-alpha color-map: too few entries"); + + cmap_entries = make_rgb_colormap(display); + + png_create_colormap_entry(display, cmap_entries, back_r, + back_g, back_b, 0/*unused*/, output_encoding); + + if (output_encoding == P_LINEAR) + { + r = PNG_sRGB_FROM_LINEAR(back_r * 255); + g = PNG_sRGB_FROM_LINEAR(back_g * 255); + b = PNG_sRGB_FROM_LINEAR(back_b * 255); + } + + else + { + r = back_r; + g = back_g; + b = back_g; + } + + /* Compare the newly-created color-map entry with the one the + * PNG_CMAP_RGB algorithm will use. If the two entries don't + * match, add the new one and set this as the background + * index. + */ + if (memcmp((png_const_bytep)display->colormap + + sample_size * cmap_entries, + (png_const_bytep)display->colormap + + sample_size * PNG_RGB_INDEX(r,g,b), + sample_size) != 0) + { + /* The background color must be added. */ + background_index = cmap_entries++; + + /* Add 27 r,g,b entries each with created by composing with + * the background at alpha 0.5. + */ + for (r=0; r<256; r = (r << 1) | 0x7f) + { + for (g=0; g<256; g = (g << 1) | 0x7f) + { + /* This generates components with the values 0, 127 + * and 255 + */ + for (b=0; b<256; b = (b << 1) | 0x7f) + png_create_colormap_entry(display, cmap_entries++, + png_colormap_compose(display, r, P_sRGB, 128, + back_r, output_encoding), + png_colormap_compose(display, g, P_sRGB, 128, + back_g, output_encoding), + png_colormap_compose(display, b, P_sRGB, 128, + back_b, output_encoding), + 0/*unused*/, output_encoding); + } + } + + expand_tRNS = 1; + output_processing = PNG_CMAP_RGB_ALPHA; + } + + else /* background color is in the standard color-map */ + { + png_color_16 c; + + c.index = 0; /*unused*/ + c.red = (png_uint_16)back_r; + c.gray = c.green = (png_uint_16)back_g; + c.blue = (png_uint_16)back_b; + + png_set_background_fixed(png_ptr, &c, + PNG_BACKGROUND_GAMMA_SCREEN, 0/*need_expand*/, + 0/*gamma: not used*/); + + output_processing = PNG_CMAP_RGB; + } + } + } + + else /* no alpha or transparency in the input */ + { + /* Alpha in the output is irrelevant, simply map the opaque input + * pixels to the 6x6x6 color-map. + */ + if (PNG_RGB_COLORMAP_ENTRIES > image->colormap_entries) + png_error(png_ptr, "rgb color-map: too few entries"); + + cmap_entries = make_rgb_colormap(display); + output_processing = PNG_CMAP_RGB; + } + } + break; + + case PNG_COLOR_TYPE_PALETTE: + /* It's already got a color-map. It may be necessary to eliminate the + * tRNS entries though. + */ + { + unsigned int num_trans = png_ptr->num_trans; + png_const_bytep trans = num_trans > 0 ? png_ptr->trans_alpha : NULL; + png_const_colorp colormap = png_ptr->palette; + const int do_background = trans != NULL && + (output_format & PNG_FORMAT_FLAG_ALPHA) == 0; + unsigned int i; + + /* Just in case: */ + if (trans == NULL) + num_trans = 0; + + output_processing = PNG_CMAP_NONE; + data_encoding = P_FILE; /* Don't change from color-map indices */ + cmap_entries = png_ptr->num_palette; + if (cmap_entries > 256) + cmap_entries = 256; + + if (cmap_entries > image->colormap_entries) + png_error(png_ptr, "palette color-map: too few entries"); + + for (i=0; i < cmap_entries; ++i) + { + if (do_background != 0 && i < num_trans && trans[i] < 255) + { + if (trans[i] == 0) + png_create_colormap_entry(display, i, back_r, back_g, + back_b, 0, output_encoding); + + else + { + /* Must compose the PNG file color in the color-map entry + * on the sRGB color in 'back'. + */ + png_create_colormap_entry(display, i, + png_colormap_compose(display, colormap[i].red, P_FILE, + trans[i], back_r, output_encoding), + png_colormap_compose(display, colormap[i].green, P_FILE, + trans[i], back_g, output_encoding), + png_colormap_compose(display, colormap[i].blue, P_FILE, + trans[i], back_b, output_encoding), + output_encoding == P_LINEAR ? trans[i] * 257U : + trans[i], + output_encoding); + } + } + + else + png_create_colormap_entry(display, i, colormap[i].red, + colormap[i].green, colormap[i].blue, + i < num_trans ? trans[i] : 255U, P_FILE/*8-bit*/); + } + + /* The PNG data may have indices packed in fewer than 8 bits, it + * must be expanded if so. + */ + if (png_ptr->bit_depth < 8) + png_set_packing(png_ptr); + } + break; + + default: + png_error(png_ptr, "invalid PNG color type"); + /*NOT REACHED*/ + break; + } + + /* Now deal with the output processing */ + if (expand_tRNS != 0 && png_ptr->num_trans > 0 && + (png_ptr->color_type & PNG_COLOR_MASK_ALPHA) == 0) + png_set_tRNS_to_alpha(png_ptr); + + switch (data_encoding) + { + default: + png_error(png_ptr, "bad data option (internal error)"); + break; + + case P_sRGB: + /* Change to 8-bit sRGB */ + png_set_alpha_mode_fixed(png_ptr, PNG_ALPHA_PNG, PNG_GAMMA_sRGB); + /* FALL THROUGH */ + + case P_FILE: + if (png_ptr->bit_depth > 8) + png_set_scale_16(png_ptr); + break; + } + + if (cmap_entries > 256 || cmap_entries > image->colormap_entries) + png_error(png_ptr, "color map overflow (BAD internal error)"); + + image->colormap_entries = cmap_entries; + + /* Double check using the recorded background index */ + switch (output_processing) + { + case PNG_CMAP_NONE: + if (background_index != PNG_CMAP_NONE_BACKGROUND) + goto bad_background; + break; + + case PNG_CMAP_GA: + if (background_index != PNG_CMAP_GA_BACKGROUND) + goto bad_background; + break; + + case PNG_CMAP_TRANS: + if (background_index >= cmap_entries || + background_index != PNG_CMAP_TRANS_BACKGROUND) + goto bad_background; + break; + + case PNG_CMAP_RGB: + if (background_index != PNG_CMAP_RGB_BACKGROUND) + goto bad_background; + break; + + case PNG_CMAP_RGB_ALPHA: + if (background_index != PNG_CMAP_RGB_ALPHA_BACKGROUND) + goto bad_background; + break; + + default: + png_error(png_ptr, "bad processing option (internal error)"); + + bad_background: + png_error(png_ptr, "bad background index (internal error)"); + } + + display->colormap_processing = output_processing; + + return 1/*ok*/; +} + +/* The final part of the color-map read called from png_image_finish_read. */ +static int +png_image_read_and_map(png_voidp argument) +{ + png_image_read_control *display = png_voidcast(png_image_read_control*, + argument); + png_imagep image = display->image; + png_structrp png_ptr = image->opaque->png_ptr; + int passes; + + /* Called when the libpng data must be transformed into the color-mapped + * form. There is a local row buffer in display->local and this routine must + * do the interlace handling. + */ + switch (png_ptr->interlaced) + { + case PNG_INTERLACE_NONE: + passes = 1; + break; + + case PNG_INTERLACE_ADAM7: + passes = PNG_INTERLACE_ADAM7_PASSES; + break; + + default: + png_error(png_ptr, "unknown interlace type"); + } + + { + png_uint_32 height = image->height; + png_uint_32 width = image->width; + int proc = display->colormap_processing; + png_bytep first_row = png_voidcast(png_bytep, display->first_row); + ptrdiff_t step_row = display->row_bytes; + int pass; + + for (pass = 0; pass < passes; ++pass) + { + unsigned int startx, stepx, stepy; + png_uint_32 y; + + if (png_ptr->interlaced == PNG_INTERLACE_ADAM7) + { + /* The row may be empty for a short image: */ + if (PNG_PASS_COLS(width, pass) == 0) + continue; + + startx = PNG_PASS_START_COL(pass); + stepx = PNG_PASS_COL_OFFSET(pass); + y = PNG_PASS_START_ROW(pass); + stepy = PNG_PASS_ROW_OFFSET(pass); + } + + else + { + y = 0; + startx = 0; + stepx = stepy = 1; + } + + for (; ylocal_row); + png_bytep outrow = first_row + y * step_row; + png_const_bytep end_row = outrow + width; + + /* Read read the libpng data into the temporary buffer. */ + png_read_row(png_ptr, inrow, NULL); + + /* Now process the row according to the processing option, note + * that the caller verifies that the format of the libpng output + * data is as required. + */ + outrow += startx; + switch (proc) + { + case PNG_CMAP_GA: + for (; outrow < end_row; outrow += stepx) + { + /* The data is always in the PNG order */ + unsigned int gray = *inrow++; + unsigned int alpha = *inrow++; + unsigned int entry; + + /* NOTE: this code is copied as a comment in + * make_ga_colormap above. Please update the + * comment if you change this code! + */ + if (alpha > 229) /* opaque */ + { + entry = (231 * gray + 128) >> 8; + } + else if (alpha < 26) /* transparent */ + { + entry = 231; + } + else /* partially opaque */ + { + entry = 226 + 6 * PNG_DIV51(alpha) + PNG_DIV51(gray); + } + + *outrow = (png_byte)entry; + } + break; + + case PNG_CMAP_TRANS: + for (; outrow < end_row; outrow += stepx) + { + png_byte gray = *inrow++; + png_byte alpha = *inrow++; + + if (alpha == 0) + *outrow = PNG_CMAP_TRANS_BACKGROUND; + + else if (gray != PNG_CMAP_TRANS_BACKGROUND) + *outrow = gray; + + else + *outrow = (png_byte)(PNG_CMAP_TRANS_BACKGROUND+1); + } + break; + + case PNG_CMAP_RGB: + for (; outrow < end_row; outrow += stepx) + { + *outrow = PNG_RGB_INDEX(inrow[0], inrow[1], inrow[2]); + inrow += 3; + } + break; + + case PNG_CMAP_RGB_ALPHA: + for (; outrow < end_row; outrow += stepx) + { + unsigned int alpha = inrow[3]; + + /* Because the alpha entries only hold alpha==0.5 values + * split the processing at alpha==0.25 (64) and 0.75 + * (196). + */ + + if (alpha >= 196) + *outrow = PNG_RGB_INDEX(inrow[0], inrow[1], + inrow[2]); + + else if (alpha < 64) + *outrow = PNG_CMAP_RGB_ALPHA_BACKGROUND; + + else + { + /* Likewise there are three entries for each of r, g + * and b. We could select the entry by popcount on + * the top two bits on those architectures that + * support it, this is what the code below does, + * crudely. + */ + unsigned int back_i = PNG_CMAP_RGB_ALPHA_BACKGROUND+1; + + /* Here are how the values map: + * + * 0x00 .. 0x3f -> 0 + * 0x40 .. 0xbf -> 1 + * 0xc0 .. 0xff -> 2 + * + * So, as above with the explicit alpha checks, the + * breakpoints are at 64 and 196. + */ + if (inrow[0] & 0x80) back_i += 9; /* red */ + if (inrow[0] & 0x40) back_i += 9; + if (inrow[0] & 0x80) back_i += 3; /* green */ + if (inrow[0] & 0x40) back_i += 3; + if (inrow[0] & 0x80) back_i += 1; /* blue */ + if (inrow[0] & 0x40) back_i += 1; + + *outrow = (png_byte)back_i; + } + + inrow += 4; + } + break; + + default: + break; + } + } + } + } + + return 1; +} + +static int +png_image_read_colormapped(png_voidp argument) +{ + png_image_read_control *display = png_voidcast(png_image_read_control*, + argument); + png_imagep image = display->image; + png_controlp control = image->opaque; + png_structrp png_ptr = control->png_ptr; + png_inforp info_ptr = control->info_ptr; + + int passes = 0; /* As a flag */ + + PNG_SKIP_CHUNKS(png_ptr); + + /* Update the 'info' structure and make sure the result is as required; first + * make sure to turn on the interlace handling if it will be required + * (because it can't be turned on *after* the call to png_read_update_info!) + */ + if (display->colormap_processing == PNG_CMAP_NONE) + passes = png_set_interlace_handling(png_ptr); + + png_read_update_info(png_ptr, info_ptr); + + /* The expected output can be deduced from the colormap_processing option. */ + switch (display->colormap_processing) + { + case PNG_CMAP_NONE: + /* Output must be one channel and one byte per pixel, the output + * encoding can be anything. + */ + if ((info_ptr->color_type == PNG_COLOR_TYPE_PALETTE || + info_ptr->color_type == PNG_COLOR_TYPE_GRAY) && + info_ptr->bit_depth == 8) + break; + + goto bad_output; + + case PNG_CMAP_TRANS: + case PNG_CMAP_GA: + /* Output must be two channels and the 'G' one must be sRGB, the latter + * can be checked with an exact number because it should have been set + * to this number above! + */ + if (info_ptr->color_type == PNG_COLOR_TYPE_GRAY_ALPHA && + info_ptr->bit_depth == 8 && + png_ptr->screen_gamma == PNG_GAMMA_sRGB && + image->colormap_entries == 256) + break; + + goto bad_output; + + case PNG_CMAP_RGB: + /* Output must be 8-bit sRGB encoded RGB */ + if (info_ptr->color_type == PNG_COLOR_TYPE_RGB && + info_ptr->bit_depth == 8 && + png_ptr->screen_gamma == PNG_GAMMA_sRGB && + image->colormap_entries == 216) + break; + + goto bad_output; + + case PNG_CMAP_RGB_ALPHA: + /* Output must be 8-bit sRGB encoded RGBA */ + if (info_ptr->color_type == PNG_COLOR_TYPE_RGB_ALPHA && + info_ptr->bit_depth == 8 && + png_ptr->screen_gamma == PNG_GAMMA_sRGB && + image->colormap_entries == 244 /* 216 + 1 + 27 */) + break; + + /* goto bad_output; */ + /* FALL THROUGH */ + + default: + bad_output: + png_error(png_ptr, "bad color-map processing (internal error)"); + } + + /* Now read the rows. Do this here if it is possible to read directly into + * the output buffer, otherwise allocate a local row buffer of the maximum + * size libpng requires and call the relevant processing routine safely. + */ + { + png_voidp first_row = display->buffer; + ptrdiff_t row_bytes = display->row_stride; + + /* The following expression is designed to work correctly whether it gives + * a signed or an unsigned result. + */ + if (row_bytes < 0) + { + char *ptr = png_voidcast(char*, first_row); + ptr += (image->height-1) * (-row_bytes); + first_row = png_voidcast(png_voidp, ptr); + } + + display->first_row = first_row; + display->row_bytes = row_bytes; + } + + if (passes == 0) + { + int result; + png_voidp row = png_malloc(png_ptr, png_get_rowbytes(png_ptr, info_ptr)); + + display->local_row = row; + result = png_safe_execute(image, png_image_read_and_map, display); + display->local_row = NULL; + png_free(png_ptr, row); + + return result; + } + + else + { + png_alloc_size_t row_bytes = display->row_bytes; + + while (--passes >= 0) + { + png_uint_32 y = image->height; + png_bytep row = png_voidcast(png_bytep, display->first_row); + + while (y-- > 0) + { + png_read_row(png_ptr, row, NULL); + row += row_bytes; + } + } + + return 1; + } +} + +/* Just the row reading part of png_image_read. */ +static int +png_image_read_composite(png_voidp argument) +{ + png_image_read_control *display = png_voidcast(png_image_read_control*, + argument); + png_imagep image = display->image; + png_structrp png_ptr = image->opaque->png_ptr; + int passes; + + switch (png_ptr->interlaced) + { + case PNG_INTERLACE_NONE: + passes = 1; + break; + + case PNG_INTERLACE_ADAM7: + passes = PNG_INTERLACE_ADAM7_PASSES; + break; + + default: + png_error(png_ptr, "unknown interlace type"); + } + + { + png_uint_32 height = image->height; + png_uint_32 width = image->width; + ptrdiff_t step_row = display->row_bytes; + unsigned int channels = + (image->format & PNG_FORMAT_FLAG_COLOR) != 0 ? 3 : 1; + int pass; + + for (pass = 0; pass < passes; ++pass) + { + unsigned int startx, stepx, stepy; + png_uint_32 y; + + if (png_ptr->interlaced == PNG_INTERLACE_ADAM7) + { + /* The row may be empty for a short image: */ + if (PNG_PASS_COLS(width, pass) == 0) + continue; + + startx = PNG_PASS_START_COL(pass) * channels; + stepx = PNG_PASS_COL_OFFSET(pass) * channels; + y = PNG_PASS_START_ROW(pass); + stepy = PNG_PASS_ROW_OFFSET(pass); + } + + else + { + y = 0; + startx = 0; + stepx = channels; + stepy = 1; + } + + for (; ylocal_row); + png_bytep outrow; + png_const_bytep end_row; + + /* Read the row, which is packed: */ + png_read_row(png_ptr, inrow, NULL); + + outrow = png_voidcast(png_bytep, display->first_row); + outrow += y * step_row; + end_row = outrow + width * channels; + + /* Now do the composition on each pixel in this row. */ + outrow += startx; + for (; outrow < end_row; outrow += stepx) + { + png_byte alpha = inrow[channels]; + + if (alpha > 0) /* else no change to the output */ + { + unsigned int c; + + for (c=0; cimage; + png_structrp png_ptr = image->opaque->png_ptr; + png_inforp info_ptr = image->opaque->info_ptr; + png_uint_32 height = image->height; + png_uint_32 width = image->width; + int pass, passes; + + /* Double check the convoluted logic below. We expect to get here with + * libpng doing rgb to gray and gamma correction but background processing + * left to the png_image_read_background function. The rows libpng produce + * might be 8 or 16-bit but should always have two channels; gray plus alpha. + */ + if ((png_ptr->transformations & PNG_RGB_TO_GRAY) == 0) + png_error(png_ptr, "lost rgb to gray"); + + if ((png_ptr->transformations & PNG_COMPOSE) != 0) + png_error(png_ptr, "unexpected compose"); + + if (png_get_channels(png_ptr, info_ptr) != 2) + png_error(png_ptr, "lost/gained channels"); + + /* Expect the 8-bit case to always remove the alpha channel */ + if ((image->format & PNG_FORMAT_FLAG_LINEAR) == 0 && + (image->format & PNG_FORMAT_FLAG_ALPHA) != 0) + png_error(png_ptr, "unexpected 8-bit transformation"); + + switch (png_ptr->interlaced) + { + case PNG_INTERLACE_NONE: + passes = 1; + break; + + case PNG_INTERLACE_ADAM7: + passes = PNG_INTERLACE_ADAM7_PASSES; + break; + + default: + png_error(png_ptr, "unknown interlace type"); + } + + /* Use direct access to info_ptr here because otherwise the simplified API + * would require PNG_EASY_ACCESS_SUPPORTED (just for this.) Note this is + * checking the value after libpng expansions, not the original value in the + * PNG. + */ + switch (info_ptr->bit_depth) + { + default: + png_error(png_ptr, "unexpected bit depth"); + break; + + case 8: + /* 8-bit sRGB gray values with an alpha channel; the alpha channel is + * to be removed by composing on a background: either the row if + * display->background is NULL or display->background->green if not. + * Unlike the code above ALPHA_OPTIMIZED has *not* been done. + */ + { + png_bytep first_row = png_voidcast(png_bytep, display->first_row); + ptrdiff_t step_row = display->row_bytes; + + for (pass = 0; pass < passes; ++pass) + { + png_bytep row = png_voidcast(png_bytep, + display->first_row); + unsigned int startx, stepx, stepy; + png_uint_32 y; + + if (png_ptr->interlaced == PNG_INTERLACE_ADAM7) + { + /* The row may be empty for a short image: */ + if (PNG_PASS_COLS(width, pass) == 0) + continue; + + startx = PNG_PASS_START_COL(pass); + stepx = PNG_PASS_COL_OFFSET(pass); + y = PNG_PASS_START_ROW(pass); + stepy = PNG_PASS_ROW_OFFSET(pass); + } + + else + { + y = 0; + startx = 0; + stepx = stepy = 1; + } + + if (display->background == NULL) + { + for (; ylocal_row); + png_bytep outrow = first_row + y * step_row; + png_const_bytep end_row = outrow + width; + + /* Read the row, which is packed: */ + png_read_row(png_ptr, inrow, NULL); + + /* Now do the composition on each pixel in this row. */ + outrow += startx; + for (; outrow < end_row; outrow += stepx) + { + png_byte alpha = inrow[1]; + + if (alpha > 0) /* else no change to the output */ + { + png_uint_32 component = inrow[0]; + + if (alpha < 255) /* else just use component */ + { + /* Since PNG_OPTIMIZED_ALPHA was not set it is + * necessary to invert the sRGB transfer + * function and multiply the alpha out. + */ + component = png_sRGB_table[component] * alpha; + component += png_sRGB_table[outrow[0]] * + (255-alpha); + component = PNG_sRGB_FROM_LINEAR(component); + } + + outrow[0] = (png_byte)component; + } + + inrow += 2; /* gray and alpha channel */ + } + } + } + + else /* constant background value */ + { + png_byte background8 = display->background->green; + png_uint_16 background = png_sRGB_table[background8]; + + for (; ylocal_row); + png_bytep outrow = first_row + y * step_row; + png_const_bytep end_row = outrow + width; + + /* Read the row, which is packed: */ + png_read_row(png_ptr, inrow, NULL); + + /* Now do the composition on each pixel in this row. */ + outrow += startx; + for (; outrow < end_row; outrow += stepx) + { + png_byte alpha = inrow[1]; + + if (alpha > 0) /* else use background */ + { + png_uint_32 component = inrow[0]; + + if (alpha < 255) /* else just use component */ + { + component = png_sRGB_table[component] * alpha; + component += background * (255-alpha); + component = PNG_sRGB_FROM_LINEAR(component); + } + + outrow[0] = (png_byte)component; + } + + else + outrow[0] = background8; + + inrow += 2; /* gray and alpha channel */ + } + + row += display->row_bytes; + } + } + } + } + break; + + case 16: + /* 16-bit linear with pre-multiplied alpha; the pre-multiplication must + * still be done and, maybe, the alpha channel removed. This code also + * handles the alpha-first option. + */ + { + png_uint_16p first_row = png_voidcast(png_uint_16p, + display->first_row); + /* The division by two is safe because the caller passed in a + * stride which was multiplied by 2 (below) to get row_bytes. + */ + ptrdiff_t step_row = display->row_bytes / 2; + int preserve_alpha = (image->format & PNG_FORMAT_FLAG_ALPHA) != 0; + unsigned int outchannels = 1+preserve_alpha; + int swap_alpha = 0; + +# ifdef PNG_SIMPLIFIED_READ_AFIRST_SUPPORTED + if (preserve_alpha != 0 && + (image->format & PNG_FORMAT_FLAG_AFIRST) != 0) + swap_alpha = 1; +# endif + + for (pass = 0; pass < passes; ++pass) + { + unsigned int startx, stepx, stepy; + png_uint_32 y; + + /* The 'x' start and step are adjusted to output components here. + */ + if (png_ptr->interlaced == PNG_INTERLACE_ADAM7) + { + /* The row may be empty for a short image: */ + if (PNG_PASS_COLS(width, pass) == 0) + continue; + + startx = PNG_PASS_START_COL(pass) * outchannels; + stepx = PNG_PASS_COL_OFFSET(pass) * outchannels; + y = PNG_PASS_START_ROW(pass); + stepy = PNG_PASS_ROW_OFFSET(pass); + } + + else + { + y = 0; + startx = 0; + stepx = outchannels; + stepy = 1; + } + + for (; ylocal_row), NULL); + inrow = png_voidcast(png_const_uint_16p, display->local_row); + + /* Now do the pre-multiplication on each pixel in this row. + */ + outrow += startx; + for (; outrow < end_row; outrow += stepx) + { + png_uint_32 component = inrow[0]; + png_uint_16 alpha = inrow[1]; + + if (alpha > 0) /* else 0 */ + { + if (alpha < 65535) /* else just use component */ + { + component *= alpha; + component += 32767; + component /= 65535; + } + } + + else + component = 0; + + outrow[swap_alpha] = (png_uint_16)component; + if (preserve_alpha != 0) + outrow[1 ^ swap_alpha] = alpha; + + inrow += 2; /* components and alpha channel */ + } + } + } + } + break; + } + + return 1; +} + +/* The guts of png_image_finish_read as a png_safe_execute callback. */ +static int +png_image_read_direct(png_voidp argument) +{ + png_image_read_control *display = png_voidcast(png_image_read_control*, + argument); + png_imagep image = display->image; + png_structrp png_ptr = image->opaque->png_ptr; + png_inforp info_ptr = image->opaque->info_ptr; + + png_uint_32 format = image->format; + int linear = (format & PNG_FORMAT_FLAG_LINEAR) != 0; + int do_local_compose = 0; + int do_local_background = 0; /* to avoid double gamma correction bug */ + int passes = 0; + + /* Add transforms to ensure the correct output format is produced then check + * that the required implementation support is there. Always expand; always + * need 8 bits minimum, no palette and expanded tRNS. + */ + png_set_expand(png_ptr); + + /* Now check the format to see if it was modified. */ + { + png_uint_32 base_format = png_image_format(png_ptr) & + ~PNG_FORMAT_FLAG_COLORMAP /* removed by png_set_expand */; + png_uint_32 change = format ^ base_format; + png_fixed_point output_gamma; + int mode; /* alpha mode */ + + /* Do this first so that we have a record if rgb to gray is happening. */ + if ((change & PNG_FORMAT_FLAG_COLOR) != 0) + { + /* gray<->color transformation required. */ + if ((format & PNG_FORMAT_FLAG_COLOR) != 0) + png_set_gray_to_rgb(png_ptr); + + else + { + /* libpng can't do both rgb to gray and + * background/pre-multiplication if there is also significant gamma + * correction, because both operations require linear colors and + * the code only supports one transform doing the gamma correction. + * Handle this by doing the pre-multiplication or background + * operation in this code, if necessary. + * + * TODO: fix this by rewriting pngrtran.c (!) + * + * For the moment (given that fixing this in pngrtran.c is an + * enormous change) 'do_local_background' is used to indicate that + * the problem exists. + */ + if ((base_format & PNG_FORMAT_FLAG_ALPHA) != 0) + do_local_background = 1/*maybe*/; + + png_set_rgb_to_gray_fixed(png_ptr, PNG_ERROR_ACTION_NONE, + PNG_RGB_TO_GRAY_DEFAULT, PNG_RGB_TO_GRAY_DEFAULT); + } + + change &= ~PNG_FORMAT_FLAG_COLOR; + } + + /* Set the gamma appropriately, linear for 16-bit input, sRGB otherwise. + */ + { + png_fixed_point input_gamma_default; + + if ((base_format & PNG_FORMAT_FLAG_LINEAR) != 0 && + (image->flags & PNG_IMAGE_FLAG_16BIT_sRGB) == 0) + input_gamma_default = PNG_GAMMA_LINEAR; + else + input_gamma_default = PNG_DEFAULT_sRGB; + + /* Call png_set_alpha_mode to set the default for the input gamma; the + * output gamma is set by a second call below. + */ + png_set_alpha_mode_fixed(png_ptr, PNG_ALPHA_PNG, input_gamma_default); + } + + if (linear != 0) + { + /* If there *is* an alpha channel in the input it must be multiplied + * out; use PNG_ALPHA_STANDARD, otherwise just use PNG_ALPHA_PNG. + */ + if ((base_format & PNG_FORMAT_FLAG_ALPHA) != 0) + mode = PNG_ALPHA_STANDARD; /* associated alpha */ + + else + mode = PNG_ALPHA_PNG; + + output_gamma = PNG_GAMMA_LINEAR; + } + + else + { + mode = PNG_ALPHA_PNG; + output_gamma = PNG_DEFAULT_sRGB; + } + + /* If 'do_local_background' is set check for the presence of gamma + * correction; this is part of the work-round for the libpng bug + * described above. + * + * TODO: fix libpng and remove this. + */ + if (do_local_background != 0) + { + png_fixed_point gtest; + + /* This is 'png_gamma_threshold' from pngrtran.c; the test used for + * gamma correction, the screen gamma hasn't been set on png_struct + * yet; it's set below. png_struct::gamma, however, is set to the + * final value. + */ + if (png_muldiv(>est, output_gamma, png_ptr->colorspace.gamma, + PNG_FP_1) != 0 && png_gamma_significant(gtest) == 0) + do_local_background = 0; + + else if (mode == PNG_ALPHA_STANDARD) + { + do_local_background = 2/*required*/; + mode = PNG_ALPHA_PNG; /* prevent libpng doing it */ + } + + /* else leave as 1 for the checks below */ + } + + /* If the bit-depth changes then handle that here. */ + if ((change & PNG_FORMAT_FLAG_LINEAR) != 0) + { + if (linear != 0 /*16-bit output*/) + png_set_expand_16(png_ptr); + + else /* 8-bit output */ + png_set_scale_16(png_ptr); + + change &= ~PNG_FORMAT_FLAG_LINEAR; + } + + /* Now the background/alpha channel changes. */ + if ((change & PNG_FORMAT_FLAG_ALPHA) != 0) + { + /* Removing an alpha channel requires composition for the 8-bit + * formats; for the 16-bit it is already done, above, by the + * pre-multiplication and the channel just needs to be stripped. + */ + if ((base_format & PNG_FORMAT_FLAG_ALPHA) != 0) + { + /* If RGB->gray is happening the alpha channel must be left and the + * operation completed locally. + * + * TODO: fix libpng and remove this. + */ + if (do_local_background != 0) + do_local_background = 2/*required*/; + + /* 16-bit output: just remove the channel */ + else if (linear != 0) /* compose on black (well, pre-multiply) */ + png_set_strip_alpha(png_ptr); + + /* 8-bit output: do an appropriate compose */ + else if (display->background != NULL) + { + png_color_16 c; + + c.index = 0; /*unused*/ + c.red = display->background->red; + c.green = display->background->green; + c.blue = display->background->blue; + c.gray = display->background->green; + + /* This is always an 8-bit sRGB value, using the 'green' channel + * for gray is much better than calculating the luminance here; + * we can get off-by-one errors in that calculation relative to + * the app expectations and that will show up in transparent + * pixels. + */ + png_set_background_fixed(png_ptr, &c, + PNG_BACKGROUND_GAMMA_SCREEN, 0/*need_expand*/, + 0/*gamma: not used*/); + } + + else /* compose on row: implemented below. */ + { + do_local_compose = 1; + /* This leaves the alpha channel in the output, so it has to be + * removed by the code below. Set the encoding to the 'OPTIMIZE' + * one so the code only has to hack on the pixels that require + * composition. + */ + mode = PNG_ALPHA_OPTIMIZED; + } + } + + else /* output needs an alpha channel */ + { + /* This is tricky because it happens before the swap operation has + * been accomplished; however, the swap does *not* swap the added + * alpha channel (weird API), so it must be added in the correct + * place. + */ + png_uint_32 filler; /* opaque filler */ + int where; + + if (linear != 0) + filler = 65535; + + else + filler = 255; + +# ifdef PNG_FORMAT_AFIRST_SUPPORTED + if ((format & PNG_FORMAT_FLAG_AFIRST) != 0) + { + where = PNG_FILLER_BEFORE; + change &= ~PNG_FORMAT_FLAG_AFIRST; + } + + else +# endif + where = PNG_FILLER_AFTER; + + png_set_add_alpha(png_ptr, filler, where); + } + + /* This stops the (irrelevant) call to swap_alpha below. */ + change &= ~PNG_FORMAT_FLAG_ALPHA; + } + + /* Now set the alpha mode correctly; this is always done, even if there is + * no alpha channel in either the input or the output because it correctly + * sets the output gamma. + */ + png_set_alpha_mode_fixed(png_ptr, mode, output_gamma); + +# ifdef PNG_FORMAT_BGR_SUPPORTED + if ((change & PNG_FORMAT_FLAG_BGR) != 0) + { + /* Check only the output format; PNG is never BGR; don't do this if + * the output is gray, but fix up the 'format' value in that case. + */ + if ((format & PNG_FORMAT_FLAG_COLOR) != 0) + png_set_bgr(png_ptr); + + else + format &= ~PNG_FORMAT_FLAG_BGR; + + change &= ~PNG_FORMAT_FLAG_BGR; + } +# endif + +# ifdef PNG_FORMAT_AFIRST_SUPPORTED + if ((change & PNG_FORMAT_FLAG_AFIRST) != 0) + { + /* Only relevant if there is an alpha channel - it's particularly + * important to handle this correctly because do_local_compose may + * be set above and then libpng will keep the alpha channel for this + * code to remove. + */ + if ((format & PNG_FORMAT_FLAG_ALPHA) != 0) + { + /* Disable this if doing a local background, + * TODO: remove this when local background is no longer required. + */ + if (do_local_background != 2) + png_set_swap_alpha(png_ptr); + } + + else + format &= ~PNG_FORMAT_FLAG_AFIRST; + + change &= ~PNG_FORMAT_FLAG_AFIRST; + } +# endif + + /* If the *output* is 16-bit then we need to check for a byte-swap on this + * architecture. + */ + if (linear != 0) + { + PNG_CONST png_uint_16 le = 0x0001; + + if ((*(png_const_bytep) & le) != 0) + png_set_swap(png_ptr); + } + + /* If change is not now 0 some transformation is missing - error out. */ + if (change != 0) + png_error(png_ptr, "png_read_image: unsupported transformation"); + } + + PNG_SKIP_CHUNKS(png_ptr); + + /* Update the 'info' structure and make sure the result is as required; first + * make sure to turn on the interlace handling if it will be required + * (because it can't be turned on *after* the call to png_read_update_info!) + * + * TODO: remove the do_local_background fixup below. + */ + if (do_local_compose == 0 && do_local_background != 2) + passes = png_set_interlace_handling(png_ptr); + + png_read_update_info(png_ptr, info_ptr); + + { + png_uint_32 info_format = 0; + + if ((info_ptr->color_type & PNG_COLOR_MASK_COLOR) != 0) + info_format |= PNG_FORMAT_FLAG_COLOR; + + if ((info_ptr->color_type & PNG_COLOR_MASK_ALPHA) != 0) + { + /* do_local_compose removes this channel below. */ + if (do_local_compose == 0) + { + /* do_local_background does the same if required. */ + if (do_local_background != 2 || + (format & PNG_FORMAT_FLAG_ALPHA) != 0) + info_format |= PNG_FORMAT_FLAG_ALPHA; + } + } + + else if (do_local_compose != 0) /* internal error */ + png_error(png_ptr, "png_image_read: alpha channel lost"); + + if (info_ptr->bit_depth == 16) + info_format |= PNG_FORMAT_FLAG_LINEAR; + +# ifdef PNG_FORMAT_BGR_SUPPORTED + if ((png_ptr->transformations & PNG_BGR) != 0) + info_format |= PNG_FORMAT_FLAG_BGR; +# endif + +# ifdef PNG_FORMAT_AFIRST_SUPPORTED + if (do_local_background == 2) + { + if ((format & PNG_FORMAT_FLAG_AFIRST) != 0) + info_format |= PNG_FORMAT_FLAG_AFIRST; + } + + if ((png_ptr->transformations & PNG_SWAP_ALPHA) != 0 || + ((png_ptr->transformations & PNG_ADD_ALPHA) != 0 && + (png_ptr->flags & PNG_FLAG_FILLER_AFTER) == 0)) + { + if (do_local_background == 2) + png_error(png_ptr, "unexpected alpha swap transformation"); + + info_format |= PNG_FORMAT_FLAG_AFIRST; + } +# endif + + /* This is actually an internal error. */ + if (info_format != format) + png_error(png_ptr, "png_read_image: invalid transformations"); + } + + /* Now read the rows. If do_local_compose is set then it is necessary to use + * a local row buffer. The output will be GA, RGBA or BGRA and must be + * converted to G, RGB or BGR as appropriate. The 'local_row' member of the + * display acts as a flag. + */ + { + png_voidp first_row = display->buffer; + ptrdiff_t row_bytes = display->row_stride; + + if (linear != 0) + row_bytes *= 2; + + /* The following expression is designed to work correctly whether it gives + * a signed or an unsigned result. + */ + if (row_bytes < 0) + { + char *ptr = png_voidcast(char*, first_row); + ptr += (image->height-1) * (-row_bytes); + first_row = png_voidcast(png_voidp, ptr); + } + + display->first_row = first_row; + display->row_bytes = row_bytes; + } + + if (do_local_compose != 0) + { + int result; + png_voidp row = png_malloc(png_ptr, png_get_rowbytes(png_ptr, info_ptr)); + + display->local_row = row; + result = png_safe_execute(image, png_image_read_composite, display); + display->local_row = NULL; + png_free(png_ptr, row); + + return result; + } + + else if (do_local_background == 2) + { + int result; + png_voidp row = png_malloc(png_ptr, png_get_rowbytes(png_ptr, info_ptr)); + + display->local_row = row; + result = png_safe_execute(image, png_image_read_background, display); + display->local_row = NULL; + png_free(png_ptr, row); + + return result; + } + + else + { + png_alloc_size_t row_bytes = display->row_bytes; + + while (--passes >= 0) + { + png_uint_32 y = image->height; + png_bytep row = png_voidcast(png_bytep, display->first_row); + + while (y-- > 0) + { + png_read_row(png_ptr, row, NULL); + row += row_bytes; + } + } + + return 1; + } +} + +int PNGAPI +png_image_finish_read(png_imagep image, png_const_colorp background, + void *buffer, png_int_32 row_stride, void *colormap) +{ + if (image != NULL && image->version == PNG_IMAGE_VERSION) + { + png_uint_32 check; + + if (row_stride == 0) + row_stride = PNG_IMAGE_ROW_STRIDE(*image); + + if (row_stride < 0) + check = -row_stride; + + else + check = row_stride; + + if (image->opaque != NULL && buffer != NULL && + check >= PNG_IMAGE_ROW_STRIDE(*image)) + { + if ((image->format & PNG_FORMAT_FLAG_COLORMAP) == 0 || + (image->colormap_entries > 0 && colormap != NULL)) + { + int result; + png_image_read_control display; + + memset(&display, 0, (sizeof display)); + display.image = image; + display.buffer = buffer; + display.row_stride = row_stride; + display.colormap = colormap; + display.background = background; + display.local_row = NULL; + + /* Choose the correct 'end' routine; for the color-map case all the + * setup has already been done. + */ + if ((image->format & PNG_FORMAT_FLAG_COLORMAP) != 0) + result = + png_safe_execute(image, png_image_read_colormap, &display) && + png_safe_execute(image, png_image_read_colormapped, &display); + + else + result = + png_safe_execute(image, png_image_read_direct, &display); + + png_image_free(image); + return result; + } + + else + return png_image_error(image, + "png_image_finish_read[color-map]: no color-map"); + } + + else + return png_image_error(image, + "png_image_finish_read: invalid argument"); + } + + else if (image != NULL) + return png_image_error(image, + "png_image_finish_read: damaged PNG_IMAGE_VERSION"); + + return 0; +} + +#endif /* SIMPLIFIED_READ */ +#endif /* READ */ diff --git a/app/src/main/jni/libpng/pngrio.c b/app/src/main/jni/libpng/pngrio.c new file mode 100644 index 000000000..1e98395fd --- /dev/null +++ b/app/src/main/jni/libpng/pngrio.c @@ -0,0 +1,120 @@ + +/* pngrio.c - functions for data input + * + * Last changed in libpng 1.6.15 [November 20, 2014] + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + * + * This file provides a location for all input. Users who need + * special handling are expected to write a function that has the same + * arguments as this and performs a similar function, but that possibly + * has a different input method. Note that you shouldn't change this + * function, but rather write a replacement function and then make + * libpng use it at run time with png_set_read_fn(...). + */ + +#include "pngpriv.h" + +#ifdef PNG_READ_SUPPORTED + +/* Read the data from whatever input you are using. The default routine + * reads from a file pointer. Note that this routine sometimes gets called + * with very small lengths, so you should implement some kind of simple + * buffering if you are using unbuffered reads. This should never be asked + * to read more then 64K on a 16 bit machine. + */ +void /* PRIVATE */ +png_read_data(png_structrp png_ptr, png_bytep data, png_size_t length) +{ + png_debug1(4, "reading %d bytes", (int)length); + + if (png_ptr->read_data_fn != NULL) + (*(png_ptr->read_data_fn))(png_ptr, data, length); + + else + png_error(png_ptr, "Call to NULL read function"); +} + +#ifdef PNG_STDIO_SUPPORTED +/* This is the function that does the actual reading of data. If you are + * not reading from a standard C stream, you should create a replacement + * read_data function and use it at run time with png_set_read_fn(), rather + * than changing the library. + */ +void PNGCBAPI +png_default_read_data(png_structp png_ptr, png_bytep data, png_size_t length) +{ + png_size_t check; + + if (png_ptr == NULL) + return; + + /* fread() returns 0 on error, so it is OK to store this in a png_size_t + * instead of an int, which is what fread() actually returns. + */ + check = fread(data, 1, length, png_voidcast(png_FILE_p, png_ptr->io_ptr)); + + if (check != length) + png_error(png_ptr, "Read Error"); +} +#endif + +/* This function allows the application to supply a new input function + * for libpng if standard C streams aren't being used. + * + * This function takes as its arguments: + * + * png_ptr - pointer to a png input data structure + * + * io_ptr - pointer to user supplied structure containing info about + * the input functions. May be NULL. + * + * read_data_fn - pointer to a new input function that takes as its + * arguments a pointer to a png_struct, a pointer to + * a location where input data can be stored, and a 32-bit + * unsigned int that is the number of bytes to be read. + * To exit and output any fatal error messages the new write + * function should call png_error(png_ptr, "Error msg"). + * May be NULL, in which case libpng's default function will + * be used. + */ +void PNGAPI +png_set_read_fn(png_structrp png_ptr, png_voidp io_ptr, + png_rw_ptr read_data_fn) +{ + if (png_ptr == NULL) + return; + + png_ptr->io_ptr = io_ptr; + +#ifdef PNG_STDIO_SUPPORTED + if (read_data_fn != NULL) + png_ptr->read_data_fn = read_data_fn; + + else + png_ptr->read_data_fn = png_default_read_data; +#else + png_ptr->read_data_fn = read_data_fn; +#endif + +#ifdef PNG_WRITE_SUPPORTED + /* It is an error to write to a read device */ + if (png_ptr->write_data_fn != NULL) + { + png_ptr->write_data_fn = NULL; + png_warning(png_ptr, + "Can't set both read_data_fn and write_data_fn in the" + " same structure"); + } +#endif + +#ifdef PNG_WRITE_FLUSH_SUPPORTED + png_ptr->output_flush_fn = NULL; +#endif +} +#endif /* READ */ diff --git a/app/src/main/jni/libpng/pngrtran.c b/app/src/main/jni/libpng/pngrtran.c new file mode 100644 index 000000000..e6fe9fdbf --- /dev/null +++ b/app/src/main/jni/libpng/pngrtran.c @@ -0,0 +1,4994 @@ + +/* pngrtran.c - transforms the data in a row for PNG readers + * + * Last changed in libpng 1.6.15 [November 20, 2014] + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + * + * This file contains functions optionally called by an application + * in order to tell libpng how to handle data when reading a PNG. + * Transformations that are used in both reading and writing are + * in pngtrans.c. + */ + +#include "pngpriv.h" + +#ifdef PNG_READ_SUPPORTED + +/* Set the action on getting a CRC error for an ancillary or critical chunk. */ +void PNGAPI +png_set_crc_action(png_structrp png_ptr, int crit_action, int ancil_action) +{ + png_debug(1, "in png_set_crc_action"); + + if (png_ptr == NULL) + return; + + /* Tell libpng how we react to CRC errors in critical chunks */ + switch (crit_action) + { + case PNG_CRC_NO_CHANGE: /* Leave setting as is */ + break; + + case PNG_CRC_WARN_USE: /* Warn/use data */ + png_ptr->flags &= ~PNG_FLAG_CRC_CRITICAL_MASK; + png_ptr->flags |= PNG_FLAG_CRC_CRITICAL_USE; + break; + + case PNG_CRC_QUIET_USE: /* Quiet/use data */ + png_ptr->flags &= ~PNG_FLAG_CRC_CRITICAL_MASK; + png_ptr->flags |= PNG_FLAG_CRC_CRITICAL_USE | + PNG_FLAG_CRC_CRITICAL_IGNORE; + break; + + case PNG_CRC_WARN_DISCARD: /* Not a valid action for critical data */ + png_warning(png_ptr, + "Can't discard critical data on CRC error"); + case PNG_CRC_ERROR_QUIT: /* Error/quit */ + + case PNG_CRC_DEFAULT: + default: + png_ptr->flags &= ~PNG_FLAG_CRC_CRITICAL_MASK; + break; + } + + /* Tell libpng how we react to CRC errors in ancillary chunks */ + switch (ancil_action) + { + case PNG_CRC_NO_CHANGE: /* Leave setting as is */ + break; + + case PNG_CRC_WARN_USE: /* Warn/use data */ + png_ptr->flags &= ~PNG_FLAG_CRC_ANCILLARY_MASK; + png_ptr->flags |= PNG_FLAG_CRC_ANCILLARY_USE; + break; + + case PNG_CRC_QUIET_USE: /* Quiet/use data */ + png_ptr->flags &= ~PNG_FLAG_CRC_ANCILLARY_MASK; + png_ptr->flags |= PNG_FLAG_CRC_ANCILLARY_USE | + PNG_FLAG_CRC_ANCILLARY_NOWARN; + break; + + case PNG_CRC_ERROR_QUIT: /* Error/quit */ + png_ptr->flags &= ~PNG_FLAG_CRC_ANCILLARY_MASK; + png_ptr->flags |= PNG_FLAG_CRC_ANCILLARY_NOWARN; + break; + + case PNG_CRC_WARN_DISCARD: /* Warn/discard data */ + + case PNG_CRC_DEFAULT: + default: + png_ptr->flags &= ~PNG_FLAG_CRC_ANCILLARY_MASK; + break; + } +} + +#ifdef PNG_READ_TRANSFORMS_SUPPORTED +/* Is it OK to set a transformation now? Only if png_start_read_image or + * png_read_update_info have not been called. It is not necessary for the IHDR + * to have been read in all cases; the need_IHDR parameter allows for this + * check too. + */ +static int +png_rtran_ok(png_structrp png_ptr, int need_IHDR) +{ + if (png_ptr != NULL) + { + if ((png_ptr->flags & PNG_FLAG_ROW_INIT) != 0) + png_app_error(png_ptr, + "invalid after png_start_read_image or png_read_update_info"); + + else if (need_IHDR && (png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_app_error(png_ptr, "invalid before the PNG header has been read"); + + else + { + /* Turn on failure to initialize correctly for all transforms. */ + png_ptr->flags |= PNG_FLAG_DETECT_UNINITIALIZED; + + return 1; /* Ok */ + } + } + + return 0; /* no png_error possible! */ +} +#endif + +#ifdef PNG_READ_BACKGROUND_SUPPORTED +/* Handle alpha and tRNS via a background color */ +void PNGFAPI +png_set_background_fixed(png_structrp png_ptr, + png_const_color_16p background_color, int background_gamma_code, + int need_expand, png_fixed_point background_gamma) +{ + png_debug(1, "in png_set_background_fixed"); + + if (png_rtran_ok(png_ptr, 0) == 0 || background_color == NULL) + return; + + if (background_gamma_code == PNG_BACKGROUND_GAMMA_UNKNOWN) + { + png_warning(png_ptr, "Application must supply a known background gamma"); + return; + } + + png_ptr->transformations |= PNG_COMPOSE | PNG_STRIP_ALPHA; + png_ptr->transformations &= ~PNG_ENCODE_ALPHA; + png_ptr->flags &= ~PNG_FLAG_OPTIMIZE_ALPHA; + + png_ptr->background = *background_color; + png_ptr->background_gamma = background_gamma; + png_ptr->background_gamma_type = (png_byte)(background_gamma_code); + if (need_expand != 0) + png_ptr->transformations |= PNG_BACKGROUND_EXPAND; + else + png_ptr->transformations &= ~PNG_BACKGROUND_EXPAND; +} + +# ifdef PNG_FLOATING_POINT_SUPPORTED +void PNGAPI +png_set_background(png_structrp png_ptr, + png_const_color_16p background_color, int background_gamma_code, + int need_expand, double background_gamma) +{ + png_set_background_fixed(png_ptr, background_color, background_gamma_code, + need_expand, png_fixed(png_ptr, background_gamma, "png_set_background")); +} +# endif /* FLOATING_POINT */ +#endif /* READ_BACKGROUND */ + +/* Scale 16-bit depth files to 8-bit depth. If both of these are set then the + * one that pngrtran does first (scale) happens. This is necessary to allow the + * TRANSFORM and API behavior to be somewhat consistent, and it's simpler. + */ +#ifdef PNG_READ_SCALE_16_TO_8_SUPPORTED +void PNGAPI +png_set_scale_16(png_structrp png_ptr) +{ + png_debug(1, "in png_set_scale_16"); + + if (png_rtran_ok(png_ptr, 0) == 0) + return; + + png_ptr->transformations |= PNG_SCALE_16_TO_8; +} +#endif + +#ifdef PNG_READ_STRIP_16_TO_8_SUPPORTED +/* Chop 16-bit depth files to 8-bit depth */ +void PNGAPI +png_set_strip_16(png_structrp png_ptr) +{ + png_debug(1, "in png_set_strip_16"); + + if (png_rtran_ok(png_ptr, 0) == 0) + return; + + png_ptr->transformations |= PNG_16_TO_8; +} +#endif + +#ifdef PNG_READ_STRIP_ALPHA_SUPPORTED +void PNGAPI +png_set_strip_alpha(png_structrp png_ptr) +{ + png_debug(1, "in png_set_strip_alpha"); + + if (png_rtran_ok(png_ptr, 0) == 0) + return; + + png_ptr->transformations |= PNG_STRIP_ALPHA; +} +#endif + +#if defined(PNG_READ_ALPHA_MODE_SUPPORTED) || defined(PNG_READ_GAMMA_SUPPORTED) +static png_fixed_point +translate_gamma_flags(png_structrp png_ptr, png_fixed_point output_gamma, + int is_screen) +{ + /* Check for flag values. The main reason for having the old Mac value as a + * flag is that it is pretty near impossible to work out what the correct + * value is from Apple documentation - a working Mac system is needed to + * discover the value! + */ + if (output_gamma == PNG_DEFAULT_sRGB || + output_gamma == PNG_FP_1 / PNG_DEFAULT_sRGB) + { + /* If there is no sRGB support this just sets the gamma to the standard + * sRGB value. (This is a side effect of using this function!) + */ +# ifdef PNG_READ_sRGB_SUPPORTED + png_ptr->flags |= PNG_FLAG_ASSUME_sRGB; +# else + PNG_UNUSED(png_ptr) +# endif + if (is_screen != 0) + output_gamma = PNG_GAMMA_sRGB; + else + output_gamma = PNG_GAMMA_sRGB_INVERSE; + } + + else if (output_gamma == PNG_GAMMA_MAC_18 || + output_gamma == PNG_FP_1 / PNG_GAMMA_MAC_18) + { + if (is_screen != 0) + output_gamma = PNG_GAMMA_MAC_OLD; + else + output_gamma = PNG_GAMMA_MAC_INVERSE; + } + + return output_gamma; +} + +# ifdef PNG_FLOATING_POINT_SUPPORTED +static png_fixed_point +convert_gamma_value(png_structrp png_ptr, double output_gamma) +{ + /* The following silently ignores cases where fixed point (times 100,000) + * gamma values are passed to the floating point API. This is safe and it + * means the fixed point constants work just fine with the floating point + * API. The alternative would just lead to undetected errors and spurious + * bug reports. Negative values fail inside the _fixed API unless they + * correspond to the flag values. + */ + if (output_gamma > 0 && output_gamma < 128) + output_gamma *= PNG_FP_1; + + /* This preserves -1 and -2 exactly: */ + output_gamma = floor(output_gamma + .5); + + if (output_gamma > PNG_FP_MAX || output_gamma < PNG_FP_MIN) + png_fixed_error(png_ptr, "gamma value"); + + return (png_fixed_point)output_gamma; +} +# endif +#endif /* READ_ALPHA_MODE || READ_GAMMA */ + +#ifdef PNG_READ_ALPHA_MODE_SUPPORTED +void PNGFAPI +png_set_alpha_mode_fixed(png_structrp png_ptr, int mode, + png_fixed_point output_gamma) +{ + int compose = 0; + png_fixed_point file_gamma; + + png_debug(1, "in png_set_alpha_mode"); + + if (png_rtran_ok(png_ptr, 0) == 0) + return; + + output_gamma = translate_gamma_flags(png_ptr, output_gamma, 1/*screen*/); + + /* Validate the value to ensure it is in a reasonable range. The value + * is expected to be 1 or greater, but this range test allows for some + * viewing correction values. The intent is to weed out users of this API + * who use the inverse of the gamma value accidentally! Since some of these + * values are reasonable this may have to be changed. + */ + if (output_gamma < 70000 || output_gamma > 300000) + png_error(png_ptr, "output gamma out of expected range"); + + /* The default file gamma is the inverse of the output gamma; the output + * gamma may be changed below so get the file value first: + */ + file_gamma = png_reciprocal(output_gamma); + + /* There are really 8 possibilities here, composed of any combination + * of: + * + * premultiply the color channels + * do not encode non-opaque pixels + * encode the alpha as well as the color channels + * + * The differences disappear if the input/output ('screen') gamma is 1.0, + * because then the encoding is a no-op and there is only the choice of + * premultiplying the color channels or not. + * + * png_set_alpha_mode and png_set_background interact because both use + * png_compose to do the work. Calling both is only useful when + * png_set_alpha_mode is used to set the default mode - PNG_ALPHA_PNG - along + * with a default gamma value. Otherwise PNG_COMPOSE must not be set. + */ + switch (mode) + { + case PNG_ALPHA_PNG: /* default: png standard */ + /* No compose, but it may be set by png_set_background! */ + png_ptr->transformations &= ~PNG_ENCODE_ALPHA; + png_ptr->flags &= ~PNG_FLAG_OPTIMIZE_ALPHA; + break; + + case PNG_ALPHA_ASSOCIATED: /* color channels premultiplied */ + compose = 1; + png_ptr->transformations &= ~PNG_ENCODE_ALPHA; + png_ptr->flags &= ~PNG_FLAG_OPTIMIZE_ALPHA; + /* The output is linear: */ + output_gamma = PNG_FP_1; + break; + + case PNG_ALPHA_OPTIMIZED: /* associated, non-opaque pixels linear */ + compose = 1; + png_ptr->transformations &= ~PNG_ENCODE_ALPHA; + png_ptr->flags |= PNG_FLAG_OPTIMIZE_ALPHA; + /* output_gamma records the encoding of opaque pixels! */ + break; + + case PNG_ALPHA_BROKEN: /* associated, non-linear, alpha encoded */ + compose = 1; + png_ptr->transformations |= PNG_ENCODE_ALPHA; + png_ptr->flags &= ~PNG_FLAG_OPTIMIZE_ALPHA; + break; + + default: + png_error(png_ptr, "invalid alpha mode"); + } + + /* Only set the default gamma if the file gamma has not been set (this has + * the side effect that the gamma in a second call to png_set_alpha_mode will + * be ignored.) + */ + if (png_ptr->colorspace.gamma == 0) + { + png_ptr->colorspace.gamma = file_gamma; + png_ptr->colorspace.flags |= PNG_COLORSPACE_HAVE_GAMMA; + } + + /* But always set the output gamma: */ + png_ptr->screen_gamma = output_gamma; + + /* Finally, if pre-multiplying, set the background fields to achieve the + * desired result. + */ + if (compose != 0) + { + /* And obtain alpha pre-multiplication by composing on black: */ + memset(&png_ptr->background, 0, (sizeof png_ptr->background)); + png_ptr->background_gamma = png_ptr->colorspace.gamma; /* just in case */ + png_ptr->background_gamma_type = PNG_BACKGROUND_GAMMA_FILE; + png_ptr->transformations &= ~PNG_BACKGROUND_EXPAND; + + if ((png_ptr->transformations & PNG_COMPOSE) != 0) + png_error(png_ptr, + "conflicting calls to set alpha mode and background"); + + png_ptr->transformations |= PNG_COMPOSE; + } +} + +# ifdef PNG_FLOATING_POINT_SUPPORTED +void PNGAPI +png_set_alpha_mode(png_structrp png_ptr, int mode, double output_gamma) +{ + png_set_alpha_mode_fixed(png_ptr, mode, convert_gamma_value(png_ptr, + output_gamma)); +} +# endif +#endif + +#ifdef PNG_READ_QUANTIZE_SUPPORTED +/* Dither file to 8-bit. Supply a palette, the current number + * of elements in the palette, the maximum number of elements + * allowed, and a histogram if possible. If the current number + * of colors is greater then the maximum number, the palette will be + * modified to fit in the maximum number. "full_quantize" indicates + * whether we need a quantizing cube set up for RGB images, or if we + * simply are reducing the number of colors in a paletted image. + */ + +typedef struct png_dsort_struct +{ + struct png_dsort_struct * next; + png_byte left; + png_byte right; +} png_dsort; +typedef png_dsort * png_dsortp; +typedef png_dsort * * png_dsortpp; + +void PNGAPI +png_set_quantize(png_structrp png_ptr, png_colorp palette, + int num_palette, int maximum_colors, png_const_uint_16p histogram, + int full_quantize) +{ + png_debug(1, "in png_set_quantize"); + + if (png_rtran_ok(png_ptr, 0) == 0) + return; + + png_ptr->transformations |= PNG_QUANTIZE; + + if (full_quantize == 0) + { + int i; + + png_ptr->quantize_index = (png_bytep)png_malloc(png_ptr, + (png_uint_32)(num_palette * (sizeof (png_byte)))); + for (i = 0; i < num_palette; i++) + png_ptr->quantize_index[i] = (png_byte)i; + } + + if (num_palette > maximum_colors) + { + if (histogram != NULL) + { + /* This is easy enough, just throw out the least used colors. + * Perhaps not the best solution, but good enough. + */ + + int i; + + /* Initialize an array to sort colors */ + png_ptr->quantize_sort = (png_bytep)png_malloc(png_ptr, + (png_uint_32)(num_palette * (sizeof (png_byte)))); + + /* Initialize the quantize_sort array */ + for (i = 0; i < num_palette; i++) + png_ptr->quantize_sort[i] = (png_byte)i; + + /* Find the least used palette entries by starting a + * bubble sort, and running it until we have sorted + * out enough colors. Note that we don't care about + * sorting all the colors, just finding which are + * least used. + */ + + for (i = num_palette - 1; i >= maximum_colors; i--) + { + int done; /* To stop early if the list is pre-sorted */ + int j; + + done = 1; + for (j = 0; j < i; j++) + { + if (histogram[png_ptr->quantize_sort[j]] + < histogram[png_ptr->quantize_sort[j + 1]]) + { + png_byte t; + + t = png_ptr->quantize_sort[j]; + png_ptr->quantize_sort[j] = png_ptr->quantize_sort[j + 1]; + png_ptr->quantize_sort[j + 1] = t; + done = 0; + } + } + + if (done != 0) + break; + } + + /* Swap the palette around, and set up a table, if necessary */ + if (full_quantize != 0) + { + int j = num_palette; + + /* Put all the useful colors within the max, but don't + * move the others. + */ + for (i = 0; i < maximum_colors; i++) + { + if ((int)png_ptr->quantize_sort[i] >= maximum_colors) + { + do + j--; + while ((int)png_ptr->quantize_sort[j] >= maximum_colors); + + palette[i] = palette[j]; + } + } + } + else + { + int j = num_palette; + + /* Move all the used colors inside the max limit, and + * develop a translation table. + */ + for (i = 0; i < maximum_colors; i++) + { + /* Only move the colors we need to */ + if ((int)png_ptr->quantize_sort[i] >= maximum_colors) + { + png_color tmp_color; + + do + j--; + while ((int)png_ptr->quantize_sort[j] >= maximum_colors); + + tmp_color = palette[j]; + palette[j] = palette[i]; + palette[i] = tmp_color; + /* Indicate where the color went */ + png_ptr->quantize_index[j] = (png_byte)i; + png_ptr->quantize_index[i] = (png_byte)j; + } + } + + /* Find closest color for those colors we are not using */ + for (i = 0; i < num_palette; i++) + { + if ((int)png_ptr->quantize_index[i] >= maximum_colors) + { + int min_d, k, min_k, d_index; + + /* Find the closest color to one we threw out */ + d_index = png_ptr->quantize_index[i]; + min_d = PNG_COLOR_DIST(palette[d_index], palette[0]); + for (k = 1, min_k = 0; k < maximum_colors; k++) + { + int d; + + d = PNG_COLOR_DIST(palette[d_index], palette[k]); + + if (d < min_d) + { + min_d = d; + min_k = k; + } + } + /* Point to closest color */ + png_ptr->quantize_index[i] = (png_byte)min_k; + } + } + } + png_free(png_ptr, png_ptr->quantize_sort); + png_ptr->quantize_sort = NULL; + } + else + { + /* This is much harder to do simply (and quickly). Perhaps + * we need to go through a median cut routine, but those + * don't always behave themselves with only a few colors + * as input. So we will just find the closest two colors, + * and throw out one of them (chosen somewhat randomly). + * [We don't understand this at all, so if someone wants to + * work on improving it, be our guest - AED, GRP] + */ + int i; + int max_d; + int num_new_palette; + png_dsortp t; + png_dsortpp hash; + + t = NULL; + + /* Initialize palette index arrays */ + png_ptr->index_to_palette = (png_bytep)png_malloc(png_ptr, + (png_uint_32)(num_palette * (sizeof (png_byte)))); + png_ptr->palette_to_index = (png_bytep)png_malloc(png_ptr, + (png_uint_32)(num_palette * (sizeof (png_byte)))); + + /* Initialize the sort array */ + for (i = 0; i < num_palette; i++) + { + png_ptr->index_to_palette[i] = (png_byte)i; + png_ptr->palette_to_index[i] = (png_byte)i; + } + + hash = (png_dsortpp)png_calloc(png_ptr, (png_uint_32)(769 * + (sizeof (png_dsortp)))); + + num_new_palette = num_palette; + + /* Initial wild guess at how far apart the farthest pixel + * pair we will be eliminating will be. Larger + * numbers mean more areas will be allocated, Smaller + * numbers run the risk of not saving enough data, and + * having to do this all over again. + * + * I have not done extensive checking on this number. + */ + max_d = 96; + + while (num_new_palette > maximum_colors) + { + for (i = 0; i < num_new_palette - 1; i++) + { + int j; + + for (j = i + 1; j < num_new_palette; j++) + { + int d; + + d = PNG_COLOR_DIST(palette[i], palette[j]); + + if (d <= max_d) + { + + t = (png_dsortp)png_malloc_warn(png_ptr, + (png_uint_32)(sizeof (png_dsort))); + + if (t == NULL) + break; + + t->next = hash[d]; + t->left = (png_byte)i; + t->right = (png_byte)j; + hash[d] = t; + } + } + if (t == NULL) + break; + } + + if (t != NULL) + for (i = 0; i <= max_d; i++) + { + if (hash[i] != NULL) + { + png_dsortp p; + + for (p = hash[i]; p; p = p->next) + { + if ((int)png_ptr->index_to_palette[p->left] + < num_new_palette && + (int)png_ptr->index_to_palette[p->right] + < num_new_palette) + { + int j, next_j; + + if (num_new_palette & 0x01) + { + j = p->left; + next_j = p->right; + } + else + { + j = p->right; + next_j = p->left; + } + + num_new_palette--; + palette[png_ptr->index_to_palette[j]] + = palette[num_new_palette]; + if (full_quantize == 0) + { + int k; + + for (k = 0; k < num_palette; k++) + { + if (png_ptr->quantize_index[k] == + png_ptr->index_to_palette[j]) + png_ptr->quantize_index[k] = + png_ptr->index_to_palette[next_j]; + + if ((int)png_ptr->quantize_index[k] == + num_new_palette) + png_ptr->quantize_index[k] = + png_ptr->index_to_palette[j]; + } + } + + png_ptr->index_to_palette[png_ptr->palette_to_index + [num_new_palette]] = png_ptr->index_to_palette[j]; + + png_ptr->palette_to_index[png_ptr->index_to_palette[j]] + = png_ptr->palette_to_index[num_new_palette]; + + png_ptr->index_to_palette[j] = + (png_byte)num_new_palette; + + png_ptr->palette_to_index[num_new_palette] = + (png_byte)j; + } + if (num_new_palette <= maximum_colors) + break; + } + if (num_new_palette <= maximum_colors) + break; + } + } + + for (i = 0; i < 769; i++) + { + if (hash[i] != NULL) + { + png_dsortp p = hash[i]; + while (p) + { + t = p->next; + png_free(png_ptr, p); + p = t; + } + } + hash[i] = 0; + } + max_d += 96; + } + png_free(png_ptr, hash); + png_free(png_ptr, png_ptr->palette_to_index); + png_free(png_ptr, png_ptr->index_to_palette); + png_ptr->palette_to_index = NULL; + png_ptr->index_to_palette = NULL; + } + num_palette = maximum_colors; + } + if (png_ptr->palette == NULL) + { + png_ptr->palette = palette; + } + png_ptr->num_palette = (png_uint_16)num_palette; + + if (full_quantize != 0) + { + int i; + png_bytep distance; + int total_bits = PNG_QUANTIZE_RED_BITS + PNG_QUANTIZE_GREEN_BITS + + PNG_QUANTIZE_BLUE_BITS; + int num_red = (1 << PNG_QUANTIZE_RED_BITS); + int num_green = (1 << PNG_QUANTIZE_GREEN_BITS); + int num_blue = (1 << PNG_QUANTIZE_BLUE_BITS); + png_size_t num_entries = ((png_size_t)1 << total_bits); + + png_ptr->palette_lookup = (png_bytep)png_calloc(png_ptr, + (png_uint_32)(num_entries * (sizeof (png_byte)))); + + distance = (png_bytep)png_malloc(png_ptr, (png_uint_32)(num_entries * + (sizeof (png_byte)))); + + memset(distance, 0xff, num_entries * (sizeof (png_byte))); + + for (i = 0; i < num_palette; i++) + { + int ir, ig, ib; + int r = (palette[i].red >> (8 - PNG_QUANTIZE_RED_BITS)); + int g = (palette[i].green >> (8 - PNG_QUANTIZE_GREEN_BITS)); + int b = (palette[i].blue >> (8 - PNG_QUANTIZE_BLUE_BITS)); + + for (ir = 0; ir < num_red; ir++) + { + /* int dr = abs(ir - r); */ + int dr = ((ir > r) ? ir - r : r - ir); + int index_r = (ir << (PNG_QUANTIZE_BLUE_BITS + + PNG_QUANTIZE_GREEN_BITS)); + + for (ig = 0; ig < num_green; ig++) + { + /* int dg = abs(ig - g); */ + int dg = ((ig > g) ? ig - g : g - ig); + int dt = dr + dg; + int dm = ((dr > dg) ? dr : dg); + int index_g = index_r | (ig << PNG_QUANTIZE_BLUE_BITS); + + for (ib = 0; ib < num_blue; ib++) + { + int d_index = index_g | ib; + /* int db = abs(ib - b); */ + int db = ((ib > b) ? ib - b : b - ib); + int dmax = ((dm > db) ? dm : db); + int d = dmax + dt + db; + + if (d < (int)distance[d_index]) + { + distance[d_index] = (png_byte)d; + png_ptr->palette_lookup[d_index] = (png_byte)i; + } + } + } + } + } + + png_free(png_ptr, distance); + } +} +#endif /* READ_QUANTIZE */ + +#ifdef PNG_READ_GAMMA_SUPPORTED +void PNGFAPI +png_set_gamma_fixed(png_structrp png_ptr, png_fixed_point scrn_gamma, + png_fixed_point file_gamma) +{ + png_debug(1, "in png_set_gamma_fixed"); + + if (png_rtran_ok(png_ptr, 0) == 0) + return; + + /* New in libpng-1.5.4 - reserve particular negative values as flags. */ + scrn_gamma = translate_gamma_flags(png_ptr, scrn_gamma, 1/*screen*/); + file_gamma = translate_gamma_flags(png_ptr, file_gamma, 0/*file*/); + + /* Checking the gamma values for being >0 was added in 1.5.4 along with the + * premultiplied alpha support; this actually hides an undocumented feature + * of the previous implementation which allowed gamma processing to be + * disabled in background handling. There is no evidence (so far) that this + * was being used; however, png_set_background itself accepted and must still + * accept '0' for the gamma value it takes, because it isn't always used. + * + * Since this is an API change (albeit a very minor one that removes an + * undocumented API feature) the following checks were only enabled in + * libpng-1.6.0. + */ + if (file_gamma <= 0) + png_error(png_ptr, "invalid file gamma in png_set_gamma"); + + if (scrn_gamma <= 0) + png_error(png_ptr, "invalid screen gamma in png_set_gamma"); + + /* Set the gamma values unconditionally - this overrides the value in the PNG + * file if a gAMA chunk was present. png_set_alpha_mode provides a + * different, easier, way to default the file gamma. + */ + png_ptr->colorspace.gamma = file_gamma; + png_ptr->colorspace.flags |= PNG_COLORSPACE_HAVE_GAMMA; + png_ptr->screen_gamma = scrn_gamma; +} + +# ifdef PNG_FLOATING_POINT_SUPPORTED +void PNGAPI +png_set_gamma(png_structrp png_ptr, double scrn_gamma, double file_gamma) +{ + png_set_gamma_fixed(png_ptr, convert_gamma_value(png_ptr, scrn_gamma), + convert_gamma_value(png_ptr, file_gamma)); +} +# endif /* FLOATING_POINT */ +#endif /* READ_GAMMA */ + +#ifdef PNG_READ_EXPAND_SUPPORTED +/* Expand paletted images to RGB, expand grayscale images of + * less than 8-bit depth to 8-bit depth, and expand tRNS chunks + * to alpha channels. + */ +void PNGAPI +png_set_expand(png_structrp png_ptr) +{ + png_debug(1, "in png_set_expand"); + + if (png_rtran_ok(png_ptr, 0) == 0) + return; + + png_ptr->transformations |= (PNG_EXPAND | PNG_EXPAND_tRNS); +} + +/* GRR 19990627: the following three functions currently are identical + * to png_set_expand(). However, it is entirely reasonable that someone + * might wish to expand an indexed image to RGB but *not* expand a single, + * fully transparent palette entry to a full alpha channel--perhaps instead + * convert tRNS to the grayscale/RGB format (16-bit RGB value), or replace + * the transparent color with a particular RGB value, or drop tRNS entirely. + * IOW, a future version of the library may make the transformations flag + * a bit more fine-grained, with separate bits for each of these three + * functions. + * + * More to the point, these functions make it obvious what libpng will be + * doing, whereas "expand" can (and does) mean any number of things. + * + * GRP 20060307: In libpng-1.2.9, png_set_gray_1_2_4_to_8() was modified + * to expand only the sample depth but not to expand the tRNS to alpha + * and its name was changed to png_set_expand_gray_1_2_4_to_8(). + */ + +/* Expand paletted images to RGB. */ +void PNGAPI +png_set_palette_to_rgb(png_structrp png_ptr) +{ + png_debug(1, "in png_set_palette_to_rgb"); + + if (png_rtran_ok(png_ptr, 0) == 0) + return; + + png_ptr->transformations |= (PNG_EXPAND | PNG_EXPAND_tRNS); +} + +/* Expand grayscale images of less than 8-bit depth to 8 bits. */ +void PNGAPI +png_set_expand_gray_1_2_4_to_8(png_structrp png_ptr) +{ + png_debug(1, "in png_set_expand_gray_1_2_4_to_8"); + + if (png_rtran_ok(png_ptr, 0) == 0) + return; + + png_ptr->transformations |= PNG_EXPAND; +} + +/* Expand tRNS chunks to alpha channels. */ +void PNGAPI +png_set_tRNS_to_alpha(png_structrp png_ptr) +{ + png_debug(1, "in png_set_tRNS_to_alpha"); + + if (png_rtran_ok(png_ptr, 0) == 0) + return; + + png_ptr->transformations |= (PNG_EXPAND | PNG_EXPAND_tRNS); +} +#endif /* READ_EXPAND */ + +#ifdef PNG_READ_EXPAND_16_SUPPORTED +/* Expand to 16-bit channels, expand the tRNS chunk too (because otherwise + * it may not work correctly.) + */ +void PNGAPI +png_set_expand_16(png_structrp png_ptr) +{ + png_debug(1, "in png_set_expand_16"); + + if (png_rtran_ok(png_ptr, 0) == 0) + return; + + png_ptr->transformations |= (PNG_EXPAND_16 | PNG_EXPAND | PNG_EXPAND_tRNS); +} +#endif + +#ifdef PNG_READ_GRAY_TO_RGB_SUPPORTED +void PNGAPI +png_set_gray_to_rgb(png_structrp png_ptr) +{ + png_debug(1, "in png_set_gray_to_rgb"); + + if (png_rtran_ok(png_ptr, 0) == 0) + return; + + /* Because rgb must be 8 bits or more: */ + png_set_expand_gray_1_2_4_to_8(png_ptr); + png_ptr->transformations |= PNG_GRAY_TO_RGB; +} +#endif + +#ifdef PNG_READ_RGB_TO_GRAY_SUPPORTED +void PNGFAPI +png_set_rgb_to_gray_fixed(png_structrp png_ptr, int error_action, + png_fixed_point red, png_fixed_point green) +{ + png_debug(1, "in png_set_rgb_to_gray"); + + /* Need the IHDR here because of the check on color_type below. */ + /* TODO: fix this */ + if (png_rtran_ok(png_ptr, 1) == 0) + return; + + switch (error_action) + { + case PNG_ERROR_ACTION_NONE: + png_ptr->transformations |= PNG_RGB_TO_GRAY; + break; + + case PNG_ERROR_ACTION_WARN: + png_ptr->transformations |= PNG_RGB_TO_GRAY_WARN; + break; + + case PNG_ERROR_ACTION_ERROR: + png_ptr->transformations |= PNG_RGB_TO_GRAY_ERR; + break; + + default: + png_error(png_ptr, "invalid error action to rgb_to_gray"); + break; + } + + if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE) +#ifdef PNG_READ_EXPAND_SUPPORTED + png_ptr->transformations |= PNG_EXPAND; +#else + { + /* Make this an error in 1.6 because otherwise the application may assume + * that it just worked and get a memory overwrite. + */ + png_error(png_ptr, + "Cannot do RGB_TO_GRAY without EXPAND_SUPPORTED"); + + /* png_ptr->transformations &= ~PNG_RGB_TO_GRAY; */ + } +#endif + { + if (red >= 0 && green >= 0 && red + green <= PNG_FP_1) + { + png_uint_16 red_int, green_int; + + /* NOTE: this calculation does not round, but this behavior is retained + * for consistency; the inaccuracy is very small. The code here always + * overwrites the coefficients, regardless of whether they have been + * defaulted or set already. + */ + red_int = (png_uint_16)(((png_uint_32)red*32768)/100000); + green_int = (png_uint_16)(((png_uint_32)green*32768)/100000); + + png_ptr->rgb_to_gray_red_coeff = red_int; + png_ptr->rgb_to_gray_green_coeff = green_int; + png_ptr->rgb_to_gray_coefficients_set = 1; + } + + else + { + if (red >= 0 && green >= 0) + png_app_warning(png_ptr, + "ignoring out of range rgb_to_gray coefficients"); + + /* Use the defaults, from the cHRM chunk if set, else the historical + * values which are close to the sRGB/HDTV/ITU-Rec 709 values. See + * png_do_rgb_to_gray for more discussion of the values. In this case + * the coefficients are not marked as 'set' and are not overwritten if + * something has already provided a default. + */ + if (png_ptr->rgb_to_gray_red_coeff == 0 && + png_ptr->rgb_to_gray_green_coeff == 0) + { + png_ptr->rgb_to_gray_red_coeff = 6968; + png_ptr->rgb_to_gray_green_coeff = 23434; + /* png_ptr->rgb_to_gray_blue_coeff = 2366; */ + } + } + } +} + +#ifdef PNG_FLOATING_POINT_SUPPORTED +/* Convert a RGB image to a grayscale of the same width. This allows us, + * for example, to convert a 24 bpp RGB image into an 8 bpp grayscale image. + */ + +void PNGAPI +png_set_rgb_to_gray(png_structrp png_ptr, int error_action, double red, + double green) +{ + png_set_rgb_to_gray_fixed(png_ptr, error_action, + png_fixed(png_ptr, red, "rgb to gray red coefficient"), + png_fixed(png_ptr, green, "rgb to gray green coefficient")); +} +#endif /* FLOATING POINT */ + +#endif /* RGB_TO_GRAY */ + +#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED) || \ + defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED) +void PNGAPI +png_set_read_user_transform_fn(png_structrp png_ptr, png_user_transform_ptr + read_user_transform_fn) +{ + png_debug(1, "in png_set_read_user_transform_fn"); + +#ifdef PNG_READ_USER_TRANSFORM_SUPPORTED + png_ptr->transformations |= PNG_USER_TRANSFORM; + png_ptr->read_user_transform_fn = read_user_transform_fn; +#endif +} +#endif + +#ifdef PNG_READ_TRANSFORMS_SUPPORTED +#ifdef PNG_READ_GAMMA_SUPPORTED +/* In the case of gamma transformations only do transformations on images where + * the [file] gamma and screen_gamma are not close reciprocals, otherwise it + * slows things down slightly, and also needlessly introduces small errors. + */ +static int /* PRIVATE */ +png_gamma_threshold(png_fixed_point screen_gamma, png_fixed_point file_gamma) +{ + /* PNG_GAMMA_THRESHOLD is the threshold for performing gamma + * correction as a difference of the overall transform from 1.0 + * + * We want to compare the threshold with s*f - 1, if we get + * overflow here it is because of wacky gamma values so we + * turn on processing anyway. + */ + png_fixed_point gtest; + return !png_muldiv(>est, screen_gamma, file_gamma, PNG_FP_1) || + png_gamma_significant(gtest); +} +#endif + +/* Initialize everything needed for the read. This includes modifying + * the palette. + */ + +/* For the moment 'png_init_palette_transformations' and + * 'png_init_rgb_transformations' only do some flag canceling optimizations. + * The intent is that these two routines should have palette or rgb operations + * extracted from 'png_init_read_transformations'. + */ +static void /* PRIVATE */ +png_init_palette_transformations(png_structrp png_ptr) +{ + /* Called to handle the (input) palette case. In png_do_read_transformations + * the first step is to expand the palette if requested, so this code must + * take care to only make changes that are invariant with respect to the + * palette expansion, or only do them if there is no expansion. + * + * STRIP_ALPHA has already been handled in the caller (by setting num_trans + * to 0.) + */ + int input_has_alpha = 0; + int input_has_transparency = 0; + + if (png_ptr->num_trans > 0) + { + int i; + + /* Ignore if all the entries are opaque (unlikely!) */ + for (i=0; inum_trans; ++i) + { + if (png_ptr->trans_alpha[i] == 255) + continue; + else if (png_ptr->trans_alpha[i] == 0) + input_has_transparency = 1; + else + { + input_has_transparency = 1; + input_has_alpha = 1; + break; + } + } + } + + /* If no alpha we can optimize. */ + if (input_has_alpha == 0) + { + /* Any alpha means background and associative alpha processing is + * required, however if the alpha is 0 or 1 throughout OPTIMIZE_ALPHA + * and ENCODE_ALPHA are irrelevant. + */ + png_ptr->transformations &= ~PNG_ENCODE_ALPHA; + png_ptr->flags &= ~PNG_FLAG_OPTIMIZE_ALPHA; + + if (input_has_transparency == 0) + png_ptr->transformations &= ~(PNG_COMPOSE | PNG_BACKGROUND_EXPAND); + } + +#if defined(PNG_READ_EXPAND_SUPPORTED) && defined(PNG_READ_BACKGROUND_SUPPORTED) + /* png_set_background handling - deals with the complexity of whether the + * background color is in the file format or the screen format in the case + * where an 'expand' will happen. + */ + + /* The following code cannot be entered in the alpha pre-multiplication case + * because PNG_BACKGROUND_EXPAND is cancelled below. + */ + if ((png_ptr->transformations & PNG_BACKGROUND_EXPAND) != 0 && + (png_ptr->transformations & PNG_EXPAND) != 0) + { + { + png_ptr->background.red = + png_ptr->palette[png_ptr->background.index].red; + png_ptr->background.green = + png_ptr->palette[png_ptr->background.index].green; + png_ptr->background.blue = + png_ptr->palette[png_ptr->background.index].blue; + +#ifdef PNG_READ_INVERT_ALPHA_SUPPORTED + if ((png_ptr->transformations & PNG_INVERT_ALPHA) != 0) + { + if ((png_ptr->transformations & PNG_EXPAND_tRNS) == 0) + { + /* Invert the alpha channel (in tRNS) unless the pixels are + * going to be expanded, in which case leave it for later + */ + int i, istop = png_ptr->num_trans; + + for (i=0; itrans_alpha[i] = (png_byte)(255 - + png_ptr->trans_alpha[i]); + } + } +#endif /* READ_INVERT_ALPHA */ + } + } /* background expand and (therefore) no alpha association. */ +#endif /* READ_EXPAND && READ_BACKGROUND */ +} + +static void /* PRIVATE */ +png_init_rgb_transformations(png_structrp png_ptr) +{ + /* Added to libpng-1.5.4: check the color type to determine whether there + * is any alpha or transparency in the image and simply cancel the + * background and alpha mode stuff if there isn't. + */ + int input_has_alpha = (png_ptr->color_type & PNG_COLOR_MASK_ALPHA) != 0; + int input_has_transparency = png_ptr->num_trans > 0; + + /* If no alpha we can optimize. */ + if (input_has_alpha == 0) + { + /* Any alpha means background and associative alpha processing is + * required, however if the alpha is 0 or 1 throughout OPTIMIZE_ALPHA + * and ENCODE_ALPHA are irrelevant. + */ +# ifdef PNG_READ_ALPHA_MODE_SUPPORTED + png_ptr->transformations &= ~PNG_ENCODE_ALPHA; + png_ptr->flags &= ~PNG_FLAG_OPTIMIZE_ALPHA; +# endif + + if (input_has_transparency == 0) + png_ptr->transformations &= ~(PNG_COMPOSE | PNG_BACKGROUND_EXPAND); + } + +#if defined(PNG_READ_EXPAND_SUPPORTED) && defined(PNG_READ_BACKGROUND_SUPPORTED) + /* png_set_background handling - deals with the complexity of whether the + * background color is in the file format or the screen format in the case + * where an 'expand' will happen. + */ + + /* The following code cannot be entered in the alpha pre-multiplication case + * because PNG_BACKGROUND_EXPAND is cancelled below. + */ + if ((png_ptr->transformations & PNG_BACKGROUND_EXPAND) != 0 && + (png_ptr->transformations & PNG_EXPAND) != 0 && + (png_ptr->color_type & PNG_COLOR_MASK_COLOR) == 0) + /* i.e., GRAY or GRAY_ALPHA */ + { + { + /* Expand background and tRNS chunks */ + int gray = png_ptr->background.gray; + int trans_gray = png_ptr->trans_color.gray; + + switch (png_ptr->bit_depth) + { + case 1: + gray *= 0xff; + trans_gray *= 0xff; + break; + + case 2: + gray *= 0x55; + trans_gray *= 0x55; + break; + + case 4: + gray *= 0x11; + trans_gray *= 0x11; + break; + + default: + + case 8: + /* FALL THROUGH (Already 8 bits) */ + + case 16: + /* Already a full 16 bits */ + break; + } + + png_ptr->background.red = png_ptr->background.green = + png_ptr->background.blue = (png_uint_16)gray; + + if ((png_ptr->transformations & PNG_EXPAND_tRNS) == 0) + { + png_ptr->trans_color.red = png_ptr->trans_color.green = + png_ptr->trans_color.blue = (png_uint_16)trans_gray; + } + } + } /* background expand and (therefore) no alpha association. */ +#endif /* READ_EXPAND && READ_BACKGROUND */ +} + +void /* PRIVATE */ +png_init_read_transformations(png_structrp png_ptr) +{ + png_debug(1, "in png_init_read_transformations"); + + /* This internal function is called from png_read_start_row in pngrutil.c + * and it is called before the 'rowbytes' calculation is done, so the code + * in here can change or update the transformations flags. + * + * First do updates that do not depend on the details of the PNG image data + * being processed. + */ + +#ifdef PNG_READ_GAMMA_SUPPORTED + /* Prior to 1.5.4 these tests were performed from png_set_gamma, 1.5.4 adds + * png_set_alpha_mode and this is another source for a default file gamma so + * the test needs to be performed later - here. In addition prior to 1.5.4 + * the tests were repeated for the PALETTE color type here - this is no + * longer necessary (and doesn't seem to have been necessary before.) + */ + { + /* The following temporary indicates if overall gamma correction is + * required. + */ + int gamma_correction = 0; + + if (png_ptr->colorspace.gamma != 0) /* has been set */ + { + if (png_ptr->screen_gamma != 0) /* screen set too */ + gamma_correction = png_gamma_threshold(png_ptr->colorspace.gamma, + png_ptr->screen_gamma); + + else + /* Assume the output matches the input; a long time default behavior + * of libpng, although the standard has nothing to say about this. + */ + png_ptr->screen_gamma = png_reciprocal(png_ptr->colorspace.gamma); + } + + else if (png_ptr->screen_gamma != 0) + /* The converse - assume the file matches the screen, note that this + * perhaps undesireable default can (from 1.5.4) be changed by calling + * png_set_alpha_mode (even if the alpha handling mode isn't required + * or isn't changed from the default.) + */ + png_ptr->colorspace.gamma = png_reciprocal(png_ptr->screen_gamma); + + else /* neither are set */ + /* Just in case the following prevents any processing - file and screen + * are both assumed to be linear and there is no way to introduce a + * third gamma value other than png_set_background with 'UNIQUE', and, + * prior to 1.5.4 + */ + png_ptr->screen_gamma = png_ptr->colorspace.gamma = PNG_FP_1; + + /* We have a gamma value now. */ + png_ptr->colorspace.flags |= PNG_COLORSPACE_HAVE_GAMMA; + + /* Now turn the gamma transformation on or off as appropriate. Notice + * that PNG_GAMMA just refers to the file->screen correction. Alpha + * composition may independently cause gamma correction because it needs + * linear data (e.g. if the file has a gAMA chunk but the screen gamma + * hasn't been specified.) In any case this flag may get turned off in + * the code immediately below if the transform can be handled outside the + * row loop. + */ + if (gamma_correction != 0) + png_ptr->transformations |= PNG_GAMMA; + + else + png_ptr->transformations &= ~PNG_GAMMA; + } +#endif + + /* Certain transformations have the effect of preventing other + * transformations that happen afterward in png_do_read_transformations; + * resolve the interdependencies here. From the code of + * png_do_read_transformations the order is: + * + * 1) PNG_EXPAND (including PNG_EXPAND_tRNS) + * 2) PNG_STRIP_ALPHA (if no compose) + * 3) PNG_RGB_TO_GRAY + * 4) PNG_GRAY_TO_RGB iff !PNG_BACKGROUND_IS_GRAY + * 5) PNG_COMPOSE + * 6) PNG_GAMMA + * 7) PNG_STRIP_ALPHA (if compose) + * 8) PNG_ENCODE_ALPHA + * 9) PNG_SCALE_16_TO_8 + * 10) PNG_16_TO_8 + * 11) PNG_QUANTIZE (converts to palette) + * 12) PNG_EXPAND_16 + * 13) PNG_GRAY_TO_RGB iff PNG_BACKGROUND_IS_GRAY + * 14) PNG_INVERT_MONO + * 15) PNG_INVERT_ALPHA + * 16) PNG_SHIFT + * 17) PNG_PACK + * 18) PNG_BGR + * 19) PNG_PACKSWAP + * 20) PNG_FILLER (includes PNG_ADD_ALPHA) + * 21) PNG_SWAP_ALPHA + * 22) PNG_SWAP_BYTES + * 23) PNG_USER_TRANSFORM [must be last] + */ +#ifdef PNG_READ_STRIP_ALPHA_SUPPORTED + if ((png_ptr->transformations & PNG_STRIP_ALPHA) != 0 && + (png_ptr->transformations & PNG_COMPOSE) == 0) + { + /* Stripping the alpha channel happens immediately after the 'expand' + * transformations, before all other transformation, so it cancels out + * the alpha handling. It has the side effect negating the effect of + * PNG_EXPAND_tRNS too: + */ + png_ptr->transformations &= ~(PNG_BACKGROUND_EXPAND | PNG_ENCODE_ALPHA | + PNG_EXPAND_tRNS); + png_ptr->flags &= ~PNG_FLAG_OPTIMIZE_ALPHA; + + /* Kill the tRNS chunk itself too. Prior to 1.5.4 this did not happen + * so transparency information would remain just so long as it wasn't + * expanded. This produces unexpected API changes if the set of things + * that do PNG_EXPAND_tRNS changes (perfectly possible given the + * documentation - which says ask for what you want, accept what you + * get.) This makes the behavior consistent from 1.5.4: + */ + png_ptr->num_trans = 0; + } +#endif /* STRIP_ALPHA supported, no COMPOSE */ + +#ifdef PNG_READ_ALPHA_MODE_SUPPORTED + /* If the screen gamma is about 1.0 then the OPTIMIZE_ALPHA and ENCODE_ALPHA + * settings will have no effect. + */ + if (png_gamma_significant(png_ptr->screen_gamma) == 0) + { + png_ptr->transformations &= ~PNG_ENCODE_ALPHA; + png_ptr->flags &= ~PNG_FLAG_OPTIMIZE_ALPHA; + } +#endif + +#ifdef PNG_READ_RGB_TO_GRAY_SUPPORTED + /* Make sure the coefficients for the rgb to gray conversion are set + * appropriately. + */ + if ((png_ptr->transformations & PNG_RGB_TO_GRAY) != 0) + png_colorspace_set_rgb_coefficients(png_ptr); +#endif + +#ifdef PNG_READ_GRAY_TO_RGB_SUPPORTED +#if defined(PNG_READ_EXPAND_SUPPORTED) && defined(PNG_READ_BACKGROUND_SUPPORTED) + /* Detect gray background and attempt to enable optimization for + * gray --> RGB case. + * + * Note: if PNG_BACKGROUND_EXPAND is set and color_type is either RGB or + * RGB_ALPHA (in which case need_expand is superfluous anyway), the + * background color might actually be gray yet not be flagged as such. + * This is not a problem for the current code, which uses + * PNG_BACKGROUND_IS_GRAY only to decide when to do the + * png_do_gray_to_rgb() transformation. + * + * TODO: this code needs to be revised to avoid the complexity and + * interdependencies. The color type of the background should be recorded in + * png_set_background, along with the bit depth, then the code has a record + * of exactly what color space the background is currently in. + */ + if ((png_ptr->transformations & PNG_BACKGROUND_EXPAND) != 0) + { + /* PNG_BACKGROUND_EXPAND: the background is in the file color space, so if + * the file was grayscale the background value is gray. + */ + if ((png_ptr->color_type & PNG_COLOR_MASK_COLOR) == 0) + png_ptr->mode |= PNG_BACKGROUND_IS_GRAY; + } + + else if ((png_ptr->transformations & PNG_COMPOSE) != 0) + { + /* PNG_COMPOSE: png_set_background was called with need_expand false, + * so the color is in the color space of the output or png_set_alpha_mode + * was called and the color is black. Ignore RGB_TO_GRAY because that + * happens before GRAY_TO_RGB. + */ + if ((png_ptr->transformations & PNG_GRAY_TO_RGB) != 0) + { + if (png_ptr->background.red == png_ptr->background.green && + png_ptr->background.red == png_ptr->background.blue) + { + png_ptr->mode |= PNG_BACKGROUND_IS_GRAY; + png_ptr->background.gray = png_ptr->background.red; + } + } + } +#endif /* READ_EXPAND && READ_BACKGROUND */ +#endif /* READ_GRAY_TO_RGB */ + + /* For indexed PNG data (PNG_COLOR_TYPE_PALETTE) many of the transformations + * can be performed directly on the palette, and some (such as rgb to gray) + * can be optimized inside the palette. This is particularly true of the + * composite (background and alpha) stuff, which can be pretty much all done + * in the palette even if the result is expanded to RGB or gray afterward. + * + * NOTE: this is Not Yet Implemented, the code behaves as in 1.5.1 and + * earlier and the palette stuff is actually handled on the first row. This + * leads to the reported bug that the palette returned by png_get_PLTE is not + * updated. + */ + if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE) + png_init_palette_transformations(png_ptr); + + else + png_init_rgb_transformations(png_ptr); + +#if defined(PNG_READ_BACKGROUND_SUPPORTED) && \ + defined(PNG_READ_EXPAND_16_SUPPORTED) + if ((png_ptr->transformations & PNG_EXPAND_16) != 0 && + (png_ptr->transformations & PNG_COMPOSE) != 0 && + (png_ptr->transformations & PNG_BACKGROUND_EXPAND) == 0 && + png_ptr->bit_depth != 16) + { + /* TODO: fix this. Because the expand_16 operation is after the compose + * handling the background color must be 8, not 16, bits deep, but the + * application will supply a 16-bit value so reduce it here. + * + * The PNG_BACKGROUND_EXPAND code above does not expand to 16 bits at + * present, so that case is ok (until do_expand_16 is moved.) + * + * NOTE: this discards the low 16 bits of the user supplied background + * color, but until expand_16 works properly there is no choice! + */ +# define CHOP(x) (x)=((png_uint_16)PNG_DIV257(x)) + CHOP(png_ptr->background.red); + CHOP(png_ptr->background.green); + CHOP(png_ptr->background.blue); + CHOP(png_ptr->background.gray); +# undef CHOP + } +#endif /* READ_BACKGROUND && READ_EXPAND_16 */ + +#if defined(PNG_READ_BACKGROUND_SUPPORTED) && \ + (defined(PNG_READ_SCALE_16_TO_8_SUPPORTED) || \ + defined(PNG_READ_STRIP_16_TO_8_SUPPORTED)) + if ((png_ptr->transformations & (PNG_16_TO_8|PNG_SCALE_16_TO_8)) != 0 && + (png_ptr->transformations & PNG_COMPOSE) != 0 && + (png_ptr->transformations & PNG_BACKGROUND_EXPAND) == 0 && + png_ptr->bit_depth == 16) + { + /* On the other hand, if a 16-bit file is to be reduced to 8-bits per + * component this will also happen after PNG_COMPOSE and so the background + * color must be pre-expanded here. + * + * TODO: fix this too. + */ + png_ptr->background.red = (png_uint_16)(png_ptr->background.red * 257); + png_ptr->background.green = + (png_uint_16)(png_ptr->background.green * 257); + png_ptr->background.blue = (png_uint_16)(png_ptr->background.blue * 257); + png_ptr->background.gray = (png_uint_16)(png_ptr->background.gray * 257); + } +#endif + + /* NOTE: below 'PNG_READ_ALPHA_MODE_SUPPORTED' is presumed to also enable the + * background support (see the comments in scripts/pnglibconf.dfa), this + * allows pre-multiplication of the alpha channel to be implemented as + * compositing on black. This is probably sub-optimal and has been done in + * 1.5.4 betas simply to enable external critique and testing (i.e. to + * implement the new API quickly, without lots of internal changes.) + */ + +#ifdef PNG_READ_GAMMA_SUPPORTED +# ifdef PNG_READ_BACKGROUND_SUPPORTED + /* Includes ALPHA_MODE */ + png_ptr->background_1 = png_ptr->background; +# endif + + /* This needs to change - in the palette image case a whole set of tables are + * built when it would be quicker to just calculate the correct value for + * each palette entry directly. Also, the test is too tricky - why check + * PNG_RGB_TO_GRAY if PNG_GAMMA is not set? The answer seems to be that + * PNG_GAMMA is cancelled even if the gamma is known? The test excludes the + * PNG_COMPOSE case, so apparently if there is no *overall* gamma correction + * the gamma tables will not be built even if composition is required on a + * gamma encoded value. + * + * In 1.5.4 this is addressed below by an additional check on the individual + * file gamma - if it is not 1.0 both RGB_TO_GRAY and COMPOSE need the + * tables. + */ + if ((png_ptr->transformations & PNG_GAMMA) != 0 || + ((png_ptr->transformations & PNG_RGB_TO_GRAY) != 0 && + (png_gamma_significant(png_ptr->colorspace.gamma) != 0 || + png_gamma_significant(png_ptr->screen_gamma) != 0)) || + ((png_ptr->transformations & PNG_COMPOSE) != 0 && + (png_gamma_significant(png_ptr->colorspace.gamma) != 0 || + png_gamma_significant(png_ptr->screen_gamma) != 0 +# ifdef PNG_READ_BACKGROUND_SUPPORTED + || (png_ptr->background_gamma_type == PNG_BACKGROUND_GAMMA_UNIQUE && + png_gamma_significant(png_ptr->background_gamma) != 0) +# endif + )) || ((png_ptr->transformations & PNG_ENCODE_ALPHA) != 0 && + png_gamma_significant(png_ptr->screen_gamma) != 0)) + { + png_build_gamma_table(png_ptr, png_ptr->bit_depth); + +#ifdef PNG_READ_BACKGROUND_SUPPORTED + if ((png_ptr->transformations & PNG_COMPOSE) != 0) + { + /* Issue a warning about this combination: because RGB_TO_GRAY is + * optimized to do the gamma transform if present yet do_background has + * to do the same thing if both options are set a + * double-gamma-correction happens. This is true in all versions of + * libpng to date. + */ + if ((png_ptr->transformations & PNG_RGB_TO_GRAY) != 0) + png_warning(png_ptr, + "libpng does not support gamma+background+rgb_to_gray"); + + if ((png_ptr->color_type == PNG_COLOR_TYPE_PALETTE) != 0) + { + /* We don't get to here unless there is a tRNS chunk with non-opaque + * entries - see the checking code at the start of this function. + */ + png_color back, back_1; + png_colorp palette = png_ptr->palette; + int num_palette = png_ptr->num_palette; + int i; + if (png_ptr->background_gamma_type == PNG_BACKGROUND_GAMMA_FILE) + { + + back.red = png_ptr->gamma_table[png_ptr->background.red]; + back.green = png_ptr->gamma_table[png_ptr->background.green]; + back.blue = png_ptr->gamma_table[png_ptr->background.blue]; + + back_1.red = png_ptr->gamma_to_1[png_ptr->background.red]; + back_1.green = png_ptr->gamma_to_1[png_ptr->background.green]; + back_1.blue = png_ptr->gamma_to_1[png_ptr->background.blue]; + } + else + { + png_fixed_point g, gs; + + switch (png_ptr->background_gamma_type) + { + case PNG_BACKGROUND_GAMMA_SCREEN: + g = (png_ptr->screen_gamma); + gs = PNG_FP_1; + break; + + case PNG_BACKGROUND_GAMMA_FILE: + g = png_reciprocal(png_ptr->colorspace.gamma); + gs = png_reciprocal2(png_ptr->colorspace.gamma, + png_ptr->screen_gamma); + break; + + case PNG_BACKGROUND_GAMMA_UNIQUE: + g = png_reciprocal(png_ptr->background_gamma); + gs = png_reciprocal2(png_ptr->background_gamma, + png_ptr->screen_gamma); + break; + default: + g = PNG_FP_1; /* back_1 */ + gs = PNG_FP_1; /* back */ + break; + } + + if (png_gamma_significant(gs) != 0) + { + back.red = png_gamma_8bit_correct(png_ptr->background.red, + gs); + back.green = png_gamma_8bit_correct(png_ptr->background.green, + gs); + back.blue = png_gamma_8bit_correct(png_ptr->background.blue, + gs); + } + + else + { + back.red = (png_byte)png_ptr->background.red; + back.green = (png_byte)png_ptr->background.green; + back.blue = (png_byte)png_ptr->background.blue; + } + + if (png_gamma_significant(g) != 0) + { + back_1.red = png_gamma_8bit_correct(png_ptr->background.red, + g); + back_1.green = png_gamma_8bit_correct( + png_ptr->background.green, g); + back_1.blue = png_gamma_8bit_correct(png_ptr->background.blue, + g); + } + + else + { + back_1.red = (png_byte)png_ptr->background.red; + back_1.green = (png_byte)png_ptr->background.green; + back_1.blue = (png_byte)png_ptr->background.blue; + } + } + + for (i = 0; i < num_palette; i++) + { + if (i < (int)png_ptr->num_trans && + png_ptr->trans_alpha[i] != 0xff) + { + if (png_ptr->trans_alpha[i] == 0) + { + palette[i] = back; + } + else /* if (png_ptr->trans_alpha[i] != 0xff) */ + { + png_byte v, w; + + v = png_ptr->gamma_to_1[palette[i].red]; + png_composite(w, v, png_ptr->trans_alpha[i], back_1.red); + palette[i].red = png_ptr->gamma_from_1[w]; + + v = png_ptr->gamma_to_1[palette[i].green]; + png_composite(w, v, png_ptr->trans_alpha[i], back_1.green); + palette[i].green = png_ptr->gamma_from_1[w]; + + v = png_ptr->gamma_to_1[palette[i].blue]; + png_composite(w, v, png_ptr->trans_alpha[i], back_1.blue); + palette[i].blue = png_ptr->gamma_from_1[w]; + } + } + else + { + palette[i].red = png_ptr->gamma_table[palette[i].red]; + palette[i].green = png_ptr->gamma_table[palette[i].green]; + palette[i].blue = png_ptr->gamma_table[palette[i].blue]; + } + } + + /* Prevent the transformations being done again. + * + * NOTE: this is highly dubious; it removes the transformations in + * place. This seems inconsistent with the general treatment of the + * transformations elsewhere. + */ + png_ptr->transformations &= ~(PNG_COMPOSE | PNG_GAMMA); + } /* color_type == PNG_COLOR_TYPE_PALETTE */ + + /* if (png_ptr->background_gamma_type!