Repository navigation
Conversation
finish, submit, writeBuffer/writeTexture and present now run in order on one thread per device instead of inside the JS call: - encoder.finish() returns at once; the command buffer is encoded on the queue thread just before the submit that uses it. - writeBuffer/writeTexture copy their data and are recorded without waking the thread; recorded work is sent with the next submit, present or drain. - The page draws into a drawing-buffer texture. Present acquires the swapchain image, copies the frame into it and presents on the queue thread, so getCurrentTexture never waits on the swapchain. The copy replaces the readback copy present already made; readbacks read the drawing buffer. Each frame starts cleared through Texture::discard_contents (wgpu bump to 04e9fef72). The page runs at most one frame ahead. - Error scopes are mirrored onto the queue thread, so errors from queued work reach the scope open when the call was made. Uncaptured errors go to the JS thread: through its loop on Android and Apple, otherwise on its next WebGPU call. - getCurrentTexture (non-drawing-buffer paths), configure/resize, readbacks, mapAsync, destroy and the copy helpers wait for queued work first.
|
Important Review skippedAuto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the
 |
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stacked on #198.
What
encoder.finish(),queue.submit(),writeBuffer/writeTextureand present now run in order on one thread per device, instead of inside the JS call.finish()returns at once. The command buffer is encoded on the queue thread just before the submit that uses it.writeBuffer/writeTexturecopy their data on the JS thread and are recorded without waking the queue thread. Recorded work is sent with the next submit, present or drain, so a frame costs a few wakeups rather than one per write.getCurrentTexturenever waits on the swapchain.toDataURL/drawImage(canvas)read the drawing buffer.Texture::discard_contents, which is why wgpu goes to04e9fef72. It is cleared lazily, as a new swapchain image is.mapAsync, buffer/texture/devicedestroy, and the image-copy helpers.getCurrentTexturealso waits on paths without a drawing buffer.This applies on every platform. On Linux, frames still reach GTK through
glib::MainContext::invoke.Measurements
Galaxy S25 Ultra at 60 Hz, three.js r184
WebGPURenderer, mixed-material scene. Same session: this PR vs #198. Times are medianrender().A PixiJS scene with 32,768 sprites rendered 54.9 fps and displayed 55.2 on SurfaceFlinger, so no frames are dropped between rendering and display.
Testing
webgpu.canvastiming specs and the 2webgpu.videospecs fail on the base too.