[webkit-reviews] review granted: [Bug 191990] [WebGPU] Begin implementation of WebGPURenderPassEncoder and barebones WebGPURenderPassDescriptor : [Attachment 355819] Patch

Wed Nov 28 10:42:31 PST 2018

Dean Jackson <dino at apple.com> has granted Justin Fan <justin_fan at apple.com>'s
request for review:
Bug 191990: [WebGPU] Begin implementation of WebGPURenderPassEncoder and
barebones WebGPURenderPassDescriptor
https://bugs.webkit.org/show_bug.cgi?id=191990

Attachment 355819: Patch

https://bugs.webkit.org/attachment.cgi?id=355819&action=review

--- Comment #5 from Dean Jackson <dino at apple.com> ---
Comment on attachment 355819
  --> https://bugs.webkit.org/attachment.cgi?id=355819
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=355819&action=review

r+ but don't commit the new scheme.

> Source/WebCore/platform/graphics/gpu/cocoa/GPURenderPassEncoderMetal.mm:73
> +    // The command encoder must not be expecting more commands when it is
destroyed.

I think we should be a bit more descriptive in this comment. e.g. // The
MTLRenderCommandEncoder object must have finished encoding before it can be
released, whether or not any commands were encoded.

Then also add // FIXME: Remember if we've ended encoding and only call this if
we haven't.

> LayoutTests/webgpu/render-passes.html:2
> +<!DOCTYPE html>
> +<html>

I was trying to start writing tests that used the WPT approach with the idea of
contributing it back to the WebGPU group. This test is fine for now, but I
think we should keep this goal in mind.