=PNG_BACKGROUND_GAMMA_UNKNOWN) */ + else /* color_type != PNG_COLOR_TYPE_PALETTE */ + { + int gs_sig, g_sig; + png_fixed_point g = PNG_FP_1; /* Correction to linear */ + png_fixed_point gs = PNG_FP_1; /* Correction to screen */ + + switch (png_ptr->background_gamma_type) + { + case PNG_BACKGROUND_GAMMA_SCREEN: + g = png_ptr->screen_gamma; + /* gs = PNG_FP_1; */ + break; + + case PNG_BACKGROUND_GAMMA_FILE: + g = png_reciprocal(png_ptr->colorspace.gamma); + gs = png_reciprocal2(png_ptr->colorspace.gamma, + png_ptr->screen_gamma); + break; + + case PNG_BACKGROUND_GAMMA_UNIQUE: + g = png_reciprocal(png_ptr->background_gamma); + gs = png_reciprocal2(png_ptr->background_gamma, + png_ptr->screen_gamma); + break; + + default: + png_error(png_ptr, "invalid background gamma type"); + } + + g_sig = png_gamma_significant(g); + gs_sig = png_gamma_significant(gs); + + if (g_sig != 0) + png_ptr->background_1.gray = png_gamma_correct(png_ptr, + png_ptr->background.gray, g); + + if (gs_sig != 0) + png_ptr->background.gray = png_gamma_correct(png_ptr, + png_ptr->background.gray, gs); + + if ((png_ptr->background.red != png_ptr->background.green) || + (png_ptr->background.red != png_ptr->background.blue) || + (png_ptr->background.red != png_ptr->background.gray)) + { + /* RGB or RGBA with color background */ + if (g_sig != 0) + { + png_ptr->background_1.red = png_gamma_correct(png_ptr, + png_ptr->background.red, g); + + png_ptr->background_1.green = png_gamma_correct(png_ptr, + png_ptr->background.green, g); + + png_ptr->background_1.blue = png_gamma_correct(png_ptr, + png_ptr->background.blue, g); + } + + if (gs_sig != 0) + { + png_ptr->background.red = png_gamma_correct(png_ptr, + png_ptr->background.red, gs); + + png_ptr->background.green = png_gamma_correct(png_ptr, + png_ptr->background.green, gs); + + png_ptr->background.blue = png_gamma_correct(png_ptr, + png_ptr->background.blue, gs); + } + } + + else + { + /* GRAY, GRAY ALPHA, RGB, or RGBA with gray background */ + png_ptr->background_1.red = png_ptr->background_1.green + = png_ptr->background_1.blue = png_ptr->background_1.gray; + + png_ptr->background.red = png_ptr->background.green + = png_ptr->background.blue = png_ptr->background.gray; + } + + /* The background is now in screen gamma: */ + png_ptr->background_gamma_type = PNG_BACKGROUND_GAMMA_SCREEN; + } /* color_type != PNG_COLOR_TYPE_PALETTE */ + }/* png_ptr->transformations & PNG_BACKGROUND */ + + else + /* Transformation does not include PNG_BACKGROUND */ +#endif /* READ_BACKGROUND */ + if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE +#ifdef PNG_READ_RGB_TO_GRAY_SUPPORTED + /* RGB_TO_GRAY needs to have non-gamma-corrected values! */ + && ((png_ptr->transformations & PNG_EXPAND) == 0 || + (png_ptr->transformations & PNG_RGB_TO_GRAY) == 0) +#endif + ) + { + png_colorp palette = png_ptr->palette; + int num_palette = png_ptr->num_palette; + int i; + + /* NOTE: there are other transformations that should probably be in + * here too. + */ + for (i = 0; i < num_palette; i++) + { + palette[i].red = png_ptr->gamma_table[palette[i].red]; + palette[i].green = png_ptr->gamma_table[palette[i].green]; + palette[i].blue = png_ptr->gamma_table[palette[i].blue]; + } + + /* Done the gamma correction. */ + png_ptr->transformations &= ~PNG_GAMMA; + } /* color_type == PALETTE && !PNG_BACKGROUND transformation */ + } +#ifdef PNG_READ_BACKGROUND_SUPPORTED + else +#endif +#endif /* READ_GAMMA */ + +#ifdef PNG_READ_BACKGROUND_SUPPORTED + /* No GAMMA transformation (see the hanging else 4 lines above) */ + if ((png_ptr->transformations & PNG_COMPOSE) != 0 && + (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)) + { + int i; + int istop = (int)png_ptr->num_trans; + png_color back; + png_colorp palette = png_ptr->palette; + + back.red = (png_byte)png_ptr->background.red; + back.green = (png_byte)png_ptr->background.green; + back.blue = (png_byte)png_ptr->background.blue; + + for (i = 0; i < istop; i++) + { + if (png_ptr->trans_alpha[i] == 0) + { + palette[i] = back; + } + + else if (png_ptr->trans_alpha[i] != 0xff) + { + /* The png_composite() macro is defined in png.h */ + png_composite(palette[i].red, palette[i].red, + png_ptr->trans_alpha[i], back.red); + + png_composite(palette[i].green, palette[i].green, + png_ptr->trans_alpha[i], back.green); + + png_composite(palette[i].blue, palette[i].blue, + png_ptr->trans_alpha[i], back.blue); + } + } + + png_ptr->transformations &= ~PNG_COMPOSE; + } +#endif /* READ_BACKGROUND */ + +#ifdef PNG_READ_SHIFT_SUPPORTED + if ((png_ptr->transformations & PNG_SHIFT) != 0 && + (png_ptr->transformations & PNG_EXPAND) == 0 && + (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)) + { + int i; + int istop = png_ptr->num_palette; + int shift = 8 - png_ptr->sig_bit.red; + + png_ptr->transformations &= ~PNG_SHIFT; + + /* significant bits can be in the range 1 to 7 for a meaninful result, if + * the number of significant bits is 0 then no shift is done (this is an + * error condition which is silently ignored.) + */ + if (shift > 0 && shift < 8) + for (i=0; ipalette[i].red; + + component >>= shift; + png_ptr->palette[i].red = (png_byte)component; + } + + shift = 8 - png_ptr->sig_bit.green; + if (shift > 0 && shift < 8) + for (i=0; ipalette[i].green; + + component >>= shift; + png_ptr->palette[i].green = (png_byte)component; + } + + shift = 8 - png_ptr->sig_bit.blue; + if (shift > 0 && shift < 8) + for (i=0; ipalette[i].blue; + + component >>= shift; + png_ptr->palette[i].blue = (png_byte)component; + } + } +#endif /* READ_SHIFT */ +} + +/* Modify the info structure to reflect the transformations. The + * info should be updated so a PNG file could be written with it, + * assuming the transformations result in valid PNG data. + */ +void /* PRIVATE */ +png_read_transform_info(png_structrp png_ptr, png_inforp info_ptr) +{ + png_debug(1, "in png_read_transform_info"); + +#ifdef PNG_READ_EXPAND_SUPPORTED + if ((png_ptr->transformations & PNG_EXPAND) != 0) + { + if (info_ptr->color_type == PNG_COLOR_TYPE_PALETTE) + { + /* This check must match what actually happens in + * png_do_expand_palette; if it ever checks the tRNS chunk to see if + * it is all opaque we must do the same (at present it does not.) + */ + if (png_ptr->num_trans > 0) + info_ptr->color_type = PNG_COLOR_TYPE_RGB_ALPHA; + + else + info_ptr->color_type = PNG_COLOR_TYPE_RGB; + + info_ptr->bit_depth = 8; + info_ptr->num_trans = 0; + + if (png_ptr->palette == NULL) + png_error (png_ptr, "Palette is NULL in indexed image"); + } + else + { + if (png_ptr->num_trans != 0) + { + if ((png_ptr->transformations & PNG_EXPAND_tRNS) != 0) + info_ptr->color_type |= PNG_COLOR_MASK_ALPHA; + } + if (info_ptr->bit_depth < 8) + info_ptr->bit_depth = 8; + + info_ptr->num_trans = 0; + } + } +#endif + +#if defined(PNG_READ_BACKGROUND_SUPPORTED) ||\ + defined(PNG_READ_ALPHA_MODE_SUPPORTED) + /* The following is almost certainly wrong unless the background value is in + * the screen space! + */ + if ((png_ptr->transformations & PNG_COMPOSE) != 0) + info_ptr->background = png_ptr->background; +#endif + +#ifdef PNG_READ_GAMMA_SUPPORTED + /* The following used to be conditional on PNG_GAMMA (prior to 1.5.4), + * however it seems that the code in png_init_read_transformations, which has + * been called before this from png_read_update_info->png_read_start_row + * sometimes does the gamma transform and cancels the flag. + * + * TODO: this looks wrong; the info_ptr should end up with a gamma equal to + * the screen_gamma value. The following probably results in weirdness if + * the info_ptr is used by the app after the rows have been read. + */ + info_ptr->colorspace.gamma = png_ptr->colorspace.gamma; +#endif + + if (info_ptr->bit_depth == 16) + { +# ifdef PNG_READ_16BIT_SUPPORTED +# ifdef PNG_READ_SCALE_16_TO_8_SUPPORTED + if ((png_ptr->transformations & PNG_SCALE_16_TO_8) != 0) + info_ptr->bit_depth = 8; +# endif + +# ifdef PNG_READ_STRIP_16_TO_8_SUPPORTED + if ((png_ptr->transformations & PNG_16_TO_8) != 0) + info_ptr->bit_depth = 8; +# endif + +# else + /* No 16 bit support: force chopping 16-bit input down to 8, in this case + * the app program can chose if both APIs are available by setting the + * correct scaling to use. + */ +# ifdef PNG_READ_STRIP_16_TO_8_SUPPORTED + /* For compatibility with previous versions use the strip method by + * default. This code works because if PNG_SCALE_16_TO_8 is already + * set the code below will do that in preference to the chop. + */ + png_ptr->transformations |= PNG_16_TO_8; + info_ptr->bit_depth = 8; +# else + +# ifdef PNG_READ_SCALE_16_TO_8_SUPPORTED + png_ptr->transformations |= PNG_SCALE_16_TO_8; + info_ptr->bit_depth = 8; +# else + + CONFIGURATION ERROR: you must enable at least one 16 to 8 method +# endif +# endif +#endif /* !READ_16BIT */ + } + +#ifdef PNG_READ_GRAY_TO_RGB_SUPPORTED + if ((png_ptr->transformations & PNG_GRAY_TO_RGB) != 0) + info_ptr->color_type = (png_byte)(info_ptr->color_type | + PNG_COLOR_MASK_COLOR); +#endif + +#ifdef PNG_READ_RGB_TO_GRAY_SUPPORTED + if ((png_ptr->transformations & PNG_RGB_TO_GRAY) != 0) + info_ptr->color_type = (png_byte)(info_ptr->color_type & + ~PNG_COLOR_MASK_COLOR); +#endif + +#ifdef PNG_READ_QUANTIZE_SUPPORTED + if ((png_ptr->transformations & PNG_QUANTIZE) != 0) + { + if (((info_ptr->color_type == PNG_COLOR_TYPE_RGB) || + (info_ptr->color_type == PNG_COLOR_TYPE_RGB_ALPHA)) && + png_ptr->palette_lookup != 0 && info_ptr->bit_depth == 8) + { + info_ptr->color_type = PNG_COLOR_TYPE_PALETTE; + } + } +#endif + +#ifdef PNG_READ_EXPAND_16_SUPPORTED + if ((png_ptr->transformations & PNG_EXPAND_16) != 0 && + info_ptr->bit_depth == 8 && + info_ptr->color_type != PNG_COLOR_TYPE_PALETTE) + { + info_ptr->bit_depth = 16; + } +#endif + +#ifdef PNG_READ_PACK_SUPPORTED + if ((png_ptr->transformations & PNG_PACK) != 0 && + (info_ptr->bit_depth < 8)) + info_ptr->bit_depth = 8; +#endif + + if (info_ptr->color_type == PNG_COLOR_TYPE_PALETTE) + info_ptr->channels = 1; + + else if ((info_ptr->color_type & PNG_COLOR_MASK_COLOR) != 0) + info_ptr->channels = 3; + + else + info_ptr->channels = 1; + +#ifdef PNG_READ_STRIP_ALPHA_SUPPORTED + if ((png_ptr->transformations & PNG_STRIP_ALPHA) != 0) + { + info_ptr->color_type = (png_byte)(info_ptr->color_type & + ~PNG_COLOR_MASK_ALPHA); + info_ptr->num_trans = 0; + } +#endif + + if ((info_ptr->color_type & PNG_COLOR_MASK_ALPHA) != 0) + info_ptr->channels++; + +#ifdef PNG_READ_FILLER_SUPPORTED + /* STRIP_ALPHA and FILLER allowed: MASK_ALPHA bit stripped above */ + if ((png_ptr->transformations & PNG_FILLER) != 0 && + (info_ptr->color_type == PNG_COLOR_TYPE_RGB || + info_ptr->color_type == PNG_COLOR_TYPE_GRAY)) + { + info_ptr->channels++; + /* If adding a true alpha channel not just filler */ + if ((png_ptr->transformations & PNG_ADD_ALPHA) != 0) + info_ptr->color_type |= PNG_COLOR_MASK_ALPHA; + } +#endif + +#if defined(PNG_USER_TRANSFORM_PTR_SUPPORTED) && \ +defined(PNG_READ_USER_TRANSFORM_SUPPORTED) + if ((png_ptr->transformations & PNG_USER_TRANSFORM) != 0) + { + if (info_ptr->bit_depth < png_ptr->user_transform_depth) + info_ptr->bit_depth = png_ptr->user_transform_depth; + + if (info_ptr->channels < png_ptr->user_transform_channels) + info_ptr->channels = png_ptr->user_transform_channels; + } +#endif + + info_ptr->pixel_depth = (png_byte)(info_ptr->channels * + info_ptr->bit_depth); + + info_ptr->rowbytes = PNG_ROWBYTES(info_ptr->pixel_depth, info_ptr->width); + + /* Adding in 1.5.4: cache the above value in png_struct so that we can later + * check in png_rowbytes that the user buffer won't get overwritten. Note + * that the field is not always set - if png_read_update_info isn't called + * the application has to either not do any transforms or get the calculation + * right itself. + */ + png_ptr->info_rowbytes = info_ptr->rowbytes; + +#ifndef PNG_READ_EXPAND_SUPPORTED + if (png_ptr != NULL) + return; +#endif +} + +#ifdef PNG_READ_PACK_SUPPORTED +/* Unpack pixels of 1, 2, or 4 bits per pixel into 1 byte per pixel, + * without changing the actual values. Thus, if you had a row with + * a bit depth of 1, you would end up with bytes that only contained + * the numbers 0 or 1. If you would rather they contain 0 and 255, use + * png_do_shift() after this. + */ +static void +png_do_unpack(png_row_infop row_info, png_bytep row) +{ + png_debug(1, "in png_do_unpack"); + + if (row_info->bit_depth < 8) + { + png_uint_32 i; + png_uint_32 row_width=row_info->width; + + switch (row_info->bit_depth) + { + case 1: + { + png_bytep sp = row + (png_size_t)((row_width - 1) >> 3); + png_bytep dp = row + (png_size_t)row_width - 1; + png_uint_32 shift = 7 - (int)((row_width + 7) & 0x07); + for (i = 0; i < row_width; i++) + { + *dp = (png_byte)((*sp >> shift) & 0x01); + + if (shift == 7) + { + shift = 0; + sp--; + } + + else + shift++; + + dp--; + } + break; + } + + case 2: + { + + png_bytep sp = row + (png_size_t)((row_width - 1) >> 2); + png_bytep dp = row + (png_size_t)row_width - 1; + png_uint_32 shift = (int)((3 - ((row_width + 3) & 0x03)) << 1); + for (i = 0; i < row_width; i++) + { + *dp = (png_byte)((*sp >> shift) & 0x03); + + if (shift == 6) + { + shift = 0; + sp--; + } + + else + shift += 2; + + dp--; + } + break; + } + + case 4: + { + png_bytep sp = row + (png_size_t)((row_width - 1) >> 1); + png_bytep dp = row + (png_size_t)row_width - 1; + png_uint_32 shift = (int)((1 - ((row_width + 1) & 0x01)) << 2); + for (i = 0; i < row_width; i++) + { + *dp = (png_byte)((*sp >> shift) & 0x0f); + + if (shift == 4) + { + shift = 0; + sp--; + } + + else + shift = 4; + + dp--; + } + break; + } + + default: + break; + } + row_info->bit_depth = 8; + row_info->pixel_depth = (png_byte)(8 * row_info->channels); + row_info->rowbytes = row_width * row_info->channels; + } +} +#endif + +#ifdef PNG_READ_SHIFT_SUPPORTED +/* Reverse the effects of png_do_shift. This routine merely shifts the + * pixels back to their significant bits values. Thus, if you have + * a row of bit depth 8, but only 5 are significant, this will shift + * the values back to 0 through 31. + */ +static void +png_do_unshift(png_row_infop row_info, png_bytep row, + png_const_color_8p sig_bits) +{ + int color_type; + + png_debug(1, "in png_do_unshift"); + + /* The palette case has already been handled in the _init routine. */ + color_type = row_info->color_type; + + if (color_type != PNG_COLOR_TYPE_PALETTE) + { + int shift[4]; + int channels = 0; + int bit_depth = row_info->bit_depth; + + if ((color_type & PNG_COLOR_MASK_COLOR) != 0) + { + shift[channels++] = bit_depth - sig_bits->red; + shift[channels++] = bit_depth - sig_bits->green; + shift[channels++] = bit_depth - sig_bits->blue; + } + + else + { + shift[channels++] = bit_depth - sig_bits->gray; + } + + if ((color_type & PNG_COLOR_MASK_ALPHA) != 0) + { + shift[channels++] = bit_depth - sig_bits->alpha; + } + + { + int c, have_shift; + + for (c = have_shift = 0; c < channels; ++c) + { + /* A shift of more than the bit depth is an error condition but it + * gets ignored here. + */ + if (shift[c] <= 0 || shift[c] >= bit_depth) + shift[c] = 0; + + else + have_shift = 1; + } + + if (have_shift == 0) + return; + } + + switch (bit_depth) + { + default: + /* Must be 1bpp gray: should not be here! */ + /* NOTREACHED */ + break; + + case 2: + /* Must be 2bpp gray */ + /* assert(channels == 1 && shift[0] == 1) */ + { + png_bytep bp = row; + png_bytep bp_end = bp + row_info->rowbytes; + + while (bp < bp_end) + { + int b = (*bp >> 1) & 0x55; + *bp++ = (png_byte)b; + } + break; + } + + case 4: + /* Must be 4bpp gray */ + /* assert(channels == 1) */ + { + png_bytep bp = row; + png_bytep bp_end = bp + row_info->rowbytes; + int gray_shift = shift[0]; + int mask = 0xf >> gray_shift; + + mask |= mask << 4; + + while (bp < bp_end) + { + int b = (*bp >> gray_shift) & mask; + *bp++ = (png_byte)b; + } + break; + } + + case 8: + /* Single byte components, G, GA, RGB, RGBA */ + { + png_bytep bp = row; + png_bytep bp_end = bp + row_info->rowbytes; + int channel = 0; + + while (bp < bp_end) + { + int b = *bp >> shift[channel]; + if (++channel >= channels) + channel = 0; + *bp++ = (png_byte)b; + } + break; + } + +#ifdef PNG_READ_16BIT_SUPPORTED + case 16: + /* Double byte components, G, GA, RGB, RGBA */ + { + png_bytep bp = row; + png_bytep bp_end = bp + row_info->rowbytes; + int channel = 0; + + while (bp < bp_end) + { + int value = (bp[0] << 8) + bp[1]; + + value >>= shift[channel]; + if (++channel >= channels) + channel = 0; + *bp++ = (png_byte)(value >> 8); + *bp++ = (png_byte)(value & 0xff); + } + break; + } +#endif + } + } +} +#endif + +#ifdef PNG_READ_SCALE_16_TO_8_SUPPORTED +/* Scale rows of bit depth 16 down to 8 accurately */ +static void +png_do_scale_16_to_8(png_row_infop row_info, png_bytep row) +{ + png_debug(1, "in png_do_scale_16_to_8"); + + if (row_info->bit_depth == 16) + { + png_bytep sp = row; /* source */ + png_bytep dp = row; /* destination */ + png_bytep ep = sp + row_info->rowbytes; /* end+1 */ + + while (sp < ep) + { + /* The input is an array of 16 bit components, these must be scaled to + * 8 bits each. For a 16 bit value V the required value (from the PNG + * specification) is: + * + * (V * 255) / 65535 + * + * This reduces to round(V / 257), or floor((V + 128.5)/257) + * + * Represent V as the two byte value vhi.vlo. Make a guess that the + * result is the top byte of V, vhi, then the correction to this value + * is: + * + * error = floor(((V-vhi.vhi) + 128.5) / 257) + * = floor(((vlo-vhi) + 128.5) / 257) + * + * This can be approximated using integer arithmetic (and a signed + * shift): + * + * error = (vlo-vhi+128) >> 8; + * + * The approximate differs from the exact answer only when (vlo-vhi) is + * 128; it then gives a correction of +1 when the exact correction is + * 0. This gives 128 errors. The exact answer (correct for all 16 bit + * input values) is: + * + * error = (vlo-vhi+128)*65535 >> 24; + * + * An alternative arithmetic calculation which also gives no errors is: + * + * (V * 255 + 32895) >> 16 + */ + + png_int_32 tmp = *sp++; /* must be signed! */ + tmp += (((int)*sp++ - tmp + 128) * 65535) >> 24; + *dp++ = (png_byte)tmp; + } + + row_info->bit_depth = 8; + row_info->pixel_depth = (png_byte)(8 * row_info->channels); + row_info->rowbytes = row_info->width * row_info->channels; + } +} +#endif + +#ifdef PNG_READ_STRIP_16_TO_8_SUPPORTED +static void +/* Simply discard the low byte. This was the default behavior prior + * to libpng-1.5.4. + */ +png_do_chop(png_row_infop row_info, png_bytep row) +{ + png_debug(1, "in png_do_chop"); + + if (row_info->bit_depth == 16) + { + png_bytep sp = row; /* source */ + png_bytep dp = row; /* destination */ + png_bytep ep = sp + row_info->rowbytes; /* end+1 */ + + while (sp < ep) + { + *dp++ = *sp; + sp += 2; /* skip low byte */ + } + + row_info->bit_depth = 8; + row_info->pixel_depth = (png_byte)(8 * row_info->channels); + row_info->rowbytes = row_info->width * row_info->channels; + } +} +#endif + +#ifdef PNG_READ_SWAP_ALPHA_SUPPORTED +static void +png_do_read_swap_alpha(png_row_infop row_info, png_bytep row) +{ + png_debug(1, "in png_do_read_swap_alpha"); + + { + png_uint_32 row_width = row_info->width; + if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA) + { + /* This converts from RGBA to ARGB */ + if (row_info->bit_depth == 8) + { + png_bytep sp = row + row_info->rowbytes; + png_bytep dp = sp; + png_byte save; + png_uint_32 i; + + for (i = 0; i < row_width; i++) + { + save = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = save; + } + } + +#ifdef PNG_READ_16BIT_SUPPORTED + /* This converts from RRGGBBAA to AARRGGBB */ + else + { + png_bytep sp = row + row_info->rowbytes; + png_bytep dp = sp; + png_byte save[2]; + png_uint_32 i; + + for (i = 0; i < row_width; i++) + { + save[0] = *(--sp); + save[1] = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = save[0]; + *(--dp) = save[1]; + } + } +#endif + } + + else if (row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA) + { + /* This converts from GA to AG */ + if (row_info->bit_depth == 8) + { + png_bytep sp = row + row_info->rowbytes; + png_bytep dp = sp; + png_byte save; + png_uint_32 i; + + for (i = 0; i < row_width; i++) + { + save = *(--sp); + *(--dp) = *(--sp); + *(--dp) = save; + } + } + +#ifdef PNG_READ_16BIT_SUPPORTED + /* This converts from GGAA to AAGG */ + else + { + png_bytep sp = row + row_info->rowbytes; + png_bytep dp = sp; + png_byte save[2]; + png_uint_32 i; + + for (i = 0; i < row_width; i++) + { + save[0] = *(--sp); + save[1] = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = save[0]; + *(--dp) = save[1]; + } + } +#endif + } + } +} +#endif + +#ifdef PNG_READ_INVERT_ALPHA_SUPPORTED +static void +png_do_read_invert_alpha(png_row_infop row_info, png_bytep row) +{ + png_uint_32 row_width; + png_debug(1, "in png_do_read_invert_alpha"); + + row_width = row_info->width; + if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA) + { + if (row_info->bit_depth == 8) + { + /* This inverts the alpha channel in RGBA */ + png_bytep sp = row + row_info->rowbytes; + png_bytep dp = sp; + png_uint_32 i; + + for (i = 0; i < row_width; i++) + { + *(--dp) = (png_byte)(255 - *(--sp)); + +/* This does nothing: + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + We can replace it with: +*/ + sp-=3; + dp=sp; + } + } + +#ifdef PNG_READ_16BIT_SUPPORTED + /* This inverts the alpha channel in RRGGBBAA */ + else + { + png_bytep sp = row + row_info->rowbytes; + png_bytep dp = sp; + png_uint_32 i; + + for (i = 0; i < row_width; i++) + { + *(--dp) = (png_byte)(255 - *(--sp)); + *(--dp) = (png_byte)(255 - *(--sp)); + +/* This does nothing: + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + We can replace it with: +*/ + sp-=6; + dp=sp; + } + } +#endif + } + else if (row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA) + { + if (row_info->bit_depth == 8) + { + /* This inverts the alpha channel in GA */ + png_bytep sp = row + row_info->rowbytes; + png_bytep dp = sp; + png_uint_32 i; + + for (i = 0; i < row_width; i++) + { + *(--dp) = (png_byte)(255 - *(--sp)); + *(--dp) = *(--sp); + } + } + +#ifdef PNG_READ_16BIT_SUPPORTED + else + { + /* This inverts the alpha channel in GGAA */ + png_bytep sp = row + row_info->rowbytes; + png_bytep dp = sp; + png_uint_32 i; + + for (i = 0; i < row_width; i++) + { + *(--dp) = (png_byte)(255 - *(--sp)); + *(--dp) = (png_byte)(255 - *(--sp)); +/* + *(--dp) = *(--sp); + *(--dp) = *(--sp); +*/ + sp-=2; + dp=sp; + } + } +#endif + } +} +#endif + +#ifdef PNG_READ_FILLER_SUPPORTED +/* Add filler channel if we have RGB color */ +static void +png_do_read_filler(png_row_infop row_info, png_bytep row, + png_uint_32 filler, png_uint_32 flags) +{ + png_uint_32 i; + png_uint_32 row_width = row_info->width; + +#ifdef PNG_READ_16BIT_SUPPORTED + png_byte hi_filler = (png_byte)((filler>>8) & 0xff); +#endif + png_byte lo_filler = (png_byte)(filler & 0xff); + + png_debug(1, "in png_do_read_filler"); + + if ( + row_info->color_type == PNG_COLOR_TYPE_GRAY) + { + if (row_info->bit_depth == 8) + { + if ((flags & PNG_FLAG_FILLER_AFTER) != 0) + { + /* This changes the data from G to GX */ + png_bytep sp = row + (png_size_t)row_width; + png_bytep dp = sp + (png_size_t)row_width; + for (i = 1; i < row_width; i++) + { + *(--dp) = lo_filler; + *(--dp) = *(--sp); + } + *(--dp) = lo_filler; + row_info->channels = 2; + row_info->pixel_depth = 16; + row_info->rowbytes = row_width * 2; + } + + else + { + /* This changes the data from G to XG */ + png_bytep sp = row + (png_size_t)row_width; + png_bytep dp = sp + (png_size_t)row_width; + for (i = 0; i < row_width; i++) + { + *(--dp) = *(--sp); + *(--dp) = lo_filler; + } + row_info->channels = 2; + row_info->pixel_depth = 16; + row_info->rowbytes = row_width * 2; + } + } + +#ifdef PNG_READ_16BIT_SUPPORTED + else if (row_info->bit_depth == 16) + { + if ((flags & PNG_FLAG_FILLER_AFTER) != 0) + { + /* This changes the data from GG to GGXX */ + png_bytep sp = row + (png_size_t)row_width * 2; + png_bytep dp = sp + (png_size_t)row_width * 2; + for (i = 1; i < row_width; i++) + { + *(--dp) = hi_filler; + *(--dp) = lo_filler; + *(--dp) = *(--sp); + *(--dp) = *(--sp); + } + *(--dp) = hi_filler; + *(--dp) = lo_filler; + row_info->channels = 2; + row_info->pixel_depth = 32; + row_info->rowbytes = row_width * 4; + } + + else + { + /* This changes the data from GG to XXGG */ + png_bytep sp = row + (png_size_t)row_width * 2; + png_bytep dp = sp + (png_size_t)row_width * 2; + for (i = 0; i < row_width; i++) + { + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = hi_filler; + *(--dp) = lo_filler; + } + row_info->channels = 2; + row_info->pixel_depth = 32; + row_info->rowbytes = row_width * 4; + } + } +#endif + } /* COLOR_TYPE == GRAY */ + else if (row_info->color_type == PNG_COLOR_TYPE_RGB) + { + if (row_info->bit_depth == 8) + { + if ((flags & PNG_FLAG_FILLER_AFTER) != 0) + { + /* This changes the data from RGB to RGBX */ + png_bytep sp = row + (png_size_t)row_width * 3; + png_bytep dp = sp + (png_size_t)row_width; + for (i = 1; i < row_width; i++) + { + *(--dp) = lo_filler; + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + } + *(--dp) = lo_filler; + row_info->channels = 4; + row_info->pixel_depth = 32; + row_info->rowbytes = row_width * 4; + } + + else + { + /* This changes the data from RGB to XRGB */ + png_bytep sp = row + (png_size_t)row_width * 3; + png_bytep dp = sp + (png_size_t)row_width; + for (i = 0; i < row_width; i++) + { + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = lo_filler; + } + row_info->channels = 4; + row_info->pixel_depth = 32; + row_info->rowbytes = row_width * 4; + } + } + +#ifdef PNG_READ_16BIT_SUPPORTED + else if (row_info->bit_depth == 16) + { + if ((flags & PNG_FLAG_FILLER_AFTER) != 0) + { + /* This changes the data from RRGGBB to RRGGBBXX */ + png_bytep sp = row + (png_size_t)row_width * 6; + png_bytep dp = sp + (png_size_t)row_width * 2; + for (i = 1; i < row_width; i++) + { + *(--dp) = hi_filler; + *(--dp) = lo_filler; + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + } + *(--dp) = hi_filler; + *(--dp) = lo_filler; + row_info->channels = 4; + row_info->pixel_depth = 64; + row_info->rowbytes = row_width * 8; + } + + else + { + /* This changes the data from RRGGBB to XXRRGGBB */ + png_bytep sp = row + (png_size_t)row_width * 6; + png_bytep dp = sp + (png_size_t)row_width * 2; + for (i = 0; i < row_width; i++) + { + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = *(--sp); + *(--dp) = hi_filler; + *(--dp) = lo_filler; + } + + row_info->channels = 4; + row_info->pixel_depth = 64; + row_info->rowbytes = row_width * 8; + } + } +#endif + } /* COLOR_TYPE == RGB */ +} +#endif + +#ifdef PNG_READ_GRAY_TO_RGB_SUPPORTED +/* Expand grayscale files to RGB, with or without alpha */ +static void +png_do_gray_to_rgb(png_row_infop row_info, png_bytep row) +{ + png_uint_32 i; + png_uint_32 row_width = row_info->width; + + png_debug(1, "in png_do_gray_to_rgb"); + + if (row_info->bit_depth >= 8 && + (row_info->color_type & PNG_COLOR_MASK_COLOR) == 0) + { + if (row_info->color_type == PNG_COLOR_TYPE_GRAY) + { + if (row_info->bit_depth == 8) + { + /* This changes G to RGB */ + png_bytep sp = row + (png_size_t)row_width - 1; + png_bytep dp = sp + (png_size_t)row_width * 2; + for (i = 0; i < row_width; i++) + { + *(dp--) = *sp; + *(dp--) = *sp; + *(dp--) = *(sp--); + } + } + + else + { + /* This changes GG to RRGGBB */ + png_bytep sp = row + (png_size_t)row_width * 2 - 1; + png_bytep dp = sp + (png_size_t)row_width * 4; + for (i = 0; i < row_width; i++) + { + *(dp--) = *sp; + *(dp--) = *(sp - 1); + *(dp--) = *sp; + *(dp--) = *(sp - 1); + *(dp--) = *(sp--); + *(dp--) = *(sp--); + } + } + } + + else if (row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA) + { + if (row_info->bit_depth == 8) + { + /* This changes GA to RGBA */ + png_bytep sp = row + (png_size_t)row_width * 2 - 1; + png_bytep dp = sp + (png_size_t)row_width * 2; + for (i = 0; i < row_width; i++) + { + *(dp--) = *(sp--); + *(dp--) = *sp; + *(dp--) = *sp; + *(dp--) = *(sp--); + } + } + + else + { + /* This changes GGAA to RRGGBBAA */ + png_bytep sp = row + (png_size_t)row_width * 4 - 1; + png_bytep dp = sp + (png_size_t)row_width * 4; + for (i = 0; i < row_width; i++) + { + *(dp--) = *(sp--); + *(dp--) = *(sp--); + *(dp--) = *sp; + *(dp--) = *(sp - 1); + *(dp--) = *sp; + *(dp--) = *(sp - 1); + *(dp--) = *(sp--); + *(dp--) = *(sp--); + } + } + } + row_info->channels = (png_byte)(row_info->channels + 2); + row_info->color_type |= PNG_COLOR_MASK_COLOR; + row_info->pixel_depth = (png_byte)(row_info->channels * + row_info->bit_depth); + row_info->rowbytes = PNG_ROWBYTES(row_info->pixel_depth, row_width); + } +} +#endif + +#ifdef PNG_READ_RGB_TO_GRAY_SUPPORTED +/* Reduce RGB files to grayscale, with or without alpha + * using the equation given in Poynton's ColorFAQ of 1998-01-04 at + * (THIS LINK IS DEAD June 2008 but + * versions dated 1998 through November 2002 have been archived at + * http://web.archive.org/web/20000816232553/http://www.inforamp.net/ + * ~poynton/notes/colour_and_gamma/ColorFAQ.txt ) + * Charles Poynton poynton at poynton.com + * + * Y = 0.212671 * R + 0.715160 * G + 0.072169 * B + * + * which can be expressed with integers as + * + * Y = (6969 * R + 23434 * G + 2365 * B)/32768 + * + * Poynton's current link (as of January 2003 through July 2011): + * + * has changed the numbers slightly: + * + * Y = 0.2126*R + 0.7152*G + 0.0722*B + * + * which can be expressed with integers as + * + * Y = (6966 * R + 23436 * G + 2366 * B)/32768 + * + * Historically, however, libpng uses numbers derived from the ITU-R Rec 709 + * end point chromaticities and the D65 white point. Depending on the + * precision used for the D65 white point this produces a variety of different + * numbers, however if the four decimal place value used in ITU-R Rec 709 is + * used (0.3127,0.3290) the Y calculation would be: + * + * Y = (6968 * R + 23435 * G + 2366 * B)/32768 + * + * While this is correct the rounding results in an overflow for white, because + * the sum of the rounded coefficients is 32769, not 32768. Consequently + * libpng uses, instead, the closest non-overflowing approximation: + * + * Y = (6968 * R + 23434 * G + 2366 * B)/32768 + * + * Starting with libpng-1.5.5, if the image being converted has a cHRM chunk + * (including an sRGB chunk) then the chromaticities are used to calculate the + * coefficients. See the chunk handling in pngrutil.c for more information. + * + * In all cases the calculation is to be done in a linear colorspace. If no + * gamma information is available to correct the encoding of the original RGB + * values this results in an implicit assumption that the original PNG RGB + * values were linear. + * + * Other integer coefficents can be used via png_set_rgb_to_gray(). Because + * the API takes just red and green coefficients the blue coefficient is + * calculated to make the sum 32768. This will result in different rounding + * to that used above. + */ +static int +png_do_rgb_to_gray(png_structrp png_ptr, png_row_infop row_info, png_bytep row) + +{ + int rgb_error = 0; + + png_debug(1, "in png_do_rgb_to_gray"); + + if ((row_info->color_type & PNG_COLOR_MASK_PALETTE) == 0 && + (row_info->color_type & PNG_COLOR_MASK_COLOR) != 0) + { + PNG_CONST png_uint_32 rc = png_ptr->rgb_to_gray_red_coeff; + PNG_CONST png_uint_32 gc = png_ptr->rgb_to_gray_green_coeff; + PNG_CONST png_uint_32 bc = 32768 - rc - gc; + PNG_CONST png_uint_32 row_width = row_info->width; + PNG_CONST int have_alpha = + (row_info->color_type & PNG_COLOR_MASK_ALPHA) != 0; + + if (row_info->bit_depth == 8) + { +#ifdef PNG_READ_GAMMA_SUPPORTED + /* Notice that gamma to/from 1 are not necessarily inverses (if + * there is an overall gamma correction). Prior to 1.5.5 this code + * checked the linearized values for equality; this doesn't match + * the documentation, the original values must be checked. + */ + if (png_ptr->gamma_from_1 != NULL && png_ptr->gamma_to_1 != NULL) + { + png_bytep sp = row; + png_bytep dp = row; + png_uint_32 i; + + for (i = 0; i < row_width; i++) + { + png_byte red = *(sp++); + png_byte green = *(sp++); + png_byte blue = *(sp++); + + if (red != green || red != blue) + { + red = png_ptr->gamma_to_1[red]; + green = png_ptr->gamma_to_1[green]; + blue = png_ptr->gamma_to_1[blue]; + + rgb_error |= 1; + *(dp++) = png_ptr->gamma_from_1[ + (rc*red + gc*green + bc*blue + 16384)>>15]; + } + + else + { + /* If there is no overall correction the table will not be + * set. + */ + if (png_ptr->gamma_table != NULL) + red = png_ptr->gamma_table[red]; + + *(dp++) = red; + } + + if (have_alpha != 0) + *(dp++) = *(sp++); + } + } + else +#endif + { + png_bytep sp = row; + png_bytep dp = row; + png_uint_32 i; + + for (i = 0; i < row_width; i++) + { + png_byte red = *(sp++); + png_byte green = *(sp++); + png_byte blue = *(sp++); + + if (red != green || red != blue) + { + rgb_error |= 1; + /* NOTE: this is the historical approach which simply + * truncates the results. + */ + *(dp++) = (png_byte)((rc*red + gc*green + bc*blue)>>15); + } + + else + *(dp++) = red; + + if (have_alpha != 0) + *(dp++) = *(sp++); + } + } + } + + else /* RGB bit_depth == 16 */ + { +#ifdef PNG_READ_GAMMA_SUPPORTED + if (png_ptr->gamma_16_to_1 != NULL && png_ptr->gamma_16_from_1 != NULL) + { + png_bytep sp = row; + png_bytep dp = row; + png_uint_32 i; + + for (i = 0; i < row_width; i++) + { + png_uint_16 red, green, blue, w; + + red = (png_uint_16)(((*(sp)) << 8) | *(sp + 1)); sp += 2; + green = (png_uint_16)(((*(sp)) << 8) | *(sp + 1)); sp += 2; + blue = (png_uint_16)(((*(sp)) << 8) | *(sp + 1)); sp += 2; + + if (red == green && red == blue) + { + if (png_ptr->gamma_16_table != NULL) + w = png_ptr->gamma_16_table[(red & 0xff) + >> png_ptr->gamma_shift][red >> 8]; + + else + w = red; + } + + else + { + png_uint_16 red_1 = png_ptr->gamma_16_to_1[(red&0xff) + >> png_ptr->gamma_shift][red>>8]; + png_uint_16 green_1 = + png_ptr->gamma_16_to_1[(green&0xff) >> + png_ptr->gamma_shift][green>>8]; + png_uint_16 blue_1 = png_ptr->gamma_16_to_1[(blue&0xff) + >> png_ptr->gamma_shift][blue>>8]; + png_uint_16 gray16 = (png_uint_16)((rc*red_1 + gc*green_1 + + bc*blue_1 + 16384)>>15); + w = png_ptr->gamma_16_from_1[(gray16&0xff) >> + png_ptr->gamma_shift][gray16 >> 8]; + rgb_error |= 1; + } + + *(dp++) = (png_byte)((w>>8) & 0xff); + *(dp++) = (png_byte)(w & 0xff); + + if (have_alpha != 0) + { + *(dp++) = *(sp++); + *(dp++) = *(sp++); + } + } + } + else +#endif + { + png_bytep sp = row; + png_bytep dp = row; + png_uint_32 i; + + for (i = 0; i < row_width; i++) + { + png_uint_16 red, green, blue, gray16; + + red = (png_uint_16)(((*(sp)) << 8) | *(sp + 1)); sp += 2; + green = (png_uint_16)(((*(sp)) << 8) | *(sp + 1)); sp += 2; + blue = (png_uint_16)(((*(sp)) << 8) | *(sp + 1)); sp += 2; + + if (red != green || red != blue) + rgb_error |= 1; + + /* From 1.5.5 in the 16 bit case do the accurate conversion even + * in the 'fast' case - this is because this is where the code + * ends up when handling linear 16 bit data. + */ + gray16 = (png_uint_16)((rc*red + gc*green + bc*blue + 16384) >> + 15); + *(dp++) = (png_byte)((gray16 >> 8) & 0xff); + *(dp++) = (png_byte)(gray16 & 0xff); + + if (have_alpha != 0) + { + *(dp++) = *(sp++); + *(dp++) = *(sp++); + } + } + } + } + + row_info->channels = (png_byte)(row_info->channels - 2); + row_info->color_type = (png_byte)(row_info->color_type & + ~PNG_COLOR_MASK_COLOR); + row_info->pixel_depth = (png_byte)(row_info->channels * + row_info->bit_depth); + row_info->rowbytes = PNG_ROWBYTES(row_info->pixel_depth, row_width); + } + return rgb_error; +} +#endif + +#if defined(PNG_READ_BACKGROUND_SUPPORTED) ||\ + defined(PNG_READ_ALPHA_MODE_SUPPORTED) +/* Replace any alpha or transparency with the supplied background color. + * "background" is already in the screen gamma, while "background_1" is + * at a gamma of 1.0. Paletted files have already been taken care of. + */ +static void +png_do_compose(png_row_infop row_info, png_bytep row, png_structrp png_ptr) +{ +#ifdef PNG_READ_GAMMA_SUPPORTED + png_const_bytep gamma_table = png_ptr->gamma_table; + png_const_bytep gamma_from_1 = png_ptr->gamma_from_1; + png_const_bytep gamma_to_1 = png_ptr->gamma_to_1; + png_const_uint_16pp gamma_16 = png_ptr->gamma_16_table; + png_const_uint_16pp gamma_16_from_1 = png_ptr->gamma_16_from_1; + png_const_uint_16pp gamma_16_to_1 = png_ptr->gamma_16_to_1; + int gamma_shift = png_ptr->gamma_shift; + int optimize = (png_ptr->flags & PNG_FLAG_OPTIMIZE_ALPHA) != 0; +#endif + + png_bytep sp; + png_uint_32 i; + png_uint_32 row_width = row_info->width; + int shift; + + png_debug(1, "in png_do_compose"); + + { + switch (row_info->color_type) + { + case PNG_COLOR_TYPE_GRAY: + { + switch (row_info->bit_depth) + { + case 1: + { + sp = row; + shift = 7; + for (i = 0; i < row_width; i++) + { + if ((png_uint_16)((*sp >> shift) & 0x01) + == png_ptr->trans_color.gray) + { + unsigned int tmp = *sp & (0x7f7f >> (7 - shift)); + tmp |= png_ptr->background.gray << shift; + *sp = (png_byte)(tmp & 0xff); + } + + if (shift == 0) + { + shift = 7; + sp++; + } + + else + shift--; + } + break; + } + + case 2: + { +#ifdef PNG_READ_GAMMA_SUPPORTED + if (gamma_table != NULL) + { + sp = row; + shift = 6; + for (i = 0; i < row_width; i++) + { + if ((png_uint_16)((*sp >> shift) & 0x03) + == png_ptr->trans_color.gray) + { + unsigned int tmp = *sp & (0x3f3f >> (6 - shift)); + tmp |= png_ptr->background.gray << shift; + *sp = (png_byte)(tmp & 0xff); + } + + else + { + unsigned int p = (*sp >> shift) & 0x03; + unsigned int g = (gamma_table [p | (p << 2) | + (p << 4) | (p << 6)] >> 6) & 0x03; + unsigned int tmp = *sp & (0x3f3f >> (6 - shift)); + tmp |= g << shift; + *sp = (png_byte)(tmp & 0xff); + } + + if (shift == 0) + { + shift = 6; + sp++; + } + + else + shift -= 2; + } + } + + else +#endif + { + sp = row; + shift = 6; + for (i = 0; i < row_width; i++) + { + if ((png_uint_16)((*sp >> shift) & 0x03) + == png_ptr->trans_color.gray) + { + unsigned int tmp = *sp & (0x3f3f >> (6 - shift)); + tmp |= png_ptr->background.gray << shift; + *sp = (png_byte)(tmp & 0xff); + } + + if (shift == 0) + { + shift = 6; + sp++; + } + + else + shift -= 2; + } + } + break; + } + + case 4: + { +#ifdef PNG_READ_GAMMA_SUPPORTED + if (gamma_table != NULL) + { + sp = row; + shift = 4; + for (i = 0; i < row_width; i++) + { + if ((png_uint_16)((*sp >> shift) & 0x0f) + == png_ptr->trans_color.gray) + { + unsigned int tmp = *sp & (0xf0f >> (4 - shift)); + tmp |= png_ptr->background.gray << shift; + *sp = (png_byte)(tmp & 0xff); + } + + else + { + unsigned int p = (*sp >> shift) & 0x0f; + unsigned int g = (gamma_table[p | (p << 4)] >> 4) & + 0x0f; + unsigned int tmp = *sp & (0xf0f >> (4 - shift)); + tmp |= g << shift; + *sp = (png_byte)(tmp & 0xff); + } + + if (shift == 0) + { + shift = 4; + sp++; + } + + else + shift -= 4; + } + } + + else +#endif + { + sp = row; + shift = 4; + for (i = 0; i < row_width; i++) + { + if ((png_uint_16)((*sp >> shift) & 0x0f) + == png_ptr->trans_color.gray) + { + unsigned int tmp = *sp & (0xf0f >> (4 - shift)); + tmp |= png_ptr->background.gray << shift; + *sp = (png_byte)(tmp & 0xff); + } + + if (shift == 0) + { + shift = 4; + sp++; + } + + else + shift -= 4; + } + } + break; + } + + case 8: + { +#ifdef PNG_READ_GAMMA_SUPPORTED + if (gamma_table != NULL) + { + sp = row; + for (i = 0; i < row_width; i++, sp++) + { + if (*sp == png_ptr->trans_color.gray) + *sp = (png_byte)png_ptr->background.gray; + + else + *sp = gamma_table[*sp]; + } + } + else +#endif + { + sp = row; + for (i = 0; i < row_width; i++, sp++) + { + if (*sp == png_ptr->trans_color.gray) + *sp = (png_byte)png_ptr->background.gray; + } + } + break; + } + + case 16: + { +#ifdef PNG_READ_GAMMA_SUPPORTED + if (gamma_16 != NULL) + { + sp = row; + for (i = 0; i < row_width; i++, sp += 2) + { + png_uint_16 v; + + v = (png_uint_16)(((*sp) << 8) + *(sp + 1)); + + if (v == png_ptr->trans_color.gray) + { + /* Background is already in screen gamma */ + *sp = (png_byte)((png_ptr->background.gray >> 8) + & 0xff); + *(sp + 1) = (png_byte)(png_ptr->background.gray + & 0xff); + } + + else + { + v = gamma_16[*(sp + 1) >> gamma_shift][*sp]; + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + } + } + } + else +#endif + { + sp = row; + for (i = 0; i < row_width; i++, sp += 2) + { + png_uint_16 v; + + v = (png_uint_16)(((*sp) << 8) + *(sp + 1)); + + if (v == png_ptr->trans_color.gray) + { + *sp = (png_byte)((png_ptr->background.gray >> 8) + & 0xff); + *(sp + 1) = (png_byte)(png_ptr->background.gray + & 0xff); + } + } + } + break; + } + + default: + break; + } + break; + } + + case PNG_COLOR_TYPE_RGB: + { + if (row_info->bit_depth == 8) + { +#ifdef PNG_READ_GAMMA_SUPPORTED + if (gamma_table != NULL) + { + sp = row; + for (i = 0; i < row_width; i++, sp += 3) + { + if (*sp == png_ptr->trans_color.red && + *(sp + 1) == png_ptr->trans_color.green && + *(sp + 2) == png_ptr->trans_color.blue) + { + *sp = (png_byte)png_ptr->background.red; + *(sp + 1) = (png_byte)png_ptr->background.green; + *(sp + 2) = (png_byte)png_ptr->background.blue; + } + + else + { + *sp = gamma_table[*sp]; + *(sp + 1) = gamma_table[*(sp + 1)]; + *(sp + 2) = gamma_table[*(sp + 2)]; + } + } + } + else +#endif + { + sp = row; + for (i = 0; i < row_width; i++, sp += 3) + { + if (*sp == png_ptr->trans_color.red && + *(sp + 1) == png_ptr->trans_color.green && + *(sp + 2) == png_ptr->trans_color.blue) + { + *sp = (png_byte)png_ptr->background.red; + *(sp + 1) = (png_byte)png_ptr->background.green; + *(sp + 2) = (png_byte)png_ptr->background.blue; + } + } + } + } + else /* if (row_info->bit_depth == 16) */ + { +#ifdef PNG_READ_GAMMA_SUPPORTED + if (gamma_16 != NULL) + { + sp = row; + for (i = 0; i < row_width; i++, sp += 6) + { + png_uint_16 r = (png_uint_16)(((*sp) << 8) + *(sp + 1)); + + png_uint_16 g = (png_uint_16)(((*(sp + 2)) << 8) + + *(sp + 3)); + + png_uint_16 b = (png_uint_16)(((*(sp + 4)) << 8) + + *(sp + 5)); + + if (r == png_ptr->trans_color.red && + g == png_ptr->trans_color.green && + b == png_ptr->trans_color.blue) + { + /* Background is already in screen gamma */ + *sp = (png_byte)((png_ptr->background.red >> 8) & 0xff); + *(sp + 1) = (png_byte)(png_ptr->background.red & 0xff); + *(sp + 2) = (png_byte)((png_ptr->background.green >> 8) + & 0xff); + *(sp + 3) = (png_byte)(png_ptr->background.green + & 0xff); + *(sp + 4) = (png_byte)((png_ptr->background.blue >> 8) + & 0xff); + *(sp + 5) = (png_byte)(png_ptr->background.blue & 0xff); + } + + else + { + png_uint_16 v = gamma_16[*(sp + 1) >> gamma_shift][*sp]; + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + + v = gamma_16[*(sp + 3) >> gamma_shift][*(sp + 2)]; + *(sp + 2) = (png_byte)((v >> 8) & 0xff); + *(sp + 3) = (png_byte)(v & 0xff); + + v = gamma_16[*(sp + 5) >> gamma_shift][*(sp + 4)]; + *(sp + 4) = (png_byte)((v >> 8) & 0xff); + *(sp + 5) = (png_byte)(v & 0xff); + } + } + } + + else +#endif + { + sp = row; + for (i = 0; i < row_width; i++, sp += 6) + { + png_uint_16 r = (png_uint_16)(((*sp) << 8) + *(sp + 1)); + + png_uint_16 g = (png_uint_16)(((*(sp + 2)) << 8) + + *(sp + 3)); + + png_uint_16 b = (png_uint_16)(((*(sp + 4)) << 8) + + *(sp + 5)); + + if (r == png_ptr->trans_color.red && + g == png_ptr->trans_color.green && + b == png_ptr->trans_color.blue) + { + *sp = (png_byte)((png_ptr->background.red >> 8) & 0xff); + *(sp + 1) = (png_byte)(png_ptr->background.red & 0xff); + *(sp + 2) = (png_byte)((png_ptr->background.green >> 8) + & 0xff); + *(sp + 3) = (png_byte)(png_ptr->background.green + & 0xff); + *(sp + 4) = (png_byte)((png_ptr->background.blue >> 8) + & 0xff); + *(sp + 5) = (png_byte)(png_ptr->background.blue & 0xff); + } + } + } + } + break; + } + + case PNG_COLOR_TYPE_GRAY_ALPHA: + { + if (row_info->bit_depth == 8) + { +#ifdef PNG_READ_GAMMA_SUPPORTED + if (gamma_to_1 != NULL && gamma_from_1 != NULL && + gamma_table != NULL) + { + sp = row; + for (i = 0; i < row_width; i++, sp += 2) + { + png_uint_16 a = *(sp + 1); + + if (a == 0xff) + *sp = gamma_table[*sp]; + + else if (a == 0) + { + /* Background is already in screen gamma */ + *sp = (png_byte)png_ptr->background.gray; + } + + else + { + png_byte v, w; + + v = gamma_to_1[*sp]; + png_composite(w, v, a, png_ptr->background_1.gray); + if (optimize == 0) + w = gamma_from_1[w]; + *sp = w; + } + } + } + else +#endif + { + sp = row; + for (i = 0; i < row_width; i++, sp += 2) + { + png_byte a = *(sp + 1); + + if (a == 0) + *sp = (png_byte)png_ptr->background.gray; + + else if (a < 0xff) + png_composite(*sp, *sp, a, png_ptr->background.gray); + } + } + } + else /* if (png_ptr->bit_depth == 16) */ + { +#ifdef PNG_READ_GAMMA_SUPPORTED + if (gamma_16 != NULL && gamma_16_from_1 != NULL && + gamma_16_to_1 != NULL) + { + sp = row; + for (i = 0; i < row_width; i++, sp += 4) + { + png_uint_16 a = (png_uint_16)(((*(sp + 2)) << 8) + + *(sp + 3)); + + if (a == (png_uint_16)0xffff) + { + png_uint_16 v; + + v = gamma_16[*(sp + 1) >> gamma_shift][*sp]; + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + } + + else if (a == 0) + { + /* Background is already in screen gamma */ + *sp = (png_byte)((png_ptr->background.gray >> 8) + & 0xff); + *(sp + 1) = (png_byte)(png_ptr->background.gray & 0xff); + } + + else + { + png_uint_16 g, v, w; + + g = gamma_16_to_1[*(sp + 1) >> gamma_shift][*sp]; + png_composite_16(v, g, a, png_ptr->background_1.gray); + if (optimize != 0) + w = v; + else + w = gamma_16_from_1[(v&0xff) >> gamma_shift][v >> 8]; + *sp = (png_byte)((w >> 8) & 0xff); + *(sp + 1) = (png_byte)(w & 0xff); + } + } + } + else +#endif + { + sp = row; + for (i = 0; i < row_width; i++, sp += 4) + { + png_uint_16 a = (png_uint_16)(((*(sp + 2)) << 8) + + *(sp + 3)); + + if (a == 0) + { + *sp = (png_byte)((png_ptr->background.gray >> 8) + & 0xff); + *(sp + 1) = (png_byte)(png_ptr->background.gray & 0xff); + } + + else if (a < 0xffff) + { + png_uint_16 g, v; + + g = (png_uint_16)(((*sp) << 8) + *(sp + 1)); + png_composite_16(v, g, a, png_ptr->background.gray); + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + } + } + } + } + break; + } + + case PNG_COLOR_TYPE_RGB_ALPHA: + { + if (row_info->bit_depth == 8) + { +#ifdef PNG_READ_GAMMA_SUPPORTED + if (gamma_to_1 != NULL && gamma_from_1 != NULL && + gamma_table != NULL) + { + sp = row; + for (i = 0; i < row_width; i++, sp += 4) + { + png_byte a = *(sp + 3); + + if (a == 0xff) + { + *sp = gamma_table[*sp]; + *(sp + 1) = gamma_table[*(sp + 1)]; + *(sp + 2) = gamma_table[*(sp + 2)]; + } + + else if (a == 0) + { + /* Background is already in screen gamma */ + *sp = (png_byte)png_ptr->background.red; + *(sp + 1) = (png_byte)png_ptr->background.green; + *(sp + 2) = (png_byte)png_ptr->background.blue; + } + + else + { + png_byte v, w; + + v = gamma_to_1[*sp]; + png_composite(w, v, a, png_ptr->background_1.red); + if (optimize == 0) w = gamma_from_1[w]; + *sp = w; + + v = gamma_to_1[*(sp + 1)]; + png_composite(w, v, a, png_ptr->background_1.green); + if (optimize == 0) w = gamma_from_1[w]; + *(sp + 1) = w; + + v = gamma_to_1[*(sp + 2)]; + png_composite(w, v, a, png_ptr->background_1.blue); + if (optimize == 0) w = gamma_from_1[w]; + *(sp + 2) = w; + } + } + } + else +#endif + { + sp = row; + for (i = 0; i < row_width; i++, sp += 4) + { + png_byte a = *(sp + 3); + + if (a == 0) + { + *sp = (png_byte)png_ptr->background.red; + *(sp + 1) = (png_byte)png_ptr->background.green; + *(sp + 2) = (png_byte)png_ptr->background.blue; + } + + else if (a < 0xff) + { + png_composite(*sp, *sp, a, png_ptr->background.red); + + png_composite(*(sp + 1), *(sp + 1), a, + png_ptr->background.green); + + png_composite(*(sp + 2), *(sp + 2), a, + png_ptr->background.blue); + } + } + } + } + else /* if (row_info->bit_depth == 16) */ + { +#ifdef PNG_READ_GAMMA_SUPPORTED + if (gamma_16 != NULL && gamma_16_from_1 != NULL && + gamma_16_to_1 != NULL) + { + sp = row; + for (i = 0; i < row_width; i++, sp += 8) + { + png_uint_16 a = (png_uint_16)(((png_uint_16)(*(sp + 6)) + << 8) + (png_uint_16)(*(sp + 7))); + + if (a == (png_uint_16)0xffff) + { + png_uint_16 v; + + v = gamma_16[*(sp + 1) >> gamma_shift][*sp]; + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + + v = gamma_16[*(sp + 3) >> gamma_shift][*(sp + 2)]; + *(sp + 2) = (png_byte)((v >> 8) & 0xff); + *(sp + 3) = (png_byte)(v & 0xff); + + v = gamma_16[*(sp + 5) >> gamma_shift][*(sp + 4)]; + *(sp + 4) = (png_byte)((v >> 8) & 0xff); + *(sp + 5) = (png_byte)(v & 0xff); + } + + else if (a == 0) + { + /* Background is already in screen gamma */ + *sp = (png_byte)((png_ptr->background.red >> 8) & 0xff); + *(sp + 1) = (png_byte)(png_ptr->background.red & 0xff); + *(sp + 2) = (png_byte)((png_ptr->background.green >> 8) + & 0xff); + *(sp + 3) = (png_byte)(png_ptr->background.green + & 0xff); + *(sp + 4) = (png_byte)((png_ptr->background.blue >> 8) + & 0xff); + *(sp + 5) = (png_byte)(png_ptr->background.blue & 0xff); + } + + else + { + png_uint_16 v, w; + + v = gamma_16_to_1[*(sp + 1) >> gamma_shift][*sp]; + png_composite_16(w, v, a, png_ptr->background_1.red); + if (optimize == 0) + w = gamma_16_from_1[((w&0xff) >> gamma_shift)][w >> + 8]; + *sp = (png_byte)((w >> 8) & 0xff); + *(sp + 1) = (png_byte)(w & 0xff); + + v = gamma_16_to_1[*(sp + 3) >> gamma_shift][*(sp + 2)]; + png_composite_16(w, v, a, png_ptr->background_1.green); + if (optimize == 0) + w = gamma_16_from_1[((w&0xff) >> gamma_shift)][w >> + 8]; + + *(sp + 2) = (png_byte)((w >> 8) & 0xff); + *(sp + 3) = (png_byte)(w & 0xff); + + v = gamma_16_to_1[*(sp + 5) >> gamma_shift][*(sp + 4)]; + png_composite_16(w, v, a, png_ptr->background_1.blue); + if (optimize == 0) + w = gamma_16_from_1[((w&0xff) >> gamma_shift)][w >> + 8]; + + *(sp + 4) = (png_byte)((w >> 8) & 0xff); + *(sp + 5) = (png_byte)(w & 0xff); + } + } + } + + else +#endif + { + sp = row; + for (i = 0; i < row_width; i++, sp += 8) + { + png_uint_16 a = (png_uint_16)(((png_uint_16)(*(sp + 6)) + << 8) + (png_uint_16)(*(sp + 7))); + + if (a == 0) + { + *sp = (png_byte)((png_ptr->background.red >> 8) & 0xff); + *(sp + 1) = (png_byte)(png_ptr->background.red & 0xff); + *(sp + 2) = (png_byte)((png_ptr->background.green >> 8) + & 0xff); + *(sp + 3) = (png_byte)(png_ptr->background.green + & 0xff); + *(sp + 4) = (png_byte)((png_ptr->background.blue >> 8) + & 0xff); + *(sp + 5) = (png_byte)(png_ptr->background.blue & 0xff); + } + + else if (a < 0xffff) + { + png_uint_16 v; + + png_uint_16 r = (png_uint_16)(((*sp) << 8) + *(sp + 1)); + png_uint_16 g = (png_uint_16)(((*(sp + 2)) << 8) + + *(sp + 3)); + png_uint_16 b = (png_uint_16)(((*(sp + 4)) << 8) + + *(sp + 5)); + + png_composite_16(v, r, a, png_ptr->background.red); + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + + png_composite_16(v, g, a, png_ptr->background.green); + *(sp + 2) = (png_byte)((v >> 8) & 0xff); + *(sp + 3) = (png_byte)(v & 0xff); + + png_composite_16(v, b, a, png_ptr->background.blue); + *(sp + 4) = (png_byte)((v >> 8) & 0xff); + *(sp + 5) = (png_byte)(v & 0xff); + } + } + } + } + break; + } + + default: + break; + } + } +} +#endif /* READ_BACKGROUND || READ_ALPHA_MODE */ + +#ifdef PNG_READ_GAMMA_SUPPORTED +/* Gamma correct the image, avoiding the alpha channel. Make sure + * you do this after you deal with the transparency issue on grayscale + * or RGB images. If your bit depth is 8, use gamma_table, if it + * is 16, use gamma_16_table and gamma_shift. Build these with + * build_gamma_table(). + */ +static void +png_do_gamma(png_row_infop row_info, png_bytep row, png_structrp png_ptr) +{ + png_const_bytep gamma_table = png_ptr->gamma_table; + png_const_uint_16pp gamma_16_table = png_ptr->gamma_16_table; + int gamma_shift = png_ptr->gamma_shift; + + png_bytep sp; + png_uint_32 i; + png_uint_32 row_width=row_info->width; + + png_debug(1, "in png_do_gamma"); + + if (((row_info->bit_depth <= 8 && gamma_table != NULL) || + (row_info->bit_depth == 16 && gamma_16_table != NULL))) + { + switch (row_info->color_type) + { + case PNG_COLOR_TYPE_RGB: + { + if (row_info->bit_depth == 8) + { + sp = row; + for (i = 0; i < row_width; i++) + { + *sp = gamma_table[*sp]; + sp++; + *sp = gamma_table[*sp]; + sp++; + *sp = gamma_table[*sp]; + sp++; + } + } + + else /* if (row_info->bit_depth == 16) */ + { + sp = row; + for (i = 0; i < row_width; i++) + { + png_uint_16 v; + + v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp]; + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + sp += 2; + + v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp]; + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + sp += 2; + + v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp]; + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + sp += 2; + } + } + break; + } + + case PNG_COLOR_TYPE_RGB_ALPHA: + { + if (row_info->bit_depth == 8) + { + sp = row; + for (i = 0; i < row_width; i++) + { + *sp = gamma_table[*sp]; + sp++; + + *sp = gamma_table[*sp]; + sp++; + + *sp = gamma_table[*sp]; + sp++; + + sp++; + } + } + + else /* if (row_info->bit_depth == 16) */ + { + sp = row; + for (i = 0; i < row_width; i++) + { + png_uint_16 v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp]; + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + sp += 2; + + v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp]; + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + sp += 2; + + v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp]; + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + sp += 4; + } + } + break; + } + + case PNG_COLOR_TYPE_GRAY_ALPHA: + { + if (row_info->bit_depth == 8) + { + sp = row; + for (i = 0; i < row_width; i++) + { + *sp = gamma_table[*sp]; + sp += 2; + } + } + + else /* if (row_info->bit_depth == 16) */ + { + sp = row; + for (i = 0; i < row_width; i++) + { + png_uint_16 v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp]; + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + sp += 4; + } + } + break; + } + + case PNG_COLOR_TYPE_GRAY: + { + if (row_info->bit_depth == 2) + { + sp = row; + for (i = 0; i < row_width; i += 4) + { + int a = *sp & 0xc0; + int b = *sp & 0x30; + int c = *sp & 0x0c; + int d = *sp & 0x03; + + *sp = (png_byte)( + ((((int)gamma_table[a|(a>>2)|(a>>4)|(a>>6)]) ) & 0xc0)| + ((((int)gamma_table[(b<<2)|b|(b>>2)|(b>>4)])>>2) & 0x30)| + ((((int)gamma_table[(c<<4)|(c<<2)|c|(c>>2)])>>4) & 0x0c)| + ((((int)gamma_table[(d<<6)|(d<<4)|(d<<2)|d])>>6) )); + sp++; + } + } + + if (row_info->bit_depth == 4) + { + sp = row; + for (i = 0; i < row_width; i += 2) + { + int msb = *sp & 0xf0; + int lsb = *sp & 0x0f; + + *sp = (png_byte)((((int)gamma_table[msb | (msb >> 4)]) & 0xf0) + | (((int)gamma_table[(lsb << 4) | lsb]) >> 4)); + sp++; + } + } + + else if (row_info->bit_depth == 8) + { + sp = row; + for (i = 0; i < row_width; i++) + { + *sp = gamma_table[*sp]; + sp++; + } + } + + else if (row_info->bit_depth == 16) + { + sp = row; + for (i = 0; i < row_width; i++) + { + png_uint_16 v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp]; + *sp = (png_byte)((v >> 8) & 0xff); + *(sp + 1) = (png_byte)(v & 0xff); + sp += 2; + } + } + break; + } + + default: + break; + } + } +} +#endif + +#ifdef PNG_READ_ALPHA_MODE_SUPPORTED +/* Encode the alpha channel to the output gamma (the input channel is always + * linear.) Called only with color types that have an alpha channel. Needs the + * from_1 tables. + */ +static void +png_do_encode_alpha(png_row_infop row_info, png_bytep row, png_structrp png_ptr) +{ + png_uint_32 row_width = row_info->width; + + png_debug(1, "in png_do_encode_alpha"); + + if ((row_info->color_type & PNG_COLOR_MASK_ALPHA) != 0) + { + if (row_info->bit_depth == 8) + { + PNG_CONST png_bytep table = png_ptr->gamma_from_1; + + if (table != NULL) + { + PNG_CONST int step = + (row_info->color_type & PNG_COLOR_MASK_COLOR) ? 4 : 2; + + /* The alpha channel is the last component: */ + row += step - 1; + + for (; row_width > 0; --row_width, row += step) + *row = table[*row]; + + return; + } + } + + else if (row_info->bit_depth == 16) + { + PNG_CONST png_uint_16pp table = png_ptr->gamma_16_from_1; + PNG_CONST int gamma_shift = png_ptr->gamma_shift; + + if (table != NULL) + { + PNG_CONST int step = + (row_info->color_type & PNG_COLOR_MASK_COLOR) ? 8 : 4; + + /* The alpha channel is the last component: */ + row += step - 2; + + for (; row_width > 0; --row_width, row += step) + { + png_uint_16 v; + + v = table[*(row + 1) >> gamma_shift][*row]; + *row = (png_byte)((v >> 8) & 0xff); + *(row + 1) = (png_byte)(v & 0xff); + } + + return; + } + } + } + + /* Only get to here if called with a weird row_info; no harm has been done, + * so just issue a warning. + */ + png_warning(png_ptr, "png_do_encode_alpha: unexpected call"); +} +#endif + +#ifdef PNG_READ_EXPAND_SUPPORTED +/* Expands a palette row to an RGB or RGBA row depending + * upon whether you supply trans and num_trans. + */ +static void +png_do_expand_palette(png_row_infop row_info, png_bytep row, + png_const_colorp palette, png_const_bytep trans_alpha, int num_trans) +{ + int shift, value; + png_bytep sp, dp; + png_uint_32 i; + png_uint_32 row_width=row_info->width; + + png_debug(1, "in png_do_expand_palette"); + + if (row_info->color_type == PNG_COLOR_TYPE_PALETTE) + { + if (row_info->bit_depth < 8) + { + switch (row_info->bit_depth) + { + case 1: + { + sp = row + (png_size_t)((row_width - 1) >> 3); + dp = row + (png_size_t)row_width - 1; + shift = 7 - (int)((row_width + 7) & 0x07); + for (i = 0; i < row_width; i++) + { + if ((*sp >> shift) & 0x01) + *dp = 1; + + else + *dp = 0; + + if (shift == 7) + { + shift = 0; + sp--; + } + + else + shift++; + + dp--; + } + break; + } + + case 2: + { + sp = row + (png_size_t)((row_width - 1) >> 2); + dp = row + (png_size_t)row_width - 1; + shift = (int)((3 - ((row_width + 3) & 0x03)) << 1); + for (i = 0; i < row_width; i++) + { + value = (*sp >> shift) & 0x03; + *dp = (png_byte)value; + if (shift == 6) + { + shift = 0; + sp--; + } + + else + shift += 2; + + dp--; + } + break; + } + + case 4: + { + sp = row + (png_size_t)((row_width - 1) >> 1); + dp = row + (png_size_t)row_width - 1; + shift = (int)((row_width & 0x01) << 2); + for (i = 0; i < row_width; i++) + { + value = (*sp >> shift) & 0x0f; + *dp = (png_byte)value; + if (shift == 4) + { + shift = 0; + sp--; + } + + else + shift += 4; + + dp--; + } + break; + } + + default: + break; + } + row_info->bit_depth = 8; + row_info->pixel_depth = 8; + row_info->rowbytes = row_width; + } + + if (row_info->bit_depth == 8) + { + { + if (num_trans > 0) + { + sp = row + (png_size_t)row_width - 1; + dp = row + (png_size_t)(row_width << 2) - 1; + + for (i = 0; i < row_width; i++) + { + if ((int)(*sp) >= num_trans) + *dp-- = 0xff; + + else + *dp-- = trans_alpha[*sp]; + + *dp-- = palette[*sp].blue; + *dp-- = palette[*sp].green; + *dp-- = palette[*sp].red; + sp--; + } + row_info->bit_depth = 8; + row_info->pixel_depth = 32; + row_info->rowbytes = row_width * 4; + row_info->color_type = 6; + row_info->channels = 4; + } + + else + { + sp = row + (png_size_t)row_width - 1; + dp = row + (png_size_t)(row_width * 3) - 1; + + for (i = 0; i < row_width; i++) + { + *dp-- = palette[*sp].blue; + *dp-- = palette[*sp].green; + *dp-- = palette[*sp].red; + sp--; + } + + row_info->bit_depth = 8; + row_info->pixel_depth = 24; + row_info->rowbytes = row_width * 3; + row_info->color_type = 2; + row_info->channels = 3; + } + } + } + } +} + +/* If the bit depth < 8, it is expanded to 8. Also, if the already + * expanded transparency value is supplied, an alpha channel is built. + */ +static void +png_do_expand(png_row_infop row_info, png_bytep row, + png_const_color_16p trans_color) +{ + int shift, value; + png_bytep sp, dp; + png_uint_32 i; + png_uint_32 row_width=row_info->width; + + png_debug(1, "in png_do_expand"); + + { + if (row_info->color_type == PNG_COLOR_TYPE_GRAY) + { + unsigned int gray = trans_color != NULL ? trans_color->gray : 0; + + if (row_info->bit_depth < 8) + { + switch (row_info->bit_depth) + { + case 1: + { + gray = (gray & 0x01) * 0xff; + sp = row + (png_size_t)((row_width - 1) >> 3); + dp = row + (png_size_t)row_width - 1; + shift = 7 - (int)((row_width + 7) & 0x07); + for (i = 0; i < row_width; i++) + { + if ((*sp >> shift) & 0x01) + *dp = 0xff; + + else + *dp = 0; + + if (shift == 7) + { + shift = 0; + sp--; + } + + else + shift++; + + dp--; + } + break; + } + + case 2: + { + gray = (gray & 0x03) * 0x55; + sp = row + (png_size_t)((row_width - 1) >> 2); + dp = row + (png_size_t)row_width - 1; + shift = (int)((3 - ((row_width + 3) & 0x03)) << 1); + for (i = 0; i < row_width; i++) + { + value = (*sp >> shift) & 0x03; + *dp = (png_byte)(value | (value << 2) | (value << 4) | + (value << 6)); + if (shift == 6) + { + shift = 0; + sp--; + } + + else + shift += 2; + + dp--; + } + break; + } + + case 4: + { + gray = (gray & 0x0f) * 0x11; + sp = row + (png_size_t)((row_width - 1) >> 1); + dp = row + (png_size_t)row_width - 1; + shift = (int)((1 - ((row_width + 1) & 0x01)) << 2); + for (i = 0; i < row_width; i++) + { + value = (*sp >> shift) & 0x0f; + *dp = (png_byte)(value | (value << 4)); + if (shift == 4) + { + shift = 0; + sp--; + } + + else + shift = 4; + + dp--; + } + break; + } + + default: + break; + } + + row_info->bit_depth = 8; + row_info->pixel_depth = 8; + row_info->rowbytes = row_width; + } + + if (trans_color != NULL) + { + if (row_info->bit_depth == 8) + { + gray = gray & 0xff; + sp = row + (png_size_t)row_width - 1; + dp = row + (png_size_t)(row_width << 1) - 1; + + for (i = 0; i < row_width; i++) + { + if (*sp == gray) + *dp-- = 0; + + else + *dp-- = 0xff; + + *dp-- = *sp--; + } + } + + else if (row_info->bit_depth == 16) + { + unsigned int gray_high = (gray >> 8) & 0xff; + unsigned int gray_low = gray & 0xff; + sp = row + row_info->rowbytes - 1; + dp = row + (row_info->rowbytes << 1) - 1; + for (i = 0; i < row_width; i++) + { + if (*(sp - 1) == gray_high && *(sp) == gray_low) + { + *dp-- = 0; + *dp-- = 0; + } + + else + { + *dp-- = 0xff; + *dp-- = 0xff; + } + + *dp-- = *sp--; + *dp-- = *sp--; + } + } + + row_info->color_type = PNG_COLOR_TYPE_GRAY_ALPHA; + row_info->channels = 2; + row_info->pixel_depth = (png_byte)(row_info->bit_depth << 1); + row_info->rowbytes = PNG_ROWBYTES(row_info->pixel_depth, + row_width); + } + } + else if (row_info->color_type == PNG_COLOR_TYPE_RGB && + trans_color != NULL) + { + if (row_info->bit_depth == 8) + { + png_byte red = (png_byte)(trans_color->red & 0xff); + png_byte green = (png_byte)(trans_color->green & 0xff); + png_byte blue = (png_byte)(trans_color->blue & 0xff); + sp = row + (png_size_t)row_info->rowbytes - 1; + dp = row + (png_size_t)(row_width << 2) - 1; + for (i = 0; i < row_width; i++) + { + if (*(sp - 2) == red && *(sp - 1) == green && *(sp) == blue) + *dp-- = 0; + + else + *dp-- = 0xff; + + *dp-- = *sp--; + *dp-- = *sp--; + *dp-- = *sp--; + } + } + else if (row_info->bit_depth == 16) + { + png_byte red_high = (png_byte)((trans_color->red >> 8) & 0xff); + png_byte green_high = (png_byte)((trans_color->green >> 8) & 0xff); + png_byte blue_high = (png_byte)((trans_color->blue >> 8) & 0xff); + png_byte red_low = (png_byte)(trans_color->red & 0xff); + png_byte green_low = (png_byte)(trans_color->green & 0xff); + png_byte blue_low = (png_byte)(trans_color->blue & 0xff); + sp = row + row_info->rowbytes - 1; + dp = row + (png_size_t)(row_width << 3) - 1; + for (i = 0; i < row_width; i++) + { + if (*(sp - 5) == red_high && + *(sp - 4) == red_low && + *(sp - 3) == green_high && + *(sp - 2) == green_low && + *(sp - 1) == blue_high && + *(sp ) == blue_low) + { + *dp-- = 0; + *dp-- = 0; + } + + else + { + *dp-- = 0xff; + *dp-- = 0xff; + } + + *dp-- = *sp--; + *dp-- = *sp--; + *dp-- = *sp--; + *dp-- = *sp--; + *dp-- = *sp--; + *dp-- = *sp--; + } + } + row_info->color_type = PNG_COLOR_TYPE_RGB_ALPHA; + row_info->channels = 4; + row_info->pixel_depth = (png_byte)(row_info->bit_depth << 2); + row_info->rowbytes = PNG_ROWBYTES(row_info->pixel_depth, row_width); + } + } +} +#endif + +#ifdef PNG_READ_EXPAND_16_SUPPORTED +/* If the bit depth is 8 and the color type is not a palette type expand the + * whole row to 16 bits. Has no effect otherwise. + */ +static void +png_do_expand_16(png_row_infop row_info, png_bytep row) +{ + if (row_info->bit_depth == 8 && + row_info->color_type != PNG_COLOR_TYPE_PALETTE) + { + /* The row have a sequence of bytes containing [0..255] and we need + * to turn it into another row containing [0..65535], to do this we + * calculate: + * + * (input / 255) * 65535 + * + * Which happens to be exactly input * 257 and this can be achieved + * simply by byte replication in place (copying backwards). + */ + png_byte *sp = row + row_info->rowbytes; /* source, last byte + 1 */ + png_byte *dp = sp + row_info->rowbytes; /* destination, end + 1 */ + while (dp > sp) + dp[-2] = dp[-1] = *--sp, dp -= 2; + + row_info->rowbytes *= 2; + row_info->bit_depth = 16; + row_info->pixel_depth = (png_byte)(row_info->channels * 16); + } +} +#endif + +#ifdef PNG_READ_QUANTIZE_SUPPORTED +static void +png_do_quantize(png_row_infop row_info, png_bytep row, + png_const_bytep palette_lookup, png_const_bytep quantize_lookup) +{ + png_bytep sp, dp; + png_uint_32 i; + png_uint_32 row_width=row_info->width; + + png_debug(1, "in png_do_quantize"); + + if (row_info->bit_depth == 8) + { + if (row_info->color_type == PNG_COLOR_TYPE_RGB && palette_lookup) + { + int r, g, b, p; + sp = row; + dp = row; + for (i = 0; i < row_width; i++) + { + r = *sp++; + g = *sp++; + b = *sp++; + + /* This looks real messy, but the compiler will reduce + * it down to a reasonable formula. For example, with + * 5 bits per color, we get: + * p = (((r >> 3) & 0x1f) << 10) | + * (((g >> 3) & 0x1f) << 5) | + * ((b >> 3) & 0x1f); + */ + p = (((r >> (8 - PNG_QUANTIZE_RED_BITS)) & + ((1 << PNG_QUANTIZE_RED_BITS) - 1)) << + (PNG_QUANTIZE_GREEN_BITS + PNG_QUANTIZE_BLUE_BITS)) | + (((g >> (8 - PNG_QUANTIZE_GREEN_BITS)) & + ((1 << PNG_QUANTIZE_GREEN_BITS) - 1)) << + (PNG_QUANTIZE_BLUE_BITS)) | + ((b >> (8 - PNG_QUANTIZE_BLUE_BITS)) & + ((1 << PNG_QUANTIZE_BLUE_BITS) - 1)); + + *dp++ = palette_lookup[p]; + } + + row_info->color_type = PNG_COLOR_TYPE_PALETTE; + row_info->channels = 1; + row_info->pixel_depth = row_info->bit_depth; + row_info->rowbytes = PNG_ROWBYTES(row_info->pixel_depth, row_width); + } + + else if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA && + palette_lookup != NULL) + { + int r, g, b, p; + sp = row; + dp = row; + for (i = 0; i < row_width; i++) + { + r = *sp++; + g = *sp++; + b = *sp++; + sp++; + + p = (((r >> (8 - PNG_QUANTIZE_RED_BITS)) & + ((1 << PNG_QUANTIZE_RED_BITS) - 1)) << + (PNG_QUANTIZE_GREEN_BITS + PNG_QUANTIZE_BLUE_BITS)) | + (((g >> (8 - PNG_QUANTIZE_GREEN_BITS)) & + ((1 << PNG_QUANTIZE_GREEN_BITS) - 1)) << + (PNG_QUANTIZE_BLUE_BITS)) | + ((b >> (8 - PNG_QUANTIZE_BLUE_BITS)) & + ((1 << PNG_QUANTIZE_BLUE_BITS) - 1)); + + *dp++ = palette_lookup[p]; + } + + row_info->color_type = PNG_COLOR_TYPE_PALETTE; + row_info->channels = 1; + row_info->pixel_depth = row_info->bit_depth; + row_info->rowbytes = PNG_ROWBYTES(row_info->pixel_depth, row_width); + } + + else if (row_info->color_type == PNG_COLOR_TYPE_PALETTE && + quantize_lookup) + { + sp = row; + + for (i = 0; i < row_width; i++, sp++) + { + *sp = quantize_lookup[*sp]; + } + } + } +} +#endif /* READ_QUANTIZE */ + +/* Transform the row. The order of transformations is significant, + * and is very touchy. If you add a transformation, take care to + * decide how it fits in with the other transformations here. + */ +void /* PRIVATE */ +png_do_read_transformations(png_structrp png_ptr, png_row_infop row_info) +{ + png_debug(1, "in png_do_read_transformations"); + + if (png_ptr->row_buf == NULL) + { + /* Prior to 1.5.4 this output row/pass where the NULL pointer is, but this + * error is incredibly rare and incredibly easy to debug without this + * information. + */ + png_error(png_ptr, "NULL row buffer"); + } + + /* The following is debugging; prior to 1.5.4 the code was never compiled in; + * in 1.5.4 PNG_FLAG_DETECT_UNINITIALIZED was added and the macro + * PNG_WARN_UNINITIALIZED_ROW removed. In 1.6 the new flag is set only for + * all transformations, however in practice the ROW_INIT always gets done on + * demand, if necessary. + */ + if ((png_ptr->flags & PNG_FLAG_DETECT_UNINITIALIZED) != 0 && + (png_ptr->flags & PNG_FLAG_ROW_INIT) == 0) + { + /* Application has failed to call either png_read_start_image() or + * png_read_update_info() after setting transforms that expand pixels. + * This check added to libpng-1.2.19 (but not enabled until 1.5.4). + */ + png_error(png_ptr, "Uninitialized row"); + } + +#ifdef PNG_READ_EXPAND_SUPPORTED + if ((png_ptr->transformations & PNG_EXPAND) != 0) + { + if (row_info->color_type == PNG_COLOR_TYPE_PALETTE) + { + png_do_expand_palette(row_info, png_ptr->row_buf + 1, + png_ptr->palette, png_ptr->trans_alpha, png_ptr->num_trans); + } + + else + { + if (png_ptr->num_trans != 0 && + (png_ptr->transformations & PNG_EXPAND_tRNS) != 0) + png_do_expand(row_info, png_ptr->row_buf + 1, + &(png_ptr->trans_color)); + + else + png_do_expand(row_info, png_ptr->row_buf + 1, + NULL); + } + } +#endif + +#ifdef PNG_READ_STRIP_ALPHA_SUPPORTED + if ((png_ptr->transformations & PNG_STRIP_ALPHA) != 0 && + (png_ptr->transformations & PNG_COMPOSE) == 0 && + (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA || + row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA)) + png_do_strip_channel(row_info, png_ptr->row_buf + 1, + 0 /* at_start == false, because SWAP_ALPHA happens later */); +#endif + +#ifdef PNG_READ_RGB_TO_GRAY_SUPPORTED + if ((png_ptr->transformations & PNG_RGB_TO_GRAY) != 0) + { + int rgb_error = + png_do_rgb_to_gray(png_ptr, row_info, + png_ptr->row_buf + 1); + + if (rgb_error != 0) + { + png_ptr->rgb_to_gray_status=1; + if ((png_ptr->transformations & PNG_RGB_TO_GRAY) == + PNG_RGB_TO_GRAY_WARN) + png_warning(png_ptr, "png_do_rgb_to_gray found nongray pixel"); + + if ((png_ptr->transformations & PNG_RGB_TO_GRAY) == + PNG_RGB_TO_GRAY_ERR) + png_error(png_ptr, "png_do_rgb_to_gray found nongray pixel"); + } + } +#endif + +/* From Andreas Dilger e-mail to png-implement, 26 March 1998: + * + * In most cases, the "simple transparency" should be done prior to doing + * gray-to-RGB, or you will have to test 3x as many bytes to check if a + * pixel is transparent. You would also need to make sure that the + * transparency information is upgraded to RGB. + * + * To summarize, the current flow is: + * - Gray + simple transparency -> compare 1 or 2 gray bytes and composite + * with background "in place" if transparent, + * convert to RGB if necessary + * - Gray + alpha -> composite with gray background and remove alpha bytes, + * convert to RGB if necessary + * + * To support RGB backgrounds for gray images we need: + * - Gray + simple transparency -> convert to RGB + simple transparency, + * compare 3 or 6 bytes and composite with + * background "in place" if transparent + * (3x compare/pixel compared to doing + * composite with gray bkgrnd) + * - Gray + alpha -> convert to RGB + alpha, composite with background and + * remove alpha bytes (3x float + * operations/pixel compared with composite + * on gray background) + * + * Greg's change will do this. The reason it wasn't done before is for + * performance, as this increases the per-pixel operations. If we would check + * in advance if the background was gray or RGB, and position the gray-to-RGB + * transform appropriately, then it would save a lot of work/time. + */ + +#ifdef PNG_READ_GRAY_TO_RGB_SUPPORTED + /* If gray -> RGB, do so now only if background is non-gray; else do later + * for performance reasons + */ + if ((png_ptr->transformations & PNG_GRAY_TO_RGB) != 0 && + (png_ptr->mode & PNG_BACKGROUND_IS_GRAY) == 0) + png_do_gray_to_rgb(row_info, png_ptr->row_buf + 1); +#endif + +#if defined(PNG_READ_BACKGROUND_SUPPORTED) ||\ + defined(PNG_READ_ALPHA_MODE_SUPPORTED) + if ((png_ptr->transformations & PNG_COMPOSE) != 0) + png_do_compose(row_info, png_ptr->row_buf + 1, png_ptr); +#endif + +#ifdef PNG_READ_GAMMA_SUPPORTED + if ((png_ptr->transformations & PNG_GAMMA) != 0 && +#ifdef PNG_READ_RGB_TO_GRAY_SUPPORTED + /* Because RGB_TO_GRAY does the gamma transform. */ + (png_ptr->transformations & PNG_RGB_TO_GRAY) == 0 && +#endif +#if defined(PNG_READ_BACKGROUND_SUPPORTED) ||\ + defined(PNG_READ_ALPHA_MODE_SUPPORTED) + /* Because PNG_COMPOSE does the gamma transform if there is something to + * do (if there is an alpha channel or transparency.) + */ + !