diff --git a/.gitignore b/.gitignore
index 5d9e8921a..5713d5e10 100644
--- a/.gitignore
+++ b/.gitignore
@@ -8,5 +8,7 @@ lint.xml
 libs/armeabi
 libs/armeabi-v7a
 libs/arm64-v8a
-libs/mips
 libs/x86
+libs/x86_64
+libs/mips
+libs/mips64
diff --git a/README.md b/README.md
index 02bcd77f8..25bc6e01f 100644
--- a/README.md
+++ b/README.md
@@ -6,8 +6,105 @@ EhViewer is an E-Hentai Application for Android. Visit [EhViewer](http://www.ehv
 # Build
 
 本项目使用 Eclipse + ADT 作为开发环境，如果你不清楚如何配置，请自行搜索。clone 之后直接导入 Eclipse 即可编译。我没试过将其迁至 Android Studio，如果你喜欢使用 Android Studio，你需要自己研究下，项目中有 NDK 部分，迁移时要注意。<br>
+目前 Android NDK (r10d) 只对 x86 使用 yasm，而 x86\_64 没有启用 yasm，但是 libjpeg-turbo 使用了汇编代码。如果要编译 x86_64 的 jni 部分需要对 Android NDK 中的脚本进行修改。<br>
 Just import into Eclipse and build. Do it youself if you prefer Android Studio.
 
