Bug 257644
| Summary: | GPUP WebGL flush() and finish() are asynchronous | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Kimmo Kinnunen <kkinnunen> |
| Component: | WebGL | Assignee: | Kimmo Kinnunen <kkinnunen> |
| Status: | NEW | ||
| Severity: | Normal | CC: | dino, kbr, kkinnunen, webkit-bug-importer |
| Priority: | P2 | Keywords: | InRadar |
| Version: | WebKit Local Build | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
| See Also: | https://bugs.webkit.org/show_bug.cgi?id=252876 | ||
Kimmo Kinnunen
GPUP WebGL flush() and finish() are asynchronous
Should be synchronous
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Radar WebKit Bug Importer
<rdar://problem/110163403>
Kimmo Kinnunen
Pull request: https://github.com/WebKit/WebKit/pull/14623
Kimmo Kinnunen
Blocks bug 252876 for benchmarking before/after read pixels perf
Kenneth Russell
The side-effects of flush() and finish() can't be observed by application code in the WebGL context. Even with finish(), there's a WebGL spec difference compared to OpenGL ES that the status of objects like queries can't change unless control is returned to the browser. For this reason it should be safe to leave these asynchronous, and I think WebKit might incur a performance penalty compared to other implementations by making them synchronous. If it's necessary to use these for synchronization with other subsystems in WebKit like the media stack, I'd suggest adding another internal primitive that can't be called by the web platform - Chromium does this.
Kimmo Kinnunen
Think Chrome's implementation is also synchronous, e.g. the call between renderer and GPU process.
I think previously we discussed wrt "real finish vs gl.finish as gl.flush()", which cannot be observed.
Kimmo Kinnunen
Or, maybe not maybe it is indeed async?
Kenneth Russell
In Chrome at the lowest level, Finish is synchronous:
https://source.chromium.org/chromium/chromium/src/+/refs/heads/main:gpu/command_buffer/client/cmd_buffer_helper.cc;l=220;drc=b8a0323a84f483b25e94b3a24d80fda16c5dd1ae;bpv=1;bpt=1
but Flush isn't:
https://source.chromium.org/chromium/chromium/src/+/refs/heads/main:gpu/command_buffer/client/cmd_buffer_helper.cc;l=173;drc=b8a0323a84f483b25e94b3a24d80fda16c5dd1ae;bpv=0;bpt=1
However, the WebGL implementation never calls the underlying Finish, only Flush:
https://source.chromium.org/chromium/chromium/src/+/main:third_party/blink/renderer/modules/webgl/webgl_rendering_context_base.cc;l=3034?q=webgl_rendering_context_base.cc
Kimmo Kinnunen
> However, the WebGL implementation never calls the underlying Finish, only Flush:
Ah, right! I keep on forgetting.