((png_ptr->transformations & PNG_COMPOSE) && + ((png_ptr->num_trans != 0) || + (png_ptr->color_type & PNG_COLOR_MASK_ALPHA) != 0)) && +#endif + /* Because png_init_read_transformations transforms the palette, unless + * RGB_TO_GRAY will do the transform. + */ + (png_ptr->color_type != PNG_COLOR_TYPE_PALETTE)) + png_do_gamma(row_info, png_ptr->row_buf + 1, png_ptr); +#endif + +#ifdef PNG_READ_STRIP_ALPHA_SUPPORTED + if ((png_ptr->transformations & PNG_STRIP_ALPHA) != 0 && + (png_ptr->transformations & PNG_COMPOSE) != 0 && + (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA || + row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA)) + png_do_strip_channel(row_info, png_ptr->row_buf + 1, + 0 /* at_start == false, because SWAP_ALPHA happens later */); +#endif + +#ifdef PNG_READ_ALPHA_MODE_SUPPORTED + if ((png_ptr->transformations & PNG_ENCODE_ALPHA) != 0 && + (row_info->color_type & PNG_COLOR_MASK_ALPHA) != 0) + png_do_encode_alpha(row_info, png_ptr->row_buf + 1, png_ptr); +#endif + +#ifdef PNG_READ_SCALE_16_TO_8_SUPPORTED + if ((png_ptr->transformations & PNG_SCALE_16_TO_8) != 0) + png_do_scale_16_to_8(row_info, png_ptr->row_buf + 1); +#endif + +#ifdef PNG_READ_STRIP_16_TO_8_SUPPORTED + /* There is no harm in doing both of these because only one has any effect, + * by putting the 'scale' option first if the app asks for scale (either by + * calling the API or in a TRANSFORM flag) this is what happens. + */ + if ((png_ptr->transformations & PNG_16_TO_8) != 0) + png_do_chop(row_info, png_ptr->row_buf + 1); +#endif + +#ifdef PNG_READ_QUANTIZE_SUPPORTED + if ((png_ptr->transformations & PNG_QUANTIZE) != 0) + { + png_do_quantize(row_info, png_ptr->row_buf + 1, + png_ptr->palette_lookup, png_ptr->quantize_index); + + if (row_info->rowbytes == 0) + png_error(png_ptr, "png_do_quantize returned rowbytes=0"); + } +#endif /* READ_QUANTIZE */ + +#ifdef PNG_READ_EXPAND_16_SUPPORTED + /* Do the expansion now, after all the arithmetic has been done. Notice + * that previous transformations can handle the PNG_EXPAND_16 flag if this + * is efficient (particularly true in the case of gamma correction, where + * better accuracy results faster!) + */ + if ((png_ptr->transformations & PNG_EXPAND_16) != 0) + png_do_expand_16(row_info, png_ptr->row_buf + 1); +#endif + +#ifdef PNG_READ_GRAY_TO_RGB_SUPPORTED + /* NOTE: moved here in 1.5.4 (from much later in this list.) */ + if ((png_ptr->transformations & PNG_GRAY_TO_RGB) != 0 && + (png_ptr->mode & PNG_BACKGROUND_IS_GRAY) != 0) + png_do_gray_to_rgb(row_info, png_ptr->row_buf + 1); +#endif + +#ifdef PNG_READ_INVERT_SUPPORTED + if ((png_ptr->transformations & PNG_INVERT_MONO) != 0) + png_do_invert(row_info, png_ptr->row_buf + 1); +#endif + +#ifdef PNG_READ_INVERT_ALPHA_SUPPORTED + if ((png_ptr->transformations & PNG_INVERT_ALPHA) != 0) + png_do_read_invert_alpha(row_info, png_ptr->row_buf + 1); +#endif + +#ifdef PNG_READ_SHIFT_SUPPORTED + if ((png_ptr->transformations & PNG_SHIFT) != 0) + png_do_unshift(row_info, png_ptr->row_buf + 1, + &(png_ptr->shift)); +#endif + +#ifdef PNG_READ_PACK_SUPPORTED + if ((png_ptr->transformations & PNG_PACK) != 0) + png_do_unpack(row_info, png_ptr->row_buf + 1); +#endif + +#ifdef PNG_READ_CHECK_FOR_INVALID_INDEX_SUPPORTED + /* Added at libpng-1.5.10 */ + if (row_info->color_type == PNG_COLOR_TYPE_PALETTE && + png_ptr->num_palette_max >= 0) + png_do_check_palette_indexes(png_ptr, row_info); +#endif + +#ifdef PNG_READ_BGR_SUPPORTED + if ((png_ptr->transformations & PNG_BGR) != 0) + png_do_bgr(row_info, png_ptr->row_buf + 1); +#endif + +#ifdef PNG_READ_PACKSWAP_SUPPORTED + if ((png_ptr->transformations & PNG_PACKSWAP) != 0) + png_do_packswap(row_info, png_ptr->row_buf + 1); +#endif + +#ifdef PNG_READ_FILLER_SUPPORTED + if ((png_ptr->transformations & PNG_FILLER) != 0) + png_do_read_filler(row_info, png_ptr->row_buf + 1, + (png_uint_32)png_ptr->filler, png_ptr->flags); +#endif + +#ifdef PNG_READ_SWAP_ALPHA_SUPPORTED + if ((png_ptr->transformations & PNG_SWAP_ALPHA) != 0) + png_do_read_swap_alpha(row_info, png_ptr->row_buf + 1); +#endif + +#ifdef PNG_READ_16BIT_SUPPORTED +#ifdef PNG_READ_SWAP_SUPPORTED + if ((png_ptr->transformations & PNG_SWAP_BYTES) != 0) + png_do_swap(row_info, png_ptr->row_buf + 1); +#endif +#endif + +#ifdef PNG_READ_USER_TRANSFORM_SUPPORTED + if ((png_ptr->transformations & PNG_USER_TRANSFORM) != 0) + { + if (png_ptr->read_user_transform_fn != NULL) + (*(png_ptr->read_user_transform_fn)) /* User read transform function */ + (png_ptr, /* png_ptr */ + row_info, /* row_info: */ + /* png_uint_32 width; width of row */ + /* png_size_t rowbytes; number of bytes in row */ + /* png_byte color_type; color type of pixels */ + /* png_byte bit_depth; bit depth of samples */ + /* png_byte channels; number of channels (1-4) */ + /* png_byte pixel_depth; bits per pixel (depth*channels) */ + png_ptr->row_buf + 1); /* start of pixel data for row */ +#ifdef PNG_USER_TRANSFORM_PTR_SUPPORTED + if (png_ptr->user_transform_depth != 0) + row_info->bit_depth = png_ptr->user_transform_depth; + + if (png_ptr->user_transform_channels != 0) + row_info->channels = png_ptr->user_transform_channels; +#endif + row_info->pixel_depth = (png_byte)(row_info->bit_depth * + row_info->channels); + + row_info->rowbytes = PNG_ROWBYTES(row_info->pixel_depth, row_info->width); + } +#endif +} + +#endif /* READ_TRANSFORMS */ +#endif /* READ */ diff --git a/app/src/main/jni/libpng/pngrutil.c b/app/src/main/jni/libpng/pngrutil.c new file mode 100644 index 000000000..4c26be48c --- /dev/null +++ b/app/src/main/jni/libpng/pngrutil.c @@ -0,0 +1,4474 @@ + +/* pngrutil.c - utilities to read a PNG file + * + * Last changed in libpng 1.6.15 [November 20, 2014] + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + * + * This file contains routines that are only called from within + * libpng itself during the course of reading an image. + */ + +#include "pngpriv.h" + +#ifdef PNG_READ_SUPPORTED + +png_uint_32 PNGAPI +png_get_uint_31(png_const_structrp png_ptr, png_const_bytep buf) +{ + png_uint_32 uval = png_get_uint_32(buf); + + if (uval > PNG_UINT_31_MAX) + png_error(png_ptr, "PNG unsigned integer out of range"); + + return (uval); +} + +#if defined(PNG_READ_gAMA_SUPPORTED) || defined(PNG_READ_cHRM_SUPPORTED) +/* The following is a variation on the above for use with the fixed + * point values used for gAMA and cHRM. Instead of png_error it + * issues a warning and returns (-1) - an invalid value because both + * gAMA and cHRM use *unsigned* integers for fixed point values. + */ +#define PNG_FIXED_ERROR (-1) + +static png_fixed_point /* PRIVATE */ +png_get_fixed_point(png_structrp png_ptr, png_const_bytep buf) +{ + png_uint_32 uval = png_get_uint_32(buf); + + if (uval <= PNG_UINT_31_MAX) + return (png_fixed_point)uval; /* known to be in range */ + + /* The caller can turn off the warning by passing NULL. */ + if (png_ptr != NULL) + png_warning(png_ptr, "PNG fixed point integer out of range"); + + return PNG_FIXED_ERROR; +} +#endif + +#ifdef PNG_READ_INT_FUNCTIONS_SUPPORTED +/* NOTE: the read macros will obscure these definitions, so that if + * PNG_USE_READ_MACROS is set the library will not use them internally, + * but the APIs will still be available externally. + * + * The parentheses around "PNGAPI function_name" in the following three + * functions are necessary because they allow the macros to co-exist with + * these (unused but exported) functions. + */ + +/* Grab an unsigned 32-bit integer from a buffer in big-endian format. */ +png_uint_32 (PNGAPI +png_get_uint_32)(png_const_bytep buf) +{ + png_uint_32 uval = + ((png_uint_32)(*(buf )) << 24) + + ((png_uint_32)(*(buf + 1)) << 16) + + ((png_uint_32)(*(buf + 2)) << 8) + + ((png_uint_32)(*(buf + 3)) ) ; + + return uval; +} + +/* Grab a signed 32-bit integer from a buffer in big-endian format. The + * data is stored in the PNG file in two's complement format and there + * is no guarantee that a 'png_int_32' is exactly 32 bits, therefore + * the following code does a two's complement to native conversion. + */ +png_int_32 (PNGAPI +png_get_int_32)(png_const_bytep buf) +{ + png_uint_32 uval = png_get_uint_32(buf); + if ((uval & 0x80000000) == 0) /* non-negative */ + return uval; + + uval = (uval ^ 0xffffffff) + 1; /* 2's complement: -x = ~x+1 */ + return -(png_int_32)uval; +} + +/* Grab an unsigned 16-bit integer from a buffer in big-endian format. */ +png_uint_16 (PNGAPI +png_get_uint_16)(png_const_bytep buf) +{ + /* ANSI-C requires an int value to accomodate at least 16 bits so this + * works and allows the compiler not to worry about possible narrowing + * on 32 bit systems. (Pre-ANSI systems did not make integers smaller + * than 16 bits either.) + */ + unsigned int val = + ((unsigned int)(*buf) << 8) + + ((unsigned int)(*(buf + 1))); + + return (png_uint_16)val; +} + +#endif /* READ_INT_FUNCTIONS */ + +/* Read and check the PNG file signature */ +void /* PRIVATE */ +png_read_sig(png_structrp png_ptr, png_inforp info_ptr) +{ + png_size_t num_checked, num_to_check; + + /* Exit if the user application does not expect a signature. */ + if (png_ptr->sig_bytes >= 8) + return; + + num_checked = png_ptr->sig_bytes; + num_to_check = 8 - num_checked; + +#ifdef PNG_IO_STATE_SUPPORTED + png_ptr->io_state = PNG_IO_READING | PNG_IO_SIGNATURE; +#endif + + /* The signature must be serialized in a single I/O call. */ + png_read_data(png_ptr, &(info_ptr->signature[num_checked]), num_to_check); + png_ptr->sig_bytes = 8; + + if (png_sig_cmp(info_ptr->signature, num_checked, num_to_check) != 0) + { + if (num_checked < 4 && + png_sig_cmp(info_ptr->signature, num_checked, num_to_check - 4)) + png_error(png_ptr, "Not a PNG file"); + else + png_error(png_ptr, "PNG file corrupted by ASCII conversion"); + } + if (num_checked < 3) + png_ptr->mode |= PNG_HAVE_PNG_SIGNATURE; +} + +/* Read the chunk header (length + type name). + * Put the type name into png_ptr->chunk_name, and return the length. + */ +png_uint_32 /* PRIVATE */ +png_read_chunk_header(png_structrp png_ptr) +{ + png_byte buf[8]; + png_uint_32 length; + +#ifdef PNG_IO_STATE_SUPPORTED + png_ptr->io_state = PNG_IO_READING | PNG_IO_CHUNK_HDR; +#endif + + /* Read the length and the chunk name. + * This must be performed in a single I/O call. + */ + png_read_data(png_ptr, buf, 8); + length = png_get_uint_31(png_ptr, buf); + + /* Put the chunk name into png_ptr->chunk_name. */ + png_ptr->chunk_name = PNG_CHUNK_FROM_STRING(buf+4); + + png_debug2(0, "Reading %lx chunk, length = %lu", + (unsigned long)png_ptr->chunk_name, (unsigned long)length); + + /* Reset the crc and run it over the chunk name. */ + png_reset_crc(png_ptr); + png_calculate_crc(png_ptr, buf + 4, 4); + + /* Check to see if chunk name is valid. */ + png_check_chunk_name(png_ptr, png_ptr->chunk_name); + +#ifdef PNG_IO_STATE_SUPPORTED + png_ptr->io_state = PNG_IO_READING | PNG_IO_CHUNK_DATA; +#endif + + return length; +} + +/* Read data, and (optionally) run it through the CRC. */ +void /* PRIVATE */ +png_crc_read(png_structrp png_ptr, png_bytep buf, png_uint_32 length) +{ + if (png_ptr == NULL) + return; + + png_read_data(png_ptr, buf, length); + png_calculate_crc(png_ptr, buf, length); +} + +/* Optionally skip data and then check the CRC. Depending on whether we + * are reading an ancillary or critical chunk, and how the program has set + * things up, we may calculate the CRC on the data and print a message. + * Returns '1' if there was a CRC error, '0' otherwise. + */ +int /* PRIVATE */ +png_crc_finish(png_structrp png_ptr, png_uint_32 skip) +{ + /* The size of the local buffer for inflate is a good guess as to a + * reasonable size to use for buffering reads from the application. + */ + while (skip > 0) + { + png_uint_32 len; + png_byte tmpbuf[PNG_INFLATE_BUF_SIZE]; + + len = (sizeof tmpbuf); + if (len > skip) + len = skip; + skip -= len; + + png_crc_read(png_ptr, tmpbuf, len); + } + + if (png_crc_error(png_ptr) != 0) + { + if (PNG_CHUNK_ANCILLARY(png_ptr->chunk_name) != 0 ? + (png_ptr->flags & PNG_FLAG_CRC_ANCILLARY_NOWARN) == 0 : + (png_ptr->flags & PNG_FLAG_CRC_CRITICAL_USE) != 0) + { + png_chunk_warning(png_ptr, "CRC error"); + } + + else + png_chunk_error(png_ptr, "CRC error"); + + return (1); + } + + return (0); +} + +/* Compare the CRC stored in the PNG file with that calculated by libpng from + * the data it has read thus far. + */ +int /* PRIVATE */ +png_crc_error(png_structrp png_ptr) +{ + png_byte crc_bytes[4]; + png_uint_32 crc; + int need_crc = 1; + + if (PNG_CHUNK_ANCILLARY(png_ptr->chunk_name) != 0) + { + if ((png_ptr->flags & PNG_FLAG_CRC_ANCILLARY_MASK) == + (PNG_FLAG_CRC_ANCILLARY_USE | PNG_FLAG_CRC_ANCILLARY_NOWARN)) + need_crc = 0; + } + + else /* critical */ + { + if ((png_ptr->flags & PNG_FLAG_CRC_CRITICAL_IGNORE) != 0) + need_crc = 0; + } + +#ifdef PNG_IO_STATE_SUPPORTED + png_ptr->io_state = PNG_IO_READING | PNG_IO_CHUNK_CRC; +#endif + + /* The chunk CRC must be serialized in a single I/O call. */ + png_read_data(png_ptr, crc_bytes, 4); + + if (need_crc != 0) + { + crc = png_get_uint_32(crc_bytes); + return ((int)(crc != png_ptr->crc)); + } + + else + return (0); +} + +#if defined(PNG_READ_iCCP_SUPPORTED) || defined(PNG_READ_iTXt_SUPPORTED) ||\ + defined(PNG_READ_pCAL_SUPPORTED) || defined(PNG_READ_sCAL_SUPPORTED) ||\ + defined(PNG_READ_sPLT_SUPPORTED) || defined(PNG_READ_tEXt_SUPPORTED) ||\ + defined(PNG_READ_zTXt_SUPPORTED) || defined(PNG_SEQUENTIAL_READ_SUPPORTED) +/* Manage the read buffer; this simply reallocates the buffer if it is not small + * enough (or if it is not allocated). The routine returns a pointer to the + * buffer; if an error occurs and 'warn' is set the routine returns NULL, else + * it will call png_error (via png_malloc) on failure. (warn == 2 means + * 'silent'). + */ +static png_bytep +png_read_buffer(png_structrp png_ptr, png_alloc_size_t new_size, int warn) +{ + png_bytep buffer = png_ptr->read_buffer; + + if (buffer != NULL && new_size > png_ptr->read_buffer_size) + { + png_ptr->read_buffer = NULL; + png_ptr->read_buffer = NULL; + png_ptr->read_buffer_size = 0; + png_free(png_ptr, buffer); + buffer = NULL; + } + + if (buffer == NULL) + { + buffer = png_voidcast(png_bytep, png_malloc_base(png_ptr, new_size)); + + if (buffer != NULL) + { + png_ptr->read_buffer = buffer; + png_ptr->read_buffer_size = new_size; + } + + else if (warn < 2) /* else silent */ + { + if (warn != 0) + png_chunk_warning(png_ptr, "insufficient memory to read chunk"); + + else + png_chunk_error(png_ptr, "insufficient memory to read chunk"); + } + } + + return buffer; +} +#endif /* READ_iCCP|iTXt|pCAL|sCAL|sPLT|tEXt|zTXt|SEQUENTIAL_READ */ + +/* png_inflate_claim: claim the zstream for some nefarious purpose that involves + * decompression. Returns Z_OK on success, else a zlib error code. It checks + * the owner but, in final release builds, just issues a warning if some other + * chunk apparently owns the stream. Prior to release it does a png_error. + */ +static int +png_inflate_claim(png_structrp png_ptr, png_uint_32 owner) +{ + if (png_ptr->zowner != 0) + { + char msg[64]; + + PNG_STRING_FROM_CHUNK(msg, png_ptr->zowner); + /* So the message that results is " using zstream"; this is an + * internal error, but is very useful for debugging. i18n requirements + * are minimal. + */ + (void)png_safecat(msg, (sizeof msg), 4, " using zstream"); +#if PNG_LIBPNG_BUILD_BASE_TYPE >= PNG_LIBPNG_BUILD_RC + png_chunk_warning(png_ptr, msg); + png_ptr->zowner = 0; +#else + png_chunk_error(png_ptr, msg); +#endif + } + + /* Implementation note: unlike 'png_deflate_claim' this internal function + * does not take the size of the data as an argument. Some efficiency could + * be gained by using this when it is known *if* the zlib stream itself does + * not record the number; however, this is an illusion: the original writer + * of the PNG may have selected a lower window size, and we really must + * follow that because, for systems with with limited capabilities, we + * would otherwise reject the application's attempts to use a smaller window + * size (zlib doesn't have an interface to say "this or lower"!). + * + * inflateReset2 was added to zlib 1.2.4; before this the window could not be + * reset, therefore it is necessary to always allocate the maximum window + * size with earlier zlibs just in case later compressed chunks need it. + */ + { + int ret; /* zlib return code */ +#if PNG_ZLIB_VERNUM >= 0x1240 + +# if defined(PNG_SET_OPTION_SUPPORTED) && defined(PNG_MAXIMUM_INFLATE_WINDOW) + int window_bits; + + if (((png_ptr->options >> PNG_MAXIMUM_INFLATE_WINDOW) & 3) == + PNG_OPTION_ON) + window_bits = 15; + + else + window_bits = 0; +# else +# define window_bits 0 +# endif +#endif + + /* Set this for safety, just in case the previous owner left pointers to + * memory allocations. + */ + png_ptr->zstream.next_in = NULL; + png_ptr->zstream.avail_in = 0; + png_ptr->zstream.next_out = NULL; + png_ptr->zstream.avail_out = 0; + + if ((png_ptr->flags & PNG_FLAG_ZSTREAM_INITIALIZED) != 0) + { +#if PNG_ZLIB_VERNUM < 0x1240 + ret = inflateReset(&png_ptr->zstream); +#else + ret = inflateReset2(&png_ptr->zstream, window_bits); +#endif + } + + else + { +#if PNG_ZLIB_VERNUM < 0x1240 + ret = inflateInit(&png_ptr->zstream); +#else + ret = inflateInit2(&png_ptr->zstream, window_bits); +#endif + + if (ret == Z_OK) + png_ptr->flags |= PNG_FLAG_ZSTREAM_INITIALIZED; + } + + if (ret == Z_OK) + png_ptr->zowner = owner; + + else + png_zstream_error(png_ptr, ret); + + return ret; + } + +#ifdef window_bits +# undef window_bits +#endif +} + +#ifdef PNG_READ_COMPRESSED_TEXT_SUPPORTED +/* png_inflate now returns zlib error codes including Z_OK and Z_STREAM_END to + * allow the caller to do multiple calls if required. If the 'finish' flag is + * set Z_FINISH will be passed to the final inflate() call and Z_STREAM_END must + * be returned or there has been a problem, otherwise Z_SYNC_FLUSH is used and + * Z_OK or Z_STREAM_END will be returned on success. + * + * The input and output sizes are updated to the actual amounts of data consumed + * or written, not the amount available (as in a z_stream). The data pointers + * are not changed, so the next input is (data+input_size) and the next + * available output is (output+output_size). + */ +static int +png_inflate(png_structrp png_ptr, png_uint_32 owner, int finish, + /* INPUT: */ png_const_bytep input, png_uint_32p input_size_ptr, + /* OUTPUT: */ png_bytep output, png_alloc_size_t *output_size_ptr) +{ + if (png_ptr->zowner == owner) /* Else not claimed */ + { + int ret; + png_alloc_size_t avail_out = *output_size_ptr; + png_uint_32 avail_in = *input_size_ptr; + + /* zlib can't necessarily handle more than 65535 bytes at once (i.e. it + * can't even necessarily handle 65536 bytes) because the type uInt is + * "16 bits or more". Consequently it is necessary to chunk the input to + * zlib. This code uses ZLIB_IO_MAX, from pngpriv.h, as the maximum (the + * maximum value that can be stored in a uInt.) It is possible to set + * ZLIB_IO_MAX to a lower value in pngpriv.h and this may sometimes have + * a performance advantage, because it reduces the amount of data accessed + * at each step and that may give the OS more time to page it in. + */ + png_ptr->zstream.next_in = PNGZ_INPUT_CAST(input); + /* avail_in and avail_out are set below from 'size' */ + png_ptr->zstream.avail_in = 0; + png_ptr->zstream.avail_out = 0; + + /* Read directly into the output if it is available (this is set to + * a local buffer below if output is NULL). + */ + if (output != NULL) + png_ptr->zstream.next_out = output; + + do + { + uInt avail; + Byte local_buffer[PNG_INFLATE_BUF_SIZE]; + + /* zlib INPUT BUFFER */ + /* The setting of 'avail_in' used to be outside the loop; by setting it + * inside it is possible to chunk the input to zlib and simply rely on + * zlib to advance the 'next_in' pointer. This allows arbitrary + * amounts of data to be passed through zlib at the unavoidable cost of + * requiring a window save (memcpy of up to 32768 output bytes) + * every ZLIB_IO_MAX input bytes. + */ + avail_in += png_ptr->zstream.avail_in; /* not consumed last time */ + + avail = ZLIB_IO_MAX; + + if (avail_in < avail) + avail = (uInt)avail_in; /* safe: < than ZLIB_IO_MAX */ + + avail_in -= avail; + png_ptr->zstream.avail_in = avail; + + /* zlib OUTPUT BUFFER */ + avail_out += png_ptr->zstream.avail_out; /* not written last time */ + + avail = ZLIB_IO_MAX; /* maximum zlib can process */ + + if (output == NULL) + { + /* Reset the output buffer each time round if output is NULL and + * make available the full buffer, up to 'remaining_space' + */ + png_ptr->zstream.next_out = local_buffer; + if ((sizeof local_buffer) < avail) + avail = (sizeof local_buffer); + } + + if (avail_out < avail) + avail = (uInt)avail_out; /* safe: < ZLIB_IO_MAX */ + + png_ptr->zstream.avail_out = avail; + avail_out -= avail; + + /* zlib inflate call */ + /* In fact 'avail_out' may be 0 at this point, that happens at the end + * of the read when the final LZ end code was not passed at the end of + * the previous chunk of input data. Tell zlib if we have reached the + * end of the output buffer. + */ + ret = inflate(&png_ptr->zstream, avail_out > 0 ? Z_NO_FLUSH : + (finish ? Z_FINISH : Z_SYNC_FLUSH)); + } while (ret == Z_OK); + + /* For safety kill the local buffer pointer now */ + if (output == NULL) + png_ptr->zstream.next_out = NULL; + + /* Claw back the 'size' and 'remaining_space' byte counts. */ + avail_in += png_ptr->zstream.avail_in; + avail_out += png_ptr->zstream.avail_out; + + /* Update the input and output sizes; the updated values are the amount + * consumed or written, effectively the inverse of what zlib uses. + */ + if (avail_out > 0) + *output_size_ptr -= avail_out; + + if (avail_in > 0) + *input_size_ptr -= avail_in; + + /* Ensure png_ptr->zstream.msg is set (even in the success case!) */ + png_zstream_error(png_ptr, ret); + return ret; + } + + else + { + /* This is a bad internal error. The recovery assigns to the zstream msg + * pointer, which is not owned by the caller, but this is safe; it's only + * used on errors! + */ + png_ptr->zstream.msg = PNGZ_MSG_CAST("zstream unclaimed"); + return Z_STREAM_ERROR; + } +} + +/* + * Decompress trailing data in a chunk. The assumption is that read_buffer + * points at an allocated area holding the contents of a chunk with a + * trailing compressed part. What we get back is an allocated area + * holding the original prefix part and an uncompressed version of the + * trailing part (the malloc area passed in is freed). + */ +static int +png_decompress_chunk(png_structrp png_ptr, + png_uint_32 chunklength, png_uint_32 prefix_size, + png_alloc_size_t *newlength /* must be initialized to the maximum! */, + int terminate /*add a '\0' to the end of the uncompressed data*/) +{ + /* TODO: implement different limits for different types of chunk. + * + * The caller supplies *newlength set to the maximum length of the + * uncompressed data, but this routine allocates space for the prefix and + * maybe a '\0' terminator too. We have to assume that 'prefix_size' is + * limited only by the maximum chunk size. + */ + png_alloc_size_t limit = PNG_SIZE_MAX; + +# ifdef PNG_SET_CHUNK_MALLOC_LIMIT_SUPPORTED + if (png_ptr->user_chunk_malloc_max > 0 && + png_ptr->user_chunk_malloc_max < limit) + limit = png_ptr->user_chunk_malloc_max; +# elif PNG_USER_CHUNK_MALLOC_MAX > 0 + if (PNG_USER_CHUNK_MALLOC_MAX < limit) + limit = PNG_USER_CHUNK_MALLOC_MAX; +# endif + + if (limit >= prefix_size + (terminate != 0)) + { + int ret; + + limit -= prefix_size + (terminate != 0); + + if (limit < *newlength) + *newlength = limit; + + /* Now try to claim the stream. */ + ret = png_inflate_claim(png_ptr, png_ptr->chunk_name); + + if (ret == Z_OK) + { + png_uint_32 lzsize = chunklength - prefix_size; + + ret = png_inflate(png_ptr, png_ptr->chunk_name, 1/*finish*/, + /* input: */ png_ptr->read_buffer + prefix_size, &lzsize, + /* output: */ NULL, newlength); + + if (ret == Z_STREAM_END) + { + /* Use 'inflateReset' here, not 'inflateReset2' because this + * preserves the previously decided window size (otherwise it would + * be necessary to store the previous window size.) In practice + * this doesn't matter anyway, because png_inflate will call inflate + * with Z_FINISH in almost all cases, so the window will not be + * maintained. + */ + if (inflateReset(&png_ptr->zstream) == Z_OK) + { + /* Because of the limit checks above we know that the new, + * expanded, size will fit in a size_t (let alone an + * png_alloc_size_t). Use png_malloc_base here to avoid an + * extra OOM message. + */ + png_alloc_size_t new_size = *newlength; + png_alloc_size_t buffer_size = prefix_size + new_size + + (terminate != 0); + png_bytep text = png_voidcast(png_bytep, png_malloc_base(png_ptr, + buffer_size)); + + if (text != NULL) + { + ret = png_inflate(png_ptr, png_ptr->chunk_name, 1/*finish*/, + png_ptr->read_buffer + prefix_size, &lzsize, + text + prefix_size, newlength); + + if (ret == Z_STREAM_END) + { + if (new_size == *newlength) + { + if (terminate != 0) + text[prefix_size + *newlength] = 0; + + if (prefix_size > 0) + memcpy(text, png_ptr->read_buffer, prefix_size); + + { + png_bytep old_ptr = png_ptr->read_buffer; + + png_ptr->read_buffer = text; + png_ptr->read_buffer_size = buffer_size; + text = old_ptr; /* freed below */ + } + } + + else + { + /* The size changed on the second read, there can be no + * guarantee that anything is correct at this point. + * The 'msg' pointer has been set to "unexpected end of + * LZ stream", which is fine, but return an error code + * that the caller won't accept. + */ + ret = PNG_UNEXPECTED_ZLIB_RETURN; + } + } + + else if (ret == Z_OK) + ret = PNG_UNEXPECTED_ZLIB_RETURN; /* for safety */ + + /* Free the text pointer (this is the old read_buffer on + * success) + */ + png_free(png_ptr, text); + text = NULL; + + /* This really is very benign, but it's still an error because + * the extra space may otherwise be used as a Trojan Horse. + */ + if (ret == Z_STREAM_END && + chunklength - prefix_size != lzsize) + png_chunk_benign_error(png_ptr, "extra compressed data"); + } + + else + { + /* Out of memory allocating the buffer */ + ret = Z_MEM_ERROR; + png_zstream_error(png_ptr, Z_MEM_ERROR); + } + } + + else + { + /* inflateReset failed, store the error message */ + png_zstream_error(png_ptr, ret); + + if (ret == Z_STREAM_END) + ret = PNG_UNEXPECTED_ZLIB_RETURN; + } + } + + else if (ret == Z_OK) + ret = PNG_UNEXPECTED_ZLIB_RETURN; + + /* Release the claimed stream */ + png_ptr->zowner = 0; + } + + else /* the claim failed */ if (ret == Z_STREAM_END) /* impossible! */ + ret = PNG_UNEXPECTED_ZLIB_RETURN; + + return ret; + } + + else + { + /* Application/configuration limits exceeded */ + png_zstream_error(png_ptr, Z_MEM_ERROR); + return Z_MEM_ERROR; + } +} +#endif /* READ_COMPRESSED_TEXT */ + +#ifdef PNG_READ_iCCP_SUPPORTED +/* Perform a partial read and decompress, producing 'avail_out' bytes and + * reading from the current chunk as required. + */ +static int +png_inflate_read(png_structrp png_ptr, png_bytep read_buffer, uInt read_size, + png_uint_32p chunk_bytes, png_bytep next_out, png_alloc_size_t *out_size, + int finish) +{ + if (png_ptr->zowner == png_ptr->chunk_name) + { + int ret; + + /* next_in and avail_in must have been initialized by the caller. */ + png_ptr->zstream.next_out = next_out; + png_ptr->zstream.avail_out = 0; /* set in the loop */ + + do + { + if (png_ptr->zstream.avail_in == 0) + { + if (read_size > *chunk_bytes) + read_size = (uInt)*chunk_bytes; + *chunk_bytes -= read_size; + + if (read_size > 0) + png_crc_read(png_ptr, read_buffer, read_size); + + png_ptr->zstream.next_in = read_buffer; + png_ptr->zstream.avail_in = read_size; + } + + if (png_ptr->zstream.avail_out == 0) + { + uInt avail = ZLIB_IO_MAX; + if (avail > *out_size) + avail = (uInt)*out_size; + *out_size -= avail; + + png_ptr->zstream.avail_out = avail; + } + + /* Use Z_SYNC_FLUSH when there is no more chunk data to ensure that all + * the available output is produced; this allows reading of truncated + * streams. + */ + ret = inflate(&png_ptr->zstream, + *chunk_bytes > 0 ? Z_NO_FLUSH : (finish ? Z_FINISH : Z_SYNC_FLUSH)); + } + while (ret == Z_OK && (*out_size > 0 || png_ptr->zstream.avail_out > 0)); + + *out_size += png_ptr->zstream.avail_out; + png_ptr->zstream.avail_out = 0; /* Should not be required, but is safe */ + + /* Ensure the error message pointer is always set: */ + png_zstream_error(png_ptr, ret); + return ret; + } + + else + { + png_ptr->zstream.msg = PNGZ_MSG_CAST("zstream unclaimed"); + return Z_STREAM_ERROR; + } +} +#endif + +/* Read and check the IDHR chunk */ + +void /* PRIVATE */ +png_handle_IHDR(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_byte buf[13]; + png_uint_32 width, height; + int bit_depth, color_type, compression_type, filter_type; + int interlace_type; + + png_debug(1, "in png_handle_IHDR"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) != 0) + png_chunk_error(png_ptr, "out of place"); + + /* Check the length */ + if (length != 13) + png_chunk_error(png_ptr, "invalid"); + + png_ptr->mode |= PNG_HAVE_IHDR; + + png_crc_read(png_ptr, buf, 13); + png_crc_finish(png_ptr, 0); + + width = png_get_uint_31(png_ptr, buf); + height = png_get_uint_31(png_ptr, buf + 4); + bit_depth = buf[8]; + color_type = buf[9]; + compression_type = buf[10]; + filter_type = buf[11]; + interlace_type = buf[12]; + + /* Set internal variables */ + png_ptr->width = width; + png_ptr->height = height; + png_ptr->bit_depth = (png_byte)bit_depth; + png_ptr->interlaced = (png_byte)interlace_type; + png_ptr->color_type = (png_byte)color_type; +#ifdef PNG_MNG_FEATURES_SUPPORTED + png_ptr->filter_type = (png_byte)filter_type; +#endif + png_ptr->compression_type = (png_byte)compression_type; + + /* Find number of channels */ + switch (png_ptr->color_type) + { + default: /* invalid, png_set_IHDR calls png_error */ + case PNG_COLOR_TYPE_GRAY: + case PNG_COLOR_TYPE_PALETTE: + png_ptr->channels = 1; + break; + + case PNG_COLOR_TYPE_RGB: + png_ptr->channels = 3; + break; + + case PNG_COLOR_TYPE_GRAY_ALPHA: + png_ptr->channels = 2; + break; + + case PNG_COLOR_TYPE_RGB_ALPHA: + png_ptr->channels = 4; + break; + } + + /* Set up other useful info */ + png_ptr->pixel_depth = (png_byte)(png_ptr->bit_depth * png_ptr->channels); + png_ptr->rowbytes = PNG_ROWBYTES(png_ptr->pixel_depth, png_ptr->width); + png_debug1(3, "bit_depth = %d", png_ptr->bit_depth); + png_debug1(3, "channels = %d", png_ptr->channels); + png_debug1(3, "rowbytes = %lu", (unsigned long)png_ptr->rowbytes); + png_set_IHDR(png_ptr, info_ptr, width, height, bit_depth, + color_type, interlace_type, compression_type, filter_type); +} + +/* Read and check the palette */ +void /* PRIVATE */ +png_handle_PLTE(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_color palette[PNG_MAX_PALETTE_LENGTH]; + int num, i; +#ifdef PNG_POINTER_INDEXING_SUPPORTED + png_colorp pal_ptr; +#endif + + png_debug(1, "in png_handle_PLTE"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + /* Moved to before the 'after IDAT' check below because otherwise duplicate + * PLTE chunks are potentially ignored (the spec says there shall not be more + * than one PLTE, the error is not treated as benign, so this check trumps + * the requirement that PLTE appears before IDAT.) + */ + else if ((png_ptr->mode & PNG_HAVE_PLTE) != 0) + png_chunk_error(png_ptr, "duplicate"); + + else if ((png_ptr->mode & PNG_HAVE_IDAT) != 0) + { + /* This is benign because the non-benign error happened before, when an + * IDAT was encountered in a color-mapped image with no PLTE. + */ + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + png_ptr->mode |= PNG_HAVE_PLTE; + + if ((png_ptr->color_type & PNG_COLOR_MASK_COLOR) == 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "ignored in grayscale PNG"); + return; + } + +#ifndef PNG_READ_OPT_PLTE_SUPPORTED + if (png_ptr->color_type != PNG_COLOR_TYPE_PALETTE) + { + png_crc_finish(png_ptr, length); + return; + } +#endif + + if (length > 3*PNG_MAX_PALETTE_LENGTH || length % 3) + { + png_crc_finish(png_ptr, length); + + if (png_ptr->color_type != PNG_COLOR_TYPE_PALETTE) + png_chunk_benign_error(png_ptr, "invalid"); + + else + png_chunk_error(png_ptr, "invalid"); + + return; + } + + /* The cast is safe because 'length' is less than 3*PNG_MAX_PALETTE_LENGTH */ + num = (int)length / 3; + +#ifdef PNG_POINTER_INDEXING_SUPPORTED + for (i = 0, pal_ptr = palette; i < num; i++, pal_ptr++) + { + png_byte buf[3]; + + png_crc_read(png_ptr, buf, 3); + pal_ptr->red = buf[0]; + pal_ptr->green = buf[1]; + pal_ptr->blue = buf[2]; + } +#else + for (i = 0; i < num; i++) + { + png_byte buf[3]; + + png_crc_read(png_ptr, buf, 3); + /* Don't depend upon png_color being any order */ + palette[i].red = buf[0]; + palette[i].green = buf[1]; + palette[i].blue = buf[2]; + } +#endif + + /* If we actually need the PLTE chunk (ie for a paletted image), we do + * whatever the normal CRC configuration tells us. However, if we + * have an RGB image, the PLTE can be considered ancillary, so + * we will act as though it is. + */ +#ifndef PNG_READ_OPT_PLTE_SUPPORTED + if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE) +#endif + { + png_crc_finish(png_ptr, 0); + } + +#ifndef PNG_READ_OPT_PLTE_SUPPORTED + else if (png_crc_error(png_ptr) != 0) /* Only if we have a CRC error */ + { + /* If we don't want to use the data from an ancillary chunk, + * we have two options: an error abort, or a warning and we + * ignore the data in this chunk (which should be OK, since + * it's considered ancillary for a RGB or RGBA image). + * + * IMPLEMENTATION NOTE: this is only here because png_crc_finish uses the + * chunk type to determine whether to check the ancillary or the critical + * flags. + */ + if ((png_ptr->flags & PNG_FLAG_CRC_ANCILLARY_USE) == 0) + { + if ((png_ptr->flags & PNG_FLAG_CRC_ANCILLARY_NOWARN) != 0) + return; + + else + png_chunk_error(png_ptr, "CRC error"); + } + + /* Otherwise, we (optionally) emit a warning and use the chunk. */ + else if ((png_ptr->flags & PNG_FLAG_CRC_ANCILLARY_NOWARN) == 0) + png_chunk_warning(png_ptr, "CRC error"); + } +#endif + + /* TODO: png_set_PLTE has the side effect of setting png_ptr->palette to its + * own copy of the palette. This has the side effect that when png_start_row + * is called (this happens after any call to png_read_update_info) the + * info_ptr palette gets changed. This is extremely unexpected and + * confusing. + * + * Fix this by not sharing the palette in this way. + */ + png_set_PLTE(png_ptr, info_ptr, palette, num); + + /* The three chunks, bKGD, hIST and tRNS *must* appear after PLTE and before + * IDAT. Prior to 1.6.0 this was not checked; instead the code merely + * checked the apparent validity of a tRNS chunk inserted before PLTE on a + * palette PNG. 1.6.0 attempts to rigorously follow the standard and + * therefore does a benign error if the erroneous condition is detected *and* + * cancels the tRNS if the benign error returns. The alternative is to + * amend the standard since it would be rather hypocritical of the standards + * maintainers to ignore it. + */ +#ifdef PNG_READ_tRNS_SUPPORTED + if (png_ptr->num_trans > 0 || + (info_ptr != NULL && (info_ptr->valid & PNG_INFO_tRNS) != 0)) + { + /* Cancel this because otherwise it would be used if the transforms + * require it. Don't cancel the 'valid' flag because this would prevent + * detection of duplicate chunks. + */ + png_ptr->num_trans = 0; + + if (info_ptr != NULL) + info_ptr->num_trans = 0; + + png_chunk_benign_error(png_ptr, "tRNS must be after"); + } +#endif + +#ifdef PNG_READ_hIST_SUPPORTED + if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_hIST) != 0) + png_chunk_benign_error(png_ptr, "hIST must be after"); +#endif + +#ifdef PNG_READ_bKGD_SUPPORTED + if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_bKGD) != 0) + png_chunk_benign_error(png_ptr, "bKGD must be after"); +#endif +} + +void /* PRIVATE */ +png_handle_IEND(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_debug(1, "in png_handle_IEND"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0 || + (png_ptr->mode & PNG_HAVE_IDAT) == 0) + png_chunk_error(png_ptr, "out of place"); + + png_ptr->mode |= (PNG_AFTER_IDAT | PNG_HAVE_IEND); + + png_crc_finish(png_ptr, length); + + if (length != 0) + png_chunk_benign_error(png_ptr, "invalid"); + + PNG_UNUSED(info_ptr) +} + +#ifdef PNG_READ_gAMA_SUPPORTED +void /* PRIVATE */ +png_handle_gAMA(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_fixed_point igamma; + png_byte buf[4]; + + png_debug(1, "in png_handle_gAMA"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if ((png_ptr->mode & (PNG_HAVE_IDAT|PNG_HAVE_PLTE)) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + if (length != 4) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + png_crc_read(png_ptr, buf, 4); + + if (png_crc_finish(png_ptr, 0) != 0) + return; + + igamma = png_get_fixed_point(NULL, buf); + + png_colorspace_set_gamma(png_ptr, &png_ptr->colorspace, igamma); + png_colorspace_sync(png_ptr, info_ptr); +} +#endif + +#ifdef PNG_READ_sBIT_SUPPORTED +void /* PRIVATE */ +png_handle_sBIT(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + unsigned int truelen, i; + png_byte sample_depth; + png_byte buf[4]; + + png_debug(1, "in png_handle_sBIT"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if ((png_ptr->mode & (PNG_HAVE_IDAT|PNG_HAVE_PLTE)) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_sBIT) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "duplicate"); + return; + } + + if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE) + { + truelen = 3; + sample_depth = 8; + } + + else + { + truelen = png_ptr->channels; + sample_depth = png_ptr->bit_depth; + } + + if (length != truelen || length > 4) + { + png_chunk_benign_error(png_ptr, "invalid"); + png_crc_finish(png_ptr, length); + return; + } + + buf[0] = buf[1] = buf[2] = buf[3] = sample_depth; + png_crc_read(png_ptr, buf, truelen); + + if (png_crc_finish(png_ptr, 0) != 0) + return; + + for (i=0; i sample_depth) + { + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + if ((png_ptr->color_type & PNG_COLOR_MASK_COLOR) != 0) + { + png_ptr->sig_bit.red = buf[0]; + png_ptr->sig_bit.green = buf[1]; + png_ptr->sig_bit.blue = buf[2]; + png_ptr->sig_bit.alpha = buf[3]; + } + + else + { + png_ptr->sig_bit.gray = buf[0]; + png_ptr->sig_bit.red = buf[0]; + png_ptr->sig_bit.green = buf[0]; + png_ptr->sig_bit.blue = buf[0]; + png_ptr->sig_bit.alpha = buf[1]; + } + + png_set_sBIT(png_ptr, info_ptr, &(png_ptr->sig_bit)); +} +#endif + +#ifdef PNG_READ_cHRM_SUPPORTED +void /* PRIVATE */ +png_handle_cHRM(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_byte buf[32]; + png_xy xy; + + png_debug(1, "in png_handle_cHRM"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if ((png_ptr->mode & (PNG_HAVE_IDAT|PNG_HAVE_PLTE)) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + if (length != 32) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + png_crc_read(png_ptr, buf, 32); + + if (png_crc_finish(png_ptr, 0) != 0) + return; + + xy.whitex = png_get_fixed_point(NULL, buf); + xy.whitey = png_get_fixed_point(NULL, buf + 4); + xy.redx = png_get_fixed_point(NULL, buf + 8); + xy.redy = png_get_fixed_point(NULL, buf + 12); + xy.greenx = png_get_fixed_point(NULL, buf + 16); + xy.greeny = png_get_fixed_point(NULL, buf + 20); + xy.bluex = png_get_fixed_point(NULL, buf + 24); + xy.bluey = png_get_fixed_point(NULL, buf + 28); + + if (xy.whitex == PNG_FIXED_ERROR || + xy.whitey == PNG_FIXED_ERROR || + xy.redx == PNG_FIXED_ERROR || + xy.redy == PNG_FIXED_ERROR || + xy.greenx == PNG_FIXED_ERROR || + xy.greeny == PNG_FIXED_ERROR || + xy.bluex == PNG_FIXED_ERROR || + xy.bluey == PNG_FIXED_ERROR) + { + png_chunk_benign_error(png_ptr, "invalid values"); + return; + } + + /* If a colorspace error has already been output skip this chunk */ + if ((png_ptr->colorspace.flags & PNG_COLORSPACE_INVALID) != 0) + return; + + if ((png_ptr->colorspace.flags & PNG_COLORSPACE_FROM_cHRM) != 0) + { + png_ptr->colorspace.flags |= PNG_COLORSPACE_INVALID; + png_colorspace_sync(png_ptr, info_ptr); + png_chunk_benign_error(png_ptr, "duplicate"); + return; + } + + png_ptr->colorspace.flags |= PNG_COLORSPACE_FROM_cHRM; + (void)png_colorspace_set_chromaticities(png_ptr, &png_ptr->colorspace, &xy, + 1/*prefer cHRM values*/); + png_colorspace_sync(png_ptr, info_ptr); +} +#endif + +#ifdef PNG_READ_sRGB_SUPPORTED +void /* PRIVATE */ +png_handle_sRGB(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_byte intent; + + png_debug(1, "in png_handle_sRGB"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if ((png_ptr->mode & (PNG_HAVE_IDAT|PNG_HAVE_PLTE)) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + if (length != 1) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + png_crc_read(png_ptr, &intent, 1); + + if (png_crc_finish(png_ptr, 0) != 0) + return; + + /* If a colorspace error has already been output skip this chunk */ + if ((png_ptr->colorspace.flags & PNG_COLORSPACE_INVALID) != 0) + return; + + /* Only one sRGB or iCCP chunk is allowed, use the HAVE_INTENT flag to detect + * this. + */ + if ((png_ptr->colorspace.flags & PNG_COLORSPACE_HAVE_INTENT) != 0) + { + png_ptr->colorspace.flags |= PNG_COLORSPACE_INVALID; + png_colorspace_sync(png_ptr, info_ptr); + png_chunk_benign_error(png_ptr, "too many profiles"); + return; + } + + (void)png_colorspace_set_sRGB(png_ptr, &png_ptr->colorspace, intent); + png_colorspace_sync(png_ptr, info_ptr); +} +#endif /* READ_sRGB */ + +#ifdef PNG_READ_iCCP_SUPPORTED +void /* PRIVATE */ +png_handle_iCCP(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +/* Note: this does not properly handle profiles that are > 64K under DOS */ +{ + png_const_charp errmsg = NULL; /* error message output, or no error */ + int finished = 0; /* crc checked */ + + png_debug(1, "in png_handle_iCCP"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if ((png_ptr->mode & (PNG_HAVE_IDAT|PNG_HAVE_PLTE)) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + /* Consistent with all the above colorspace handling an obviously *invalid* + * chunk is just ignored, so does not invalidate the color space. An + * alternative is to set the 'invalid' flags at the start of this routine + * and only clear them in they were not set before and all the tests pass. + * The minimum 'deflate' stream is assumed to be just the 2 byte header and + * 4 byte checksum. The keyword must be at least one character and there is + * a terminator (0) byte and the compression method. + */ + if (length < 9) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "too short"); + return; + } + + /* If a colorspace error has already been output skip this chunk */ + if ((png_ptr->colorspace.flags & PNG_COLORSPACE_INVALID) != 0) + { + png_crc_finish(png_ptr, length); + return; + } + + /* Only one sRGB or iCCP chunk is allowed, use the HAVE_INTENT flag to detect + * this. + */ + if ((png_ptr->colorspace.flags & PNG_COLORSPACE_HAVE_INTENT) == 0) + { + uInt read_length, keyword_length; + char keyword[81]; + + /* Find the keyword; the keyword plus separator and compression method + * bytes can be at most 81 characters long. + */ + read_length = 81; /* maximum */ + if (read_length > length) + read_length = (uInt)length; + + png_crc_read(png_ptr, (png_bytep)keyword, read_length); + length -= read_length; + + keyword_length = 0; + while (keyword_length < 80 && keyword_length < read_length && + keyword[keyword_length] != 0) + ++keyword_length; + + /* TODO: make the keyword checking common */ + if (keyword_length >= 1 && keyword_length <= 79) + { + /* We only understand '0' compression - deflate - so if we get a + * different value we can't safely decode the chunk. + */ + if (keyword_length+1 < read_length && + keyword[keyword_length+1] == PNG_COMPRESSION_TYPE_BASE) + { + read_length -= keyword_length+2; + + if (png_inflate_claim(png_ptr, png_iCCP) == Z_OK) + { + Byte profile_header[132]; + Byte local_buffer[PNG_INFLATE_BUF_SIZE]; + png_alloc_size_t size = (sizeof profile_header); + + png_ptr->zstream.next_in = (Bytef*)keyword + (keyword_length+2); + png_ptr->zstream.avail_in = read_length; + (void)png_inflate_read(png_ptr, local_buffer, + (sizeof local_buffer), &length, profile_header, &size, + 0/*finish: don't, because the output is too small*/); + + if (size == 0) + { + /* We have the ICC profile header; do the basic header checks. + */ + const png_uint_32 profile_length = + png_get_uint_32(profile_header); + + if (png_icc_check_length(png_ptr, &png_ptr->colorspace, + keyword, profile_length) != 0) + { + /* The length is apparently ok, so we can check the 132 + * byte header. + */ + if (png_icc_check_header(png_ptr, &png_ptr->colorspace, + keyword, profile_length, profile_header, + png_ptr->color_type) != 0) + { + /* Now read the tag table; a variable size buffer is + * needed at this point, allocate one for the whole + * profile. The header check has already validated + * that none of these stuff will overflow. + */ + const png_uint_32 tag_count = png_get_uint_32( + profile_header+128); + png_bytep profile = png_read_buffer(png_ptr, + profile_length, 2/*silent*/); + + if (profile != NULL) + { + memcpy(profile, profile_header, + (sizeof profile_header)); + + size = 12 * tag_count; + + (void)png_inflate_read(png_ptr, local_buffer, + (sizeof local_buffer), &length, + profile + (sizeof profile_header), &size, 0); + + /* Still expect a buffer error because we expect + * there to be some tag data! + */ + if (size == 0) + { + if (png_icc_check_tag_table(png_ptr, + &png_ptr->colorspace, keyword, profile_length, + profile) != 0) + { + /* The profile has been validated for basic + * security issues, so read the whole thing in. + */ + size = profile_length - (sizeof profile_header) + - 12 * tag_count; + + (void)png_inflate_read(png_ptr, local_buffer, + (sizeof local_buffer), &length, + profile + (sizeof profile_header) + + 12 * tag_count, &size, 1/*finish*/); + + if (length > 0 && !(png_ptr->flags & + PNG_FLAG_BENIGN_ERRORS_WARN)) + errmsg = "extra compressed data"; + + /* But otherwise allow extra data: */ + else if (size == 0) + { + if (length > 0) + { + /* This can be handled completely, so + * keep going. + */ + png_chunk_warning(png_ptr, + "extra compressed data"); + } + + png_crc_finish(png_ptr, length); + finished = 1; + +# ifdef PNG_sRGB_SUPPORTED + /* Check for a match against sRGB */ + png_icc_set_sRGB(png_ptr, + &png_ptr->colorspace, profile, + png_ptr->zstream.adler); +# endif + + /* Steal the profile for info_ptr. */ + if (info_ptr != NULL) + { + png_free_data(png_ptr, info_ptr, + PNG_FREE_ICCP, 0); + + info_ptr->iccp_name = png_voidcast(char*, + png_malloc_base(png_ptr, + keyword_length+1)); + if (info_ptr->iccp_name != NULL) + { + memcpy(info_ptr->iccp_name, keyword, + keyword_length+1); + info_ptr->iccp_proflen = + profile_length; + info_ptr->iccp_profile = profile; + png_ptr->read_buffer = NULL; /*steal*/ + info_ptr->free_me |= PNG_FREE_ICCP; + info_ptr->valid |= PNG_INFO_iCCP; + } + + else + { + png_ptr->colorspace.flags |= + PNG_COLORSPACE_INVALID; + errmsg = "out of memory"; + } + } + + /* else the profile remains in the read + * buffer which gets reused for subsequent + * chunks. + */ + + if (info_ptr != NULL) + png_colorspace_sync(png_ptr, info_ptr); + + if (errmsg == NULL) + { + png_ptr->zowner = 0; + return; + } + } + + else if (size > 0) + errmsg = "truncated"; + + else + errmsg = png_ptr->zstream.msg; + } + + /* else png_icc_check_tag_table output an error */ + } + + else /* profile truncated */ + errmsg = png_ptr->zstream.msg; + } + + else + errmsg = "out of memory"; + } + + /* else png_icc_check_header output an error */ + } + + /* else png_icc_check_length output an error */ + } + + else /* profile truncated */ + errmsg = png_ptr->zstream.msg; + + /* Release the stream */ + png_ptr->zowner = 0; + } + + else /* png_inflate_claim failed */ + errmsg = png_ptr->zstream.msg; + } + + else + errmsg = "bad compression method"; /* or missing */ + } + + else + errmsg = "bad keyword"; + } + + else + errmsg = "too many profiles"; + + /* Failure: the reason is in 'errmsg' */ + if (finished == 0) + png_crc_finish(png_ptr, length); + + png_ptr->colorspace.flags |= PNG_COLORSPACE_INVALID; + png_colorspace_sync(png_ptr, info_ptr); + if (errmsg != NULL) /* else already output */ + png_chunk_benign_error(png_ptr, errmsg); +} +#endif /* READ_iCCP */ + +#ifdef PNG_READ_sPLT_SUPPORTED +void /* PRIVATE */ +png_handle_sPLT(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +/* Note: this does not properly handle chunks that are > 64K under DOS */ +{ + png_bytep entry_start, buffer; + png_sPLT_t new_palette; + png_sPLT_entryp pp; + png_uint_32 data_length; + int entry_size, i; + png_uint_32 skip = 0; + png_uint_32 dl; + png_size_t max_dl; + + png_debug(1, "in png_handle_sPLT"); + +#ifdef PNG_USER_LIMITS_SUPPORTED + if (png_ptr->user_chunk_cache_max != 0) + { + if (png_ptr->user_chunk_cache_max == 1) + { + png_crc_finish(png_ptr, length); + return; + } + + if (--png_ptr->user_chunk_cache_max == 1) + { + png_warning(png_ptr, "No space in chunk cache for sPLT"); + png_crc_finish(png_ptr, length); + return; + } + } +#endif + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if ((png_ptr->mode & PNG_HAVE_IDAT) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + +#ifdef PNG_MAX_MALLOC_64K + if (length > 65535U) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "too large to fit in memory"); + return; + } +#endif + + buffer = png_read_buffer(png_ptr, length+1, 2/*silent*/); + if (buffer == NULL) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of memory"); + return; + } + + + /* WARNING: this may break if size_t is less than 32 bits; it is assumed + * that the PNG_MAX_MALLOC_64K test is enabled in this case, but this is a + * potential breakage point if the types in pngconf.h aren't exactly right. + */ + png_crc_read(png_ptr, buffer, length); + + if (png_crc_finish(png_ptr, skip) != 0) + return; + + buffer[length] = 0; + + for (entry_start = buffer; *entry_start; entry_start++) + /* Empty loop to find end of name */ ; + + ++entry_start; + + /* A sample depth should follow the separator, and we should be on it */ + if (entry_start > buffer + length - 2) + { + png_warning(png_ptr, "malformed sPLT chunk"); + return; + } + + new_palette.depth = *entry_start++; + entry_size = (new_palette.depth == 8 ? 6 : 10); + /* This must fit in a png_uint_32 because it is derived from the original + * chunk data length. + */ + data_length = length - (png_uint_32)(entry_start - buffer); + + /* Integrity-check the data length */ + if ((data_length % entry_size) != 0) + { + png_warning(png_ptr, "sPLT chunk has bad length"); + return; + } + + dl = (png_int_32)(data_length / entry_size); + max_dl = PNG_SIZE_MAX / (sizeof (png_sPLT_entry)); + + if (dl > max_dl) + { + png_warning(png_ptr, "sPLT chunk too long"); + return; + } + + new_palette.nentries = (png_int_32)(data_length / entry_size); + + new_palette.entries = (png_sPLT_entryp)png_malloc_warn( + png_ptr, new_palette.nentries * (sizeof (png_sPLT_entry))); + + if (new_palette.entries == NULL) + { + png_warning(png_ptr, "sPLT chunk requires too much memory"); + return; + } + +#ifdef PNG_POINTER_INDEXING_SUPPORTED + for (i = 0; i < new_palette.nentries; i++) + { + pp = new_palette.entries + i; + + if (new_palette.depth == 8) + { + pp->red = *entry_start++; + pp->green = *entry_start++; + pp->blue = *entry_start++; + pp->alpha = *entry_start++; + } + + else + { + pp->red = png_get_uint_16(entry_start); entry_start += 2; + pp->green = png_get_uint_16(entry_start); entry_start += 2; + pp->blue = png_get_uint_16(entry_start); entry_start += 2; + pp->alpha = png_get_uint_16(entry_start); entry_start += 2; + } + + pp->frequency = png_get_uint_16(entry_start); entry_start += 2; + } +#else + pp = new_palette.entries; + + for (i = 0; i < new_palette.nentries; i++) + { + + if (new_palette.depth == 8) + { + pp[i].red = *entry_start++; + pp[i].green = *entry_start++; + pp[i].blue = *entry_start++; + pp[i].alpha = *entry_start++; + } + + else + { + pp[i].red = png_get_uint_16(entry_start); entry_start += 2; + pp[i].green = png_get_uint_16(entry_start); entry_start += 2; + pp[i].blue = png_get_uint_16(entry_start); entry_start += 2; + pp[i].alpha = png_get_uint_16(entry_start); entry_start += 2; + } + + pp[i].frequency = png_get_uint_16(entry_start); entry_start += 2; + } +#endif + + /* Discard all chunk data except the name and stash that */ + new_palette.name = (png_charp)buffer; + + png_set_sPLT(png_ptr, info_ptr, &new_palette, 1); + + png_free(png_ptr, new_palette.entries); +} +#endif /* READ_sPLT */ + +#ifdef PNG_READ_tRNS_SUPPORTED +void /* PRIVATE */ +png_handle_tRNS(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_byte readbuf[PNG_MAX_PALETTE_LENGTH]; + + png_debug(1, "in png_handle_tRNS"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if ((png_ptr->mode & PNG_HAVE_IDAT) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_tRNS) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "duplicate"); + return; + } + + if (png_ptr->color_type == PNG_COLOR_TYPE_GRAY) + { + png_byte buf[2]; + + if (length != 2) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + png_crc_read(png_ptr, buf, 2); + png_ptr->num_trans = 1; + png_ptr->trans_color.gray = png_get_uint_16(buf); + } + + else if (png_ptr->color_type == PNG_COLOR_TYPE_RGB) + { + png_byte buf[6]; + + if (length != 6) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + png_crc_read(png_ptr, buf, length); + png_ptr->num_trans = 1; + png_ptr->trans_color.red = png_get_uint_16(buf); + png_ptr->trans_color.green = png_get_uint_16(buf + 2); + png_ptr->trans_color.blue = png_get_uint_16(buf + 4); + } + + else if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE) + { + if ((png_ptr->mode & PNG_HAVE_PLTE) == 0) + { + /* TODO: is this actually an error in the ISO spec? */ + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + if (length > png_ptr->num_palette || length > PNG_MAX_PALETTE_LENGTH || + length == 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + png_crc_read(png_ptr, readbuf, length); + png_ptr->num_trans = (png_uint_16)length; + } + + else + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "invalid with alpha channel"); + return; + } + + if (png_crc_finish(png_ptr, 0) != 0) + { + png_ptr->num_trans = 0; + return; + } + + /* TODO: this is a horrible side effect in the palette case because the + * png_struct ends up with a pointer to the tRNS buffer owned by the + * png_info. Fix this. + */ + png_set_tRNS(png_ptr, info_ptr, readbuf, png_ptr->num_trans, + &(png_ptr->trans_color)); +} +#endif + +#ifdef PNG_READ_bKGD_SUPPORTED +void /* PRIVATE */ +png_handle_bKGD(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + unsigned int truelen; + png_byte buf[6]; + png_color_16 background; + + png_debug(1, "in png_handle_bKGD"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if ((png_ptr->mode & PNG_HAVE_IDAT) != 0 || + (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE && + (png_ptr->mode & PNG_HAVE_PLTE) == 0)) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_bKGD) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "duplicate"); + return; + } + + if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE) + truelen = 1; + + else if ((png_ptr->color_type & PNG_COLOR_MASK_COLOR) != 0) + truelen = 6; + + else + truelen = 2; + + if (length != truelen) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + png_crc_read(png_ptr, buf, truelen); + + if (png_crc_finish(png_ptr, 0) != 0) + return; + + /* We convert the index value into RGB components so that we can allow + * arbitrary RGB values for background when we have transparency, and + * so it is easy to determine the RGB values of the background color + * from the info_ptr struct. + */ + if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE) + { + background.index = buf[0]; + + if (info_ptr != NULL && info_ptr->num_palette != 0) + { + if (buf[0] >= info_ptr->num_palette) + { + png_chunk_benign_error(png_ptr, "invalid index"); + return; + } + + background.red = (png_uint_16)png_ptr->palette[buf[0]].red; + background.green = (png_uint_16)png_ptr->palette[buf[0]].green; + background.blue = (png_uint_16)png_ptr->palette[buf[0]].blue; + } + + else + background.red = background.green = background.blue = 0; + + background.gray = 0; + } + + else if ((png_ptr->color_type & PNG_COLOR_MASK_COLOR) == 0) /* GRAY */ + { + background.index = 0; + background.red = + background.green = + background.blue = + background.gray = png_get_uint_16(buf); + } + + else + { + background.index = 0; + background.red = png_get_uint_16(buf); + background.green = png_get_uint_16(buf + 2); + background.blue = png_get_uint_16(buf + 4); + background.gray = 0; + } + + png_set_bKGD(png_ptr, info_ptr, &background); +} +#endif + +#ifdef PNG_READ_hIST_SUPPORTED +void /* PRIVATE */ +png_handle_hIST(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + unsigned int num, i; + png_uint_16 readbuf[PNG_MAX_PALETTE_LENGTH]; + + png_debug(1, "in png_handle_hIST"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if ((png_ptr->mode & PNG_HAVE_IDAT) != 0 || + (png_ptr->mode & PNG_HAVE_PLTE) == 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_hIST) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "duplicate"); + return; + } + + num = length / 2 ; + + if (num != png_ptr->num_palette || num > PNG_MAX_PALETTE_LENGTH) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + for (i = 0; i < num; i++) + { + png_byte buf[2]; + + png_crc_read(png_ptr, buf, 2); + readbuf[i] = png_get_uint_16(buf); + } + + if (png_crc_finish(png_ptr, 0) != 0) + return; + + png_set_hIST(png_ptr, info_ptr, readbuf); +} +#endif + +#ifdef PNG_READ_pHYs_SUPPORTED +void /* PRIVATE */ +png_handle_pHYs(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_byte buf[9]; + png_uint_32 res_x, res_y; + int unit_type; + + png_debug(1, "in png_handle_pHYs"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if ((png_ptr->mode & PNG_HAVE_IDAT) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_pHYs) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "duplicate"); + return; + } + + if (length != 9) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + png_crc_read(png_ptr, buf, 9); + + if (png_crc_finish(png_ptr, 0) != 0) + return; + + res_x = png_get_uint_32(buf); + res_y = png_get_uint_32(buf + 4); + unit_type = buf[8]; + png_set_pHYs(png_ptr, info_ptr, res_x, res_y, unit_type); +} +#endif + +#ifdef PNG_READ_oFFs_SUPPORTED +void /* PRIVATE */ +png_handle_oFFs(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_byte buf[9]; + png_int_32 offset_x, offset_y; + int unit_type; + + png_debug(1, "in png_handle_oFFs"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if ((png_ptr->mode & PNG_HAVE_IDAT) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_oFFs) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "duplicate"); + return; + } + + if (length != 9) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + png_crc_read(png_ptr, buf, 9); + + if (png_crc_finish(png_ptr, 0) != 0) + return; + + offset_x = png_get_int_32(buf); + offset_y = png_get_int_32(buf + 4); + unit_type = buf[8]; + png_set_oFFs(png_ptr, info_ptr, offset_x, offset_y, unit_type); +} +#endif + +#ifdef PNG_READ_pCAL_SUPPORTED +/* Read the pCAL chunk (described in the PNG Extensions document) */ +void /* PRIVATE */ +png_handle_pCAL(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_int_32 X0, X1; + png_byte type, nparams; + png_bytep buffer, buf, units, endptr; + png_charpp params; + int i; + + png_debug(1, "in png_handle_pCAL"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if ((png_ptr->mode & PNG_HAVE_IDAT) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_pCAL) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "duplicate"); + return; + } + + png_debug1(2, "Allocating and reading pCAL chunk data (%u bytes)", + length + 1); + + buffer = png_read_buffer(png_ptr, length+1, 2/*silent*/); + + if (buffer == NULL) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of memory"); + return; + } + + png_crc_read(png_ptr, buffer, length); + + if (png_crc_finish(png_ptr, 0) != 0) + return; + + buffer[length] = 0; /* Null terminate the last string */ + + png_debug(3, "Finding end of pCAL purpose string"); + for (buf = buffer; *buf; buf++) + /* Empty loop */ ; + + endptr = buffer + length; + + /* We need to have at least 12 bytes after the purpose string + * in order to get the parameter information. + */ + if (endptr <= buf + 12) + { + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + png_debug(3, "Reading pCAL X0, X1, type, nparams, and units"); + X0 = png_get_int_32((png_bytep)buf+1); + X1 = png_get_int_32((png_bytep)buf+5); + type = buf[9]; + nparams = buf[10]; + units = buf + 11; + + png_debug(3, "Checking pCAL equation type and number of parameters"); + /* Check that we have the right number of parameters for known + * equation types. + */ + if ((type == PNG_EQUATION_LINEAR && nparams != 2) || + (type == PNG_EQUATION_BASE_E && nparams != 3) || + (type == PNG_EQUATION_ARBITRARY && nparams != 3) || + (type == PNG_EQUATION_HYPERBOLIC && nparams != 4)) + { + png_chunk_benign_error(png_ptr, "invalid parameter count"); + return; + } + + else if (type >= PNG_EQUATION_LAST) + { + png_chunk_benign_error(png_ptr, "unrecognized equation type"); + } + + for (buf = units; *buf; buf++) + /* Empty loop to move past the units string. */ ; + + png_debug(3, "Allocating pCAL parameters array"); + + params = png_voidcast(png_charpp, png_malloc_warn(png_ptr, + nparams * (sizeof (png_charp)))); + + if (params == NULL) + { + png_chunk_benign_error(png_ptr, "out of memory"); + return; + } + + /* Get pointers to the start of each parameter string. */ + for (i = 0; i < nparams; i++) + { + buf++; /* Skip the null string terminator from previous parameter. */ + + png_debug1(3, "Reading pCAL parameter %d", i); + + for (params[i] = (png_charp)buf; buf <= endptr && *buf != 0; buf++) + /* Empty loop to move past each parameter string */ ; + + /* Make sure we haven't run out of data yet */ + if (buf > endptr) + { + png_free(png_ptr, params); + png_chunk_benign_error(png_ptr, "invalid data"); + return; + } + } + + png_set_pCAL(png_ptr, info_ptr, (png_charp)buffer, X0, X1, type, nparams, + (png_charp)units, params); + + png_free(png_ptr, params); +} +#endif + +#ifdef PNG_READ_sCAL_SUPPORTED +/* Read the sCAL chunk */ +void /* PRIVATE */ +png_handle_sCAL(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_bytep buffer; + png_size_t i; + int state; + + png_debug(1, "in png_handle_sCAL"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if ((png_ptr->mode & PNG_HAVE_IDAT) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of place"); + return; + } + + else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_sCAL) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "duplicate"); + return; + } + + /* Need unit type, width, \0, height: minimum 4 bytes */ + else if (length < 4) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + png_debug1(2, "Allocating and reading sCAL chunk data (%u bytes)", + length + 1); + + buffer = png_read_buffer(png_ptr, length+1, 2/*silent*/); + + if (buffer == NULL) + { + png_chunk_benign_error(png_ptr, "out of memory"); + png_crc_finish(png_ptr, length); + return; + } + + png_crc_read(png_ptr, buffer, length); + buffer[length] = 0; /* Null terminate the last string */ + + if (png_crc_finish(png_ptr, 0) != 0) + return; + + /* Validate the unit. */ + if (buffer[0] != 1 && buffer[0] != 2) + { + png_chunk_benign_error(png_ptr, "invalid unit"); + return; + } + + /* Validate the ASCII numbers, need two ASCII numbers separated by + * a '\0' and they need to fit exactly in the chunk data. + */ + i = 1; + state = 0; + + if (png_check_fp_number((png_const_charp)buffer, length, &state, &i) == 0 || + i >= length || buffer[i++] != 0) + png_chunk_benign_error(png_ptr, "bad width format"); + + else if (PNG_FP_IS_POSITIVE(state) == 0) + png_chunk_benign_error(png_ptr, "non-positive width"); + + else + { + png_size_t heighti = i; + + state = 0; + if (png_check_fp_number((png_const_charp)buffer, length, + &state, &i) == 0 || i != length) + png_chunk_benign_error(png_ptr, "bad height format"); + + else if (PNG_FP_IS_POSITIVE(state) == 0) + png_chunk_benign_error(png_ptr, "non-positive height"); + + else + /* This is the (only) success case. */ + png_set_sCAL_s(png_ptr, info_ptr, buffer[0], + (png_charp)buffer+1, (png_charp)buffer+heighti); + } +} +#endif + +#ifdef PNG_READ_tIME_SUPPORTED +void /* PRIVATE */ +png_handle_tIME(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_byte buf[7]; + png_time mod_time; + + png_debug(1, "in png_handle_tIME"); + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_tIME) != 0) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "duplicate"); + return; + } + + if ((png_ptr->mode & PNG_HAVE_IDAT) != 0) + png_ptr->mode |= PNG_AFTER_IDAT; + + if (length != 7) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "invalid"); + return; + } + + png_crc_read(png_ptr, buf, 7); + + if (png_crc_finish(png_ptr, 0) != 0) + return; + + mod_time.second = buf[6]; + mod_time.minute = buf[5]; + mod_time.hour = buf[4]; + mod_time.day = buf[3]; + mod_time.month = buf[2]; + mod_time.year = png_get_uint_16(buf); + + png_set_tIME(png_ptr, info_ptr, &mod_time); +} +#endif + +#ifdef PNG_READ_tEXt_SUPPORTED +/* Note: this does not properly handle chunks that are > 64K under DOS */ +void /* PRIVATE */ +png_handle_tEXt(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_text text_info; + png_bytep buffer; + png_charp key; + png_charp text; + png_uint_32 skip = 0; + + png_debug(1, "in png_handle_tEXt"); + +#ifdef PNG_USER_LIMITS_SUPPORTED + if (png_ptr->user_chunk_cache_max != 0) + { + if (png_ptr->user_chunk_cache_max == 1) + { + png_crc_finish(png_ptr, length); + return; + } + + if (--png_ptr->user_chunk_cache_max == 1) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "no space in chunk cache"); + return; + } + } +#endif + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + if ((png_ptr->mode & PNG_HAVE_IDAT) != 0) + png_ptr->mode |= PNG_AFTER_IDAT; + +#ifdef PNG_MAX_MALLOC_64K + if (length > 65535U) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "too large to fit in memory"); + return; + } +#endif + + buffer = png_read_buffer(png_ptr, length+1, 1/*warn*/); + + if (buffer == NULL) + { + png_chunk_benign_error(png_ptr, "out of memory"); + return; + } + + png_crc_read(png_ptr, buffer, length); + + if (png_crc_finish(png_ptr, skip) != 0) + return; + + key = (png_charp)buffer; + key[length] = 0; + + for (text = key; *text; text++) + /* Empty loop to find end of key */ ; + + if (text != key + length) + text++; + + text_info.compression = PNG_TEXT_COMPRESSION_NONE; + text_info.key = key; + text_info.lang = NULL; + text_info.lang_key = NULL; + text_info.itxt_length = 0; + text_info.text = text; + text_info.text_length = strlen(text); + + if (png_set_text_2(png_ptr, info_ptr, &text_info, 1) != 0) + png_warning(png_ptr, "Insufficient memory to process text chunk"); +} +#endif + +#ifdef PNG_READ_zTXt_SUPPORTED +/* Note: this does not correctly handle chunks that are > 64K under DOS */ +void /* PRIVATE */ +png_handle_zTXt(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_const_charp errmsg = NULL; + png_bytep buffer; + png_uint_32 keyword_length; + + png_debug(1, "in png_handle_zTXt"); + +#ifdef PNG_USER_LIMITS_SUPPORTED + if (png_ptr->user_chunk_cache_max != 0) + { + if (png_ptr->user_chunk_cache_max == 1) + { + png_crc_finish(png_ptr, length); + return; + } + + if (--png_ptr->user_chunk_cache_max == 1) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "no space in chunk cache"); + return; + } + } +#endif + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + if ((png_ptr->mode & PNG_HAVE_IDAT) != 0) + png_ptr->mode |= PNG_AFTER_IDAT; + + buffer = png_read_buffer(png_ptr, length, 2/*silent*/); + + if (buffer == NULL) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of memory"); + return; + } + + png_crc_read(png_ptr, buffer, length); + + if (png_crc_finish(png_ptr, 0) != 0) + return; + + /* TODO: also check that the keyword contents match the spec! */ + for (keyword_length = 0; + keyword_length < length && buffer[keyword_length] != 0; + ++keyword_length) + /* Empty loop to find end of name */ ; + + if (keyword_length > 79 || keyword_length < 1) + errmsg = "bad keyword"; + + /* zTXt must have some LZ data after the keyword, although it may expand to + * zero bytes; we need a '\0' at the end of the keyword, the compression type + * then the LZ data: + */ + else if (keyword_length + 3 > length) + errmsg = "truncated"; + + else if (buffer[keyword_length+1] != PNG_COMPRESSION_TYPE_BASE) + errmsg = "unknown compression type"; + + else + { + png_alloc_size_t uncompressed_length = PNG_SIZE_MAX; + + /* TODO: at present png_decompress_chunk imposes a single application + * level memory limit, this should be split to different values for iCCP + * and text chunks. + */ + if (png_decompress_chunk(png_ptr, length, keyword_length+2, + &uncompressed_length, 1/*terminate*/) == Z_STREAM_END) + { + png_text text; + + /* It worked; png_ptr->read_buffer now looks like a tEXt chunk except + * for the extra compression type byte and the fact that it isn't + * necessarily '\0' terminated. + */ + buffer = png_ptr->read_buffer; + buffer[uncompressed_length+(keyword_length+2)] = 0; + + text.compression = PNG_TEXT_COMPRESSION_zTXt; + text.key = (png_charp)buffer; + text.text = (png_charp)(buffer + keyword_length+2); + text.text_length = uncompressed_length; + text.itxt_length = 0; + text.lang = NULL; + text.lang_key = NULL; + + if (png_set_text_2(png_ptr, info_ptr, &text, 1) != 0) + errmsg = "insufficient memory"; + } + + else + errmsg = png_ptr->zstream.msg; + } + + if (errmsg != NULL) + png_chunk_benign_error(png_ptr, errmsg); +} +#endif + +#ifdef PNG_READ_iTXt_SUPPORTED +/* Note: this does not correctly handle chunks that are > 64K under DOS */ +void /* PRIVATE */ +png_handle_iTXt(png_structrp png_ptr, png_inforp info_ptr, png_uint_32 length) +{ + png_const_charp errmsg = NULL; + png_bytep buffer; + png_uint_32 prefix_length; + + png_debug(1, "in png_handle_iTXt"); + +#ifdef PNG_USER_LIMITS_SUPPORTED + if (png_ptr->user_chunk_cache_max != 0) + { + if (png_ptr->user_chunk_cache_max == 1) + { + png_crc_finish(png_ptr, length); + return; + } + + if (--png_ptr->user_chunk_cache_max == 1) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "no space in chunk cache"); + return; + } + } +#endif + + if ((png_ptr->mode & PNG_HAVE_IHDR) == 0) + png_chunk_error(png_ptr, "missing IHDR"); + + if ((png_ptr->mode & PNG_HAVE_IDAT) != 0) + png_ptr->mode |= PNG_AFTER_IDAT; + + buffer = png_read_buffer(png_ptr, length+1, 1/*warn*/); + + if (buffer == NULL) + { + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "out of memory"); + return; + } + + png_crc_read(png_ptr, buffer, length); + + if (png_crc_finish(png_ptr, 0) != 0) + return; + + /* First the keyword. */ + for (prefix_length=0; + prefix_length < length && buffer[prefix_length] != 0; + ++prefix_length) + /* Empty loop */ ; + + /* Perform a basic check on the keyword length here. */ + if (prefix_length > 79 || prefix_length < 1) + errmsg = "bad keyword"; + + /* Expect keyword, compression flag, compression type, language, translated + * keyword (both may be empty but are 0 terminated) then the text, which may + * be empty. + */ + else if (prefix_length + 5 > length) + errmsg = "truncated"; + + else if (buffer[prefix_length+1] == 0 || + (buffer[prefix_length+1] == 1 && + buffer[prefix_length+2] == PNG_COMPRESSION_TYPE_BASE)) + { + int compressed = buffer[prefix_length+1] != 0; + png_uint_32 language_offset, translated_keyword_offset; + png_alloc_size_t uncompressed_length = 0; + + /* Now the language tag */ + prefix_length += 3; + language_offset = prefix_length; + + for (; prefix_length < length && buffer[prefix_length] != 0; + ++prefix_length) + /* Empty loop */ ; + + /* WARNING: the length may be invalid here, this is checked below. */ + translated_keyword_offset = ++prefix_length; + + for (; prefix_length < length && buffer[prefix_length] != 0; + ++prefix_length) + /* Empty loop */ ; + + /* prefix_length should now be at the trailing '\0' of the translated + * keyword, but it may already be over the end. None of this arithmetic + * can overflow because chunks are at most 2^31 bytes long, but on 16-bit + * systems the available allocation may overflow. + */ + ++prefix_length; + + if (compressed == 0 && prefix_length <= length) + uncompressed_length = length - prefix_length; + + else if (compressed != 0 && prefix_length < length) + { + uncompressed_length = PNG_SIZE_MAX; + + /* TODO: at present png_decompress_chunk imposes a single application + * level memory limit, this should be split to different values for + * iCCP and text chunks. + */ + if (png_decompress_chunk(png_ptr, length, prefix_length, + &uncompressed_length, 1/*terminate*/) == Z_STREAM_END) + buffer = png_ptr->read_buffer; + + else + errmsg = png_ptr->zstream.msg; + } + + else + errmsg = "truncated"; + + if (errmsg == NULL) + { + png_text text; + + buffer[uncompressed_length+prefix_length] = 0; + + if (compressed == 0) + text.compression = PNG_ITXT_COMPRESSION_NONE; + + else + text.compression = PNG_ITXT_COMPRESSION_zTXt; + + text.key = (png_charp)buffer; + text.lang = (png_charp)buffer + language_offset; + text.lang_key = (png_charp)buffer + translated_keyword_offset; + text.text = (png_charp)buffer + prefix_length; + text.text_length = 0; + text.itxt_length = uncompressed_length; + + if (png_set_text_2(png_ptr, info_ptr, &text, 1) != 0) + errmsg = "insufficient memory"; + } + } + + else + errmsg = "bad compression info"; + + if (errmsg != NULL) + png_chunk_benign_error(png_ptr, errmsg); +} +#endif + +#ifdef PNG_READ_UNKNOWN_CHUNKS_SUPPORTED +/* Utility function for png_handle_unknown; set up png_ptr::unknown_chunk */ +static int +png_cache_unknown_chunk(png_structrp png_ptr, png_uint_32 length) +{ + png_alloc_size_t limit = PNG_SIZE_MAX; + + if (png_ptr->unknown_chunk.data != NULL) + { + png_free(png_ptr, png_ptr->unknown_chunk.data); + png_ptr->unknown_chunk.data = NULL; + } + +# ifdef PNG_SET_CHUNK_MALLOC_LIMIT_SUPPORTED + if (png_ptr->user_chunk_malloc_max > 0 && + png_ptr->user_chunk_malloc_max < limit) + limit = png_ptr->user_chunk_malloc_max; + +# elif PNG_USER_CHUNK_MALLOC_MAX > 0 + if (PNG_USER_CHUNK_MALLOC_MAX < limit) + limit = PNG_USER_CHUNK_MALLOC_MAX; +# endif + + if (length <= limit) + { + PNG_CSTRING_FROM_CHUNK(png_ptr->unknown_chunk.name, png_ptr->chunk_name); + /* The following is safe because of the PNG_SIZE_MAX init above */ + png_ptr->unknown_chunk.size = (png_size_t)length/*SAFE*/; + /* 'mode' is a flag array, only the bottom four bits matter here */ + png_ptr->unknown_chunk.location = (png_byte)png_ptr->mode/*SAFE*/; + + if (length == 0) + png_ptr->unknown_chunk.data = NULL; + + else + { + /* Do a 'warn' here - it is handled below. */ + png_ptr->unknown_chunk.data = png_voidcast(png_bytep, + png_malloc_warn(png_ptr, length)); + } + } + + if (png_ptr->unknown_chunk.data == NULL && length > 0) + { + /* This is benign because we clean up correctly */ + png_crc_finish(png_ptr, length); + png_chunk_benign_error(png_ptr, "unknown chunk exceeds memory limits"); + return 0; + } + + else + { + if (length > 0) + png_crc_read(png_ptr, png_ptr->unknown_chunk.data, length); + png_crc_finish(png_ptr, 0); + return 1; + } +} +#endif /* READ_UNKNOWN_CHUNKS */ + +/* Handle an unknown, or known but disabled, chunk */ +void /* PRIVATE */ +png_handle_unknown(png_structrp png_ptr, png_inforp info_ptr, + png_uint_32 length, int keep) +{ + int handled = 0; /* the chunk was handled */ + + png_debug(1, "in png_handle_unknown"); + +#ifdef PNG_READ_UNKNOWN_CHUNKS_SUPPORTED + /* NOTE: this code is based on the code in libpng-1.4.12 except for fixing + * the bug which meant that setting a non-default behavior for a specific + * chunk would be ignored (the default was always used unless a user + * callback was installed). + * + * 'keep' is the value from the png_chunk_unknown_handling, the setting for + * this specific chunk_name, if PNG_HANDLE_AS_UNKNOWN_SUPPORTED, if not it + * will always be PNG_HANDLE_CHUNK_AS_DEFAULT and it needs to be set here. + * This is just an optimization to avoid multiple calls to the lookup + * function. + */ +# ifndef PNG_HANDLE_AS_UNKNOWN_SUPPORTED +# ifdef PNG_SET_UNKNOWN_CHUNKS_SUPPORTED + keep = png_chunk_unknown_handling(png_ptr, png_ptr->chunk_name); +# endif +# endif + + /* One of the following methods will read the chunk or skip it (at least one + * of these is always defined because this is the only way to switch on + * PNG_READ_UNKNOWN_CHUNKS_SUPPORTED) + */ +# ifdef PNG_READ_USER_CHUNKS_SUPPORTED + /* The user callback takes precedence over the chunk keep value, but the + * keep value is still required to validate a save of a critical chunk. + */ + if (png_ptr->read_user_chunk_fn != NULL) + { + if (png_cache_unknown_chunk(png_ptr, length) != 0) + { + /* Callback to user unknown chunk handler */ + int ret = (*(png_ptr->read_user_chunk_fn))(png_ptr, + &png_ptr->unknown_chunk); + + /* ret is: + * negative: An error occured, png_chunk_error will be called. + * zero: The chunk was not handled, the chunk will be discarded + * unless png_set_keep_unknown_chunks has been used to set + * a 'keep' behavior for this particular chunk, in which + * case that will be used. A critical chunk will cause an + * error at this point unless it is to be saved. + * positive: The chunk was handled, libpng will ignore/discard it. + */ + if (ret < 0) + png_chunk_error(png_ptr, "error in user chunk"); + + else if (ret == 0) + { + /* If the keep value is 'default' or 'never' override it, but + * still error out on critical chunks unless the keep value is + * 'always' While this is weird it is the behavior in 1.4.12. + * A possible improvement would be to obey the value set for the + * chunk, but this would be an API change that would probably + * damage some applications. + * + * The png_app_warning below catches the case that matters, where + * the application has not set specific save or ignore for this + * chunk or global save or ignore. + */ + if (keep < PNG_HANDLE_CHUNK_IF_SAFE) + { +# ifdef PNG_SET_UNKNOWN_CHUNKS_SUPPORTED + if (png_ptr->unknown_default < PNG_HANDLE_CHUNK_IF_SAFE) + { + png_chunk_warning(png_ptr, "Saving unknown chunk:"); + png_app_warning(png_ptr, + "forcing save of an unhandled chunk;" + " please call png_set_keep_unknown_chunks"); + /* with keep = PNG_HANDLE_CHUNK_IF_SAFE */ + } +# endif + keep = PNG_HANDLE_CHUNK_IF_SAFE; + } + } + + else /* chunk was handled */ + { + handled = 1; + /* Critical chunks can be safely discarded at this point. */ + keep = PNG_HANDLE_CHUNK_NEVER; + } + } + + else + keep = PNG_HANDLE_CHUNK_NEVER; /* insufficient memory */ + } + + else + /* Use the SAVE_UNKNOWN_CHUNKS code or skip the chunk */ +# endif /* READ_USER_CHUNKS */ + +# ifdef PNG_SAVE_UNKNOWN_CHUNKS_SUPPORTED + { + /* keep is currently just the per-chunk setting, if there was no + * setting change it to the global default now (not that this may + * still be AS_DEFAULT) then obtain the cache of the chunk if required, + * if not simply skip the chunk. + */ + if (keep == PNG_HANDLE_CHUNK_AS_DEFAULT) + keep = png_ptr->unknown_default; + + if (keep == PNG_HANDLE_CHUNK_ALWAYS || + (keep == PNG_HANDLE_CHUNK_IF_SAFE && + PNG_CHUNK_ANCILLARY(png_ptr->chunk_name))) + { + if (png_cache_unknown_chunk(png_ptr, length) == 0) + keep = PNG_HANDLE_CHUNK_NEVER; + } + + else + png_crc_finish(png_ptr, length); + } +# else +# ifndef PNG_READ_USER_CHUNKS_SUPPORTED +# error no method to support READ_UNKNOWN_CHUNKS +# endif + + { + /* If here there is no read callback pointer set and no support is + * compiled in to just save the unknown chunks, so simply skip this + * chunk. If 'keep' is something other than AS_DEFAULT or NEVER then + * the app has erroneously asked for unknown chunk saving when there + * is no support. + */ + if (keep > PNG_HANDLE_CHUNK_NEVER) + png_app_error(png_ptr, "no unknown chunk support available"); + + png_crc_finish(png_ptr, length); + } +# endif + +# ifdef PNG_STORE_UNKNOWN_CHUNKS_SUPPORTED + /* Now store the chunk in the chunk list if appropriate, and if the limits + * permit it. + */ + if (keep == PNG_HANDLE_CHUNK_ALWAYS || + (keep == PNG_HANDLE_CHUNK_IF_SAFE && + PNG_CHUNK_ANCILLARY(png_ptr->chunk_name))) + { +# ifdef PNG_USER_LIMITS_SUPPORTED + switch (png_ptr->user_chunk_cache_max) + { + case 2: + png_ptr->user_chunk_cache_max = 1; + png_chunk_benign_error(png_ptr, "no space in chunk cache"); + /* FALL THROUGH */ + case 1: + /* NOTE: prior to 1.6.0 this case resulted in an unknown critical + * chunk being skipped, now there will be a hard error below. + */ + break; + + default: /* not at limit */ + --(png_ptr->user_chunk_cache_max); + /* FALL THROUGH */ + case 0: /* no limit */ +# endif /* USER_LIMITS */ + /* Here when the limit isn't reached or when limits are compiled + * out; store the chunk. + */ + png_set_unknown_chunks(png_ptr, info_ptr, + &png_ptr->unknown_chunk, 1); + handled = 1; +# ifdef PNG_USER_LIMITS_SUPPORTED + break; + } +# endif + } +# else /* no store support: the chunk must be handled by the user callback */ + PNG_UNUSED(info_ptr) +# endif + + /* Regardless of the error handling below the cached data (if any) can be + * freed now. Notice that the data is not freed if there is a png_error, but + * it will be freed by destroy_read_struct. + */ + if (png_ptr->unknown_chunk.data != NULL) + png_free(png_ptr, png_ptr->unknown_chunk.data); + png_ptr->unknown_chunk.data = NULL; + +#else /* !PNG_READ_UNKNOWN_CHUNKS_SUPPORTED */ + /* There is no support to read an unknown chunk, so just skip it. */ + png_crc_finish(png_ptr, length); + PNG_UNUSED(info_ptr) + PNG_UNUSED(keep) +#endif /* !READ_UNKNOWN_CHUNKS */ + + /* Check for unhandled critical chunks */ + if (handled == 0 && PNG_CHUNK_CRITICAL(png_ptr->chunk_name)) + png_chunk_error(png_ptr, "unhandled critical chunk"); +} + +/* This function is called to verify that a chunk name is valid. + * This function can't have the "critical chunk check" incorporated + * into it, since in the future we will need to be able to call user + * functions to handle unknown critical chunks after we check that + * the chunk name itself is valid. + */ + +/* Bit hacking: the test for an invalid byte in the 4 byte chunk name is: + * + * ((c) < 65 || (c) > 122 || ((c) > 90 && (c) < 97)) + */ + +void /* PRIVATE */ +png_check_chunk_name(png_structrp png_ptr, png_uint_32 chunk_name) +{ + int i; + + png_debug(1, "in png_check_chunk_name"); + + for (i=1; i<=4; ++i) + { + int c = chunk_name & 0xff; + + if (c < 65 || c > 122 || (c > 90 && c < 97)) + png_chunk_error(png_ptr, "invalid chunk type"); + + chunk_name >>= 8; + } +} + +/* Combines the row recently read in with the existing pixels in the row. This + * routine takes care of alpha and transparency if requested. This routine also + * handles the two methods of progressive display of interlaced images, + * depending on the 'display' value; if 'display' is true then the whole row + * (dp) is filled from the start by replicating the available pixels. If + * 'display' is false only those pixels present in the pass are filled in. + */ +void /* PRIVATE */ +png_combine_row(png_const_structrp png_ptr, png_bytep dp, int display) +{ + unsigned int pixel_depth = png_ptr->transformed_pixel_depth; + png_const_bytep sp = png_ptr->row_buf + 1; + png_alloc_size_t row_width = png_ptr->width; + unsigned int pass = png_ptr->pass; + png_bytep end_ptr = 0; + png_byte end_byte = 0; + unsigned int end_mask; + + png_debug(1, "in png_combine_row"); + + /* Added in 1.5.6: it should not be possible to enter this routine until at + * least one row has been read from the PNG data and transformed. + */ + if (pixel_depth == 0) + png_error(png_ptr, "internal row logic error"); + + /* Added in 1.5.4: the pixel depth should match the information returned by + * any call to png_read_update_info at this point. Do not continue if we got + * this wrong. + */ + if (png_ptr->info_rowbytes != 0 && png_ptr->info_rowbytes != + PNG_ROWBYTES(pixel_depth, row_width)) + png_error(png_ptr, "internal row size calculation error"); + + /* Don't expect this to ever happen: */ + if (row_width == 0) + png_error(png_ptr, "internal row width error"); + + /* Preserve the last byte in cases where only part of it will be overwritten, + * the multiply below may overflow, we don't care because ANSI-C guarantees + * we get the low bits. + */ + end_mask = (pixel_depth * row_width) & 7; + if (end_mask != 0) + { + /* end_ptr == NULL is a flag to say do nothing */ + end_ptr = dp + PNG_ROWBYTES(pixel_depth, row_width) - 1; + end_byte = *end_ptr; +# ifdef PNG_READ_PACKSWAP_SUPPORTED + if ((png_ptr->transformations & PNG_PACKSWAP) != 0) + /* little-endian byte */ + end_mask = 0xff << end_mask; + + else /* big-endian byte */ +# endif + end_mask = 0xff >> end_mask; + /* end_mask is now the bits to *keep* from the destination row */ + } + + /* For non-interlaced images this reduces to a memcpy(). A memcpy() + * will also happen if interlacing isn't supported or if the application + * does not call png_set_interlace_handling(). In the latter cases the + * caller just gets a sequence of the unexpanded rows from each interlace + * pass. + */ +#ifdef PNG_READ_INTERLACING_SUPPORTED + if (png_ptr->interlaced != 0 && + (png_ptr->transformations & PNG_INTERLACE) != 0 && + pass < 6 && (display == 0 || + /* The following copies everything for 'display' on passes 0, 2 and 4. */ + (display == 1 && (pass & 1) != 0))) + { + /* Narrow images may have no bits in a pass; the caller should handle + * this, but this test is cheap: + */ + if (row_width <= PNG_PASS_START_COL(pass)) + return; + + if (pixel_depth < 8) + { + /* For pixel depths up to 4 bpp the 8-pixel mask can be expanded to fit + * into 32 bits, then a single loop over the bytes using the four byte + * values in the 32-bit mask can be used. For the 'display' option the + * expanded mask may also not require any masking within a byte. To + * make this work the PACKSWAP option must be taken into account - it + * simply requires the pixels to be reversed in each byte. + * + * The 'regular' case requires a mask for each of the first 6 passes, + * the 'display' case does a copy for the even passes in the range + * 0..6. This has already been handled in the test above. + * + * The masks are arranged as four bytes with the first byte to use in + * the lowest bits (little-endian) regardless of the order (PACKSWAP or + * not) of the pixels in each byte. + * + * NOTE: the whole of this logic depends on the caller of this function + * only calling it on rows appropriate to the pass. This function only + * understands the 'x' logic; the 'y' logic is handled by the caller. + * + * The following defines allow generation of compile time constant bit + * masks for each pixel depth and each possibility of swapped or not + * swapped bytes. Pass 'p' is in the range 0..6; 'x', a pixel index, + * is in the range 0..7; and the result is 1 if the pixel is to be + * copied in the pass, 0 if not. 