yenru0/zed

0 Fork 0

Files

T

Ben Brandt d4d6abd439 Fix DirectX atlas panic after GPU device recovery (#55878 )

## Problem

A Sentry-reported crash on Windows (Intel Iris Xe Graphics, v1.0.1):

```
index out of bounds: the len is 1 but the index is 1
```

panicking at `DirectXAtlasState::texture` in

[`crates/gpui_windows/src/directx_atlas.rs`](https://github.com/zed-industries/zed/blob/main/crates/gpui_windows/src/directx_atlas.rs):

```rust
AtlasTextureKind::Subpixel => {
    &self.subpixel_textures[id.index as usize].as_ref().unwrap()
}
```

## Root cause

After a GPU device-lost recovery, GPUI's view cache replays stale
`AtlasTile`
references from the previous frame's `paint_operations` via
`Scene::replay`.

1. **Atlas grows past one texture.** A long enough session pushes
`subpixel_textures.textures.len() ≥ 2` (easy on Iris Xe at the default
1024×1024 atlas size). Top-level views in Zed use `cached(...)`, so
their
`AnyViewState.paint_range` records into
`rendered_frame.scene.paint_operations`,
   referencing both index `0` and index `1`.
2. **Device lost.** `handle_device_lost` clears every `AtlasTextureList`
(`textures.len() == 0`)
   and `tiles_by_key`, then sets `skip_draws = true`.
3. **`WM_GPUI_FORCE_UPDATE_WINDOW` arrives.** `mark_drawable()` flips
`skip_draws` back to `false` and `request_frame` runs with
`force_render: true`.
4. **The cache hit.** Inside `Window::draw`, `AnyView::prepaint`'s cache
check (`!dirty_views.contains(...) && !window.refreshing`) succeeds for
every cached view because the recovery doesn't touch invalidator state
and
`force_render` doesn't propagate into `Window`. `AnyView::paint` calls
`window.reuse_paint` → `Scene::replay` → `primitive.clone()`, which
(since
   `SubpixelSprite`/`AtlasTile` are `Copy`) verbatim copies a
`Primitive::SubpixelSprite { tile: { texture_id: { index: 1, ... }, ...
} }`
   into `next_frame.scene`.
5. **Atlas regrows to one.** Dirty/uncached parts of the same frame
   (caret, animations, anything that called `cx.notify`) fall through to
   `paint_glyph` → `get_or_insert_with` → `push_texture`, growing
`subpixel_textures.textures` from `0` to **`1`** with index `0` valid.
6. **Panic.** After `mem::swap`, `rendered_frame.scene` contains a mix
of
   fresh `index = 0` and replayed `index = 1` sprites. `Scene::batches`
emits separate batches per `texture_id`; the `index = 1` batch reaches
   `atlas.get_texture_view` → `subpixel_textures[1]` → panic with
   `len = 1, index = 1`.

The two earlier related fixes do not catch this:

- **#52389 / c8f20c78** (`if force_render { mark_drawable }`) protects
the
200 ms recovery sleep — pending `WM_PAINT`s carry `force_render = false`
  and so do not clear `skip_draws`. But `WM_GPUI_FORCE_UPDATE_WINDOW`
  carries `force_render = true`, so `mark_drawable` runs, then
  `Window::draw`'s `reuse_paint` still reproduces stale tiles.
- The unmerged Windows draft `2e5d890e37`
(`force_render_after_recovery`)
  similarly only forces the forced-render branch — it doesn't bypass the
  view cache.

## Fix

Two parts:

**1. Bypass the view cache on a forced draw (cross-platform).**

In the platform-agnostic `request_frame` closure in `Window::new`, call
`window.refresh()` whenever `RequestFrameOptions::force_render` is
`true`.
`Window::refresh` is the documented escape hatch for cached views (per
the
`AnyView::cached` docs: *"The one exception is when [Window::refresh] is
called, in which case caching is ignored."*). With `refreshing = true`
every `AnyView::prepaint` cache check fails, every cached view fully
repaints, and `paint_glyph` allocates fresh tiles for every glyph, so
`rendered_frame.scene` ends up free of stale `AtlasTile`s.

**2. Add the `force_render_after_recovery` flag on Windows.**

Mirror the Linux fix from #52389: a per-window `Cell<bool>` set after
`WindowsWindowInner::handle_device_lost` succeeds and consumed at the
top
of `draw_window`. Together with the GPUI change above, the first frame
after recovery (whether a stray `WM_PAINT` during the 200 ms recovery
sleep or the explicit `WM_GPUI_FORCE_UPDATE_WINDOW`) is treated as a
forced render that both clears `skip_draws` and bypasses the view cache.

## Testing

- `script/clippy -p gpui` is clean.
- I do not have a Windows toolchain available locally, so I have not
  cross-compiled `gpui_windows`. Reviewers with Windows access — please
  smoke-test on a machine where the device-lost path can be exercised
  (Intel iGPU, suspend/resume, or running a TDR-inducing test on a GPU
  driver).