+##针对 x86\_64 yasm 所做的修改
+build/core/build-binary.mk<br>
+将（注意第三行）<br>
+```markdown
+# all_source_patterns contains the list of filename patterns that correspond
+# to source files recognized by our build system
+ifeq ($(TARGET_ARCH_ABI),x86)
+all_source_extensions := .c .s .S .asm $(LOCAL_CPP_EXTENSION) $(LOCAL_RS_EXTENSION)
+else
+all_source_extensions := .c .s .S $(LOCAL_CPP_EXTENSION) $(LOCAL_RS_EXTENSION)
+endif
+all_source_patterns   := $(foreach _ext,$(all_source_extensions),%$(_ext))
+all_cpp_patterns      := $(foreach _ext,$(LOCAL_CPP_EXTENSION),%$(_ext))
+all_rs_patterns       := $(foreach _ext,$(LOCAL_RS_EXTENSION),%$(_ext))
+```
+改为<br>
+```markdown
+# all_source_patterns contains the list of filename patterns that correspond
+# to source files recognized by our build system
+ifeq ($(TARGET_ARCH_ABI),$(filter $(TARGET_ARCH_ABI), x86 x86_64))
+all_source_extensions := .c .s .S .asm $(LOCAL_CPP_EXTENSION) $(LOCAL_RS_EXTENSION)
+else
+all_source_extensions := .c .s .S $(LOCAL_CPP_EXTENSION) $(LOCAL_RS_EXTENSION)
+endif
+all_source_patterns   := $(foreach _ext,$(all_source_extensions),%$(_ext))
+all_cpp_patterns      := $(foreach _ext,$(LOCAL_CPP_EXTENSION),%$(_ext))
+all_rs_patterns       := $(foreach _ext,$(LOCAL_RS_EXTENSION),%$(_ext))
+```
+然后在<br>
+```markdown
+ifeq ($(TARGET_ARCH_ABI),x86)
+$(foreach src,$(filter %.asm,$(LOCAL_SRC_FILES)), $(call compile-asm-source,$(src),$(call get-object-name,$(src))))
+endif
+```
+的下方添加<br>
+```markdown
+ifeq ($(TARGET_ARCH_ABI),x86_64)
+$(foreach src,$(filter %.asm,$(LOCAL_SRC_FILES)), $(call compile-asm-64-source,$(src),$(call get-object-name,$(src))))
+endif
+```
+build/core/definitions.mk<br>
+添加
+```markdown
+# -----------------------------------------------------------------------------
+# Template  : ev-compile-asm-64-source
+# Arguments : 1: single ASM source file name (relative to LOCAL_PATH)
+#             2: target object file (without path)
+# Returns   : None
+# Usage     : $(eval $(call ev-compile-asm-64-source,<srcfile>,<objfile>)
+# Rationale : Internal template evaluated by compile-asm-64-source
+# -----------------------------------------------------------------------------
+define  ev-compile-asm-64-source
+_SRC:=$$(call local-source-file-path,$(1))
+_OBJ:=$$(LOCAL_OBJS_DIR:%/=%)/$(2)
+
+_FLAGS := $$(call host-c-includes,$$(LOCAL_C_INCLUDES) $$(LOCAL_PATH)) \
+          $$(LOCAL_ASMFLAGS) \
+          $$(NDK_APP_ASMFLAGS) \
+          $$(call host-c-includes,$$($(my)C_INCLUDES)) \
+          -f elf64 -m amd64 -DELF -D__x86_64__ -DPIC
+
+_TEXT := Assemble $$(call get-src-file-text,$1)
+_CC   := $$(NDK_CCACHE) $$(TARGET_ASM)
+
+$$(_OBJ): PRIVATE_ABI      := $$(TARGET_ARCH_ABI)
+$$(_OBJ): PRIVATE_SRC      := $$(_SRC)
+$$(_OBJ): PRIVATE_OBJ      := $$(_OBJ)
+$$(_OBJ): PRIVATE_MODULE   := $$(LOCAL_MODULE)
+$$(_OBJ): PRIVATE_TEXT     := $$(_TEXT)
+$$(_OBJ): PRIVATE_CC       := $$(_CC)
+$$(_OBJ): PRIVATE_CFLAGS   := $$(_FLAGS)
+
+ifeq ($$(LOCAL_SHORT_COMMANDS),true)
+_OPTIONS_LISTFILE := $$(_OBJ).cflags
+$$(_OBJ): $$(call generate-list-file,$$(_FLAGS),$$(_OPTIONS_LISTFILE))
+$$(_OBJ): PRIVATE_CFLAGS := @$$(call host-path,$$(_OPTIONS_LISTFILE))
+$$(_OBJ): $$(_OPTIONS_LISTFILE)
+endif
+
+$$(call generate-file-dir,$$(_OBJ))
+$$(_OBJ): $$(_SRC) $$(LOCAL_MAKEFILE) $$(NDK_APP_APPLICATION_MK) $$(NDK_DEPENDENCIES_CONVERTER) $(LOCAL_RS_OBJECTS)
+    $$(call host-echo-build-step,$$(PRIVATE_ABI),$$(PRIVATE_TEXT)) "$$(PRIVATE_MODULE) <= $$(notdir $$(PRIVATE_SRC))"
+    $$(hide) $$(PRIVATE_CC) $$(PRIVATE_CFLAGS) $$(call host-path,$$(PRIVATE_SRC)) -o $$(call host-path,$$(PRIVATE_OBJ))
+endef
+
+# -----------------------------------------------------------------------------
+# Function  : compile-asm-64-source
+# Arguments : 1: single Assembly source file name (relative to LOCAL_PATH)
+#             2: object file
+# Returns   : None
+# Usage     : $(call compile-asm-64-source,<srcfile>,<objfile>)
+# Rationale : Setup everything required to build a single Assembly source file
+# -----------------------------------------------------------------------------
+compile-asm-64-source = $(eval $(call ev-compile-asm-64-source,$1,$2))
+```
+
 # Thanks
 
 - [android-gif-drawable](https://github.com/koral--/android-gif-drawable)
diff --git a/jni/Application.mk b/jni/Application.mk
index 38a7dde6b..7eed0c0fe 100644
--- a/jni/Application.mk
+++ b/jni/Application.mk
@@ -1,3 +1,3 @@
-APP_ABI := armeabi armeabi-v7a arm64-v8a x86 mips
+APP_ABI := armeabi armeabi-v7a arm64-v8a x86 x86_64 mips
 APP_PLATFORM := android-7
 APP_OPTIM := release
diff --git a/jni/libjpeg-turbo/README.md b/jni/libjpeg-turbo/README.md
index 20b810b3e..fcdb395d5 100644
--- a/jni/libjpeg-turbo/README.md
+++ b/jni/libjpeg-turbo/README.md
@@ -1,9 +1,9 @@
 # simd/jsimdext.inc
 