'S' is for the sparkle method, 'B' + * for the block method. + * + * With some compilers a compile time expression of the general form: + * + * (shift >= 32) ? (a >> (shift-32)) : (b >> shift) + * + * Produces warnings with values of 'shift' in the range 33 to 63 + * because the right hand side of the ?: expression is evaluated by + * the compiler even though it isn't used. Microsoft Visual C (various + * versions) and the Intel C compiler are known to do this. To avoid + * this the following macros are used in 1.5.6. This is a temporary + * solution to avoid destabilizing the code during the release process. + */ +# if PNG_USE_COMPILE_TIME_MASKS +# define PNG_LSR(x,s) ((x)>>((s) & 0x1f)) +# define PNG_LSL(x,s) ((x)<<((s) & 0x1f)) +# else +# define PNG_LSR(x,s) ((x)>>(s)) +# define PNG_LSL(x,s) ((x)<<(s)) +# endif +# define S_COPY(p,x) (((p)<4 ? PNG_LSR(0x80088822,(3-(p))*8+(7-(x))) :\ + PNG_LSR(0xaa55ff00,(7-(p))*8+(7-(x)))) & 1) +# define B_COPY(p,x) (((p)<4 ? PNG_LSR(0xff0fff33,(3-(p))*8+(7-(x))) :\ + PNG_LSR(0xff55ff00,(7-(p))*8+(7-(x)))) & 1) + + /* Return a mask for pass 'p' pixel 'x' at depth 'd'. The mask is + * little endian - the first pixel is at bit 0 - however the extra + * parameter 's' can be set to cause the mask position to be swapped + * within each byte, to match the PNG format. This is done by XOR of + * the shift with 7, 6 or 4 for bit depths 1, 2 and 4. + */ +# define PIXEL_MASK(p,x,d,s) \ + (PNG_LSL(((PNG_LSL(1U,(d)))-1),(((x)*(d))^((s)?8-(d):0)))) + + /* Hence generate the appropriate 'block' or 'sparkle' pixel copy mask. + */ +# define S_MASKx(p,x,d,s) (S_COPY(p,x)?PIXEL_MASK(p,x,d,s):0) +# define B_MASKx(p,x,d,s) (B_COPY(p,x)?PIXEL_MASK(p,x,d,s):0) + + /* Combine 8 of these to get the full mask. For the 1-bpp and 2-bpp + * cases the result needs replicating, for the 4-bpp case the above + * generates a full 32 bits. + */ +# define MASK_EXPAND(m,d) ((m)*((d)==1?0x01010101:((d)==2?0x00010001:1))) + +# define S_MASK(p,d,s) MASK_EXPAND(S_MASKx(p,0,d,s) + S_MASKx(p,1,d,s) +\ + S_MASKx(p,2,d,s) + S_MASKx(p,3,d,s) + S_MASKx(p,4,d,s) +\ + S_MASKx(p,5,d,s) + S_MASKx(p,6,d,s) + S_MASKx(p,7,d,s), d) + +# define B_MASK(p,d,s) MASK_EXPAND(B_MASKx(p,0,d,s) + B_MASKx(p,1,d,s) +\ + B_MASKx(p,2,d,s) + B_MASKx(p,3,d,s) + B_MASKx(p,4,d,s) +\ + B_MASKx(p,5,d,s) + B_MASKx(p,6,d,s) + B_MASKx(p,7,d,s), d) + +#if PNG_USE_COMPILE_TIME_MASKS + /* Utility macros to construct all the masks for a depth/swap + * combination. The 's' parameter says whether the format is PNG + * (big endian bytes) or not. Only the three odd-numbered passes are + * required for the display/block algorithm. + */ +# define S_MASKS(d,s) { S_MASK(0,d,s), S_MASK(1,d,s), S_MASK(2,d,s),\ + S_MASK(3,d,s), S_MASK(4,d,s), S_MASK(5,d,s) } + +# define B_MASKS(d,s) { B_MASK(1,d,s), B_MASK(3,d,s), B_MASK(5,d,s) } + +# define DEPTH_INDEX(d) ((d)==1?0:((d)==2?1:2)) + + /* Hence the pre-compiled masks indexed by PACKSWAP (or not), depth and + * then pass: + */ + static PNG_CONST png_uint_32 row_mask[2/*PACKSWAP*/][3/*depth*/][6] = + { + /* Little-endian byte masks for PACKSWAP */ + { S_MASKS(1,0), S_MASKS(2,0), S_MASKS(4,0) }, + /* Normal (big-endian byte) masks - PNG format */ + { S_MASKS(1,1), S_MASKS(2,1), S_MASKS(4,1) } + }; + + /* display_mask has only three entries for the odd passes, so index by + * pass>>1. + */ + static PNG_CONST png_uint_32 display_mask[2][3][3] = + { + /* Little-endian byte masks for PACKSWAP */ + { B_MASKS(1,0), B_MASKS(2,0), B_MASKS(4,0) }, + /* Normal (big-endian byte) masks - PNG format */ + { B_MASKS(1,1), B_MASKS(2,1), B_MASKS(4,1) } + }; + +# define MASK(pass,depth,display,png)\ + ((display)?display_mask[png][DEPTH_INDEX(depth)][pass>>1]:\ + row_mask[png][DEPTH_INDEX(depth)][pass]) + +#else /* !PNG_USE_COMPILE_TIME_MASKS */ + /* This is the runtime alternative: it seems unlikely that this will + * ever be either smaller or faster than the compile time approach. + */ +# define MASK(pass,depth,display,png)\ + ((display)?B_MASK(pass,depth,png):S_MASK(pass,depth,png)) +#endif /* !USE_COMPILE_TIME_MASKS */ + + /* Use the appropriate mask to copy the required bits. In some cases + * the byte mask will be 0 or 0xff; optimize these cases. row_width is + * the number of pixels, but the code copies bytes, so it is necessary + * to special case the end. + */ + png_uint_32 pixels_per_byte = 8 / pixel_depth; + png_uint_32 mask; + +# ifdef PNG_READ_PACKSWAP_SUPPORTED + if ((png_ptr->transformations & PNG_PACKSWAP) != 0) + mask = MASK(pass, pixel_depth, display, 0); + + else +# endif + mask = MASK(pass, pixel_depth, display, 1); + + for (;;) + { + png_uint_32 m; + + /* It doesn't matter in the following if png_uint_32 has more than + * 32 bits because the high bits always match those in m<<24; it is, + * however, essential to use OR here, not +, because of this. + */ + m = mask; + mask = (m >> 8) | (m << 24); /* rotate right to good compilers */ + m &= 0xff; + + if (m != 0) /* something to copy */ + { + if (m != 0xff) + *dp = (png_byte)((*dp & ~m) | (*sp & m)); + else + *dp = *sp; + } + + /* NOTE: this may overwrite the last byte with garbage if the image + * is not an exact number of bytes wide; libpng has always done + * this. + */ + if (row_width <= pixels_per_byte) + break; /* May need to restore part of the last byte */ + + row_width -= pixels_per_byte; + ++dp; + ++sp; + } + } + + else /* pixel_depth >= 8 */ + { + unsigned int bytes_to_copy, bytes_to_jump; + + /* Validate the depth - it must be a multiple of 8 */ + if (pixel_depth & 7) + png_error(png_ptr, "invalid user transform pixel depth"); + + pixel_depth >>= 3; /* now in bytes */ + row_width *= pixel_depth; + + /* Regardless of pass number the Adam 7 interlace always results in a + * fixed number of pixels to copy then to skip. There may be a + * different number of pixels to skip at the start though. + */ + { + unsigned int offset = PNG_PASS_START_COL(pass) * pixel_depth; + + row_width -= offset; + dp += offset; + sp += offset; + } + + /* Work out the bytes to copy. */ + if (display != 0) + { + /* When doing the 'block' algorithm the pixel in the pass gets + * replicated to adjacent pixels. This is why the even (0,2,4,6) + * passes are skipped above - the entire expanded row is copied. + */ + bytes_to_copy = (1<<((6-pass)>>1)) * pixel_depth; + + /* But don't allow this number to exceed the actual row width. */ + if (bytes_to_copy > row_width) + bytes_to_copy = (unsigned int)/*SAFE*/row_width; + } + + else /* normal row; Adam7 only ever gives us one pixel to copy. */ + bytes_to_copy = pixel_depth; + + /* In Adam7 there is a constant offset between where the pixels go. */ + bytes_to_jump = PNG_PASS_COL_OFFSET(pass) * pixel_depth; + + /* And simply copy these bytes. Some optimization is possible here, + * depending on the value of 'bytes_to_copy'. Special case the low + * byte counts, which we know to be frequent. + * + * Notice that these cases all 'return' rather than 'break' - this + * avoids an unnecessary test on whether to restore the last byte + * below. + */ + switch (bytes_to_copy) + { + case 1: + for (;;) + { + *dp = *sp; + + if (row_width <= bytes_to_jump) + return; + + dp += bytes_to_jump; + sp += bytes_to_jump; + row_width -= bytes_to_jump; + } + + case 2: + /* There is a possibility of a partial copy at the end here; this + * slows the code down somewhat. + */ + do + { + dp[0] = sp[0], dp[1] = sp[1]; + + if (row_width <= bytes_to_jump) + return; + + sp += bytes_to_jump; + dp += bytes_to_jump; + row_width -= bytes_to_jump; + } + while (row_width > 1); + + /* And there can only be one byte left at this point: */ + *dp = *sp; + return; + + case 3: + /* This can only be the RGB case, so each copy is exactly one + * pixel and it is not necessary to check for a partial copy. + */ + for (;;) + { + dp[0] = sp[0], dp[1] = sp[1], dp[2] = sp[2]; + + if (row_width <= bytes_to_jump) + return; + + sp += bytes_to_jump; + dp += bytes_to_jump; + row_width -= bytes_to_jump; + } + + default: +#if PNG_ALIGN_TYPE != PNG_ALIGN_NONE + /* Check for double byte alignment and, if possible, use a + * 16-bit copy. Don't attempt this for narrow images - ones that + * are less than an interlace panel wide. Don't attempt it for + * wide bytes_to_copy either - use the memcpy there. + */ + if (bytes_to_copy < 16 /*else use memcpy*/ && + png_isaligned(dp, png_uint_16) && + png_isaligned(sp, png_uint_16) && + bytes_to_copy % (sizeof (png_uint_16)) == 0 && + bytes_to_jump % (sizeof (png_uint_16)) == 0) + { + /* Everything is aligned for png_uint_16 copies, but try for + * png_uint_32 first. + */ + if (png_isaligned(dp, png_uint_32) != 0 && + png_isaligned(sp, png_uint_32) != 0 && + bytes_to_copy % (sizeof (png_uint_32)) == 0 && + bytes_to_jump % (sizeof (png_uint_32)) == 0) + { + png_uint_32p dp32 = png_aligncast(png_uint_32p,dp); + png_const_uint_32p sp32 = png_aligncastconst( + png_const_uint_32p, sp); + size_t skip = (bytes_to_jump-bytes_to_copy) / + (sizeof (png_uint_32)); + + do + { + size_t c = bytes_to_copy; + do + { + *dp32++ = *sp32++; + c -= (sizeof (png_uint_32)); + } + while (c > 0); + + if (row_width <= bytes_to_jump) + return; + + dp32 += skip; + sp32 += skip; + row_width -= bytes_to_jump; + } + while (bytes_to_copy <= row_width); + + /* Get to here when the row_width truncates the final copy. + * There will be 1-3 bytes left to copy, so don't try the + * 16-bit loop below. + */ + dp = (png_bytep)dp32; + sp = (png_const_bytep)sp32; + do + *dp++ = *sp++; + while (--row_width > 0); + return; + } + + /* Else do it in 16-bit quantities, but only if the size is + * not too large. + */ + else + { + png_uint_16p dp16 = png_aligncast(png_uint_16p, dp); + png_const_uint_16p sp16 = png_aligncastconst( + png_const_uint_16p, sp); + size_t skip = (bytes_to_jump-bytes_to_copy) / + (sizeof (png_uint_16)); + + do + { + size_t c = bytes_to_copy; + do + { + *dp16++ = *sp16++; + c -= (sizeof (png_uint_16)); + } + while (c > 0); + + if (row_width <= bytes_to_jump) + return; + + dp16 += skip; + sp16 += skip; + row_width -= bytes_to_jump; + } + while (bytes_to_copy <= row_width); + + /* End of row - 1 byte left, bytes_to_copy > row_width: */ + dp = (png_bytep)dp16; + sp = (png_const_bytep)sp16; + do + *dp++ = *sp++; + while (--row_width > 0); + return; + } + } +#endif /* ALIGN_TYPE code */ + + /* The true default - use a memcpy: */ + for (;;) + { + memcpy(dp, sp, bytes_to_copy); + + if (row_width <= bytes_to_jump) + return; + + sp += bytes_to_jump; + dp += bytes_to_jump; + row_width -= bytes_to_jump; + if (bytes_to_copy > row_width) + bytes_to_copy = (unsigned int)/*SAFE*/row_width; + } + } + + /* NOT REACHED*/ + } /* pixel_depth >= 8 */ + + /* Here if pixel_depth < 8 to check 'end_ptr' below. */ + } + else +#endif /* READ_INTERLACING */ + + /* If here then the switch above wasn't used so just memcpy the whole row + * from the temporary row buffer (notice that this overwrites the end of the + * destination row if it is a partial byte.) + */ + memcpy(dp, sp, PNG_ROWBYTES(pixel_depth, row_width)); + + /* Restore the overwritten bits from the last byte if necessary. */ + if (end_ptr != NULL) + *end_ptr = (png_byte)((end_byte & end_mask) | (*end_ptr & ~end_mask)); +} + +#ifdef PNG_READ_INTERLACING_SUPPORTED +void /* PRIVATE */ +png_do_read_interlace(png_row_infop row_info, png_bytep row, int pass, + png_uint_32 transformations /* Because these may affect the byte layout */) +{ + /* Arrays to facilitate easy interlacing - use pass (0 - 6) as index */ + /* Offset to next interlace block */ + static PNG_CONST int png_pass_inc[7] = {8, 8, 4, 4, 2, 2, 1}; + + png_debug(1, "in png_do_read_interlace"); + if (row != NULL && row_info != NULL) + { + png_uint_32 final_width; + + final_width = row_info->width * png_pass_inc[pass]; + + switch (row_info->pixel_depth) + { + case 1: + { + png_bytep sp = row + (png_size_t)((row_info->width - 1) >> 3); + png_bytep dp = row + (png_size_t)((final_width - 1) >> 3); + int sshift, dshift; + int s_start, s_end, s_inc; + int jstop = png_pass_inc[pass]; + png_byte v; + png_uint_32 i; + int j; + +#ifdef PNG_READ_PACKSWAP_SUPPORTED + if ((transformations & PNG_PACKSWAP) != 0) + { + sshift = (int)((row_info->width + 7) & 0x07); + dshift = (int)((final_width + 7) & 0x07); + s_start = 7; + s_end = 0; + s_inc = -1; + } + + else +#endif + { + sshift = 7 - (int)((row_info->width + 7) & 0x07); + dshift = 7 - (int)((final_width + 7) & 0x07); + s_start = 0; + s_end = 7; + s_inc = 1; + } + + for (i = 0; i < row_info->width; i++) + { + v = (png_byte)((*sp >> sshift) & 0x01); + for (j = 0; j < jstop; j++) + { + unsigned int tmp = *dp & (0x7f7f >> (7 - dshift)); + tmp |= v << dshift; + *dp = (png_byte)(tmp & 0xff); + + if (dshift == s_end) + { + dshift = s_start; + dp--; + } + + else + dshift += s_inc; + } + + if (sshift == s_end) + { + sshift = s_start; + sp--; + } + + else + sshift += s_inc; + } + break; + } + + case 2: + { + png_bytep sp = row + (png_uint_32)((row_info->width - 1) >> 2); + png_bytep dp = row + (png_uint_32)((final_width - 1) >> 2); + int sshift, dshift; + int s_start, s_end, s_inc; + int jstop = png_pass_inc[pass]; + png_uint_32 i; + +#ifdef PNG_READ_PACKSWAP_SUPPORTED + if ((transformations & PNG_PACKSWAP) != 0) + { + sshift = (int)(((row_info->width + 3) & 0x03) << 1); + dshift = (int)(((final_width + 3) & 0x03) << 1); + s_start = 6; + s_end = 0; + s_inc = -2; + } + + else +#endif + { + sshift = (int)((3 - ((row_info->width + 3) & 0x03)) << 1); + dshift = (int)((3 - ((final_width + 3) & 0x03)) << 1); + s_start = 0; + s_end = 6; + s_inc = 2; + } + + for (i = 0; i < row_info->width; i++) + { + png_byte v; + int j; + + v = (png_byte)((*sp >> sshift) & 0x03); + for (j = 0; j < jstop; j++) + { + unsigned int tmp = *dp & (0x3f3f >> (6 - dshift)); + tmp |= v << dshift; + *dp = (png_byte)(tmp & 0xff); + + if (dshift == s_end) + { + dshift = s_start; + dp--; + } + + else + dshift += s_inc; + } + + if (sshift == s_end) + { + sshift = s_start; + sp--; + } + + else + sshift += s_inc; + } + break; + } + + case 4: + { + png_bytep sp = row + (png_size_t)((row_info->width - 1) >> 1); + png_bytep dp = row + (png_size_t)((final_width - 1) >> 1); + int sshift, dshift; + int s_start, s_end, s_inc; + png_uint_32 i; + int jstop = png_pass_inc[pass]; + +#ifdef PNG_READ_PACKSWAP_SUPPORTED + if ((transformations & PNG_PACKSWAP) != 0) + { + sshift = (int)(((row_info->width + 1) & 0x01) << 2); + dshift = (int)(((final_width + 1) & 0x01) << 2); + s_start = 4; + s_end = 0; + s_inc = -4; + } + + else +#endif + { + sshift = (int)((1 - ((row_info->width + 1) & 0x01)) << 2); + dshift = (int)((1 - ((final_width + 1) & 0x01)) << 2); + s_start = 0; + s_end = 4; + s_inc = 4; + } + + for (i = 0; i < row_info->width; i++) + { + png_byte v = (png_byte)((*sp >> sshift) & 0x0f); + int j; + + for (j = 0; j < jstop; j++) + { + unsigned int tmp = *dp & (0xf0f >> (4 - dshift)); + tmp |= v << dshift; + *dp = (png_byte)(tmp & 0xff); + + if (dshift == s_end) + { + dshift = s_start; + dp--; + } + + else + dshift += s_inc; + } + + if (sshift == s_end) + { + sshift = s_start; + sp--; + } + + else + sshift += s_inc; + } + break; + } + + default: + { + png_size_t pixel_bytes = (row_info->pixel_depth >> 3); + + png_bytep sp = row + (png_size_t)(row_info->width - 1) + * pixel_bytes; + + png_bytep dp = row + (png_size_t)(final_width - 1) * pixel_bytes; + + int jstop = png_pass_inc[pass]; + png_uint_32 i; + + for (i = 0; i < row_info->width; i++) + { + png_byte v[8]; /* SAFE; pixel_depth does not exceed 64 */ + int j; + + memcpy(v, sp, pixel_bytes); + + for (j = 0; j < jstop; j++) + { + memcpy(dp, v, pixel_bytes); + dp -= pixel_bytes; + } + + sp -= pixel_bytes; + } + break; + } + } + + row_info->width = final_width; + row_info->rowbytes = PNG_ROWBYTES(row_info->pixel_depth, final_width); + } +#ifndef PNG_READ_PACKSWAP_SUPPORTED + PNG_UNUSED(transformations) /* Silence compiler warning */ +#endif +} +#endif /* READ_INTERLACING */ + +static void +png_read_filter_row_sub(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + png_size_t i; + png_size_t istop = row_info->rowbytes; + unsigned int bpp = (row_info->pixel_depth + 7) >> 3; + png_bytep rp = row + bpp; + + PNG_UNUSED(prev_row) + + for (i = bpp; i < istop; i++) + { + *rp = (png_byte)(((int)(*rp) + (int)(*(rp-bpp))) & 0xff); + rp++; + } +} + +static void +png_read_filter_row_up(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + png_size_t i; + png_size_t istop = row_info->rowbytes; + png_bytep rp = row; + png_const_bytep pp = prev_row; + + for (i = 0; i < istop; i++) + { + *rp = (png_byte)(((int)(*rp) + (int)(*pp++)) & 0xff); + rp++; + } +} + +static void +png_read_filter_row_avg(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + png_size_t i; + png_bytep rp = row; + png_const_bytep pp = prev_row; + unsigned int bpp = (row_info->pixel_depth + 7) >> 3; + png_size_t istop = row_info->rowbytes - bpp; + + for (i = 0; i < bpp; i++) + { + *rp = (png_byte)(((int)(*rp) + + ((int)(*pp++) / 2 )) & 0xff); + + rp++; + } + + for (i = 0; i < istop; i++) + { + *rp = (png_byte)(((int)(*rp) + + (int)(*pp++ + *(rp-bpp)) / 2 ) & 0xff); + + rp++; + } +} + +static void +png_read_filter_row_paeth_1byte_pixel(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + png_bytep rp_end = row + row_info->rowbytes; + int a, c; + + /* First pixel/byte */ + c = *prev_row++; + a = *row + c; + *row++ = (png_byte)a; + + /* Remainder */ + while (row < rp_end) + { + int b, pa, pb, pc, p; + + a &= 0xff; /* From previous iteration or start */ + b = *prev_row++; + + p = b - c; + pc = a - c; + +# ifdef PNG_USE_ABS + pa = abs(p); + pb = abs(pc); + pc = abs(p + pc); +# else + pa = p < 0 ? -p : p; + pb = pc < 0 ? -pc : pc; + pc = (p + pc) < 0 ? -(p + pc) : p + pc; +# endif + + /* Find the best predictor, the least of pa, pb, pc favoring the earlier + * ones in the case of a tie. + */ + if (pb < pa) pa = pb, a = b; + if (pc < pa) a = c; + + /* Calculate the current pixel in a, and move the previous row pixel to c + * for the next time round the loop + */ + c = b; + a += *row; + *row++ = (png_byte)a; + } +} + +static void +png_read_filter_row_paeth_multibyte_pixel(png_row_infop row_info, png_bytep row, + png_const_bytep prev_row) +{ + int bpp = (row_info->pixel_depth + 7) >> 3; + png_bytep rp_end = row + bpp; + + /* Process the first pixel in the row completely (this is the same as 'up' + * because there is only one candidate predictor for the first row). + */ + while (row < rp_end) + { + int a = *row + *prev_row++; + *row++ = (png_byte)a; + } + + /* Remainder */ + rp_end += row_info->rowbytes - bpp; + + while (row < rp_end) + { + int a, b, c, pa, pb, pc, p; + + c = *(prev_row - bpp); + a = *(row - bpp); + b = *prev_row++; + + p = b - c; + pc = a - c; + +# ifdef PNG_USE_ABS + pa = abs(p); + pb = abs(pc); + pc = abs(p + pc); +# else + pa = p < 0 ? -p : p; + pb = pc < 0 ? -pc : pc; + pc = (p + pc) < 0 ? -(p + pc) : p + pc; +# endif + + if (pb < pa) pa = pb, a = b; + if (pc < pa) a = c; + + a += *row; + *row++ = (png_byte)a; + } +} + +static void +png_init_filter_functions(png_structrp pp) + /* This function is called once for every PNG image (except for PNG images + * that only use PNG_FILTER_VALUE_NONE for all rows) to set the + * implementations required to reverse the filtering of PNG rows. Reversing + * the filter is the first transformation performed on the row data. It is + * performed in place, therefore an implementation can be selected based on + * the image pixel format. If the implementation depends on image width then + * take care to ensure that it works correctly if the image is interlaced - + * interlacing causes the actual row width to vary. + */ +{ + unsigned int bpp = (pp->pixel_depth + 7) >> 3; + + pp->read_filter[PNG_FILTER_VALUE_SUB-1] = png_read_filter_row_sub; + pp->read_filter[PNG_FILTER_VALUE_UP-1] = png_read_filter_row_up; + pp->read_filter[PNG_FILTER_VALUE_AVG-1] = png_read_filter_row_avg; + if (bpp == 1) + pp->read_filter[PNG_FILTER_VALUE_PAETH-1] = + png_read_filter_row_paeth_1byte_pixel; + else + pp->read_filter[PNG_FILTER_VALUE_PAETH-1] = + png_read_filter_row_paeth_multibyte_pixel; + +#ifdef PNG_FILTER_OPTIMIZATIONS + /* To use this define PNG_FILTER_OPTIMIZATIONS as the name of a function to + * call to install hardware optimizations for the above functions; simply + * replace whatever elements of the pp->read_filter[] array with a hardware + * specific (or, for that matter, generic) optimization. + * + * To see an example of this examine what configure.ac does when + * --enable-arm-neon is specified on the command line. + */ + PNG_FILTER_OPTIMIZATIONS(pp, bpp); +#endif +} + +void /* PRIVATE */ +png_read_filter_row(png_structrp pp, png_row_infop row_info, png_bytep row, + png_const_bytep prev_row, int filter) +{ + /* OPTIMIZATION: DO NOT MODIFY THIS FUNCTION, instead #define + * PNG_FILTER_OPTIMIZATIONS to a function that overrides the generic + * implementations. See png_init_filter_functions above. + */ + if (filter > PNG_FILTER_VALUE_NONE && filter < PNG_FILTER_VALUE_LAST) + { + if (pp->read_filter[0] == NULL) + png_init_filter_functions(pp); + + pp->read_filter[filter-1](row_info, row, prev_row); + } +} + +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED +void /* PRIVATE */ +png_read_IDAT_data(png_structrp png_ptr, png_bytep output, + png_alloc_size_t avail_out) +{ + /* Loop reading IDATs and decompressing the result into output[avail_out] */ + png_ptr->zstream.next_out = output; + png_ptr->zstream.avail_out = 0; /* safety: set below */ + + if (output == NULL) + avail_out = 0; + + do + { + int ret; + png_byte tmpbuf[PNG_INFLATE_BUF_SIZE]; + + if (png_ptr->zstream.avail_in == 0) + { + uInt avail_in; + png_bytep buffer; + + while (png_ptr->idat_size == 0) + { + png_crc_finish(png_ptr, 0); + + png_ptr->idat_size = png_read_chunk_header(png_ptr); + /* This is an error even in the 'check' case because the code just + * consumed a non-IDAT header. + */ + if (png_ptr->chunk_name != png_IDAT) + png_error(png_ptr, "Not enough image data"); + } + + avail_in = png_ptr->IDAT_read_size; + + if (avail_in > png_ptr->idat_size) + avail_in = (uInt)png_ptr->idat_size; + + /* A PNG with a gradually increasing IDAT size will defeat this attempt + * to minimize memory usage by causing lots of re-allocs, but + * realistically doing IDAT_read_size re-allocs is not likely to be a + * big problem. + */ + buffer = png_read_buffer(png_ptr, avail_in, 0/*error*/); + + png_crc_read(png_ptr, buffer, avail_in); + png_ptr->idat_size -= avail_in; + + png_ptr->zstream.next_in = buffer; + png_ptr->zstream.avail_in = avail_in; + } + + /* And set up the output side. */ + if (output != NULL) /* standard read */ + { + uInt out = ZLIB_IO_MAX; + + if (out > avail_out) + out = (uInt)avail_out; + + avail_out -= out; + png_ptr->zstream.avail_out = out; + } + + else /* after last row, checking for end */ + { + png_ptr->zstream.next_out = tmpbuf; + png_ptr->zstream.avail_out = (sizeof tmpbuf); + } + + /* Use NO_FLUSH; this gives zlib the maximum opportunity to optimize the + * process. If the LZ stream is truncated the sequential reader will + * terminally damage the stream, above, by reading the chunk header of the + * following chunk (it then exits with png_error). + * + * TODO: deal more elegantly with truncated IDAT lists. + */ + ret = inflate(&png_ptr->zstream, Z_NO_FLUSH); + + /* Take the unconsumed output back. */ + if (output != NULL) + avail_out += png_ptr->zstream.avail_out; + + else /* avail_out counts the extra bytes */ + avail_out += (sizeof tmpbuf) - png_ptr->zstream.avail_out; + + png_ptr->zstream.avail_out = 0; + + if (ret == Z_STREAM_END) + { + /* Do this for safety; we won't read any more into this row. */ + png_ptr->zstream.next_out = NULL; + + png_ptr->mode |= PNG_AFTER_IDAT; + png_ptr->flags |= PNG_FLAG_ZSTREAM_ENDED; + + if (png_ptr->zstream.avail_in > 0 || png_ptr->idat_size > 0) + png_chunk_benign_error(png_ptr, "Extra compressed data"); + break; + } + + if (ret != Z_OK) + { + png_zstream_error(png_ptr, ret); + + if (output != NULL) + png_chunk_error(png_ptr, png_ptr->zstream.msg); + + else /* checking */ + { + png_chunk_benign_error(png_ptr, png_ptr->zstream.msg); + return; + } + } + } while (avail_out > 0); + + if (avail_out > 0) + { + /* The stream ended before the image; this is the same as too few IDATs so + * should be handled the same way. + */ + if (output != NULL) + png_error(png_ptr, "Not enough image data"); + + else /* the deflate stream contained extra data */ + png_chunk_benign_error(png_ptr, "Too much image data"); + } +} + +void /* PRIVATE */ +png_read_finish_IDAT(png_structrp png_ptr) +{ + /* We don't need any more data and the stream should have ended, however the + * LZ end code may actually not have been processed. In this case we must + * read it otherwise stray unread IDAT data or, more likely, an IDAT chunk + * may still remain to be consumed. + */ + if ((png_ptr->flags & PNG_FLAG_ZSTREAM_ENDED) == 0) + { + /* The NULL causes png_read_IDAT_data to swallow any remaining bytes in + * the compressed stream, but the stream may be damaged too, so even after + * this call we may need to terminate the zstream ownership. + */ + png_read_IDAT_data(png_ptr, NULL, 0); + png_ptr->zstream.next_out = NULL; /* safety */ + + /* Now clear everything out for safety; the following may not have been + * done. + */ + if ((png_ptr->flags & PNG_FLAG_ZSTREAM_ENDED) == 0) + { + png_ptr->mode |= PNG_AFTER_IDAT; + png_ptr->flags |= PNG_FLAG_ZSTREAM_ENDED; + } + } + + /* If the zstream has not been released do it now *and* terminate the reading + * of the final IDAT chunk. + */ + if (png_ptr->zowner == png_IDAT) + { + /* Always do this; the pointers otherwise point into the read buffer. */ + png_ptr->zstream.next_in = NULL; + png_ptr->zstream.avail_in = 0; + + /* Now we no longer own the zstream. */ + png_ptr->zowner = 0; + + /* The slightly weird semantics of the sequential IDAT reading is that we + * are always in or at the end of an IDAT chunk, so we always need to do a + * crc_finish here. If idat_size is non-zero we also need to read the + * spurious bytes at the end of the chunk now. + */ + (void)png_crc_finish(png_ptr, png_ptr->idat_size); + } +} + +void /* PRIVATE */ +png_read_finish_row(png_structrp png_ptr) +{ + /* Arrays to facilitate easy interlacing - use pass (0 - 6) as index */ + + /* Start of interlace block */ + static PNG_CONST png_byte png_pass_start[7] = {0, 4, 0, 2, 0, 1, 0}; + + /* Offset to next interlace block */ + static PNG_CONST png_byte png_pass_inc[7] = {8, 8, 4, 4, 2, 2, 1}; + + /* Start of interlace block in the y direction */ + static PNG_CONST png_byte png_pass_ystart[7] = {0, 0, 4, 0, 2, 0, 1}; + + /* Offset to next interlace block in the y direction */ + static PNG_CONST png_byte png_pass_yinc[7] = {8, 8, 8, 4, 4, 2, 2}; + + png_debug(1, "in png_read_finish_row"); + png_ptr->row_number++; + if (png_ptr->row_number < png_ptr->num_rows) + return; + + if (png_ptr->interlaced != 0) + { + png_ptr->row_number = 0; + + /* TO DO: don't do this if prev_row isn't needed (requires + * read-ahead of the next row's filter byte. + */ + memset(png_ptr->prev_row, 0, png_ptr->rowbytes + 1); + + do + { + png_ptr->pass++; + + if (png_ptr->pass >= 7) + break; + + png_ptr->iwidth = (png_ptr->width + + png_pass_inc[png_ptr->pass] - 1 - + png_pass_start[png_ptr->pass]) / + png_pass_inc[png_ptr->pass]; + + if ((png_ptr->transformations & PNG_INTERLACE) == 0) + { + png_ptr->num_rows = (png_ptr->height + + png_pass_yinc[png_ptr->pass] - 1 - + png_pass_ystart[png_ptr->pass]) / + png_pass_yinc[png_ptr->pass]; + } + + else /* if (png_ptr->transformations & PNG_INTERLACE) */ + break; /* libpng deinterlacing sees every row */ + + } while (png_ptr->num_rows == 0 || png_ptr->iwidth == 0); + + if (png_ptr->pass < 7) + return; + } + + /* Here after at the end of the last row of the last pass. */ + png_read_finish_IDAT(png_ptr); +} +#endif /* SEQUENTIAL_READ */ + +void /* PRIVATE */ +png_read_start_row(png_structrp png_ptr) +{ + /* Arrays to facilitate easy interlacing - use pass (0 - 6) as index */ + + /* Start of interlace block */ + static PNG_CONST png_byte png_pass_start[7] = {0, 4, 0, 2, 0, 1, 0}; + + /* Offset to next interlace block */ + static PNG_CONST png_byte png_pass_inc[7] = {8, 8, 4, 4, 2, 2, 1}; + + /* Start of interlace block in the y direction */ + static PNG_CONST png_byte png_pass_ystart[7] = {0, 0, 4, 0, 2, 0, 1}; + + /* Offset to next interlace block in the y direction */ + static PNG_CONST png_byte png_pass_yinc[7] = {8, 8, 8, 4, 4, 2, 2}; + + int max_pixel_depth; + png_size_t row_bytes; + + png_debug(1, "in png_read_start_row"); + +#ifdef PNG_READ_TRANSFORMS_SUPPORTED + png_init_read_transformations(png_ptr); +#endif + if (png_ptr->interlaced != 0) + { + if ((png_ptr->transformations & PNG_INTERLACE) == 0) + png_ptr->num_rows = (png_ptr->height + png_pass_yinc[0] - 1 - + png_pass_ystart[0]) / png_pass_yinc[0]; + + else + png_ptr->num_rows = png_ptr->height; + + png_ptr->iwidth = (png_ptr->width + + png_pass_inc[png_ptr->pass] - 1 - + png_pass_start[png_ptr->pass]) / + png_pass_inc[png_ptr->pass]; + } + + else + { + png_ptr->num_rows = png_ptr->height; + png_ptr->iwidth = png_ptr->width; + } + + max_pixel_depth = png_ptr->pixel_depth; + + /* WARNING: * png_read_transform_info (pngrtran.c) performs a simpler set of + * calculations to calculate the final pixel depth, then + * png_do_read_transforms actually does the transforms. This means that the + * code which effectively calculates this value is actually repeated in three + * separate places. They must all match. Innocent changes to the order of + * transformations can and will break libpng in a way that causes memory + * overwrites. + * + * TODO: fix this. + */ +#ifdef PNG_READ_PACK_SUPPORTED + if ((png_ptr->transformations & PNG_PACK) != 0 && png_ptr->bit_depth < 8) + max_pixel_depth = 8; +#endif + +#ifdef PNG_READ_EXPAND_SUPPORTED + if ((png_ptr->transformations & PNG_EXPAND) != 0) + { + if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE) + { + if (png_ptr->num_trans != 0) + max_pixel_depth = 32; + + else + max_pixel_depth = 24; + } + + else if (png_ptr->color_type == PNG_COLOR_TYPE_GRAY) + { + if (max_pixel_depth < 8) + max_pixel_depth = 8; + + if (png_ptr->num_trans != 0) + max_pixel_depth *= 2; + } + + else if (png_ptr->color_type == PNG_COLOR_TYPE_RGB) + { + if (png_ptr->num_trans != 0) + { + max_pixel_depth *= 4; + max_pixel_depth /= 3; + } + } + } +#endif + +#ifdef PNG_READ_EXPAND_16_SUPPORTED + if ((png_ptr->transformations & PNG_EXPAND_16) != 0) + { +# ifdef PNG_READ_EXPAND_SUPPORTED + /* In fact it is an error if it isn't supported, but checking is + * the safe way. + */ + if ((png_ptr->transformations & PNG_EXPAND) != 0) + { + if (png_ptr->bit_depth < 16) + max_pixel_depth *= 2; + } + else +# endif + png_ptr->transformations &= ~PNG_EXPAND_16; + } +#endif + +#ifdef PNG_READ_FILLER_SUPPORTED + if ((png_ptr->transformations & (PNG_FILLER)) != 0) + { + if (png_ptr->color_type == PNG_COLOR_TYPE_GRAY) + { + if (max_pixel_depth <= 8) + max_pixel_depth = 16; + + else + max_pixel_depth = 32; + } + + else if (png_ptr->color_type == PNG_COLOR_TYPE_RGB || + png_ptr->color_type == PNG_COLOR_TYPE_PALETTE) + { + if (max_pixel_depth <= 32) + max_pixel_depth = 32; + + else + max_pixel_depth = 64; + } + } +#endif + +#ifdef PNG_READ_GRAY_TO_RGB_SUPPORTED + if ((png_ptr->transformations & PNG_GRAY_TO_RGB) != 0) + { + if ( +#ifdef PNG_READ_EXPAND_SUPPORTED + (png_ptr->num_trans != 0 && + (png_ptr->transformations & PNG_EXPAND) != 0) || +#endif +#ifdef PNG_READ_FILLER_SUPPORTED + (png_ptr->transformations & (PNG_FILLER)) != 0 || +#endif + png_ptr->color_type == PNG_COLOR_TYPE_GRAY_ALPHA) + { + if (max_pixel_depth <= 16) + max_pixel_depth = 32; + + else + max_pixel_depth = 64; + } + + else + { + if (max_pixel_depth <= 8) + { + if (png_ptr->color_type == PNG_COLOR_TYPE_RGB_ALPHA) + max_pixel_depth = 32; + + else + max_pixel_depth = 24; + } + + else if (png_ptr->color_type == PNG_COLOR_TYPE_RGB_ALPHA) + max_pixel_depth = 64; + + else + max_pixel_depth = 48; + } + } +#endif + +#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED) && \ +defined(PNG_USER_TRANSFORM_PTR_SUPPORTED) + if ((png_ptr->transformations & PNG_USER_TRANSFORM) != 0) + { + int user_pixel_depth = png_ptr->user_transform_depth * + png_ptr->user_transform_channels; + + if (user_pixel_depth > max_pixel_depth) + max_pixel_depth = user_pixel_depth; + } +#endif + + /* This value is stored in png_struct and double checked in the row read + * code. + */ + png_ptr->maximum_pixel_depth = (png_byte)max_pixel_depth; + png_ptr->transformed_pixel_depth = 0; /* calculated on demand */ + + /* Align the width on the next larger 8 pixels. Mainly used + * for interlacing + */ + row_bytes = ((png_ptr->width + 7) & ~((png_uint_32)7)); + /* Calculate the maximum bytes needed, adding a byte and a pixel + * for safety's sake + */ + row_bytes = PNG_ROWBYTES(max_pixel_depth, row_bytes) + + 1 + ((max_pixel_depth + 7) >> 3); + +#ifdef PNG_MAX_MALLOC_64K + if (row_bytes > (png_uint_32)65536L) + png_error(png_ptr, "This image requires a row greater than 64KB"); +#endif + + if (row_bytes + 48 > png_ptr->old_big_row_buf_size) + { + png_free(png_ptr, png_ptr->big_row_buf); + png_free(png_ptr, png_ptr->big_prev_row); + + if (png_ptr->interlaced != 0) + png_ptr->big_row_buf = (png_bytep)png_calloc(png_ptr, + row_bytes + 48); + + else + png_ptr->big_row_buf = (png_bytep)png_malloc(png_ptr, row_bytes + 48); + + png_ptr->big_prev_row = (png_bytep)png_malloc(png_ptr, row_bytes + 48); + +#ifdef PNG_ALIGNED_MEMORY_SUPPORTED + /* Use 16-byte aligned memory for row_buf with at least 16 bytes + * of padding before and after row_buf; treat prev_row similarly. + * NOTE: the alignment is to the start of the pixels, one beyond the start + * of the buffer, because of the filter byte. Prior to libpng 1.5.6 this + * was incorrect; the filter byte was aligned, which had the exact + * opposite effect of that intended. + */ + { + png_bytep temp = png_ptr->big_row_buf + 32; + int extra = (int)((temp - (png_bytep)0) & 0x0f); + png_ptr->row_buf = temp - extra - 1/*filter byte*/; + + temp = png_ptr->big_prev_row + 32; + extra = (int)((temp - (png_bytep)0) & 0x0f); + png_ptr->prev_row = temp - extra - 1/*filter byte*/; + } + +#else + /* Use 31 bytes of padding before and 17 bytes after row_buf. */ + png_ptr->row_buf = png_ptr->big_row_buf + 31; + png_ptr->prev_row = png_ptr->big_prev_row + 31; +#endif + png_ptr->old_big_row_buf_size = row_bytes + 48; + } + +#ifdef PNG_MAX_MALLOC_64K + if (png_ptr->rowbytes > 65535) + png_error(png_ptr, "This image requires a row greater than 64KB"); + +#endif + if (png_ptr->rowbytes > (PNG_SIZE_MAX - 1)) + png_error(png_ptr, "Row has too many bytes to allocate in memory"); + + memset(png_ptr->prev_row, 0, png_ptr->rowbytes + 1); + + png_debug1(3, "width = %u,", png_ptr->width); + png_debug1(3, "height = %u,", png_ptr->height); + png_debug1(3, "iwidth = %u,", png_ptr->iwidth); + png_debug1(3, "num_rows = %u,", png_ptr->num_rows); + png_debug1(3, "rowbytes = %lu,", (unsigned long)png_ptr->rowbytes); + png_debug1(3, "irowbytes = %lu", + (unsigned long)PNG_ROWBYTES(png_ptr->pixel_depth, png_ptr->iwidth) + 1); + + /* The sequential reader needs a buffer for IDAT, but the progressive reader + * does not, so free the read buffer now regardless; the sequential reader + * reallocates it on demand. + */ + if (png_ptr->read_buffer != 0) + { + png_bytep buffer = png_ptr->read_buffer; + + png_ptr->read_buffer_size = 0; + png_ptr->read_buffer = NULL; + png_free(png_ptr, buffer); + } + + /* Finally claim the zstream for the inflate of the IDAT data, use the bits + * value from the stream (note that this will result in a fatal error if the + * IDAT stream has a bogus deflate header window_bits value, but this should + * not be happening any longer!) + */ + if (png_inflate_claim(png_ptr, png_IDAT) != Z_OK) + png_error(png_ptr, png_ptr->zstream.msg); + + png_ptr->flags |= PNG_FLAG_ROW_INIT; +} +#endif /* READ */ diff --git a/app/src/main/jni/libpng/pngset.c b/app/src/main/jni/libpng/pngset.c new file mode 100644 index 000000000..2f1f550da --- /dev/null +++ b/app/src/main/jni/libpng/pngset.c @@ -0,0 +1,1611 @@ + +/* pngset.c - storage of image information into info struct + * + * Last changed in libpng 1.6.15 [November 20, 2014] + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + * + * The functions here are used during reads to store data from the file + * into the info struct, and during writes to store application data + * into the info struct for writing into the file. This abstracts the + * info struct and allows us to change the structure in the future. + */ + +#include "pngpriv.h" + +#if defined(PNG_READ_SUPPORTED) || defined(PNG_WRITE_SUPPORTED) + +#ifdef PNG_bKGD_SUPPORTED +void PNGAPI +png_set_bKGD(png_const_structrp png_ptr, png_inforp info_ptr, + png_const_color_16p background) +{ + png_debug1(1, "in %s storage function", "bKGD"); + + if (png_ptr == NULL || info_ptr == NULL || background == NULL) + return; + + info_ptr->background = *background; + info_ptr->valid |= PNG_INFO_bKGD; +} +#endif + +#ifdef PNG_cHRM_SUPPORTED +void PNGFAPI +png_set_cHRM_fixed(png_const_structrp png_ptr, png_inforp info_ptr, + png_fixed_point white_x, png_fixed_point white_y, png_fixed_point red_x, + png_fixed_point red_y, png_fixed_point green_x, png_fixed_point green_y, + png_fixed_point blue_x, png_fixed_point blue_y) +{ + png_xy xy; + + png_debug1(1, "in %s storage function", "cHRM fixed"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + xy.redx = red_x; + xy.redy = red_y; + xy.greenx = green_x; + xy.greeny = green_y; + xy.bluex = blue_x; + xy.bluey = blue_y; + xy.whitex = white_x; + xy.whitey = white_y; + + if (png_colorspace_set_chromaticities(png_ptr, &info_ptr->colorspace, &xy, + 2/* override with app values*/) != 0) + info_ptr->colorspace.flags |= PNG_COLORSPACE_FROM_cHRM; + + png_colorspace_sync_info(png_ptr, info_ptr); +} + +void PNGFAPI +png_set_cHRM_XYZ_fixed(png_const_structrp png_ptr, png_inforp info_ptr, + png_fixed_point int_red_X, png_fixed_point int_red_Y, + png_fixed_point int_red_Z, png_fixed_point int_green_X, + png_fixed_point int_green_Y, png_fixed_point int_green_Z, + png_fixed_point int_blue_X, png_fixed_point int_blue_Y, + png_fixed_point int_blue_Z) +{ + png_XYZ XYZ; + + png_debug1(1, "in %s storage function", "cHRM XYZ fixed"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + XYZ.red_X = int_red_X; + XYZ.red_Y = int_red_Y; + XYZ.red_Z = int_red_Z; + XYZ.green_X = int_green_X; + XYZ.green_Y = int_green_Y; + XYZ.green_Z = int_green_Z; + XYZ.blue_X = int_blue_X; + XYZ.blue_Y = int_blue_Y; + XYZ.blue_Z = int_blue_Z; + + if (png_colorspace_set_endpoints(png_ptr, &info_ptr->colorspace, + &XYZ, 2) != 0) + info_ptr->colorspace.flags |= PNG_COLORSPACE_FROM_cHRM; + + png_colorspace_sync_info(png_ptr, info_ptr); +} + +# ifdef PNG_FLOATING_POINT_SUPPORTED +void PNGAPI +png_set_cHRM(png_const_structrp png_ptr, png_inforp info_ptr, + double white_x, double white_y, double red_x, double red_y, + double green_x, double green_y, double blue_x, double blue_y) +{ + png_set_cHRM_fixed(png_ptr, info_ptr, + png_fixed(png_ptr, white_x, "cHRM White X"), + png_fixed(png_ptr, white_y, "cHRM White Y"), + png_fixed(png_ptr, red_x, "cHRM Red X"), + png_fixed(png_ptr, red_y, "cHRM Red Y"), + png_fixed(png_ptr, green_x, "cHRM Green X"), + png_fixed(png_ptr, green_y, "cHRM Green Y"), + png_fixed(png_ptr, blue_x, "cHRM Blue X"), + png_fixed(png_ptr, blue_y, "cHRM Blue Y")); +} + +void PNGAPI +png_set_cHRM_XYZ(png_const_structrp png_ptr, png_inforp info_ptr, double red_X, + double red_Y, double red_Z, double green_X, double green_Y, double green_Z, + double blue_X, double blue_Y, double blue_Z) +{ + png_set_cHRM_XYZ_fixed(png_ptr, info_ptr, + png_fixed(png_ptr, red_X, "cHRM Red X"), + png_fixed(png_ptr, red_Y, "cHRM Red Y"), + png_fixed(png_ptr, red_Z, "cHRM Red Z"), + png_fixed(png_ptr, green_X, "cHRM Red X"), + png_fixed(png_ptr, green_Y, "cHRM Red Y"), + png_fixed(png_ptr, green_Z, "cHRM Red Z"), + png_fixed(png_ptr, blue_X, "cHRM Red X"), + png_fixed(png_ptr, blue_Y, "cHRM Red Y"), + png_fixed(png_ptr, blue_Z, "cHRM Red Z")); +} +# endif /* FLOATING_POINT */ + +#endif /* cHRM */ + +#ifdef PNG_gAMA_SUPPORTED +void PNGFAPI +png_set_gAMA_fixed(png_const_structrp png_ptr, png_inforp info_ptr, + png_fixed_point file_gamma) +{ + png_debug1(1, "in %s storage function", "gAMA"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + png_colorspace_set_gamma(png_ptr, &info_ptr->colorspace, file_gamma); + png_colorspace_sync_info(png_ptr, info_ptr); +} + +# ifdef PNG_FLOATING_POINT_SUPPORTED +void PNGAPI +png_set_gAMA(png_const_structrp png_ptr, png_inforp info_ptr, double file_gamma) +{ + png_set_gAMA_fixed(png_ptr, info_ptr, png_fixed(png_ptr, file_gamma, + "png_set_gAMA")); +} +# endif +#endif + +#ifdef PNG_hIST_SUPPORTED +void PNGAPI +png_set_hIST(png_const_structrp png_ptr, png_inforp info_ptr, + png_const_uint_16p hist) +{ + int i; + + png_debug1(1, "in %s storage function", "hIST"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + if (info_ptr->num_palette == 0 || info_ptr->num_palette + > PNG_MAX_PALETTE_LENGTH) + { + png_warning(png_ptr, + "Invalid palette size, hIST allocation skipped"); + + return; + } + + png_free_data(png_ptr, info_ptr, PNG_FREE_HIST, 0); + + /* Changed from info->num_palette to PNG_MAX_PALETTE_LENGTH in + * version 1.2.1 + */ + info_ptr->hist = png_voidcast(png_uint_16p, png_malloc_warn(png_ptr, + PNG_MAX_PALETTE_LENGTH * (sizeof (png_uint_16)))); + + if (info_ptr->hist == NULL) + { + png_warning(png_ptr, "Insufficient memory for hIST chunk data"); + return; + } + + info_ptr->free_me |= PNG_FREE_HIST; + + for (i = 0; i < info_ptr->num_palette; i++) + info_ptr->hist[i] = hist[i]; + + info_ptr->valid |= PNG_INFO_hIST; +} +#endif + +void PNGAPI +png_set_IHDR(png_const_structrp png_ptr, png_inforp info_ptr, + png_uint_32 width, png_uint_32 height, int bit_depth, + int color_type, int interlace_type, int compression_type, + int filter_type) +{ + png_debug1(1, "in %s storage function", "IHDR"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + info_ptr->width = width; + info_ptr->height = height; + info_ptr->bit_depth = (png_byte)bit_depth; + info_ptr->color_type = (png_byte)color_type; + info_ptr->compression_type = (png_byte)compression_type; + info_ptr->filter_type = (png_byte)filter_type; + info_ptr->interlace_type = (png_byte)interlace_type; + + png_check_IHDR (png_ptr, info_ptr->width, info_ptr->height, + info_ptr->bit_depth, info_ptr->color_type, info_ptr->interlace_type, + info_ptr->compression_type, info_ptr->filter_type); + + if (info_ptr->color_type == PNG_COLOR_TYPE_PALETTE) + info_ptr->channels = 1; + + else if ((info_ptr->color_type & PNG_COLOR_MASK_COLOR) != 0) + info_ptr->channels = 3; + + else + info_ptr->channels = 1; + + if ((info_ptr->color_type & PNG_COLOR_MASK_ALPHA) != 0) + info_ptr->channels++; + + info_ptr->pixel_depth = (png_byte)(info_ptr->channels * info_ptr->bit_depth); + + info_ptr->rowbytes = PNG_ROWBYTES(info_ptr->pixel_depth, width); +} + +#ifdef PNG_oFFs_SUPPORTED +void PNGAPI +png_set_oFFs(png_const_structrp png_ptr, png_inforp info_ptr, + png_int_32 offset_x, png_int_32 offset_y, int unit_type) +{ + png_debug1(1, "in %s storage function", "oFFs"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + info_ptr->x_offset = offset_x; + info_ptr->y_offset = offset_y; + info_ptr->offset_unit_type = (png_byte)unit_type; + info_ptr->valid |= PNG_INFO_oFFs; +} +#endif + +#ifdef PNG_pCAL_SUPPORTED +void PNGAPI +png_set_pCAL(png_const_structrp png_ptr, png_inforp info_ptr, + png_const_charp purpose, png_int_32 X0, png_int_32 X1, int type, + int nparams, png_const_charp units, png_charpp params) +{ + png_size_t length; + int i; + + png_debug1(1, "in %s storage function", "pCAL"); + + if (png_ptr == NULL || info_ptr == NULL || purpose == NULL || units == NULL + || (nparams > 0 && params == NULL)) + return; + + length = strlen(purpose) + 1; + png_debug1(3, "allocating purpose for info (%lu bytes)", + (unsigned long)length); + + /* TODO: validate format of calibration name and unit name */ + + /* Check that the type matches the specification. */ + if (type < 0 || type > 3) + png_error(png_ptr, "Invalid pCAL equation type"); + + if (nparams < 0 || nparams > 255) + png_error(png_ptr, "Invalid pCAL parameter count"); + + /* Validate params[nparams] */ + for (i=0; ipcal_purpose = png_voidcast(png_charp, + png_malloc_warn(png_ptr, length)); + + if (info_ptr->pcal_purpose == NULL) + { + png_warning(png_ptr, "Insufficient memory for pCAL purpose"); + return; + } + + memcpy(info_ptr->pcal_purpose, purpose, length); + + png_debug(3, "storing X0, X1, type, and nparams in info"); + info_ptr->pcal_X0 = X0; + info_ptr->pcal_X1 = X1; + info_ptr->pcal_type = (png_byte)type; + info_ptr->pcal_nparams = (png_byte)nparams; + + length = strlen(units) + 1; + png_debug1(3, "allocating units for info (%lu bytes)", + (unsigned long)length); + + info_ptr->pcal_units = png_voidcast(png_charp, + png_malloc_warn(png_ptr, length)); + + if (info_ptr->pcal_units == NULL) + { + png_warning(png_ptr, "Insufficient memory for pCAL units"); + return; + } + + memcpy(info_ptr->pcal_units, units, length); + + info_ptr->pcal_params = png_voidcast(png_charpp, png_malloc_warn(png_ptr, + (png_size_t)((nparams + 1) * (sizeof (png_charp))))); + + if (info_ptr->pcal_params == NULL) + { + png_warning(png_ptr, "Insufficient memory for pCAL params"); + return; + } + + memset(info_ptr->pcal_params, 0, (nparams + 1) * (sizeof (png_charp))); + + for (i = 0; i < nparams; i++) + { + length = strlen(params[i]) + 1; + png_debug2(3, "allocating parameter %d for info (%lu bytes)", i, + (unsigned long)length); + + info_ptr->pcal_params[i] = (png_charp)png_malloc_warn(png_ptr, length); + + if (info_ptr->pcal_params[i] == NULL) + { + png_warning(png_ptr, "Insufficient memory for pCAL parameter"); + return; + } + + memcpy(info_ptr->pcal_params[i], params[i], length); + } + + info_ptr->valid |= PNG_INFO_pCAL; + info_ptr->free_me |= PNG_FREE_PCAL; +} +#endif + +#ifdef PNG_sCAL_SUPPORTED +void PNGAPI +png_set_sCAL_s(png_const_structrp png_ptr, png_inforp info_ptr, + int unit, png_const_charp swidth, png_const_charp sheight) +{ + png_size_t lengthw = 0, lengthh = 0; + + png_debug1(1, "in %s storage function", "sCAL"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + /* Double check the unit (should never get here with an invalid + * unit unless this is an API call.) + */ + if (unit != 1 && unit != 2) + png_error(png_ptr, "Invalid sCAL unit"); + + if (swidth == NULL || (lengthw = strlen(swidth)) == 0 || + swidth[0] == 45 /* '-' */ || !png_check_fp_string(swidth, lengthw)) + png_error(png_ptr, "Invalid sCAL width"); + + if (sheight == NULL || (lengthh = strlen(sheight)) == 0 || + sheight[0] == 45 /* '-' */ || !png_check_fp_string(sheight, lengthh)) + png_error(png_ptr, "Invalid sCAL height"); + + info_ptr->scal_unit = (png_byte)unit; + + ++lengthw; + + png_debug1(3, "allocating unit for info (%u bytes)", (unsigned int)lengthw); + + info_ptr->scal_s_width = png_voidcast(png_charp, + png_malloc_warn(png_ptr, lengthw)); + + if (info_ptr->scal_s_width == NULL) + { + png_warning(png_ptr, "Memory allocation failed while processing sCAL"); + return; + } + + memcpy(info_ptr->scal_s_width, swidth, lengthw); + + ++lengthh; + + png_debug1(3, "allocating unit for info (%u bytes)", (unsigned int)lengthh); + + info_ptr->scal_s_height = png_voidcast(png_charp, + png_malloc_warn(png_ptr, lengthh)); + + if (info_ptr->scal_s_height == NULL) + { + png_free (png_ptr, info_ptr->scal_s_width); + info_ptr->scal_s_width = NULL; + + png_warning(png_ptr, "Memory allocation failed while processing sCAL"); + return; + } + + memcpy(info_ptr->scal_s_height, sheight, lengthh); + + info_ptr->valid |= PNG_INFO_sCAL; + info_ptr->free_me |= PNG_FREE_SCAL; +} + +# ifdef PNG_FLOATING_POINT_SUPPORTED +void PNGAPI +png_set_sCAL(png_const_structrp png_ptr, png_inforp info_ptr, int unit, + double width, double height) +{ + png_debug1(1, "in %s storage function", "sCAL"); + + /* Check the arguments. */ + if (width <= 0) + png_warning(png_ptr, "Invalid sCAL width ignored"); + + else if (height <= 0) + png_warning(png_ptr, "Invalid sCAL height ignored"); + + else + { + /* Convert 'width' and 'height' to ASCII. */ + char swidth[PNG_sCAL_MAX_DIGITS+1]; + char sheight[PNG_sCAL_MAX_DIGITS+1]; + + png_ascii_from_fp(png_ptr, swidth, (sizeof swidth), width, + PNG_sCAL_PRECISION); + png_ascii_from_fp(png_ptr, sheight, (sizeof sheight), height, + PNG_sCAL_PRECISION); + + png_set_sCAL_s(png_ptr, info_ptr, unit, swidth, sheight); + } +} +# endif + +# ifdef PNG_FIXED_POINT_SUPPORTED +void PNGAPI +png_set_sCAL_fixed(png_const_structrp png_ptr, png_inforp info_ptr, int unit, + png_fixed_point width, png_fixed_point height) +{ + png_debug1(1, "in %s storage function", "sCAL"); + + /* Check the arguments. */ + if (width <= 0) + png_warning(png_ptr, "Invalid sCAL width ignored"); + + else if (height <= 0) + png_warning(png_ptr, "Invalid sCAL height ignored"); + + else + { + /* Convert 'width' and 'height' to ASCII. */ + char swidth[PNG_sCAL_MAX_DIGITS+1]; + char sheight[PNG_sCAL_MAX_DIGITS+1]; + + png_ascii_from_fixed(png_ptr, swidth, (sizeof swidth), width); + png_ascii_from_fixed(png_ptr, sheight, (sizeof sheight), height); + + png_set_sCAL_s(png_ptr, info_ptr, unit, swidth, sheight); + } +} +# endif +#endif + +#ifdef PNG_pHYs_SUPPORTED +void PNGAPI +png_set_pHYs(png_const_structrp png_ptr, png_inforp info_ptr, + png_uint_32 res_x, png_uint_32 res_y, int unit_type) +{ + png_debug1(1, "in %s storage function", "pHYs"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + info_ptr->x_pixels_per_unit = res_x; + info_ptr->y_pixels_per_unit = res_y; + info_ptr->phys_unit_type = (png_byte)unit_type; + info_ptr->valid |= PNG_INFO_pHYs; +} +#endif + +void PNGAPI +png_set_PLTE(png_structrp png_ptr, png_inforp info_ptr, + png_const_colorp palette, int num_palette) +{ + + png_debug1(1, "in %s storage function", "PLTE"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + if (num_palette < 0 || num_palette > PNG_MAX_PALETTE_LENGTH) + { + if (info_ptr->color_type == PNG_COLOR_TYPE_PALETTE) + png_error(png_ptr, "Invalid palette length"); + + else + { + png_warning(png_ptr, "Invalid palette length"); + return; + } + } + + if ((num_palette > 0 && palette == NULL) || + (num_palette == 0 +# ifdef PNG_MNG_FEATURES_SUPPORTED + && (png_ptr->mng_features_permitted & PNG_FLAG_MNG_EMPTY_PLTE) == 0 +# endif + )) + { + png_error(png_ptr, "Invalid palette"); + return; + } + + /* It may not actually be necessary to set png_ptr->palette here; + * we do it for backward compatibility with the way the png_handle_tRNS + * function used to do the allocation. + * + * 1.6.0: the above statement appears to be incorrect; something has to set + * the palette inside png_struct on read. + */ + png_free_data(png_ptr, info_ptr, PNG_FREE_PLTE, 0); + + /* Changed in libpng-1.2.1 to allocate PNG_MAX_PALETTE_LENGTH instead + * of num_palette entries, in case of an invalid PNG file that has + * too-large sample values. + */ + png_ptr->palette = png_voidcast(png_colorp, png_calloc(png_ptr, + PNG_MAX_PALETTE_LENGTH * (sizeof (png_color)))); + + if (num_palette > 0) + memcpy(png_ptr->palette, palette, num_palette * (sizeof (png_color))); + info_ptr->palette = png_ptr->palette; + info_ptr->num_palette = png_ptr->num_palette = (png_uint_16)num_palette; + + info_ptr->free_me |= PNG_FREE_PLTE; + + info_ptr->valid |= PNG_INFO_PLTE; +} + +#ifdef PNG_sBIT_SUPPORTED +void PNGAPI +png_set_sBIT(png_const_structrp png_ptr, png_inforp info_ptr, + png_const_color_8p sig_bit) +{ + png_debug1(1, "in %s storage function", "sBIT"); + + if (png_ptr == NULL || info_ptr == NULL || sig_bit == NULL) + return; + + info_ptr->sig_bit = *sig_bit; + info_ptr->valid |= PNG_INFO_sBIT; +} +#endif + +#ifdef PNG_sRGB_SUPPORTED +void PNGAPI +png_set_sRGB(png_const_structrp png_ptr, png_inforp info_ptr, int srgb_intent) +{ + png_debug1(1, "in %s storage function", "sRGB"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + (void)png_colorspace_set_sRGB(png_ptr, &info_ptr->colorspace, srgb_intent); + png_colorspace_sync_info(png_ptr, info_ptr); +} + +void PNGAPI +png_set_sRGB_gAMA_and_cHRM(png_const_structrp png_ptr, png_inforp info_ptr, + int srgb_intent) +{ + png_debug1(1, "in %s storage function", "sRGB_gAMA_and_cHRM"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + if (png_colorspace_set_sRGB(png_ptr, &info_ptr->colorspace, + srgb_intent) != 0) + { + /* This causes the gAMA and cHRM to be written too */ + info_ptr->colorspace.flags |= + PNG_COLORSPACE_FROM_gAMA|PNG_COLORSPACE_FROM_cHRM; + } + + png_colorspace_sync_info(png_ptr, info_ptr); +} +#endif /* sRGB */ + + +#ifdef PNG_iCCP_SUPPORTED +void PNGAPI +png_set_iCCP(png_const_structrp png_ptr, png_inforp info_ptr, + png_const_charp name, int compression_type, + png_const_bytep profile, png_uint_32 proflen) +{ + png_charp new_iccp_name; + png_bytep new_iccp_profile; + png_size_t length; + + png_debug1(1, "in %s storage function", "iCCP"); + + if (png_ptr == NULL || info_ptr == NULL || name == NULL || profile == NULL) + return; + + if (compression_type != PNG_COMPRESSION_TYPE_BASE) + png_app_error(png_ptr, "Invalid iCCP compression method"); + + /* Set the colorspace first because this validates the profile; do not + * override previously set app cHRM or gAMA here (because likely as not the + * application knows better than libpng what the correct values are.) Pass + * the info_ptr color_type field to png_colorspace_set_ICC because in the + * write case it has not yet been stored in png_ptr. + */ + { + int result = png_colorspace_set_ICC(png_ptr, &info_ptr->colorspace, name, + proflen, profile, info_ptr->color_type); + + png_colorspace_sync_info(png_ptr, info_ptr); + + /* Don't do any of the copying if the profile was bad, or inconsistent. */ + if (result == 0) + return; + + /* But do write the gAMA and cHRM chunks from the profile. */ + info_ptr->colorspace.flags |= + PNG_COLORSPACE_FROM_gAMA|PNG_COLORSPACE_FROM_cHRM; + } + + length = strlen(name)+1; + new_iccp_name = png_voidcast(png_charp, png_malloc_warn(png_ptr, length)); + + if (new_iccp_name == NULL) + { + png_benign_error(png_ptr, "Insufficient memory to process iCCP chunk"); + return; + } + + memcpy(new_iccp_name, name, length); + new_iccp_profile = png_voidcast(png_bytep, + png_malloc_warn(png_ptr, proflen)); + + if (new_iccp_profile == NULL) + { + png_free(png_ptr, new_iccp_name); + new_iccp_name = NULL; + png_benign_error(png_ptr, + "Insufficient memory to process iCCP profile"); + return; + } + + memcpy(new_iccp_profile, profile, proflen); + + png_free_data(png_ptr, info_ptr, PNG_FREE_ICCP, 0); + + info_ptr->iccp_proflen = proflen; + info_ptr->iccp_name = new_iccp_name; + info_ptr->iccp_profile = new_iccp_profile; + info_ptr->free_me |= PNG_FREE_ICCP; + info_ptr->valid |= PNG_INFO_iCCP; +} +#endif + +#ifdef PNG_TEXT_SUPPORTED +void PNGAPI +png_set_text(png_const_structrp png_ptr, png_inforp info_ptr, + png_const_textp text_ptr, int num_text) +{ + int ret; + ret = png_set_text_2(png_ptr, info_ptr, text_ptr, num_text); + + if (ret != 0) + png_error(png_ptr, "Insufficient memory to store text"); +} + +int /* PRIVATE */ +png_set_text_2(png_const_structrp png_ptr, png_inforp info_ptr, + png_const_textp text_ptr, int num_text) +{ + int i; + + png_debug1(1, "in %lx storage function", png_ptr == NULL ? "unexpected" : + (unsigned long)png_ptr->chunk_name); + + if (png_ptr == NULL || info_ptr == NULL || num_text <= 0 || text_ptr == NULL) + return(0); + + /* Make sure we have enough space in the "text" array in info_struct + * to hold all of the incoming text_ptr objects. This compare can't overflow + * because max_text >= num_text (anyway, subtract of two positive integers + * can't overflow in any case.) + */ + if (num_text > info_ptr->max_text - info_ptr->num_text) + { + int old_num_text = info_ptr->num_text; + int max_text; + png_textp new_text = NULL; + + /* Calculate an appropriate max_text, checking for overflow. */ + max_text = old_num_text; + if (num_text <= INT_MAX - max_text) + { + max_text += num_text; + + /* Round up to a multiple of 8 */ + if (max_text < INT_MAX-8) + max_text = (max_text + 8) & ~0x7; + + else + max_text = INT_MAX; + + /* Now allocate a new array and copy the old members in; this does all + * the overflow checks. + */ + new_text = png_voidcast(png_textp,png_realloc_array(png_ptr, + info_ptr->text, old_num_text, max_text-old_num_text, + sizeof *new_text)); + } + + if (new_text == NULL) + { + png_chunk_report(png_ptr, "too many text chunks", + PNG_CHUNK_WRITE_ERROR); + return 1; + } + + png_free(png_ptr, info_ptr->text); + + info_ptr->text = new_text; + info_ptr->free_me |= PNG_FREE_TEXT; + info_ptr->max_text = max_text; + /* num_text is adjusted below as the entries are copied in */ + + png_debug1(3, "allocated %d entries for info_ptr->text", max_text); + } + + for (i = 0; i < num_text; i++) + { + size_t text_length, key_len; + size_t lang_len, lang_key_len; + png_textp textp = &(info_ptr->text[info_ptr->num_text]); + + if (text_ptr[i].key == NULL) + continue; + + if (text_ptr[i].compression < PNG_TEXT_COMPRESSION_NONE || + text_ptr[i].compression >= PNG_TEXT_COMPRESSION_LAST) + { + png_chunk_report(png_ptr, "text compression mode is out of range", + PNG_CHUNK_WRITE_ERROR); + continue; + } + + key_len = strlen(text_ptr[i].key); + + if (text_ptr[i].compression <= 0) + { + lang_len = 0; + lang_key_len = 0; + } + + else +# ifdef PNG_iTXt_SUPPORTED + { + /* Set iTXt data */ + + if (text_ptr[i].lang != NULL) + lang_len = strlen(text_ptr[i].lang); + + else + lang_len = 0; + + if (text_ptr[i].lang_key != NULL) + lang_key_len = strlen(text_ptr[i].lang_key); + + else + lang_key_len = 0; + } +# else /* PNG_iTXt_SUPPORTED */ + { + png_chunk_report(png_ptr, "iTXt chunk not supported", + PNG_CHUNK_WRITE_ERROR); + continue; + } +# endif + + if (text_ptr[i].text == NULL || text_ptr[i].text[0] == '\0') + { + text_length = 0; +# ifdef PNG_iTXt_SUPPORTED + if (text_ptr[i].compression > 0) + textp->compression = PNG_ITXT_COMPRESSION_NONE; + + else +# endif + textp->compression = PNG_TEXT_COMPRESSION_NONE; + } + + else + { + text_length = strlen(text_ptr[i].text); + textp->compression = text_ptr[i].compression; + } + + textp->key = png_voidcast(png_charp,png_malloc_base(png_ptr, + key_len + text_length + lang_len + lang_key_len + 4)); + + if (textp->key == NULL) + { + png_chunk_report(png_ptr, "text chunk: out of memory", + PNG_CHUNK_WRITE_ERROR); + return 1; + } + + png_debug2(2, "Allocated %lu bytes at %p in png_set_text", + (unsigned long)(png_uint_32) + (key_len + lang_len + lang_key_len + text_length + 4), + textp->key); + + memcpy(textp->key, text_ptr[i].key, key_len); + *(textp->key + key_len) = '\0'; + + if (text_ptr[i].compression > 0) + { + textp->lang = textp->key + key_len + 1; + memcpy(textp->lang, text_ptr[i].lang, lang_len); + *(textp->lang + lang_len) = '\0'; + textp->lang_key = textp->lang + lang_len + 1; + memcpy(textp->lang_key, text_ptr[i].lang_key, lang_key_len); + *(textp->lang_key + lang_key_len) = '\0'; + textp->text = textp->lang_key + lang_key_len + 1; + } + + else + { + textp->lang=NULL; + textp->lang_key=NULL; + textp->text = textp->key + key_len + 1; + } + + if (text_length != 0) + memcpy(textp->text, text_ptr[i].text, text_length); + + *(textp->text + text_length) = '\0'; + +# ifdef PNG_iTXt_SUPPORTED + if (textp->compression > 0) + { + textp->text_length = 0; + textp->itxt_length = text_length; + } + + else +# endif + { + textp->text_length = text_length; + textp->itxt_length = 0; + } + + info_ptr->num_text++; + png_debug1(3, "transferred text chunk %d", info_ptr->num_text); + } + + return(0); +} +#endif + +#ifdef PNG_tIME_SUPPORTED +void PNGAPI +png_set_tIME(png_const_structrp png_ptr, png_inforp info_ptr, + png_const_timep mod_time) +{ + png_debug1(1, "in %s storage function", "tIME"); + + if (png_ptr == NULL || info_ptr == NULL || mod_time == NULL || + (png_ptr->mode & PNG_WROTE_tIME) != 0) + return; + + if (mod_time->month == 0 || mod_time->month > 12 || + mod_time->day == 0 || mod_time->day > 31 || + mod_time->hour > 23 || mod_time->minute > 59 || + mod_time->second > 60) + { + png_warning(png_ptr, "Ignoring invalid time value"); + return; + } + + info_ptr->mod_time = *mod_time; + info_ptr->valid |= PNG_INFO_tIME; +} +#endif + +#ifdef PNG_tRNS_SUPPORTED +void PNGAPI +png_set_tRNS(png_structrp png_ptr, png_inforp info_ptr, + png_const_bytep trans_alpha, int num_trans, png_const_color_16p trans_color) +{ + png_debug1(1, "in %s storage function", "tRNS"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + if (trans_alpha != NULL) + { + /* It may not actually be necessary to set png_ptr->trans_alpha here; + * we do it for backward compatibility with the way the png_handle_tRNS + * function used to do the allocation. + * + * 1.6.0: The above statement is incorrect; png_handle_tRNS effectively + * relies on png_set_tRNS storing the information in png_struct + * (otherwise it won't be there for the code in pngrtran.c). + */ + + png_free_data(png_ptr, info_ptr, PNG_FREE_TRNS, 0); + + /* Changed from num_trans to PNG_MAX_PALETTE_LENGTH in version 1.2.1 */ + png_ptr->trans_alpha = info_ptr->trans_alpha = png_voidcast(png_bytep, + png_malloc(png_ptr, PNG_MAX_PALETTE_LENGTH)); + + if (num_trans > 0 && num_trans <= PNG_MAX_PALETTE_LENGTH) + memcpy(info_ptr->trans_alpha, trans_alpha, (png_size_t)num_trans); + } + + if (trans_color != NULL) + { + int sample_max = (1 << info_ptr->bit_depth); + + if ((info_ptr->color_type == PNG_COLOR_TYPE_GRAY && + trans_color->gray > sample_max) || + (info_ptr->color_type == PNG_COLOR_TYPE_RGB && + (trans_color->red > sample_max || + trans_color->green > sample_max || + trans_color->blue > sample_max))) + png_warning(png_ptr, + "tRNS chunk has out-of-range samples for bit_depth"); + + info_ptr->trans_color = *trans_color; + + if (num_trans == 0) + num_trans = 1; + } + + info_ptr->num_trans = (png_uint_16)num_trans; + + if (num_trans != 0) + { + info_ptr->valid |= PNG_INFO_tRNS; + info_ptr->free_me |= PNG_FREE_TRNS; + } +} +#endif + +#ifdef PNG_sPLT_SUPPORTED +void PNGAPI +png_set_sPLT(png_const_structrp png_ptr, + png_inforp info_ptr, png_const_sPLT_tp entries, int nentries) +/* + * entries - array of png_sPLT_t structures + * to be added to the list of palettes + * in the info structure. + * + * nentries - number of palette structures to be + * added. + */ +{ + png_sPLT_tp np; + + if (png_ptr == NULL || info_ptr == NULL || nentries <= 0 || entries == NULL) + return; + + /* Use the internal realloc function, which checks for all the possible + * overflows. Notice that the parameters are (int) and (size_t) + */ + np = png_voidcast(png_sPLT_tp,png_realloc_array(png_ptr, + info_ptr->splt_palettes, info_ptr->splt_palettes_num, nentries, + sizeof *np)); + + if (np == NULL) + { + /* Out of memory or too many chunks */ + png_chunk_report(png_ptr, "too many sPLT chunks", PNG_CHUNK_WRITE_ERROR); + return; + } + + png_free(png_ptr, info_ptr->splt_palettes); + info_ptr->splt_palettes = np; + info_ptr->free_me |= PNG_FREE_SPLT; + + np += info_ptr->splt_palettes_num; + + do + { + png_size_t length; + + /* Skip invalid input entries */ + if (entries->name == NULL || entries->entries == NULL) + { + /* png_handle_sPLT doesn't do this, so this is an app error */ + png_app_error(png_ptr, "png_set_sPLT: invalid sPLT"); + /* Just skip the invalid entry */ + continue; + } + + np->depth = entries->depth; + + /* In the event of out-of-memory just return - there's no point keeping + * on trying to add sPLT chunks. + */ + length = strlen(entries->name) + 1; + np->name = png_voidcast(png_charp, png_malloc_base(png_ptr, length)); + + if (np->name == NULL) + break; + + memcpy(np->name, entries->name, length); + + /* IMPORTANT: we have memory now that won't get freed if something else + * goes wrong; this code must free it. png_malloc_array produces no + * warnings; use a png_chunk_report (below) if there is an error. + */ + np->entries = png_voidcast(png_sPLT_entryp, png_malloc_array(png_ptr, + entries->nentries, sizeof (png_sPLT_entry))); + + if (np->entries == NULL) + { + png_free(png_ptr, np->name); + np->name = NULL; + break; + } + + np->nentries = entries->nentries; + /* This multiply can't overflow because png_malloc_array has already + * checked it when doing the allocation. + */ + memcpy(np->entries, entries->entries, + entries->nentries * sizeof (png_sPLT_entry)); + + /* Note that 'continue' skips the advance of the out pointer and out + * count, so an invalid entry is not added. + */ + info_ptr->valid |= PNG_INFO_sPLT; + ++(info_ptr->splt_palettes_num); + ++np; + } + while (++entries, --nentries); + + if (nentries > 0) + png_chunk_report(png_ptr, "sPLT out of memory", PNG_CHUNK_WRITE_ERROR); +} +#endif /* sPLT */ + +#ifdef PNG_STORE_UNKNOWN_CHUNKS_SUPPORTED +static png_byte +check_location(png_const_structrp png_ptr, int location) +{ + location &= (PNG_HAVE_IHDR|PNG_HAVE_PLTE|PNG_AFTER_IDAT); + + /* New in 1.6.0; copy the location and check it. This is an API + * change; previously the app had to use the + * png_set_unknown_chunk_location API below for each chunk. + */ + if (location == 0 && (png_ptr->mode & PNG_IS_READ_STRUCT) == 0) + { + /* Write struct, so unknown chunks come from the app */ + png_app_warning(png_ptr, + "png_set_unknown_chunks now expects a valid location"); + /* Use the old behavior */ + location = (png_byte)(png_ptr->mode & + (PNG_HAVE_IHDR|PNG_HAVE_PLTE|PNG_AFTER_IDAT)); + } + + /* This need not be an internal error - if the app calls + * png_set_unknown_chunks on a read pointer it must get the location right. + */ + if (location == 0) + png_error(png_ptr, "invalid location in png_set_unknown_chunks"); + + /* Now reduce the location to the top-most set bit by removing each least + * significant bit in turn. + */ + while (location != (location & -location)) + location &= ~(location & -location); + + /* The cast is safe because 'location' is a bit mask and only the low four + * bits are significant. + */ + return (png_byte)location; +} + +void PNGAPI +png_set_unknown_chunks(png_const_structrp png_ptr, + png_inforp info_ptr, png_const_unknown_chunkp unknowns, int num_unknowns) +{ + png_unknown_chunkp np; + + if (png_ptr == NULL || info_ptr == NULL || num_unknowns <= 0 || + unknowns == NULL) + return; + + /* Check for the failure cases where support has been disabled at compile + * time. This code is hardly ever compiled - it's here because + * STORE_UNKNOWN_CHUNKS is set by both read and write code (compiling in this + * code) but may be meaningless if the read or write handling of unknown + * chunks is not compiled in. + */ +# if !defined(PNG_READ_UNKNOWN_CHUNKS_SUPPORTED) && \ + defined(PNG_READ_SUPPORTED) + if ((png_ptr->mode & PNG_IS_READ_STRUCT) != 0) + { + png_app_error(png_ptr, "no unknown chunk support on read"); + return; + } +# endif +# if !defined(PNG_WRITE_UNKNOWN_CHUNKS_SUPPORTED) && \ + defined(PNG_WRITE_SUPPORTED) + if ((png_ptr->mode & PNG_IS_READ_STRUCT) == 0) + { + png_app_error(png_ptr, "no unknown chunk support on write"); + return; + } +# endif + + /* Prior to 1.6.0 this code used png_malloc_warn; however, this meant that + * unknown critical chunks could be lost with just a warning resulting in + * undefined behavior. Now png_chunk_report is used to provide behavior + * appropriate to read or write. + */ + np = png_voidcast(png_unknown_chunkp, png_realloc_array(png_ptr, + info_ptr->unknown_chunks, info_ptr->unknown_chunks_num, num_unknowns, + sizeof *np)); + + if (np == NULL) + { + png_chunk_report(png_ptr, "too many unknown chunks", + PNG_CHUNK_WRITE_ERROR); + return; + } + + png_free(png_ptr, info_ptr->unknown_chunks); + info_ptr->unknown_chunks = np; /* safe because it is initialized */ + info_ptr->free_me |= PNG_FREE_UNKN; + + np += info_ptr->unknown_chunks_num; + + /* Increment unknown_chunks_num each time round the loop to protect the + * just-allocated chunk data. + */ + for (; num_unknowns > 0; --num_unknowns, ++unknowns) + { + memcpy(np->name, unknowns->name, (sizeof np->name)); + np->name[(sizeof np->name)-1] = '\0'; + np->location = check_location(png_ptr, unknowns->location); + + if (unknowns->size == 0) + { + np->data = NULL; + np->size = 0; + } + + else + { + np->data = png_voidcast(png_bytep, + png_malloc_base(png_ptr, unknowns->size)); + + if (np->data == NULL) + { + png_chunk_report(png_ptr, "unknown chunk: out of memory", + PNG_CHUNK_WRITE_ERROR); + /* But just skip storing the unknown chunk */ + continue; + } + + memcpy(np->data, unknowns->data, unknowns->size); + np->size = unknowns->size; + } + + /* These increments are skipped on out-of-memory for the data - the + * unknown chunk entry gets overwritten if the png_chunk_report returns. + * This is correct in the read case (the chunk is just dropped.) + */ + ++np; + ++(info_ptr->unknown_chunks_num); + } +} + +void PNGAPI +png_set_unknown_chunk_location(png_const_structrp png_ptr, png_inforp info_ptr, + int chunk, int location) +{ + /* This API is pretty pointless in 1.6.0 because the location can be set + * before the call to png_set_unknown_chunks. + * + * TODO: add a png_app_warning in 1.7 + */ + if (png_ptr != NULL && info_ptr != NULL && chunk >= 0 && + chunk < info_ptr->unknown_chunks_num) + { + if ((location & (PNG_HAVE_IHDR|PNG_HAVE_PLTE|PNG_AFTER_IDAT)) == 0) + { + png_app_error(png_ptr, "invalid unknown chunk location"); + /* Fake out the pre 1.6.0 behavior: */ + if ((location & PNG_HAVE_IDAT) != 0) /* undocumented! */ + location = PNG_AFTER_IDAT; + + else + location = PNG_HAVE_IHDR; /* also undocumented */ + } + + info_ptr->unknown_chunks[chunk].location = + check_location(png_ptr, location); + } +} +#endif + + +#ifdef PNG_MNG_FEATURES_SUPPORTED +png_uint_32 PNGAPI +png_permit_mng_features (png_structrp png_ptr, png_uint_32 mng_features) +{ + png_debug(1, "in png_permit_mng_features"); + + if (png_ptr == NULL) + return 0; + + png_ptr->mng_features_permitted = mng_features & PNG_ALL_MNG_FEATURES; + + return png_ptr->mng_features_permitted; +} +#endif + +#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED +static unsigned int +add_one_chunk(png_bytep list, unsigned int count, png_const_bytep add, int keep) +{ + unsigned int i; + + /* Utility function: update the 'keep' state of a chunk if it is already in + * the list, otherwise add it to the list. + */ + for (i=0; i= PNG_HANDLE_CHUNK_LAST) + { + png_app_error(png_ptr, "png_set_keep_unknown_chunks: invalid keep"); + return; + } + + if (num_chunks_in <= 0) + { + png_ptr->unknown_default = keep; + + /* '0' means just set the flags, so stop here */ + if (num_chunks_in == 0) + return; + } + + if (num_chunks_in < 0) + { + /* Ignore all unknown chunks and all chunks recognized by + * libpng except for IHDR, PLTE, tRNS, IDAT, and IEND + */ + static PNG_CONST png_byte chunks_to_ignore[] = { + 98, 75, 71, 68, '\0', /* bKGD */ + 99, 72, 82, 77, '\0', /* cHRM */ + 103, 65, 77, 65, '\0', /* gAMA */ + 104, 73, 83, 84, '\0', /* hIST */ + 105, 67, 67, 80, '\0', /* iCCP */ + 105, 84, 88, 116, '\0', /* iTXt */ + 111, 70, 70, 115, '\0', /* oFFs */ + 112, 67, 65, 76, '\0', /* pCAL */ + 112, 72, 89, 115, '\0', /* pHYs */ + 115, 66, 73, 84, '\0', /* sBIT */ + 115, 67, 65, 76, '\0', /* sCAL */ + 115, 80, 76, 84, '\0', /* sPLT */ + 115, 84, 69, 82, '\0', /* sTER */ + 115, 82, 71, 66, '\0', /* sRGB */ + 116, 69, 88, 116, '\0', /* tEXt */ + 116, 73, 77, 69, '\0', /* tIME */ + 122, 84, 88, 116, '\0' /* zTXt */ + }; + + chunk_list = chunks_to_ignore; + num_chunks = (unsigned int)/*SAFE*/(sizeof chunks_to_ignore)/5U; + } + + else /* num_chunks_in > 0 */ + { + if (chunk_list == NULL) + { + /* Prior to 1.6.0 this was silently ignored, now it is an app_error + * which can be switched off. + */ + png_app_error(png_ptr, "png_set_keep_unknown_chunks: no chunk list"); + return; + } + + num_chunks = num_chunks_in; + } + + old_num_chunks = png_ptr->num_chunk_list; + if (png_ptr->chunk_list == NULL) + old_num_chunks = 0; + + /* Since num_chunks is always restricted to UINT_MAX/5 this can't overflow. + */ + if (num_chunks + old_num_chunks > UINT_MAX/5) + { + png_app_error(png_ptr, "png_set_keep_unknown_chunks: too many chunks"); + return; + } + + /* If these chunks are being reset to the default then no more memory is + * required because add_one_chunk above doesn't extend the list if the 'keep' + * parameter is the default. + */ + if (keep != 0) + { + new_list = png_voidcast(png_bytep, png_malloc(png_ptr, + 5 * (num_chunks + old_num_chunks))); + + if (old_num_chunks > 0) + memcpy(new_list, png_ptr->chunk_list, 5*old_num_chunks); + } + + else if (old_num_chunks > 0) + new_list = png_ptr->chunk_list; + + else + new_list = NULL; + + /* Add the new chunks together with each one's handling code. If the chunk + * already exists the code is updated, otherwise the chunk is added to the + * end. (In libpng 1.6.0 order no longer matters because this code enforces + * the earlier convention that the last setting is the one that is used.) + */ + if (new_list != NULL) + { + png_const_bytep inlist; + png_bytep outlist; + unsigned int i; + + for (i=0; ichunk_list != new_list) + png_free(png_ptr, new_list); + + new_list = NULL; + } + } + + else + num_chunks = 0; + + png_ptr->num_chunk_list = num_chunks; + + if (png_ptr->chunk_list != new_list) + { + if (png_ptr->chunk_list != NULL) + png_free(png_ptr, png_ptr->chunk_list); + + png_ptr->chunk_list = new_list; + } +} +#endif + +#ifdef PNG_READ_USER_CHUNKS_SUPPORTED +void PNGAPI +png_set_read_user_chunk_fn(png_structrp png_ptr, png_voidp user_chunk_ptr, + png_user_chunk_ptr read_user_chunk_fn) +{ + png_debug(1, "in png_set_read_user_chunk_fn"); + + if (png_ptr == NULL) + return; + + png_ptr->read_user_chunk_fn = read_user_chunk_fn; + png_ptr->user_chunk_ptr = user_chunk_ptr; +} +#endif + +#ifdef PNG_INFO_IMAGE_SUPPORTED +void PNGAPI +png_set_rows(png_const_structrp png_ptr, png_inforp info_ptr, + png_bytepp row_pointers) +{ + png_debug1(1, "in %s storage function", "rows"); + + if (png_ptr == NULL || info_ptr == NULL) + return; + + if (info_ptr->row_pointers != NULL && + (info_ptr->row_pointers != row_pointers)) + png_free_data(png_ptr, info_ptr, PNG_FREE_ROWS, 0); + + info_ptr->row_pointers = row_pointers; + + if (row_pointers != NULL) + info_ptr->valid |= PNG_INFO_IDAT; +} +#endif + +void PNGAPI +png_set_compression_buffer_size(png_structrp png_ptr, png_size_t size) +{ + if (png_ptr == NULL) + return; + + if (size == 0 || size > PNG_UINT_31_MAX) + png_error(png_ptr, "invalid compression buffer size"); + +# ifdef PNG_SEQUENTIAL_READ_SUPPORTED + if ((png_ptr->mode & PNG_IS_READ_STRUCT) != 0) + { + png_ptr->IDAT_read_size = (png_uint_32)size; /* checked above */ + return; + } +# endif + +# ifdef PNG_WRITE_SUPPORTED + if ((png_ptr->mode & PNG_IS_READ_STRUCT) == 0) + { + if (png_ptr->zowner != 0) + { + png_warning(png_ptr, + "Compression buffer size cannot be changed because it is in use"); + return; + } + + if (size > ZLIB_IO_MAX) + { + png_warning(png_ptr, + "Compression buffer size limited to system maximum"); + size = ZLIB_IO_MAX; /* must fit */ + } + + else if (size < 6) + { + /* Deflate will potentially go into an infinite loop on a SYNC_FLUSH + * if this is permitted. + */ + png_warning(png_ptr, + "Compression buffer size cannot be reduced below 6"); + return; + } + + if (png_ptr->zbuffer_size != size) + { + png_free_buffer_list(png_ptr, &png_ptr->zbuffer_list); + png_ptr->zbuffer_size = (uInt)size; + } + } +# endif +} + +void PNGAPI +png_set_invalid(png_const_structrp png_ptr, png_inforp info_ptr, int mask) +{ + if (png_ptr != NULL && info_ptr != NULL) + info_ptr->valid &= ~mask; +} + + +#ifdef PNG_SET_USER_LIMITS_SUPPORTED +/* This function was added to libpng 1.2.6 */ +void PNGAPI +png_set_user_limits (png_structrp png_ptr, png_uint_32 user_width_max, + png_uint_32 user_height_max) +{ + /* Images with dimensions larger than these limits will be + * rejected by png_set_IHDR(). To accept any PNG datastream + * regardless of dimensions, set both limits to 0x7ffffffL. + */ + if (png_ptr == NULL) + return; + + png_ptr->user_width_max = user_width_max; + png_ptr->user_height_max = user_height_max; +} + +/* This function was added to libpng 1.4.0 */ +void PNGAPI +png_set_chunk_cache_max (png_structrp png_ptr, png_uint_32 user_chunk_cache_max) +{ + if (png_ptr != NULL) + png_ptr->user_chunk_cache_max = user_chunk_cache_max; +} + +/* This function was added to libpng 1.4.1 */ +void PNGAPI +png_set_chunk_malloc_max (png_structrp png_ptr, + png_alloc_size_t user_chunk_malloc_max) +{ + if (png_ptr != NULL) + png_ptr->user_chunk_malloc_max = user_chunk_malloc_max; +} +#endif /* ?SET_USER_LIMITS */ + + +#ifdef PNG_BENIGN_ERRORS_SUPPORTED +void PNGAPI +png_set_benign_errors(png_structrp png_ptr, int allowed) +{ + png_debug(1, "in png_set_benign_errors"); + + /* If allowed is 1, png_benign_error() is treated as a warning. + * + * If allowed is 0, png_benign_error() is treated as an error (which + * is the default behavior if png_set_benign_errors() is not called). + */ + + if (allowed != 0) + png_ptr->flags |= PNG_FLAG_BENIGN_ERRORS_WARN | + PNG_FLAG_APP_WARNINGS_WARN | PNG_FLAG_APP_ERRORS_WARN; + + else + png_ptr->flags &= ~(PNG_FLAG_BENIGN_ERRORS_WARN | + PNG_FLAG_APP_WARNINGS_WARN | PNG_FLAG_APP_ERRORS_WARN); +} +#endif /* BENIGN_ERRORS */ + +#ifdef PNG_CHECK_FOR_INVALID_INDEX_SUPPORTED + /* Whether to report invalid palette index; added at libng-1.5.10. + * It is possible for an indexed (color-type==3) PNG file to contain + * pixels with invalid (out-of-range) indexes if the PLTE chunk has + * fewer entries than the image's bit-depth would allow. We recover + * from this gracefully by filling any incomplete palette with zeros + * (opaque black). By default, when this occurs libpng will issue + * a benign error. This API can be used to override that behavior. + */ +void PNGAPI +png_set_check_for_invalid_index(png_structrp png_ptr, int allowed) +{ + png_debug(1, "in png_set_check_for_invalid_index"); + + if (allowed > 0) + png_ptr->num_palette_max = 0; + + else + png_ptr->num_palette_max = -1; +} +#endif +#endif /* READ || WRITE */ diff --git a/app/src/main/jni/libpng/pngstruct.h b/app/src/main/jni/libpng/pngstruct.h new file mode 100644 index 000000000..c1dd026a9 --- /dev/null +++ b/app/src/main/jni/libpng/pngstruct.h @@ -0,0 +1,489 @@ + +/* pngstruct.h - header file for PNG reference library + * + * Copyright (c) 1998-2013 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * Last changed in libpng 1.6.1 [March 28, 2013] + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + */ + +/* The structure that holds the information to read and write PNG files. + * The only people who need to care about what is inside of this are the + * people who will be modifying the library for their own special needs. + * It should NOT be accessed directly by an application. + */ + +#ifndef PNGSTRUCT_H +#define PNGSTRUCT_H +/* zlib.h defines the structure z_stream, an instance of which is included + * in this structure and is required for decompressing the LZ compressed + * data in PNG files. + */ +#ifndef ZLIB_CONST + /* We must ensure that zlib uses 'const' in declarations. */ +# define ZLIB_CONST +#endif +#include "zlib.h" +#ifdef const + /* zlib.h sometimes #defines const to nothing, undo this. */ +# undef const +#endif + +/* zlib.h has mediocre z_const use before 1.2.6, this stuff is for compatibility + * with older builds. + */ +#if ZLIB_VERNUM < 0x1260 +# define PNGZ_MSG_CAST(s) png_constcast(char*,s) +# define PNGZ_INPUT_CAST(b) png_constcast(png_bytep,b) +#else +# define PNGZ_MSG_CAST(s) (s) +# define PNGZ_INPUT_CAST(b) (b) +#endif + +/* zlib.h declares a magic type 'uInt' that limits the amount of data that zlib + * can handle at once. This type need be no larger than 16 bits (so maximum of + * 65535), this define allows us to discover how big it is, but limited by the + * maximuum for png_size_t. The value can be overriden in a library build + * (pngusr.h, or set it in CPPFLAGS) and it works to set it to a considerably + * lower value (e.g. 255 works). A lower value may help memory usage (slightly) + * and may even improve performance on some systems (and degrade it on others.) + */ +#ifndef ZLIB_IO_MAX +# define ZLIB_IO_MAX ((uInt)-1) +#endif + +#ifdef PNG_WRITE_SUPPORTED +/* The type of a compression buffer list used by the write code. */ +typedef struct png_compression_buffer +{ + struct png_compression_buffer *next; + png_byte output[1]; /* actually zbuf_size */ +} png_compression_buffer, *png_compression_bufferp; + +#define PNG_COMPRESSION_BUFFER_SIZE(pp)\ + (offsetof(png_compression_buffer, output) + (pp)->zbuffer_size) +#endif + +/* Colorspace support; structures used in png_struct, png_info and in internal + * functions to hold and communicate information about the color space. + * + * PNG_COLORSPACE_SUPPORTED is only required if the application will perform + * colorspace corrections, otherwise all the colorspace information can be + * skipped and the size of libpng can be reduced (significantly) by compiling + * out the colorspace support. + */ +#ifdef PNG_COLORSPACE_SUPPORTED +/* The chromaticities of the red, green and blue colorants and the chromaticity + * of the corresponding white point (i.e. of rgb(1.0,1.0,1.0)). + */ +typedef struct png_xy +{ + png_fixed_point redx, redy; + png_fixed_point greenx, greeny; + png_fixed_point bluex, bluey; + png_fixed_point whitex, whitey; +} png_xy; + +/* The same data as above but encoded as CIE XYZ values. When this data comes + * from chromaticities the sum of the Y values is assumed to be 1.0 + */ +typedef struct png_XYZ +{ + png_fixed_point red_X, red_Y, red_Z; + png_fixed_point green_X, green_Y, green_Z; + png_fixed_point blue_X, blue_Y, blue_Z; +} png_XYZ; +#endif /* COLORSPACE */ + +#if defined(PNG_COLORSPACE_SUPPORTED) || defined(PNG_GAMMA_SUPPORTED) +/* A colorspace is all the above plus, potentially, profile information, + * however at present libpng does not use the profile internally so it is only + * stored in the png_info struct (if iCCP is supported.) The rendering intent + * is retained here and is checked. + * + * The file gamma encoding information is also stored here and gamma correction + * is done by libpng, whereas color correction must currently be done by the + * application. + */ +typedef struct png_colorspace +{ +#ifdef PNG_GAMMA_SUPPORTED + png_fixed_point gamma; /* File gamma */ +#endif + +#ifdef PNG_COLORSPACE_SUPPORTED + png_xy end_points_xy; /* End points as chromaticities */ + png_XYZ end_points_XYZ; /* End points as CIE XYZ colorant values */ + png_uint_16 rendering_intent; /* Rendering intent of a profile */ +#endif + + /* Flags are always defined to simplify the code. */ + png_uint_16 flags; /* As defined below */ +} png_colorspace, * PNG_RESTRICT png_colorspacerp; + +typedef const png_colorspace * PNG_RESTRICT png_const_colorspacerp; + +/* General flags for the 'flags' field */ +#define PNG_COLORSPACE_HAVE_GAMMA 0x0001 +#define PNG_COLORSPACE_HAVE_ENDPOINTS 0x0002 +#define PNG_COLORSPACE_HAVE_INTENT 0x0004 +#define PNG_COLORSPACE_FROM_gAMA 0x0008 +#define PNG_COLORSPACE_FROM_cHRM 0x0010 +#define PNG_COLORSPACE_FROM_sRGB 0x0020 +#define PNG_COLORSPACE_ENDPOINTS_MATCH_sRGB 0x0040 +#define PNG_COLORSPACE_MATCHES_sRGB 0x0080 /* exact match on profile */ +#define PNG_COLORSPACE_INVALID 0x8000 +#define PNG_COLORSPACE_CANCEL(flags) (0xffff ^ (flags)) +#endif /* COLORSPACE || GAMMA */ + +struct png_struct_def +{ +#ifdef PNG_SETJMP_SUPPORTED + jmp_buf jmp_buf_local; /* New name in 1.6.0 for jmp_buf in png_struct */ + png_longjmp_ptr longjmp_fn;/* setjmp non-local goto function. */ + jmp_buf *jmp_buf_ptr; /* passed to longjmp_fn */ + size_t jmp_buf_size; /* size of the above, if allocated */ +#endif + png_error_ptr error_fn; /* function for printing errors and aborting */ +#ifdef PNG_WARNINGS_SUPPORTED + png_error_ptr warning_fn; /* function for printing warnings */ +#endif + png_voidp error_ptr; /* user supplied struct for error functions */ + png_rw_ptr write_data_fn; /* function for writing output data */ + png_rw_ptr read_data_fn; /* function for reading input data */ + png_voidp io_ptr; /* ptr to application struct for I/O functions */ + +#ifdef PNG_READ_USER_TRANSFORM_SUPPORTED + png_user_transform_ptr read_user_transform_fn; /* user read transform */ +#endif + +#ifdef PNG_WRITE_USER_TRANSFORM_SUPPORTED + png_user_transform_ptr write_user_transform_fn; /* user write transform */ +#endif + +/* These were added in libpng-1.0.2 */ +#ifdef PNG_USER_TRANSFORM_PTR_SUPPORTED +#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED) || \ + defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED) + png_voidp user_transform_ptr; /* user supplied struct for user transform */ + png_byte user_transform_depth; /* bit depth of user transformed pixels */ + png_byte user_transform_channels; /* channels in user transformed pixels */ +#endif +#endif + + png_uint_32 mode; /* tells us where we are in the PNG file */ + png_uint_32 flags; /* flags indicating various things to libpng */ + png_uint_32 transformations; /* which transformations to perform */ + + png_uint_32 zowner; /* ID (chunk type) of zstream owner, 0 if none */ + z_stream zstream; /* decompression structure */ + +#ifdef PNG_WRITE_SUPPORTED + png_compression_bufferp zbuffer_list; /* Created on demand during write */ + uInt zbuffer_size; /* size of the actual buffer */ + + int zlib_level; /* holds zlib compression level */ + int zlib_method; /* holds zlib compression method */ + int zlib_window_bits; /* holds zlib compression window bits */ + int zlib_mem_level; /* holds zlib compression memory level */ + int zlib_strategy; /* holds zlib compression strategy */ +#endif +/* Added at libpng 1.5.4 */ +#ifdef PNG_WRITE_CUSTOMIZE_ZTXT_COMPRESSION_SUPPORTED + int zlib_text_level; /* holds zlib compression level */ + int zlib_text_method; /* holds zlib compression method */ + int zlib_text_window_bits; /* holds zlib compression window bits */ + int zlib_text_mem_level; /* holds zlib compression memory level */ + int zlib_text_strategy; /* holds zlib compression strategy */ +#endif +/* End of material added at libpng 1.5.4 */ +/* Added at libpng 1.6.0 */ +#ifdef PNG_WRITE_SUPPORTED + int zlib_set_level; /* Actual values set into the zstream on write */ + int zlib_set_method; + int zlib_set_window_bits; + int zlib_set_mem_level; + int zlib_set_strategy; +#endif + + png_uint_32 width; /* width of image in pixels */ + png_uint_32 height; /* height of image in pixels */ + png_uint_32 num_rows; /* number of rows in current pass */ + png_uint_32 usr_width; /* width of row at start of write */ + png_size_t rowbytes; /* size of row in bytes */ + png_uint_32 iwidth; /* width of current interlaced row in pixels */ + png_uint_32 row_number; /* current row in interlace pass */ + png_uint_32 chunk_name; /* PNG_CHUNK() id of current chunk */ + png_bytep prev_row; /* buffer to save previous (unfiltered) row. + * This is a pointer into big_prev_row + */ + png_bytep row_buf; /* buffer to save current (unfiltered) row. + * This is a pointer into big_row_buf + */ +#ifdef PNG_WRITE_SUPPORTED + png_bytep sub_row; /* buffer to save "sub" row when filtering */ + png_bytep up_row; /* buffer to save "up" row when filtering */ + png_bytep avg_row; /* buffer to save "avg" row when filtering */ + png_bytep paeth_row; /* buffer to save "Paeth" row when filtering */ +#endif + png_size_t info_rowbytes; /* Added in 1.5.4: cache of updated row bytes */ + + png_uint_32 idat_size; /* current IDAT size for read */ + png_uint_32 crc; /* current chunk CRC value */ + png_colorp palette; /* palette from the input file */ + png_uint_16 num_palette; /* number of color entries in palette */ + +/* Added at libpng-1.5.10 */ +#ifdef PNG_CHECK_FOR_INVALID_INDEX_SUPPORTED + int num_palette_max; /* maximum palette index found in IDAT */ +#endif + + png_uint_16 num_trans; /* number of transparency values */ + png_byte compression; /* file compression type (always 0) */ + png_byte filter; /* file filter type (always 0) */ + png_byte interlaced; /* PNG_INTERLACE_NONE, PNG_INTERLACE_ADAM7 */ + png_byte pass; /* current interlace pass (0 - 6) */ + png_byte do_filter; /* row filter flags (see PNG_FILTER_ below ) */ + png_byte color_type; /* color type of file */ + png_byte bit_depth; /* bit depth of file */ + png_byte usr_bit_depth; /* bit depth of users row: write only */ + png_byte pixel_depth; /* number of bits per pixel */ + png_byte channels; /* number of channels in file */ +#ifdef PNG_WRITE_SUPPORTED + png_byte usr_channels; /* channels at start of write: write only */ +#endif + png_byte sig_bytes; /* magic bytes read/written from start of file */ + png_byte maximum_pixel_depth; + /* pixel depth used for the row buffers */ + png_byte transformed_pixel_depth; + /* pixel depth after read/write transforms */ +#if defined(PNG_READ_FILLER_SUPPORTED) || defined(PNG_WRITE_FILLER_SUPPORTED) + png_uint_16 filler; /* filler bytes for pixel expansion */ +#endif + +#if defined(PNG_bKGD_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED) ||\ + defined(PNG_READ_ALPHA_MODE_SUPPORTED) + png_byte background_gamma_type; + png_fixed_point background_gamma; + png_color_16 background; /* background color in screen gamma space */ +#ifdef PNG_READ_GAMMA_SUPPORTED + png_color_16 background_1; /* background normalized to gamma 1.0 */ +#endif +#endif /* bKGD */ + +#ifdef PNG_WRITE_FLUSH_SUPPORTED + png_flush_ptr output_flush_fn; /* Function for flushing output */ + png_uint_32 flush_dist; /* how many rows apart to flush, 0 - no flush */ + png_uint_32 flush_rows; /* number of rows written since last flush */ +#endif + +#ifdef PNG_READ_GAMMA_SUPPORTED + int gamma_shift; /* number of "insignificant" bits in 16-bit gamma */ + png_fixed_point screen_gamma; /* screen gamma value (display_exponent) */ + + png_bytep gamma_table; /* gamma table for 8-bit depth files */ + png_uint_16pp gamma_16_table; /* gamma table for 16-bit depth files */ +#if defined(PNG_READ_BACKGROUND_SUPPORTED) || \ + defined(PNG_READ_ALPHA_MODE_SUPPORTED) || \ + defined(PNG_READ_RGB_TO_GRAY_SUPPORTED) + png_bytep gamma_from_1; /* converts from 1.0 to screen */ + png_bytep gamma_to_1; /* converts from file to 1.0 */ + png_uint_16pp gamma_16_from_1; /* converts from 1.0 to screen */ + png_uint_16pp gamma_16_to_1; /* converts from file to 1.0 */ +#endif /* READ_BACKGROUND || READ_ALPHA_MODE || RGB_TO_GRAY */ +#endif + +#if defined(PNG_READ_GAMMA_SUPPORTED) || defined(PNG_sBIT_SUPPORTED) + png_color_8 sig_bit; /* significant bits in each available channel */ +#endif + +#if defined(PNG_READ_SHIFT_SUPPORTED) || defined(PNG_WRITE_SHIFT_SUPPORTED) + png_color_8 shift; /* shift for significant bit tranformation */ +#endif + +#if defined(PNG_tRNS_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED) \ + || defined(PNG_READ_EXPAND_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED) + png_bytep trans_alpha; /* alpha values for paletted files */ + png_color_16 trans_color; /* transparent color for non-paletted files */ +#endif + + png_read_status_ptr read_row_fn; /* called after each row is decoded */ + png_write_status_ptr write_row_fn; /* called after each row is encoded */ +#ifdef PNG_PROGRESSIVE_READ_SUPPORTED + png_progressive_info_ptr info_fn; /* called after header data fully read */ + png_progressive_row_ptr row_fn; /* called after a prog. row is decoded */ + png_progressive_end_ptr end_fn; /* called after image is complete */ + png_bytep save_buffer_ptr; /* current location in save_buffer */ + png_bytep save_buffer; /* buffer for previously read data */ + png_bytep current_buffer_ptr; /* current location in current_buffer */ + png_bytep current_buffer; /* buffer for recently used data */ + png_uint_32 push_length; /* size of current input chunk */ + png_uint_32 skip_length; /* bytes to skip in input data */ + png_size_t save_buffer_size; /* amount of data now in save_buffer */ + png_size_t save_buffer_max; /* total size of save_buffer */ + png_size_t buffer_size; /* total amount of available input data */ + png_size_t current_buffer_size; /* amount of data now in current_buffer */ + int process_mode; /* what push library is currently doing */ + int cur_palette; /* current push library palette index */ + +#endif /* PROGRESSIVE_READ */ + +#if defined(__TURBOC__) && !defined(_Windows) && !defined(__FLAT__) +/* For the Borland special 64K segment handler */ + png_bytepp offset_table_ptr; + png_bytep offset_table; + png_uint_16 offset_table_number; + png_uint_16 offset_table_count; + png_uint_16 offset_table_count_free; +#endif + +#ifdef PNG_READ_QUANTIZE_SUPPORTED + png_bytep palette_lookup; /* lookup table for quantizing */ + png_bytep quantize_index; /* index translation for palette files */ +#endif + +#ifdef PNG_WRITE_WEIGHTED_FILTER_SUPPORTED + png_byte heuristic_method; /* heuristic for row filter selection */ + png_byte num_prev_filters; /* number of weights for previous rows */ + png_bytep prev_filters; /* filter type(s) of previous row(s) */ + png_uint_16p filter_weights; /* weight(s) for previous line(s) */ + png_uint_16p inv_filter_weights; /* 1/weight(s) for previous line(s) */ + png_uint_16p filter_costs; /* relative filter calculation cost */ + png_uint_16p inv_filter_costs; /* 1/relative filter calculation cost */ +#endif + + /* Options */ +#ifdef PNG_SET_OPTION_SUPPORTED + png_byte options; /* On/off state (up to 4 options) */ +#endif + +#if PNG_LIBPNG_VER < 10700 +/* To do: remove this from libpng-1.7 */ +#ifdef PNG_TIME_RFC1123_SUPPORTED + char time_buffer[29]; /* String to hold RFC 1123 time text */ +#endif +#endif + +/* New members added in libpng-1.0.6 */ + + png_uint_32 free_me; /* flags items libpng is responsible for freeing */ + +#ifdef PNG_USER_CHUNKS_SUPPORTED + png_voidp user_chunk_ptr; +#ifdef PNG_READ_USER_CHUNKS_SUPPORTED + png_user_chunk_ptr read_user_chunk_fn; /* user read chunk handler */ +#endif +#endif + +#ifdef PNG_SET_UNKNOWN_CHUNKS_SUPPORTED + int unknown_default; /* As PNG_HANDLE_* */ + unsigned int num_chunk_list; /* Number of entries in the list */ + png_bytep chunk_list; /* List of png_byte[5]; the textual chunk name + * followed by a PNG_HANDLE_* byte */ +#endif + +/* New members added in libpng-1.0.3 */ +#ifdef PNG_READ_RGB_TO_GRAY_SUPPORTED + png_byte rgb_to_gray_status; + /* Added in libpng 1.5.5 to record setting of coefficients: */ + png_byte rgb_to_gray_coefficients_set; + /* These were changed from png_byte in libpng-1.0.6 */ + png_uint_16 rgb_to_gray_red_coeff; + png_uint_16 rgb_to_gray_green_coeff; + /* deleted in 1.5.5: rgb_to_gray_blue_coeff; */ +#endif + +/* New member added in libpng-1.0.4 (renamed in 1.0.9) */ +#if defined(PNG_MNG_FEATURES_SUPPORTED) +/* Changed from png_byte to png_uint_32 at version 1.2.0 */ + png_uint_32 mng_features_permitted; +#endif + +/* New member added in libpng-1.0.9, ifdef'ed out in 1.0.12, enabled in 1.2.0 */ +#ifdef PNG_MNG_FEATURES_SUPPORTED + png_byte filter_type; +#endif + +/* New members added in libpng-1.2.0 */ + +/* New members added in libpng-1.0.2 but first enabled by default in 1.2.0 */ +#ifdef PNG_USER_MEM_SUPPORTED + png_voidp mem_ptr; /* user supplied struct for mem functions */ + png_malloc_ptr malloc_fn; /* function for allocating memory */ + png_free_ptr free_fn; /* function for freeing memory */ +#endif + +/* New member added in libpng-1.0.13 and 1.2.0 */ + png_bytep big_row_buf; /* buffer to save current (unfiltered) row */ + +#ifdef PNG_READ_QUANTIZE_SUPPORTED +/* The following three members were added at version 1.0.14 and 1.2.4 */ + png_bytep quantize_sort; /* working sort array */ + png_bytep index_to_palette; /* where the original index currently is + in the palette */ + png_bytep palette_to_index; /* which original index points to this + palette color */ +#endif + +/* New members added in libpng-1.0.16 and 1.2.6 */ + png_byte compression_type; + +#ifdef PNG_USER_LIMITS_SUPPORTED + png_uint_32 user_width_max; + png_uint_32 user_height_max; + + /* Added in libpng-1.4.0: Total number of sPLT, text, and unknown + * chunks that can be stored (0 means unlimited). + */ + png_uint_32 user_chunk_cache_max; + + /* Total memory that a zTXt, sPLT, iTXt, iCCP, or unknown chunk + * can occupy when decompressed. 