## Related

- Sentry issue ID 7457971403 (DirectX subpixel atlas crash, Intel Iris
Xe).
- Builds on / fixes the residual gap in #52389 (`gpui_linux: Force scene
  rebuild after GPU device recovery"). The GPUI change here also hardens
  the corresponding Linux path against the same `reuse_paint` mechanism.

Release Notes:

- Fixed a crash on Windows when the GPU device is lost and recovered
during use (typically driver crash, suspend/resume, or display
reconfiguration, most commonly on Intel iGPUs)

2026-05-07 10:22:52 +00:00

docs

docs: Change render function's return type (#27229 )

2025-03-20 22:48:22 -06:00

examples

gpui: Add examples index (#55154 )

2026-04-29 12:36:29 +00:00

resources/windows

Enable segment heap for Zed (#54538 )

2026-04-22 16:58:55 -04:00

src

Fix DirectX atlas panic after GPU device recovery (#55878 )

2026-05-07 10:22:52 +00:00

tests

Use serde 1.0.221 instead of serde_derive hackery (#38137 )

2025-09-14 14:01:04 +02:00

build.rs

GPUI on the web (#50228 )

2026-02-26 18:36:50 +01:00

Cargo.toml

gpui: Remove naga dependency (#55070 )

2026-04-28 10:00:07 +00:00

LICENSE-APACHE

chore: Add crate licenses. (#4158 )

2024-01-23 16:56:22 +01:00

README.md

gpui: Add examples index (#55154 )

2026-04-29 12:36:29 +00:00

README.md

Welcome to GPUI!

GPUI is a hybrid immediate and retained mode, GPU accelerated, UI framework for Rust, designed to support a wide variety of applications.

Getting Started

GPUI is still in active development as we work on the Zed code editor, and is still pre-1.0. There will often be breaking changes between versions. You'll also need to use the latest version of stable Rust and be on macOS or Linux. Add the following to your Cargo.toml:

gpui = { version = "*" }

Ownership and data flow

Everything in GPUI starts with an Application. You can create one with Application::new(), and kick off your application by passing a callback to Application::run(). Inside this callback, you can create a new window with App::open_window(), and register your first root view. See gpui.rs for a complete example.

Dependencies

GPUI has various system dependencies that it needs in order to work.

macOS

On macOS, GPUI uses Metal for rendering. In order to use Metal, you need to do the following:

Install Xcode from the macOS App Store, or from the Apple Developer website. Note this requires a developer account.

Ensure you launch Xcode after installing, and install the macOS components, which is the default option.

Install Xcode command line tools
```
xcode-select --install
```
Ensure that the Xcode command line tools are using your newly installed copy of Xcode:
```
sudo xcode-select --switch /Applications/Xcode.app/Contents/Developer
```

The Big Picture

GPUI offers three different registers depending on your needs:

State management and communication with Entity's. Whenever you need to store application state that communicates between different parts of your application, you'll want to use GPUI's entities. Entities are owned by GPUI and are only accessible through an owned smart pointer similar to an Rc. See the app::context module for more information.
High level, declarative UI with views. All UI in GPUI starts with a view. A view is simply an Entity that can be rendered, by implementing the Render trait. At the start of each frame, GPUI will call this render method on the root view of a given window. Views build a tree of elements, lay them out and style them with a tailwind-style API, and then give them to GPUI to turn into pixels. See the div element for an all purpose swiss-army knife of rendering.
Low level, imperative UI with Elements. Elements are the building blocks of UI in GPUI, and they provide a nice wrapper around an imperative API that provides as much flexibility and control as you need. Elements have total control over how they and their child elements are rendered and can be used for making efficient views into large lists, implement custom layouting for a code editor, and anything else you can think of. See the element module for more information.

Each of these registers has one or more corresponding contexts that can be accessed from all GPUI services. This context is your main interface to GPUI, and is used extensively throughout the framework.

Other Resources

In addition to the systems above, GPUI provides a range of smaller services that are useful for building complex applications:

Actions are user-defined structs that are used for converting keystrokes into logical operations in your UI. Use this for implementing keyboard shortcuts, such as cmd-q. See the action module for more information.
Platform services, such as quit the app or open a URL are available as methods on the app::App.
An async executor that is integrated with the platform's event loop. See the executor module for more information.,
The [gpui::test] macro provides a convenient way to write tests for your GPUI applications. Tests also have their own kind of context, a TestAppContext which provides ways of simulating common platform input. See app::test_context and test modules for more details.

Currently, the best way to learn about these APIs is to read the Zed source code or drop a question in the Zed Discord. We're working on improving the documentation, creating more examples, and will be publishing more guides to GPUI on our blog.