-%define EXTN(name)   _ %+ name
-to
-%define EXTN(name)   name
-or ld error for x86
+%define EXTN(name)   _ %+ name<br>
+to<br>
+%define EXTN(name)   name<br>
+or ld error for x86 and x86_64
 
 # Add cmyk to rgb, cmyk to gray
 
diff --git a/jni/libjpeg-turbo/simd/x86_64/jccolext-sse2-64.asm b/jni/libjpeg-turbo/simd/jccolext-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jccolext-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jccolext-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jccolor-sse2-64.asm b/jni/libjpeg-turbo/simd/jccolor-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jccolor-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jccolor-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jcgray-sse2-64.asm b/jni/libjpeg-turbo/simd/jcgray-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jcgray-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jcgray-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jcgryext-sse2-64.asm b/jni/libjpeg-turbo/simd/jcgryext-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jcgryext-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jcgryext-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jcsample-sse2-64.asm b/jni/libjpeg-turbo/simd/jcsample-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jcsample-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jcsample-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jdcolext-sse2-64.asm b/jni/libjpeg-turbo/simd/jdcolext-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jdcolext-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jdcolext-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jdcolor-sse2-64.asm b/jni/libjpeg-turbo/simd/jdcolor-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jdcolor-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jdcolor-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jdmerge-sse2-64.asm b/jni/libjpeg-turbo/simd/jdmerge-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jdmerge-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jdmerge-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jdmrgext-sse2-64.asm b/jni/libjpeg-turbo/simd/jdmrgext-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jdmrgext-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jdmrgext-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jdsample-sse2-64.asm b/jni/libjpeg-turbo/simd/jdsample-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jdsample-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jdsample-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jfdctflt-sse-64.asm b/jni/libjpeg-turbo/simd/jfdctflt-sse-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jfdctflt-sse-64.asm
rename to jni/libjpeg-turbo/simd/jfdctflt-sse-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jfdctfst-sse2-64.asm b/jni/libjpeg-turbo/simd/jfdctfst-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jfdctfst-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jfdctfst-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jfdctint-sse2-64.asm b/jni/libjpeg-turbo/simd/jfdctint-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jfdctint-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jfdctint-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jidctflt-sse2-64.asm b/jni/libjpeg-turbo/simd/jidctflt-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jidctflt-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jidctflt-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jidctfst-sse2-64.asm b/jni/libjpeg-turbo/simd/jidctfst-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jidctfst-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jidctfst-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jidctint-sse2-64.asm b/jni/libjpeg-turbo/simd/jidctint-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jidctint-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jidctint-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jidctred-sse2-64.asm b/jni/libjpeg-turbo/simd/jidctred-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jidctred-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jidctred-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jquantf-sse2-64.asm b/jni/libjpeg-turbo/simd/jquantf-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jquantf-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jquantf-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/x86_64/jquanti-sse2-64.asm b/jni/libjpeg-turbo/simd/jquanti-sse2-64.asm
similarity index 100%
rename from jni/libjpeg-turbo/simd/x86_64/jquanti-sse2-64.asm
rename to jni/libjpeg-turbo/simd/jquanti-sse2-64.asm
diff --git a/jni/libjpeg-turbo/simd/jsimd_x86_64.c b/jni/libjpeg-turbo/simd/jsimd_x86_64.c
new file mode 100644
index 000000000..2c47a7f36
--- /dev/null
+++ b/jni/libjpeg-turbo/simd/jsimd_x86_64.c
@@ -0,0 +1,757 @@
+/*
+ * jsimd_x86_64.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> 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);
+}
+