0 means unlimited. + */ + png_alloc_size_t user_chunk_malloc_max; +#endif + +/* New member added in libpng-1.0.25 and 1.2.17 */ +#ifdef PNG_READ_UNKNOWN_CHUNKS_SUPPORTED + /* Temporary storage for unknown chunk that the library doesn't recognize, + * used while reading the chunk. + */ + png_unknown_chunk unknown_chunk; +#endif + +/* New member added in libpng-1.2.26 */ + png_size_t old_big_row_buf_size; + +#ifdef PNG_READ_SUPPORTED +/* New member added in libpng-1.2.30 */ + png_bytep read_buffer; /* buffer for reading chunk data */ + png_alloc_size_t read_buffer_size; /* current size of the buffer */ +#endif +#ifdef PNG_SEQUENTIAL_READ_SUPPORTED + uInt IDAT_read_size; /* limit on read buffer size for IDAT */ +#endif + +#ifdef PNG_IO_STATE_SUPPORTED +/* New member added in libpng-1.4.0 */ + png_uint_32 io_state; +#endif + +/* New member added in libpng-1.5.6 */ + png_bytep big_prev_row; + +/* New member added in libpng-1.5.7 */ + void (*read_filter[PNG_FILTER_VALUE_LAST-1])(png_row_infop row_info, + png_bytep row, png_const_bytep prev_row); + +#ifdef PNG_READ_SUPPORTED +#if defined(PNG_COLORSPACE_SUPPORTED) || defined(PNG_GAMMA_SUPPORTED) + png_colorspace colorspace; +#endif +#endif +}; +#endif /* PNGSTRUCT_H */ diff --git a/app/src/main/jni/libpng/pngtrans.c b/app/src/main/jni/libpng/pngtrans.c new file mode 100644 index 000000000..56856b480 --- /dev/null +++ b/app/src/main/jni/libpng/pngtrans.c @@ -0,0 +1,849 @@ + +/* pngtrans.c - transforms the data in a row (used by both readers and writers) + * + * Last changed in libpng 1.6.15 [November 20, 2014] + * Copyright (c) 1998-2014 Glenn Randers-Pehrson + * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger) + * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.) + * + * This code is released under the libpng license. + * For conditions of distribution and use, see the disclaimer + * and license in png.h + */ + +#include "pngpriv.h" + +#if defined(PNG_READ_SUPPORTED) || defined(PNG_WRITE_SUPPORTED) + +#if defined(PNG_READ_BGR_SUPPORTED) || defined(PNG_WRITE_BGR_SUPPORTED) +/* Turn on BGR-to-RGB mapping */ +void PNGAPI +png_set_bgr(png_structrp png_ptr) +{ + png_debug(1, "in png_set_bgr"); + + if (png_ptr == NULL) + return; + + png_ptr->transformations |= PNG_BGR; +} +#endif + +#if defined(PNG_READ_SWAP_SUPPORTED) || defined(PNG_WRITE_SWAP_SUPPORTED) +/* Turn on 16 bit byte swapping */ +void PNGAPI +png_set_swap(png_structrp png_ptr) +{ + png_debug(1, "in png_set_swap"); + + if (png_ptr == NULL) + return; + + if (png_ptr->bit_depth == 16) + png_ptr->transformations |= PNG_SWAP_BYTES; +} +#endif + +#if defined(PNG_READ_PACK_SUPPORTED) || defined(PNG_WRITE_PACK_SUPPORTED) +/* Turn on pixel packing */ +void PNGAPI +png_set_packing(png_structrp png_ptr) +{ + png_debug(1, "in png_set_packing"); + + if (png_ptr == NULL) + return; + + if (png_ptr->bit_depth < 8) + { + png_ptr->transformations |= PNG_PACK; +# ifdef PNG_WRITE_SUPPORTED + png_ptr->usr_bit_depth = 8; +# endif + } +} +#endif + +#if defined(PNG_READ_PACKSWAP_SUPPORTED)||defined(PNG_WRITE_PACKSWAP_SUPPORTED) +/* Turn on packed pixel swapping */ +void PNGAPI +png_set_packswap(png_structrp png_ptr) +{ + png_debug(1, "in png_set_packswap"); + + if (png_ptr == NULL) + return; + + if (png_ptr->bit_depth < 8) + png_ptr->transformations |= PNG_PACKSWAP; +} +#endif + +#if defined(PNG_READ_SHIFT_SUPPORTED) || defined(PNG_WRITE_SHIFT_SUPPORTED) +void PNGAPI +png_set_shift(png_structrp png_ptr, png_const_color_8p true_bits) +{ + png_debug(1, "in png_set_shift"); + + if (png_ptr == NULL) + return; + + png_ptr->transformations |= PNG_SHIFT; + png_ptr->shift = *true_bits; +} +#endif + +#if defined(PNG_READ_INTERLACING_SUPPORTED) || \ + defined(PNG_WRITE_INTERLACING_SUPPORTED) +int PNGAPI +png_set_interlace_handling(png_structrp png_ptr) +{ + png_debug(1, "in png_set_interlace handling"); + + if (png_ptr != 0 && png_ptr->interlaced != 0) + { + png_ptr->transformations |= PNG_INTERLACE; + return (7); + } + + return (1); +} +#endif + +#if defined(PNG_READ_FILLER_SUPPORTED) || defined(PNG_WRITE_FILLER_SUPPORTED) +/* Add a filler byte on read, or remove a filler or alpha byte on write. + * The filler type has changed in v0.95 to allow future 2-byte fillers + * for 48-bit input data, as well as to avoid problems with some compilers + * that don't like bytes as parameters. + */ +void PNGAPI +png_set_filler(png_structrp png_ptr, png_uint_32 filler, int filler_loc) +{ + png_debug(1, "in png_set_filler"); + + if (png_ptr == NULL) + return; + + /* In libpng 1.6 it is possible to determine whether this is a read or write + * operation and therefore to do more checking here for a valid call. + */ + if ((png_ptr->mode & PNG_IS_READ_STRUCT) != 0) + { +# ifdef PNG_READ_FILLER_SUPPORTED + /* On read png_set_filler is always valid, regardless of the base PNG + * format, because other transformations can give a format where the + * filler code can execute (basically an 8 or 16-bit component RGB or G + * format.) + * + * NOTE: usr_channels is not used by the read code! (This has led to + * confusion in the past.) The filler is only used in the read code. + */ + png_ptr->filler = (png_uint_16)filler; +# else + png_app_error(png_ptr, "png_set_filler not supported on read"); + PNG_UNUSED(filler) /* not used in the write case */ + return; +# endif + } + + else /* write */ + { +# ifdef PNG_WRITE_FILLER_SUPPORTED + /* On write the usr_channels parameter must be set correctly at the + * start to record the number of channels in the app-supplied data. + */ + switch (png_ptr->color_type) + { + case PNG_COLOR_TYPE_RGB: + png_ptr->usr_channels = 4; + break; + + case PNG_COLOR_TYPE_GRAY: + if (png_ptr->bit_depth >= 8) + { + png_ptr->usr_channels = 2; + break; + } + + else + { + /* There simply isn't any code in libpng to strip out bits + * from bytes when the components are less than a byte in + * size! + */ + png_app_error(png_ptr, + "png_set_filler is invalid for low bit depth gray output"); + return; + } + + default: + png_app_error(png_ptr, + "png_set_filler: inappropriate color type"); + return; + } +# else + png_app_error(png_ptr, "png_set_filler not supported on write"); + return; +# endif + } + + /* Here on success - libpng supports the operation, set the transformation + * and the flag to say where the filler channel is. + */ + png_ptr->transformations |= PNG_FILLER; + + if (filler_loc == PNG_FILLER_AFTER) + png_ptr->flags |= PNG_FLAG_FILLER_AFTER; + + else + png_ptr->flags &= ~PNG_FLAG_FILLER_AFTER; +} + +/* Added to libpng-1.2.7 */ +void PNGAPI +png_set_add_alpha(png_structrp png_ptr, png_uint_32 filler, int filler_loc) +{ + png_debug(1, "in png_set_add_alpha"); + + if (png_ptr == NULL) + return; + + png_set_filler(png_ptr, filler, filler_loc); + /* The above may fail to do anything. */ + if ((png_ptr->transformations & PNG_FILLER) != 0) + png_ptr->transformations |= PNG_ADD_ALPHA; +} + +#endif + +#if defined(PNG_READ_SWAP_ALPHA_SUPPORTED) || \ + defined(PNG_WRITE_SWAP_ALPHA_SUPPORTED) +void PNGAPI +png_set_swap_alpha(png_structrp png_ptr) +{ + png_debug(1, "in png_set_swap_alpha"); + + if (png_ptr == NULL) + return; + + png_ptr->transformations |= PNG_SWAP_ALPHA; +} +#endif + +#if defined(PNG_READ_INVERT_ALPHA_SUPPORTED) || \ + defined(PNG_WRITE_INVERT_ALPHA_SUPPORTED) +void PNGAPI +png_set_invert_alpha(png_structrp png_ptr) +{ + png_debug(1, "in png_set_invert_alpha"); + + if (png_ptr == NULL) + return; + + png_ptr->transformations |= PNG_INVERT_ALPHA; +} +#endif + +#if defined(PNG_READ_INVERT_SUPPORTED) || defined(PNG_WRITE_INVERT_SUPPORTED) +void PNGAPI +png_set_invert_mono(png_structrp png_ptr) +{ + png_debug(1, "in png_set_invert_mono"); + + if (png_ptr == NULL) + return; + + png_ptr->transformations |= PNG_INVERT_MONO; +} + +/* Invert monochrome grayscale data */ +void /* PRIVATE */ +png_do_invert(png_row_infop row_info, png_bytep row) +{ + png_debug(1, "in png_do_invert"); + + /* This test removed from libpng version 1.0.13 and 1.2.0: + * if (row_info->bit_depth == 1 && + */ + if (row_info->color_type == PNG_COLOR_TYPE_GRAY) + { + png_bytep rp = row; + png_size_t i; + png_size_t istop = row_info->rowbytes; + + for (i = 0; i < istop; i++) + { + *rp = (png_byte)(~(*rp)); + rp++; + } + } + + else if (row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA && + row_info->bit_depth == 8) + { + png_bytep rp = row; + png_size_t i; + png_size_t istop = row_info->rowbytes; + + for (i = 0; i < istop; i += 2) + { + *rp = (png_byte)(~(*rp)); + rp += 2; + } + } + +#ifdef PNG_16BIT_SUPPORTED + else if (row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA && + row_info->bit_depth == 16) + { + png_bytep rp = row; + png_size_t i; + png_size_t istop = row_info->rowbytes; + + for (i = 0; i < istop; i += 4) + { + *rp = (png_byte)(~(*rp)); + *(rp + 1) = (png_byte)(~(*(rp + 1))); + rp += 4; + } + } +#endif +} +#endif + +#ifdef PNG_16BIT_SUPPORTED +#if defined(PNG_READ_SWAP_SUPPORTED) || defined(PNG_WRITE_SWAP_SUPPORTED) +/* Swaps byte order on 16 bit depth images */ +void /* PRIVATE */ +png_do_swap(png_row_infop row_info, png_bytep row) +{ + png_debug(1, "in png_do_swap"); + + if (row_info->bit_depth == 16) + { + png_bytep rp = row; + png_uint_32 i; + png_uint_32 istop= row_info->width * row_info->channels; + + for (i = 0; i < istop; i++, rp += 2) + { +#ifdef PNG_BUILTIN_BSWAP16_SUPPORTED + /* Feature added to libpng-1.6.11 for testing purposes, not + * enabled by default. + */ + *(png_uint_16*)rp = __builtin_bswap16(*(png_uint_16*)rp); +#else + png_byte t = *rp; + *rp = *(rp + 1); + *(rp + 1) = t; +#endif + } + } +} +#endif +#endif + +#if defined(PNG_READ_PACKSWAP_SUPPORTED)||defined(PNG_WRITE_PACKSWAP_SUPPORTED) +static PNG_CONST png_byte onebppswaptable[256] = { + 0x00, 0x80, 0x40, 0xC0, 0x20, 0xA0, 0x60, 0xE0, + 0x10, 0x90, 0x50, 0xD0, 0x30, 0xB0, 0x70, 0xF0, + 0x08, 0x88, 0x48, 0xC8, 0x28, 0xA8, 0x68, 0xE8, + 0x18, 0x98, 0x58, 0xD8, 0x38, 0xB8, 0x78, 0xF8, + 0x04, 0x84, 0x44, 0xC4, 0x24, 0xA4, 0x64, 0xE4, + 0x14, 0x94, 0x54, 0xD4, 0x34, 0xB4, 0x74, 0xF4, + 0x0C, 0x8C, 0x4C, 0xCC, 0x2C, 0xAC, 0x6C, 0xEC, + 0x1C, 0x9C, 0x5C, 0xDC, 0x3C, 0xBC, 0x7C, 0xFC, + 0x02, 0x82, 0x42, 0xC2, 0x22, 0xA2, 0x62, 0xE2, + 0x12, 0x92, 0x52, 0xD2, 0x32, 0xB2, 0x72, 0xF2, + 0x0A, 0x8A, 0x4A, 0xCA, 0x2A, 0xAA, 0x6A, 0xEA, + 0x1A, 0x9A, 0x5A, 0xDA, 0x3A, 0xBA, 0x7A, 0xFA, + 0x06, 0x86, 0x46, 0xC6, 0x26, 0xA6, 0x66, 0xE6, + 0x16, 0x96, 0x56, 0xD6, 0x36, 0xB6, 0x76, 0xF6, + 0x0E, 0x8E, 0x4E, 0xCE, 0x2E, 0xAE, 0x6E, 0xEE, + 0x1E, 0x9E, 0x5E, 0xDE, 0x3E, 0xBE, 0x7E, 0xFE, + 0x01, 0x81, 0x41, 0xC1, 0x21, 0xA1, 0x61, 0xE1, + 0x11, 0x91, 0x51, 0xD1, 0x31, 0xB1, 0x71, 0xF1, + 0x09, 0x89, 0x49, 0xC9, 0x29, 0xA9, 0x69, 0xE9, + 0x19, 0x99, 0x59, 0xD9, 0x39, 0xB9, 0x79, 0xF9, + 0x05, 0x85, 0x45, 0xC5, 0x25, 0xA5, 0x65, 0xE5, + 0x15, 0x95, 0x55, 0xD5, 0x35, 0xB5, 0x75, 0xF5, + 0x0D, 0x8D, 0x4D, 0xCD, 0x2D, 0xAD, 0x6D, 0xED, + 0x1D, 0x9D, 0x5D, 0xDD, 0x3D, 0xBD, 0x7D, 0xFD, + 0x03, 0x83, 0x43, 0xC3, 0x23, 0xA3, 0x63, 0xE3, + 0x13, 0x93, 0x53, 0xD3, 0x33, 0xB3, 0x73, 0xF3, + 0x0B, 0x8B, 0x4B, 0xCB, 0x2B, 0xAB, 0x6B, 0xEB, + 0x1B, 0x9B, 0x5B, 0xDB, 0x3B, 0xBB, 0x7B, 0xFB, + 0x07, 0x87, 0x47, 0xC7, 0x27, 0xA7, 0x67, 0xE7, + 0x17, 0x97, 0x57, 0xD7, 0x37, 0xB7, 0x77, 0xF7, + 0x0F, 0x8F, 0x4F, 0xCF, 0x2F, 0xAF, 0x6F, 0xEF, + 0x1F, 0x9F, 0x5F, 0xDF, 0x3F, 0xBF, 0x7F, 0xFF +}; + +static PNG_CONST png_byte twobppswaptable[256] = { + 0x00, 0x40, 0x80, 0xC0, 0x10, 0x50, 0x90, 0xD0, + 0x20, 0x60, 0xA0, 0xE0, 0x30, 0x70, 0xB0, 0xF0, + 0x04, 0x44, 0x84, 0xC4, 0x14, 0x54, 0x94, 0xD4, + 0x24, 0x64, 0xA4, 0xE4, 0x34, 0x74, 0xB4, 0xF4, + 0x08, 0x48, 0x88, 0xC8, 0x18, 0x58, 0x98, 0xD8, + 0x28, 0x68, 0xA8, 0xE8, 0x38, 0x78, 0xB8, 0xF8, + 0x0C, 0x4C, 0x8C, 0xCC, 0x1C, 0x5C, 0x9C, 0xDC, + 0x2C, 0x6C, 0xAC, 0xEC, 0x3C, 0x7C, 0xBC, 0xFC, + 0x01, 0x41, 0x81, 0xC1, 0x11, 0x51, 0x91, 0xD1, + 0x21, 0x61, 0xA1, 0xE1, 0x31, 0x71, 0xB1, 0xF1, + 0x05, 0x45, 0x85, 0xC5, 0x15, 0x55, 0x95, 0xD5, + 0x25, 0x65, 0xA5, 0xE5, 0x35, 0x75, 0xB5, 0xF5, + 0x09, 0x49, 0x89, 0xC9, 0x19, 0x59, 0x99, 0xD9, + 0x29, 0x69, 0xA9, 0xE9, 0x39, 0x79, 0xB9, 0xF9, + 0x0D, 0x4D, 0x8D, 0xCD, 0x1D, 0x5D, 0x9D, 0xDD, + 0x2D, 0x6D, 0xAD, 0xED, 0x3D, 0x7D, 0xBD, 0xFD, + 0x02, 0x42, 0x82, 0xC2, 0x12, 0x52, 0x92, 0xD2, + 0x22, 0x62, 0xA2, 0xE2, 0x32, 0x72, 0xB2, 0xF2, + 0x06, 0x46, 0x86, 0xC6, 0x16, 0x56, 0x96, 0xD6, + 0x26, 0x66, 0xA6, 0xE6, 0x36, 0x76, 0xB6, 0xF6, + 0x0A, 0x4A, 0x8A, 0xCA, 0x1A, 0x5A, 0x9A, 0xDA, + 0x2A, 0x6A, 0xAA, 0xEA, 0x3A, 0x7A, 0xBA, 0xFA, + 0x0E, 0x4E, 0x8E, 0xCE, 0x1E, 0x5E, 0x9E, 0xDE, + 0x2E, 0x6E, 0xAE, 0xEE, 0x3E, 0x7E, 0xBE, 0xFE, + 0x03, 0x43, 0x83, 0xC3, 0x13, 0x53, 0x93, 0xD3, + 0x23, 0x63, 0xA3, 0xE3, 0x33, 0x73, 0xB3, 0xF3, + 0x07, 0x47, 0x87, 0xC7, 0x17, 0x57, 0x97, 0xD7, + 0x27, 0x67, 0xA7, 0xE7, 0x37, 0x77, 0xB7, 0xF7, + 0x0B, 0x4B, 0x8B, 0xCB, 0x1B, 0x5B, 0x9B, 0xDB, + 0x2B, 0x6B, 0xAB, 0xEB, 0x3B, 0x7B, 0xBB, 0xFB, + 0x0F, 0x4F, 0x8F, 0xCF, 0x1F, 0x5F, 0x9F, 0xDF, + 0x2F, 0x6F, 0xAF, 0xEF, 0x3F, 0x7F, 0xBF, 0xFF +}; + +static PNG_CONST png_byte fourbppswaptable[256] = { + 0x00, 0x10, 0x20, 0x30, 0x40, 0x50, 0x60, 0x70, + 0x80, 0x90, 0xA0, 0xB0, 0xC0, 0xD0, 0xE0, 0xF0, + 0x01, 0x11, 0x21, 0x31, 0x41, 0x51, 0x61, 0x71, + 0x81, 0x91, 0xA1, 0xB1, 0xC1, 0xD1, 0xE1, 0xF1, + 0x02, 0x12, 0x22, 0x32, 0x42, 0x52, 0x62, 0x72, + 0x82, 0x92, 0xA2, 0xB2, 0xC2, 0xD2, 0xE2, 0xF2, + 0x03, 0x13, 0x23, 0x33, 0x43, 0x53, 0x63, 0x73, + 0x83, 0x93, 0xA3, 0xB3, 0xC3, 0xD3, 0xE3, 0xF3, + 0x04, 0x14, 0x24, 0x34, 0x44, 0x54, 0x64, 0x74, + 0x84, 0x94, 0xA4, 0xB4, 0xC4, 0xD4, 0xE4, 0xF4, + 0x05, 0x15, 0x25, 0x35, 0x45, 0x55, 0x65, 0x75, + 0x85, 0x95, 0xA5, 0xB5, 0xC5, 0xD5, 0xE5, 0xF5, + 0x06, 0x16, 0x26, 0x36, 0x46, 0x56, 0x66, 0x76, + 0x86, 0x96, 0xA6, 0xB6, 0xC6, 0xD6, 0xE6, 0xF6, + 0x07, 0x17, 0x27, 0x37, 0x47, 0x57, 0x67, 0x77, + 0x87, 0x97, 0xA7, 0xB7, 0xC7, 0xD7, 0xE7, 0xF7, + 0x08, 0x18, 0x28, 0x38, 0x48, 0x58, 0x68, 0x78, + 0x88, 0x98, 0xA8, 0xB8, 0xC8, 0xD8, 0xE8, 0xF8, + 0x09, 0x19, 0x29, 0x39, 0x49, 0x59, 0x69, 0x79, + 0x89, 0x99, 0xA9, 0xB9, 0xC9, 0xD9, 0xE9, 0xF9, + 0x0A, 0x1A, 0x2A, 0x3A, 0x4A, 0x5A, 0x6A, 0x7A, + 0x8A, 0x9A, 0xAA, 0xBA, 0xCA, 0xDA, 0xEA, 0xFA, + 0x0B, 0x1B, 0x2B, 0x3B, 0x4B, 0x5B, 0x6B, 0x7B, + 0x8B, 0x9B, 0xAB, 0xBB, 0xCB, 0xDB, 0xEB, 0xFB, + 0x0C, 0x1C, 0x2C, 0x3C, 0x4C, 0x5C, 0x6C, 0x7C, + 0x8C, 0x9C, 0xAC, 0xBC, 0xCC, 0xDC, 0xEC, 0xFC, + 0x0D, 0x1D, 0x2D, 0x3D, 0x4D, 0x5D, 0x6D, 0x7D, + 0x8D, 0x9D, 0xAD, 0xBD, 0xCD, 0xDD, 0xED, 0xFD, + 0x0E, 0x1E, 0x2E, 0x3E, 0x4E, 0x5E, 0x6E, 0x7E, + 0x8E, 0x9E, 0xAE, 0xBE, 0xCE, 0xDE, 0xEE, 0xFE, + 0x0F, 0x1F, 0x2F, 0x3F, 0x4F, 0x5F, 0x6F, 0x7F, + 0x8F, 0x9F, 0xAF, 0xBF, 0xCF, 0xDF, 0xEF, 0xFF +}; + +/* Swaps pixel packing order within bytes */ +void /* PRIVATE */ +png_do_packswap(png_row_infop row_info, png_bytep row) +{ + png_debug(1, "in png_do_packswap"); + + if (row_info->bit_depth < 8) + { + png_bytep rp; + png_const_bytep end, table; + + end = row + row_info->rowbytes; + + if (row_info->bit_depth == 1) + table = onebppswaptable; + + else if (row_info->bit_depth == 2) + table = twobppswaptable; + + else if (row_info->bit_depth == 4) + table = fourbppswaptable; + + else + return; + + for (rp = row; rp < end; rp++) + *rp = table[*rp]; + } +} +#endif /* PACKSWAP || WRITE_PACKSWAP */ + +#if defined(PNG_WRITE_FILLER_SUPPORTED) || \ + defined(PNG_READ_STRIP_ALPHA_SUPPORTED) +/* Remove a channel - this used to be 'png_do_strip_filler' but it used a + * somewhat weird combination of flags to determine what to do. All the calls + * to png_do_strip_filler are changed in 1.5.2 to call this instead with the + * correct arguments. + * + * The routine isn't general - the channel must be the channel at the start or + * end (not in the middle) of each pixel. + */ +void /* PRIVATE */ +png_do_strip_channel(png_row_infop row_info, png_bytep row, int at_start) +{ + png_bytep sp = row; /* source pointer */ + png_bytep dp = row; /* destination pointer */ + png_bytep ep = row + row_info->rowbytes; /* One beyond end of row */ + + /* At the start sp will point to the first byte to copy and dp to where + * it is copied to. ep always points just beyond the end of the row, so + * the loop simply copies (channels-1) channels until sp reaches ep. + * + * at_start: 0 -- convert AG, XG, ARGB, XRGB, AAGG, XXGG, etc. + * nonzero -- convert GA, GX, RGBA, RGBX, GGAA, RRGGBBXX, etc. + */ + + /* GA, GX, XG cases */ + if (row_info->channels == 2) + { + if (row_info->bit_depth == 8) + { + if (at_start != 0) /* Skip initial filler */ + ++sp; + else /* Skip initial channel and, for sp, the filler */ + sp += 2, ++dp; + + /* For a 1 pixel wide image there is nothing to do */ + while (sp < ep) + *dp++ = *sp, sp += 2; + + row_info->pixel_depth = 8; + } + + else if (row_info->bit_depth == 16) + { + if (at_start != 0) /* Skip initial filler */ + sp += 2; + else /* Skip initial channel and, for sp, the filler */ + sp += 4, dp += 2; + + while (sp < ep) + *dp++ = *sp++, *dp++ = *sp, sp += 3; + + row_info->pixel_depth = 16; + } + + else + return; /* bad bit depth */ + + row_info->channels = 1; + + /* Finally fix the color type if it records an alpha channel */ + if (row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA) + row_info->color_type = PNG_COLOR_TYPE_GRAY; + } + + /* RGBA, RGBX, XRGB cases */ + else if (row_info->channels == 4) + { + if (row_info->bit_depth == 8) + { + if (at_start != 0) /* Skip initial filler */ + ++sp; + else /* Skip initial channels and, for sp, the filler */ + sp += 4, dp += 3; + + /* Note that the loop adds 3 to dp and 4 to sp each time. */ + while (sp < ep) + *dp++ = *sp++, *dp++ = *sp++, *dp++ = *sp, sp += 2; + + row_info->pixel_depth = 24; + } + + else if (row_info->bit_depth == 16) + { + if (at_start != 0) /* Skip initial filler */ + sp += 2; + else /* Skip initial channels and, for sp, the filler */ + sp += 8, dp += 6; + + while (sp < ep) + { + /* Copy 6 bytes, skip 2 */ + *dp++ = *sp++, *dp++ = *sp++; + *dp++ = *sp++, *dp++ = *sp++; + *dp++ = *sp++, *dp++ = *sp, sp += 3; + } + + row_info->pixel_depth = 48; + } + + else + return; /* bad bit depth */ + + row_info->channels = 3; + + /* Finally fix the color type if it records an alpha channel */ + if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA) + row_info->color_type = PNG_COLOR_TYPE_RGB; + } + + else + return; /* The filler channel has gone already */ + + /* Fix the rowbytes value. */ + row_info->rowbytes = dp-row; +} +#endif + +#if defined(PNG_READ_BGR_SUPPORTED) || defined(PNG_WRITE_BGR_SUPPORTED) +/* Swaps red and blue bytes within a pixel */ +void /* PRIVATE */ +png_do_bgr(png_row_infop row_info, png_bytep row) +{ + png_debug(1, "in png_do_bgr"); + + if ((row_info->color_type & PNG_COLOR_MASK_COLOR) != 0) + { + png_uint_32 row_width = row_info->width; + if (row_info->bit_depth == 8) + { + if (row_info->color_type == PNG_COLOR_TYPE_RGB) + { + png_bytep rp; + png_uint_32 i; + + for (i = 0, rp = row; i < row_width; i++, rp += 3) + { + png_byte save = *rp; + *rp = *(rp + 2); + *(rp + 2) = save; + } + } + + else if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA) + { + png_bytep rp; + png_uint_32 i; + + for (i = 0, rp = row; i < row_width; i++, rp += 4) + { + png_byte save = *rp; + *rp = *(rp + 2); + *(rp + 2) = save; + } + } + } + +#ifdef PNG_16BIT_SUPPORTED + else if (row_info->bit_depth == 16) + { + if (row_info->color_type == PNG_COLOR_TYPE_RGB) + { + png_bytep rp; + png_uint_32 i; + + for (i = 0, rp = row; i < row_width; i++, rp += 6) + { + png_byte save = *rp; + *rp = *(rp + 4); + *(rp + 4) = save; + save = *(rp + 1); + *(rp + 1) = *(rp + 5); + *(rp + 5) = save; + } + } + + else if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA) + { + png_bytep rp; + png_uint_32 i; + + for (i = 0, rp = row; i < row_width; i++, rp += 8) + { + png_byte save = *rp; + *rp = *(rp + 4); + *(rp + 4) = save; + save = *(rp + 1); + *(rp + 1) = *(rp + 5); + *(rp + 5) = save; + } + } + } +#endif + } +} +#endif /* READ_BGR || WRITE_BGR */ + +#if defined(PNG_READ_CHECK_FOR_INVALID_INDEX_SUPPORTED) || \ + defined(PNG_WRITE_CHECK_FOR_INVALID_INDEX_SUPPORTED) +/* Added at libpng-1.5.10 */ +void /* PRIVATE */ +png_do_check_palette_indexes(png_structrp png_ptr, png_row_infop row_info) +{ + if (png_ptr->num_palette < (1 << row_info->bit_depth) && + png_ptr->num_palette > 0) /* num_palette can be 0 in MNG files */ + { + /* Calculations moved outside switch in an attempt to stop different + * compiler warnings. 'padding' is in *bits* within the last byte, it is + * an 'int' because pixel_depth becomes an 'int' in the expression below, + * and this calculation is used because it avoids warnings that other + * forms produced on either GCC or MSVC. + */ + int padding = (-row_info->pixel_depth * row_info->width) & 7; + png_bytep rp = png_ptr->row_buf + row_info->rowbytes; + + switch (row_info->bit_depth) + { + case 1: + { + /* in this case, all bytes must be 0 so we don't need + * to unpack the pixels except for the rightmost one. + */ + for (; rp > png_ptr->row_buf; rp--) + { + if (*rp >> padding != 0) + png_ptr->num_palette_max = 1; + padding = 0; + } + + break; + } + + case 2: + { + for (; rp > png_ptr->row_buf; rp--) + { + int i = ((*rp >> padding) & 0x03); + + if (i > png_ptr->num_palette_max) + png_ptr->num_palette_max = i; + + i = (((*rp >> padding) >> 2) & 0x03); + + if (i > png_ptr->num_palette_max) + png_ptr->num_palette_max = i; + + i = (((*rp >> padding) >> 4) & 0x03); + + if (i > png_ptr->num_palette_max) + png_ptr->num_palette_max = i; + + i = (((*rp >> padding) >> 6) & 0x03); + + if (i > png_ptr->num_palette_max) + png_ptr->num_palette_max = i; + + padding = 0; + } + + break; + } + + case 4: + { + for (; rp > png_ptr->row_buf; rp--) + { + int i = ((*rp >> padding) & 0x0f); + + if (i > png_ptr->num_palette_max) + png_ptr->num_palette_max = i; + + i = (((*rp >> padding) >> 4) & 0x0f); + + if (i > png_ptr->num_palette_max) + png_ptr->num_palette_max = i; + + padding = 0; + } + + break; + } + + case 8: + { + for (; rp > png_ptr->row_buf; rp--) + { + if (*rp > png_ptr->num_palette_max) + png_ptr->num_palette_max = (int) *rp; + } + + break; + } + + default: + break; + } + } +} +#endif /* CHECK_FOR_INVALID_INDEX */ + +#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED) || \ + defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED) +#ifdef PNG_USER_TRANSFORM_PTR_SUPPORTED +void PNGAPI +png_set_user_transform_info(png_structrp png_ptr, png_voidp + user_transform_ptr, int user_transform_depth, int user_transform_channels) +{ + png_debug(1, "in png_set_user_transform_info"); + + if (png_ptr == NULL) + return; + +#ifdef PNG_READ_USER_TRANSFORM_SUPPORTED + if ((png_ptr->mode & PNG_IS_READ_STRUCT) != 0 && + (png_ptr->flags & PNG_FLAG_ROW_INIT) != 0) + { + png_app_error(png_ptr, + "info change after png_start_read_image or png_read_update_info"); + return; + } +#endif + + png_ptr->user_transform_ptr = user_transform_ptr; + png_ptr->user_transform_depth = (png_byte)user_transform_depth; + png_ptr->user_transform_channels = (png_byte)user_transform_channels; +} +#endif + +/* This function returns a pointer to the user_transform_ptr associated with + * the user transform functions. The application should free any memory + * associated with this pointer before png_write_destroy and png_read_destroy + * are called. + */ +#ifdef PNG_USER_TRANSFORM_PTR_SUPPORTED +png_voidp PNGAPI +png_get_user_transform_ptr(png_const_structrp png_ptr) +{ + if (png_ptr == NULL) + return (NULL); + + return png_ptr->user_transform_ptr; +} +#endif + +#ifdef PNG_USER_TRANSFORM_INFO_SUPPORTED +png_uint_32 PNGAPI +png_get_current_row_number(png_const_structrp png_ptr) +{ + /* See the comments in png.h - this is the sub-image row when reading an + * interlaced image. + */ + if (png_ptr != NULL) + return png_ptr->row_number; + + return PNG_UINT_32_MAX; /* help the app not to fail silently */ +} + +png_byte PNGAPI +png_get_current_pass_number(png_const_structrp png_ptr) +{ + if (png_ptr != NULL) + return png_ptr->pass; + return 8; /* invalid */ +} +#endif /* USER_TRANSFORM_INFO */ +#endif /* READ_USER_TRANSFORM || WRITE_USER_TRANSFORM */ +#endif /* READ || WRITE */ diff --git a/app/src/main/jni/png.c b/app/src/main/jni/png.c new file mode 100644 index 000000000..baa4d59e9 --- /dev/null +++ b/app/src/main/jni/png.c @@ -0,0 +1,109 @@ +#include "png.h" +#include "libpng/png.h" + +jobject PNG_DecodeStream(JNIEnv* env, jobject is, jint format) { + // TODO + return NULL; +} + +jobject PNG_DecodeFileHandler(JNIEnv* env, FILE* fp, jint format) { + int width, height; + png_image image; + png_bytep buffer; + PNG* png; + jclass imageClazz; + jmethodID constructor; + int fakeWidth; + int fakeStride; + + png = (PNG*) malloc(sizeof(PNG)); + if (png == NULL) + return NULL; + + memset(&image, 0, sizeof image); + image.version = PNG_IMAGE_VERSION; + + if (!png_image_begin_read_from_stdio(&image, fp)) + return NULL; + + // Get png format + switch (format) { + case FORMAT_AUTO: + if (image.format == PNG_FORMAT_GRAY) { + format = FORMAT_GRAY; + } else if (image.format == PNG_FORMAT_GA) { + format = FORMAT_GRAY_ALPHA; + } else if (image.format == PNG_FORMAT_RGB) { + format = FORMAT_RGB; + } else if (image.format == PNG_FORMAT_RGBA) { + format = FORMAT_RGBA; + } else { + format = FORMAT_RGB; + image.format = PNG_FORMAT_RGB; + } + break; + case FORMAT_RGB: + image.format = PNG_FORMAT_RGB; + break; + case FORMAT_RGBA: + image.format = PNG_FORMAT_RGBA; + break; + case FORMAT_GRAY: + image.format = PNG_FORMAT_GRAY; + break; + case FORMAT_GRAY_ALPHA: + image.format = PNG_FORMAT_GA; + break; + default: + format = FORMAT_RGB; + image.format = PNG_FORMAT_RGB; + break; + } + + fakeWidth = format != FORMAT_RGBA ? nextMulOf4(image.width) : image.width; + fakeStride = PNG_IMAGE_PIXEL_CHANNELS(image.format) * fakeWidth; + buffer = malloc(image.height * fakeStride); + if (buffer == NULL) + return NULL; + + if (!png_image_finish_read(&image, (png_const_colorp)&defaultBgColorRGB, + buffer, fakeStride, NULL)) { + free(buffer); + png_image_free(&image); + return NULL; + } + + // We get data now ~~~~ ! + png->data = buffer; + png->width = fakeWidth; + png->height = image.height; // Tell other the image file width + png->format = format; + + imageClazz = (*env)->FindClass(env, + "com/hippo/ehviewer/gallery/image/Image"); + constructor = (*env)->GetMethodID(env, imageClazz, "", + "(JIIIII)V"); + + if (constructor == 0) { + PNG_Free((JNIEnv*)NULL, png); + return NULL; + } else { + return (*env)->NewObject(env, imageClazz, constructor, (jlong) (intptr_t) png, + FILE_FORMAT_PNG, png->width, png->height, format, + DEFAULT_TYPE); + } +} + +void PNG_Render(JNIEnv* env, PNG* png, int format) { + if (format != png->format) + return; + + glTexSubImage2D(DEFAULT_TARGET, 0, 0, 0, png->width, + png->height, format, DEFAULT_TYPE, png->data); +} + +void PNG_Free(JNIEnv* env, PNG* png) { + + free(png->data); + free(png); +} diff --git a/app/src/main/jni/png.h b/app/src/main/jni/png.h new file mode 100644 index 000000000..8a535e273 --- /dev/null +++ b/app/src/main/jni/png.h @@ -0,0 +1,20 @@ +#ifndef PNG_H_ +#define PNG_H_ + +#include + +#include "image.h" + +typedef struct { + void* data; + int width; + int height; + int format; +} PNG; + +jobject PNG_DecodeStream(JNIEnv* env, jobject is, jint format); +jobject PNG_DecodeFileHandler(JNIEnv* env, FILE* fp, jint format); +void PNG_Render(JNIEnv* env, PNG* png, int format); +void PNG_Free(JNIEnv* env, PNG* png); + +#endif /* PNG_H_ */ diff --git a/app/src/main/jni/utils.c b/app/src/main/jni/utils.c new file mode 100644 index 000000000..5c64d0a73 --- /dev/null +++ b/app/src/main/jni/utils.c @@ -0,0 +1,71 @@ +#include "image.h" +#include "utils.h" + +int nextMulOf4(int x) { + return (x + 3) / 4 * 4; +} + +// TODO Need a better one +byte getVFrowRGB(byte r, byte g, byte b) { + byte t = g > b ? g : b; + return r > t ? r : t; +} + +void eraseRGB(rgb* pixels, int num, rgb color) { + int i; + for (i = 0; i < num; i++) + *(pixels + i) = color; +} + +void eraseRGBA(rgba* pixels, int num, rgba color) { + int i; + for (i = 0; i < num; i++) + *(pixels + i) = color; +} + +void eraseLUM(lum* pixels, int num, lum color) { + int i; + for (i = 0; i < num; i++) + *(pixels + i) = color; +} + +void eraseLUMA(luma* pixels, int num, luma color) { + int i; + for (i = 0; i < num; i++) + *(pixels + i) = color; +} + +void clearException(JNIEnv* env) { + if ((*env)->ExceptionOccurred(env)) + (*env)->ExceptionClear(env); +} + +StreamContainer* getStreamContainer(JNIEnv* env, jobject is) { + jclass streamCls = (*env)->GetObjectClass(env, is); + jmethodID readMID = (*env)->GetMethodID(env, streamCls, "read", "([BII)I"); + jmethodID closeMID = (*env)->GetMethodID(env, streamCls, "close", "()V"); + + if (readMID == 0 || closeMID == 0) + return NULL ; + + // Create stream container + StreamContainer* sc = (StreamContainer*) malloc(sizeof(StreamContainer)); + sc->readMID = readMID; + sc->closeMID = closeMID; + sc->jvm = g_jvm; + sc->stream = (*env)->NewGlobalRef(env, is); + sc->buffer = NULL; + + return sc; +} + +void closeStreamContainer(JNIEnv* env, StreamContainer* sc) { + (*env)->CallVoidMethod(env, sc->stream, sc->closeMID); +} + +void freeStreamContainer(JNIEnv* env, StreamContainer* sc) { + (*env)->DeleteGlobalRef(env, sc->stream); + if (sc->buffer != NULL) + (*env)->DeleteGlobalRef(env, sc->buffer); + free(sc); +} diff --git a/app/src/main/jni/utils.h b/app/src/main/jni/utils.h new file mode 100644 index 000000000..37a0c6340 --- /dev/null +++ b/app/src/main/jni/utils.h @@ -0,0 +1,29 @@ +#ifndef UTILS_H_ +#define UTILS_H_ + +#include "image.h" + +typedef struct { + JavaVM* jvm; + jobject stream; + jmethodID readMID; + jmethodID closeMID; + jbyteArray buffer; +} StreamContainer; + +int nextMulOf4(int x); + +void eraseRGB(rgb* pixels, int num, rgb color); +void eraseRGBA(rgba* pixels, int num, rgba color); +void eraseLUM(lum* pixels, int num, lum color); +void eraseLUMA(luma* pixels, int num, luma color); + +byte getVFrowRGB(byte r, byte g, byte b); + +void clearException(JNIEnv* env); + +StreamContainer* getSC(JNIEnv* env, jobject is); +void closeSC(JNIEnv* env, StreamContainer* sc); +void freeSC(JNIEnv* env, StreamContainer* sc); + +#endif /* UTILS_H_ */ diff --git a/app/src/main/jni/zlib/adler32.c b/app/src/main/jni/zlib/adler32.c new file mode 100644 index 000000000..a868f073d --- /dev/null +++ b/app/src/main/jni/zlib/adler32.c @@ -0,0 +1,179 @@ +/* adler32.c -- compute the Adler-32 checksum of a data stream + * Copyright (C) 1995-2011 Mark Adler + * For conditions of distribution and use, see copyright notice in zlib.h + */ + +/* @(#) $Id$ */ + +#include "zutil.h" + +#define local static + +local uLong adler32_combine_ OF((uLong adler1, uLong adler2, z_off64_t len2)); + +#define BASE 65521 /* largest prime smaller than 65536 */ +#define NMAX 5552 +/* NMAX is the largest n such that 255n(n+1)/2 + (n+1)(BASE-1) <= 2^32-1 */ + +#define DO1(buf,i) {adler += (buf)[i]; sum2 += adler;} +#define DO2(buf,i) DO1(buf,i); DO1(buf,i+1); +#define DO4(buf,i) DO2(buf,i); DO2(buf,i+2); +#define DO8(buf,i) DO4(buf,i); DO4(buf,i+4); +#define DO16(buf) DO8(buf,0); DO8(buf,8); + +/* use NO_DIVIDE if your processor does not do division in hardware -- + try it both ways to see which is faster */ +#ifdef NO_DIVIDE +/* note that this assumes BASE is 65521, where 65536 % 65521 == 15 + (thank you to John Reiser for pointing this out) */ +# define CHOP(a) \ + do { \ + unsigned long tmp = a >> 16; \ + a &= 0xffffUL; \ + a += (tmp << 4) - tmp; \ + } while (0) +# define MOD28(a) \ + do { \ + CHOP(a); \ + if (a >= BASE) a -= BASE; \ + } while (0) +# define MOD(a) \ + do { \ + CHOP(a); \ + MOD28(a); \ + } while (0) +# define MOD63(a) \ + do { /* this assumes a is not negative */ \ + z_off64_t tmp = a >> 32; \ + a &= 0xffffffffL; \ + a += (tmp << 8) - (tmp << 5) + tmp; \ + tmp = a >> 16; \ + a &= 0xffffL; \ + a += (tmp << 4) - tmp; \ + tmp = a >> 16; \ + a &= 0xffffL; \ + a += (tmp << 4) - tmp; \ + if (a >= BASE) a -= BASE; \ + } while (0) +#else +# define MOD(a) a %= BASE +# define MOD28(a) a %= BASE +# define MOD63(a) a %= BASE +#endif + +/* ========================================================================= */ +uLong ZEXPORT adler32(adler, buf, len) + uLong adler; + const Bytef *buf; + uInt len; +{ + unsigned long sum2; + unsigned n; + + /* split Adler-32 into component sums */ + sum2 = (adler >> 16) & 0xffff; + adler &= 0xffff; + + /* in case user likes doing a byte at a time, keep it fast */ + if (len == 1) { + adler += buf[0]; + if (adler >= BASE) + adler -= BASE; + sum2 += adler; + if (sum2 >= BASE) + sum2 -= BASE; + return adler | (sum2 << 16); + } + + /* initial Adler-32 value (deferred check for len == 1 speed) */ + if (buf == Z_NULL) + return 1L; + + /* in case short lengths are provided, keep it somewhat fast */ + if (len < 16) { + while (len--) { + adler += *buf++; + sum2 += adler; + } + if (adler >= BASE) + adler -= BASE; + MOD28(sum2); /* only added so many BASE's */ + return adler | (sum2 << 16); + } + + /* do length NMAX blocks -- requires just one modulo operation */ + while (len >= NMAX) { + len -= NMAX; + n = NMAX / 16; /* NMAX is divisible by 16 */ + do { + DO16(buf); /* 16 sums unrolled */ + buf += 16; + } while (--n); + MOD(adler); + MOD(sum2); + } + + /* do remaining bytes (less than NMAX, still just one modulo) */ + if (len) { /* avoid modulos if none remaining */ + while (len >= 16) { + len -= 16; + DO16(buf); + buf += 16; + } + while (len--) { + adler += *buf++; + sum2 += adler; + } + MOD(adler); + MOD(sum2); + } + + /* return recombined sums */ + return adler | (sum2 << 16); +} + +/* ========================================================================= */ +local uLong adler32_combine_(adler1, adler2, len2) + uLong adler1; + uLong adler2; + z_off64_t len2; +{ + unsigned long sum1; + unsigned long sum2; + unsigned rem; + + /* for negative len, return invalid adler32 as a clue for debugging */ + if (len2 < 0) + return 0xffffffffUL; + + /* the derivation of this formula is left as an exercise for the reader */ + MOD63(len2); /* assumes len2 >= 0 */ + rem = (unsigned)len2; + sum1 = adler1 & 0xffff; + sum2 = rem * sum1; + MOD(sum2); + sum1 += (adler2 & 0xffff) + BASE - 1; + sum2 += ((adler1 >> 16) & 0xffff) + ((adler2 >> 16) & 0xffff) + BASE - rem; + if (sum1 >= BASE) sum1 -= BASE; + if (sum1 >= BASE) sum1 -= BASE; + if (sum2 >= (BASE << 1)) sum2 -= (BASE << 1); + if (sum2 >= BASE) sum2 -= BASE; + return sum1 | (sum2 << 16); +} + +/* ========================================================================= */ +uLong ZEXPORT adler32_combine(adler1, adler2, len2) + uLong adler1; + uLong adler2; + z_off_t len2; +{ + return adler32_combine_(adler1, adler2, len2); +} + +uLong ZEXPORT adler32_combine64(adler1, adler2, len2) + uLong adler1; + uLong adler2; + z_off64_t len2; +{ + return adler32_combine_(adler1, adler2, len2); +} diff --git a/app/src/main/jni/zlib/crc32.c b/app/src/main/jni/zlib/crc32.c new file mode 100644 index 000000000..979a7190a --- /dev/null +++ b/app/src/main/jni/zlib/crc32.c @@ -0,0 +1,425 @@ +/* crc32.c -- compute the CRC-32 of a data stream + * Copyright (C) 1995-2006, 2010, 2011, 2012 Mark Adler + * For conditions of distribution and use, see copyright notice in zlib.h + * + * Thanks to Rodney Brown for his contribution of faster + * CRC methods: exclusive-oring 32 bits of data at a time, and pre-computing + * tables for updating the shift register in one step with three exclusive-ors + * instead of four steps with four exclusive-ors. This results in about a + * factor of two increase in speed on a Power PC G4 (PPC7455) using gcc -O3. + */ + +/* @(#) $Id$ */ + +/* + Note on the use of DYNAMIC_CRC_TABLE: there is no mutex or semaphore + protection on the static variables used to control the first-use generation + of the crc tables. Therefore, if you #define DYNAMIC_CRC_TABLE, you should + first call get_crc_table() to initialize the tables before allowing more than + one thread to use crc32(). + + DYNAMIC_CRC_TABLE and MAKECRCH can be #defined to write out crc32.h. + */ + +#ifdef MAKECRCH +# include +# ifndef DYNAMIC_CRC_TABLE +# define DYNAMIC_CRC_TABLE +# endif /* !DYNAMIC_CRC_TABLE */ +#endif /* MAKECRCH */ + +#include "zutil.h" /* for STDC and FAR definitions */ + +#define local static + +/* Definitions for doing the crc four data bytes at a time. */ +#if !defined(NOBYFOUR) && defined(Z_U4) +# define BYFOUR +#endif +#ifdef BYFOUR + local unsigned long crc32_little OF((unsigned long, + const unsigned char FAR *, unsigned)); + local unsigned long crc32_big OF((unsigned long, + const unsigned char FAR *, unsigned)); +# define TBLS 8 +#else +# define TBLS 1 +#endif /* BYFOUR */ + +/* Local functions for crc concatenation */ +local unsigned long gf2_matrix_times OF((unsigned long *mat, + unsigned long vec)); +local void gf2_matrix_square OF((unsigned long *square, unsigned long *mat)); +local uLong crc32_combine_ OF((uLong crc1, uLong crc2, z_off64_t len2)); + + +#ifdef DYNAMIC_CRC_TABLE + +local volatile int crc_table_empty = 1; +local z_crc_t FAR crc_table[TBLS][256]; +local void make_crc_table OF((void)); +#ifdef MAKECRCH + local void write_table OF((FILE *, const z_crc_t FAR *)); +#endif /* MAKECRCH */ +/* + Generate tables for a byte-wise 32-bit CRC calculation on the polynomial: + x^32+x^26+x^23+x^22+x^16+x^12+x^11+x^10+x^8+x^7+x^5+x^4+x^2+x+1. + + Polynomials over GF(2) are represented in binary, one bit per coefficient, + with the lowest powers in the most significant bit. Then adding polynomials + is just exclusive-or, and multiplying a polynomial by x is a right shift by + one. If we call the above polynomial p, and represent a byte as the + polynomial q, also with the lowest power in the most significant bit (so the + byte 0xb1 is the polynomial x^7+x^3+x+1), then the CRC is (q*x^32) mod p, + where a mod b means the remainder after dividing a by b. + + This calculation is done using the shift-register method of multiplying and + taking the remainder. The register is initialized to zero, and for each + incoming bit, x^32 is added mod p to the register if the bit is a one (where + x^32 mod p is p+x^32 = x^26+...+1), and the register is multiplied mod p by + x (which is shifting right by one and adding x^32 mod p if the bit shifted + out is a one). We start with the highest power (least significant bit) of + q and repeat for all eight bits of q. + + The first table is simply the CRC of all possible eight bit values. This is + all the information needed to generate CRCs on data a byte at a time for all + combinations of CRC register values and incoming bytes. The remaining tables + allow for word-at-a-time CRC calculation for both big-endian and little- + endian machines, where a word is four bytes. +*/ +local void make_crc_table() +{ + z_crc_t c; + int n, k; + z_crc_t poly; /* polynomial exclusive-or pattern */ + /* terms of polynomial defining this crc (except x^32): */ + static volatile int first = 1; /* flag to limit concurrent making */ + static const unsigned char p[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26}; + + /* See if another task is already doing this (not thread-safe, but better + than nothing -- significantly reduces duration of vulnerability in + case the advice about DYNAMIC_CRC_TABLE is ignored) */ + if (first) { + first = 0; + + /* make exclusive-or pattern from polynomial (0xedb88320UL) */ + poly = 0; + for (n = 0; n < (int)(sizeof(p)/sizeof(unsigned char)); n++) + poly |= (z_crc_t)1 << (31 - p[n]); + + /* generate a crc for every 8-bit value */ + for (n = 0; n < 256; n++) { + c = (z_crc_t)n; + for (k = 0; k < 8; k++) + c = c & 1 ? poly ^ (c >> 1) : c >> 1; + crc_table[0][n] = c; + } + +#ifdef BYFOUR + /* generate crc for each value followed by one, two, and three zeros, + and then the byte reversal of those as well as the first table */ + for (n = 0; n < 256; n++) { + c = crc_table[0][n]; + crc_table[4][n] = ZSWAP32(c); + for (k = 1; k < 4; k++) { + c = crc_table[0][c & 0xff] ^ (c >> 8); + crc_table[k][n] = c; + crc_table[k + 4][n] = ZSWAP32(c); + } + } +#endif /* BYFOUR */ + + crc_table_empty = 0; + } + else { /* not first */ + /* wait for the other guy to finish (not efficient, but rare) */ + while (crc_table_empty) + ; + } + +#ifdef MAKECRCH + /* write out CRC tables to crc32.h */ + { + FILE *out; + + out = fopen("crc32.h", "w"); + if (out == NULL) return; + fprintf(out, "/* crc32.h -- tables for rapid CRC calculation\n"); + fprintf(out, " * Generated automatically by crc32.c\n */\n\n"); + fprintf(out, "local const z_crc_t FAR "); + fprintf(out, "crc_table[TBLS][256] =\n{\n {\n"); + write_table(out, crc_table[0]); +# ifdef BYFOUR + fprintf(out, "#ifdef BYFOUR\n"); + for (k = 1; k < 8; k++) { + fprintf(out, " },\n {\n"); + write_table(out, crc_table[k]); + } + fprintf(out, "#endif\n"); +# endif /* BYFOUR */ + fprintf(out, " }\n};\n"); + fclose(out); + } +#endif /* MAKECRCH */ +} + +#ifdef MAKECRCH +local void write_table(out, table) + FILE *out; + const z_crc_t FAR *table; +{ + int n; + + for (n = 0; n < 256; n++) + fprintf(out, "%s0x%08lxUL%s", n % 5 ? "" : " ", + (unsigned long)(table[n]), + n == 255 ? "\n" : (n % 5 == 4 ? ",\n" : ", ")); +} +#endif /* MAKECRCH */ + +#else /* !DYNAMIC_CRC_TABLE */ +/* ======================================================================== + * Tables of CRC-32s of all single-byte values, made by make_crc_table(). + */ +#include "crc32.h" +#endif /* DYNAMIC_CRC_TABLE */ + +/* ========================================================================= + * This function can be used by asm versions of crc32() + */ +const z_crc_t FAR * ZEXPORT get_crc_table() +{ +#ifdef DYNAMIC_CRC_TABLE + if (crc_table_empty) + make_crc_table(); +#endif /* DYNAMIC_CRC_TABLE */ + return (const z_crc_t FAR *)crc_table; +} + +/* ========================================================================= */ +#define DO1 crc = crc_table[0][((int)crc ^ (*buf++)) & 0xff] ^ (crc >> 8) +#define DO8 DO1; DO1; DO1; DO1; DO1; DO1; DO1; DO1 + +/* ========================================================================= */ +unsigned long ZEXPORT crc32(crc, buf, len) + unsigned long crc; + const unsigned char FAR *buf; + uInt len; +{ + if (buf == Z_NULL) return 0UL; + +#ifdef DYNAMIC_CRC_TABLE + if (crc_table_empty) + make_crc_table(); +#endif /* DYNAMIC_CRC_TABLE */ + +#ifdef BYFOUR + if (sizeof(void *) == sizeof(ptrdiff_t)) { + z_crc_t endian; + + endian = 1; + if (*((unsigned char *)(&endian))) + return crc32_little(crc, buf, len); + else + return crc32_big(crc, buf, len); + } +#endif /* BYFOUR */ + crc = crc ^ 0xffffffffUL; + while (len >= 8) { + DO8; + len -= 8; + } + if (len) do { + DO1; + } while (--len); + return crc ^ 0xffffffffUL; +} + +#ifdef BYFOUR + +/* ========================================================================= */ +#define DOLIT4 c ^= *buf4++; \ + c = crc_table[3][c & 0xff] ^ crc_table[2][(c >> 8) & 0xff] ^ \ + crc_table[1][(c >> 16) & 0xff] ^ crc_table[0][c >> 24] +#define DOLIT32 DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4 + +/* ========================================================================= */ +local unsigned long crc32_little(crc, buf, len) + unsigned long crc; + const unsigned char FAR *buf; + unsigned len; +{ + register z_crc_t c; + register const z_crc_t FAR *buf4; + + c = (z_crc_t)crc; + c = ~c; + while (len && ((ptrdiff_t)buf & 3)) { + c = crc_table[0][(c ^ *buf++) & 0xff] ^ (c >> 8); + len--; + } + + buf4 = (const z_crc_t FAR *)(const void FAR *)buf; + while (len >= 32) { + DOLIT32; + len -= 32; + } + while (len >= 4) { + DOLIT4; + len -= 4; + } + buf = (const unsigned char FAR *)buf4; + + if (len) do { + c = crc_table[0][(c ^ *buf++) & 0xff] ^ (c >> 8); + } while (--len); + c = ~c; + return (unsigned long)c; +} + +/* ========================================================================= */ +#define DOBIG4 c ^= *++buf4; \ + c = crc_table[4][c & 0xff] ^ crc_table[5][(c >> 8) & 0xff] ^ \ + crc_table[6][(c >> 16) & 0xff] ^ crc_table[7][c >> 24] +#define DOBIG32 DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4 + +/* ========================================================================= */ +local unsigned long crc32_big(crc, buf, len) + unsigned long crc; + const unsigned char FAR *buf; + unsigned len; +{ + register z_crc_t c; + register const z_crc_t FAR *buf4; + + c = ZSWAP32((z_crc_t)crc); + c = ~c; + while (len && ((ptrdiff_t)buf & 3)) { + c = crc_table[4][(c >> 24) ^ *buf++] ^ (c << 8); + len--; + } + + buf4 = (const z_crc_t FAR *)(const void FAR *)buf; + buf4--; + while (len >= 32) { + DOBIG32; + len -= 32; + } + while (len >= 4) { + DOBIG4; + len -= 4; + } + buf4++; + buf = (const unsigned char FAR *)buf4; + + if (len) do { + c = crc_table[4][(c >> 24) ^ *buf++] ^ (c << 8); + } while (--len); + c = ~c; + return (unsigned long)(ZSWAP32(c)); +} + +#endif /* BYFOUR */ + +#define GF2_DIM 32 /* dimension of GF(2) vectors (length of CRC) */ + +/* ========================================================================= */ +local unsigned long gf2_matrix_times(mat, vec) + unsigned long *mat; + unsigned long vec; +{ + unsigned long sum; + + sum = 0; + while (vec) { + if (vec & 1) + sum ^= *mat; + vec >>= 1; + mat++; + } + return sum; +} + +/* ========================================================================= */ +local void gf2_matrix_square(square, mat) + unsigned long *square; + unsigned long *mat; +{ + int n; + + for (n = 0; n < GF2_DIM; n++) + square[n] = gf2_matrix_times(mat, mat[n]); +} + +/* ========================================================================= */ +local uLong crc32_combine_(crc1, crc2, len2) + uLong crc1; + uLong crc2; + z_off64_t len2; +{ + int n; + unsigned long row; + unsigned long even[GF2_DIM]; /* even-power-of-two zeros operator */ + unsigned long odd[GF2_DIM]; /* odd-power-of-two zeros operator */ + + /* degenerate case (also disallow negative lengths) */ + if (len2 <= 0) + return crc1; + + /* put operator for one zero bit in odd */ + odd[0] = 0xedb88320UL; /* CRC-32 polynomial */ + row = 1; + for (n = 1; n < GF2_DIM; n++) { + odd[n] = row; + row <<= 1; + } + + /* put operator for two zero bits in even */ + gf2_matrix_square(even, odd); + + /* put operator for four zero bits in odd */ + gf2_matrix_square(odd, even); + + /* apply len2 zeros to crc1 (first square will put the operator for one + zero byte, eight zero bits, in even) */ + do { + /* apply zeros operator for this bit of len2 */ + gf2_matrix_square(even, odd); + if (len2 & 1) + crc1 = gf2_matrix_times(even, crc1); + len2 >>= 1; + + /* if no more bits set, then done */ + if (len2 == 0) + break; + + /* another iteration of the loop with odd and even swapped */ + gf2_matrix_square(odd, even); + if (len2 & 1) + crc1 = gf2_matrix_times(odd, crc1); + len2 >>= 1; + + /* if no more bits set, then done */ + } while (len2 != 0); + + /* return combined crc */ + crc1 ^= crc2; + return crc1; +} + +/* ========================================================================= */ +uLong ZEXPORT crc32_combine(crc1, crc2, len2) + uLong crc1; + uLong crc2; + z_off_t len2; +{ + return crc32_combine_(crc1, crc2, len2); +} + +uLong ZEXPORT crc32_combine64(crc1, crc2, len2) + uLong crc1; + uLong crc2; + z_off64_t len2; +{ + return crc32_combine_(crc1, crc2, len2); +} diff --git a/app/src/main/jni/zlib/crc32.h b/app/src/main/jni/zlib/crc32.h new file mode 100644 index 000000000..9e0c77810 --- /dev/null +++ b/app/src/main/jni/zlib/crc32.h @@ -0,0 +1,441 @@ +/* crc32.h -- tables for rapid CRC calculation + * Generated automatically by crc32.c + */ + +local const z_crc_t FAR crc_table[TBLS][256] = +{ + { + 0x00000000UL, 0x77073096UL, 0xee0e612cUL, 0x990951baUL, 0x076dc419UL, + 0x706af48fUL, 0xe963a535UL, 0x9e6495a3UL, 0x0edb8832UL, 0x79dcb8a4UL, + 0xe0d5e91eUL, 0x97d2d988UL, 0x09b64c2bUL, 0x7eb17cbdUL, 0xe7b82d07UL, + 0x90bf1d91UL, 0x1db71064UL, 0x6ab020f2UL, 0xf3b97148UL, 0x84be41deUL, + 0x1adad47dUL, 0x6ddde4ebUL, 0xf4d4b551UL, 0x83d385c7UL, 0x136c9856UL, + 0x646ba8c0UL, 0xfd62f97aUL, 0x8a65c9ecUL, 0x14015c4fUL, 0x63066cd9UL, + 0xfa0f3d63UL, 0x8d080df5UL, 0x3b6e20c8UL, 0x4c69105eUL, 0xd56041e4UL, + 0xa2677172UL, 0x3c03e4d1UL, 0x4b04d447UL, 0xd20d85fdUL, 0xa50ab56bUL, + 0x35b5a8faUL, 0x42b2986cUL, 0xdbbbc9d6UL, 0xacbcf940UL, 0x32d86ce3UL, + 0x45df5c75UL, 0xdcd60dcfUL, 0xabd13d59UL, 0x26d930acUL, 0x51de003aUL, + 0xc8d75180UL, 0xbfd06116UL, 0x21b4f4b5UL, 0x56b3c423UL, 0xcfba9599UL, + 0xb8bda50fUL, 0x2802b89eUL, 0x5f058808UL, 0xc60cd9b2UL, 0xb10be924UL, + 0x2f6f7c87UL, 0x58684c11UL, 0xc1611dabUL, 0xb6662d3dUL, 0x76dc4190UL, + 0x01db7106UL, 0x98d220bcUL, 0xefd5102aUL, 0x71b18589UL, 0x06b6b51fUL, + 0x9fbfe4a5UL, 0xe8b8d433UL, 0x7807c9a2UL, 0x0f00f934UL, 0x9609a88eUL, + 0xe10e9818UL, 0x7f6a0dbbUL, 0x086d3d2dUL, 0x91646c97UL, 0xe6635c01UL, + 0x6b6b51f4UL, 0x1c6c6162UL, 0x856530d8UL, 0xf262004eUL, 0x6c0695edUL, + 0x1b01a57bUL, 0x8208f4c1UL, 0xf50fc457UL, 0x65b0d9c6UL, 0x12b7e950UL, + 0x8bbeb8eaUL, 0xfcb9887cUL, 0x62dd1ddfUL, 0x15da2d49UL, 0x8cd37cf3UL, + 0xfbd44c65UL, 0x4db26158UL, 0x3ab551ceUL, 0xa3bc0074UL, 0xd4bb30e2UL, + 0x4adfa541UL, 0x3dd895d7UL, 0xa4d1c46dUL, 0xd3d6f4fbUL, 0x4369e96aUL, + 0x346ed9fcUL, 0xad678846UL, 0xda60b8d0UL, 0x44042d73UL, 0x33031de5UL, + 0xaa0a4c5fUL, 0xdd0d7cc9UL, 0x5005713cUL, 0x270241aaUL, 0xbe0b1010UL, + 0xc90c2086UL, 0x5768b525UL, 0x206f85b3UL, 0xb966d409UL, 0xce61e49fUL, + 0x5edef90eUL, 0x29d9c998UL, 0xb0d09822UL, 0xc7d7a8b4UL, 0x59b33d17UL, + 0x2eb40d81UL, 0xb7bd5c3bUL, 0xc0ba6cadUL, 0xedb88320UL, 0x9abfb3b6UL, + 0x03b6e20cUL, 0x74b1d29aUL, 0xead54739UL, 0x9dd277afUL, 0x04db2615UL, + 0x73dc1683UL, 0xe3630b12UL, 0x94643b84UL, 0x0d6d6a3eUL, 0x7a6a5aa8UL, + 0xe40ecf0bUL, 0x9309ff9dUL, 0x0a00ae27UL, 0x7d079eb1UL, 0xf00f9344UL, + 0x8708a3d2UL, 0x1e01f268UL, 0x6906c2feUL, 0xf762575dUL, 0x806567cbUL, + 0x196c3671UL, 0x6e6b06e7UL, 0xfed41b76UL, 0x89d32be0UL, 0x10da7a5aUL, + 0x67dd4accUL, 0xf9b9df6fUL, 0x8ebeeff9UL, 0x17b7be43UL, 0x60b08ed5UL, + 0xd6d6a3e8UL, 0xa1d1937eUL, 0x38d8c2c4UL, 0x4fdff252UL, 0xd1bb67f1UL, + 0xa6bc5767UL, 0x3fb506ddUL, 0x48b2364bUL, 0xd80d2bdaUL, 0xaf0a1b4cUL, + 0x36034af6UL, 0x41047a60UL, 0xdf60efc3UL, 0xa867df55UL, 0x316e8eefUL, + 0x4669be79UL, 0xcb61b38cUL, 0xbc66831aUL, 0x256fd2a0UL, 0x5268e236UL, + 0xcc0c7795UL, 0xbb0b4703UL, 0x220216b9UL, 0x5505262fUL, 0xc5ba3bbeUL, + 0xb2bd0b28UL, 0x2bb45a92UL, 0x5cb36a04UL, 0xc2d7ffa7UL, 0xb5d0cf31UL, + 0x2cd99e8bUL, 0x5bdeae1dUL, 0x9b64c2b0UL, 0xec63f226UL, 0x756aa39cUL, + 0x026d930aUL, 0x9c0906a9UL, 0xeb0e363fUL, 0x72076785UL, 0x05005713UL, + 0x95bf4a82UL, 0xe2b87a14UL, 0x7bb12baeUL, 0x0cb61b38UL, 0x92d28e9bUL, + 0xe5d5be0dUL, 0x7cdcefb7UL, 0x0bdbdf21UL, 0x86d3d2d4UL, 0xf1d4e242UL, + 0x68ddb3f8UL, 0x1fda836eUL, 0x81be16cdUL, 0xf6b9265bUL, 0x6fb077e1UL, + 0x18b74777UL, 0x88085ae6UL, 0xff0f6a70UL, 0x66063bcaUL, 0x11010b5cUL, + 0x8f659effUL, 0xf862ae69UL, 0x616bffd3UL, 0x166ccf45UL, 0xa00ae278UL, + 0xd70dd2eeUL, 0x4e048354UL, 0x3903b3c2UL, 0xa7672661UL, 0xd06016f7UL, + 0x4969474dUL, 0x3e6e77dbUL, 0xaed16a4aUL, 0xd9d65adcUL, 0x40df0b66UL, + 0x37d83bf0UL, 0xa9bcae53UL, 0xdebb9ec5UL, 0x47b2cf7fUL, 0x30b5ffe9UL, + 0xbdbdf21cUL, 0xcabac28aUL, 0x53b39330UL, 0x24b4a3a6UL, 0xbad03605UL, + 0xcdd70693UL, 0x54de5729UL, 0x23d967bfUL, 0xb3667a2eUL, 0xc4614ab8UL, + 0x5d681b02UL, 0x2a6f2b94UL, 0xb40bbe37UL, 0xc30c8ea1UL, 0x5a05df1bUL, + 0x2d02ef8dUL +#ifdef BYFOUR + }, + { + 0x00000000UL, 0x191b3141UL, 0x32366282UL, 0x2b2d53c3UL, 0x646cc504UL, + 0x7d77f445UL, 0x565aa786UL, 0x4f4196c7UL, 0xc8d98a08UL, 0xd1c2bb49UL, + 0xfaefe88aUL, 0xe3f4d9cbUL, 0xacb54f0cUL, 0xb5ae7e4dUL, 0x9e832d8eUL, + 0x87981ccfUL, 0x4ac21251UL, 0x53d92310UL, 0x78f470d3UL, 0x61ef4192UL, + 0x2eaed755UL, 0x37b5e614UL, 0x1c98b5d7UL, 0x05838496UL, 0x821b9859UL, + 0x9b00a918UL, 0xb02dfadbUL, 0xa936cb9aUL, 0xe6775d5dUL, 0xff6c6c1cUL, + 0xd4413fdfUL, 0xcd5a0e9eUL, 0x958424a2UL, 0x8c9f15e3UL, 0xa7b24620UL, + 0xbea97761UL, 0xf1e8e1a6UL, 0xe8f3d0e7UL, 0xc3de8324UL, 0xdac5b265UL, + 0x5d5daeaaUL, 0x44469febUL, 0x6f6bcc28UL, 0x7670fd69UL, 0x39316baeUL, + 0x202a5aefUL, 0x0b07092cUL, 0x121c386dUL, 0xdf4636f3UL, 0xc65d07b2UL, + 0xed705471UL, 0xf46b6530UL, 0xbb2af3f7UL, 0xa231c2b6UL, 0x891c9175UL, + 0x9007a034UL, 0x179fbcfbUL, 0x0e848dbaUL, 0x25a9de79UL, 0x3cb2ef38UL, + 0x73f379ffUL, 0x6ae848beUL, 0x41c51b7dUL, 0x58de2a3cUL, 0xf0794f05UL, + 0xe9627e44UL, 0xc24f2d87UL, 0xdb541cc6UL, 0x94158a01UL, 0x8d0ebb40UL, + 0xa623e883UL, 0xbf38d9c2UL, 0x38a0c50dUL, 0x21bbf44cUL, 0x0a96a78fUL, + 0x138d96ceUL, 0x5ccc0009UL, 0x45d73148UL, 0x6efa628bUL, 0x77e153caUL, + 0xbabb5d54UL, 0xa3a06c15UL, 0x888d3fd6UL, 0x91960e97UL, 0xded79850UL, + 0xc7cca911UL, 0xece1fad2UL, 0xf5facb93UL, 0x7262d75cUL, 0x6b79e61dUL, + 0x4054b5deUL, 0x594f849fUL, 0x160e1258UL, 0x0f152319UL, 0x243870daUL, + 0x3d23419bUL, 0x65fd6ba7UL, 0x7ce65ae6UL, 0x57cb0925UL, 0x4ed03864UL, + 0x0191aea3UL, 0x188a9fe2UL, 0x33a7cc21UL, 0x2abcfd60UL, 0xad24e1afUL, + 0xb43fd0eeUL, 0x9f12832dUL, 0x8609b26cUL, 0xc94824abUL, 0xd05315eaUL, + 0xfb7e4629UL, 0xe2657768UL, 0x2f3f79f6UL, 0x362448b7UL, 0x1d091b74UL, + 0x04122a35UL, 0x4b53bcf2UL, 0x52488db3UL, 0x7965de70UL, 0x607eef31UL, + 0xe7e6f3feUL, 0xfefdc2bfUL, 0xd5d0917cUL, 0xcccba03dUL, 0x838a36faUL, + 0x9a9107bbUL, 0xb1bc5478UL, 0xa8a76539UL, 0x3b83984bUL, 0x2298a90aUL, + 0x09b5fac9UL, 0x10aecb88UL, 0x5fef5d4fUL, 0x46f46c0eUL, 0x6dd93fcdUL, + 0x74c20e8cUL, 0xf35a1243UL, 0xea412302UL, 0xc16c70c1UL, 0xd8774180UL, + 0x9736d747UL, 0x8e2de606UL, 0xa500b5c5UL, 0xbc1b8484UL, 0x71418a1aUL, + 0x685abb5bUL, 0x4377e898UL, 0x5a6cd9d9UL, 0x152d4f1eUL, 0x0c367e5fUL, + 0x271b2d9cUL, 0x3e001cddUL, 0xb9980012UL, 0xa0833153UL, 0x8bae6290UL, + 0x92b553d1UL, 0xddf4c516UL, 0xc4eff457UL, 0xefc2a794UL, 0xf6d996d5UL, + 0xae07bce9UL, 0xb71c8da8UL, 0x9c31de6bUL, 0x852aef2aUL, 0xca6b79edUL, + 0xd37048acUL, 0xf85d1b6fUL, 0xe1462a2eUL, 0x66de36e1UL, 0x7fc507a0UL, + 0x54e85463UL, 0x4df36522UL, 0x02b2f3e5UL, 0x1ba9c2a4UL, 0x30849167UL, + 0x299fa026UL, 0xe4c5aeb8UL, 0xfdde9ff9UL, 0xd6f3cc3aUL, 0xcfe8fd7bUL, + 0x80a96bbcUL, 0x99b25afdUL, 0xb29f093eUL, 0xab84387fUL, 0x2c1c24b0UL, + 0x350715f1UL, 0x1e2a4632UL, 0x07317773UL, 0x4870e1b4UL, 0x516bd0f5UL, + 0x7a468336UL, 0x635db277UL, 0xcbfad74eUL, 0xd2e1e60fUL, 0xf9ccb5ccUL, + 0xe0d7848dUL, 0xaf96124aUL, 0xb68d230bUL, 0x9da070c8UL, 0x84bb4189UL, + 0x03235d46UL, 0x1a386c07UL, 0x31153fc4UL, 0x280e0e85UL, 0x674f9842UL, + 0x7e54a903UL, 0x5579fac0UL, 0x4c62cb81UL, 0x8138c51fUL, 0x9823f45eUL, + 0xb30ea79dUL, 0xaa1596dcUL, 0xe554001bUL, 0xfc4f315aUL, 0xd7626299UL, + 0xce7953d8UL, 0x49e14f17UL, 0x50fa7e56UL, 0x7bd72d95UL, 0x62cc1cd4UL, + 0x2d8d8a13UL, 0x3496bb52UL, 0x1fbbe891UL, 0x06a0d9d0UL, 0x5e7ef3ecUL, + 0x4765c2adUL, 0x6c48916eUL, 0x7553a02fUL, 0x3a1236e8UL, 0x230907a9UL, + 0x0824546aUL, 0x113f652bUL, 0x96a779e4UL, 0x8fbc48a5UL, 0xa4911b66UL, + 0xbd8a2a27UL, 0xf2cbbce0UL, 0xebd08da1UL, 0xc0fdde62UL, 0xd9e6ef23UL, + 0x14bce1bdUL, 0x0da7d0fcUL, 0x268a833fUL, 0x3f91b27eUL, 0x70d024b9UL, + 0x69cb15f8UL, 0x42e6463bUL, 0x5bfd777aUL, 0xdc656bb5UL, 0xc57e5af4UL, + 0xee530937UL, 0xf7483876UL, 0xb809aeb1UL, 0xa1129ff0UL, 0x8a3fcc33UL, + 0x9324fd72UL + }, + { + 0x00000000UL, 0x01c26a37UL, 0x0384d46eUL, 0x0246be59UL, 0x0709a8dcUL, + 0x06cbc2ebUL, 0x048d7cb2UL, 0x054f1685UL, 0x0e1351b8UL, 0x0fd13b8fUL, + 0x0d9785d6UL, 0x0c55efe1UL, 0x091af964UL, 0x08d89353UL, 0x0a9e2d0aUL, + 0x0b5c473dUL, 0x1c26a370UL, 0x1de4c947UL, 0x1fa2771eUL, 0x1e601d29UL, + 0x1b2f0bacUL, 0x1aed619bUL, 0x18abdfc2UL, 0x1969b5f5UL, 0x1235f2c8UL, + 0x13f798ffUL, 0x11b126a6UL, 0x10734c91UL, 0x153c5a14UL, 0x14fe3023UL, + 0x16b88e7aUL, 0x177ae44dUL, 0x384d46e0UL, 0x398f2cd7UL, 0x3bc9928eUL, + 0x3a0bf8b9UL, 0x3f44ee3cUL, 0x3e86840bUL, 0x3cc03a52UL, 0x3d025065UL, + 0x365e1758UL, 0x379c7d6fUL, 0x35dac336UL, 0x3418a901UL, 0x3157bf84UL, + 0x3095d5b3UL, 0x32d36beaUL, 0x331101ddUL, 0x246be590UL, 0x25a98fa7UL, + 0x27ef31feUL, 0x262d5bc9UL, 0x23624d4cUL, 0x22a0277bUL, 0x20e69922UL, + 0x2124f315UL, 0x2a78b428UL, 0x2bbade1fUL, 0x29fc6046UL, 0x283e0a71UL, + 0x2d711cf4UL, 0x2cb376c3UL, 0x2ef5c89aUL, 0x2f37a2adUL, 0x709a8dc0UL, + 0x7158e7f7UL, 0x731e59aeUL, 0x72dc3399UL, 0x7793251cUL, 0x76514f2bUL, + 0x7417f172UL, 0x75d59b45UL, 0x7e89dc78UL, 0x7f4bb64fUL, 0x7d0d0816UL, + 0x7ccf6221UL, 0x798074a4UL, 0x78421e93UL, 0x7a04a0caUL, 0x7bc6cafdUL, + 0x6cbc2eb0UL, 0x6d7e4487UL, 0x6f38fadeUL, 0x6efa90e9UL, 0x6bb5866cUL, + 0x6a77ec5bUL, 0x68315202UL, 0x69f33835UL, 0x62af7f08UL, 0x636d153fUL, + 0x612bab66UL, 0x60e9c151UL, 0x65a6d7d4UL, 0x6464bde3UL, 0x662203baUL, + 0x67e0698dUL, 0x48d7cb20UL, 0x4915a117UL, 0x4b531f4eUL, 0x4a917579UL, + 0x4fde63fcUL, 0x4e1c09cbUL, 0x4c5ab792UL, 0x4d98dda5UL, 0x46c49a98UL, + 0x4706f0afUL, 0x45404ef6UL, 0x448224c1UL, 0x41cd3244UL, 0x400f5873UL, + 0x4249e62aUL, 0x438b8c1dUL, 0x54f16850UL, 0x55330267UL, 0x5775bc3eUL, + 0x56b7d609UL, 0x53f8c08cUL, 0x523aaabbUL, 0x507c14e2UL, 0x51be7ed5UL, + 0x5ae239e8UL, 0x5b2053dfUL, 0x5966ed86UL, 0x58a487b1UL, 0x5deb9134UL, + 0x5c29fb03UL, 0x5e6f455aUL, 0x5fad2f6dUL, 0xe1351b80UL, 0xe0f771b7UL, + 0xe2b1cfeeUL, 0xe373a5d9UL, 0xe63cb35cUL, 0xe7fed96bUL, 0xe5b86732UL, + 0xe47a0d05UL, 0xef264a38UL, 0xeee4200fUL, 0xeca29e56UL, 0xed60f461UL, + 0xe82fe2e4UL, 0xe9ed88d3UL, 0xebab368aUL, 0xea695cbdUL, 0xfd13b8f0UL, + 0xfcd1d2c7UL, 0xfe976c9eUL, 0xff5506a9UL, 0xfa1a102cUL, 0xfbd87a1bUL, + 0xf99ec442UL, 0xf85cae75UL, 0xf300e948UL, 0xf2c2837fUL, 0xf0843d26UL, + 0xf1465711UL, 0xf4094194UL, 0xf5cb2ba3UL, 0xf78d95faUL, 0xf64fffcdUL, + 0xd9785d60UL, 0xd8ba3757UL, 0xdafc890eUL, 0xdb3ee339UL, 0xde71f5bcUL, + 0xdfb39f8bUL, 0xddf521d2UL, 0xdc374be5UL, 0xd76b0cd8UL, 0xd6a966efUL, + 0xd4efd8b6UL, 0xd52db281UL, 0xd062a404UL, 0xd1a0ce33UL, 0xd3e6706aUL, + 0xd2241a5dUL, 0xc55efe10UL, 0xc49c9427UL, 0xc6da2a7eUL, 0xc7184049UL, + 0xc25756ccUL, 0xc3953cfbUL, 0xc1d382a2UL, 0xc011e895UL, 0xcb4dafa8UL, + 0xca8fc59fUL, 0xc8c97bc6UL, 0xc90b11f1UL, 0xcc440774UL, 0xcd866d43UL, + 0xcfc0d31aUL, 0xce02b92dUL, 0x91af9640UL, 0x906dfc77UL, 0x922b422eUL, + 0x93e92819UL, 0x96a63e9cUL, 0x976454abUL, 0x9522eaf2UL, 0x94e080c5UL, + 0x9fbcc7f8UL, 0x9e7eadcfUL, 0x9c381396UL, 0x9dfa79a1UL, 0x98b56f24UL, + 0x99770513UL, 0x9b31bb4aUL, 0x9af3d17dUL, 0x8d893530UL, 0x8c4b5f07UL, + 0x8e0de15eUL, 0x8fcf8b69UL, 0x8a809decUL, 0x8b42f7dbUL, 0x89044982UL, + 0x88c623b5UL, 0x839a6488UL, 0x82580ebfUL, 0x801eb0e6UL, 0x81dcdad1UL, + 0x8493cc54UL, 0x8551a663UL, 0x8717183aUL, 0x86d5720dUL, 0xa9e2d0a0UL, + 0xa820ba97UL, 0xaa6604ceUL, 0xaba46ef9UL, 0xaeeb787cUL, 0xaf29124bUL, + 0xad6fac12UL, 0xacadc625UL, 0xa7f18118UL, 0xa633eb2fUL, 0xa4755576UL, + 0xa5b73f41UL, 0xa0f829c4UL, 0xa13a43f3UL, 0xa37cfdaaUL, 0xa2be979dUL, + 0xb5c473d0UL, 0xb40619e7UL, 0xb640a7beUL, 0xb782cd89UL, 0xb2cddb0cUL, + 0xb30fb13bUL, 0xb1490f62UL, 0xb08b6555UL, 0xbbd72268UL, 0xba15485fUL, + 0xb853f606UL, 0xb9919c31UL, 0xbcde8ab4UL, 0xbd1ce083UL, 0xbf5a5edaUL, + 0xbe9834edUL + }, + { + 0x00000000UL, 0xb8bc6765UL, 0xaa09c88bUL, 0x12b5afeeUL, 0x8f629757UL, + 0x37def032UL, 0x256b5fdcUL, 0x9dd738b9UL, 0xc5b428efUL, 0x7d084f8aUL, + 0x6fbde064UL, 0xd7018701UL, 0x4ad6bfb8UL, 0xf26ad8ddUL, 0xe0df7733UL, + 0x58631056UL, 0x5019579fUL, 0xe8a530faUL, 0xfa109f14UL, 0x42acf871UL, + 0xdf7bc0c8UL, 0x67c7a7adUL, 0x75720843UL, 0xcdce6f26UL, 0x95ad7f70UL, + 0x2d111815UL, 0x3fa4b7fbUL, 0x8718d09eUL, 0x1acfe827UL, 0xa2738f42UL, + 0xb0c620acUL, 0x087a47c9UL, 0xa032af3eUL, 0x188ec85bUL, 0x0a3b67b5UL, + 0xb28700d0UL, 0x2f503869UL, 0x97ec5f0cUL, 0x8559f0e2UL, 0x3de59787UL, + 0x658687d1UL, 0xdd3ae0b4UL, 0xcf8f4f5aUL, 0x7733283fUL, 0xeae41086UL, + 0x525877e3UL, 0x40edd80dUL, 0xf851bf68UL, 0xf02bf8a1UL, 0x48979fc4UL, + 0x5a22302aUL, 0xe29e574fUL, 0x7f496ff6UL, 0xc7f50893UL, 0xd540a77dUL, + 0x6dfcc018UL, 0x359fd04eUL, 0x8d23b72bUL, 0x9f9618c5UL, 0x272a7fa0UL, + 0xbafd4719UL, 0x0241207cUL, 0x10f48f92UL, 0xa848e8f7UL, 0x9b14583dUL, + 0x23a83f58UL, 0x311d90b6UL, 0x89a1f7d3UL, 0x1476cf6aUL, 0xaccaa80fUL, + 0xbe7f07e1UL, 0x06c36084UL, 0x5ea070d2UL, 0xe61c17b7UL, 0xf4a9b859UL, + 0x4c15df3cUL, 0xd1c2e785UL, 0x697e80e0UL, 0x7bcb2f0eUL, 0xc377486bUL, + 0xcb0d0fa2UL, 0x73b168c7UL, 0x6104c729UL, 0xd9b8a04cUL, 0x446f98f5UL, + 0xfcd3ff90UL, 0xee66507eUL, 0x56da371bUL, 0x0eb9274dUL, 0xb6054028UL, + 0xa4b0efc6UL, 0x1c0c88a3UL, 0x81dbb01aUL, 0x3967d77fUL, 0x2bd27891UL, + 0x936e1ff4UL, 0x3b26f703UL, 0x839a9066UL, 0x912f3f88UL, 0x299358edUL, + 0xb4446054UL, 0x0cf80731UL, 0x1e4da8dfUL, 0xa6f1cfbaUL, 0xfe92dfecUL, + 0x462eb889UL, 0x549b1767UL, 0xec277002UL, 0x71f048bbUL, 0xc94c2fdeUL, + 0xdbf98030UL, 0x6345e755UL, 0x6b3fa09cUL, 0xd383c7f9UL, 0xc1366817UL, + 0x798a0f72UL, 0xe45d37cbUL, 0x5ce150aeUL, 0x4e54ff40UL, 0xf6e89825UL, + 0xae8b8873UL, 0x1637ef16UL, 0x048240f8UL, 0xbc3e279dUL, 0x21e91f24UL, + 0x99557841UL, 0x8be0d7afUL, 0x335cb0caUL, 0xed59b63bUL, 0x55e5d15eUL, + 0x47507eb0UL, 0xffec19d5UL, 0x623b216cUL, 0xda874609UL, 0xc832e9e7UL, + 0x708e8e82UL, 0x28ed9ed4UL, 0x9051f9b1UL, 0x82e4565fUL, 0x3a58313aUL, + 0xa78f0983UL, 0x1f336ee6UL, 0x0d86c108UL, 0xb53aa66dUL, 0xbd40e1a4UL, + 0x05fc86c1UL, 0x1749292fUL, 0xaff54e4aUL, 0x322276f3UL, 0x8a9e1196UL, + 0x982bbe78UL, 0x2097d91dUL, 0x78f4c94bUL, 0xc048ae2eUL, 0xd2fd01c0UL, + 0x6a4166a5UL, 0xf7965e1cUL, 0x4f2a3979UL, 0x5d9f9697UL, 0xe523f1f2UL, + 0x4d6b1905UL, 0xf5d77e60UL, 0xe762d18eUL, 0x5fdeb6ebUL, 0xc2098e52UL, + 0x7ab5e937UL, 0x680046d9UL, 0xd0bc21bcUL, 0x88df31eaUL, 0x3063568fUL, + 0x22d6f961UL, 0x9a6a9e04UL, 0x07bda6bdUL, 0xbf01c1d8UL, 0xadb46e36UL, + 0x15080953UL, 0x1d724e9aUL, 0xa5ce29ffUL, 0xb77b8611UL, 0x0fc7e174UL, + 0x9210d9cdUL, 0x2aacbea8UL, 0x38191146UL, 0x80a57623UL, 0xd8c66675UL, + 0x607a0110UL, 0x72cfaefeUL, 0xca73c99bUL, 0x57a4f122UL, 0xef189647UL, + 0xfdad39a9UL, 0x45115eccUL, 0x764dee06UL, 0xcef18963UL, 0xdc44268dUL, + 0x64f841e8UL, 0xf92f7951UL, 0x41931e34UL, 0x5326b1daUL, 0xeb9ad6bfUL, + 0xb3f9c6e9UL, 0x0b45a18cUL, 0x19f00e62UL, 0xa14c6907UL, 0x3c9b51beUL, + 0x842736dbUL, 0x96929935UL, 0x2e2efe50UL, 0x2654b999UL, 0x9ee8defcUL, + 0x8c5d7112UL, 0x34e11677UL, 0xa9362eceUL, 0x118a49abUL, 0x033fe645UL, + 0xbb838120UL, 0xe3e09176UL, 0x5b5cf613UL, 0x49e959fdUL, 0xf1553e98UL, + 0x6c820621UL, 0xd43e6144UL, 0xc68bceaaUL, 0x7e37a9cfUL, 0xd67f4138UL, + 0x6ec3265dUL, 0x7c7689b3UL, 0xc4caeed6UL, 0x591dd66fUL, 0xe1a1b10aUL, + 0xf3141ee4UL, 0x4ba87981UL, 0x13cb69d7UL, 0xab770eb2UL, 0xb9c2a15cUL, + 0x017ec639UL, 0x9ca9fe80UL, 0x241599e5UL, 0x36a0360bUL, 0x8e1c516eUL, + 0x866616a7UL, 0x3eda71c2UL, 0x2c6fde2cUL, 0x94d3b949UL, 0x090481f0UL, + 0xb1b8e695UL, 0xa30d497bUL, 0x1bb12e1eUL, 0x43d23e48UL, 0xfb6e592dUL, + 0xe9dbf6c3UL, 0x516791a6UL, 0xccb0a91fUL, 0x740cce7aUL, 0x66b96194UL, + 0xde0506f1UL + }, + { + 0x00000000UL, 0x96300777UL, 0x2c610eeeUL, 0xba510999UL, 0x19c46d07UL, + 0x8ff46a70UL, 0x35a563e9UL, 0xa395649eUL, 0x3288db0eUL, 0xa4b8dc79UL, + 0x1ee9d5e0UL, 0x88d9d297UL, 0x2b4cb609UL, 0xbd7cb17eUL, 0x072db8e7UL, + 0x911dbf90UL, 0x6410b71dUL, 0xf220b06aUL, 0x4871b9f3UL, 0xde41be84UL, + 0x7dd4da1aUL, 0xebe4dd6dUL, 0x51b5d4f4UL, 0xc785d383UL, 0x56986c13UL, + 0xc0a86b64UL, 0x7af962fdUL, 0xecc9658aUL, 0x4f5c0114UL, 0xd96c0663UL, + 0x633d0ffaUL, 0xf50d088dUL, 0xc8206e3bUL, 0x5e10694cUL, 0xe44160d5UL, + 0x727167a2UL, 0xd1e4033cUL, 0x47d4044bUL, 0xfd850dd2UL, 0x6bb50aa5UL, + 0xfaa8b535UL, 0x6c98b242UL, 0xd6c9bbdbUL, 0x40f9bcacUL, 0xe36cd832UL, + 0x755cdf45UL, 0xcf0dd6dcUL, 0x593dd1abUL, 0xac30d926UL, 0x3a00de51UL, + 0x8051d7c8UL, 0x1661d0bfUL, 0xb5f4b421UL, 0x23c4b356UL, 0x9995bacfUL, + 0x0fa5bdb8UL, 0x9eb80228UL, 0x0888055fUL, 0xb2d90cc6UL, 0x24e90bb1UL, + 0x877c6f2fUL, 0x114c6858UL, 0xab1d61c1UL, 0x3d2d66b6UL, 0x9041dc76UL, + 0x0671db01UL, 0xbc20d298UL, 0x2a10d5efUL, 0x8985b171UL, 0x1fb5b606UL, + 0xa5e4bf9fUL, 0x33d4b8e8UL, 0xa2c90778UL, 0x34f9000fUL, 0x8ea80996UL, + 0x18980ee1UL, 0xbb0d6a7fUL, 0x2d3d6d08UL, 0x976c6491UL, 0x015c63e6UL, + 0xf4516b6bUL, 0x62616c1cUL, 0xd8306585UL, 0x4e0062f2UL, 0xed95066cUL, + 0x7ba5011bUL, 0xc1f40882UL, 0x57c40ff5UL, 0xc6d9b065UL, 0x50e9b712UL, + 0xeab8be8bUL, 0x7c88b9fcUL, 0xdf1ddd62UL, 0x492dda15UL, 0xf37cd38cUL, + 0x654cd4fbUL, 0x5861b24dUL, 0xce51b53aUL, 0x7400bca3UL, 0xe230bbd4UL, + 0x41a5df4aUL, 0xd795d83dUL, 0x6dc4d1a4UL, 0xfbf4d6d3UL, 0x6ae96943UL, + 0xfcd96e34UL, 0x468867adUL, 0xd0b860daUL, 0x732d0444UL, 0xe51d0333UL, + 0x5f4c0aaaUL, 0xc97c0dddUL, 0x3c710550UL, 0xaa410227UL, 0x10100bbeUL, + 0x86200cc9UL, 0x25b56857UL, 0xb3856f20UL, 0x09d466b9UL, 0x9fe461ceUL, + 0x0ef9de5eUL, 0x98c9d929UL, 0x2298d0b0UL, 0xb4a8d7c7UL, 0x173db359UL, + 0x810db42eUL, 0x3b5cbdb7UL, 0xad6cbac0UL, 0x2083b8edUL, 0xb6b3bf9aUL, + 0x0ce2b603UL, 0x9ad2b174UL, 0x3947d5eaUL, 0xaf77d29dUL, 0x1526db04UL, + 0x8316dc73UL, 0x120b63e3UL, 0x843b6494UL, 0x3e6a6d0dUL, 0xa85a6a7aUL, + 0x0bcf0ee4UL, 0x9dff0993UL, 0x27ae000aUL, 0xb19e077dUL, 0x44930ff0UL, + 0xd2a30887UL, 0x68f2011eUL, 0xfec20669UL, 0x5d5762f7UL, 0xcb676580UL, + 0x71366c19UL, 0xe7066b6eUL, 0x761bd4feUL, 0xe02bd389UL, 0x5a7ada10UL, + 0xcc4add67UL, 0x6fdfb9f9UL, 0xf9efbe8eUL, 0x43beb717UL, 0xd58eb060UL, + 0xe8a3d6d6UL, 0x7e93d1a1UL, 0xc4c2d838UL, 0x52f2df4fUL, 0xf167bbd1UL, + 0x6757bca6UL, 0xdd06b53fUL, 0x4b36b248UL, 0xda2b0dd8UL, 0x4c1b0aafUL, + 0xf64a0336UL, 0x607a0441UL, 0xc3ef60dfUL, 0x55df67a8UL, 0xef8e6e31UL, + 0x79be6946UL, 0x8cb361cbUL, 0x1a8366bcUL, 0xa0d26f25UL, 0x36e26852UL, + 0x95770cccUL, 0x03470bbbUL, 0xb9160222UL, 0x2f260555UL, 0xbe3bbac5UL, + 0x280bbdb2UL, 0x925ab42bUL, 0x046ab35cUL, 0xa7ffd7c2UL, 0x31cfd0b5UL, + 0x8b9ed92cUL, 0x1daede5bUL, 0xb0c2649bUL, 0x26f263ecUL, 0x9ca36a75UL, + 0x0a936d02UL, 0xa906099cUL, 0x3f360eebUL, 0x85670772UL, 0x13570005UL, + 0x824abf95UL, 0x147ab8e2UL, 0xae2bb17bUL, 0x381bb60cUL, 0x9b8ed292UL, + 0x0dbed5e5UL, 0xb7efdc7cUL, 0x21dfdb0bUL, 0xd4d2d386UL, 0x42e2d4f1UL, + 0xf8b3dd68UL, 0x6e83da1fUL, 0xcd16be81UL, 0x5b26b9f6UL, 0xe177b06fUL, + 0x7747b718UL, 0xe65a0888UL, 0x706a0fffUL, 0xca3b0666UL, 0x5c0b0111UL, + 0xff9e658fUL, 0x69ae62f8UL, 0xd3ff6b61UL, 0x45cf6c16UL, 0x78e20aa0UL, + 0xeed20dd7UL, 0x5483044eUL, 0xc2b30339UL, 0x612667a7UL, 0xf71660d0UL, + 0x4d476949UL, 0xdb776e3eUL, 0x4a6ad1aeUL, 0xdc5ad6d9UL, 0x660bdf40UL, + 0xf03bd837UL, 0x53aebca9UL, 0xc59ebbdeUL, 0x7fcfb247UL, 0xe9ffb530UL, + 0x1cf2bdbdUL, 0x8ac2bacaUL, 0x3093b353UL, 0xa6a3b424UL, 0x0536d0baUL, + 0x9306d7cdUL, 0x2957de54UL, 0xbf67d923UL, 0x2e7a66b3UL, 0xb84a61c4UL, + 0x021b685dUL, 0x942b6f2aUL, 0x37be0bb4UL, 0xa18e0cc3UL, 0x1bdf055aUL, + 0x8def022dUL + }, + { + 0x00000000UL, 0x41311b19UL, 0x82623632UL, 0xc3532d2bUL, 0x04c56c64UL, + 0x45f4777dUL, 0x86a75a56UL, 0xc796414fUL, 0x088ad9c8UL, 0x49bbc2d1UL, + 0x8ae8effaUL, 0xcbd9f4e3UL, 0x0c4fb5acUL, 0x4d7eaeb5UL, 0x8e2d839eUL, + 0xcf1c9887UL, 0x5112c24aUL, 0x1023d953UL, 0xd370f478UL, 0x9241ef61UL, + 0x55d7ae2eUL, 0x14e6b537UL, 0xd7b5981cUL, 0x96848305UL, 0x59981b82UL, + 0x18a9009bUL, 0xdbfa2db0UL, 0x9acb36a9UL, 0x5d5d77e6UL, 0x1c6c6cffUL, + 0xdf3f41d4UL, 0x9e0e5acdUL, 0xa2248495UL, 0xe3159f8cUL, 0x2046b2a7UL, + 0x6177a9beUL, 0xa6e1e8f1UL, 0xe7d0f3e8UL, 0x2483dec3UL, 0x65b2c5daUL, + 0xaaae5d5dUL, 0xeb9f4644UL, 0x28cc6b6fUL, 0x69fd7076UL, 0xae6b3139UL, + 0xef5a2a20UL, 0x2c09070bUL, 0x6d381c12UL, 0xf33646dfUL, 0xb2075dc6UL, + 0x715470edUL, 0x30656bf4UL, 0xf7f32abbUL, 0xb6c231a2UL, 0x75911c89UL, + 0x34a00790UL, 0xfbbc9f17UL, 0xba8d840eUL, 0x79dea925UL, 0x38efb23cUL, + 0xff79f373UL, 0xbe48e86aUL, 0x7d1bc541UL, 0x3c2ade58UL, 0x054f79f0UL, + 0x447e62e9UL, 0x872d4fc2UL, 0xc61c54dbUL, 0x018a1594UL, 0x40bb0e8dUL, + 0x83e823a6UL, 0xc2d938bfUL, 0x0dc5a038UL, 0x4cf4bb21UL, 0x8fa7960aUL, + 0xce968d13UL, 0x0900cc5cUL, 0x4831d745UL, 0x8b62fa6eUL, 0xca53e177UL, + 0x545dbbbaUL, 0x156ca0a3UL, 0xd63f8d88UL, 0x970e9691UL, 0x5098d7deUL, + 0x11a9ccc7UL, 0xd2fae1ecUL, 0x93cbfaf5UL, 0x5cd76272UL, 0x1de6796bUL, + 0xdeb55440UL, 0x9f844f59UL, 0x58120e16UL, 0x1923150fUL, 0xda703824UL, + 0x9b41233dUL, 0xa76bfd65UL, 0xe65ae67cUL, 0x2509cb57UL, 0x6438d04eUL, + 0xa3ae9101UL, 0xe29f8a18UL, 0x21cca733UL, 0x60fdbc2aUL, 0xafe124adUL, + 0xeed03fb4UL, 0x2d83129fUL, 0x6cb20986UL, 0xab2448c9UL, 0xea1553d0UL, + 0x29467efbUL, 0x687765e2UL, 0xf6793f2fUL, 0xb7482436UL, 0x741b091dUL, + 0x352a1204UL, 0xf2bc534bUL, 0xb38d4852UL, 0x70de6579UL, 0x31ef7e60UL, + 0xfef3e6e7UL, 0xbfc2fdfeUL, 0x7c91d0d5UL, 0x3da0cbccUL, 0xfa368a83UL, + 0xbb07919aUL, 0x7854bcb1UL, 0x3965a7a8UL, 0x4b98833bUL, 0x0aa99822UL, + 0xc9fab509UL, 0x88cbae10UL, 0x4f5def5fUL, 0x0e6cf446UL, 0xcd3fd96dUL, + 0x8c0ec274UL, 0x43125af3UL, 0x022341eaUL, 0xc1706cc1UL, 0x804177d8UL, + 0x47d73697UL, 0x06e62d8eUL, 0xc5b500a5UL, 0x84841bbcUL, 0x1a8a4171UL, + 0x5bbb5a68UL, 0x98e87743UL, 0xd9d96c5aUL, 0x1e4f2d15UL, 0x5f7e360cUL, + 0x9c2d1b27UL, 0xdd1c003eUL, 0x120098b9UL, 0x533183a0UL, 0x9062ae8bUL, + 0xd153b592UL, 0x16c5f4ddUL, 0x57f4efc4UL, 0x94a7c2efUL, 0xd596d9f6UL, + 0xe9bc07aeUL, 0xa88d1cb7UL, 0x6bde319cUL, 0x2aef2a85UL, 0xed796bcaUL, + 0xac4870d3UL, 0x6f1b5df8UL, 0x2e2a46e1UL, 0xe136de66UL, 0xa007c57fUL, + 0x6354e854UL, 0x2265f34dUL, 0xe5f3b202UL, 0xa4c2a91bUL, 0x67918430UL, + 0x26a09f29UL, 0xb8aec5e4UL, 0xf99fdefdUL, 0x3accf3d6UL, 0x7bfde8cfUL, + 0xbc6ba980UL, 0xfd5ab299UL, 0x3e099fb2UL, 0x7f3884abUL, 0xb0241c2cUL, + 0xf1150735UL, 0x32462a1eUL, 0x73773107UL, 0xb4e17048UL, 0xf5d06b51UL, + 0x3683467aUL, 0x77b25d63UL, 0x4ed7facbUL, 0x0fe6e1d2UL, 0xccb5ccf9UL, + 0x8d84d7e0UL, 0x4a1296afUL, 0x0b238db6UL, 0xc870a09dUL, 0x8941bb84UL, + 0x465d2303UL, 0x076c381aUL, 0xc43f1531UL, 0x850e0e28UL, 0x42984f67UL, + 0x03a9547eUL, 0xc0fa7955UL, 0x81cb624cUL, 0x1fc53881UL, 0x5ef42398UL, + 0x9da70eb3UL, 0xdc9615aaUL, 0x1b0054e5UL, 0x5a314ffcUL, 0x996262d7UL, + 0xd85379ceUL, 0x174fe149UL, 0x567efa50UL, 0x952dd77bUL, 0xd41ccc62UL, + 0x138a8d2dUL, 0x52bb9634UL, 0x91e8bb1fUL, 0xd0d9a006UL, 0xecf37e5eUL, + 0xadc26547UL, 0x6e91486cUL, 0x2fa05375UL, 0xe836123aUL, 0xa9070923UL, + 0x6a542408UL, 0x2b653f11UL, 0xe479a796UL, 0xa548bc8fUL, 0x661b91a4UL, + 0x272a8abdUL, 0xe0bccbf2UL, 0xa18dd0ebUL, 0x62defdc0UL, 0x23efe6d9UL, + 0xbde1bc14UL, 0xfcd0a70dUL, 0x3f838a26UL, 0x7eb2913fUL, 0xb924d070UL, + 0xf815cb69UL, 0x3b46e642UL, 0x7a77fd5bUL, 0xb56b65dcUL, 0xf45a7ec5UL, + 0x370953eeUL, 0x763848f7UL, 0xb1ae09b8UL, 0xf09f12a1UL, 0x33cc3f8aUL, + 0x72fd2493UL + }, + { + 0x00000000UL, 0x376ac201UL, 0x6ed48403UL, 0x59be4602UL, 0xdca80907UL, + 0xebc2cb06UL, 0xb27c8d04UL, 0x85164f05UL, 0xb851130eUL, 0x8f3bd10fUL, + 0xd685970dUL, 0xe1ef550cUL, 0x64f91a09UL, 0x5393d808UL, 0x0a2d9e0aUL, + 0x3d475c0bUL, 0x70a3261cUL, 0x47c9e41dUL, 0x1e77a21fUL, 0x291d601eUL, + 0xac0b2f1bUL, 0x9b61ed1aUL, 0xc2dfab18UL, 0xf5b56919UL, 0xc8f23512UL, + 0xff98f713UL, 0xa626b111UL, 0x914c7310UL, 0x145a3c15UL, 0x2330fe14UL, + 0x7a8eb816UL, 0x4de47a17UL, 0xe0464d38UL, 0xd72c8f39UL, 0x8e92c93bUL, + 0xb9f80b3aUL, 0x3cee443fUL, 0x0b84863eUL, 0x523ac03cUL, 0x6550023dUL, + 0x58175e36UL, 0x6f7d9c37UL, 0x36c3da35UL, 0x01a91834UL, 0x84bf5731UL, + 0xb3d59530UL, 0xea6bd332UL, 0xdd011133UL, 0x90e56b24UL, 0xa78fa925UL, + 0xfe31ef27UL, 0xc95b2d26UL, 0x4c4d6223UL, 0x7b27a022UL, 0x2299e620UL, + 0x15f32421UL, 0x28b4782aUL, 0x1fdeba2bUL, 0x4660fc29UL, 0x710a3e28UL, + 0xf41c712dUL, 0xc376b32cUL, 0x9ac8f52eUL, 0xada2372fUL, 0xc08d9a70UL, + 0xf7e75871UL, 0xae591e73UL, 0x9933dc72UL, 0x1c259377UL, 0x2b4f5176UL, + 0x72f11774UL, 0x459bd575UL, 0x78dc897eUL, 0x4fb64b7fUL, 0x16080d7dUL, + 0x2162cf7cUL, 0xa4748079UL, 0x931e4278UL, 0xcaa0047aUL, 0xfdcac67bUL, + 0xb02ebc6cUL, 0x87447e6dUL, 0xdefa386fUL, 0xe990fa6eUL, 0x6c86b56bUL, + 0x5bec776aUL, 0x02523168UL, 0x3538f369UL, 0x087faf62UL, 0x3f156d63UL, + 0x66ab2b61UL, 0x51c1e960UL, 0xd4d7a665UL, 0xe3bd6464UL, 0xba032266UL, + 0x8d69e067UL, 0x20cbd748UL, 0x17a11549UL, 0x4e1f534bUL, 0x7975914aUL, + 0xfc63de4fUL, 0xcb091c4eUL, 0x92b75a4cUL, 0xa5dd984dUL, 0x989ac446UL, + 0xaff00647UL, 0xf64e4045UL, 0xc1248244UL, 0x4432cd41UL, 0x73580f40UL, + 0x2ae64942UL, 0x1d8c8b43UL, 0x5068f154UL, 0x67023355UL, 0x3ebc7557UL, + 0x09d6b756UL, 0x8cc0f853UL, 0xbbaa3a52UL, 0xe2147c50UL, 0xd57ebe51UL, + 0xe839e25aUL, 0xdf53205bUL, 0x86ed6659UL, 0xb187a458UL, 0x3491eb5dUL, + 0x03fb295cUL, 0x5a456f5eUL, 0x6d2fad5fUL, 0x801b35e1UL, 0xb771f7e0UL, + 0xeecfb1e2UL, 0xd9a573e3UL, 0x5cb33ce6UL, 0x6bd9fee7UL, 0x3267b8e5UL, + 0x050d7ae4UL, 0x384a26efUL, 0x0f20e4eeUL, 0x569ea2ecUL, 0x61f460edUL, + 0xe4e22fe8UL, 0xd388ede9UL, 0x8a36abebUL, 0xbd5c69eaUL, 0xf0b813fdUL, + 0xc7d2d1fcUL, 0x9e6c97feUL, 0xa90655ffUL, 0x2c101afaUL, 0x1b7ad8fbUL, + 0x42c49ef9UL, 0x75ae5cf8UL, 0x48e900f3UL, 0x7f83c2f2UL, 0x263d84f0UL, + 0x115746f1UL, 0x944109f4UL, 0xa32bcbf5UL, 0xfa958df7UL, 0xcdff4ff6UL, + 0x605d78d9UL, 0x5737bad8UL, 0x0e89fcdaUL, 0x39e33edbUL, 0xbcf571deUL, + 0x8b9fb3dfUL, 0xd221f5ddUL, 0xe54b37dcUL, 0xd80c6bd7UL, 0xef66a9d6UL, + 0xb6d8efd4UL, 0x81b22dd5UL, 0x04a462d0UL, 0x33cea0d1UL, 0x6a70e6d3UL, + 0x5d1a24d2UL, 0x10fe5ec5UL, 0x27949cc4UL, 0x7e2adac6UL, 0x494018c7UL, + 0xcc5657c2UL, 0xfb3c95c3UL, 0xa282d3c1UL, 0x95e811c0UL, 0xa8af4dcbUL, + 0x9fc58fcaUL, 0xc67bc9c8UL, 0xf1110bc9UL, 0x740744ccUL, 0x436d86cdUL, + 0x1ad3c0cfUL, 0x2db902ceUL, 0x4096af91UL, 0x77fc6d90UL, 0x2e422b92UL, + 0x1928e993UL, 0x9c3ea696UL, 0xab546497UL, 0xf2ea2295UL, 0xc580e094UL, + 0xf8c7bc9fUL, 0xcfad7e9eUL, 0x9613389cUL, 0xa179fa9dUL, 0x246fb598UL, + 0x13057799UL, 0x4abb319bUL, 0x7dd1f39aUL, 0x3035898dUL, 0x075f4b8cUL, + 0x5ee10d8eUL, 0x698bcf8fUL, 0xec9d808aUL, 0xdbf7428bUL, 0x82490489UL, + 0xb523c688UL, 0x88649a83UL, 0xbf0e5882UL, 0xe6b01e80UL, 0xd1dadc81UL, + 0x54cc9384UL, 0x63a65185UL, 0x3a181787UL, 0x0d72d586UL, 0xa0d0e2a9UL, + 0x97ba20a8UL, 0xce0466aaUL, 0xf96ea4abUL, 0x7c78ebaeUL, 0x4b1229afUL, + 0x12ac6fadUL, 0x25c6adacUL, 0x1881f1a7UL, 0x2feb33a6UL, 0x765575a4UL, + 0x413fb7a5UL, 0xc429f8a0UL, 0xf3433aa1UL, 0xaafd7ca3UL, 0x9d97bea2UL, + 0xd073c4b5UL, 0xe71906b4UL, 0xbea740b6UL, 0x89cd82b7UL, 0x0cdbcdb2UL, + 0x3bb10fb3UL, 0x620f49b1UL, 0x55658bb0UL, 0x6822d7bbUL, 0x5f4815baUL, + 0x06f653b8UL, 0x319c91b9UL, 0xb48adebcUL, 0x83e01cbdUL, 0xda5e5abfUL, + 0xed3498beUL + }, + { + 0x00000000UL, 0x6567bcb8UL, 0x8bc809aaUL, 0xeeafb512UL, 0x5797628fUL, + 0x32f0de37UL, 0xdc5f6b25UL, 0xb938d79dUL, 0xef28b4c5UL, 0x8a4f087dUL, + 0x64e0bd6fUL, 0x018701d7UL, 0xb8bfd64aUL, 0xddd86af2UL, 0x3377dfe0UL, + 0x56106358UL, 0x9f571950UL, 0xfa30a5e8UL, 0x149f10faUL, 0x71f8ac42UL, + 0xc8c07bdfUL, 0xada7c767UL, 0x43087275UL, 0x266fcecdUL, 0x707fad95UL, + 0x1518112dUL, 0xfbb7a43fUL, 0x9ed01887UL, 0x27e8cf1aUL, 0x428f73a2UL, + 0xac20c6b0UL, 0xc9477a08UL, 0x3eaf32a0UL, 0x5bc88e18UL, 0xb5673b0aUL, + 0xd00087b2UL, 0x6938502fUL, 0x0c5fec97UL, 0xe2f05985UL, 0x8797e53dUL, + 0xd1878665UL, 0xb4e03addUL, 0x5a4f8fcfUL, 0x3f283377UL, 0x8610e4eaUL, + 0xe3775852UL, 0x0dd8ed40UL, 0x68bf51f8UL, 0xa1f82bf0UL, 0xc49f9748UL, + 0x2a30225aUL, 0x4f579ee2UL, 0xf66f497fUL, 0x9308f5c7UL, 0x7da740d5UL, + 0x18c0fc6dUL, 0x4ed09f35UL, 0x2bb7238dUL, 0xc518969fUL, 0xa07f2a27UL, + 0x1947fdbaUL, 0x7c204102UL, 0x928ff410UL, 0xf7e848a8UL, 0x3d58149bUL, + 0x583fa823UL, 0xb6901d31UL, 0xd3f7a189UL, 0x6acf7614UL, 0x0fa8caacUL, + 0xe1077fbeUL, 0x8460c306UL, 0xd270a05eUL, 0xb7171ce6UL, 0x59b8a9f4UL, + 0x3cdf154cUL, 0x85e7c2d1UL, 0xe0807e69UL, 0x0e2fcb7bUL, 0x6b4877c3UL, + 0xa20f0dcbUL, 0xc768b173UL, 0x29c70461UL, 0x4ca0b8d9UL, 0xf5986f44UL, + 0x90ffd3fcUL, 0x7e5066eeUL, 0x1b37da56UL, 0x4d27b90eUL, 0x284005b6UL, + 0xc6efb0a4UL, 0xa3880c1cUL, 0x1ab0db81UL, 0x7fd76739UL, 0x9178d22bUL, + 0xf41f6e93UL, 0x03f7263bUL, 0x66909a83UL, 0x883f2f91UL, 0xed589329UL, + 0x546044b4UL, 0x3107f80cUL, 0xdfa84d1eUL, 0xbacff1a6UL, 0xecdf92feUL, + 0x89b82e46UL, 0x67179b54UL, 0x027027ecUL, 0xbb48f071UL, 0xde2f4cc9UL, + 0x3080f9dbUL, 0x55e74563UL, 0x9ca03f6bUL, 0xf9c783d3UL, 0x176836c1UL, + 0x720f8a79UL, 0xcb375de4UL, 0xae50e15cUL, 0x40ff544eUL, 0x2598e8f6UL, + 0x73888baeUL, 0x16ef3716UL, 0xf8408204UL, 0x9d273ebcUL, 0x241fe921UL, + 0x41785599UL, 0xafd7e08bUL, 0xcab05c33UL, 0x3bb659edUL, 0x5ed1e555UL, + 0xb07e5047UL, 0xd519ecffUL, 0x6c213b62UL, 0x094687daUL, 0xe7e932c8UL, + 0x828e8e70UL, 0xd49eed28UL, 0xb1f95190UL, 0x5f56e482UL, 0x3a31583aUL, + 0x83098fa7UL, 0xe66e331fUL, 0x08c1860dUL, 0x6da63ab5UL, 0xa4e140bdUL, + 0xc186fc05UL, 0x2f294917UL, 0x4a4ef5afUL, 0xf3762232UL, 0x96119e8aUL, + 0x78be2b98UL, 0x1dd99720UL, 0x4bc9f478UL, 0x2eae48c0UL, 0xc001fdd2UL, + 0xa566416aUL, 0x1c5e96f7UL, 0x79392a4fUL, 0x97969f5dUL, 0xf2f123e5UL, + 0x05196b4dUL, 0x607ed7f5UL, 0x8ed162e7UL, 0xebb6de5fUL, 0x528e09c2UL, + 0x37e9b57aUL, 0xd9460068UL, 0xbc21bcd0UL, 0xea31df88UL, 0x8f566330UL, + 0x61f9d622UL, 0x049e6a9aUL, 0xbda6bd07UL, 0xd8c101bfUL, 0x366eb4adUL, + 0x53090815UL, 0x9a4e721dUL, 0xff29cea5UL, 0x11867bb7UL, 0x74e1c70fUL, + 0xcdd91092UL, 0xa8beac2aUL, 0x46111938UL, 0x2376a580UL, 0x7566c6d8UL, + 0x10017a60UL, 0xfeaecf72UL, 0x9bc973caUL, 0x22f1a457UL, 0x479618efUL, + 0xa939adfdUL, 0xcc5e1145UL, 0x06ee4d76UL, 0x6389f1ceUL, 0x8d2644dcUL, + 0xe841f864UL, 0x51792ff9UL, 0x341e9341UL, 0xdab12653UL, 0xbfd69aebUL, + 0xe9c6f9b3UL, 0x8ca1450bUL, 0x620ef019UL, 0x07694ca1UL, 0xbe519b3cUL, + 0xdb362784UL, 0x35999296UL, 0x50fe2e2eUL, 0x99b95426UL, 0xfcdee89eUL, + 0x12715d8cUL, 0x7716e134UL, 0xce2e36a9UL, 0xab498a11UL, 0x45e63f03UL, + 0x208183bbUL, 0x7691e0e3UL, 0x13f65c5bUL, 0xfd59e949UL, 0x983e55f1UL, + 0x2106826cUL, 0x44613ed4UL, 0xaace8bc6UL, 0xcfa9377eUL, 0x38417fd6UL, + 0x5d26c36eUL, 0xb389767cUL, 0xd6eecac4UL, 0x6fd61d59UL, 0x0ab1a1e1UL, + 0xe41e14f3UL, 0x8179a84bUL, 0xd769cb13UL, 0xb20e77abUL, 0x5ca1c2b9UL, + 0x39c67e01UL, 0x80fea99cUL, 0xe5991524UL, 0x0b36a036UL, 0x6e511c8eUL, + 0xa7166686UL, 0xc271da3eUL, 0x2cde6f2cUL, 0x49b9d394UL, 0xf0810409UL, + 0x95e6b8b1UL, 0x7b490da3UL, 0x1e2eb11bUL, 0x483ed243UL, 0x2d596efbUL, + 0xc3f6dbe9UL, 0xa6916751UL, 0x1fa9b0ccUL, 0x7ace0c74UL, 0x9461b966UL, + 0xf10605deUL +#endif + } +}; diff --git a/app/src/main/jni/zlib/deflate.h b/app/src/main/jni/zlib/deflate.h new file mode 100644 index 000000000..ce0299edd --- /dev/null +++ b/app/src/main/jni/zlib/deflate.h @@ -0,0 +1,346 @@ +/* deflate.h -- internal compression state + * Copyright (C) 1995-2012 Jean-loup Gailly + * For conditions of distribution and use, see copyright notice in zlib.h + */ + +/* WARNING: this file should *not* be used by applications. It is + part of the implementation of the compression library and is + subject to change. Applications should only use zlib.h. + */ + +/* @(#) $Id$ */ + +#ifndef DEFLATE_H +#define DEFLATE_H + +#include "zutil.h" + +/* define NO_GZIP when compiling if you want to disable gzip header and + trailer creation by deflate(). NO_GZIP would be used to avoid linking in + the crc code when it is not needed. For shared libraries, gzip encoding + should be left enabled. */ +#ifndef NO_GZIP +# define GZIP +#endif + +/* =========================================================================== + * Internal compression state. + */ + +#define LENGTH_CODES 29 +/* number of length codes, not counting the special END_BLOCK code */ + +#define LITERALS 256 +/* number of literal bytes 0..255 */ + +#define L_CODES (LITERALS+1+LENGTH_CODES) +/* number of Literal or Length codes, including the END_BLOCK code */ + +#define D_CODES 30 +/* number of distance codes */ + +#define BL_CODES 19 +/* number of codes used to transfer the bit lengths */ + +#define HEAP_SIZE (2*L_CODES+1) +/* maximum heap size */ + +#define MAX_BITS 15 +/* All codes must not exceed MAX_BITS bits */ + +#define Buf_size 16 +/* size of bit buffer in bi_buf */ + +#define INIT_STATE 42 +#define EXTRA_STATE 69 +#define NAME_STATE 73 +#define COMMENT_STATE 91 +#define HCRC_STATE 103 +#define BUSY_STATE 113 +#define FINISH_STATE 666 +/* Stream status */ + + +/* Data structure describing a single value and its code string. */ +typedef struct ct_data_s { + union { + ush freq; /* frequency count */ + ush code; /* bit string */ + } fc; + union { + ush dad; /* father node in Huffman tree */ + ush len; /* length of bit string */ + } dl; +} FAR ct_data; + +#define Freq fc.freq +#define Code fc.code +#define Dad dl.dad +#define Len dl.len + +typedef struct static_tree_desc_s static_tree_desc; + +typedef struct tree_desc_s { + ct_data *dyn_tree; /* the dynamic tree */ + int max_code; /* largest code with non zero frequency */ + static_tree_desc *stat_desc; /* the corresponding static tree */ +} FAR tree_desc; + +typedef ush Pos; +typedef Pos FAR Posf; +typedef unsigned IPos; + +/* A Pos is an index in the character window. We use short instead of int to + * save space in the various tables. IPos is used only for parameter passing. + */ + +typedef struct internal_state { + z_streamp strm; /* pointer back to this zlib stream */ + int status; /* as the name implies */ + Bytef *pending_buf; /* output still pending */ + ulg pending_buf_size; /* size of pending_buf */ + Bytef *pending_out; /* next pending byte to output to the stream */ + uInt pending; /* nb of bytes in the pending buffer */ + int wrap; /* bit 0 true for zlib, bit 1 true for gzip */ + gz_headerp gzhead; /* gzip header information to write */ + uInt gzindex; /* where in extra, name, or comment */ + Byte method; /* can only be DEFLATED */ + int last_flush; /* value of flush param for previous deflate call */ + + /* used by deflate.c: */ + + uInt w_size; /* LZ77 window size (32K by default) */ + uInt w_bits; /* log2(w_size) (8..16) */ + uInt w_mask; /* w_size - 1 */ + + Bytef *window; + /* Sliding window. Input bytes are read into the second half of the window, + * and move to the first half later to keep a dictionary of at least wSize + * bytes. With this organization, matches are limited to a distance of + * wSize-MAX_MATCH bytes, but this ensures that IO is always + * performed with a length multiple of the block size. Also, it limits + * the window size to 64K, which is quite useful on MSDOS. + * To do: use the user input buffer as sliding window. + */ + + ulg window_size; + /* Actual size of window: 2*wSize, except when the user input buffer + * is directly used as sliding window. + */ + + Posf *prev; + /* Link to older string with same hash index. To limit the size of this + * array to 64K, this link is maintained only for the last 32K strings. + * An index in this array is thus a window index modulo 32K. + */ + + Posf *head; /* Heads of the hash chains or NIL. */ + + uInt ins_h; /* hash index of string to be inserted */ + uInt hash_size; /* number of elements in hash table */ + uInt hash_bits; /* log2(hash_size) */ + uInt hash_mask; /* hash_size-1 */ + + uInt hash_shift; + /* Number of bits by which ins_h must be shifted at each input + * step. It must be such that after MIN_MATCH steps, the oldest + * byte no longer takes part in the hash key, that is: + * hash_shift * MIN_MATCH >= hash_bits + */ + + long block_start; + /* Window position at the beginning of the current output block. Gets + * negative when the window is moved backwards. + */ + + uInt match_length; /* length of best match */ + IPos prev_match; /* previous match */ + int match_available; /* set if previous match exists */ + uInt strstart; /* start of string to insert */ + uInt match_start; /* start of matching string */ + uInt lookahead; /* number of valid bytes ahead in window */ + + uInt prev_length; + /* Length of the best match at previous step. Matches not greater than this + * are discarded. This is used in the lazy match evaluation. + */ + + uInt max_chain_length; + /* To speed up deflation, hash chains are never searched beyond this + * length. A higher limit improves compression ratio but degrades the + * speed. + */ + + uInt max_lazy_match; + /* Attempt to find a better match only when the current match is strictly + * smaller than this value. This mechanism is used only for compression + * levels >= 4. + */ +# define max_insert_length max_lazy_match + /* Insert new strings in the hash table only if the match length is not + * greater than this length. This saves time but degrades compression. + * max_insert_length is used only for compression levels <= 3. + */ + + int level; /* compression level (1..9) */ + int strategy; /* favor or force Huffman coding*/ + + uInt good_match; + /* Use a faster search when the previous match is longer than this */ + + int nice_match; /* Stop searching when current match exceeds this */ + + /* used by trees.c: */ + /* Didn't use ct_data typedef below to suppress compiler warning */ + struct ct_data_s dyn_ltree[HEAP_SIZE]; /* literal and length tree */ + struct ct_data_s dyn_dtree[2*D_CODES+1]; /* distance tree */ + struct ct_data_s bl_tree[2*BL_CODES+1]; /* Huffman tree for bit lengths */ + + struct tree_desc_s l_desc; /* desc. for literal tree */ + struct tree_desc_s d_desc; /* desc. for distance tree */ + struct tree_desc_s bl_desc; /* desc. for bit length tree */ + + ush bl_count[MAX_BITS+1]; + /* number of codes at each bit length for an optimal tree */ + + int heap[2*L_CODES+1]; /* heap used to build the Huffman trees */ + int heap_len; /* number of elements in the heap */ + int heap_max; /* element of largest frequency */ + /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used. + * The same heap array is used to build all trees. + */ + + uch depth[2*L_CODES+1]; + /* Depth of each subtree used as tie breaker for trees of equal frequency + */ + + uchf *l_buf; /* buffer for literals or lengths */ + + uInt lit_bufsize; + /* Size of match buffer for literals/lengths. There are 4 reasons for + * limiting lit_bufsize to 64K: + * - frequencies can be kept in 16 bit counters + * - if compression is not successful for the first block, all input + * data is still in the window so we can still emit a stored block even + * when input comes from standard input. (This can also be done for + * all blocks if lit_bufsize is not greater than 32K.) + * - if compression is not successful for a file smaller than 64K, we can + * even emit a stored file instead of a stored block (saving 5 bytes). + * This is applicable only for zip (not gzip or zlib). + * - creating new Huffman trees less frequently may not provide fast + * adaptation to changes in the input data statistics. (Take for + * example a binary file with poorly compressible code followed by + * a highly compressible string table.) Smaller buffer sizes give + * fast adaptation but have of course the overhead of transmitting + * trees more frequently. + * - I can't count above 4 + */ + + uInt last_lit; /* running index in l_buf */ + + ushf *d_buf; + /* Buffer for distances. To simplify the code, d_buf and l_buf have + * the same number of elements. To use different lengths, an extra flag + * array would be necessary. + */ + + ulg opt_len; /* bit length of current block with optimal trees */ + ulg static_len; /* bit length of current block with static trees */ + uInt matches; /* number of string matches in current block */ + uInt insert; /* bytes at end of window left to insert */ + +#ifdef DEBUG + ulg compressed_len; /* total bit length of compressed file mod 2^32 */ + ulg bits_sent; /* bit length of compressed data sent mod 2^32 */ +#endif + + ush bi_buf; + /* Output buffer. bits are inserted starting at the bottom (least + * significant bits). + */ + int bi_valid; + /* Number of valid bits in bi_buf. All bits above the last valid bit + * are always zero. + */ + + ulg high_water; + /* High water mark offset in window for initialized bytes -- bytes above + * this are set to zero in order to avoid memory check warnings when + * longest match routines access bytes past the input. This is then + * updated to the new high water mark. + */ + +} FAR deflate_state; + +/* Output a byte on the stream. + * IN assertion: there is enough room in pending_buf. + */ +#define put_byte(s, c) {s->pending_buf[s->pending++] = (c);} + + +#define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1) +/* Minimum amount of lookahead, except at the end of the input file. + * See deflate.c for comments about the MIN_MATCH+1. + */ + +#define MAX_DIST(s) ((s)->w_size-MIN_LOOKAHEAD) +/* In order to simplify the code, particularly on 16 bit machines, match + * distances are limited to MAX_DIST instead of WSIZE. + */ + +#define WIN_INIT MAX_MATCH +/* Number of bytes after end of data in window to initialize in order to avoid + memory checker errors from longest match routines */ + + /* in trees.c */ +void ZLIB_INTERNAL _tr_init OF((deflate_state *s)); +int ZLIB_INTERNAL _tr_tally OF((deflate_state *s, unsigned dist, unsigned lc)); +void ZLIB_INTERNAL _tr_flush_block OF((deflate_state *s, charf *buf, + ulg stored_len, int last)); +void ZLIB_INTERNAL _tr_flush_bits OF((deflate_state *s)); +void ZLIB_INTERNAL _tr_align OF((deflate_state *s)); +void ZLIB_INTERNAL _tr_stored_block OF((deflate_state *s, charf *buf, + ulg stored_len, int last)); + +#define d_code(dist) \ + ((dist) < 256 ? _dist_code[dist] : _dist_code[256+((dist)>>7)]) +/* Mapping from a distance to a distance code. dist is the distance - 1 and + * must not have side effects. _dist_code[256] and _dist_code[257] are never + * used. + */ + +#ifndef DEBUG +/* Inline versions of _tr_tally for speed: */ + +#if defined(GEN_TREES_H) || !defined(STDC) + extern uch ZLIB_INTERNAL _length_code[]; + extern uch ZLIB_INTERNAL _dist_code[]; +#else + extern const uch ZLIB_INTERNAL _length_code[]; + extern const uch ZLIB_INTERNAL _dist_code[]; +#endif + +# define _tr_tally_lit(s, c, flush) \ + { uch cc = (c); \ + s->d_buf[s->last_lit] = 0; \ + s->l_buf[s->last_lit++] = cc; \ + s->dyn_ltree[cc].Freq++; \ + flush = (s->last_lit == s->lit_bufsize-1); \ + } +# define _tr_tally_dist(s, distance, length, flush) \ + { uch len = (length); \ + ush dist = (distance); \ + s->d_buf[s->last_lit] = dist; \ + s->l_buf[s->last_lit++] = len; \ + dist--; \ + s->dyn_ltree[_length_code[len]+LITERALS+1].Freq++; \ + s->dyn_dtree[d_code(dist)].Freq++; \ + flush = (s->last_lit == s->lit_bufsize-1); \ + } +#else +# define _tr_tally_lit(s, c, flush) flush = _tr_tally(s, 0, c) +# define _tr_tally_dist(s, distance, length, flush) \ + flush = _tr_tally(s, distance, length) +#endif + +#endif /* DEFLATE_H */ diff --git a/app/src/main/jni/zlib/gzguts.h b/app/src/main/jni/zlib/gzguts.h new file mode 100644 index 000000000..d87659d03 --- /dev/null +++ b/app/src/main/jni/zlib/gzguts.h @@ -0,0 +1,209 @@ +/* gzguts.h -- zlib internal header definitions for gz* operations + * Copyright (C) 2004, 2005, 2010, 2011, 2012, 2013 Mark Adler + * For conditions of distribution and use, see copyright notice in zlib.h + */ + +#ifdef _LARGEFILE64_SOURCE +# ifndef _LARGEFILE_SOURCE +# define _LARGEFILE_SOURCE 1 +# endif +# ifdef _FILE_OFFSET_BITS +# undef _FILE_OFFSET_BITS +# endif +#endif + +#ifdef HAVE_HIDDEN +# define ZLIB_INTERNAL __attribute__((visibility ("hidden"))) +#else +# define ZLIB_INTERNAL +#endif + +#include +#include "zlib.h" +#ifdef STDC +# include +# include +# include +#endif +#include + +#ifdef _WIN32 +# include +#endif + +#if defined(__TURBOC__) || defined(_MSC_VER) || defined(_WIN32) +# include +#endif + +#ifdef WINAPI_FAMILY +# define open _open +# define read _read +# define write _write +# define close _close +#endif + +#ifdef NO_DEFLATE /* for compatibility with old definition */ +# define NO_GZCOMPRESS +#endif + +#if defined(STDC99) || (defined(__TURBOC__) && __TURBOC__ >= 0x550) +# ifndef HAVE_VSNPRINTF +# define HAVE_VSNPRINTF +# endif +#endif + +#if defined(__CYGWIN__) +# ifndef HAVE_VSNPRINTF +# define HAVE_VSNPRINTF +# endif +#endif + +#if defined(MSDOS) && defined(__BORLANDC__) && (BORLANDC > 0x410) +# ifndef HAVE_VSNPRINTF +# define HAVE_VSNPRINTF +# endif +#endif + +#ifndef HAVE_VSNPRINTF +# ifdef MSDOS +/* vsnprintf may exist on some MS-DOS compilers (DJGPP?), + but for now we just assume it doesn't. */ +# define NO_vsnprintf +# endif +# ifdef __TURBOC__ +# define NO_vsnprintf +# endif +# ifdef WIN32 +/* In Win32, vsnprintf is available as the "non-ANSI" _vsnprintf. */ +# if !defined(vsnprintf) && !defined(NO_vsnprintf) +# if !defined(_MSC_VER) || ( defined(_MSC_VER) && _MSC_VER < 1500 ) +# define vsnprintf _vsnprintf +# endif +# endif +# endif +# ifdef __SASC +# define NO_vsnprintf +# endif +# ifdef VMS +# define NO_vsnprintf +# endif +# ifdef __OS400__ +# define NO_vsnprintf +# endif +# ifdef __MVS__ +# define NO_vsnprintf +# endif +#endif + +/* unlike snprintf (which is required in C99, yet still not supported by + Microsoft more than a decade later!), _snprintf does not guarantee null + termination of the result -- however this is only used in gzlib.c where + the result is assured to fit in the space provided */ +#ifdef _MSC_VER +# define snprintf _snprintf +#endif + +#ifndef local +# define local static +#endif +/* compile with -Dlocal if your debugger can't find static symbols */ + +/* gz* functions always use library allocation functions */ +#ifndef STDC + extern voidp malloc OF((uInt size)); + extern void free OF((voidpf ptr)); +#endif + +/* get errno and strerror definition */ +#if defined UNDER_CE +# include +# define zstrerror() gz_strwinerror((DWORD)GetLastError()) +#else +# ifndef NO_STRERROR +# include +# define zstrerror() strerror(errno) +# else +# define zstrerror() "stdio error (consult errno)" +# endif +#endif + +/* provide prototypes for these when building zlib without LFS */ +#if !defined(_LARGEFILE64_SOURCE) || _LFS64_LARGEFILE-0 == 0 + ZEXTERN gzFile ZEXPORT gzopen64 OF((const char *, const char *)); + ZEXTERN z_off64_t ZEXPORT gzseek64 OF((gzFile, z_off64_t, int)); + ZEXTERN z_off64_t ZEXPORT gztell64 OF((gzFile)); + ZEXTERN z_off64_t ZEXPORT gzoffset64 OF((gzFile)); +#endif + +/* default memLevel */ +#if MAX_MEM_LEVEL >= 8 +# define DEF_MEM_LEVEL 8 +#else +# define DEF_MEM_LEVEL MAX_MEM_LEVEL +#endif + +/* default i/o buffer size -- double this for output when reading (this and + twice this must be able to fit in an unsigned type) */ +#define GZBUFSIZE 8192 + +/* gzip modes, also provide a little integrity check on the passed structure */ +#define GZ_NONE 0 +#define GZ_READ 7247 +#define GZ_WRITE 31153 +#define GZ_APPEND 1 /* mode set to GZ_WRITE after the file is opened */ + +/* values for gz_state how */ +#define LOOK 0 /* look for a gzip header */ +#define COPY 1 /* copy input directly */ +#define GZIP 2 /* decompress a gzip stream */ + +/* internal gzip file state data structure */ +typedef struct { + /* exposed contents for gzgetc() macro */ + struct gzFile_s x; /* "x" for exposed */ + /* x.have: number of bytes available at x.next */ + /* x.next: next output data to deliver or write */ + /* x.pos: current position in uncompressed data */ + /* used for both reading and writing */ + int mode; /* see gzip modes above */ + int fd; /* file descriptor */ + char *path; /* path or fd for error messages */ + unsigned size; /* buffer size, zero if not allocated yet */ + unsigned want; /* requested buffer size, default is GZBUFSIZE */ + unsigned char *in; /* input buffer */ + unsigned char *out; /* output buffer (double-sized when reading) */ + int direct; /* 0 if processing gzip, 1 if transparent */ + /* just for reading */ + int how; /* 0: get header, 1: copy, 2: decompress */ + z_off64_t start; /* where the gzip data started, for rewinding */ + int eof; /* true if end of input file reached */ + int past; /* true if read requested past end */ + /* just for writing */ + int level; /* compression level */ + int strategy; /* compression strategy */ + /* seek request */ + z_off64_t skip; /* amount to skip (already rewound if backwards) */ + int seek; /* true if seek request pending */ + /* error information */ + int err; /* error code */ + char *msg; /* error message */ + /* zlib inflate or deflate stream */ + z_stream strm; /* stream structure in-place (not a pointer) */ +} gz_state; +typedef gz_state FAR *gz_statep; + +/* shared functions */ +void ZLIB_INTERNAL gz_error OF((gz_statep, int, const char *)); +#if defined UNDER_CE +char ZLIB_INTERNAL *gz_strwinerror OF((DWORD error)); +#endif + +/* GT_OFF(x), where x is an unsigned value, is true if x > maximum z_off64_t + value -- needed when comparing unsigned to z_off64_t, which is signed + (possible z_off64_t types off_t, off64_t, and long are all signed) */ +#ifdef INT_MAX +# define GT_OFF(x) (sizeof(int) == sizeof(z_off64_t) && (x) > INT_MAX) +#else +unsigned ZLIB_INTERNAL gz_intmax OF((void)); +# define GT_OFF(x) (sizeof(int) == sizeof(z_off64_t) && (x) > gz_intmax()) +#endif diff --git a/app/src/main/jni/zlib/inffast.c b/app/src/main/jni/zlib/inffast.c new file mode 100644 index 000000000..bda59ceb6 --- /dev/null +++ b/app/src/main/jni/zlib/inffast.c @@ -0,0 +1,340 @@ +/* inffast.c -- fast decoding + * Copyright (C) 1995-2008, 2010, 2013 Mark Adler + * For conditions of distribution and use, see copyright notice in zlib.h + */ + +#include "zutil.h" +#include "inftrees.h" +#include "inflate.h" +#include "inffast.h" + +#ifndef ASMINF + +/* Allow machine dependent optimization for post-increment or pre-increment. + Based on testing to date, + Pre-increment preferred for: + - PowerPC G3 (Adler) + - MIPS R5000 (Randers-Pehrson) + Post-increment preferred for: + - none + No measurable difference: + - Pentium III (Anderson) + - M68060 (Nikl) + */ +#ifdef POSTINC +# define OFF 0 +# define PUP(a) *(a)++ +#else +# define OFF 1 +# define PUP(a) *++(a) +#endif + +/* + Decode literal, length, and distance codes and write out the resulting + literal and match bytes until either not enough input or output is + available, an end-of-block is encountered, or a data error is encountered. + When large enough input and output buffers are supplied to inflate(), for + example, a 16K input buffer and a 64K output buffer, more than 95% of the + inflate execution time is spent in this routine. + + Entry assumptions: + + state->mode == LEN + strm->avail_in >= 6 + strm->avail_out >= 258 + start >= strm->avail_out + state->bits < 8 + + On return, state->mode is one of: + + LEN -- ran out of enough output space or enough available input + TYPE -- reached end of block code, inflate() to interpret next block + BAD -- error in block data + + Notes: + + - The maximum input bits used by a length/distance pair is 15 bits for the + length code, 5 bits for the length extra, 15 bits for the distance code, + and 13 bits for the distance extra. This totals 48 bits, or six bytes. + Therefore if strm->avail_in >= 6, then there is enough input to avoid + checking for available input while decoding. + + - The maximum bytes that a single length/distance pair can output is 258 + bytes, which is the maximum length that can be coded. inflate_fast() + requires strm->avail_out >= 258 for each loop to avoid checking for + output space. + */ +void ZLIB_INTERNAL inflate_fast(strm, start) +z_streamp strm; +unsigned start; /* inflate()'s starting value for strm->avail_out */ +{ + struct inflate_state FAR *state; + z_const unsigned char FAR *in; /* local strm->next_in */ + z_const unsigned char FAR *last; /* have enough input while in < last */ + unsigned char FAR *out; /* local strm->next_out */ + unsigned char FAR *beg; /* inflate()'s initial strm->next_out */ + unsigned char FAR *end; /* while out < end, enough space available */ +#ifdef INFLATE_STRICT + unsigned dmax; /* maximum distance from zlib header */ +#endif + unsigned wsize; /* window size or zero if not using window */ + unsigned whave; /* valid bytes in the window */ + unsigned wnext; /* window write index */ + unsigned char FAR *window; /* allocated sliding window, if wsize != 0 */ + unsigned long hold; /* local strm->hold */ + unsigned bits; /* local strm->bits */ + code const FAR *lcode; /* local strm->lencode */ + code const FAR *dcode; /* local strm->distcode */ + unsigned lmask; /* mask for first level of length codes */ + unsigned dmask; /* mask for first level of distance codes */ + code here; /* retrieved table entry */ + unsigned op; /* code bits, operation, extra bits, or */ + /* window position, window bytes to copy */ + unsigned len; /* match length, unused bytes */ + unsigned dist; /* match distance */ + unsigned char FAR *from; /* where to copy match from */ + + /* copy state to local variables */ + state = (struct inflate_state FAR *)strm->state; + in = strm->next_in - OFF; + last = in + (strm->avail_in - 5); + out = strm->next_out - OFF; + beg = out - (start - strm->avail_out); + end = out + (strm->avail_out - 257); +#ifdef INFLATE_STRICT + dmax = state->dmax; +#endif + wsize = state->wsize; + whave = state->whave; + wnext = state->wnext; + window = state->window; + hold = state->hold; + bits = state->bits; + lcode = state->lencode; + dcode = state->distcode; + lmask = (1U << state->lenbits) - 1; + dmask = (1U << state->distbits) - 1; + + /* decode literals and length/distances until end-of-block or not enough + input data or output space */ + do { + if (bits < 15) { + hold += (unsigned long)(PUP(in)) << bits; + bits += 8; + hold += (unsigned long)(PUP(in)) << bits; + bits += 8; + } + here = lcode[hold & lmask]; + dolen: + op = (unsigned)(here.bits); + hold >>= op; + bits -= op; + op = (unsigned)(here.op); + if (op == 0) { /* literal */ + Tracevv((stderr, here.val >= 0x20 && here.val < 0x7f ? + "inflate: literal '%c'\n" : + "inflate: literal 0x%02x\n", here.val)); + PUP(out) = (unsigned char)(here.val); + } + else if (op & 16) { /* length base */ + len = (unsigned)(here.val); + op &= 15; /* number of extra bits */ + if (op) { + if (bits < op) { + hold += (unsigned long)(PUP(in)) << bits; + bits += 8; + } + len += (unsigned)hold & ((1U << op) - 1); + hold >>= op; + bits -= op; + } + Tracevv((stderr, "inflate: length %u\n", len)); + if (bits < 15) { + hold += (unsigned long)(PUP(in)) << bits; + bits += 8; + hold += (unsigned long)(PUP(in)) << bits; + bits += 8; + } + here = dcode[hold & dmask]; + dodist: + op = (unsigned)(here.bits); + hold >>= op; + bits -= op; + op = (unsigned)(here.op); + if (op & 16) { /* distance base */ + dist = (unsigned)(here.val); + op &= 15; /* number of extra bits */ + if (bits < op) { + hold += (unsigned long)(PUP(in)) << bits; + bits += 8; + if (bits < op) { + hold += (unsigned long)(PUP(in)) << bits; + bits += 8; + } + } + dist += (unsigned)hold & ((1U << op) - 1); +#ifdef INFLATE_STRICT + if (dist > dmax) { + strm->msg = (char *)"invalid distance too far back"; + state->mode = BAD; + break; + } +#endif + hold >>= op; + bits -= op; + Tracevv((stderr, "inflate: distance %u\n", dist)); + op = (unsigned)(out - beg); /* max distance in output */ + if (dist > op) { /* see if copy from window */ + op = dist - op; /* distance back in window */ + if (op > whave) { + if (state->sane) { + strm->msg = + (char *)"invalid distance too far back"; + state->mode = BAD; + break; + } +#ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR + if (len <= op - whave) { + do { + PUP(out) = 0; + } while (--len); + continue; + } + len -= op - whave; + do { + PUP(out) = 0; + } while (--op > whave); + if (op == 0) { + from = out - dist; + do { + PUP(out) = PUP(from); + } while (--len); + continue; + } +#endif + } + from = window - OFF; + if (wnext == 0) { /* very common case */ + from += wsize - op; + if (op < len) { /* some from window */ + len -= op; + do { + PUP(out) = PUP(from); + } while (--op); + from = out - dist; /* rest from output */ + } + } + else if (wnext < op) { /* wrap around window */ + from += wsize + wnext - op; + op -= wnext; + if (op < len) { /* some from end of window */ + len -= op; + do { + PUP(out) = PUP(from); + } while (--op); + from = window - OFF; + if (wnext < len) { /* some from start of window */ + op = wnext; + len -= op; + do { + PUP(out) = PUP(from); + } while (--op); + from = out - dist; /* rest from output */ + } + } + } + else { /* contiguous in window */ + from += wnext - op; + if (op < len) { /* some from window */ + len -= op; + do { + PUP(out) = PUP(from); + } while (--op); + from = out - dist; /* rest from output */ + } + } + while (len > 2) { + PUP(out) = PUP(from); + PUP(out) = PUP(from); + PUP(out) = PUP(from); + len -= 3; + } + if (len) { + PUP(out) = PUP(from); + if (len > 1) + PUP(out) = PUP(from); + } + } + else { + from = out - dist; /* copy direct from output */ + do { /* minimum length is three */ + PUP(out) = PUP(from); + PUP(out) = PUP(from); + PUP(out) = PUP(from); + len -= 3; + } while (len > 2); + if (len) { + PUP(out) = PUP(from); + if (len > 1) + PUP(out) = PUP(from); + } + } + } + else if ((op & 64) == 0) { /* 2nd level distance code */ + here = dcode[here.val + (hold & ((1U << op) - 1))]; + goto dodist; + } + else { + strm->msg = (char *)"invalid distance code"; + state->mode = BAD; + break; + } + } + else if ((op & 64) == 0) { /* 2nd level length code */ + here = lcode[here.val + (hold & ((1U << op) - 1))]; + goto dolen; + } + else if (op & 32) { /* end-of-block */ + Tracevv((stderr, "inflate: end of block\n")); + state->mode = TYPE; + break; + } + else { + strm->msg = (char *)"invalid literal/length code"; + state->mode = BAD; + break; + } + } while (in < last && out < end); + + /* return unused bytes (on entry, bits < 8, so in won't go too far back) */ + len = bits >> 3; + in -= len; + bits -= len << 3; + hold &= (1U << bits) - 1; + + /* update state and return */ + strm->next_in = in + OFF; + strm->next_out = out + OFF; + strm->avail_in = (unsigned)(in < last ? 5 + (last - in) : 5 - (in - last)); + strm->avail_out = (unsigned)(out < end ? + 257 + (end - out) : 257 - (out - end)); + state->hold = hold; + state->bits = bits; + return; +} + +/* + inflate_fast() speedups that turned out slower (on a PowerPC G3 750CXe): + - Using bit fields for code structure + - Different op definition to avoid & for extra bits (do & for table bits) + - Three separate decoding do-loops for direct, window, and wnext == 0 + - Special case for distance > 1 copies to do overlapped load and store copy + - Explicit branch predictions (based on measured branch probabilities) + - Deferring match copy and interspersed it with decoding subsequent codes + - Swapping literal/length else + - Swapping window/direct else + - Larger unrolled copy loops (three is about right) + - Moving len -= 3 statement into middle of loop + */ + +#endif /* !ASMINF */ diff --git a/app/src/main/jni/zlib/inffast.h b/app/src/main/jni/zlib/inffast.h new file mode 100644 index 000000000..e5c1aa4ca --- /dev/null +++ b/app/src/main/jni/zlib/inffast.h @@ -0,0 +1,11 @@ +/* inffast.h -- header to use inffast.c + * Copyright (C) 1995-2003, 2010 Mark Adler + * For conditions of distribution and use, see copyright notice in zlib.h + */ + +/* WARNING: this file should *not* be used by applications. It is + part of the implementation of the compression library and is + subject to change. Applications should only use zlib.h. + */ + +void ZLIB_INTERNAL inflate_fast OF((z_streamp strm, unsigned start)); diff --git a/app/src/main/jni/zlib/inffixed.h b/app/src/main/jni/zlib/inffixed.h new file mode 100644 index 000000000..d62832776 --- /dev/null +++ b/app/src/main/jni/zlib/inffixed.h @@ -0,0 +1,94 @@ + /* inffixed.h -- table for decoding fixed codes + * Generated automatically by makefixed(). + */ + + /* WARNING: this file should *not* be used by applications. + It is part of the implementation of this library and is + subject to change. Applications should only use zlib.h. + */ + + static const code lenfix[512] = { + {96,7,0},{0,8,80},{0,8,16},{20,8,115},{18,7,31},{0,8,112},{0,8,48}, + {0,9,192},{16,7,10},{0,8,96},{0,8,32},{0,9,160},{0,8,0},{0,8,128}, + {0,8,64},{0,9,224},{16,7,6},{0,8,88},{0,8,24},{0,9,144},{19,7,59}, + {0,8,120},{0,8,56},{0,9,208},{17,7,17},{0,8,104},{0,8,40},{0,9,176}, + {0,8,8},{0,8,136},{0,8,72},{0,9,240},{16,7,4},{0,8,84},{0,8,20}, + {21,8,227},{19,7,43},{0,8,116},{0,8,52},{0,9,200},{17,7,13},{0,8,100}, + {0,8,36},{0,9,168},{0,8,4},{0,8,132},{0,8,68},{0,9,232},{16,7,8}, + {0,8,92},{0,8,28},{0,9,152},{20,7,83},{0,8,124},{0,8,60},{0,9,216}, + {18,7,23},{0,8,108},{0,8,44},{0,9,184},{0,8,12},{0,8,140},{0,8,76}, + {0,9,248},{16,7,3},{0,8,82},{0,8,18},{21,8,163},{19,7,35},{0,8,114}, + {0,8,50},{0,9,196},{17,7,11},{0,8,98},{0,8,34},{0,9,164},{0,8,2}, + {0,8,130},{0,8,66},{0,9,228},{16,7,7},{0,8,90},{0,8,26},{0,9,148}, + {20,7,67},{0,8,122},{0,8,58},{0,9,212},{18,7,19},{0,8,106},{0,8,42}, + {0,9,180},{0,8,10},{0,8,138},{0,8,74},{0,9,244},{16,7,5},{0,8,86}, + {0,8,22},{64,8,0},{19,7,51},{0,8,118},{0,8,54},{0,9,204},{17,7,15}, + {0,8,102},{0,8,38},{0,9,172},{0,8,6},{0,8,134},{0,8,70},{0,9,236}, + {16,7,9},{0,8,94},{0,8,30},{0,9,156},{20,7,99},{0,8,126},{0,8,62}, + {0,9,220},{18,7,27},{0,8,110},{0,8,46},{0,9,188},{0,8,14},{0,8,142}, + {0,8,78},{0,9,252},{96,7,0},{0,8,81},{0,8,17},{21,8,131},{18,7,31}, + {0,8,113},{0,8,49},{0,9,194},{16,7,10},{0,8,97},{0,8,33},{0,9,162}, + {0,8,1},{0,8,129},{0,8,65},{0,9,226},{16,7,6},{0,8,89},{0,8,25}, + {0,9,146},{19,7,59},{0,8,121},{0,8,57},{0,9,210},{17,7,17},{0,8,105}, + {0,8,41},{0,9,178},{0,8,9},{0,8,137},{0,8,73},{0,9,242},{16,7,4}, + {0,8,85},{0,8,21},{16,8,258},{19,7,43},{0,8,117},{0,8,53},{0,9,202}, + {17,7,13},{0,8,101},{0,8,37},{0,9,170},{0,8,5},{0,8,133},{0,8,69}, + {0,9,234},{16,7,8},{0,8,93},{0,8,29},{0,9,154},{20,7,83},{0,8,125}, + {0,8,61},{0,9,218},{18,7,23},{0,8,109},{0,8,45},{0,9,186},{0,8,13}, + {0,8,141},{0,8,77},{0,9,250},{16,7,3},{0,8,83},{0,8,19},{21,8,195}, + {19,7,35},{0,8,115},{0,8,51},{0,9,198},{17,7,11},{0,8,99},{0,8,35}, + {0,9,166},{0,8,3},{0,8,131},{0,8,67},{0,9,230},{16,7,7},{0,8,91}, + {0,8,27},{0,9,150},{20,7,67},{0,8,123},{0,8,59},{0,9,214},{18,7,19}, + {0,8,107},{0,8,43},{0,9,182},{0,8,11},{0,8,139},{0,8,75},{0,9,246}, + {16,7,5},{0,8,87},{0,8,23},{64,8,0},{19,7,51},{0,8,119},{0,8,55}, + {0,9,206},{17,7,15},{0,8,103},{0,8,39},{0,9,174},{0,8,7},{0,8,135}, + {0,8,71},{0,9,238},{16,7,9},{0,8,95},{0,8,31},{0,9,158},{20,7,99}, + {0,8,127},{0,8,63},{0,9,222},{18,7,27},{0,8,111},{0,8,47},{0,9,190}, + {0,8,15},{0,8,143},{0,8,79},{0,9,254},{96,7,0},{0,8,80},{0,8,16}, + {20,8,115},{18,7,31},{0,8,112},{0,8,48},{0,9,193},{16,7,10},{0,8,96}, + {0,8,32},{0,9,161},{0,8,0},{0,8,128},{0,8,64},{0,9,225},{16,7,6}, + {0,8,88},{0,8,24},{0,9,145},{19,7,59},{0,8,120},{0,8,56},{0,9,209}, + {17,7,17},{0,8,104},{0,8,40},{0,9,177},{0,8,8},{0,8,136},{0,8,72}, + {0,9,241},{16,7,4},{0,8,84},{0,8,20},{21,8,227},{19,7,43},{0,8,116}, + {0,8,52},{0,9,201},{17,7,13},{0,8,100},{0,8,36},{0,9,169},{0,8,4}, + {0,8,132},{0,8,68},{0,9,233},{16,7,8},{0,8,92},{0,8,28},{0,9,153}, + {20,7,83},{0,8,124},{0,8,60},{0,9,217},{18,7,23},{0,8,108},{0,8,44}, + {0,9,185},{0,8,12},{0,8,140},{0,8,76},{0,9,249},{16,7,3},{0,8,82}, + {0,8,18},{21,8,163},{19,7,35},{0,8,114},{0,8,50},{0,9,197},{17,7,11}, + {0,8,98},{0,8,34},{0,9,165},{0,8,2},{0,8,130},{0,8,66},{0,9,229}, + {16,7,7},{0,8,90},{0,8,26},{0,9,149},{20,7,67},{0,8,122},{0,8,58}, + {0,9,213},{18,7,19},{0,8,106},{0,8,42},{0,9,181},{0,8,10},{0,8,138}, + {0,8,74},{0,9,245},{16,7,5},{0,8,86},{0,8,22},{64,8,0},{19,7,51}, + {0,8,118},{0,8,54},{0,9,205},{17,7,15},{0,8,102},{0,8,38},{0,9,173}, + {0,8,6},{0,8,134},{0,8,70},{0,9,237},{16,7,9},{0,8,94},{0,8,30}, + {0,9,157},{20,7,99},{0,8,126},{0,8,62},{0,9,221},{18,7,27},{0,8,110}, + {0,8,46},{0,9,189},{0,8,14},{0,8,142},{0,8,78},{0,9,253},{96,7,0}, + {0,8,81},{0,8,17},{21,8,131},{18,7,31},{0,8,113},{0,8,49},{0,9,195}, + {16,7,10},{0,8,97},{0,8,33},{0,9,163},{0,8,1},{0,8,129},{0,8,65}, + {0,9,227},{16,7,6},{0,8,89},{0,8,25},{0,9,147},{19,7,59},{0,8,121}, + {0,8,57},{0,9,211},{17,7,17},{0,8,105},{0,8,41},{0,9,179},{0,8,9}, + {0,8,137},{0,8,73},{0,9,243},{16,7,4},{0,8,85},{0,8,21},{16,8,258}, + {19,7,43},{0,8,117},{0,8,53},{0,9,203},{17,7,13},{0,8,101},{0,8,37}, + {0,9,171},{0,8,5},{0,8,133},{0,8,69},{0,9,235},{16,7,8},{0,8,93}, + {0,8,29},{0,9,155},{20,7,83},{0,8,125},{0,8,61},{0,9,219},{18,7,23}, + {0,8,109},{0,8,45},{0,9,187},{0,8,13},{0,8,141},{0,8,77},{0,9,251}, + {16,7,3},{0,8,83},{0,8,19},{21,8,195},{19,7,35},{0,8,115},{0,8,51}, + {0,9,199},{17,7,11},{0,8,99},{0,8,35},{0,9,167},{0,8,3},{0,8,131}, + {0,8,67},{0,9,231},{16,7,7},{0,8,91},{0,8,27},{0,9,151},{20,7,67}, + {0,8,123},{0,8,59},{0,9,215},{18,7,19},{0,8,107},{0,8,43},{0,9,183}, + {0,8,11},{0,8,139},{0,8,75},{0,9,247},{16,7,5},{0,8,87},{0,8,23}, + {64,8,0},{19,7,51},{0,8,119},{0,8,55},{0,9,207},{17,7,15},{0,8,103}, + {0,8,39},{0,9,175},{0,8,7},{0,8,135},{0,8,71},{0,9,239},{16,7,9}, + {0,8,95},{0,8,31},{0,9,159},{20,7,99},{0,8,127},{0,8,63},{0,9,223}, + {18,7,27},{0,8,111},{0,8,47},{0,9,191},{0,8,15},{0,8,143},{0,8,79}, + {0,9,255} + }; + + static const code distfix[32] = { + {16,5,1},{23,5,257},{19,5,17},{27,5,4097},{17,5,5},{25,5,1025}, + {21,5,65},{29,5,16385},{16,5,3},{24,5,513},{20,5,33},{28,5,8193}, + {18,5,9},{26,5,2049},{22,5,129},{64,5,0},{16,5,2},{23,5,385}, + {19,5,25},{27,5,6145},{17,5,7},{25,5,1537},{21,5,97},{29,5,24577}, + {16,5,4},{24,5,769},{20,5,49},{28,5,12289},{18,5,13},{26,5,3073}, + {22,5,193},{64,5,0} + }; diff --git a/app/src/main/jni/zlib/inflate.c b/app/src/main/jni/zlib/inflate.c new file mode 100644 index 000000000..870f89bb4 --- /dev/null +++ b/app/src/main/jni/zlib/inflate.c @@ -0,0 +1,1512 @@ +/* inflate.c -- zlib decompression + * Copyright (C) 1995-2012 Mark Adler + * For conditions of distribution and use, see copyright notice in zlib.h + */ + +/* + * Change history: + * + * 1.2.beta0 24 Nov 2002 + * - First version -- complete rewrite of inflate to simplify code, avoid + * creation of window when not needed, minimize use of window when it is + * needed, make inffast.c even faster, implement gzip decoding, and to + * improve code readability and style over the previous zlib inflate code + * + * 1.2.beta1 25 Nov 2002 + * - Use pointers for available input and output checking in inffast.c + * - Remove input and output counters in inffast.c + * - Change inffast.c entry and loop from avail_in >= 7 to >= 6 + * - Remove unnecessary second byte pull from length extra in inffast.c + * - Unroll direct copy to three copies per loop in inffast.c + * + * 1.2.beta2 4 Dec 2002 + * - Change external routine names to reduce potential conflicts + * - Correct filename to inffixed.h for fixed tables in inflate.c + * - Make hbuf[] unsigned char to match parameter type in inflate.c + * - Change strm->next_out[-state->offset] to *(strm->next_out - state->offset) + * to avoid negation problem on Alphas (64 bit) in inflate.c + * + * 1.2.beta3 22 Dec 2002 + * - Add comments on state->bits assertion in inffast.c + * - Add comments on op field in inftrees.h + * - Fix bug in reuse of allocated window after inflateReset() + * - Remove bit fields--back to byte structure for speed + * - Remove distance extra == 0 check in inflate_fast()--only helps for lengths + * - Change post-increments to pre-increments in inflate_fast(), PPC biased? + * - Add compile time option, POSTINC, to use post-increments instead (Intel?) + * - Make MATCH copy in inflate() much faster for when inflate_fast() not used + * - Use local copies of stream next and avail values, as well as local bit + * buffer and bit count in inflate()--for speed when inflate_fast() not used + * + * 1.2.beta4 1 Jan 2003 + * - Split ptr - 257 statements in inflate_table() to avoid compiler warnings + * - Move a comment on output buffer sizes from inffast.c to inflate.c + * - Add comments in inffast.c to introduce the inflate_fast() routine + * - Rearrange window copies in inflate_fast() for speed and simplification + * - Unroll last copy for window match in inflate_fast() + * - Use local copies of window variables in inflate_fast() for speed + * - Pull out common wnext == 0 case for speed in inflate_fast() + * - Make op and len in inflate_fast() unsigned for consistency + * - Add FAR to lcode and dcode declarations in inflate_fast() + * - Simplified bad distance check in inflate_fast() + * - Added inflateBackInit(), inflateBack(), and inflateBackEnd() in new + * source file infback.c to provide a call-back interface to inflate for + * programs like gzip and unzip -- uses window as output buffer to avoid + * window copying + * + * 1.2.beta5 1 Jan 2003 + * - Improved inflateBack() interface to allow the caller to provide initial + * input in strm. + * - Fixed stored blocks bug in inflateBack() + * + * 1.2.beta6 4 Jan 2003 + * - Added comments in inffast.c on effectiveness of POSTINC + * - Typecasting all around to reduce compiler warnings + * - Changed loops from while (1) or do {} while (1) to for (;;), again to + * make compilers happy + * - Changed type of window in inflateBackInit() to unsigned char * + * + * 1.2.beta7 27 Jan 2003 + * - Changed many types to unsigned or unsigned short to avoid warnings + * - Added inflateCopy() function + * + * 1.2.0 9 Mar 2003 + * - Changed inflateBack() interface to provide separate opaque descriptors + * for the in() and out() functions + * - Changed inflateBack() argument and in_func typedef to swap the length + * and buffer address return values for the input function + * - Check next_in and next_out for Z_NULL on entry to inflate() + * + * The history for versions after 1.2.0 are in ChangeLog in zlib distribution. + */ + +#include "zutil.h" +#include "inftrees.h" +#include "inflate.h" +#include "inffast.h" + +#ifdef MAKEFIXED +# ifndef BUILDFIXED +# define BUILDFIXED +# endif +#endif + +/* function prototypes */ +local void fixedtables OF((struct inflate_state FAR *state)); +local int updatewindow OF((z_streamp strm, const unsigned char FAR *end, + unsigned copy)); +#ifdef BUILDFIXED + void makefixed OF((void)); +#endif +local unsigned syncsearch OF((unsigned FAR *have, const unsigned char FAR *buf, + unsigned len)); + +int ZEXPORT inflateResetKeep(strm) +z_streamp strm; +{ + struct inflate_state FAR *state; + + if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; + state = (struct inflate_state FAR *)strm->state; + strm->total_in = strm->total_out = state->total = 0; + strm->msg = Z_NULL; + if (state->wrap) /* to support ill-conceived Java test suite */ + strm->adler = state->wrap & 1; + state->mode = HEAD; + state->last = 0; + state->havedict = 0; + state->dmax = 32768U; + state->head = Z_NULL; + state->hold = 0; + state->bits = 0; + state->lencode = state->distcode = state->next = state->codes; + state->sane = 1; + state->back = -1; + Tracev((stderr, "inflate: reset\n")); + return Z_OK; +} + +int ZEXPORT inflateReset(strm) +z_streamp strm; +{ + struct inflate_state FAR *state; + + if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; + state = (struct inflate_state FAR *)strm->state; + state->wsize = 0; + state->whave = 0; + state->wnext = 0; + return inflateResetKeep(strm); +} + +int ZEXPORT inflateReset2(strm, windowBits) +z_streamp strm; +int windowBits; +{ + int wrap; + struct inflate_state FAR *state; + + /* get the state */ + if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; + state = (struct inflate_state FAR *)strm->state; + + /* extract wrap request from windowBits parameter */ + if (windowBits < 0) { + wrap = 0; + windowBits = -windowBits; + } + else { + wrap = (windowBits >> 4) + 1; +#ifdef GUNZIP + if (windowBits < 48) + windowBits &= 15; +#endif + } + + /* set number of window bits, free window if different */ + if (windowBits && (windowBits < 8 || windowBits > 15)) + return Z_STREAM_ERROR; + if (state->window != Z_NULL && state->wbits != (unsigned)windowBits) { + ZFREE(strm, state->window); + state->window = Z_NULL; + } + + /* update state and reset the rest of it */ + state->wrap = wrap; + state->wbits = (unsigned)windowBits; + return inflateReset(strm); +} + +int ZEXPORT inflateInit2_(strm, windowBits, version, stream_size) +z_streamp strm; +int windowBits; +const char *version; +int stream_size; +{ + int ret; + struct inflate_state FAR *state; + + if (version == Z_NULL || version[0] != ZLIB_VERSION[0] || + stream_size != (int)(sizeof(z_stream))) + return Z_VERSION_ERROR; + if (strm == Z_NULL) return Z_STREAM_ERROR; + strm->msg = Z_NULL; /* in case we return an error */ + if (strm->zalloc == (alloc_func)0) { +#ifdef Z_SOLO + return Z_STREAM_ERROR; +#else + strm->zalloc = zcalloc; + strm->opaque = (voidpf)0; +#endif + } + if (strm->zfree == (free_func)0) +#ifdef Z_SOLO + return Z_STREAM_ERROR; +#else + strm->zfree = zcfree; +#endif + state = (struct inflate_state FAR *) + ZALLOC(strm, 1, sizeof(struct inflate_state)); + if (state == Z_NULL) return Z_MEM_ERROR; + Tracev((stderr, "inflate: allocated\n")); + strm->state = (struct internal_state FAR *)state; + state->window = Z_NULL; + ret = inflateReset2(strm, windowBits); + if (ret != Z_OK) { + ZFREE(strm, state); + strm->state = Z_NULL; + } + return ret; +} + +int ZEXPORT inflateInit_(strm, version, stream_size) +z_streamp strm; +const char *version; +int stream_size; +{ + return inflateInit2_(strm, DEF_WBITS, version, stream_size); +} + +int ZEXPORT inflatePrime(strm, bits, value) +z_streamp strm; +int bits; +int value; +{ + struct inflate_state FAR *state; + + if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; + state = (struct inflate_state FAR *)strm->state; + if (bits < 0) { + state->hold = 0; + state->bits = 0; + return Z_OK; + } + if (bits > 16 || state->bits + bits > 32) return Z_STREAM_ERROR; + value &= (1L << bits) - 1; + state->hold += value << state->bits; + state->bits += bits; + return Z_OK; +} + +/* + Return state with length and distance decoding tables and index sizes set to + fixed code decoding. Normally this returns fixed tables from inffixed.h. + If BUILDFIXED is defined, then instead this routine builds the tables the + first time it's called, and returns those tables the first time and + thereafter. This reduces the size of the code by about 2K bytes, in + exchange for a little execution time. However, BUILDFIXED should not be + used for threaded applications, since the rewriting of the tables and virgin + may not be thread-safe. + */ +local void fixedtables(state) +struct inflate_state FAR *state; +{ +#ifdef BUILDFIXED + static int virgin = 1; + static code *lenfix, *distfix; + static code fixed[544]; + + /* build fixed huffman tables if first call (may not be thread safe) */ + if (virgin) { + unsigned sym, bits; + static code *next; + + /* literal/length table */ + sym = 0; + while (sym < 144) state->lens[sym++] = 8; + while (sym < 256) state->lens[sym++] = 9; + while (sym < 280) state->lens[sym++] = 7; + while (sym < 288) state->lens[sym++] = 8; + next = fixed; + lenfix = next; + bits = 9; + inflate_table(LENS, state->lens, 288, &(next), &(bits), state->work); + + /* distance table */ + sym = 0; + while (sym < 32) state->lens[sym++] = 5; + distfix = next; + bits = 5; + inflate_table(DISTS, state->lens, 32, &(next), &(bits), state->work); + + /* do this just once */ + virgin = 0; + } +#else /* !BUILDFIXED */ +# include "inffixed.h" +#endif /* BUILDFIXED */ + state->lencode = lenfix; + state->lenbits = 9; + state->distcode = distfix; + state->distbits = 5; +} + +#ifdef MAKEFIXED +#include + +/* + Write out the inffixed.h that is #include'd above. Defining MAKEFIXED also + defines BUILDFIXED, so the tables are built on the fly. makefixed() writes + those tables to stdout, which would be piped to inffixed.h. A small program + can simply call makefixed to do this: + + void makefixed(void); + + int main(void) + { + makefixed(); + return 0; + } + + Then that can be linked with zlib built with MAKEFIXED defined and run: + + a.out > inffixed.h + */ +void makefixed() +{ + unsigned low, size; + struct inflate_state state; + + fixedtables(&state); + puts(" /* inffixed.h -- table for decoding fixed codes"); + puts(" * Generated automatically by makefixed()."); + puts(" */"); + puts(""); + puts(" /* WARNING: this file should *not* be used by applications."); + puts(" It is part of the implementation of this library and is"); + puts(" subject to change. Applications should only use zlib.h."); + puts(" */"); + puts(""); + size = 1U << 9; + printf(" static const code lenfix[%u] = {", size); + low = 0; + for (;;) { + if ((low % 7) == 0) printf("\n "); + printf("{%u,%u,%d}", (low & 127) == 99 ? 64 : state.lencode[low].op, + state.lencode[low].bits, state.lencode[low].val); + if (++low == size) break; + putchar(','); + } + puts("\n };"); + size = 1U << 5; + printf("\n static const code distfix[%u] = {", size); + low = 0; + for (;;) { + if ((low % 6) == 0) printf("\n "); + printf("{%u,%u,%d}", state.distcode[low].op, state.distcode[low].bits, + state.distcode[low].val); + if (++low == size) break; + putchar(','); + } + puts("\n };"); +} +#endif /* MAKEFIXED */ + +/* + Update the window with the last wsize (normally 32K) bytes written before + returning. If window does not exist yet, create it. This is only called + when a window is already in use, or when output has been written during this + inflate call, but the end of the deflate stream has not been reached yet. + It is also called to create a window for dictionary data when a dictionary + is loaded. + + Providing output buffers larger than 32K to inflate() should provide a speed + advantage, since only the last 32K of output is copied to the sliding window + upon return from inflate(), and since all distances after the first 32K of + output will fall in the output data, making match copies simpler and faster. + The advantage may be dependent on the size of the processor's data caches. + */ +local int updatewindow(strm, end, copy) +z_streamp strm; +const Bytef *end; +unsigned copy; +{ + struct inflate_state FAR *state; + unsigned dist; + + state = (struct inflate_state FAR *)strm->state; + + /* if it hasn't been done already, allocate space for the window */ + if (state->window == Z_NULL) { + state->window = (unsigned char FAR *) + ZALLOC(strm, 1U << state->wbits, + sizeof(unsigned char)); + if (state->window == Z_NULL) return 1; + } + + /* if window not in use yet, initialize */ + if (state->wsize == 0) { + state->wsize = 1U << state->wbits; + state->wnext = 0; + state->whave = 0; + } + + /* copy state->wsize or less output bytes into the circular window */ + if (copy >= state->wsize) { + zmemcpy(state->window, end - state->wsize, state->wsize); + state->wnext = 0; + state->whave = state->wsize; + } + else { + dist = state->wsize - state->wnext; + if (dist > copy) dist = copy; + zmemcpy(state->window + state->wnext, end - copy, dist); + copy -= dist; + if (copy) { + zmemcpy(state->window, end - copy, copy); + state->wnext = copy; + state->whave = state->wsize; + } + else { + state->wnext += dist; + if (state->wnext == state->wsize) state->wnext = 0; + if (state->whave < state->wsize) state->whave += dist; + } + } + return 0; +} + +/* Macros for inflate(): */ + +/* check function to use adler32() for zlib or crc32() for gzip */ +#ifdef GUNZIP +# define UPDATE(check, buf, len) \ + (state->flags ? crc32(check, buf, len) : adler32(check, buf, len)) +#else +# define UPDATE(check, buf, len) adler32(check, buf, len) +#endif + +/* check macros for header crc */ +#ifdef GUNZIP +# define CRC2(check, word) \ + do { \ + hbuf[0] = (unsigned char)(word); \ + hbuf[1] = (unsigned char)((word) >> 8); \ + check = crc32(check, hbuf, 2); \ + } while (0) + +# define CRC4(check, word) \ + do { \ + hbuf[0] = (unsigned char)(word); \ + hbuf[1] = (unsigned char)((word) >> 8); \ + hbuf[2] = (unsigned char)((word) >> 16); \ + hbuf[3] = (unsigned char)((word) >> 24); \ + check = crc32(check, hbuf, 4); \ + } while (0) +#endif + +/* Load registers with state in inflate() for speed */ +#define LOAD() \ + do { \ + put = strm->next_out; \ + left = strm->avail_out; \ + next = strm->next_in; \ + have = strm->avail_in; \ + hold = state->hold; \ + bits = state->bits; \ + } while (0) + +/* Restore state from registers in inflate() */ +#define RESTORE() \ + do { \ + strm->next_out = put; \ + strm->avail_out = left; \ + strm->next_in = next; \ + strm->avail_in = have; \ + state->hold = hold; \ + state->bits = bits; \ + } while (0) + +/* Clear the input bit accumulator */ +#define INITBITS() \ + do { \ + hold = 0; \ + bits = 0; \ + } while (0) + +/* Get a byte of input into the bit accumulator, or return from inflate() + if there is no input available. */ +#define PULLBYTE() \ + do { \ + if (have == 0) goto inf_leave; \ + have--; \ + hold += (unsigned long)(*next++) << bits; \ + bits += 8; \ + } while (0) + +/* Assure that there are at least n bits in the bit accumulator. If there is + not enough available input to do that, then return from inflate(). */ +#define NEEDBITS(n) \ + do { \ + while (bits < (unsigned)(n)) \ + PULLBYTE(); \ + } while (0) + +/* Return the low n bits of the bit accumulator (n < 16) */ +#define BITS(n) \ + ((unsigned)hold & ((1U << (n)) - 1)) + +/* Remove n bits from the bit accumulator */ +#define DROPBITS(n) \ + do { \ + hold >>= (n); \ + bits -= (unsigned)(n); \ + } while (0) + +/* Remove zero to seven bits as needed to go to a byte boundary */ +#define BYTEBITS() \ + do { \ + hold >>= bits & 7; \ + bits -= bits & 7; \ + } while (0) + +/* + inflate() uses a state machine to process as much input data and generate as + much output data as possible before returning. The state machine is + structured roughly as follows: + + for (;;) switch (state) { + ... + case STATEn: + if (not enough input data or output space to make progress) + return; + ... make progress ... + state = STATEm; + break; + ... + } + + so when inflate() is called again, the same case is attempted again, and + if the appropriate resources are provided, the machine proceeds to the + next state. The NEEDBITS() macro is usually the way the state evaluates + whether it can proceed or should return. NEEDBITS() does the return if + the requested bits are not available. The typical use of the BITS macros + is: + + NEEDBITS(n); + ... do something with BITS(n) ... + DROPBITS(n); + + where NEEDBITS(n) either returns from inflate() if there isn't enough + input left to load n bits into the accumulator, or it continues. BITS(n) + gives the low n bits in the accumulator. When done, DROPBITS(n) drops + the low n bits off the accumulator. INITBITS() clears the accumulator + and sets the number of available bits to zero. BYTEBITS() discards just + enough bits to put the accumulator on a byte boundary. After BYTEBITS() + and a NEEDBITS(8), then BITS(8) would return the next byte in the stream. + + NEEDBITS(n) uses PULLBYTE() to get an available byte of input, or to return + if there is no input available. The decoding of variable length codes uses + PULLBYTE() directly in order to pull just enough bytes to decode the next + code, and no more. + + Some states loop until they get enough input, making sure that enough + state information is maintained to continue the loop where it left off + if NEEDBITS() returns in the loop. For example, want, need, and keep + would all have to actually be part of the saved state in case NEEDBITS() + returns: + + case STATEw: + while (want < need) { + NEEDBITS(n); + keep[want++] = BITS(n); + DROPBITS(n); + } + state = STATEx; + case STATEx: + + As shown above, if the next state is also the next case, then the break + is omitted. + + A state may also return if there is not enough output space available to + complete that state. Those states are copying stored data, writing a + literal byte, and copying a matching string. + + When returning, a "goto inf_leave" is used to update the total counters, + update the check value, and determine whether any progress has been made + during that inflate() call in order to return the proper return code. + Progress is defined as a change in either strm->avail_in or strm->avail_out. + When there is a window, goto inf_leave will update the window with the last + output written. If a goto inf_leave occurs in the middle of decompression + and there is no window currently, goto inf_leave will create one and copy + output to the window for the next call of inflate(). + + In this implementation, the flush parameter of inflate() only affects the + return code (per zlib.h). inflate() always writes as much as possible to + strm->next_out, given the space available and the provided input--the effect + documented in zlib.h of Z_SYNC_FLUSH. Furthermore, inflate() always defers + the allocation of and copying into a sliding window until necessary, which + provides the effect documented in zlib.h for Z_FINISH when the entire input + stream available. So the only thing the flush parameter actually does is: + when flush is set to Z_FINISH, inflate() cannot return Z_OK. Instead it + will return Z_BUF_ERROR if it has not reached the end of the stream. + */ + +int ZEXPORT inflate(strm, flush) +z_streamp strm; +int flush; +{ + struct inflate_state FAR *state; + z_const unsigned char FAR *next; /* next input */ + unsigned char FAR *put; /* next output */ + unsigned have, left; /* available input and output */ + unsigned long hold; /* bit buffer */ + unsigned bits; /* bits in bit buffer */ + unsigned in, out; /* save starting available input and output */ + unsigned copy; /* number of stored or match bytes to copy */ + unsigned char FAR *from; /* where to copy match bytes from */ + code here; /* current decoding table entry */ + code last; /* parent table entry */ + unsigned len; /* length to copy for repeats, bits to drop */ + int ret; /* return code */ +#ifdef GUNZIP + unsigned char hbuf[4]; /* buffer for gzip header crc calculation */ +#endif + static const unsigned short order[19] = /* permutation of code lengths */ + {16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15}; + + if (strm == Z_NULL || strm->state == Z_NULL || strm->next_out == Z_NULL || + (strm->next_in == Z_NULL && strm->avail_in != 0)) + return Z_STREAM_ERROR; + + state = (struct inflate_state FAR *)strm->state; + if (state->mode == TYPE) state->mode = TYPEDO; /* skip check */ + LOAD(); + in = have; + out = left; + ret = Z_OK; + for (;;) + switch (state->mode) { + case HEAD: + if (state->wrap == 0) { + state->mode = TYPEDO; + break; + } + NEEDBITS(16); +#ifdef GUNZIP + if ((state->wrap & 2) && hold == 0x8b1f) { /* gzip header */ + state->check = crc32(0L, Z_NULL, 0); + CRC2(state->check, hold); + INITBITS(); + state->mode = FLAGS; + break; + } + state->flags = 0; /* expect zlib header */ + if (state->head != Z_NULL) + state->head->done = -1; + if (!(state->wrap & 1) || /* check if zlib header allowed */ +#else + if ( +#endif + ((BITS(8) << 8) + (hold >> 8)) % 31) { + strm->msg = (char *)"incorrect header check"; + state->mode = BAD; + break; + } + if (BITS(4) != Z_DEFLATED) { + strm->msg = (char *)"unknown compression method"; + state->mode = BAD; + break; + } + DROPBITS(4); + len = BITS(4) + 8; + if (state->wbits == 0) + state->wbits = len; + else if (len > state->wbits) { + strm->msg = (char *)"invalid window size"; + state->mode = BAD; + break; + } + state->dmax = 1U << len; + Tracev((stderr, "inflate: zlib header ok\n")); + strm->adler = state->check = adler32(0L, Z_NULL, 0); + state->mode = hold & 0x200 ? DICTID : TYPE; + INITBITS(); + break; +#ifdef GUNZIP + case FLAGS: + NEEDBITS(16); + state->flags = (int)(hold); + if ((state->flags & 0xff) != Z_DEFLATED) { + strm->msg = (char *)"unknown compression method"; + state->mode = BAD; + break; + } + if (state->flags & 0xe000) { + strm->msg = (char *)"unknown header flags set"; + state->mode = BAD; + break; + } + if (state->head != Z_NULL) + state->head->text = (int)((hold >> 8) & 1); + if (state->flags & 0x0200) CRC2(state->check, hold); + INITBITS(); + state->mode = TIME; + case TIME: + NEEDBITS(32); + if (state->head != Z_NULL) + state->head->time = hold; + if (state->flags & 0x0200) CRC4(state->check, hold); + INITBITS(); + state->mode = OS; + case OS: + NEEDBITS(16); + if (state->head != Z_NULL) { + state->head->xflags = (int)(hold & 0xff); + state->head->os = (int)(hold >> 8); + } + if (state->flags & 0x0200) CRC2(state->check, hold); + INITBITS(); + state->mode = EXLEN; + case EXLEN: + if (state->flags & 0x0400) { + NEEDBITS(16); + state->length = (unsigned)(hold); + if (state->head != Z_NULL) + state->head->extra_len = (unsigned)hold; + if (state->flags & 0x0200) CRC2(state->check, hold); + INITBITS(); + } + else if (state->head != Z_NULL) + state->head->extra = Z_NULL; + state->mode = EXTRA; + case EXTRA: + if (state->flags & 0x0400) { + copy = state->length; + if (copy > have) copy = have; + if (copy) { + if (state->head != Z_NULL && + state->head->extra != Z_NULL) { + len = state->head->extra_len - state->length; + zmemcpy(state->head->extra + len, next, + len + copy > state->head->extra_max ? + state->head->extra_max - len : copy); + } + if (state->flags & 0x0200) + state->check = crc32(state->check, next, copy); + have -= copy; + next += copy; + state->length -= copy; + } + if (state->length) goto inf_leave; + } + state->length = 0; + state->mode = NAME; + case NAME: + if (state->flags & 0x0800) { + if (have == 0) goto inf_leave; + copy = 0; + do { + len = (unsigned)(next[copy++]); + if (state->head != Z_NULL && + state->head->name != Z_NULL && + state->length < state->head->name_max) + state->head->name[state->length++] = len; + } while (len && copy < have); + if (state->flags & 0x0200) + state->check = crc32(state->check, next, copy); + have -= copy; + next += copy; + if (len) goto inf_leave; + } + else if (state->head != Z_NULL) + state->head->name = Z_NULL; + state->length = 0; + state->mode = COMMENT; + case COMMENT: + if (state->flags & 0x1000) { + if (have == 0) goto inf_leave; + copy = 0; + do { + len = (unsigned)(next[copy++]); + if (state->head != Z_NULL && + state->head->comment != Z_NULL && + state->length < state->head->comm_max) + state->head->comment[state->length++] = len; + } while (len && copy < have); + if (state->flags & 0x0200) + state->check = crc32(state->check, next, copy); + have -= copy; + next += copy; + if (len) goto inf_leave; + } + else if (state->head != Z_NULL) + state->head->comment = Z_NULL; + state->mode = HCRC; + case HCRC: + if (state->flags & 0x0200) { + NEEDBITS(16); + if (hold != (state->check & 0xffff)) { + strm->msg = (char *)"header crc mismatch"; + state->mode = BAD; + break; + } + INITBITS(); + } + if (state->head != Z_NULL) { + state->head->hcrc = (int)((state->flags >> 9) & 1); + state->head->done = 1; + } + strm->adler = state->check = crc32(0L, Z_NULL, 0); + state->mode = TYPE; + break; +#endif + case DICTID: + NEEDBITS(32); + strm->adler = state->check = ZSWAP32(hold); + INITBITS(); + state->mode = DICT; + case DICT: + if (state->havedict == 0) { + RESTORE(); + return Z_NEED_DICT; + } + strm->adler = state->check = adler32(0L, Z_NULL, 0); + state->mode = TYPE; + case TYPE: + if (flush == Z_BLOCK || flush == Z_TREES) goto inf_leave; + case TYPEDO: + if (state->last) { + BYTEBITS(); + state->mode = CHECK; + break; + } + NEEDBITS(3); + state->last = BITS(1); + DROPBITS(1); + switch (BITS(2)) { + case 0: /* stored block */ + Tracev((stderr, "inflate: stored block%s\n", + state->last ? " (last)" : "")); + state->mode = STORED; + break; + case 1: /* fixed block */ + fixedtables(state); + Tracev((stderr, "inflate: fixed codes block%s\n", + state->last ? " (last)" : "")); + state->mode = LEN_; /* decode codes */ + if (flush == Z_TREES) { + DROPBITS(2); + goto inf_leave; + } + break; + case 2: /* dynamic block */ + Tracev((stderr, "inflate: dynamic codes block%s\n", + state->last ? " (last)" : "")); + state->mode = TABLE; + break; + case 3: + strm->msg = (char *)"invalid block type"; + state->mode = BAD; + } + DROPBITS(2); + break; + case STORED: + BYTEBITS(); /* go to byte boundary */ + NEEDBITS(32); + if ((hold & 0xffff) != ((hold >> 16) ^ 0xffff)) { + strm->msg = (char *)"invalid stored block lengths"; + state->mode = BAD; + break; + } + state->length = (unsigned)hold & 0xffff; + Tracev((stderr, "inflate: stored length %u\n", + state->length)); + INITBITS(); + state->mode = COPY_; + if (flush == Z_TREES) goto inf_leave; + case COPY_: + state->mode = COPY; + case COPY: + copy = state->length; + if (copy) { + if (copy > have) copy = have; + if (copy > left) copy = left; + if (copy == 0) goto inf_leave; + zmemcpy(put, next, copy); + have -= copy; + next += copy; + left -= copy; + put += copy; + state->length -= copy; + break; + } + Tracev((stderr, "inflate: stored end\n")); + state->mode = TYPE; + break; + case TABLE: + NEEDBITS(14); + state->nlen = BITS(5) + 257; + DROPBITS(5); + state->ndist = BITS(5) + 1; + DROPBITS(5); + state->ncode = BITS(4) + 4; + DROPBITS(4); +#ifndef PKZIP_BUG_WORKAROUND + if (state->nlen > 286 || state->ndist > 30) { + strm->msg = (char *)"too many length or distance symbols"; + state->mode = BAD; + break; + } +#endif + Tracev((stderr, "inflate: table sizes ok\n")); + state->have = 0; + state->mode = LENLENS; + case LENLENS: + while (state->have < state->ncode) { + NEEDBITS(3); + state->lens[order[state->have++]] = (unsigned short)BITS(3); + DROPBITS(3); + } + while (state->have < 19) + state->lens[order[state->have++]] = 0; + state->next = state->codes; + state->lencode = (const code FAR *)(state->next); + state->lenbits = 7; + ret = inflate_table(CODES, state->lens, 19, &(state->next), + &(state->lenbits), state->work); + if (ret) { + strm->msg = (char *)"invalid code lengths set"; + state->mode = BAD; + break; + } + Tracev((stderr, "inflate: code lengths ok\n")); + state->have = 0; + state->mode = CODELENS; + case CODELENS: + while (state->have < state->nlen + state->ndist) { + for (;;) { + here = state->lencode[BITS(state->lenbits)]; + if ((unsigned)(here.bits) <= bits) break; + PULLBYTE(); + } + if (here.val < 16) { + DROPBITS(here.bits); + state->lens[state->have++] = here.val; + } + else { + if (here.val == 16) { + NEEDBITS(here.bits + 2); + DROPBITS(here.bits); + if (state->have == 0) { + strm->msg = (char *)"invalid bit length repeat"; + state->mode = BAD; + break; + } + len = state->lens[state->have - 1]; + copy = 3 + BITS(2); + DROPBITS(2); + } + else if (here.val == 17) { + NEEDBITS(here.bits + 3); + DROPBITS(here.bits); + len = 0; + copy = 3 + BITS(3); + DROPBITS(3); + } + else { + NEEDBITS(here.bits + 7); + DROPBITS(here.bits); + len = 0; + copy = 11 + BITS(7); + DROPBITS(7); + } + if (state->have + copy > state->nlen + state->ndist) { + strm->msg = (char *)"invalid bit length repeat"; + state->mode = BAD; + break; + } + while (copy--) + state->lens[state->have++] = (unsigned short)len; + } + } + + /* handle error breaks in while */ + if (state->mode == BAD) break; + + /* check for end-of-block code (better have one) */ + if (state->lens[256] == 0) { + strm->msg = (char *)"invalid code -- missing end-of-block"; + state->mode = BAD; + break; + } + + /* build code tables -- note: do not change the lenbits or distbits + values here (9 and 6) without reading the comments in inftrees.h + concerning the ENOUGH constants, which depend on those values */ + state->next = state->codes; + state->lencode = (const code FAR *)(state->next); + state->lenbits = 9; + ret = inflate_table(LENS, state->lens, state->nlen, &(state->next), + &(state->lenbits), state->work); + if (ret) { + strm->msg = (char *)"invalid literal/lengths set"; + state->mode = BAD; + break; + } + state->distcode = (const code FAR *)(state->next); + state->distbits = 6; + ret = inflate_table(DISTS, state->lens + state->nlen, state->ndist, + &(state->next), &(state->distbits), state->work); + if (ret) { + strm->msg = (char *)"invalid distances set"; + state->mode = BAD; + break; + } + Tracev((stderr, "inflate: codes ok\n")); + state->mode = LEN_; + if (flush == Z_TREES) goto inf_leave; + case LEN_: + state->mode = LEN; + case LEN: + if (have >= 6 && left >= 258) { + RESTORE(); + inflate_fast(strm, out); + LOAD(); + if (state->mode == TYPE) + state->back = -1; + break; + } + state->back = 0; + for (;;) { + here = state->lencode[BITS(state->lenbits)]; + if ((unsigned)(here.bits) <= bits) break; + PULLBYTE(); + } + if (here.op && (here.op & 0xf0) == 0) { + last = here; + for (;;) { + here = state->lencode[last.val + + (BITS(last.bits + last.op) >> last.bits)]; + if ((unsigned)(last.bits + here.bits) <= bits) break; + PULLBYTE(); + } + DROPBITS(last.bits); + state->back += last.bits; + } + DROPBITS(here.bits); + state->back += here.bits; + state->length = (unsigned)here.val; + if ((int)(here.op) == 0) { + Tracevv((stderr, here.val >= 0x20 && here.val < 0x7f ? + "inflate: literal '%c'\n" : + "inflate: literal 0x%02x\n", here.val)); + state->mode = LIT; + break; + } + if (here.op & 32) { + Tracevv((stderr, "inflate: end of block\n")); + state->back = -1; + state->mode = TYPE; + break; + } + if (here.op & 64) { + strm->msg = (char *)"invalid literal/length code"; + state->mode = BAD; + break; + } + state->extra = (unsigned)(here.op) & 15; + state->mode = LENEXT; + case LENEXT: + if (state->extra) { + NEEDBITS(state->extra); + state->length += BITS(state->extra); + DROPBITS(state->extra); + state->back += state->extra; + } + Tracevv((stderr, "inflate: length %u\n", state->length)); + state->was = state->length; + state->mode = DIST; + case DIST: + for (;;) { + here = state->distcode[BITS(state->distbits)]; + if ((unsigned)(here.bits) <= bits) break; + PULLBYTE(); + } + if ((here.op & 0xf0) == 0) { + last = here; + for (;;) { + here = state->distcode[last.val + + (BITS(last.bits + last.op) >> last.bits)]; + if ((unsigned)(last.bits + here.bits) <= bits) break; + PULLBYTE(); + } + DROPBITS(last.bits); + state->back += last.bits; + } + DROPBITS(here.bits); + state->back += here.bits; + if (here.op & 64) { + strm->msg = (char *)"invalid distance code"; + state->mode = BAD; + break; + } + state->offset = (unsigned)here.val; + state->extra = (unsigned)(here.op) & 15; + state->mode = DISTEXT; + case DISTEXT: + if (state->extra) { + NEEDBITS(state->extra); + state->offset += BITS(state->extra); + DROPBITS(state->extra); + state->back += state->extra; + } +#ifdef INFLATE_STRICT + if (state->offset > state->dmax) { + strm->msg = (char *)"invalid distance too far back"; + state->mode = BAD; + break; + } +#endif + Tracevv((stderr, "inflate: distance %u\n", state->offset)); + state->mode = MATCH; + case MATCH: + if (left == 0) goto inf_leave; + copy = out - left; + if (state->offset > copy) { /* copy from window */ + copy = state->offset - copy; + if (copy > state->whave) { + if (state->sane) { + strm->msg = (char *)"invalid distance too far back"; + state->mode = BAD; + break; + } +#ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR + Trace((stderr, "inflate.c too far\n")); + copy -= state->whave; + if (copy > state->length) copy = state->length; + if (copy > left) copy = left; + left -= copy; + state->length -= copy; + do { + *put++ = 0; + } while (--copy); + if (state->length == 0) state->mode = LEN; + break; +#endif + } + if (copy > state->wnext) { + copy -= state->wnext; + from = state->window + (state->wsize - copy); + } + else + from = state->window + (state->wnext - copy); + if (copy > state->length) copy = state->length; + } + else { /* copy from output */ + from = put - state->offset; + copy = state->length; + } + if (copy > left) copy = left; + left -= copy; + state->length -= copy; + do { + *put++ = *from++; + } while (--copy); + if (state->length == 0) state->mode = LEN; + break; + case LIT: + if (left == 0) goto inf_leave; + *put++ = (unsigned char)(state->length); + left--; + state->mode = LEN; + break; + case CHECK: + if (state->wrap) { + NEEDBITS(32); + out -= left; + strm->total_out += out; + state->total += out; + if (out) + strm->adler = state->check = + UPDATE(state->check, put - out, out); + out = left; + if (( +#ifdef GUNZIP + state->flags ? hold : +#endif + ZSWAP32(hold)) != state->check) { + strm->msg = (char *)"incorrect data check"; + state->mode = BAD; + break; + } + INITBITS(); + Tracev((stderr, "inflate: check matches trailer\n")); + } +#ifdef GUNZIP + state->mode = LENGTH; + case LENGTH: + if (state->wrap && state->flags) { + NEEDBITS(32); + if (hold != (state->total & 0xffffffffUL)) { + strm->msg = (char *)"incorrect length check"; + state->mode = BAD; + break; + } + INITBITS(); + Tracev((stderr, "inflate: length matches trailer\n")); + } +#endif + state->mode = DONE; + case DONE: + ret = Z_STREAM_END; + goto inf_leave; + case BAD: + ret = Z_DATA_ERROR; + goto inf_leave; + case MEM: + return Z_MEM_ERROR; + case SYNC: + default: + return Z_STREAM_ERROR; + } + + /* + Return from inflate(), updating the total counts and the check value. + If there was no progress during the inflate() call, return a buffer + error. Call updatewindow() to create and/or update the window state. + Note: a memory error from inflate() is non-recoverable. + */ + inf_leave: + RESTORE(); + if (state->wsize || (out != strm->avail_out && state->mode < BAD && + (state->mode < CHECK || flush != Z_FINISH))) + if (updatewindow(strm, strm->next_out, out - strm->avail_out)) { + state->mode = MEM; + return Z_MEM_ERROR; + } + in -= strm->avail_in; + out -= strm->avail_out; + strm->total_in += in; + strm->total_out += out; + state->total += out; + if (state->wrap && out) + strm->adler = state->check = + UPDATE(state->check, strm->next_out - out, out); + strm->data_type = state->bits + (state->last ? 64 : 0) + + (state->mode == TYPE ? 128 : 0) + + (state->mode == LEN_ || state->mode == COPY_ ? 256 : 0); + if (((in == 0 && out == 0) || flush == Z_FINISH) && ret == Z_OK) + ret = Z_BUF_ERROR; + return ret; +} + +int ZEXPORT inflateEnd(strm) +z_streamp strm; +{ + struct inflate_state FAR *state; + if (strm == Z_NULL || strm->state == Z_NULL || strm->zfree == (free_func)0) + return Z_STREAM_ERROR; + state = (struct inflate_state FAR *)strm->state; + if (state->window != Z_NULL) ZFREE(strm, state->window); + ZFREE(strm, strm->state); + strm->state = Z_NULL; + Tracev((stderr, "inflate: end\n")); + return Z_OK; +} + +int ZEXPORT inflateGetDictionary(strm, dictionary, dictLength) +z_streamp strm; +Bytef *dictionary; +uInt *dictLength; +{ + struct inflate_state FAR *state; + + /* check state */ + if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; + state = (struct inflate_state FAR *)strm->state; + + /* copy dictionary */ + if (state->whave && dictionary != Z_NULL) { + zmemcpy(dictionary, state->window + state->wnext, + state->whave - state->wnext); + zmemcpy(dictionary + state->whave - state->wnext, + state->window, state->wnext); + } + if (dictLength != Z_NULL) + *dictLength = state->whave; + return Z_OK; +} + +int ZEXPORT inflateSetDictionary(strm, dictionary, dictLength) +z_streamp strm; +const Bytef *dictionary; +uInt dictLength; +{ + struct inflate_state FAR *state; + unsigned long dictid; + int ret; + + /* check state */ + if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; + state = (struct inflate_state FAR *)strm->state; + if (state->wrap != 0 && state->mode != DICT) + return Z_STREAM_ERROR; + + /* check for correct dictionary identifier */ + if (state->mode == DICT) { + dictid = adler32(0L, Z_NULL, 0); + dictid = adler32(dictid, dictionary, dictLength); + if (dictid != state->check) + return Z_DATA_ERROR; + } + + /* copy dictionary to window using updatewindow(), which will amend the + existing dictionary if appropriate */ + ret = updatewindow(strm, dictionary + dictLength, dictLength); + if (ret) { + state->mode = MEM; + return Z_MEM_ERROR; + } + state->havedict = 1; + Tracev((stderr, "inflate: dictionary set\n")); + return Z_OK; +} + +int ZEXPORT inflateGetHeader(strm, head) +z_streamp strm; +gz_headerp head; +{ + struct inflate_state FAR *state; + + /* check state */ + if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; + state = (struct inflate_state FAR *)strm->state; + if ((state->wrap & 2) == 0) return Z_STREAM_ERROR; + + /* save header structure */ + state->head = head; + head->done = 0; + return Z_OK; +} + +/* + Search buf[0..len-1] for the pattern: 0, 0, 0xff, 0xff. Return when found + or when out of input. When called, *have is the number of pattern bytes + found in order so far, in 0..3. On return *have is updated to the new + state. If on return *have equals four, then the pattern was found and the + return value is how many bytes were read including the last byte of the + pattern. If *have is less than four, then the pattern has not been found + yet and the return value is len. In the latter case, syncsearch() can be + called again with more data and the *have state. *have is initialized to + zero for the first call. + */ +local unsigned syncsearch(have, buf, len) +unsigned FAR *have; +const unsigned char FAR *buf; +unsigned len; +{ + unsigned got; + unsigned next; + + got = *have; + next = 0; + while (next < len && got < 4) { + if ((int)(buf[next]) == (got < 2 ? 0 : 0xff)) + got++; + else if (buf[next]) + got = 0; + else + got = 4 - got; + next++; + } + *have = got; + return next; +} + +int ZEXPORT inflateSync(strm) +z_streamp strm; +{ + unsigned len; /* number of bytes to look at or looked at */ + unsigned long in, out; /* temporary to save total_in and total_out */ + unsigned char buf[4]; /* to restore bit buffer to byte string */ + struct inflate_state FAR *state; + + /* check parameters */ + if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; + state = (struct inflate_state FAR *)strm->state; + if (strm->avail_in == 0 && state->bits < 8) return Z_BUF_ERROR; + + /* if first time, start search in bit buffer */ + if (state->mode != SYNC) { + state->mode = SYNC; + state->hold <<= state->bits & 7; + state->bits -= state->bits & 7; + len = 0; + while (state->bits >= 8) { + buf[len++] = (unsigned char)(state->hold); + state->hold >>= 8; + state->bits -= 8; + } + state->have = 0; + syncsearch(&(state->have), buf, len); + } + + /* search available input */ + len = syncsearch(&(state->have), strm->next_in, strm->avail_in); + strm->avail_in -= len; + strm->next_in += len; + strm->total_in += len; + + /* return no joy or set up to restart inflate() on a new block */ + if (state->have != 4) return Z_DATA_ERROR; + in = strm->total_in; out = strm->total_out; + inflateReset(strm); + strm->total_in = in; strm->total_out = out; + state->mode = TYPE; + return Z_OK; +} + +/* + Returns true if inflate is currently at the end of a block generated by + Z_SYNC_FLUSH or Z_FULL_FLUSH. This function is used by one PPP + implementation to provide an additional safety check. PPP uses + Z_SYNC_FLUSH but removes the length bytes of the resulting empty stored + block. When decompressing, PPP checks that at the end of input packet, + inflate is waiting for these length bytes. + */ +int ZEXPORT inflateSyncPoint(strm) +z_streamp strm; +{ + struct inflate_state FAR *state; + + if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; + state = (struct inflate_state FAR *)strm->state; + return state->mode == STORED && state->bits == 0; +} + +int ZEXPORT inflateCopy(dest, source) +z_streamp dest; +z_streamp source; +{ + struct inflate_state FAR *state; + struct inflate_state FAR *copy; + unsigned char FAR *window; + unsigned wsize; + + /* check input */ + if (dest == Z_NULL || source == Z_NULL || source->state == Z_NULL || + source->zalloc == (alloc_func)0 || source->zfree == (free_func)0) + return Z_STREAM_ERROR; + state = (struct inflate_state FAR *)source->state; + + /* allocate space */ + copy = (struct inflate_state FAR *) + ZALLOC(source, 1, sizeof(struct inflate_state)); + if (copy == Z_NULL) return Z_MEM_ERROR; + window = Z_NULL; + if (state->window != Z_NULL) { + window = (unsigned char FAR *) + ZALLOC(source, 1U << state->wbits, sizeof(unsigned char)); + if (window == Z_NULL) { + ZFREE(source, copy); + return Z_MEM_ERROR; + } + } + + /* copy state */ + zmemcpy((voidpf)dest, (voidpf)source, sizeof(z_stream)); + zmemcpy((voidpf)copy, (voidpf)state, sizeof(struct inflate_state)); + if (state->lencode >= state->codes && + state->lencode <= state->codes + ENOUGH - 1) { + copy->lencode = copy->codes + (state->lencode - state->codes); + copy->distcode = copy->codes + (state->distcode - state->codes); + } + copy->next = copy->codes + (state->next - state->codes); + if (window != Z_NULL) { + wsize = 1U << state->wbits; + zmemcpy(window, state->window, wsize); + } + copy->window = window; + dest->state = (struct internal_state FAR *)copy; + return Z_OK; +} + +int ZEXPORT inflateUndermine(strm, subvert) +z_streamp strm; +int subvert; +{ + struct inflate_state FAR *state; + + if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; + state = (struct inflate_state FAR *)strm->state; + state->sane = !subvert; +#ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR + return Z_OK; +#else + state->sane = 1; + return Z_DATA_ERROR; +#endif +} + +long ZEXPORT inflateMark(strm) +z_streamp strm; +{ + struct inflate_state FAR *state; + + if (strm == Z_NULL || strm->state == Z_NULL) return -1L << 16; + state = (struct inflate_state FAR *)strm->state; + return ((long)(state->back) << 16) + + (state->mode == COPY ? state->length : + (state->mode == MATCH ? state->was - state->length : 0)); +} diff --git a/app/src/main/jni/zlib/inflate.h b/app/src/main/jni/zlib/inflate.h new file mode 100644 index 000000000..95f4986d4 --- /dev/null +++ b/app/src/main/jni/zlib/inflate.h @@ -0,0 +1,122 @@ +/* inflate.h -- internal inflate state definition + * Copyright (C) 1995-2009 Mark Adler + * For conditions of distribution and use, see copyright notice in zlib.h + */ + +/* WARNING: this file should *not* be used by applications. It is + part of the implementation of the compression library and is + subject to change. Applications should only use zlib.h. + */ + +/* define NO_GZIP when compiling if you want to disable gzip header and + trailer decoding by inflate(). NO_GZIP would be used to avoid linking in + the crc code when it is not needed. For shared libraries, gzip decoding + should be left enabled. */ +#ifndef NO_GZIP +# define GUNZIP +#endif + +/* Possible inflate modes between inflate() calls */ +typedef enum { + HEAD, /* i: waiting for magic header */ + FLAGS, /* i: waiting for method and flags (gzip) */ + TIME, /* i: waiting for modification time (gzip) */ + OS, /* i: waiting for extra flags and operating system (gzip) */ + EXLEN, /* i: waiting for extra length (gzip) */ + EXTRA, /* i: waiting for extra bytes (gzip) */ + NAME, /* i: waiting for end of file name (gzip) */ + COMMENT, /* i: waiting for end of comment (gzip) */ + HCRC, /* i: waiting for header crc (gzip) */ + DICTID, /* i: waiting for dictionary check value */ + DICT, /* waiting for inflateSetDictionary() call */ + TYPE, /* i: waiting for type bits, including last-flag bit */ + TYPEDO, /* i: same, but skip check to exit inflate on new block */ + STORED, /* i: waiting for stored size (length and complement) */ + COPY_, /* i/o: same as COPY below, but only first time in */ + COPY, /* i/o: waiting for input or output to copy stored block */ + TABLE, /* i: waiting for dynamic block table lengths */ + LENLENS, /* i: waiting for code length code lengths */ + CODELENS, /* i: waiting for length/lit and distance code lengths */ + LEN_, /* i: same as LEN below, but only first time in */ + LEN, /* i: waiting for length/lit/eob code */ + LENEXT, /* i: waiting for length extra bits */ + DIST, /* i: waiting for distance code */ + DISTEXT, /* i: waiting for distance extra bits */ + MATCH, /* o: waiting for output space to copy string */ + LIT, /* o: waiting for output space to write literal */ + CHECK, /* i: waiting for 32-bit check value */ + LENGTH, /* i: waiting for 32-bit length (gzip) */ + DONE, /* finished check, done -- remain here until reset */ + BAD, /* got a data error -- remain here until reset */ + MEM, /* got an inflate() memory error -- remain here until reset */ + SYNC /* looking for synchronization bytes to restart inflate() */ +} inflate_mode; + +/* + State transitions between above modes - + + (most modes can go to BAD or MEM on error -- not shown for clarity) + + Process header: + HEAD -> (gzip) or (zlib) or (raw) + (gzip) -> FLAGS -> TIME -> OS -> EXLEN -> EXTRA -> NAME -> COMMENT -> + HCRC -> TYPE + (zlib) -> DICTID or TYPE + DICTID -> DICT -> TYPE + (raw) -> TYPEDO + Read deflate blocks: + TYPE -> TYPEDO -> STORED or TABLE or LEN_ or CHECK + STORED -> COPY_ -> COPY -> TYPE + TABLE -> LENLENS -> CODELENS -> LEN_ + LEN_ -> LEN + Read deflate codes in fixed or dynamic block: + LEN -> LENEXT or LIT or TYPE + LENEXT -> DIST -> DISTEXT -> MATCH -> LEN + LIT -> LEN + Process trailer: + CHECK -> LENGTH -> DONE + */ + +/* state maintained between inflate() calls. Approximately 10K bytes. */ +struct inflate_state { + inflate_mode mode; /* current inflate mode */ + int last; /* true if processing last block */ + int wrap; /* bit 0 true for zlib, bit 1 true for gzip */ + int havedict; /* true if dictionary provided */ + int flags; /* gzip header method and flags (0 if zlib) */ + unsigned dmax; /* zlib header max distance (INFLATE_STRICT) */ + unsigned long check; /* protected copy of check value */ + unsigned long total; /* protected copy of output count */ + gz_headerp head; /* where to save gzip header information */ + /* sliding window */ + unsigned wbits; /* log base 2 of requested window size */ + unsigned wsize; /* window size or zero if not using window */ + unsigned whave; /* valid bytes in the window */ + unsigned wnext; /* window write index */ + unsigned char FAR *window; /* allocated sliding window, if needed */ + /* bit accumulator */ + unsigned long hold; /* input bit accumulator */ + unsigned bits; /* number of bits in "in" */ + /* for string and stored block copying */ + unsigned length; /* literal or length of data to copy */ + unsigned offset; /* distance back to copy string from */ + /* for table and code decoding */ + unsigned extra; /* extra bits needed */ + /* fixed and dynamic code tables */ + code const FAR *lencode; /* starting table for length/literal codes */ + code const FAR *distcode; /* starting table for distance codes */ + unsigned lenbits; /* index bits for lencode */ + unsigned distbits; /* index bits for distcode */ + /* dynamic table building */ + unsigned ncode; /* number of code length code lengths */ + unsigned nlen; /* number of length code lengths */ + unsigned ndist; /* number of distance code lengths */ + unsigned have; /* number of code lengths in lens[] */ + code FAR *next; /* next available space in codes[] */ + unsigned short lens[320]; /* temporary storage for code lengths */ + unsigned short work[288]; /* work area for code table building */ + code codes[ENOUGH]; /* space for code tables */ + int sane; /* if false, allow invalid distance too far */ + int back; /* bits back of last unprocessed length/lit */ + unsigned was; /* initial length of match */ +}; diff --git a/app/src/main/jni/zlib/inftrees.c b/app/src/main/jni/zlib/inftrees.c new file mode 100644 index 000000000..44d89cf24 --- /dev/null +++ b/app/src/main/jni/zlib/inftrees.c @@ -0,0 +1,306 @@ +/* inftrees.c -- generate Huffman trees for efficient decoding + * Copyright (C) 1995-2013 Mark Adler + * For conditions of distribution and use, see copyright notice in zlib.h + */ + +#include "zutil.h" +#include "inftrees.h" + +#define MAXBITS 15 + +const char inflate_copyright[] = + " inflate 1.2.8 Copyright 1995-2013 Mark Adler "; +/* + If you use the zlib library in a product, an acknowledgment is welcome + in the documentation of your product. If for some reason you cannot + include such an acknowledgment, I would appreciate that you keep this + copyright string in the executable of your product. + */ + +/* + Build a set of tables to decode the provided canonical Huffman code. + The code lengths are lens[0..codes-1]. The result starts at *table, + whose indices are 0..2^bits-1. work is a writable array of at least + lens shorts, which is used as a work area. type is the type of code + to be generated, CODES, LENS, or DISTS. On return, zero is success, + -1 is an invalid code, and +1 means that ENOUGH isn't enough. table + on return points to the next available entry's address. bits is the + requested root table index bits, and on return it is the actual root + table index bits. It will differ if the request is greater than the + longest code or if it is less than the shortest code. + */ +int ZLIB_INTERNAL inflate_table(type, lens, codes, table, bits, work) +codetype type; +unsigned short FAR *lens; +unsigned codes; +code FAR * FAR *table; +unsigned FAR *bits; +unsigned short FAR *work; +{ + unsigned len; /* a code's length in bits */ + unsigned sym; /* index of code symbols */ + unsigned min, max; /* minimum and maximum code lengths */ + unsigned root; /* number of index bits for root table */ + unsigned curr; /* number of index bits for current table */ + unsigned drop; /* code bits to drop for sub-table */ + int left; /* number of prefix codes available */ + unsigned used; /* code entries in table used */ + unsigned huff; /* Huffman code */ + unsigned incr; /* for incrementing code, index */ + unsigned fill; /* index for replicating entries */ + unsigned low; /* low bits for current root entry */ + unsigned mask; /* mask for low root bits */ + code here; /* table entry for duplication */ + code FAR *next; /* next available space in table */ + const unsigned short FAR *base; /* base value table to use */ + const unsigned short FAR *extra; /* extra bits table to use */ + int end; /* use base and extra for symbol > end */ + unsigned short count[MAXBITS+1]; /* number of codes of each length */ + unsigned short offs[MAXBITS+1]; /* offsets in table for each length */ + static const unsigned short lbase[31] = { /* Length codes 257..285 base */ + 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31, + 35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0}; + static const unsigned short lext[31] = { /* Length codes 257..285 extra */ + 16, 16, 16, 16, 16, 16, 16, 16, 17, 17, 17, 17, 18, 18, 18, 18, + 19, 19, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 16, 72, 78}; + static const unsigned short dbase[32] = { /* Distance codes 0..29 base */ + 1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, + 257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145, + 8193, 12289, 16385, 24577, 0, 0}; + static const unsigned short dext[32] = { /* Distance codes 0..29 extra */ + 16, 16, 16, 16, 17, 17, 18, 18, 19, 19, 20, 20, 21, 21, 22, 22, + 23, 23, 24, 24, 25, 25, 26, 26, 27, 27, + 28, 28, 29, 29, 64, 64}; + + /* + Process a set of code lengths to create a canonical Huffman code. The + code lengths are lens[0..codes-1]. Each length corresponds to the + symbols 0..codes-1. The Huffman code is generated by first sorting the + symbols by length from short to long, and retaining the symbol order + for codes with equal lengths. Then the code starts with all zero bits + for the first code of the shortest length, and the codes are integer + increments for the same length, and zeros are appended as the length + increases. For the deflate format, these bits are stored backwards + from their more natural integer increment ordering, and so when the + decoding tables are built in the large loop below, the integer codes + are incremented backwards. + + This routine assumes, but does not check, that all of the entries in + lens[] are in the range 0..MAXBITS. The caller must assure this. + 1..MAXBITS is interpreted as that code length. zero means that that + symbol does not occur in this code. + + The codes are sorted by computing a count of codes for each length, + creating from that a table of starting indices for each length in the + sorted table, and then entering the symbols in order in the sorted + table. The sorted table is work[], with that space being provided by + the caller. + + The length counts are used for other purposes as well, i.e. finding + the minimum and maximum length codes, determining if there are any + codes at all, checking for a valid set of lengths, and looking ahead + at length counts to determine sub-table sizes when building the + decoding tables. + */ + + /* accumulate lengths for codes (assumes lens[] all in 0..MAXBITS) */ + for (len = 0; len <= MAXBITS; len++) + count[len] = 0; + for (sym = 0; sym < codes; sym++) + count[lens[sym]]++; + + /* bound code lengths, force root to be within code lengths */ + root = *bits; + for (max = MAXBITS; max >= 1; max--) + if (count[max] != 0) break; + if (root > max) root = max; + if (max == 0) { /* no symbols to code at all */ + here.op = (unsigned char)64; /* invalid code marker */ + here.bits = (unsigned char)1; + here.val = (unsigned short)0; + *(*table)++ = here; /* make a table to force an error */ + *(*table)++ = here; + *bits = 1; + return 0; /* no symbols, but wait for decoding to report error */ + } + for (min = 1; min < max; min++) + if (count[min] != 0) break; + if (root < min) root = min; + + /* check for an over-subscribed or incomplete set of lengths */ + left = 1; + for (len = 1; len <= MAXBITS; len++) { + left <<= 1; + left -= count[len]; + if (left < 0) return -1; /* over-subscribed */ + } + if (left > 0 && (type == CODES || max != 1)) + return -1; /* incomplete set */ + + /* generate offsets into symbol table for each length for sorting */ + offs[1] = 0; + for (len = 1; len < MAXBITS; len++) + offs[len + 1] = offs[len] + count[len]; + + /* sort symbols by length, by symbol order within each length */ + for (sym = 0; sym < codes; sym++) + if (lens[sym] != 0) work[offs[lens[sym]]++] = (unsigned short)sym; + + /* + Create and fill in decoding tables. In this loop, the table being + filled is at next and has curr index bits. The code being used is huff + with length len. That code is converted to an index by dropping drop + bits off of the bottom. For codes where len is less than drop + curr, + those top drop + curr - len bits are incremented through all values to + fill the table with replicated entries. + + root is the number of index bits for the root table. When len exceeds + root, sub-tables are created pointed to by the root entry with an index + of the low root bits of huff. This is saved in low to check for when a + new sub-table should be started. drop is zero when the root table is + being filled, and drop is root when sub-tables are being filled. + + When a new sub-table is needed, it is necessary to look ahead in the + code lengths to determine what size sub-table is needed. The length + counts are used for this, and so count[] is decremented as codes are + entered in the tables. + + used keeps track of how many table entries have been allocated from the + provided *table space. It is checked for LENS and DIST tables against + the constants ENOUGH_LENS and ENOUGH_DISTS to guard against changes in + the initial root table size constants. See the comments in inftrees.h + for more information. + + sym increments through all symbols, and the loop terminates when + all codes of length max, i.e. all codes, have been processed. This + routine permits incomplete codes, so another loop after this one fills + in the rest of the decoding tables with invalid code markers. + */ + + /* set up for code type */ + switch (type) { + case CODES: + base = extra = work; /* dummy value--not used */ + end = 19; + break; + case LENS: + base = lbase; + base -= 257; + extra = lext; + extra -= 257; + end = 256; + break; + default: /* DISTS */ + base = dbase; + extra = dext; + end = -1; + } + + /* initialize state for loop */ + huff = 0; /* starting code */ + sym = 0; /* starting code symbol */ + len = min; /* starting code length */ + next = *table; /* current table to fill in */ + curr = root; /* current table index bits */ + drop = 0; /* current bits to drop from code for index */ + low = (unsigned)(-1); /* trigger new sub-table when len > root */ + used = 1U << root; /* use root table entries */ + mask = used - 1; /* mask for comparing low */ + + /* check available table space */ + if ((type == LENS && used > ENOUGH_LENS) || + (type == DISTS && used > ENOUGH_DISTS)) + return 1; + + /* process all codes and make table entries */ + for (;;) { + /* create table entry */ + here.bits = (unsigned char)(len - drop); + if ((int)(work[sym]) < end) { + here.op = (unsigned char)0; + here.val = work[sym]; + } + else if ((int)(work[sym]) > end) { + here.op = (unsigned char)(extra[work[sym]]); + here.val = base[work[sym]]; + } + else { + here.op = (unsigned char)(32 + 64); /* end of block */ + here.val = 0; + } + + /* replicate for those indices with low len bits equal to huff */ + incr = 1U << (len - drop); + fill = 1U << curr; + min = fill; /* save offset to next table */ + do { + fill -= incr; + next[(huff >> drop) + fill] = here; + } while (fill != 0); + + /* backwards increment the len-bit code huff */ + incr = 1U << (len - 1); + while (huff & incr) + incr >>= 1; + if (incr != 0) { + huff &= incr - 1; + huff += incr; + } + else + huff = 0; + + /* go to next symbol, update count, len */ + sym++; + if (--(count[len]) == 0) { + if (len == max) break; + len = lens[work[sym]]; + } + + /* create new sub-table if needed */ + if (len > root && (huff & mask) != low) { + /* if first time, transition to sub-tables */ + if (drop == 0) + drop = root; + + /* increment past last table */ + next += min; /* here min is 1 << curr */ + + /* determine length of next table */ + curr = len - drop; + left = (int)(1 << curr); + while (curr + drop < max) { + left -= count[curr + drop]; + if (left <= 0) break; + curr++; + left <<= 1; + } + + /* check for enough space */ + used += 1U << curr; + if ((type == LENS && used > ENOUGH_LENS) || + (type == DISTS && used > ENOUGH_DISTS)) + return 1; + + /* point entry in root table to sub-table */ + low = huff & mask; + (*table)[low].op = (unsigned char)curr; + (*table)[low].bits = (unsigned char)root; + (*table)[low].val = (unsigned short)(next - *table); + } + } + + /* fill in remaining table entry if code is incomplete (guaranteed to have + at most one remaining entry, since if the code is incomplete, the + maximum code length that was allowed to get this far is one bit) */ + if (huff != 0) { + here.op = (unsigned char)64; /* invalid code marker */ + here.bits = (unsigned char)(len - drop); + here.val = (unsigned short)0; + next[huff] = here; + } + + /* set return parameters */ + *table += used; + *bits = root; + return 0; +} diff --git a/app/src/main/jni/zlib/inftrees.h b/app/src/main/jni/zlib/inftrees.h new file mode 100644 index 000000000..baa53a0b1 --- /dev/null +++ b/app/src/main/jni/zlib/inftrees.h @@ -0,0 +1,62 @@ +/* inftrees.h -- header to use inftrees.c + * Copyright (C) 1995-2005, 2010 Mark Adler + * For conditions of distribution and use, see copyright notice in zlib.h + */ + +/* WARNING: this file should *not* be used by applications. It is + part of the implementation of the compression library and is + subject to change. Applications should only use zlib.h. + */ + +/* Structure for decoding tables. Each entry provides either the + information needed to do the operation requested by the code that + indexed that table entry, or it provides a pointer to another + table that indexes more bits of the code. op indicates whether + the entry is a pointer to another table, a literal, a length or + distance, an end-of-block, or an invalid code. For a table + pointer, the low four bits of op is the number of index bits of + that table. For a length or distance, the low four bits of op + is the number of extra bits to get after the code. bits is + the number of bits in this code or part of the code to drop off + of the bit buffer. val is the actual byte to output in the case + of a literal, the base length or distance, or the offset from + the current table to the next table. Each entry is four bytes. */ +typedef struct { + unsigned char op; /* operation, extra bits, table bits */ + unsigned char bits; /* bits in this part of the code */ + unsigned short val; /* offset in table or code value */ +} code; + +/* op values as set by inflate_table(): + 00000000 - literal + 0000tttt - table link, tttt != 0 is the number of table index bits + 0001eeee - length or distance, eeee is the number of extra bits + 01100000 - end of block + 01000000 - invalid code + */ + +/* Maximum size of the dynamic table. The maximum number of code structures is + 1444, which is the sum of 852 for literal/length codes and 592 for distance + codes. These values were found by exhaustive searches using the program + examples/enough.c found in the zlib distribtution. The arguments to that + program are the number of symbols, the initial root table size, and the + maximum bit length of a code. "enough 286 9 15" for literal/length codes + returns returns 852, and "enough 30 6 15" for distance codes returns 592. + The initial root table size (9 or 6) is found in the fifth argument of the + inflate_table() calls in inflate.c and infback.c. If the root table size is + changed, then these maximum sizes would be need to be recalculated and + updated. */ +#define ENOUGH_LENS 852 +#define ENOUGH_DISTS 592 +#define ENOUGH (ENOUGH_LENS+ENOUGH_DISTS) + +/* Type of code to build for inflate_table() */ +typedef enum { + CODES, + LENS, + DISTS +} codetype; + +int ZLIB_INTERNAL inflate_table OF((codetype type, unsigned short FAR *lens, + unsigned codes, code FAR * FAR *table, + unsigned FAR *bits, unsigned short FAR *work)); diff --git a/app/src/main/jni/zlib/trees.c b/app/src/main/jni/zlib/trees.c new file mode 100644 index 000000000..1fd7759ef --- /dev/null +++ b/app/src/main/jni/zlib/trees.c @@ -0,0 +1,1226 @@ +/* trees.c -- output deflated data using Huffman coding + * Copyright (C) 1995-2012 Jean-loup Gailly + * detect_data_type() function provided freely by Cosmin Truta, 2006 + * For conditions of distribution and use, see copyright notice in zlib.h + */ + +/* + * ALGORITHM + * + * The "deflation" process uses several Huffman trees. The more + * common source values are represented by shorter bit sequences. + * + * Each code tree is stored in a compressed form which is itself + * a Huffman encoding of the lengths of all the code strings (in + * ascending order by source values). The actual code strings are + * reconstructed from the lengths in the inflate process, as described + * in the deflate specification. + * + * REFERENCES + * + * Deutsch, L.P.,"'Deflate' Compressed Data Format Specification". + * Available in ftp.uu.net:/pub/archiving/zip/doc/deflate-1.1.doc + * + * Storer, James A. + * Data Compression: Methods and Theory, pp. 49-50. + * Computer Science Press, 1988. ISBN 0-7167-8156-5. + * + * Sedgewick, R. + * Algorithms, p290. + * Addison-Wesley, 1983. ISBN 0-201-06672-6. + */ + +/* @(#) $Id$ */ + +/* #define GEN_TREES_H */ + +#include "deflate.h" + +#ifdef DEBUG +# include +#endif + +/* =========================================================================== + * Constants + */ + +#define MAX_BL_BITS 7 +/* Bit length codes must not exceed MAX_BL_BITS bits */ + +#define END_BLOCK 256 +/* end of block literal code */ + +#define REP_3_6 16 +/* repeat previous bit length 3-6 times (2 bits of repeat count) */ + +#define REPZ_3_10 17 +/* repeat a zero length 3-10 times (3 bits of repeat count) */ + +#define REPZ_11_138 18 +/* repeat a zero length 11-138 times (7 bits of repeat count) */ + +local const int extra_lbits[LENGTH_CODES] /* extra bits for each length code */ + = {0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0}; + +local const int extra_dbits[D_CODES] /* extra bits for each distance code */ + = {0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13}; + +local const int extra_blbits[BL_CODES]/* extra bits for each bit length code */ + = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,2,3,7}; + +local const uch bl_order[BL_CODES] + = {16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15}; +/* The lengths of the bit length codes are sent in order of decreasing + * probability, to avoid transmitting the lengths for unused bit length codes. + */ + +/* =========================================================================== + * Local data. These are initialized only once. + */ + +#define DIST_CODE_LEN 512 /* see definition of array dist_code below */ + +#if defined(GEN_TREES_H) || !defined(STDC) +/* non ANSI compilers may not accept trees.h */ + +local ct_data static_ltree[L_CODES+2]; +/* The static literal tree. Since the bit lengths are imposed, there is no + * need for the L_CODES extra codes used during heap construction. However + * The codes 286 and 287 are needed to build a canonical tree (see _tr_init + * below). + */ + +local ct_data static_dtree[D_CODES]; +/* The static distance tree. (Actually a trivial tree since all codes use + * 5 bits.) + */ + +uch _dist_code[DIST_CODE_LEN]; +/* Distance codes. The first 256 values correspond to the distances + * 3 .. 258, the last 256 values correspond to the top 8 bits of + * the 15 bit distances. + */ + +uch _length_code[MAX_MATCH-MIN_MATCH+1]; +/* length code for each normalized match length (0 == MIN_MATCH) */ + +local int base_length[LENGTH_CODES]; +/* First normalized length for each code (0 = MIN_MATCH) */ + +local int base_dist[D_CODES]; +/* First normalized distance for each code (0 = distance of 1) */ + +#else +# include "trees.h" +#endif /* GEN_TREES_H */ + +struct static_tree_desc_s { + const ct_data *static_tree; /* static tree or NULL */ + const intf *extra_bits; /* extra bits for each code or NULL */ + int extra_base; /* base index for extra_bits */ + int elems; /* max number of elements in the tree */ + int max_length; /* max bit length for the codes */ +}; + +local static_tree_desc static_l_desc = +{static_ltree, extra_lbits, LITERALS+1, L_CODES, MAX_BITS}; + +local static_tree_desc static_d_desc = +{static_dtree, extra_dbits, 0, D_CODES, MAX_BITS}; + +local static_tree_desc static_bl_desc = +{(const ct_data *)0, extra_blbits, 0, BL_CODES, MAX_BL_BITS}; + +/* =========================================================================== + * Local (static) routines in this file. + */ + +local void tr_static_init OF((void)); +local void init_block OF((deflate_state *s)); +local void pqdownheap OF((deflate_state *s, ct_data *tree, int k)); +local void gen_bitlen OF((deflate_state *s, tree_desc *desc)); +local void gen_codes OF((ct_data *tree, int max_code, ushf *bl_count)); +local void build_tree OF((deflate_state *s, tree_desc *desc)); +local void scan_tree OF((deflate_state *s, ct_data *tree, int max_code)); +local void send_tree OF((deflate_state *s, ct_data *tree, int max_code)); +local int build_bl_tree OF((deflate_state *s)); +local void send_all_trees OF((deflate_state *s, int lcodes, int dcodes, + int blcodes)); +local void compress_block OF((deflate_state *s, const ct_data *ltree, + const ct_data *dtree)); +local int detect_data_type OF((deflate_state *s)); +local unsigned bi_reverse OF((unsigned value, int length)); +local void bi_windup OF((deflate_state *s)); +local void bi_flush OF((deflate_state *s)); +local void copy_block OF((deflate_state *s, charf *buf, unsigned len, + int header)); + +#ifdef GEN_TREES_H +local void gen_trees_header OF((void)); +#endif + +#ifndef DEBUG +# define send_code(s, c, tree) send_bits(s, tree[c].Code, tree[c].Len) + /* Send a code of the given tree. c and tree must not have side effects */ + +#else /* DEBUG */ +# define send_code(s, c, tree) \ + { if (z_verbose>2) fprintf(stderr,"\ncd %3d ",(c)); \ + send_bits(s, tree[c].Code, tree[c].Len); } +#endif + +/* =========================================================================== + * Output a short LSB first on the stream. + * IN assertion: there is enough room in pendingBuf. + */ +#define put_short(s, w) { \ + put_byte(s, (uch)((w) & 0xff)); \ + put_byte(s, (uch)((ush)(w) >> 8)); \ +} + +/* =========================================================================== + * Send a value on a given number of bits. + * IN assertion: length <= 16 and value fits in length bits. + */ +#ifdef DEBUG +local void send_bits OF((deflate_state *s, int value, int length)); + +local void send_bits(s, value, length) + deflate_state *s; + int value; /* value to send */ + int length; /* number of bits */ +{ + Tracevv((stderr," l %2d v %4x ", length, value)); + Assert(length > 0 && length <= 15, "invalid length"); + s->bits_sent += (ulg)length; + + /* If not enough room in bi_buf, use (valid) bits from bi_buf and + * (16 - bi_valid) bits from value, leaving (width - (16-bi_valid)) + * unused bits in value. + */ + if (s->bi_valid > (int)Buf_size - length) { + s->bi_buf |= (ush)value << s->bi_valid; + put_short(s, s->bi_buf); + s->bi_buf = (ush)value >> (Buf_size - s->bi_valid); + s->bi_valid += length - Buf_size; + } else { + s->bi_buf |= (ush)value << s->bi_valid; + s->bi_valid += length; + } +} +#else /* !DEBUG */ + +#define send_bits(s, value, length) \ +{ int len = length;\ + if (s->bi_valid > (int)Buf_size - len) {\ + int val = value;\ + s->bi_buf |= (ush)val << s->bi_valid;\ + put_short(s, s->bi_buf);\ + s->bi_buf = (ush)val >> (Buf_size - s->bi_valid);\ + s->bi_valid += len - Buf_size;\ + } else {\ + s->bi_buf |= (ush)(value) << s->bi_valid;\ + s->bi_valid += len;\ + }\ +} +#endif /* DEBUG */ + + +/* the arguments must not have side effects */ + +/* =========================================================================== + * Initialize the various 'constant' tables. + */ +local void tr_static_init() +{ +#if defined(GEN_TREES_H) || !defined(STDC) + static int static_init_done = 0; + int n; /* iterates over tree elements */ + int bits; /* bit counter */ + int length; /* length value */ + int code; /* code value */ + int dist; /* distance index */ + ush bl_count[MAX_BITS+1]; + /* number of codes at each bit length for an optimal tree */ + + if (static_init_done) return; + + /* For some embedded targets, global variables are not initialized: */ +#ifdef NO_INIT_GLOBAL_POINTERS + static_l_desc.static_tree = static_ltree; + static_l_desc.extra_bits = extra_lbits; + static_d_desc.static_tree = static_dtree; + static_d_desc.extra_bits = extra_dbits; + static_bl_desc.extra_bits = extra_blbits; +#endif + + /* Initialize the mapping length (0..255) -> length code (0..28) */ + length = 0; + for (code = 0; code < LENGTH_CODES-1; code++) { + base_length[code] = length; + for (n = 0; n < (1< dist code (0..29) */ + dist = 0; + for (code = 0 ; code < 16; code++) { + base_dist[code] = dist; + for (n = 0; n < (1<>= 7; /* from now on, all distances are divided by 128 */ + for ( ; code < D_CODES; code++) { + base_dist[code] = dist << 7; + for (n = 0; n < (1<<(extra_dbits[code]-7)); n++) { + _dist_code[256 + dist++] = (uch)code; + } + } + Assert (dist == 256, "tr_static_init: 256+dist != 512"); + + /* Construct the codes of the static literal tree */ + for (bits = 0; bits <= MAX_BITS; bits++) bl_count[bits] = 0; + n = 0; + while (n <= 143) static_ltree[n++].Len = 8, bl_count[8]++; + while (n <= 255) static_ltree[n++].Len = 9, bl_count[9]++; + while (n <= 279) static_ltree[n++].Len = 7, bl_count[7]++; + while (n <= 287) static_ltree[n++].Len = 8, bl_count[8]++; + /* Codes 286 and 287 do not exist, but we must include them in the + * tree construction to get a canonical Huffman tree (longest code + * all ones) + */ + gen_codes((ct_data *)static_ltree, L_CODES+1, bl_count); + + /* The static distance tree is trivial: */ + for (n = 0; n < D_CODES; n++) { + static_dtree[n].Len = 5; + static_dtree[n].Code = bi_reverse((unsigned)n, 5); + } + static_init_done = 1; + +# ifdef GEN_TREES_H + gen_trees_header(); +# endif +#endif /* defined(GEN_TREES_H) || !defined(STDC) */ +} + +/* =========================================================================== + * Genererate the file trees.h describing the static trees. + */ +#ifdef GEN_TREES_H +# ifndef DEBUG +# include +# endif + +# define SEPARATOR(i, last, width) \ + ((i) == (last)? "\n};\n\n" : \ + ((i) % (width) == (width)-1 ? ",\n" : ", ")) + +void gen_trees_header() +{ + FILE *header = fopen("trees.h", "w"); + int i; + + Assert (header != NULL, "Can't open trees.h"); + fprintf(header, + "/* header created automatically with -DGEN_TREES_H */\n\n"); + + fprintf(header, "local const ct_data static_ltree[L_CODES+2] = {\n"); + for (i = 0; i < L_CODES+2; i++) { + fprintf(header, "{{%3u},{%3u}}%s", static_ltree[i].Code, + static_ltree[i].Len, SEPARATOR(i, L_CODES+1, 5)); + } + + fprintf(header, "local const ct_data static_dtree[D_CODES] = {\n"); + for (i = 0; i < D_CODES; i++) { + fprintf(header, "{{%2u},{%2u}}%s", static_dtree[i].Code, + static_dtree[i].Len, SEPARATOR(i, D_CODES-1, 5)); + } + + fprintf(header, "const uch ZLIB_INTERNAL _dist_code[DIST_CODE_LEN] = {\n"); + for (i = 0; i < DIST_CODE_LEN; i++) { + fprintf(header, "%2u%s", _dist_code[i], + SEPARATOR(i, DIST_CODE_LEN-1, 20)); + } + + fprintf(header, + "const uch ZLIB_INTERNAL _length_code[MAX_MATCH-MIN_MATCH+1]= {\n"); + for (i = 0; i < MAX_MATCH-MIN_MATCH+1; i++) { + fprintf(header, "%2u%s", _length_code[i], + SEPARATOR(i, MAX_MATCH-MIN_MATCH, 20)); + } + + fprintf(header, "local const int base_length[LENGTH_CODES] = {\n"); + for (i = 0; i < LENGTH_CODES; i++) { + fprintf(header, "%1u%s", base_length[i], + SEPARATOR(i, LENGTH_CODES-1, 20)); + } + + fprintf(header, "local const int base_dist[D_CODES] = {\n"); + for (i = 0; i < D_CODES; i++) { + fprintf(header, "%5u%s", base_dist[i], + SEPARATOR(i, D_CODES-1, 10)); + } + + fclose(header); +} +#endif /* GEN_TREES_H */ + +/* =========================================================================== + * Initialize the tree data structures for a new zlib stream. + */ +void ZLIB_INTERNAL _tr_init(s) + deflate_state *s; +{ + tr_static_init(); + + s->l_desc.dyn_tree = s->dyn_ltree; + s->l_desc.stat_desc = &static_l_desc; + + s->d_desc.dyn_tree = s->dyn_dtree; + s->d_desc.stat_desc = &static_d_desc; + + s->bl_desc.dyn_tree = s->bl_tree; + s->bl_desc.stat_desc = &static_bl_desc; + + s->bi_buf = 0; + s->bi_valid = 0; +#ifdef DEBUG + s->compressed_len = 0L; + s->bits_sent = 0L; +#endif + + /* Initialize the first block of the first file: */ + init_block(s); +} + +/* =========================================================================== + * Initialize a new block. + */ +local void init_block(s) + deflate_state *s; +{ + int n; /* iterates over tree elements */ + + /* Initialize the trees. */ + for (n = 0; n < L_CODES; n++) s->dyn_ltree[n].Freq = 0; + for (n = 0; n < D_CODES; n++) s->dyn_dtree[n].Freq = 0; + for (n = 0; n < BL_CODES; n++) s->bl_tree[n].Freq = 0; + + s->dyn_ltree[END_BLOCK].Freq = 1; + s->opt_len = s->static_len = 0L; + s->last_lit = s->matches = 0; +} + +#define SMALLEST 1 +/* Index within the heap array of least frequent node in the Huffman tree */ + + +/* =========================================================================== + * Remove the smallest element from the heap and recreate the heap with + * one less element. Updates heap and heap_len. + */ +#define pqremove(s, tree, top) \ +{\ + top = s->heap[SMALLEST]; \ + s->heap[SMALLEST] = s->heap[s->heap_len--]; \ + pqdownheap(s, tree, SMALLEST); \ +} + +/* =========================================================================== + * Compares to subtrees, using the tree depth as tie breaker when + * the subtrees have equal frequency. This minimizes the worst case length. + */ +#define smaller(tree, n, m, depth) \ + (tree[n].Freq < tree[m].Freq || \ + (tree[n].Freq == tree[m].Freq && depth[n] <= depth[m])) + +/* =========================================================================== + * Restore the heap property by moving down the tree starting at node k, + * exchanging a node with the smallest of its two sons if necessary, stopping + * when the heap property is re-established (each father smaller than its + * two sons). + */ +local void pqdownheap(s, tree, k) + deflate_state *s; + ct_data *tree; /* the tree to restore */ + int k; /* node to move down */ +{ + int v = s->heap[k]; + int j = k << 1; /* left son of k */ + while (j <= s->heap_len) { + /* Set j to the smallest of the two sons: */ + if (j < s->heap_len && + smaller(tree, s->heap[j+1], s->heap[j], s->depth)) { + j++; + } + /* Exit if v is smaller than both sons */ + if (smaller(tree, v, s->heap[j], s->depth)) break; + + /* Exchange v with the smallest son */ + s->heap[k] = s->heap[j]; k = j; + + /* And continue down the tree, setting j to the left son of k */ + j <<= 1; + } + s->heap[k] = v; +} + +/* =========================================================================== + * Compute the optimal bit lengths for a tree and update the total bit length + * for the current block. + * IN assertion: the fields freq and dad are set, heap[heap_max] and + * above are the tree nodes sorted by increasing frequency. + * OUT assertions: the field len is set to the optimal bit length, the + * array bl_count contains the frequencies for each bit length. + * The length opt_len is updated; static_len is also updated if stree is + * not null. + */ +local void gen_bitlen(s, desc) + deflate_state *s; + tree_desc *desc; /* the tree descriptor */ +{ + ct_data *tree = desc->dyn_tree; + int max_code = desc->max_code; + const ct_data *stree = desc->stat_desc->static_tree; + const intf *extra = desc->stat_desc->extra_bits; + int base = desc->stat_desc->extra_base; + int max_length = desc->stat_desc->max_length; + int h; /* heap index */ + int n, m; /* iterate over the tree elements */ + int bits; /* bit length */ + int xbits; /* extra bits */ + ush f; /* frequency */ + int overflow = 0; /* number of elements with bit length too large */ + + for (bits = 0; bits <= MAX_BITS; bits++) s->bl_count[bits] = 0; + + /* In a first pass, compute the optimal bit lengths (which may + * overflow in the case of the bit length tree). + */ + tree[s->heap[s->heap_max]].Len = 0; /* root of the heap */ + + for (h = s->heap_max+1; h < HEAP_SIZE; h++) { + n = s->heap[h]; + bits = tree[tree[n].Dad].Len + 1; + if (bits > max_length) bits = max_length, overflow++; + tree[n].Len = (ush)bits; + /* We overwrite tree[n].Dad which is no longer needed */ + + if (n > max_code) continue; /* not a leaf node */ + + s->bl_count[bits]++; + xbits = 0; + if (n >= base) xbits = extra[n-base]; + f = tree[n].Freq; + s->opt_len += (ulg)f * (bits + xbits); + if (stree) s->static_len += (ulg)f * (stree[n].Len + xbits); + } + if (overflow == 0) return; + + Trace((stderr,"\nbit length overflow\n")); + /* This happens for example on obj2 and pic of the Calgary corpus */ + + /* Find the first bit length which could increase: */ + do { + bits = max_length-1; + while (s->bl_count[bits] == 0) bits--; + s->bl_count[bits]--; /* move one leaf down the tree */ + s->bl_count[bits+1] += 2; /* move one overflow item as its brother */ + s->bl_count[max_length]--; + /* The brother of the overflow item also moves one step up, + * but this does not affect bl_count[max_length] + */ + overflow -= 2; + } while (overflow > 0); + + /* Now recompute all bit lengths, scanning in increasing frequency. + * h is still equal to HEAP_SIZE. (It is simpler to reconstruct all + * lengths instead of fixing only the wrong ones. This idea is taken + * from 'ar' written by Haruhiko Okumura.) + */ + for (bits = max_length; bits != 0; bits--) { + n = s->bl_count[bits]; + while (n != 0) { + m = s->heap[--h]; + if (m > max_code) continue; + if ((unsigned) tree[m].Len != (unsigned) bits) { + Trace((stderr,"code %d bits %d->%d\n", m, tree[m].Len, bits)); + s->opt_len += ((long)bits - (long)tree[m].Len) + *(long)tree[m].Freq; + tree[m].Len = (ush)bits; + } + n--; + } + } +} + +/* =========================================================================== + * Generate the codes for a given tree and bit counts (which need not be + * optimal). + * IN assertion: the array bl_count contains the bit length statistics for + * the given tree and the field len is set for all tree elements. + * OUT assertion: the field code is set for all tree elements of non + * zero code length. + */ +local void gen_codes (tree, max_code, bl_count) + ct_data *tree; /* the tree to decorate */ + int max_code; /* largest code with non zero frequency */ + ushf *bl_count; /* number of codes at each bit length */ +{ + ush next_code[MAX_BITS+1]; /* next code value for each bit length */ + ush code = 0; /* running code value */ + int bits; /* bit index */ + int n; /* code index */ + + /* The distribution counts are first used to generate the code values + * without bit reversal. + */ + for (bits = 1; bits <= MAX_BITS; bits++) { + next_code[bits] = code = (code + bl_count[bits-1]) << 1; + } + /* Check that the bit counts in bl_count are consistent. The last code + * must be all ones. + */ + Assert (code + bl_count[MAX_BITS]-1 == (1<dyn_tree; + const ct_data *stree = desc->stat_desc->static_tree; + int elems = desc->stat_desc->elems; + int n, m; /* iterate over heap elements */ + int max_code = -1; /* largest code with non zero frequency */ + int node; /* new node being created */ + + /* Construct the initial heap, with least frequent element in + * heap[SMALLEST]. The sons of heap[n] are heap[2*n] and heap[2*n+1]. + * heap[0] is not used. + */ + s->heap_len = 0, s->heap_max = HEAP_SIZE; + + for (n = 0; n < elems; n++) { + if (tree[n].Freq != 0) { + s->heap[++(s->heap_len)] = max_code = n; + s->depth[n] = 0; + } else { + tree[n].Len = 0; + } + } + + /* The pkzip format requires that at least one distance code exists, + * and that at least one bit should be sent even if there is only one + * possible code. So to avoid special checks later on we force at least + * two codes of non zero frequency. + */ + while (s->heap_len < 2) { + node = s->heap[++(s->heap_len)] = (max_code < 2 ? ++max_code : 0); + tree[node].Freq = 1; + s->depth[node] = 0; + s->opt_len--; if (stree) s->static_len -= stree[node].Len; + /* node is 0 or 1 so it does not have extra bits */ + } + desc->max_code = max_code; + + /* The elements heap[heap_len/2+1 .. heap_len] are leaves of the tree, + * establish sub-heaps of increasing lengths: + */ + for (n = s->heap_len/2; n >= 1; n--) pqdownheap(s, tree, n); + + /* Construct the Huffman tree by repeatedly combining the least two + * frequent nodes. + */ + node = elems; /* next internal node of the tree */ + do { + pqremove(s, tree, n); /* n = node of least frequency */ + m = s->heap[SMALLEST]; /* m = node of next least frequency */ + + s->heap[--(s->heap_max)] = n; /* keep the nodes sorted by frequency */ + s->heap[--(s->heap_max)] = m; + + /* Create a new node father of n and m */ + tree[node].Freq = tree[n].Freq + tree[m].Freq; + s->depth[node] = (uch)((s->depth[n] >= s->depth[m] ? + s->depth[n] : s->depth[m]) + 1); + tree[n].Dad = tree[m].Dad = (ush)node; +#ifdef DUMP_BL_TREE + if (tree == s->bl_tree) { + fprintf(stderr,"\nnode %d(%d), sons %d(%d) %d(%d)", + node, tree[node].Freq, n, tree[n].Freq, m, tree[m].Freq); + } +#endif + /* and insert the new node in the heap */ + s->heap[SMALLEST] = node++; + pqdownheap(s, tree, SMALLEST); + + } while (s->heap_len >= 2); + + s->heap[--(s->heap_max)] = s->heap[SMALLEST]; + + /* At this point, the fields freq and dad are set. We can now + * generate the bit lengths. + */ + gen_bitlen(s, (tree_desc *)desc); + + /* The field len is now set, we can generate the bit codes */ + gen_codes ((ct_data *)tree, max_code, s->bl_count); +} + +/* =========================================================================== + * Scan a literal or distance tree to determine the frequencies of the codes + * in the bit length tree. + */ +local void scan_tree (s, tree, max_code) + deflate_state *s; + ct_data *tree; /* the tree to be scanned */ + int max_code; /* and its largest code of non zero frequency */ +{ + int n; /* iterates over all tree elements */ + int prevlen = -1; /* last emitted length */ + int curlen; /* length of current code */ + int nextlen = tree[0].Len; /* length of next code */ + int count = 0; /* repeat count of the current code */ + int max_count = 7; /* max repeat count */ + int min_count = 4; /* min repeat count */ + + if (nextlen == 0) max_count = 138, min_count = 3; + tree[max_code+1].Len = (ush)0xffff; /* guard */ + + for (n = 0; n <= max_code; n++) { + curlen = nextlen; nextlen = tree[n+1].Len; + if (++count < max_count && curlen == nextlen) { + continue; + } else if (count < min_count) { + s->bl_tree[curlen].Freq += count; + } else if (curlen != 0) { + if (curlen != prevlen) s->bl_tree[curlen].Freq++; + s->bl_tree[REP_3_6].Freq++; + } else if (count <= 10) { + s->bl_tree[REPZ_3_10].Freq++; + } else { + s->bl_tree[REPZ_11_138].Freq++; + } + count = 0; prevlen = curlen; + if (nextlen == 0) { + max_count = 138, min_count = 3; + } else if (curlen == nextlen) { + max_count = 6, min_count = 3; + } else { + max_count = 7, min_count = 4; + } + } +} + +/* =========================================================================== + * Send a literal or distance tree in compressed form, using the codes in + * bl_tree. + */ +local void send_tree (s, tree, max_code) + deflate_state *s; + ct_data *tree; /* the tree to be scanned */ + int max_code; /* and its largest code of non zero frequency */ +{ + int n; /* iterates over all tree elements */ + int prevlen = -1; /* last emitted length */ + int curlen; /* length of current code */ + int nextlen = tree[0].Len; /* length of next code */ + int count = 0; /* repeat count of the current code */ + int max_count = 7; /* max repeat count */ + int min_count = 4; /* min repeat count */ + + /* tree[max_code+1].Len = -1; */ /* guard already set */ + if (nextlen == 0) max_count = 138, min_count = 3; + + for (n = 0; n <= max_code; n++) { + curlen = nextlen; nextlen = tree[n+1].Len; + if (++count < max_count && curlen == nextlen) { + continue; + } else if (count < min_count) { + do { send_code(s, curlen, s->bl_tree); } while (--count != 0); + + } else if (curlen != 0) { + if (curlen != prevlen) { + send_code(s, curlen, s->bl_tree); count--; + } + Assert(count >= 3 && count <= 6, " 3_6?"); + send_code(s, REP_3_6, s->bl_tree); send_bits(s, count-3, 2); + + } else if (count <= 10) { + send_code(s, REPZ_3_10, s->bl_tree); send_bits(s, count-3, 3); + + } else { + send_code(s, REPZ_11_138, s->bl_tree); send_bits(s, count-11, 7); + } + count = 0; prevlen = curlen; + if (nextlen == 0) { + max_count = 138, min_count = 3; + } else if (curlen == nextlen) { + max_count = 6, min_count = 3; + } else { + max_count = 7, min_count = 4; + } + } +} + +/* =========================================================================== + * Construct the Huffman tree for the bit lengths and return the index in + * bl_order of the last bit length code to send. + */ +local int build_bl_tree(s) + deflate_state *s; +{ + int max_blindex; /* index of last bit length code of non zero freq */ + + /* Determine the bit length frequencies for literal and distance trees */ + scan_tree(s, (ct_data *)s->dyn_ltree, s->l_desc.max_code); + scan_tree(s, (ct_data *)s->dyn_dtree, s->d_desc.max_code); + + /* Build the bit length tree: */ + build_tree(s, (tree_desc *)(&(s->bl_desc))); + /* opt_len now includes the length of the tree representations, except + * the lengths of the bit lengths codes and the 5+5+4 bits for the counts. + */ + + /* Determine the number of bit length codes to send. The pkzip format + * requires that at least 4 bit length codes be sent. (appnote.txt says + * 3 but the actual value used is 4.) + */ + for (max_blindex = BL_CODES-1; max_blindex >= 3; max_blindex--) { + if (s->bl_tree[bl_order[max_blindex]].Len != 0) break; + } + /* Update opt_len to include the bit length tree and counts */ + s->opt_len += 3*(max_blindex+1) + 5+5+4; + Tracev((stderr, "\ndyn trees: dyn %ld, stat %ld", + s->opt_len, s->static_len)); + + return max_blindex; +} + +/* =========================================================================== + * Send the header for a block using dynamic Huffman trees: the counts, the + * lengths of the bit length codes, the literal tree and the distance tree. + * IN assertion: lcodes >= 257, dcodes >= 1, blcodes >= 4. + */ +local void send_all_trees(s, lcodes, dcodes, blcodes) + deflate_state *s; + int lcodes, dcodes, blcodes; /* number of codes for each tree */ +{ + int rank; /* index in bl_order */ + + Assert (lcodes >= 257 && dcodes >= 1 && blcodes >= 4, "not enough codes"); + Assert (lcodes <= L_CODES && dcodes <= D_CODES && blcodes <= BL_CODES, + "too many codes"); + Tracev((stderr, "\nbl counts: ")); + send_bits(s, lcodes-257, 5); /* not +255 as stated in appnote.txt */ + send_bits(s, dcodes-1, 5); + send_bits(s, blcodes-4, 4); /* not -3 as stated in appnote.txt */ + for (rank = 0; rank < blcodes; rank++) { + Tracev((stderr, "\nbl code %2d ", bl_order[rank])); + send_bits(s, s->bl_tree[bl_order[rank]].Len, 3); + } + Tracev((stderr, "\nbl tree: sent %ld", s->bits_sent)); + + send_tree(s, (ct_data *)s->dyn_ltree, lcodes-1); /* literal tree */ + Tracev((stderr, "\nlit tree: sent %ld", s->bits_sent)); + + send_tree(s, (ct_data *)s->dyn_dtree, dcodes-1); /* distance tree */ + Tracev((stderr, "\ndist tree: sent %ld", s->bits_sent)); +} + +/* =========================================================================== + * Send a stored block + */ +void ZLIB_INTERNAL _tr_stored_block(s, buf, stored_len, last) + deflate_state *s; + charf *buf; /* input block */ + ulg stored_len; /* length of input block */ + int last; /* one if this is the last block for a file */ +{ + send_bits(s, (STORED_BLOCK<<1)+last, 3); /* send block type */ +#ifdef DEBUG + s->compressed_len = (s->compressed_len + 3 + 7) & (ulg)~7L; + s->compressed_len += (stored_len + 4) << 3; +#endif + copy_block(s, buf, (unsigned)stored_len, 1); /* with header */ +} + +/* =========================================================================== + * Flush the bits in the bit buffer to pending output (leaves at most 7 bits) + */ +void ZLIB_INTERNAL _tr_flush_bits(s) + deflate_state *s; +{ + bi_flush(s); +} + +/* =========================================================================== + * Send one empty static block to give enough lookahead for inflate. + * This takes 10 bits, of which 7 may remain in the bit buffer. + */ +void ZLIB_INTERNAL _tr_align(s) + deflate_state *s; +{ + send_bits(s, STATIC_TREES<<1, 3); + send_code(s, END_BLOCK, static_ltree); +#ifdef DEBUG + s->compressed_len += 10L; /* 3 for block type, 7 for EOB */ +#endif + bi_flush(s); +} + +/* =========================================================================== + * Determine the best encoding for the current block: dynamic trees, static + * trees or store, and output the encoded block to the zip file. + */ +void ZLIB_INTERNAL _tr_flush_block(s, buf, stored_len, last) + deflate_state *s; + charf *buf; /* input block, or NULL if too old */ + ulg stored_len; /* length of input block */ + int last; /* one if this is the last block for a file */ +{ + ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */ + int max_blindex = 0; /* index of last bit length code of non zero freq */ + + /* Build the Huffman trees unless a stored block is forced */ + if (s->level > 0) { + + /* Check if the file is binary or text */ + if (s->strm->data_type == Z_UNKNOWN) + s->strm->data_type = detect_data_type(s); + + /* Construct the literal and distance trees */ + build_tree(s, (tree_desc *)(&(s->l_desc))); + Tracev((stderr, "\nlit data: dyn %ld, stat %ld", s->opt_len, + s->static_len)); + + build_tree(s, (tree_desc *)(&(s->d_desc))); + Tracev((stderr, "\ndist data: dyn %ld, stat %ld", s->opt_len, + s->static_len)); + /* At this point, opt_len and static_len are the total bit lengths of + * the compressed block data, excluding the tree representations. + */ + + /* Build the bit length tree for the above two trees, and get the index + * in bl_order of the last bit length code to send. + */ + max_blindex = build_bl_tree(s); + + /* Determine the best encoding. Compute the block lengths in bytes. */ + opt_lenb = (s->opt_len+3+7)>>3; + static_lenb = (s->static_len+3+7)>>3; + + Tracev((stderr, "\nopt %lu(%lu) stat %lu(%lu) stored %lu lit %u ", + opt_lenb, s->opt_len, static_lenb, s->static_len, stored_len, + s->last_lit)); + + if (static_lenb <= opt_lenb) opt_lenb = static_lenb; + + } else { + Assert(buf != (char*)0, "lost buf"); + opt_lenb = static_lenb = stored_len + 5; /* force a stored block */ + } + +#ifdef FORCE_STORED + if (buf != (char*)0) { /* force stored block */ +#else + if (stored_len+4 <= opt_lenb && buf != (char*)0) { + /* 4: two words for the lengths */ +#endif + /* The test buf != NULL is only necessary if LIT_BUFSIZE > WSIZE. + * Otherwise we can't have processed more than WSIZE input bytes since + * the last block flush, because compression would have been + * successful. If LIT_BUFSIZE <= WSIZE, it is never too late to + * transform a block into a stored block. + */ + _tr_stored_block(s, buf, stored_len, last); + +#ifdef FORCE_STATIC + } else if (static_lenb >= 0) { /* force static trees */ +#else + } else if (s->strategy == Z_FIXED || static_lenb == opt_lenb) { +#endif + send_bits(s, (STATIC_TREES<<1)+last, 3); + compress_block(s, (const ct_data *)static_ltree, + (const ct_data *)static_dtree); +#ifdef DEBUG + s->compressed_len += 3 + s->static_len; +#endif + } else { + send_bits(s, (DYN_TREES<<1)+last, 3); + send_all_trees(s, s->l_desc.max_code+1, s->d_desc.max_code+1, + max_blindex+1); + compress_block(s, (const ct_data *)s->dyn_ltree, + (const ct_data *)s->dyn_dtree); +#ifdef DEBUG + s->compressed_len += 3 + s->opt_len; +#endif + } + Assert (s->compressed_len == s->bits_sent, "bad compressed size"); + /* The above check is made mod 2^32, for files larger than 512 MB + * and uLong implemented on 32 bits. + */ + init_block(s); + + if (last) { + bi_windup(s); +#ifdef DEBUG + s->compressed_len += 7; /* align on byte boundary */ +#endif + } + Tracev((stderr,"\ncomprlen %lu(%lu) ", s->compressed_len>>3, + s->compressed_len-7*last)); +} + +/* =========================================================================== + * Save the match info and tally the frequency counts. Return true if + * the current block must be flushed. + */ +int ZLIB_INTERNAL _tr_tally (s, dist, lc) + deflate_state *s; + unsigned dist; /* distance of matched string */ + unsigned lc; /* match length-MIN_MATCH or unmatched char (if dist==0) */ +{ + s->d_buf[s->last_lit] = (ush)dist; + s->l_buf[s->last_lit++] = (uch)lc; + if (dist == 0) { + /* lc is the unmatched char */ + s->dyn_ltree[lc].Freq++; + } else { + s->matches++; + /* Here, lc is the match length - MIN_MATCH */ + dist--; /* dist = match distance - 1 */ + Assert((ush)dist < (ush)MAX_DIST(s) && + (ush)lc <= (ush)(MAX_MATCH-MIN_MATCH) && + (ush)d_code(dist) < (ush)D_CODES, "_tr_tally: bad match"); + + s->dyn_ltree[_length_code[lc]+LITERALS+1].Freq++; + s->dyn_dtree[d_code(dist)].Freq++; + } + +#ifdef TRUNCATE_BLOCK + /* Try to guess if it is profitable to stop the current block here */ + if ((s->last_lit & 0x1fff) == 0 && s->level > 2) { + /* Compute an upper bound for the compressed length */ + ulg out_length = (ulg)s->last_lit*8L; + ulg in_length = (ulg)((long)s->strstart - s->block_start); + int dcode; + for (dcode = 0; dcode < D_CODES; dcode++) { + out_length += (ulg)s->dyn_dtree[dcode].Freq * + (5L+extra_dbits[dcode]); + } + out_length >>= 3; + Tracev((stderr,"\nlast_lit %u, in %ld, out ~%ld(%ld%%) ", + s->last_lit, in_length, out_length, + 100L - out_length*100L/in_length)); + if (s->matches < s->last_lit/2 && out_length < in_length/2) return 1; + } +#endif + return (s->last_lit == s->lit_bufsize-1); + /* We avoid equality with lit_bufsize because of wraparound at 64K + * on 16 bit machines and because stored blocks are restricted to + * 64K-1 bytes. + */ +} + +/* =========================================================================== + * Send the block data compressed using the given Huffman trees + */ +local void compress_block(s, ltree, dtree) + deflate_state *s; + const ct_data *ltree; /* literal tree */ + const ct_data *dtree; /* distance tree */ +{ + unsigned dist; /* distance of matched string */ + int lc; /* match length or unmatched char (if dist == 0) */ + unsigned lx = 0; /* running index in l_buf */ + unsigned code; /* the code to send */ + int extra; /* number of extra bits to send */ + + if (s->last_lit != 0) do { + dist = s->d_buf[lx]; + lc = s->l_buf[lx++]; + if (dist == 0) { + send_code(s, lc, ltree); /* send a literal byte */ + Tracecv(isgraph(lc), (stderr," '%c' ", lc)); + } else { + /* Here, lc is the match length - MIN_MATCH */ + code = _length_code[lc]; + send_code(s, code+LITERALS+1, ltree); /* send the length code */ + extra = extra_lbits[code]; + if (extra != 0) { + lc -= base_length[code]; + send_bits(s, lc, extra); /* send the extra length bits */ + } + dist--; /* dist is now the match distance - 1 */ + code = d_code(dist); + Assert (code < D_CODES, "bad d_code"); + + send_code(s, code, dtree); /* send the distance code */ + extra = extra_dbits[code]; + if (extra != 0) { + dist -= base_dist[code]; + send_bits(s, dist, extra); /* send the extra distance bits */ + } + } /* literal or match pair ? */ + + /* Check that the overlay between pending_buf and d_buf+l_buf is ok: */ + Assert((uInt)(s->pending) < s->lit_bufsize + 2*lx, + "pendingBuf overflow"); + + } while (lx < s->last_lit); + + send_code(s, END_BLOCK, ltree); +} + +/* =========================================================================== + * Check if the data type is TEXT or BINARY, using the following algorithm: + * - TEXT if the two conditions below are satisfied: + * a) There are no non-portable control characters belonging to the + * "black list" (0..6, 14..25, 28..31). + * b) There is at least one printable character belonging to the + * "white list" (9 {TAB}, 10 {LF}, 13 {CR}, 32..255). + * - BINARY otherwise. + * - The following partially-portable control characters form a + * "gray list" that is ignored in this detection algorithm: + * (7 {BEL}, 8 {BS}, 11 {VT}, 12 {FF}, 26 {SUB}, 27 {ESC}). + * IN assertion: the fields Freq of dyn_ltree are set. + */ +local int detect_data_type(s) + deflate_state *s; +{ + /* black_mask is the bit mask of black-listed bytes + * set bits 0..6, 14..25, and 28..31 + * 0xf3ffc07f = binary 11110011111111111100000001111111 + */ + unsigned long black_mask = 0xf3ffc07fUL; + int n; + + /* Check for non-textual ("black-listed") bytes. */ + for (n = 0; n <= 31; n++, black_mask >>= 1) + if ((black_mask & 1) && (s->dyn_ltree[n].Freq != 0)) + return Z_BINARY; + + /* Check for textual ("white-listed") bytes. */ + if (s->dyn_ltree[9].Freq != 0 || s->dyn_ltree[10].Freq != 0 + || s->dyn_ltree[13].Freq != 0) + return Z_TEXT; + for (n = 32; n < LITERALS; n++) + if (s->dyn_ltree[n].Freq != 0) + return Z_TEXT; + + /* There are no "black-listed" or "white-listed" bytes: + * this stream either is empty or has tolerated ("gray-listed") bytes only. + */ + return Z_BINARY; +} + +/* =========================================================================== + * Reverse the first len bits of a code, using straightforward code (a faster + * method would use a table) + * IN assertion: 1 <= len <= 15 + */ +local unsigned bi_reverse(code, len) + unsigned code; /* the value to invert */ + int len; /* its bit length */ +{ + register unsigned res = 0; + do { + res |= code & 1; + code >>= 1, res <<= 1; + } while (--len > 0); + return res >> 1; +} + +/* =========================================================================== + * Flush the bit buffer, keeping at most 7 bits in it. + */ +local void bi_flush(s) + deflate_state *s; +{ + if (s->bi_valid == 16) { + put_short(s, s->bi_buf); + s->bi_buf = 0; + s->bi_valid = 0; + } else if (s->bi_valid >= 8) { + put_byte(s, (Byte)s->bi_buf); + s->bi_buf >>= 8; + s->bi_valid -= 8; + } +} + +/* =========================================================================== + * Flush the bit buffer and align the output on a byte boundary + */ +local void bi_windup(s) + deflate_state *s; +{ + if (s->bi_valid > 8) { + put_short(s, s->bi_buf); + } else if (s->bi_valid > 0) { + put_byte(s, (Byte)s->bi_buf); + } + s->bi_buf = 0; + s->bi_valid = 0; +#ifdef DEBUG + s->bits_sent = (s->bits_sent+7) & ~7; +#endif +} + +/* =========================================================================== + * Copy a stored block, storing first the length and its + * one's complement if requested. + */ +local void copy_block(s, buf, len, header) + deflate_state *s; + charf *buf; /* the input data */ + unsigned len; /* its length */ + int header; /* true if block header must be written */ +{ + bi_windup(s); /* align on byte boundary */ + + if (header) { + put_short(s, (ush)len); + put_short(s, (ush)~len); +#ifdef DEBUG + s->bits_sent += 2*16; +#endif + } +#ifdef DEBUG + s->bits_sent += (ulg)len<<3; +#endif + while (len--) { + put_byte(s, *buf++); + } +} diff --git a/app/src/main/jni/zlib/trees.h b/app/src/main/jni/zlib/trees.h new file mode 100644 index 000000000..d35639d82 --- /dev/null +++ b/app/src/main/jni/zlib/trees.h @@ -0,0 +1,128 @@ +/* header created automatically with -DGEN_TREES_H */ + +local const ct_data static_ltree[L_CODES+2] = { +{{ 12},{ 8}}, {{140},{ 8}}, {{ 76},{ 8}}, {{204},{ 8}}, {{ 44},{ 8}}, +{{172},{ 8}}, {{108},{ 8}}, {{236},{ 8}}, {{ 28},{ 8}}, {{156},{ 8}}, +{{ 92},{ 8}}, {{220},{ 8}}, {{ 60},{ 8}}, {{188},{ 8}}, {{124},{ 8}}, +{{252},{ 8}}, {{ 2},{ 8}}, {{130},{ 8}}, {{ 66},{ 8}}, {{194},{ 8}}, +{{ 34},{ 8}}, {{162},{ 8}}, {{ 98},{ 8}}, {{226},{ 8}}, {{ 18},{ 8}}, +{{146},{ 8}}, {{ 82},{ 8}}, {{210},{ 8}}, {{ 50},{ 8}}, {{178},{ 8}}, +{{114},{ 8}}, {{242},{ 8}}, {{ 10},{ 8}}, {{138},{ 8}}, {{ 74},{ 8}}, +{{202},{ 8}}, {{ 42},{ 8}}, {{170},{ 8}}, {{106},{ 8}}, {{234},{ 8}}, +{{ 26},{ 8}}, {{154},{ 8}}, {{ 90},{ 8}}, {{218},{ 8}}, {{ 58},{ 8}}, +{{186},{ 8}}, {{122},{ 8}}, {{250},{ 8}}, {{ 6},{ 8}}, {{134},{ 8}}, +{{ 70},{ 8}}, {{198},{ 8}}, {{ 38},{ 8}}, {{166},{ 8}}, {{102},{ 8}}, +{{230},{ 8}}, {{ 22},{ 8}}, {{150},{ 8}}, {{ 86},{ 8}}, {{214},{ 8}}, +{{ 54},{ 8}}, {{182},{ 8}}, {{118},{ 8}}, {{246},{ 8}}, {{ 14},{ 8}}, +{{142},{ 8}}, {{ 78},{ 8}}, {{206},{ 8}}, {{ 46},{ 8}}, {{174},{ 8}}, +{{110},{ 8}}, {{238},{ 8}}, {{ 30},{ 8}}, {{158},{ 8}}, {{ 94},{ 8}}, +{{222},{ 8}}, {{ 62},{ 8}}, {{190},{ 8}}, {{126},{ 8}}, {{254},{ 8}}, +{{ 1},{ 8}}, {{129},{ 8}}, {{ 65},{ 8}}, {{193},{ 8}}, {{ 33},{ 8}}, +{{161},{ 8}}, {{ 97},{ 8}}, {{225},{ 8}}, {{ 17},{ 8}}, {{145},{ 8}}, +{{ 81},{ 8}}, {{209},{ 8}}, {{ 49},{ 8}}, {{177},{ 8}}, {{113},{ 8}}, +{{241},{ 8}}, {{ 9},{ 8}}, {{137},{ 8}}, {{ 73},{ 8}}, {{201},{ 8}}, +{{ 41},{ 8}}, {{169},{ 8}}, {{105},{ 8}}, {{233},{ 8}}, {{ 25},{ 8}}, +{{153},{ 8}}, {{ 89},{ 8}}, {{217},{ 8}}, {{ 57},{ 8}}, {{185},{ 8}}, +{{121},{ 8}}, {{249},{ 8}}, {{ 5},{ 8}}, {{133},{ 8}}, {{ 69},{ 8}}, +{{197},{ 8}}, {{ 37},{ 8}}, {{165},{ 8}}, {{101},{ 8}}, {{229},{ 8}}, +{{ 21},{ 8}}, {{149},{ 8}}, {{ 85},{ 8}}, {{213},{ 8}}, {{ 53},{ 8}}, +{{181},{ 8}}, {{117},{ 8}}, {{245},{ 8}}, {{ 13},{ 8}}, {{141},{ 8}}, +{{ 77},{ 8}}, {{205},{ 8}}, {{ 45},{ 8}}, {{173},{ 8}}, {{109},{ 8}}, +{{237},{ 8}}, {{ 29},{ 8}}, {{157},{ 8}}, {{ 93},{ 8}}, {{221},{ 8}}, +{{ 61},{ 8}}, {{189},{ 8}}, {{125},{ 8}}, {{253},{ 8}}, {{ 19},{ 9}}, +{{275},{ 9}}, {{147},{ 9}}, {{403},{ 9}}, {{ 83},{ 9}}, {{339},{ 9}}, +{{211},{ 9}}, {{467},{ 9}}, {{ 51},{ 9}}, {{307},{ 9}}, {{179},{ 9}}, +{{435},{ 9}}, {{115},{ 9}}, {{371},{ 9}}, {{243},{ 9}}, {{499},{ 9}}, +{{ 11},{ 9}}, {{267},{ 9}}, {{139},{ 9}}, {{395},{ 9}}, {{ 75},{ 9}}, +{{331},{ 9}}, {{203},{ 9}}, {{459},{ 9}}, {{ 43},{ 9}}, {{299},{ 9}}, +{{171},{ 9}}, {{427},{ 9}}, {{107},{ 9}}, {{363},{ 9}}, {{235},{ 9}}, +{{491},{ 9}}, {{ 27},{ 9}}, {{283},{ 9}}, {{155},{ 9}}, {{411},{ 9}}, +{{ 91},{ 9}}, {{347},{ 9}}, {{219},{ 9}}, {{475},{ 9}}, {{ 59},{ 9}}, +{{315},{ 9}}, {{187},{ 9}}, {{443},{ 9}}, {{123},{ 9}}, {{379},{ 9}}, +{{251},{ 9}}, {{507},{ 9}}, {{ 7},{ 9}}, {{263},{ 9}}, {{135},{ 9}}, +{{391},{ 9}}, {{ 71},{ 9}}, {{327},{ 9}}, {{199},{ 9}}, {{455},{ 9}}, +{{ 39},{ 9}}, {{295},{ 9}}, {{167},{ 9}}, {{423},{ 9}}, {{103},{ 9}}, +{{359},{ 9}}, {{231},{ 9}}, {{487},{ 9}}, {{ 23},{ 9}}, {{279},{ 9}}, +{{151},{ 9}}, {{407},{ 9}}, {{ 87},{ 9}}, {{343},{ 9}}, {{215},{ 9}}, +{{471},{ 9}}, {{ 55},{ 9}}, {{311},{ 9}}, {{183},{ 9}}, {{439},{ 9}}, +{{119},{ 9}}, {{375},{ 9}}, {{247},{ 9}}, {{503},{ 9}}, {{ 15},{ 9}}, +{{271},{ 9}}, {{143},{ 9}}, {{399},{ 9}}, {{ 79},{ 9}}, {{335},{ 9}}, +{{207},{ 9}}, {{463},{ 9}}, {{ 47},{ 9}}, {{303},{ 9}}, {{175},{ 9}}, +{{431},{ 9}}, {{111},{ 9}}, {{367},{ 9}}, {{239},{ 9}}, {{495},{ 9}}, +{{ 31},{ 9}}, {{287},{ 9}}, {{159},{ 9}}, {{415},{ 9}}, {{ 95},{ 9}}, +{{351},{ 9}}, {{223},{ 9}}, {{479},{ 9}}, {{ 63},{ 9}}, {{319},{ 9}}, +{{191},{ 9}}, {{447},{ 9}}, {{127},{ 9}}, {{383},{ 9}}, {{255},{ 9}}, +{{511},{ 9}}, {{ 0},{ 7}}, {{ 64},{ 7}}, {{ 32},{ 7}}, {{ 96},{ 7}}, +{{ 16},{ 7}}, {{ 80},{ 7}}, {{ 48},{ 7}}, {{112},{ 7}}, {{ 8},{ 7}}, +{{ 72},{ 7}}, {{ 40},{ 7}}, {{104},{ 7}}, {{ 24},{ 7}}, {{ 88},{ 7}}, +{{ 56},{ 7}}, {{120},{ 7}}, {{ 4},{ 7}}, {{ 68},{ 7}}, {{ 36},{ 7}}, +{{100},{ 7}}, {{ 20},{ 7}}, {{ 84},{ 7}}, {{ 52},{ 7}}, {{116},{ 7}}, +{{ 3},{ 8}}, {{131},{ 8}}, {{ 67},{ 8}}, {{195},{ 8}}, {{ 35},{ 8}}, +{{163},{ 8}}, {{ 99},{ 8}}, {{227},{ 8}} +}; + +local const ct_data static_dtree[D_CODES] = { +{{ 0},{ 5}}, {{16},{ 5}}, {{ 8},{ 5}}, {{24},{ 5}}, {{ 4},{ 5}}, +{{20},{ 5}}, {{12},{ 5}}, {{28},{ 5}}, {{ 2},{ 5}}, {{18},{ 5}}, +{{10},{ 5}}, {{26},{ 5}}, {{ 6},{ 5}}, {{22},{ 5}}, {{14},{ 5}}, +{{30},{ 5}}, {{ 1},{ 5}}, {{17},{ 5}}, {{ 9},{ 5}}, {{25},{ 5}}, +{{ 5},{ 5}}, {{21},{ 5}}, {{13},{ 5}}, {{29},{ 5}}, {{ 3},{ 5}}, +{{19},{ 5}}, {{11},{ 5}}, {{27},{ 5}}, {{ 7},{ 5}}, {{23},{ 5}} +}; + +const uch ZLIB_INTERNAL _dist_code[DIST_CODE_LEN] = { + 0, 1, 2, 3, 4, 4, 5, 5, 6, 6, 6, 6, 7, 7, 7, 7, 8, 8, 8, 8, + 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 9, 9, 10, 10, 10, 10, 10, 10, 10, 10, +10, 10, 10, 10, 10, 10, 10, 10, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, +11, 11, 11, 11, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, +12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 13, 13, 13, 13, +13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, +13, 13, 13, 13, 13, 13, 13, 13, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, +14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, +14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, +14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 15, 15, 15, 15, 15, 15, 15, 15, +15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, +15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, +15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 0, 0, 16, 17, +18, 18, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 22, 22, 22, 22, 22, 22, 22, 22, +23, 23, 23, 23, 23, 23, 23, 23, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, +24, 24, 24, 24, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, +26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, +26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 27, 27, 27, 27, 27, 27, 27, 27, +27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, +27, 27, 27, 27, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, +28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, +28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, +28, 28, 28, 28, 28, 28, 28, 28, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, +29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, +29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, +29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29 +}; + +const uch ZLIB_INTERNAL _length_code[MAX_MATCH-MIN_MATCH+1]= { + 0, 1, 2, 3, 4, 5, 6, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12, 12, 12, 12, +13, 13, 13, 13, 14, 14, 14, 14, 15, 15, 15, 15, 16, 16, 16, 16, 16, 16, 16, 16, +17, 17, 17, 17, 17, 17, 17, 17, 18, 18, 18, 18, 18, 18, 18, 18, 19, 19, 19, 19, +19, 19, 19, 19, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, +21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 22, 22, 22, 22, +22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 23, 23, 23, 23, 23, 23, 23, 23, +23, 23, 23, 23, 23, 23, 23, 23, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, +24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, +25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, +25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 26, 26, 26, 26, 26, 26, 26, 26, +26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, +26, 26, 26, 26, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, +27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 28 +}; + +local const int base_length[LENGTH_CODES] = { +0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 20, 24, 28, 32, 40, 48, 56, +64, 80, 96, 112, 128, 160, 192, 224, 0 +}; + +local const int base_dist[D_CODES] = { + 0, 1, 2, 3, 4, 6, 8, 12, 16, 24, + 32, 48, 64, 96, 128, 192, 256, 384, 512, 768, + 1024, 1536, 2048, 3072, 4096, 6144, 8192, 12288, 16384, 24576 +}; + diff --git a/app/src/main/jni/zlib/zconf.h b/app/src/main/jni/zlib/zconf.h new file mode 100644 index 000000000..996fff292 --- /dev/null +++ b/app/src/main/jni/zlib/zconf.h @@ -0,0 +1,511 @@ +/* zconf.h -- configuration of the zlib compression library + * Copyright (C) 1995-2013 Jean-loup Gailly. + * For conditions of distribution and use, see copyright notice in zlib.h + */ + +/* @(#) $Id$ */ + +#ifndef ZCONF_H +#define ZCONF_H + +/* + * If you *really* need a unique prefix for all types and library functions, + * compile with -DZ_PREFIX. The "standard" zlib should be compiled without it. + * Even better than compiling with -DZ_PREFIX would be to use configure to set + * this permanently in zconf.h using "./configure --zprefix". + */ +#ifdef Z_PREFIX /* may be set to #if 1 by ./configure */ +# define Z_PREFIX_SET + +/* all linked symbols */ +# define _dist_code z__dist_code +# define _length_code z__length_code +# define _tr_align z__tr_align +# define _tr_flush_bits z__tr_flush_bits +# define _tr_flush_block z__tr_flush_block +# define _tr_init z__tr_init +# define _tr_stored_block z__tr_stored_block +# define _tr_tally z__tr_tally +# define adler32 z_adler32 +# define adler32_combine z_adler32_combine +# define adler32_combine64 z_adler32_combine64 +# ifndef Z_SOLO +# define compress z_compress +# define compress2 z_compress2 +# define compressBound z_compressBound +# endif +# define crc32 z_crc32 +# define crc32_combine z_crc32_combine +# define crc32_combine64 z_crc32_combine64 +# define deflate z_deflate +# define deflateBound z_deflateBound +# define deflateCopy z_deflateCopy +# define deflateEnd z_deflateEnd +# define deflateInit2_ z_deflateInit2_ +# define deflateInit_ z_deflateInit_ +# define deflateParams z_deflateParams +# define deflatePending z_deflatePending +# define deflatePrime z_deflatePrime +# define deflateReset z_deflateReset +# define deflateResetKeep z_deflateResetKeep +# define deflateSetDictionary z_deflateSetDictionary +# define deflateSetHeader z_deflateSetHeader +# define deflateTune z_deflateTune +# define deflate_copyright z_deflate_copyright +# define get_crc_table z_get_crc_table +# ifndef Z_SOLO +# define gz_error z_gz_error +# define gz_intmax z_gz_intmax +# define gz_strwinerror z_gz_strwinerror +# define gzbuffer z_gzbuffer +# define gzclearerr z_gzclearerr +# define gzclose z_gzclose +# define gzclose_r z_gzclose_r +# define gzclose_w z_gzclose_w +# define gzdirect z_gzdirect +# define gzdopen z_gzdopen +# define gzeof z_gzeof +# define gzerror z_gzerror +# define gzflush z_gzflush +# define gzgetc z_gzgetc +# define gzgetc_ z_gzgetc_ +# define gzgets z_gzgets +# define gzoffset z_gzoffset +# define gzoffset64 z_gzoffset64 +# define gzopen z_gzopen +# define gzopen64 z_gzopen64 +# ifdef _WIN32 +# define gzopen_w z_gzopen_w +# endif +# define gzprintf z_gzprintf +# define gzvprintf z_gzvprintf +# define gzputc z_gzputc +# define gzputs z_gzputs +# define gzread z_gzread +# define gzrewind z_gzrewind +# define gzseek z_gzseek +# define gzseek64 z_gzseek64 +# define gzsetparams z_gzsetparams +# define gztell z_gztell +# define gztell64 z_gztell64 +# define gzungetc z_gzungetc +# define gzwrite z_gzwrite +# endif +# define inflate z_inflate +# define inflateBack z_inflateBack +# define inflateBackEnd z_inflateBackEnd +# define inflateBackInit_ z_inflateBackInit_ +# define inflateCopy z_inflateCopy +# define inflateEnd z_inflateEnd +# define inflateGetHeader z_inflateGetHeader +# define inflateInit2_ z_inflateInit2_ +# define inflateInit_ z_inflateInit_ +# define inflateMark z_inflateMark +# define inflatePrime z_inflatePrime +# define inflateReset z_inflateReset +# define inflateReset2 z_inflateReset2 +# define inflateSetDictionary z_inflateSetDictionary +# define inflateGetDictionary z_inflateGetDictionary +# define inflateSync z_inflateSync +# define inflateSyncPoint z_inflateSyncPoint +# define inflateUndermine z_inflateUndermine +# define inflateResetKeep z_inflateResetKeep +# define inflate_copyright z_inflate_copyright +# define inflate_fast z_inflate_fast +# define inflate_table z_inflate_table +# ifndef Z_SOLO +# define uncompress z_uncompress +# endif +# define zError z_zError +# ifndef Z_SOLO +# define zcalloc z_zcalloc +# define zcfree z_zcfree +# endif +# define zlibCompileFlags z_zlibCompileFlags +# define zlibVersion z_zlibVersion + +/* all zlib typedefs in zlib.h and zconf.h */ +# define Byte z_Byte +# define Bytef z_Bytef +# define alloc_func z_alloc_func +# define charf z_charf +# define free_func z_free_func +# ifndef Z_SOLO +# define gzFile z_gzFile +# endif +# define gz_header z_gz_header +# define gz_headerp z_gz_headerp +# define in_func z_in_func +# define intf z_intf +# define out_func z_out_func +# define uInt z_uInt +# define uIntf z_uIntf +# define uLong z_uLong +# define uLongf z_uLongf +# define voidp z_voidp +# define voidpc z_voidpc +# define voidpf z_voidpf + +/* all zlib structs in zlib.h and zconf.h */ +# define gz_header_s z_gz_header_s +# define internal_state z_internal_state + +#endif + +#if defined(__MSDOS__) && !defined(MSDOS) +# define MSDOS +#endif +#if (defined(OS_2) || defined(__OS2__)) && !defined(OS2) +# define OS2 +#endif +#if defined(_WINDOWS) && !defined(WINDOWS) +# define WINDOWS +#endif +#if defined(_WIN32) || defined(_WIN32_WCE) || defined(__WIN32__) +# ifndef WIN32 +# define WIN32 +# endif +#endif +#if (defined(MSDOS) || defined(OS2) || defined(WINDOWS)) && !defined(WIN32) +# if !defined(__GNUC__) && !defined(__FLAT__) && !defined(__386__) +# ifndef SYS16BIT +# define SYS16BIT +# endif +# endif +#endif + +/* + * Compile with -DMAXSEG_64K if the alloc function cannot allocate more + * than 64k bytes at a time (needed on systems with 16-bit int). + */ +#ifdef SYS16BIT +# define MAXSEG_64K +#endif +#ifdef MSDOS +# define UNALIGNED_OK +#endif + +#ifdef __STDC_VERSION__ +# ifndef STDC +# define STDC +# endif +# if __STDC_VERSION__ >= 199901L +# ifndef STDC99 +# define STDC99 +# endif +# endif +#endif +#if !defined(STDC) && (defined(__STDC__) || defined(__cplusplus)) +# define STDC +#endif +#if !defined(STDC) && (defined(__GNUC__) || defined(__BORLANDC__)) +# define STDC +#endif +#if !defined(STDC) && (defined(MSDOS) || defined(WINDOWS) || defined(WIN32)) +# define STDC +#endif +#if !defined(STDC) && (defined(OS2) || defined(__HOS_AIX__)) +# define STDC +#endif + +#if defined(__OS400__) && !defined(STDC) /* iSeries (formerly AS/400). */ +# define STDC +#endif + +#ifndef STDC +# ifndef const /* cannot use !defined(STDC) && !defined(const) on Mac */ +# define const /* note: need a more gentle solution here */ +# endif +#endif + +#if defined(ZLIB_CONST) && !defined(z_const) +# define z_const const +#else +# define z_const +#endif + +/* Some Mac compilers merge all .h files incorrectly: */ +#if defined(__MWERKS__)||defined(applec)||defined(THINK_C)||defined(__SC__) +# define NO_DUMMY_DECL +#endif + +/* Maximum value for memLevel in deflateInit2 */ +#ifndef MAX_MEM_LEVEL +# ifdef MAXSEG_64K +# define MAX_MEM_LEVEL 8 +# else +# define MAX_MEM_LEVEL 9 +# endif +#endif + +/* Maximum value for windowBits in deflateInit2 and inflateInit2. + * WARNING: reducing MAX_WBITS makes minigzip unable to extract .gz files + * created by gzip. (Files created by minigzip can still be extracted by + * gzip.) + */ +#ifndef MAX_WBITS +# define MAX_WBITS 15 /* 32K LZ77 window */ +#endif + +/* The memory requirements for deflate are (in bytes): + (1 << (windowBits+2)) + (1 << (memLevel+9)) + that is: 128K for windowBits=15 + 128K for memLevel = 8 (default values) + plus a few kilobytes for small objects. For example, if you want to reduce + the default memory requirements from 256K to 128K, compile with + make CFLAGS="-O -DMAX_WBITS=14 -DMAX_MEM_LEVEL=7" + Of course this will generally degrade compression (there's no free lunch). + + The memory requirements for inflate are (in bytes) 1 << windowBits + that is, 32K for windowBits=15 (default value) plus a few kilobytes + for small objects. +*/ + + /* Type declarations */ + +#ifndef OF /* function prototypes */ +# ifdef STDC +# define OF(args) args +# else +# define OF(args) () +# endif +#endif + +#ifndef Z_ARG /* function prototypes for stdarg */ +# if defined(STDC) || defined(Z_HAVE_STDARG_H) +# define Z_ARG(args) args +# else +# define Z_ARG(args) () +# endif +#endif + +/* The following definitions for FAR are needed only for MSDOS mixed + * model programming (small or medium model with some far allocations). + * This was tested only with MSC; for other MSDOS compilers you may have + * to define NO_MEMCPY in zutil.h. If you don't need the mixed model, + * just define FAR to be empty. + */ +#ifdef SYS16BIT +# if defined(M_I86SM) || defined(M_I86MM) + /* MSC small or medium model */ +# define SMALL_MEDIUM +# ifdef _MSC_VER +# define FAR _far +# else +# define FAR far +# endif +# endif +# if (defined(__SMALL__) || defined(__MEDIUM__)) + /* Turbo C small or medium model */ +# define SMALL_MEDIUM +# ifdef __BORLANDC__ +# define FAR _far +# else +# define FAR far +# endif +# endif +#endif + +#if defined(WINDOWS) || defined(WIN32) + /* If building or using zlib as a DLL, define ZLIB_DLL. + * This is not mandatory, but it offers a little performance increase. + */ +# ifdef ZLIB_DLL +# if defined(WIN32) && (!defined(__BORLANDC__) || (__BORLANDC__ >= 0x500)) +# ifdef ZLIB_INTERNAL +# define ZEXTERN extern __declspec(dllexport) +# else +# define ZEXTERN extern __declspec(dllimport) +# endif +# endif +# endif /* ZLIB_DLL */ + /* If building or using zlib with the WINAPI/WINAPIV calling convention, + * define ZLIB_WINAPI. + * Caution: the standard ZLIB1.DLL is NOT compiled using ZLIB_WINAPI. + */ +# ifdef ZLIB_WINAPI +# ifdef FAR +# undef FAR +# endif +# include + /* No need for _export, use ZLIB.DEF instead. */ + /* For complete Windows compatibility, use WINAPI, not __stdcall. */ +# define ZEXPORT WINAPI +# ifdef WIN32 +# define ZEXPORTVA WINAPIV +# else +# define ZEXPORTVA FAR CDECL +# endif +# endif +#endif + +#if defined (__BEOS__) +# ifdef ZLIB_DLL +# ifdef ZLIB_INTERNAL +# define ZEXPORT __declspec(dllexport) +# define ZEXPORTVA __declspec(dllexport) +# else +# define ZEXPORT __declspec(dllimport) +# define ZEXPORTVA __declspec(dllimport) +# endif +# endif +#endif + +#ifndef ZEXTERN +# define ZEXTERN extern +#endif +#ifndef ZEXPORT +# define ZEXPORT +#endif +#ifndef ZEXPORTVA +# define ZEXPORTVA +#endif + +#ifndef FAR +# define FAR +#endif + +#if !defined(__MACTYPES__) +typedef unsigned char Byte; /* 8 bits */ +#endif +typedef unsigned int uInt; /* 16 bits or more */ +typedef unsigned long uLong; /* 32 bits or more */ + +#ifdef SMALL_MEDIUM + /* Borland C/C++ and some old MSC versions ignore FAR inside typedef */ +# define Bytef Byte FAR +#else + typedef Byte FAR Bytef; +#endif +typedef char FAR charf; +typedef int FAR intf; +typedef uInt FAR uIntf; +typedef uLong FAR uLongf; + +#ifdef STDC + typedef void const *voidpc; + typedef void FAR *voidpf; + typedef void *voidp; +#else + typedef Byte const *voidpc; + typedef Byte FAR *voidpf; + typedef Byte *voidp; +#endif + +#if !defined(Z_U4) && !defined(Z_SOLO) && defined(STDC) +# include +# if (UINT_MAX == 0xffffffffUL) +# define Z_U4 unsigned +# elif (ULONG_MAX == 0xffffffffUL) +# define Z_U4 unsigned long +# elif (USHRT_MAX == 0xffffffffUL) +# define Z_U4 unsigned short +# endif +#endif + +#ifdef Z_U4 + typedef Z_U4 z_crc_t; +#else + typedef unsigned long z_crc_t; +#endif + +#if 1 /* was set to #if 1 by ./configure */ +# define Z_HAVE_UNISTD_H +#endif + +#if 1 /* was set to #if 1 by ./configure */ +# define Z_HAVE_STDARG_H +#endif + +#ifdef STDC +# ifndef Z_SOLO +# include /* for off_t */ +# endif +#endif + +#if defined(STDC) || defined(Z_HAVE_STDARG_H) +# ifndef Z_SOLO +# include /* for va_list */ +# endif +#endif + +#ifdef _WIN32 +# ifndef Z_SOLO +# include /* for wchar_t */ +# endif +#endif + +/* a little trick to accommodate both "#define _LARGEFILE64_SOURCE" and + * "#define _LARGEFILE64_SOURCE 1" as requesting 64-bit operations, (even + * though the former does not conform to the LFS document), but considering + * both "#undef _LARGEFILE64_SOURCE" and "#define _LARGEFILE64_SOURCE 0" as + * equivalently requesting no 64-bit operations + */ +#if defined(_LARGEFILE64_SOURCE) && -_LARGEFILE64_SOURCE - -1 == 1 +# undef _LARGEFILE64_SOURCE +#endif + +#if defined(__WATCOMC__) && !defined(Z_HAVE_UNISTD_H) +# define Z_HAVE_UNISTD_H +#endif +#ifndef Z_SOLO +# if defined(Z_HAVE_UNISTD_H) || defined(_LARGEFILE64_SOURCE) +# include /* for SEEK_*, off_t, and _LFS64_LARGEFILE */ +# ifdef VMS +# include /* for off_t */ +# endif +# ifndef z_off_t +# define z_off_t off_t +# endif +# endif +#endif + +#if defined(_LFS64_LARGEFILE) && _LFS64_LARGEFILE-0 +# define Z_LFS64 +#endif + +#if defined(_LARGEFILE64_SOURCE) && defined(Z_LFS64) +# define Z_LARGE64 +#endif + +#if defined(_FILE_OFFSET_BITS) && _FILE_OFFSET_BITS-0 == 64 && defined(Z_LFS64) +# define Z_WANT64 +#endif + +#if !defined(SEEK_SET) && !defined(Z_SOLO) +# define SEEK_SET 0 /* Seek from beginning of file. */ +# define SEEK_CUR 1 /* Seek from current position. */ +# define SEEK_END 2 /* Set file pointer to EOF plus "offset" */ +#endif + +#ifndef z_off_t +# define z_off_t long +#endif + +#if !defined(_WIN32) && defined(Z_LARGE64) +# define z_off64_t off64_t +#else +# if defined(_WIN32) && !defined(__GNUC__) && !defined(Z_SOLO) +# define z_off64_t __int64 +# else +# define z_off64_t z_off_t +# endif +#endif + +/* MVS linker does not support external names larger than 8 bytes */ +#if defined(__MVS__) + #pragma map(deflateInit_,"DEIN") + #pragma map(deflateInit2_,"DEIN2") + #pragma map(deflateEnd,"DEEND") + #pragma map(deflateBound,"DEBND") + #pragma map(inflateInit_,"ININ") + #pragma map(inflateInit2_,"ININ2") + #pragma map(inflateEnd,"INEND") + #pragma map(inflateSync,"INSY") + #pragma map(inflateSetDictionary,"INSEDI") + #pragma map(compressBound,"CMBND") + #pragma map(inflate_table,"INTABL") + #pragma map(inflate_fast,"INFA") + #pragma map(inflate_copyright,"INCOPY") +#endif + +#endif /* ZCONF_H */ diff --git a/app/src/main/jni/zlib/zlib.h b/app/src/main/jni/zlib/zlib.h new file mode 100644 index 000000000..9b8c8c7ea --- /dev/null +++ b/app/src/main/jni/zlib/zlib.h @@ -0,0 +1,1760 @@ +/* zlib.h -- interface of the 'zlib' general purpose compression library + version 1.2.8, April 28th, 2013 + + Copyright (C) 1995-2013 Jean-loup Gailly and Mark Adler + + This software is provided 'as-is', without any express or implied + warranty. In no event will the authors be held liable for any damages + arising from the use of this software. + + Permission is granted to anyone to use this software for any purpose, + including commercial applications, and to alter it and redistribute it + freely, subject to the following restrictions: + + 1. The origin of this software must not be misrepresented; you must not + claim that you wrote the original software. If you use this software + in a product, an acknowledgment in the product documentation would be + appreciated but is not required. + 2. Altered source versions must be plainly marked as such, and must not be + misrepresented as being the original software. + 3. This notice may not be removed or altered from any source distribution. + + Jean-loup Gailly Mark Adler + jloup@gzip.org madler@alumni.caltech.edu + + + The data format used by the zlib library is described by RFCs (Request for + Comments) 1950 to 1952 in the files http://tools.ietf.org/html/rfc1950 + (zlib format), rfc1951 (deflate format) and rfc1952 (gzip format). +*/ + +#ifndef ZLIB_H +#define ZLIB_H + +#include "zconf.h" + +#define ZLIB_VERSION "1.2.8" +#define ZLIB_VERNUM 0x1280 +#define ZLIB_VER_MAJOR 1 +#define ZLIB_VER_MINOR 2 +#define ZLIB_VER_REVISION 8 +#define ZLIB_VER_SUBREVISION 0 + +/* + The 'zlib' compression library provides in-memory compression and + decompression functions, including integrity checks of the uncompressed data. + This version of the library supports only one compression method (deflation) + but other algorithms will be added later and will have the same stream + interface. + + Compression can be done in a single step if the buffers are large enough, + or can be done by repeated calls of the compression function. In the latter + case, the application must provide more input and/or consume the output + (providing more output space) before each call. + + The compressed data format used by default by the in-memory functions is + the zlib format, which is a zlib wrapper documented in RFC 1950, wrapped + around a deflate stream, which is itself documented in RFC 1951. + + The library also supports reading and writing files in gzip (.gz) format + with an interface similar to that of stdio using the functions that start + with "gz". The gzip format is different from the zlib format. gzip is a + gzip wrapper, documented in RFC 1952, wrapped around a deflate stream. + + This library can optionally read and write gzip streams in memory as well. + + The zlib format was designed to be compact and fast for use in memory + and on communications channels. The gzip format was designed for single- + file compression on file systems, has a larger header than zlib to maintain + directory information, and uses a different, slower check method than zlib. + + The library does not install any signal handler. The decoder checks + the consistency of the compressed data, so the library should never crash + even in case of corrupted input. +*/ + +typedef voidpf (*alloc_func) OF((voidpf opaque, uInt items, uInt size)); +typedef void (*free_func) OF((voidpf opaque, voidpf address)); + +struct internal_state; + +typedef struct z_stream_s { + z_const Bytef *next_in; /* next input byte */ + uInt avail_in; /* number of bytes available at next_in */ + uLong total_in; /* total number of input bytes read so far */ + + Bytef *next_out; /* next output byte should be put there */ + uInt avail_out; /* remaining free space at next_out */ + uLong total_out; /* total number of bytes output so far */ + + z_const char *msg; /* last error message, NULL if no error */ + struct internal_state FAR *state; /* not visible by applications */ + + alloc_func zalloc; /* used to allocate the internal state */ + free_func zfree; /* used to free the internal state */ + voidpf opaque; /* private data object passed to zalloc and zfree */ + + int data_type; /* best guess about the data type: binary or text */ + uLong adler; /* adler32 value of the uncompressed data */ + uLong reserved; /* reserved for future use */ +} z_stream; + +typedef z_stream FAR *z_streamp; + +/* + gzip header information passed to and from zlib routines. See RFC 1952 + for more details on the meanings of these fields. +*/ +typedef struct gz_header_s { + int text; /* true if compressed data believed to be text */ + uLong time; /* modification time */ + int xflags; /* extra flags (not used when writing a gzip file) */ + int os; /* operating system */ + Bytef *extra; /* pointer to extra field or Z_NULL if none */ + uInt extra_len; /* extra field length (valid if extra != Z_NULL) */ + uInt extra_max; /* space at extra (only when reading header) */ + Bytef *name; /* pointer to zero-terminated file name or Z_NULL */ + uInt name_max; /* space at name (only when reading header) */ + Bytef *comment; /* pointer to zero-terminated comment or Z_NULL */ + uInt comm_max; /* space at comment (only when reading header) */ + int hcrc; /* true if there was or will be a header crc */ + int done; /* true when done reading gzip header (not used + when writing a gzip file) */ +} gz_header; + +typedef gz_header FAR *gz_headerp; + +/* + The application must update next_in and avail_in when avail_in has dropped + to zero. It must update next_out and avail_out when avail_out has dropped + to zero. The application must initialize zalloc, zfree and opaque before + calling the init function. All other fields are set by the compression + library and must not be updated by the application. + + The opaque value provided by the application will be passed as the first + parameter for calls of zalloc and zfree. This can be useful for custom + memory management. The compression library attaches no meaning to the + opaque value. + + zalloc must return Z_NULL if there is not enough memory for the object. + If zlib is used in a multi-threaded application, zalloc and zfree must be + thread safe. + + On 16-bit systems, the functions zalloc and zfree must be able to allocate + exactly 65536 bytes, but will not be required to allocate more than this if + the symbol MAXSEG_64K is defined (see zconf.h). WARNING: On MSDOS, pointers + returned by zalloc for objects of exactly 65536 bytes *must* have their + offset normalized to zero. The default allocation function provided by this + library ensures this (see zutil.c). To reduce memory requirements and avoid + any allocation of 64K objects, at the expense of compression ratio, compile + the library with -DMAX_WBITS=14 (see zconf.h). + + The fields total_in and total_out can be used for statistics or progress + reports. After compression, total_in holds the total size of the + uncompressed data and may be saved for use in the decompressor (particularly + if the decompressor wants to decompress everything in a single step). +*/ + + /* constants */ + +#define Z_NO_FLUSH 0 +#define Z_PARTIAL_FLUSH 1 +#define Z_SYNC_FLUSH 2 +#define Z_FULL_FLUSH 3 +#define Z_FINISH 4 +#define Z_BLOCK 5 +#define Z_TREES 6 +/* Allowed flush values; see deflate() and inflate() below for details */ + +#define Z_OK 0 +#define Z_STREAM_END 1 +#define Z_NEED_DICT 2 +#define Z_ERRNO (-1) +#define Z_STREAM_ERROR (-2) +#define Z_DATA_ERROR (-3) +#define Z_MEM_ERROR (-4) +#define Z_BUF_ERROR (-5) +#define Z_VERSION_ERROR (-6) +/* Return codes for the compression/decompression functions. Negative values + * are errors, positive values are used for special but normal events. + */ + +#define Z_NO_COMPRESSION 0 +#define Z_BEST_SPEED 1 +#define Z_BEST_COMPRESSION 9 +#define Z_DEFAULT_COMPRESSION (-1) +/* compression levels */ + +#define Z_FILTERED 1 +#define Z_HUFFMAN_ONLY 2 +#define Z_RLE 3 +#define Z_FIXED 4 +#define Z_DEFAULT_STRATEGY 0 +/* compression strategy; see deflateInit2() below for details */ + +#define Z_BINARY 0 +#define Z_TEXT 1 +#define Z_ASCII Z_TEXT /* for compatibility with 1.2.2 and earlier */ +#define Z_UNKNOWN 2 +/* Possible values of the data_type field (though see inflate()) */ + +#define Z_DEFLATED 8 +/* The deflate compression method (the only one supported in this version) */ + +#define Z_NULL 0 /* for initializing zalloc, zfree, opaque */ + +#define zlib_version zlibVersion() +/* for compatibility with versions < 1.0.2 */ + + + /* basic functions */ + +ZEXTERN const char * ZEXPORT zlibVersion OF((void)); +/* The application can compare zlibVersion and ZLIB_VERSION for consistency. + If the first character differs, the library code actually used is not + compatible with the zlib.h header file used by the application. This check + is automatically made by deflateInit and inflateInit. + */ + +/* +ZEXTERN int ZEXPORT deflateInit OF((z_streamp strm, int level)); + + Initializes the internal stream state for compression. The fields + zalloc, zfree and opaque must be initialized before by the caller. If + zalloc and zfree are set to Z_NULL, deflateInit updates them to use default + allocation functions. + + The compression level must be Z_DEFAULT_COMPRESSION, or between 0 and 9: + 1 gives best speed, 9 gives best compression, 0 gives no compression at all + (the input data is simply copied a block at a time). Z_DEFAULT_COMPRESSION + requests a default compromise between speed and compression (currently + equivalent to level 6). + + deflateInit returns Z_OK if success, Z_MEM_ERROR if there was not enough + memory, Z_STREAM_ERROR if level is not a valid compression level, or + Z_VERSION_ERROR if the zlib library version (zlib_version) is incompatible + with the version assumed by the caller (ZLIB_VERSION). msg is set to null + if there is no error message. deflateInit does not perform any compression: + this will be done by deflate(). +*/ + + +ZEXTERN int ZEXPORT deflate OF((z_streamp strm, int flush)); +/* + deflate compresses as much data as possible, and stops when the input + buffer becomes empty or the output buffer becomes full. It may introduce + some output latency (reading input without producing any output) except when + forced to flush. + + The detailed semantics are as follows. deflate performs one or both of the + following actions: + + - Compress more input starting at next_in and update next_in and avail_in + accordingly. If not all input can be processed (because there is not + enough room in the output buffer), next_in and avail_in are updated and + processing will resume at this point for the next call of deflate(). + + - Provide more output starting at next_out and update next_out and avail_out + accordingly. This action is forced if the parameter flush is non zero. + Forcing flush frequently degrades the compression ratio, so this parameter + should be set only when necessary (in interactive applications). Some + output may be provided even if flush is not set. + + Before the call of deflate(), the application should ensure that at least + one of the actions is possible, by providing more input and/or consuming more + output, and updating avail_in or avail_out accordingly; avail_out should + never be zero before the call. The application can consume the compressed + output when it wants, for example when the output buffer is full (avail_out + == 0), or after each call of deflate(). If deflate returns Z_OK and with + zero avail_out, it must be called again after making room in the output + buffer because there might be more output pending. + + Normally the parameter flush is set to Z_NO_FLUSH, which allows deflate to + decide how much data to accumulate before producing output, in order to + maximize compression. + + If the parameter flush is set to Z_SYNC_FLUSH, all pending output is + flushed to the output buffer and the output is aligned on a byte boundary, so + that the decompressor can get all input data available so far. (In + particular avail_in is zero after the call if enough output space has been + provided before the call.) Flushing may degrade compression for some + compression algorithms and so it should be used only when necessary. This + completes the current deflate block and follows it with an empty stored block + that is three bits plus filler bits to the next byte, followed by four bytes + (00 00 ff ff). + + If flush is set to Z_PARTIAL_FLUSH, all pending output is flushed to the + output buffer, but the output is not aligned to a byte boundary. All of the + input data so far will be available to the decompressor, as for Z_SYNC_FLUSH. + This completes the current deflate block and follows it with an empty fixed + codes block that is 10 bits long. This assures that enough bytes are output + in order for the decompressor to finish the block before the empty fixed code + block. + + If flush is set to Z_BLOCK, a deflate block is completed and emitted, as + for Z_SYNC_FLUSH, but the output is not aligned on a byte boundary, and up to + seven bits of the current block are held to be written as the next byte after + the next deflate block is completed. In this case, the decompressor may not + be provided enough bits at this point in order to complete decompression of + the data provided so far to the compressor. It may need to wait for the next + block to be emitted. This is for advanced applications that need to control + the emission of deflate blocks. + + If flush is set to Z_FULL_FLUSH, all output is flushed as with + Z_SYNC_FLUSH, and the compression state is reset so that decompression can + restart from this point if previous compressed data has been damaged or if + random access is desired. Using Z_FULL_FLUSH too often can seriously degrade + compression. + + If deflate returns with avail_out == 0, this function must be called again + with the same value of the flush parameter and more output space (updated + avail_out), until the flush is complete (deflate returns with non-zero + avail_out). In the case of a Z_FULL_FLUSH or Z_SYNC_FLUSH, make sure that + avail_out is greater than six to avoid repeated flush markers due to + avail_out == 0 on return. + + If the parameter flush is set to Z_FINISH, pending input is processed, + pending output is flushed and deflate returns with Z_STREAM_END if there was + enough output space; if deflate returns with Z_OK, this function must be + called again with Z_FINISH and more output space (updated avail_out) but no + more input data, until it returns with Z_STREAM_END or an error. After + deflate has returned Z_STREAM_END, the only possible operations on the stream + are deflateReset or deflateEnd. + + Z_FINISH can be used immediately after deflateInit if all the compression + is to be done in a single step. In this case, avail_out must be at least the + value returned by deflateBound (see below). Then deflate is guaranteed to + return Z_STREAM_END. If not enough output space is provided, deflate will + not return Z_STREAM_END, and it must be called again as described above. + + deflate() sets strm->adler to the adler32 checksum of all input read + so far (that is, total_in bytes). + + deflate() may update strm->data_type if it can make a good guess about + the input data type (Z_BINARY or Z_TEXT). In doubt, the data is considered + binary. This field is only for information purposes and does not affect the + compression algorithm in any manner. + + deflate() returns Z_OK if some progress has been made (more input + processed or more output produced), Z_STREAM_END if all input has been + consumed and all output has been produced (only when flush is set to + Z_FINISH), Z_STREAM_ERROR if the stream state was inconsistent (for example + if next_in or next_out was Z_NULL), Z_BUF_ERROR if no progress is possible + (for example avail_in or avail_out was zero). Note that Z_BUF_ERROR is not + fatal, and deflate() can be called again with more input and more output + space to continue compressing. +*/ + + +ZEXTERN int ZEXPORT deflateEnd OF((z_streamp strm)); +/* + All dynamically allocated data structures for this stream are freed. + This function discards any unprocessed input and does not flush any pending + output. + + deflateEnd returns Z_OK if success, Z_STREAM_ERROR if the + stream state was inconsistent, Z_DATA_ERROR if the stream was freed + prematurely (some input or output was discarded). In the error case, msg + may be set but then points to a static string (which must not be + deallocated). +*/ + + +/* +ZEXTERN int ZEXPORT inflateInit OF((z_streamp strm)); + + Initializes the internal stream state for decompression. The fields + next_in, avail_in, zalloc, zfree and opaque must be initialized before by + the caller. If next_in is not Z_NULL and avail_in is large enough (the + exact value depends on the compression method), inflateInit determines the + compression method from the zlib header and allocates all data structures + accordingly; otherwise the allocation will be deferred to the first call of + inflate. If zalloc and zfree are set to Z_NULL, inflateInit updates them to + use default allocation functions. + + inflateInit returns Z_OK if success, Z_MEM_ERROR if there was not enough + memory, Z_VERSION_ERROR if the zlib library version is incompatible with the + version assumed by the caller, or Z_STREAM_ERROR if the parameters are + invalid, such as a null pointer to the structure. msg is set to null if + there is no error message. inflateInit does not perform any decompression + apart from possibly reading the zlib header if present: actual decompression + will be done by inflate(). (So next_in and avail_in may be modified, but + next_out and avail_out are unused and unchanged.) The current implementation + of inflateInit() does not process any header information -- that is deferred + until inflate() is called. +*/ + + +ZEXTERN int ZEXPORT inflate OF((z_streamp strm, int flush)); +/* + inflate decompresses as much data as possible, and stops when the input + buffer becomes empty or the output buffer becomes full. It may introduce + some output latency (reading input without producing any output) except when + forced to flush. + + The detailed semantics are as follows. inflate performs one or both of the + following actions: + + - Decompress more input starting at next_in and update next_in and avail_in + accordingly. If not all input can be processed (because there is not + enough room in the output buffer), next_in is updated and processing will + resume at this point for the next call of inflate(). + + - Provide more output starting at next_out and update next_out and avail_out + accordingly. inflate() provides as much output as possible, until there is + no more input data or no more space in the output buffer (see below about + the flush parameter). + + Before the call of inflate(), the application should ensure that at least + one of the actions is possible, by providing more input and/or consuming more + output, and updating the next_* and avail_* values accordingly. The + application can consume the uncompressed output when it wants, for example + when the output buffer is full (avail_out == 0), or after each call of + inflate(). If inflate returns Z_OK and with zero avail_out, it must be + called again after making room in the output buffer because there might be + more output pending. + + The flush parameter of inflate() can be Z_NO_FLUSH, Z_SYNC_FLUSH, Z_FINISH, + Z_BLOCK, or Z_TREES. Z_SYNC_FLUSH requests that inflate() flush as much + output as possible to the output buffer. Z_BLOCK requests that inflate() + stop if and when it gets to the next deflate block boundary. When decoding + the zlib or gzip format, this will cause inflate() to return immediately + after the header and before the first block. When doing a raw inflate, + inflate() will go ahead and process the first block, and will return when it + gets to the end of that block, or when it runs out of data. + + The Z_BLOCK option assists in appending to or combining deflate streams. + Also to assist in this, on return inflate() will set strm->data_type to the + number of unused bits in the last byte taken from strm->next_in, plus 64 if + inflate() is currently decoding the last block in the deflate stream, plus + 128 if inflate() returned immediately after decoding an end-of-block code or + decoding the complete header up to just before the first byte of the deflate + stream. The end-of-block will not be indicated until all of the uncompressed + data from that block has been written to strm->next_out. The number of + unused bits may in general be greater than seven, except when bit 7 of + data_type is set, in which case the number of unused bits will be less than + eight. data_type is set as noted here every time inflate() returns for all + flush options, and so can be used to determine the amount of currently + consumed input in bits. + + The Z_TREES option behaves as Z_BLOCK does, but it also returns when the + end of each deflate block header is reached, before any actual data in that + block is decoded. This allows the caller to determine the length of the + deflate block header for later use in random access within a deflate block. + 256 is added to the value of strm->data_type when inflate() returns + immediately after reaching the end of the deflate block header. + + inflate() should normally be called until it returns Z_STREAM_END or an + error. However if all decompression is to be performed in a single step (a + single call of inflate), the parameter flush should be set to Z_FINISH. In + this case all pending input is processed and all pending output is flushed; + avail_out must be large enough to hold all of the uncompressed data for the + operation to complete. (The size of the uncompressed data may have been + saved by the compressor for this purpose.) The use of Z_FINISH is not + required to perform an inflation in one step. However it may be used to + inform inflate that a faster approach can be used for the single inflate() + call. Z_FINISH also informs inflate to not maintain a sliding window if the + stream completes, which reduces inflate's memory footprint. If the stream + does not complete, either because not all of the stream is provided or not + enough output space is provided, then a sliding window will be allocated and + inflate() can be called again to continue the operation as if Z_NO_FLUSH had + been used. + + In this implementation, inflate() always flushes as much output as + possible to the output buffer, and always uses the faster approach on the + first call. So the effects of the flush parameter in this implementation are + on the return value of inflate() as noted below, when inflate() returns early + when Z_BLOCK or Z_TREES is used, and when inflate() avoids the allocation of + memory for a sliding window when Z_FINISH is used. + + If a preset dictionary is needed after this call (see inflateSetDictionary + below), inflate sets strm->adler to the Adler-32 checksum of the dictionary + chosen by the compressor and returns Z_NEED_DICT; otherwise it sets + strm->adler to the Adler-32 checksum of all output produced so far (that is, + total_out bytes) and returns Z_OK, Z_STREAM_END or an error code as described + below. At the end of the stream, inflate() checks that its computed adler32 + checksum is equal to that saved by the compressor and returns Z_STREAM_END + only if the checksum is correct. + + inflate() can decompress and check either zlib-wrapped or gzip-wrapped + deflate data. The header type is detected automatically, if requested when + initializing with inflateInit2(). Any information contained in the gzip + header is not retained, so applications that need that information should + instead use raw inflate, see inflateInit2() below, or inflateBack() and + perform their own processing of the gzip header and trailer. When processing + gzip-wrapped deflate data, strm->adler32 is set to the CRC-32 of the output + producted so far. The CRC-32 is checked against the gzip trailer. + + inflate() returns Z_OK if some progress has been made (more input processed + or more output produced), Z_STREAM_END if the end of the compressed data has + been reached and all uncompressed output has been produced, Z_NEED_DICT if a + preset dictionary is needed at this point, Z_DATA_ERROR if the input data was + corrupted (input stream not conforming to the zlib format or incorrect check + value), Z_STREAM_ERROR if the stream structure was inconsistent (for example + next_in or next_out was Z_NULL), Z_MEM_ERROR if there was not enough memory, + Z_BUF_ERROR if no progress is possible or if there was not enough room in the + output buffer when Z_FINISH is used. Note that Z_BUF_ERROR is not fatal, and + inflate() can be called again with more input and more output space to + continue decompressing. If Z_DATA_ERROR is returned, the application may + then call inflateSync() to look for a good compression block if a partial + recovery of the data is desired. +*/ + + +ZEXTERN int ZEXPORT inflateEnd OF((z_streamp strm)); +/* + All dynamically allocated data structures for this stream are freed. + This function discards any unprocessed input and does not flush any pending + output. + + inflateEnd returns Z_OK if success, Z_STREAM_ERROR if the stream state + was inconsistent. In the error case, msg may be set but then points to a + static string (which must not be deallocated). +*/ + + + /* Advanced functions */ + +/* + The following functions are needed only in some special applications. +*/ + +/* +ZEXTERN int ZEXPORT deflateInit2 OF((z_streamp strm, + int level, + int method, + int windowBits, + int memLevel, + int strategy)); + + This is another version of deflateInit with more compression options. The + fields next_in, zalloc, zfree and opaque must be initialized before by the + caller. + + The method parameter is the compression method. It must be Z_DEFLATED in + this version of the library. + + The windowBits parameter is the base two logarithm of the window size + (the size of the history buffer). It should be in the range 8..15 for this + version of the library. Larger values of this parameter result in better + compression at the expense of memory usage. The default value is 15 if + deflateInit is used instead. + + windowBits can also be -8..-15 for raw deflate. In this case, -windowBits + determines the window size. deflate() will then generate raw deflate data + with no zlib header or trailer, and will not compute an adler32 check value. + + windowBits can also be greater than 15 for optional gzip encoding. Add + 16 to windowBits to write a simple gzip header and trailer around the + compressed data instead of a zlib wrapper. The gzip header will have no + file name, no extra data, no comment, no modification time (set to zero), no + header crc, and the operating system will be set to 255 (unknown). If a + gzip stream is being written, strm->adler is a crc32 instead of an adler32. + + The memLevel parameter specifies how much memory should be allocated + for the internal compression state. memLevel=1 uses minimum memory but is + slow and reduces compression ratio; memLevel=9 uses maximum memory for + optimal speed. The default value is 8. See zconf.h for total memory usage + as a function of windowBits and memLevel. + + The strategy parameter is used to tune the compression algorithm. Use the + value Z_DEFAULT_STRATEGY for normal data, Z_FILTERED for data produced by a + filter (or predictor), Z_HUFFMAN_ONLY to force Huffman encoding only (no + string match), or Z_RLE to limit match distances to one (run-length + encoding). Filtered data consists mostly of small values with a somewhat + random distribution. In this case, the compression algorithm is tuned to + compress them better. The effect of Z_FILTERED is to force more Huffman + coding and less string matching; it is somewhat intermediate between + Z_DEFAULT_STRATEGY and Z_HUFFMAN_ONLY. Z_RLE is designed to be almost as + fast as Z_HUFFMAN_ONLY, but give better compression for PNG image data. The + strategy parameter only affects the compression ratio but not the + correctness of the compressed output even if it is not set appropriately. + Z_FIXED prevents the use of dynamic Huffman codes, allowing for a simpler + decoder for special applications. + + deflateInit2 returns Z_OK if success, Z_MEM_ERROR if there was not enough + memory, Z_STREAM_ERROR if any parameter is invalid (such as an invalid + method), or Z_VERSION_ERROR if the zlib library version (zlib_version) is + incompatible with the version assumed by the caller (ZLIB_VERSION). msg is + set to null if there is no error message. deflateInit2 does not perform any + compression: this will be done by deflate(). +*/ + +ZEXTERN int ZEXPORT deflateSetDictionary OF((z_streamp strm, + const Bytef *dictionary, + uInt dictLength)); +/* + Initializes the compression dictionary from the given byte sequence + without producing any compressed output. When using the zlib format, this + function must be called immediately after deflateInit, deflateInit2 or + deflateReset, and before any call of deflate. When doing raw deflate, this + function must be called either before any call of deflate, or immediately + after the completion of a deflate block, i.e. after all input has been + consumed and all output has been delivered when using any of the flush + options Z_BLOCK, Z_PARTIAL_FLUSH, Z_SYNC_FLUSH, or Z_FULL_FLUSH. The + compressor and decompressor must use exactly the same dictionary (see + inflateSetDictionary). + + The dictionary should consist of strings (byte sequences) that are likely + to be encountered later in the data to be compressed, with the most commonly + used strings preferably put towards the end of the dictionary. Using a + dictionary is most useful when the data to be compressed is short and can be + predicted with good accuracy; the data can then be compressed better than + with the default empty dictionary. + + Depending on the size of the compression data structures selected by + deflateInit or deflateInit2, a part of the dictionary may in effect be + discarded, for example if the dictionary is larger than the window size + provided in deflateInit or deflateInit2. Thus the strings most likely to be + useful should be put at the end of the dictionary, not at the front. In + addition, the current implementation of deflate will use at most the window + size minus 262 bytes of the provided dictionary. + + Upon return of this function, strm->adler is set to the adler32 value + of the dictionary; the decompressor may later use this value to determine + which dictionary has been used by the compressor. (The adler32 value + applies to the whole dictionary even if only a subset of the dictionary is + actually used by the compressor.) If a raw deflate was requested, then the + adler32 value is not computed and strm->adler is not set. + + deflateSetDictionary returns Z_OK if success, or Z_STREAM_ERROR if a + parameter is invalid (e.g. dictionary being Z_NULL) or the stream state is + inconsistent (for example if deflate has already been called for this stream + or if not at a block boundary for raw deflate). deflateSetDictionary does + not perform any compression: this will be done by deflate(). +*/ + +ZEXTERN int ZEXPORT deflateCopy OF((z_streamp dest, + z_streamp source)); +/* + Sets the destination stream as a complete copy of the source stream. + + This function can be useful when several compression strategies will be + tried, for example when there are several ways of pre-processing the input + data with a filter. The streams that will be discarded should then be freed + by calling deflateEnd. Note that deflateCopy duplicates the internal + compression state which can be quite large, so this strategy is slow and can + consume lots of memory. + + deflateCopy returns Z_OK if success, Z_MEM_ERROR if there was not + enough memory, Z_STREAM_ERROR if the source stream state was inconsistent + (such as zalloc being Z_NULL). msg is left unchanged in both source and + destination. +*/ + +ZEXTERN int ZEXPORT deflateReset OF((z_streamp strm)); +/* + This function is equivalent to deflateEnd followed by deflateInit, + but does not free and reallocate all the internal compression state. The + stream will keep the same compression level and any other attributes that + may have been set by deflateInit2. + + deflateReset returns Z_OK if success, or Z_STREAM_ERROR if the source + stream state was inconsistent (such as zalloc or state being Z_NULL). +*/ + +ZEXTERN int ZEXPORT deflateParams OF((z_streamp strm, + int level, + int strategy)); +/* + Dynamically update the compression level and compression strategy. The + interpretation of level and strategy is as in deflateInit2. This can be + used to switch between compression and straight copy of the input data, or + to switch to a different kind of input data requiring a different strategy. + If the compression level is changed, the input available so far is + compressed with the old level (and may be flushed); the new level will take + effect only at the next call of deflate(). + + Before the call of deflateParams, the stream state must be set as for + a call of deflate(), since the currently available input may have to be + compressed and flushed. In particular, strm->avail_out must be non-zero. + + deflateParams returns Z_OK if success, Z_STREAM_ERROR if the source + stream state was inconsistent or if a parameter was invalid, Z_BUF_ERROR if + strm->avail_out was zero. +*/ + +ZEXTERN int ZEXPORT deflateTune OF((z_streamp strm, + int good_length, + int max_lazy, + int nice_length, + int max_chain)); +/* + Fine tune deflate's internal compression parameters. This should only be + used by someone who understands the algorithm used by zlib's deflate for + searching for the best matching string, and even then only by the most + fanatic optimizer trying to squeeze out the last compressed bit for their + specific input data. Read the deflate.c source code for the meaning of the + max_lazy, good_length, nice_length, and max_chain parameters. + + deflateTune() can be called after deflateInit() or deflateInit2(), and + returns Z_OK on success, or Z_STREAM_ERROR for an invalid deflate stream. + */ + +ZEXTERN uLong ZEXPORT deflateBound OF((z_streamp strm, + uLong sourceLen)); +/* + deflateBound() returns an upper bound on the compressed size after + deflation of sourceLen bytes. It must be called after deflateInit() or + deflateInit2(), and after deflateSetHeader(), if used. This would be used + to allocate an output buffer for deflation in a single pass, and so would be + called before deflate(). If that first deflate() call is provided the + sourceLen input bytes, an output buffer allocated to the size returned by + deflateBound(), and the flush value Z_FINISH, then deflate() is guaranteed + to return Z_STREAM_END. Note that it is possible for the compressed size to + be larger than the value returned by deflateBound() if flush options other + than Z_FINISH or Z_NO_FLUSH are used. +*/ + +ZEXTERN int ZEXPORT deflatePending OF((z_streamp strm, + unsigned *pending, + int *bits)); +/* + deflatePending() returns the number of bytes and bits of output that have + been generated, but not yet provided in the available output. The bytes not + provided would be due to the available output space having being consumed. + The number of bits of output not provided are between 0 and 7, where they + await more bits to join them in order to fill out a full byte. If pending + or bits are Z_NULL, then those values are not set. + + deflatePending returns Z_OK if success, or Z_STREAM_ERROR if the source + stream state was inconsistent. + */ + +ZEXTERN int ZEXPORT deflatePrime OF((z_streamp strm, + int bits, + int value)); +/* + deflatePrime() inserts bits in the deflate output stream. The intent + is that this function is used to start off the deflate output with the bits + leftover from a previous deflate stream when appending to it. As such, this + function can only be used for raw deflate, and must be used before the first + deflate() call after a deflateInit2() or deflateReset(). bits must be less + than or equal to 16, and that many of the least significant bits of value + will be inserted in the output. + + deflatePrime returns Z_OK if success, Z_BUF_ERROR if there was not enough + room in the internal buffer to insert the bits, or Z_STREAM_ERROR if the + source stream state was inconsistent. +*/ + +ZEXTERN int ZEXPORT deflateSetHeader OF((z_streamp strm, + gz_headerp head)); +/* + deflateSetHeader() provides gzip header information for when a gzip + stream is requested by deflateInit2(). deflateSetHeader() may be called + after deflateInit2() or deflateReset() and before the first call of + deflate(). The text, time, os, extra field, name, and comment information + in the provided gz_header structure are written to the gzip header (xflag is + ignored -- the extra flags are set according to the compression level). The + caller must assure that, if not Z_NULL, name and comment are terminated with + a zero byte, and that if extra is not Z_NULL, that extra_len bytes are + available there. If hcrc is true, a gzip header crc is included. Note that + the current versions of the command-line version of gzip (up through version + 1.3.x) do not support header crc's, and will report that it is a "multi-part + gzip file" and give up. + + If deflateSetHeader is not used, the default gzip header has text false, + the time set to zero, and os set to 255, with no extra, name, or comment + fields. The gzip header is returned to the default state by deflateReset(). + + deflateSetHeader returns Z_OK if success, or Z_STREAM_ERROR if the source + stream state was inconsistent. +*/ + +/* +ZEXTERN int ZEXPORT inflateInit2 OF((z_streamp strm, + int windowBits)); + + This is another version of inflateInit with an extra parameter. The + fields next_in, avail_in, zalloc, zfree and opaque must be initialized + before by the caller. + + The windowBits parameter is the base two logarithm of the maximum window + size (the size of the history buffer). It should be in the range 8..15 for + this version of the library. The default value is 15 if inflateInit is used + instead. windowBits must be greater than or equal to the windowBits value + provided to deflateInit2() while compressing, or it must be equal to 15 if + deflateInit2() was not used. If a compressed stream with a larger window + size is given as input, inflate() will return with the error code + Z_DATA_ERROR instead of trying to allocate a larger window. + + windowBits can also be zero to request that inflate use the window size in + the zlib header of the compressed stream. + + windowBits can also be -8..-15 for raw inflate. In this case, -windowBits + determines the window size. inflate() will then process raw deflate data, + not looking for a zlib or gzip header, not generating a check value, and not + looking for any check values for comparison at the end of the stream. This + is for use with other formats that use the deflate compressed data format + such as zip. Those formats provide their own check values. If a custom + format is developed using the raw deflate format for compressed data, it is + recommended that a check value such as an adler32 or a crc32 be applied to + the uncompressed data as is done in the zlib, gzip, and zip formats. For + most applications, the zlib format should be used as is. Note that comments + above on the use in deflateInit2() applies to the magnitude of windowBits. + + windowBits can also be greater than 15 for optional gzip decoding. Add + 32 to windowBits to enable zlib and gzip decoding with automatic header + detection, or add 16 to decode only the gzip format (the zlib format will + return a Z_DATA_ERROR). If a gzip stream is being decoded, strm->adler is a + crc32 instead of an adler32. + + inflateInit2 returns Z_OK if success, Z_MEM_ERROR if there was not enough + memory, Z_VERSION_ERROR if the zlib library version is incompatible with the + version assumed by the caller, or Z_STREAM_ERROR if the parameters are + invalid, such as a null pointer to the structure. msg is set to null if + there is no error message. inflateInit2 does not perform any decompression + apart from possibly reading the zlib header if present: actual decompression + will be done by inflate(). (So next_in and avail_in may be modified, but + next_out and avail_out are unused and unchanged.) The current implementation + of inflateInit2() does not process any header information -- that is + deferred until inflate() is called. +*/ + +ZEXTERN int ZEXPORT inflateSetDictionary OF((z_streamp strm, + const Bytef *dictionary, + uInt dictLength)); +/* + Initializes the decompression dictionary from the given uncompressed byte + sequence. This function must be called immediately after a call of inflate, + if that call returned Z_NEED_DICT. The dictionary chosen by the compressor + can be determined from the adler32 value returned by that call of inflate. + The compressor and decompressor must use exactly the same dictionary (see + deflateSetDictionary). For raw inflate, this function can be called at any + time to set the dictionary. If the provided dictionary is smaller than the + window and there is already data in the window, then the provided dictionary + will amend what's there. The application must insure that the dictionary + that was used for compression is provided. + + inflateSetDictionary returns Z_OK if success, Z_STREAM_ERROR if a + parameter is invalid (e.g. dictionary being Z_NULL) or the stream state is + inconsistent, Z_DATA_ERROR if the given dictionary doesn't match the + expected one (incorrect adler32 value). inflateSetDictionary does not + perform any decompression: this will be done by subsequent calls of + inflate(). +*/ + +ZEXTERN int ZEXPORT inflateGetDictionary OF((z_streamp strm, + Bytef *dictionary, + uInt *dictLength)); +/* + Returns the sliding dictionary being maintained by inflate. dictLength is + set to the number of bytes in the dictionary, and that many bytes are copied + to dictionary. dictionary must have enough space, where 32768 bytes is + always enough. If inflateGetDictionary() is called with dictionary equal to + Z_NULL, then only the dictionary length is returned, and nothing is copied. + Similary, if dictLength is Z_NULL, then it is not set. + + inflateGetDictionary returns Z_OK on success, or Z_STREAM_ERROR if the + stream state is inconsistent. +*/ + +ZEXTERN int ZEXPORT inflateSync OF((z_streamp strm)); +/* + Skips invalid compressed data until a possible full flush point (see above + for the description of deflate with Z_FULL_FLUSH) can be found, or until all + available input is skipped. No output is provided. + + inflateSync searches for a 00 00 FF FF pattern in the compressed data. + All full flush points have this pattern, but not all occurrences of this + pattern are full flush points. + + inflateSync returns Z_OK if a possible full flush point has been found, + Z_BUF_ERROR if no more input was provided, Z_DATA_ERROR if no flush point + has been found, or Z_STREAM_ERROR if the stream structure was inconsistent. + In the success case, the application may save the current current value of + total_in which indicates where valid compressed data was found. In the + error case, the application may repeatedly call inflateSync, providing more + input each time, until success or end of the input data. +*/ + +ZEXTERN int ZEXPORT inflateCopy OF((z_streamp dest, + z_streamp source)); +/* + Sets the destination stream as a complete copy of the source stream. + + This function can be useful when randomly accessing a large stream. The + first pass through the stream can periodically record the inflate state, + allowing restarting inflate at those points when randomly accessing the + stream. + + inflateCopy returns Z_OK if success, Z_MEM_ERROR if there was not + enough memory, Z_STREAM_ERROR if the source stream state was inconsistent + (such as zalloc being Z_NULL). msg is left unchanged in both source and + destination. +*/ + +ZEXTERN int ZEXPORT inflateReset OF((z_streamp strm)); +/* + This function is equivalent to inflateEnd followed by inflateInit, + but does not free and reallocate all the internal decompression state. The + stream will keep attributes that may have been set by inflateInit2. + + inflateReset returns Z_OK if success, or Z_STREAM_ERROR if the source + stream state was inconsistent (such as zalloc or state being Z_NULL). +*/ + +ZEXTERN int ZEXPORT inflateReset2 OF((z_streamp strm, + int windowBits)); +/* + This function is the same as inflateReset, but it also permits changing + the wrap and window size requests. The windowBits parameter is interpreted + the same as it is for inflateInit2. + + inflateReset2 returns Z_OK if success, or Z_STREAM_ERROR if the source + stream state was inconsistent (such as zalloc or state being Z_NULL), or if + the windowBits parameter is invalid. +*/ + +ZEXTERN int ZEXPORT inflatePrime OF((z_streamp strm, + int bits, + int value)); +/* + This function inserts bits in the inflate input stream. The intent is + that this function is used to start inflating at a bit position in the + middle of a byte. The provided bits will be used before any bytes are used + from next_in. This function should only be used with raw inflate, and + should be used before the first inflate() call after inflateInit2() or + inflateReset(). bits must be less than or equal to 16, and that many of the + least significant bits of value will be inserted in the input. + + If bits is negative, then the input stream bit buffer is emptied. Then + inflatePrime() can be called again to put bits in the buffer. This is used + to clear out bits leftover after feeding inflate a block description prior + to feeding inflate codes. + + inflatePrime returns Z_OK if success, or Z_STREAM_ERROR if the source + stream state was inconsistent. +*/ + +ZEXTERN long ZEXPORT inflateMark OF((z_streamp strm)); +/* + This function returns two values, one in the lower 16 bits of the return + value, and the other in the remaining upper bits, obtained by shifting the + return value down 16 bits. If the upper value is -1 and the lower value is + zero, then inflate() is currently decoding information outside of a block. + If the upper value is -1 and the lower value is non-zero, then inflate is in + the middle of a stored block, with the lower value equaling the number of + bytes from the input remaining to copy. If the upper value is not -1, then + it is the number of bits back from the current bit position in the input of + the code (literal or length/distance pair) currently being processed. In + that case the lower value is the number of bytes already emitted for that + code. + + A code is being processed if inflate is waiting for more input to complete + decoding of the code, or if it has completed decoding but is waiting for + more output space to write the literal or match data. + + inflateMark() is used to mark locations in the input data for random + access, which may be at bit positions, and to note those cases where the + output of a code may span boundaries of random access blocks. The current + location in the input stream can be determined from avail_in and data_type + as noted in the description for the Z_BLOCK flush parameter for inflate. + + inflateMark returns the value noted above or -1 << 16 if the provided + source stream state was inconsistent. +*/ + +ZEXTERN int ZEXPORT inflateGetHeader OF((z_streamp strm, + gz_headerp head)); +/* + inflateGetHeader() requests that gzip header information be stored in the + provided gz_header structure. inflateGetHeader() may be called after + inflateInit2() or inflateReset(), and before the first call of inflate(). + As inflate() processes the gzip stream, head->done is zero until the header + is completed, at which time head->done is set to one. If a zlib stream is + being decoded, then head->done is set to -1 to indicate that there will be + no gzip header information forthcoming. Note that Z_BLOCK or Z_TREES can be + used to force inflate() to return immediately after header processing is + complete and before any actual data is decompressed. + + The text, time, xflags, and os fields are filled in with the gzip header + contents. hcrc is set to true if there is a header CRC. (The header CRC + was valid if done is set to one.) If extra is not Z_NULL, then extra_max + contains the maximum number of bytes to write to extra. Once done is true, + extra_len contains the actual extra field length, and extra contains the + extra field, or that field truncated if extra_max is less than extra_len. + If name is not Z_NULL, then up to name_max characters are written there, + terminated with a zero unless the length is greater than name_max. If + comment is not Z_NULL, then up to comm_max characters are written there, + terminated with a zero unless the length is greater than comm_max. When any + of extra, name, or comment are not Z_NULL and the respective field is not + present in the header, then that field is set to Z_NULL to signal its + absence. This allows the use of deflateSetHeader() with the returned + structure to duplicate the header. However if those fields are set to + allocated memory, then the application will need to save those pointers + elsewhere so that they can be eventually freed. + + If inflateGetHeader is not used, then the header information is simply + discarded. The header is always checked for validity, including the header + CRC if present. inflateReset() will reset the process to discard the header + information. The application would need to call inflateGetHeader() again to + retrieve the header from the next gzip stream. + + inflateGetHeader returns Z_OK if success, or Z_STREAM_ERROR if the source + stream state was inconsistent. +*/ + +/* +ZEXTERN int ZEXPORT inflateBackInit OF((z_streamp strm, int windowBits, + unsigned char FAR *window)); + + Initialize the internal stream state for decompression using inflateBack() + calls. The fields zalloc, zfree and opaque in strm must be initialized + before the call. If zalloc and zfree are Z_NULL, then the default library- + derived memory allocation routines are used. windowBits is the base two + logarithm of the window size, in the range 8..15. window is a caller + supplied buffer of that size. Except for special applications where it is + assured that deflate was used with small window sizes, windowBits must be 15 + and a 32K byte window must be supplied to be able to decompress general + deflate streams. + + See inflateBack() for the usage of these routines. + + inflateBackInit will return Z_OK on success, Z_STREAM_ERROR if any of + the parameters are invalid, Z_MEM_ERROR if the internal state could not be + allocated, or Z_VERSION_ERROR if the version of the library does not match + the version of the header file. +*/ + +typedef unsigned (*in_func) OF((void FAR *, + z_const unsigned char FAR * FAR *)); +typedef int (*out_func) OF((void FAR *, unsigned char FAR *, unsigned)); + +ZEXTERN int ZEXPORT inflateBack OF((z_streamp strm, + in_func in, void FAR *in_desc, + out_func out, void FAR *out_desc)); +/* + inflateBack() does a raw inflate with a single call using a call-back + interface for input and output. This is potentially more efficient than + inflate() for file i/o applications, in that it avoids copying between the + output and the sliding window by simply making the window itself the output + buffer. inflate() can be faster on modern CPUs when used with large + buffers. inflateBack() trusts the application to not change the output + buffer passed by the output function, at least until inflateBack() returns. + + inflateBackInit() must be called first to allocate the internal state + and to initialize the state with the user-provided window buffer. + inflateBack() may then be used multiple times to inflate a complete, raw + deflate stream with each call. inflateBackEnd() is then called to free the + allocated state. + + A raw deflate stream is one with no zlib or gzip header or trailer. + This routine would normally be used in a utility that reads zip or gzip + files and writes out uncompressed files. The utility would decode the + header and process the trailer on its own, hence this routine expects only + the raw deflate stream to decompress. This is different from the normal + behavior of inflate(), which expects either a zlib or gzip header and + trailer around the deflate stream. + + inflateBack() uses two subroutines supplied by the caller that are then + called by inflateBack() for input and output. inflateBack() calls those + routines until it reads a complete deflate stream and writes out all of the + uncompressed data, or until it encounters an error. The function's + parameters and return types are defined above in the in_func and out_func + typedefs. inflateBack() will call in(in_desc, &buf) which should return the + number of bytes of provided input, and a pointer to that input in buf. If + there is no input available, in() must return zero--buf is ignored in that + case--and inflateBack() will return a buffer error. inflateBack() will call + out(out_desc, buf, len) to write the uncompressed data buf[0..len-1]. out() + should return zero on success, or non-zero on failure. If out() returns + non-zero, inflateBack() will return with an error. Neither in() nor out() + are permitted to change the contents of the window provided to + inflateBackInit(), which is also the buffer that out() uses to write from. + The length written by out() will be at most the window size. Any non-zero + amount of input may be provided by in(). + + For convenience, inflateBack() can be provided input on the first call by + setting strm->next_in and strm->avail_in. If that input is exhausted, then + in() will be called. Therefore strm->next_in must be initialized before + calling inflateBack(). If strm->next_in is Z_NULL, then in() will be called + immediately for input. If strm->next_in is not Z_NULL, then strm->avail_in + must also be initialized, and then if strm->avail_in is not zero, input will + initially be taken from strm->next_in[0 .. strm->avail_in - 1]. + + The in_desc and out_desc parameters of inflateBack() is passed as the + first parameter of in() and out() respectively when they are called. These + descriptors can be optionally used to pass any information that the caller- + supplied in() and out() functions need to do their job. + + On return, inflateBack() will set strm->next_in and strm->avail_in to + pass back any unused input that was provided by the last in() call. The + return values of inflateBack() can be Z_STREAM_END on success, Z_BUF_ERROR + if in() or out() returned an error, Z_DATA_ERROR if there was a format error + in the deflate stream (in which case strm->msg is set to indicate the nature + of the error), or Z_STREAM_ERROR if the stream was not properly initialized. + In the case of Z_BUF_ERROR, an input or output error can be distinguished + using strm->next_in which will be Z_NULL only if in() returned an error. If + strm->next_in is not Z_NULL, then the Z_BUF_ERROR was due to out() returning + non-zero. (in() will always be called before out(), so strm->next_in is + assured to be defined if out() returns non-zero.) Note that inflateBack() + cannot return Z_OK. +*/ + +ZEXTERN int ZEXPORT inflateBackEnd OF((z_streamp strm)); +/* + All memory allocated by inflateBackInit() is freed. + + inflateBackEnd() returns Z_OK on success, or Z_STREAM_ERROR if the stream + state was inconsistent. +*/ + +ZEXTERN uLong ZEXPORT zlibCompileFlags OF((void)); +/* Return flags indicating compile-time options. + + Type sizes, two bits each, 00 = 16 bits, 01 = 32, 10 = 64, 11 = other: + 1.0: size of uInt + 3.2: size of uLong + 5.4: size of voidpf (pointer) + 7.6: size of z_off_t + + Compiler, assembler, and debug options: + 8: DEBUG + 9: ASMV or ASMINF -- use ASM code + 10: ZLIB_WINAPI -- exported functions use the WINAPI calling convention + 11: 0 (reserved) + + One-time table building (smaller code, but not thread-safe if true): + 12: BUILDFIXED -- build static block decoding tables when needed + 13: DYNAMIC_CRC_TABLE -- build CRC calculation tables when needed + 14,15: 0 (reserved) + + Library content (indicates missing functionality): + 16: NO_GZCOMPRESS -- gz* functions cannot compress (to avoid linking + deflate code when not needed) + 17: NO_GZIP -- deflate can't write gzip streams, and inflate can't detect + and decode gzip streams (to avoid linking crc code) + 18-19: 0 (reserved) + + Operation variations (changes in library functionality): + 20: PKZIP_BUG_WORKAROUND -- slightly more permissive inflate + 21: FASTEST -- deflate algorithm with only one, lowest compression level + 22,23: 0 (reserved) + + The sprintf variant used by gzprintf (zero is best): + 24: 0 = vs*, 1 = s* -- 1 means limited to 20 arguments after the format + 25: 0 = *nprintf, 1 = *printf -- 1 means gzprintf() not secure! + 26: 0 = returns value, 1 = void -- 1 means inferred string length returned + + Remainder: + 27-31: 0 (reserved) + */ + +#ifndef Z_SOLO + + /* utility functions */ + +/* + The following utility functions are implemented on top of the basic + stream-oriented functions. To simplify the interface, some default options + are assumed (compression level and memory usage, standard memory allocation + functions). The source code of these utility functions can be modified if + you need special options. +*/ + +ZEXTERN int ZEXPORT compress OF((Bytef *dest, uLongf *destLen, + const Bytef *source, uLong sourceLen)); +/* + Compresses the source buffer into the destination buffer. sourceLen is + the byte length of the source buffer. Upon entry, destLen is the total size + of the destination buffer, which must be at least the value returned by + compressBound(sourceLen). Upon exit, destLen is the actual size of the + compressed buffer. + + compress returns Z_OK if success, Z_MEM_ERROR if there was not + enough memory, Z_BUF_ERROR if there was not enough room in the output + buffer. +*/ + +ZEXTERN int ZEXPORT compress2 OF((Bytef *dest, uLongf *destLen, + const Bytef *source, uLong sourceLen, + int level)); +/* + Compresses the source buffer into the destination buffer. The level + parameter has the same meaning as in deflateInit. sourceLen is the byte + length of the source buffer. Upon entry, destLen is the total size of the + destination buffer, which must be at least the value returned by + compressBound(sourceLen). Upon exit, destLen is the actual size of the + compressed buffer. + + compress2 returns Z_OK if success, Z_MEM_ERROR if there was not enough + memory, Z_BUF_ERROR if there was not enough room in the output buffer, + Z_STREAM_ERROR if the level parameter is invalid. +*/ + +ZEXTERN uLong ZEXPORT compressBound OF((uLong sourceLen)); +/* + compressBound() returns an upper bound on the compressed size after + compress() or compress2() on sourceLen bytes. It would be used before a + compress() or compress2() call to allocate the destination buffer. +*/ + +ZEXTERN int ZEXPORT uncompress OF((Bytef *dest, uLongf *destLen, + const Bytef *source, uLong sourceLen)); +/* + Decompresses the source buffer into the destination buffer. sourceLen is + the byte length of the source buffer. Upon entry, destLen is the total size + of the destination buffer, which must be large enough to hold the entire + uncompressed data. (The size of the uncompressed data must have been saved + previously by the compressor and transmitted to the decompressor by some + mechanism outside the scope of this compression library.) Upon exit, destLen + is the actual size of the uncompressed buffer. + + uncompress returns Z_OK if success, Z_MEM_ERROR if there was not + enough memory, Z_BUF_ERROR if there was not enough room in the output + buffer, or Z_DATA_ERROR if the input data was corrupted or incomplete. In + the case where there is not enough room, uncompress() will fill the output + buffer with the uncompressed data up to that point. +*/ + + /* gzip file access functions */ + +/* + This library supports reading and writing files in gzip (.gz) format with + an interface similar to that of stdio, using the functions that start with + "gz". The gzip format is different from the zlib format. gzip is a gzip + wrapper, documented in RFC 1952, wrapped around a deflate stream. +*/ + +typedef struct gzFile_s *gzFile; /* semi-opaque gzip file descriptor */ + +/* +ZEXTERN gzFile ZEXPORT gzopen OF((const char *path, const char *mode)); + + Opens a gzip (.gz) file for reading or writing. The mode parameter is as + in fopen ("rb" or "wb") but can also include a compression level ("wb9") or + a strategy: 'f' for filtered data as in "wb6f", 'h' for Huffman-only + compression as in "wb1h", 'R' for run-length encoding as in "wb1R", or 'F' + for fixed code compression as in "wb9F". (See the description of + deflateInit2 for more information about the strategy parameter.) 'T' will + request transparent writing or appending with no compression and not using + the gzip format. + + "a" can be used instead of "w" to request that the gzip stream that will + be written be appended to the file. "+" will result in an error, since + reading and writing to the same gzip file is not supported. The addition of + "x" when writing will create the file exclusively, which fails if the file + already exists. On systems that support it, the addition of "e" when + reading or writing will set the flag to close the file on an execve() call. + + These functions, as well as gzip, will read and decode a sequence of gzip + streams in a file. The append function of gzopen() can be used to create + such a file. (Also see gzflush() for another way to do this.) When + appending, gzopen does not test whether the file begins with a gzip stream, + nor does it look for the end of the gzip streams to begin appending. gzopen + will simply append a gzip stream to the existing file. + + gzopen can be used to read a file which is not in gzip format; in this + case gzread will directly read from the file without decompression. When + reading, this will be detected automatically by looking for the magic two- + byte gzip header. + + gzopen returns NULL if the file could not be opened, if there was + insufficient memory to allocate the gzFile state, or if an invalid mode was + specified (an 'r', 'w', or 'a' was not provided, or '+' was provided). + errno can be checked to determine if the reason gzopen failed was that the + file could not be opened. +*/ + +ZEXTERN gzFile ZEXPORT gzdopen OF((int fd, const char *mode)); +/* + gzdopen associates a gzFile with the file descriptor fd. File descriptors + are obtained from calls like open, dup, creat, pipe or fileno (if the file + has been previously opened with fopen). The mode parameter is as in gzopen. + + The next call of gzclose on the returned gzFile will also close the file + descriptor fd, just like fclose(fdopen(fd, mode)) closes the file descriptor + fd. If you want to keep fd open, use fd = dup(fd_keep); gz = gzdopen(fd, + mode);. The duplicated descriptor should be saved to avoid a leak, since + gzdopen does not close fd if it fails. If you are using fileno() to get the + file descriptor from a FILE *, then you will have to use dup() to avoid + double-close()ing the file descriptor. Both gzclose() and fclose() will + close the associated file descriptor, so they need to have different file + descriptors. + + gzdopen returns NULL if there was insufficient memory to allocate the + gzFile state, if an invalid mode was specified (an 'r', 'w', or 'a' was not + provided, or '+' was provided), or if fd is -1. The file descriptor is not + used until the next gz* read, write, seek, or close operation, so gzdopen + will not detect if fd is invalid (unless fd is -1). +*/ + +ZEXTERN int ZEXPORT gzbuffer OF((gzFile file, unsigned size)); +/* + Set the internal buffer size used by this library's functions. The + default buffer size is 8192 bytes. This function must be called after + gzopen() or gzdopen(), and before any other calls that read or write the + file. The buffer memory allocation is always deferred to the first read or + write. Two buffers are allocated, either both of the specified size when + writing, or one of the specified size and the other twice that size when + reading. A larger buffer size of, for example, 64K or 128K bytes will + noticeably increase the speed of decompression (reading). + + The new buffer size also affects the maximum length for gzprintf(). + + gzbuffer() returns 0 on success, or -1 on failure, such as being called + too late. +*/ + +ZEXTERN int ZEXPORT gzsetparams OF((gzFile file, int level, int strategy)); +/* + Dynamically update the compression level or strategy. See the description + of deflateInit2 for the meaning of these parameters. + + gzsetparams returns Z_OK if success, or Z_STREAM_ERROR if the file was not + opened for writing. +*/ + +ZEXTERN int ZEXPORT gzread OF((gzFile file, voidp buf, unsigned len)); +/* + Reads the given number of uncompressed bytes from the compressed file. If + the input file is not in gzip format, gzread copies the given number of + bytes into the buffer directly from the file. + + After reaching the end of a gzip stream in the input, gzread will continue + to read, looking for another gzip stream. Any number of gzip streams may be + concatenated in the input file, and will all be decompressed by gzread(). + If something other than a gzip stream is encountered after a gzip stream, + that remaining trailing garbage is ignored (and no error is returned). + + gzread can be used to read a gzip file that is being concurrently written. + Upon reaching the end of the input, gzread will return with the available + data. If the error code returned by gzerror is Z_OK or Z_BUF_ERROR, then + gzclearerr can be used to clear the end of file indicator in order to permit + gzread to be tried again. Z_OK indicates that a gzip stream was completed + on the last gzread. Z_BUF_ERROR indicates that the input file ended in the + middle of a gzip stream. Note that gzread does not return -1 in the event + of an incomplete gzip stream. This error is deferred until gzclose(), which + will return Z_BUF_ERROR if the last gzread ended in the middle of a gzip + stream. Alternatively, gzerror can be used before gzclose to detect this + case. + + gzread returns the number of uncompressed bytes actually read, less than + len for end of file, or -1 for error. +*/ + +ZEXTERN int ZEXPORT gzwrite OF((gzFile file, + voidpc buf, unsigned len)); +/* + Writes the given number of uncompressed bytes into the compressed file. + gzwrite returns the number of uncompressed bytes written or 0 in case of + error. +*/ + +ZEXTERN int ZEXPORTVA gzprintf Z_ARG((gzFile file, const char *format, ...)); +/* + Converts, formats, and writes the arguments to the compressed file under + control of the format string, as in fprintf. gzprintf returns the number of + uncompressed bytes actually written, or 0 in case of error. The number of + uncompressed bytes written is limited to 8191, or one less than the buffer + size given to gzbuffer(). The caller should assure that this limit is not + exceeded. If it is exceeded, then gzprintf() will return an error (0) with + nothing written. In this case, there may also be a buffer overflow with + unpredictable consequences, which is possible only if zlib was compiled with + the insecure functions sprintf() or vsprintf() because the secure snprintf() + or vsnprintf() functions were not available. This can be determined using + zlibCompileFlags(). +*/ + +ZEXTERN int ZEXPORT gzputs OF((gzFile file, const char *s)); +/* + Writes the given null-terminated string to the compressed file, excluding + the terminating null character. + + gzputs returns the number of characters written, or -1 in case of error. +*/ + +ZEXTERN char * ZEXPORT gzgets OF((gzFile file, char *buf, int len)); +/* + Reads bytes from the compressed file until len-1 characters are read, or a + newline character is read and transferred to buf, or an end-of-file + condition is encountered. If any characters are read or if len == 1, the + string is terminated with a null character. If no characters are read due + to an end-of-file or len < 1, then the buffer is left untouched. + + gzgets returns buf which is a null-terminated string, or it returns NULL + for end-of-file or in case of error. If there was an error, the contents at + buf are indeterminate. +*/ + +ZEXTERN int ZEXPORT gzputc OF((gzFile file, int c)); +/* + Writes c, converted to an unsigned char, into the compressed file. gzputc + returns the value that was written, or -1 in case of error. +*/ + +ZEXTERN int ZEXPORT gzgetc OF((gzFile file)); +/* + Reads one byte from the compressed file. gzgetc returns this byte or -1 + in case of end of file or error. This is implemented as a macro for speed. + As such, it does not do all of the checking the other functions do. I.e. + it does not check to see if file is NULL, nor whether the structure file + points to has been clobbered or not. +*/ + +ZEXTERN int ZEXPORT gzungetc OF((int c, gzFile file)); +/* + Push one character back onto the stream to be read as the first character + on the next read. At least one character of push-back is allowed. + gzungetc() returns the character pushed, or -1 on failure. gzungetc() will + fail if c is -1, and may fail if a character has been pushed but not read + yet. If gzungetc is used immediately after gzopen or gzdopen, at least the + output buffer size of pushed characters is allowed. (See gzbuffer above.) + The pushed character will be discarded if the stream is repositioned with + gzseek() or gzrewind(). +*/ + +ZEXTERN int ZEXPORT gzflush OF((gzFile file, int flush)); +/* + Flushes all pending output into the compressed file. The parameter flush + is as in the deflate() function. The return value is the zlib error number + (see function gzerror below). gzflush is only permitted when writing. + + If the flush parameter is Z_FINISH, the remaining data is written and the + gzip stream is completed in the output. If gzwrite() is called again, a new + gzip stream will be started in the output. gzread() is able to read such + concatented gzip streams. + + gzflush should be called only when strictly necessary because it will + degrade compression if called too often. +*/ + +/* +ZEXTERN z_off_t ZEXPORT gzseek OF((gzFile file, + z_off_t offset, int whence)); + + Sets the starting position for the next gzread or gzwrite on the given + compressed file. The offset represents a number of bytes in the + uncompressed data stream. The whence parameter is defined as in lseek(2); + the value SEEK_END is not supported. + + If the file is opened for reading, this function is emulated but can be + extremely slow. If the file is opened for writing, only forward seeks are + supported; gzseek then compresses a sequence of zeroes up to the new + starting position. + + gzseek returns the resulting offset location as measured in bytes from + the beginning of the uncompressed stream, or -1 in case of error, in + particular if the file is opened for writing and the new starting position + would be before the current position. +*/ + +ZEXTERN int ZEXPORT gzrewind OF((gzFile file)); +/* + Rewinds the given file. This function is supported only for reading. + + gzrewind(file) is equivalent to (int)gzseek(file, 0L, SEEK_SET) +*/ + +/* +ZEXTERN z_off_t ZEXPORT gztell OF((gzFile file)); + + Returns the starting position for the next gzread or gzwrite on the given + compressed file. This position represents a number of bytes in the + uncompressed data stream, and is zero when starting, even if appending or + reading a gzip stream from the middle of a file using gzdopen(). + + gztell(file) is equivalent to gzseek(file, 0L, SEEK_CUR) +*/ + +/* +ZEXTERN z_off_t ZEXPORT gzoffset OF((gzFile file)); + + Returns the current offset in the file being read or written. This offset + includes the count of bytes that precede the gzip stream, for example when + appending or when using gzdopen() for reading. When reading, the offset + does not include as yet unused buffered input. This information can be used + for a progress indicator. On error, gzoffset() returns -1. +*/ + +ZEXTERN int ZEXPORT gzeof OF((gzFile file)); +/* + Returns true (1) if the end-of-file indicator has been set while reading, + false (0) otherwise. Note that the end-of-file indicator is set only if the + read tried to go past the end of the input, but came up short. Therefore, + just like feof(), gzeof() may return false even if there is no more data to + read, in the event that the last read request was for the exact number of + bytes remaining in the input file. This will happen if the input file size + is an exact multiple of the buffer size. + + If gzeof() returns true, then the read functions will return no more data, + unless the end-of-file indicator is reset by gzclearerr() and the input file + has grown since the previous end of file was detected. +*/ + +ZEXTERN int ZEXPORT gzdirect OF((gzFile file)); +/* + Returns true (1) if file is being copied directly while reading, or false + (0) if file is a gzip stream being decompressed. + + If the input file is empty, gzdirect() will return true, since the input + does not contain a gzip stream. + + If gzdirect() is used immediately after gzopen() or gzdopen() it will + cause buffers to be allocated to allow reading the file to determine if it + is a gzip file. Therefore if gzbuffer() is used, it should be called before + gzdirect(). + + When writing, gzdirect() returns true (1) if transparent writing was + requested ("wT" for the gzopen() mode), or false (0) otherwise. (Note: + gzdirect() is not needed when writing. Transparent writing must be + explicitly requested, so the application already knows the answer. When + linking statically, using gzdirect() will include all of the zlib code for + gzip file reading and decompression, which may not be desired.) +*/ + +ZEXTERN int ZEXPORT gzclose OF((gzFile file)); +/* + Flushes all pending output if necessary, closes the compressed file and + deallocates the (de)compression state. Note that once file is closed, you + cannot call gzerror with file, since its structures have been deallocated. + gzclose must not be called more than once on the same file, just as free + must not be called more than once on the same allocation. + + gzclose will return Z_STREAM_ERROR if file is not valid, Z_ERRNO on a + file operation error, Z_MEM_ERROR if out of memory, Z_BUF_ERROR if the + last read ended in the middle of a gzip stream, or Z_OK on success. +*/ + +ZEXTERN int ZEXPORT gzclose_r OF((gzFile file)); +ZEXTERN int ZEXPORT gzclose_w OF((gzFile file)); +/* + Same as gzclose(), but gzclose_r() is only for use when reading, and + gzclose_w() is only for use when writing or appending. The advantage to + using these instead of gzclose() is that they avoid linking in zlib + compression or decompression code that is not used when only reading or only + writing respectively. If gzclose() is used, then both compression and + decompression code will be included the application when linking to a static + zlib library. +*/ + +ZEXTERN const char * ZEXPORT gzerror OF((gzFile file, int *errnum)); +/* + Returns the error message for the last error which occurred on the given + compressed file. errnum is set to zlib error number. If an error occurred + in the file system and not in the compression library, errnum is set to + Z_ERRNO and the application may consult errno to get the exact error code. + + The application must not modify the returned string. Future calls to + this function may invalidate the previously returned string. If file is + closed, then the string previously returned by gzerror will no longer be + available. + + gzerror() should be used to distinguish errors from end-of-file for those + functions above that do not distinguish those cases in their return values. +*/ + +ZEXTERN void ZEXPORT gzclearerr OF((gzFile file)); +/* + Clears the error and end-of-file flags for file. This is analogous to the + clearerr() function in stdio. This is useful for continuing to read a gzip + file that is being written concurrently. +*/ + +#endif /* !Z_SOLO */ + + /* checksum functions */ + +/* + These functions are not related to compression but are exported + anyway because they might be useful in applications using the compression + library. +*/ + +ZEXTERN uLong ZEXPORT adler32 OF((uLong adler, const Bytef *buf, uInt len)); +/* + Update a running Adler-32 checksum with the bytes buf[0..len-1] and + return the updated checksum. If buf is Z_NULL, this function returns the + required initial value for the checksum. + + An Adler-32 checksum is almost as reliable as a CRC32 but can be computed + much faster. + + Usage example: + + uLong adler = adler32(0L, Z_NULL, 0); + + while (read_buffer(buffer, length) != EOF) { + adler = adler32(adler, buffer, length); + } + if (adler != original_adler) error(); +*/ + +/* +ZEXTERN uLong ZEXPORT adler32_combine OF((uLong adler1, uLong adler2, + z_off_t len2)); + + Combine two Adler-32 checksums into one. For two sequences of bytes, seq1 + and seq2 with lengths len1 and len2, Adler-32 checksums were calculated for + each, adler1 and adler2. adler32_combine() returns the Adler-32 checksum of + seq1 and seq2 concatenated, requiring only adler1, adler2, and len2. Note + that the z_off_t type (like off_t) is a signed integer. If len2 is + negative, the result has no meaning or utility. +*/ + +ZEXTERN uLong ZEXPORT crc32 OF((uLong crc, const Bytef *buf, uInt len)); +/* + Update a running CRC-32 with the bytes buf[0..len-1] and return the + updated CRC-32. If buf is Z_NULL, this function returns the required + initial value for the crc. Pre- and post-conditioning (one's complement) is + performed within this function so it shouldn't be done by the application. + + Usage example: + + uLong crc = crc32(0L, Z_NULL, 0); + + while (read_buffer(buffer, length) != EOF) { + crc = crc32(crc, buffer, length); + } + if (crc != original_crc) error(); +*/ + +/* +ZEXTERN uLong ZEXPORT crc32_combine OF((uLong crc1, uLong crc2, z_off_t len2)); + + Combine two CRC-32 check values into one. For two sequences of bytes, + seq1 and seq2 with lengths len1 and len2, CRC-32 check values were + calculated for each, crc1 and crc2. crc32_combine() returns the CRC-32 + check value of seq1 and seq2 concatenated, requiring only crc1, crc2, and + len2. +*/ + + + /* various hacks, don't look :) */ + +/* deflateInit and inflateInit are macros to allow checking the zlib version + * and the compiler's view of z_stream: + */ +ZEXTERN int ZEXPORT deflateInit_ OF((z_streamp strm, int level, + const char *version, int stream_size)); +ZEXTERN int ZEXPORT inflateInit_ OF((z_streamp strm, + const char *version, int stream_size)); +ZEXTERN int ZEXPORT deflateInit2_ OF((z_streamp strm, int level, int method, + int windowBits, int memLevel, + int strategy, const char *version, + int stream_size)); +ZEXTERN int ZEXPORT inflateInit2_ OF((z_streamp strm, int windowBits, + const char *version, int stream_size)); +ZEXTERN int ZEXPORT inflateBackInit_ OF((z_streamp strm, int windowBits, + unsigned char FAR *window, + const char *version, + int stream_size)); +#define deflateInit(strm, level) \ + deflateInit_((strm), (level), ZLIB_VERSION, (int)sizeof(z_stream)) +#define inflateInit(strm) \ + inflateInit_((strm), ZLIB_VERSION, (int)sizeof(z_stream)) +#define deflateInit2(strm, level, method, windowBits, memLevel, strategy) \ + deflateInit2_((strm),(level),(method),(windowBits),(memLevel),\ + (strategy), ZLIB_VERSION, (int)sizeof(z_stream)) +#define inflateInit2(strm, windowBits) \ + inflateInit2_((strm), (windowBits), ZLIB_VERSION, \ + (int)sizeof(z_stream)) +#define inflateBackInit(strm, windowBits, window) \ + inflateBackInit_((strm), (windowBits), (window), \ + ZLIB_VERSION, (int)sizeof(z_stream)) + +#ifndef Z_SOLO + +/* gzgetc() macro and its supporting function and exposed data structure. Note + * that the real internal state is much larger than the exposed structure. + * This abbreviated structure exposes just enough for the gzgetc() macro. The + * user should not mess with these exposed elements, since their names or + * behavior could change in the future, perhaps even capriciously. They can + * only be used by the gzgetc() macro. You have been warned. + */ +struct gzFile_s { + unsigned have; + unsigned char *next; + z_off64_t pos; +}; +ZEXTERN int ZEXPORT gzgetc_ OF((gzFile file)); /* backward compatibility */ +#ifdef Z_PREFIX_SET +# undef z_gzgetc +# define z_gzgetc(g) \ + ((g)->have ? ((g)->have--, (g)->pos++, *((g)->next)++) : gzgetc(g)) +#else +# define gzgetc(g) \ + ((g)->have ? ((g)->have--, (g)->pos++, *((g)->next)++) : gzgetc(g)) +#endif + +/* provide 64-bit offset functions if _LARGEFILE64_SOURCE defined, and/or + * change the regular functions to 64 bits if _FILE_OFFSET_BITS is 64 (if + * both are true, the application gets the *64 functions, and the regular + * functions are changed to 64 bits) -- in case these are set on systems + * without large file support, _LFS64_LARGEFILE must also be true + */ +#ifdef Z_LARGE64 + ZEXTERN gzFile ZEXPORT gzopen64 OF((const char *, const char *)); + ZEXTERN z_off64_t ZEXPORT gzseek64 OF((gzFile, z_off64_t, int)); + ZEXTERN z_off64_t ZEXPORT gztell64 OF((gzFile)); + ZEXTERN z_off64_t ZEXPORT gzoffset64 OF((gzFile)); + ZEXTERN uLong ZEXPORT adler32_combine64 OF((uLong, uLong, z_off64_t)); + ZEXTERN uLong ZEXPORT crc32_combine64 OF((uLong, uLong, z_off64_t)); +#endif + +#if !defined(ZLIB_INTERNAL) && defined(Z_WANT64) +# ifdef Z_PREFIX_SET +# define z_gzopen z_gzopen64 +# define z_gzseek z_gzseek64 +# define z_gztell z_gztell64 +# define z_gzoffset z_gzoffset64 +# define z_adler32_combine z_adler32_combine64 +# define z_crc32_combine z_crc32_combine64 +# else +# define gzopen gzopen64 +# define gzseek gzseek64 +# define gztell gztell64 +# define gzoffset gzoffset64 +# define adler32_combine adler32_combine64 +# define crc32_combine crc32_combine64 +# endif +# ifndef Z_LARGE64 + ZEXTERN gzFile ZEXPORT gzopen64 OF((const char *, const char *)); + ZEXTERN z_off_t ZEXPORT gzseek64 OF((gzFile, z_off_t, int)); + ZEXTERN z_off_t ZEXPORT gztell64 OF((gzFile)); + ZEXTERN z_off_t ZEXPORT gzoffset64 OF((gzFile)); + ZEXTERN uLong ZEXPORT adler32_combine64 OF((uLong, uLong, z_off_t)); + ZEXTERN uLong ZEXPORT crc32_combine64 OF((uLong, uLong, z_off_t)); +# endif +#else + ZEXTERN gzFile ZEXPORT gzopen OF((const char *, const char *)); + ZEXTERN z_off_t ZEXPORT gzseek OF((gzFile, z_off_t, int)); + ZEXTERN z_off_t ZEXPORT gztell OF((gzFile)); + ZEXTERN z_off_t ZEXPORT gzoffset OF((gzFile)); + ZEXTERN uLong ZEXPORT adler32_combine OF((uLong, uLong, z_off_t)); + ZEXTERN uLong ZEXPORT crc32_combine OF((uLong, uLong, z_off_t)); +#endif + +#else /* Z_SOLO */ + + ZEXTERN uLong ZEXPORT adler32_combine OF((uLong, uLong, z_off_t)); + ZEXTERN uLong ZEXPORT crc32_combine OF((uLong, uLong, z_off_t)); + +#endif /* !Z_SOLO */ + +/* hack for buggy compilers */ +#if !defined(ZUTIL_H) && !defined(NO_DUMMY_DECL) + struct internal_state {int dummy;}; +#endif + +/* undocumented functions */ +ZEXTERN const char * ZEXPORT zError OF((int)); +ZEXTERN int ZEXPORT inflateSyncPoint OF((z_streamp)); +ZEXTERN const z_crc_t FAR * ZEXPORT get_crc_table OF((void)); +ZEXTERN int ZEXPORT inflateUndermine OF((z_streamp, int)); +ZEXTERN int ZEXPORT inflateResetKeep OF((z_streamp)); +ZEXTERN int ZEXPORT deflateResetKeep OF((z_streamp)); +#if defined(_WIN32) && !defined(Z_SOLO) +ZEXTERN gzFile ZEXPORT gzopen_w OF((const wchar_t *path, + const char *mode)); +#endif +#if defined(STDC) || defined(Z_HAVE_STDARG_H) +# ifndef Z_SOLO +ZEXTERN int ZEXPORTVA gzvprintf Z_ARG((gzFile file, + const char *format, + va_list va)); +# endif +#endif + +#endif /* ZLIB_H */ diff --git a/app/src/main/jni/zlib/zutil.c b/app/src/main/jni/zlib/zutil.c new file mode 100644 index 000000000..23d2ebef0 --- /dev/null +++ b/app/src/main/jni/zlib/zutil.c @@ -0,0 +1,324 @@ +/* zutil.c -- target dependent utility functions for the compression library + * Copyright (C) 1995-2005, 2010, 2011, 2012 Jean-loup Gailly. + * For conditions of distribution and use, see copyright notice in zlib.h + */ + +/* @(#) $Id$ */ + +#include "zutil.h" +#ifndef Z_SOLO +# include "gzguts.h" +#endif + +#ifndef NO_DUMMY_DECL +struct internal_state {int dummy;}; /* for buggy compilers */ +#endif + +z_const char * const z_errmsg[10] = { +"need dictionary", /* Z_NEED_DICT 2 */ +"stream end", /* Z_STREAM_END 1 */ +"", /* Z_OK 0 */ +"file error", /* Z_ERRNO (-1) */ +"stream error", /* Z_STREAM_ERROR (-2) */ +"data error", /* Z_DATA_ERROR (-3) */ +"insufficient memory", /* Z_MEM_ERROR (-4) */ +"buffer error", /* Z_BUF_ERROR (-5) */ +"incompatible version",/* Z_VERSION_ERROR (-6) */ +""}; + + +const char * ZEXPORT zlibVersion() +{ + return ZLIB_VERSION; +} + +uLong ZEXPORT zlibCompileFlags() +{ + uLong flags; + + flags = 0; + switch ((int)(sizeof(uInt))) { + case 2: break; + case 4: flags += 1; break; + case 8: flags += 2; break; + default: flags += 3; + } + switch ((int)(sizeof(uLong))) { + case 2: break; + case 4: flags += 1 << 2; break; + case 8: flags += 2 << 2; break; + default: flags += 3 << 2; + } + switch ((int)(sizeof(voidpf))) { + case 2: break; + case 4: flags += 1 << 4; break; + case 8: flags += 2 << 4; break; + default: flags += 3 << 4; + } + switch ((int)(sizeof(z_off_t))) { + case 2: break; + case 4: flags += 1 << 6; break; + case 8: flags += 2 << 6; break; + default: flags += 3 << 6; + } +#ifdef DEBUG + flags += 1 << 8; +#endif +#if defined(ASMV) || defined(ASMINF) + flags += 1 << 9; +#endif +#ifdef ZLIB_WINAPI + flags += 1 << 10; +#endif +#ifdef BUILDFIXED + flags += 1 << 12; +#endif +#ifdef DYNAMIC_CRC_TABLE + flags += 1 << 13; +#endif +#ifdef NO_GZCOMPRESS + flags += 1L << 16; +#endif +#ifdef NO_GZIP + flags += 1L << 17; +#endif +#ifdef PKZIP_BUG_WORKAROUND + flags += 1L << 20; +#endif +#ifdef FASTEST + flags += 1L << 21; +#endif +#if defined(STDC) || defined(Z_HAVE_STDARG_H) +# ifdef NO_vsnprintf + flags += 1L << 25; +# ifdef HAS_vsprintf_void + flags += 1L << 26; +# endif +# else +# ifdef HAS_vsnprintf_void + flags += 1L << 26; +# endif +# endif +#else + flags += 1L << 24; +# ifdef NO_snprintf + flags += 1L << 25; +# ifdef HAS_sprintf_void + flags += 1L << 26; +# endif +# else +# ifdef HAS_snprintf_void + flags += 1L << 26; +# endif +# endif +#endif + return flags; +} + +#ifdef DEBUG + +# ifndef verbose +# define verbose 0 +# endif +int ZLIB_INTERNAL z_verbose = verbose; + +void ZLIB_INTERNAL z_error (m) + char *m; +{ + fprintf(stderr, "%s\n", m); + exit(1); +} +#endif + +/* exported to allow conversion of error code to string for compress() and + * uncompress() + */ +const char * ZEXPORT zError(err) + int err; +{ + return ERR_MSG(err); +} + +#if defined(_WIN32_WCE) + /* The Microsoft C Run-Time Library for Windows CE doesn't have + * errno. We define it as a global variable to simplify porting. + * Its value is always 0 and should not be used. + */ + int errno = 0; +#endif + +#ifndef HAVE_MEMCPY + +void ZLIB_INTERNAL zmemcpy(dest, source, len) + Bytef* dest; + const Bytef* source; + uInt len; +{ + if (len == 0) return; + do { + *dest++ = *source++; /* ??? to be unrolled */ + } while (--len != 0); +} + +int ZLIB_INTERNAL zmemcmp(s1, s2, len) + const Bytef* s1; + const Bytef* s2; + uInt len; +{ + uInt j; + + for (j = 0; j < len; j++) { + if (s1[j] != s2[j]) return 2*(s1[j] > s2[j])-1; + } + return 0; +} + +void ZLIB_INTERNAL zmemzero(dest, len) + Bytef* dest; + uInt len; +{ + if (len == 0) return; + do { + *dest++ = 0; /* ??? to be unrolled */ + } while (--len != 0); +} +#endif + +#ifndef Z_SOLO + +#ifdef SYS16BIT + +#ifdef __TURBOC__ +/* Turbo C in 16-bit mode */ + +# define MY_ZCALLOC + +/* Turbo C malloc() does not allow dynamic allocation of 64K bytes + * and farmalloc(64K) returns a pointer with an offset of 8, so we + * must fix the pointer. Warning: the pointer must be put back to its + * original form in order to free it, use zcfree(). + */ + +#define MAX_PTR 10 +/* 10*64K = 640K */ + +local int next_ptr = 0; + +typedef struct ptr_table_s { + voidpf org_ptr; + voidpf new_ptr; +} ptr_table; + +local ptr_table table[MAX_PTR]; +/* This table is used to remember the original form of pointers + * to large buffers (64K). Such pointers are normalized with a zero offset. + * Since MSDOS is not a preemptive multitasking OS, this table is not + * protected from concurrent access. This hack doesn't work anyway on + * a protected system like OS/2. Use Microsoft C instead. + */ + +voidpf ZLIB_INTERNAL zcalloc (voidpf opaque, unsigned items, unsigned size) +{ + voidpf buf = opaque; /* just to make some compilers happy */ + ulg bsize = (ulg)items*size; + + /* If we allocate less than 65520 bytes, we assume that farmalloc + * will return a usable pointer which doesn't have to be normalized. + */ + if (bsize < 65520L) { + buf = farmalloc(bsize); + if (*(ush*)&buf != 0) return buf; + } else { + buf = farmalloc(bsize + 16L); + } + if (buf == NULL || next_ptr >= MAX_PTR) return NULL; + table[next_ptr].org_ptr = buf; + + /* Normalize the pointer to seg:0 */ + *((ush*)&buf+1) += ((ush)((uch*)buf-0) + 15) >> 4; + *(ush*)&buf = 0; + table[next_ptr++].new_ptr = buf; + return buf; +} + +void ZLIB_INTERNAL zcfree (voidpf opaque, voidpf ptr) +{ + int n; + if (*(ush*)&ptr != 0) { /* object < 64K */ + farfree(ptr); + return; + } + /* Find the original pointer */ + for (n = 0; n < next_ptr; n++) { + if (ptr != table[n].new_ptr) continue; + + farfree(table[n].org_ptr); + while (++n < next_ptr) { + table[n-1] = table[n]; + } + next_ptr--; + return; + } + ptr = opaque; /* just to make some compilers happy */ + Assert(0, "zcfree: ptr not found"); +} + +#endif /* __TURBOC__ */ + + +#ifdef M_I86 +/* Microsoft C in 16-bit mode */ + +# define MY_ZCALLOC + +#if (!defined(_MSC_VER) || (_MSC_VER <= 600)) +# define _halloc halloc +# define _hfree hfree +#endif + +voidpf ZLIB_INTERNAL zcalloc (voidpf opaque, uInt items, uInt size) +{ + if (opaque) opaque = 0; /* to make compiler happy */ + return _halloc((long)items, size); +} + +void ZLIB_INTERNAL zcfree (voidpf opaque, voidpf ptr) +{ + if (opaque) opaque = 0; /* to make compiler happy */ + _hfree(ptr); +} + +#endif /* M_I86 */ + +#endif /* SYS16BIT */ + + +#ifndef MY_ZCALLOC /* Any system without a special alloc function */ + +#ifndef STDC +extern voidp malloc OF((uInt size)); +extern voidp calloc OF((uInt items, uInt size)); +extern void free OF((voidpf ptr)); +#endif + +voidpf ZLIB_INTERNAL zcalloc (opaque, items, size) + voidpf opaque; + unsigned items; + unsigned size; +{ + if (opaque) items += size - size; /* make compiler happy */ + return sizeof(uInt) > 2 ? (voidpf)malloc(items * size) : + (voidpf)calloc(items, size); +} + +void ZLIB_INTERNAL zcfree (opaque, ptr) + voidpf opaque; + voidpf ptr; +{ + free(ptr); + if (opaque) return; /* make compiler happy */ +} + +#endif /* MY_ZCALLOC */ + +#endif /* !Z_SOLO */ diff --git a/app/src/main/jni/zlib/zutil.h b/app/src/main/jni/zlib/zutil.h new file mode 100644 index 000000000..24ab06b1c --- /dev/null +++ b/app/src/main/jni/zlib/zutil.h @@ -0,0 +1,253 @@ +/* zutil.h -- internal interface and configuration of the compression library + * Copyright (C) 1995-2013 Jean-loup Gailly. + * For conditions of distribution and use, see copyright notice in zlib.h + */ + +/* WARNING: this file should *not* be used by applications. It is + part of the implementation of the compression library and is + subject to change. Applications should only use zlib.h. + */ + +/* @(#) $Id$ */ + +#ifndef ZUTIL_H +#define ZUTIL_H + +#ifdef HAVE_HIDDEN +# define ZLIB_INTERNAL __attribute__((visibility ("hidden"))) +#else +# define ZLIB_INTERNAL +#endif + +#include "zlib.h" + +#if defined(STDC) && !defined(Z_SOLO) +# if !(defined(_WIN32_WCE) && defined(_MSC_VER)) +# include +# endif +# include +# include +#endif + +#ifdef Z_SOLO + typedef long ptrdiff_t; /* guess -- will be caught if guess is wrong */ +#endif + +#ifndef local +# define local static +#endif +/* compile with -Dlocal if your debugger can't find static symbols */ + +typedef unsigned char uch; +typedef uch FAR uchf; +typedef unsigned short ush; +typedef ush FAR ushf; +typedef unsigned long ulg; + +extern z_const char * const z_errmsg[10]; /* indexed by 2-zlib_error */ +/* (size given to avoid silly warnings with Visual C++) */ + +#define ERR_MSG(err) z_errmsg[Z_NEED_DICT-(err)] + +#define ERR_RETURN(strm,err) \ + return (strm->msg = ERR_MSG(err), (err)) +/* To be used only when the state is known to be valid */ + + /* common constants */ + +#ifndef DEF_WBITS +# define DEF_WBITS MAX_WBITS +#endif +/* default windowBits for decompression. MAX_WBITS is for compression only */ + +#if MAX_MEM_LEVEL >= 8 +# define DEF_MEM_LEVEL 8 +#else +# define DEF_MEM_LEVEL MAX_MEM_LEVEL +#endif +/* default memLevel */ + +#define STORED_BLOCK 0 +#define STATIC_TREES 1 +#define DYN_TREES 2 +/* The three kinds of block type */ + +#define MIN_MATCH 3 +#define MAX_MATCH 258 +/* The minimum and maximum match lengths */ + +#define PRESET_DICT 0x20 /* preset dictionary flag in zlib header */ + + /* target dependencies */ + +#if defined(MSDOS) || (defined(WINDOWS) && !defined(WIN32)) +# define OS_CODE 0x00 +# ifndef Z_SOLO +# if defined(__TURBOC__) || defined(__BORLANDC__) +# if (__STDC__ == 1) && (defined(__LARGE__) || defined(__COMPACT__)) + /* Allow compilation with ANSI keywords only enabled */ + void _Cdecl farfree( void *block ); + void *_Cdecl farmalloc( unsigned long nbytes ); +# else +# include +# endif +# else /* MSC or DJGPP */ +# include +# endif +# endif +#endif + +#ifdef AMIGA +# define OS_CODE 0x01 +#endif + +#if defined(VAXC) || defined(VMS) +# define OS_CODE 0x02 +# define F_OPEN(name, mode) \ + fopen((name), (mode), "mbc=60", "ctx=stm", "rfm=fix", "mrs=512") +#endif + +#if defined(ATARI) || defined(atarist) +# define OS_CODE 0x05 +#endif + +#ifdef OS2 +# define OS_CODE 0x06 +# if defined(M_I86) && !defined(Z_SOLO) +# include +# endif +#endif + +#if defined(MACOS) || defined(TARGET_OS_MAC) +# define OS_CODE 0x07 +# ifndef Z_SOLO +# if defined(__MWERKS__) && __dest_os != __be_os && __dest_os != __win32_os +# include /* for fdopen */ +# else +# ifndef fdopen +# define fdopen(fd,mode) NULL /* No fdopen() */ +# endif +# endif +# endif +#endif + +#ifdef TOPS20 +# define OS_CODE 0x0a +#endif + +#ifdef WIN32 +# ifndef __CYGWIN__ /* Cygwin is Unix, not Win32 */ +# define OS_CODE 0x0b +# endif +#endif + +#ifdef __50SERIES /* Prime/PRIMOS */ +# define OS_CODE 0x0f +#endif + +#if defined(_BEOS_) || defined(RISCOS) +# define fdopen(fd,mode) NULL /* No fdopen() */ +#endif + +#if (defined(_MSC_VER) && (_MSC_VER > 600)) && !defined __INTERIX +# if defined(_WIN32_WCE) +# define fdopen(fd,mode) NULL /* No fdopen() */ +# ifndef _PTRDIFF_T_DEFINED + typedef int ptrdiff_t; +# define _PTRDIFF_T_DEFINED +# endif +# else +# define fdopen(fd,type) _fdopen(fd,type) +# endif +#endif + +#if defined(__BORLANDC__) && !defined(MSDOS) + #pragma warn -8004 + #pragma warn -8008 + #pragma warn -8066 +#endif + +/* provide prototypes for these when building zlib without LFS */ +#if !defined(_WIN32) && \ + (!defined(_LARGEFILE64_SOURCE) || _LFS64_LARGEFILE-0 == 0) + ZEXTERN uLong ZEXPORT adler32_combine64 OF((uLong, uLong, z_off_t)); + ZEXTERN uLong ZEXPORT crc32_combine64 OF((uLong, uLong, z_off_t)); +#endif + + /* common defaults */ + +#ifndef OS_CODE +# define OS_CODE 0x03 /* assume Unix */ +#endif + +#ifndef F_OPEN +# define F_OPEN(name, mode) fopen((name), (mode)) +#endif + + /* functions */ + +#if defined(pyr) || defined(Z_SOLO) +# define NO_MEMCPY +#endif +#if defined(SMALL_MEDIUM) && !defined(_MSC_VER) && !defined(__SC__) + /* Use our own functions for small and medium model with MSC <= 5.0. + * You may have to use the same strategy for Borland C (untested). + * The __SC__ check is for Symantec. + */ +# define NO_MEMCPY +#endif +#if defined(STDC) && !defined(HAVE_MEMCPY) && !defined(NO_MEMCPY) +# define HAVE_MEMCPY +#endif +#ifdef HAVE_MEMCPY +# ifdef SMALL_MEDIUM /* MSDOS small or medium model */ +# define zmemcpy _fmemcpy +# define zmemcmp _fmemcmp +# define zmemzero(dest, len) _fmemset(dest, 0, len) +# else +# define zmemcpy memcpy +# define zmemcmp memcmp +# define zmemzero(dest, len) memset(dest, 0, len) +# endif +#else + void ZLIB_INTERNAL zmemcpy OF((Bytef* dest, const Bytef* source, uInt len)); + int ZLIB_INTERNAL zmemcmp OF((const Bytef* s1, const Bytef* s2, uInt len)); + void ZLIB_INTERNAL zmemzero OF((Bytef* dest, uInt len)); +#endif + +/* Diagnostic functions */ +#ifdef DEBUG +# include + extern int ZLIB_INTERNAL z_verbose; + extern void ZLIB_INTERNAL z_error OF((char *m)); +# define Assert(cond,msg) {if(!(cond)) z_error(msg);} +# define Trace(x) {if (z_verbose>=0) fprintf x ;} +# define Tracev(x) {if (z_verbose>0) fprintf x ;} +# define Tracevv(x) {if (z_verbose>1) fprintf x ;} +# define Tracec(c,x) {if (z_verbose>0 && (c)) fprintf x ;} +# define Tracecv(c,x) {if (z_verbose>1 && (c)) fprintf x ;} +#else +# define Assert(cond,msg) +# define Trace(x) +# define Tracev(x) +# define Tracevv(x) +# define Tracec(c,x) +# define Tracecv(c,x) +#endif + +#ifndef Z_SOLO + voidpf ZLIB_INTERNAL zcalloc OF((voidpf opaque, unsigned items, + unsigned size)); + void ZLIB_INTERNAL zcfree OF((voidpf opaque, voidpf ptr)); +#endif + +#define ZALLOC(strm, items, size) \ + (*((strm)->zalloc))((strm)->opaque, (items), (size)) +#define ZFREE(strm, addr) (*((strm)->zfree))((strm)->opaque, (voidpf)(addr)) +#define TRY_FREE(s, p) {if (p) ZFREE(s, p);} + +/* Reverse the bytes in a 32-bit value */ +#define ZSWAP32(q) ((((q) >> 24) & 0xff) + (((q) >> 8) & 0xff00) + \ + (((q) & 0xff00) << 8) + (((q) & 0xff) << 24)) + +#endif /* ZUTIL_H */ diff --git a/app/src/main/jniLibs/armeabi-v7a/libimage.so b/app/src/main/jniLibs/armeabi-v7a/libimage.so deleted file mode 100644 index ae80f344d..000000000 Binary files a/app/src/main/jniLibs/armeabi-v7a/libimage.so and /dev/null differ diff --git a/app/src/main/jniLibs/armeabi/libimage.so b/app/src/main/jniLibs/armeabi/libimage.so deleted file mode 100644 index ccfbde9f5..000000000 Binary files a/app/src/main/jniLibs/armeabi/libimage.so and /dev/null differ diff --git a/app/src/main/jniLibs/x86/libimage.so b/app/src/main/jniLibs/x86/libimage.so deleted file mode 100644 index a53c17bf5..000000000 Binary files a/app/src/main/jniLibs/x86/libimage.so and /dev/null differ diff --git a/gradle/wrapper/gradle-wrapper.properties b/gradle/wrapper/gradle-wrapper.properties index 0c71e760d..345761265 100644 --- a/gradle/wrapper/gradle-wrapper.properties +++ b/gradle/wrapper/gradle-wrapper.properties @@ -1,6 +1,6 @@ -#Wed Apr 10 15:27:10 PDT 2013 +#Mon Apr 13 12:32:52 CST 2015 distributionBase=GRADLE_USER_HOME distributionPath=wrapper/dists zipStoreBase=GRADLE_USER_HOME zipStorePath=wrapper/dists -distributionUrl=https\://services.gradle.org/distributions/gradle-2.2.1-all.zip +distributionUrl=https\://services.gradle.org/distributions/gradle-2.3-all.zip