Roll Emdawnwebgpu and repoint USE_WEBGPU tests at Emdawnwebgpu #25397

kainino0x · 2025-09-26T07:19:50Z

In preparation to remove USE_WEBGPU entirely (#24265). Ideally these tests would live in Dawn, but we don't have a way to automate tests in the browser yet, so for the moment we'll keep them in Emscripten.

Emdawnwebgpu roll includes:

bb3dc45 [emscripten] Use Emscripten `diagnostics.error` instead of `raise`
a3650c0 [emscripten] Error on LINK_AS_CXX=0 in emdawnwebgpu.port.py

In preparation to remove USE_WEBGPU entirely (emscripten-core#24265). Ideally these tests would live in Dawn, but we don't have a way to automate tests in the browser yet, so for the moment we'll keep them in Emscripten. DO NOT SUBMIT: Needs Emdawnwebgpu roll in order to pass

kainino0x · 2025-09-26T07:19:56Z

@lokokung PTAL

test/webgpu_required_limits.c

test/test_browser.py

tools/ports/emdawnwebgpu.py

webgpu.cpp uses C++ library functions, so it needs to be linked with C++ support, otherwise it gets a long list of missing C++ STL symbol errors. Make this an explicit error so people know what to do if they get this error (because they're using -sSTRICT or -sNO_DEFAULT_TO_CXX). Initial discussion: emscripten-core/emscripten#25397 (comment) Bug: none Change-Id: I99b766298f47b946e0ba49eacad04fe8b427f6de Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/264434 Reviewed-by: Loko Kung <lokokung@google.com> Commit-Queue: Kai Ninomiya <kainino@chromium.org>

…sn't release until after the callback)

test/test_other.py

sbc100

I'm not really qualified to read the .c/.cpp changes to the test files, so I'll have to take your word for it mostly that they good.

@juj might have some opinions?

test/test_browser.py

test/test_other.py

sbc100 · 2025-10-03T22:40:38Z

test/test_other.py

-      '-Wno-error=deprecated',
      '-sINCLUDE_FULL_LIBRARY',
-      '-sUSE_WEBGPU',
+      '--use-port=emdawnwebgpu',


Do we have any remaining tests for -sUSE_WEBGPU?

No, but I'm planning to remove it immediately after this (#25398).

sbc100 · 2025-10-03T22:41:00Z

test/test_other.py

-    'dylink': (['-sMAIN_MODULE'],),
-  })
-  def test_webgpu_compiletest(self, args):
-    self.run_process([EMXX, test_file('webgpu_jsvalstore.cpp'), '-Wno-error=deprecated', '-sUSE_WEBGPU', '-sASYNCIFY'] + args)


Why is this test no longer useful?

The whole jsvalstore thing has been removed in Emdawnwebgpu. We have another mechanism but it would need entirely new tests.

sbc100 · 2025-10-03T22:41:51Z

test/test_other.py

-    self.run_process([EMXX, test_file('test_emdawnwebgpu_link_test.cpp'), '--use-port=emdawnwebgpu', '-sASYNCIFY'] + args)
+    if config.FROZEN_CACHE:
+      # TODO(crbug.com/446944885): Make Emdawnwebgpu work with FROZEN_CACHE if possible.
+      self.skipTest("test doesn't work with frozen cache")


Why is this skipTest needed now but not before this PR?

I think it's because previously the cache was pre-populated with -sUSE_WEBGPU but we don't prepopulate it with --use-port=emdawnwebgpu (which would have to be done with several different flags too).

I can try to pre-populate the cache with the appropriate Emdawnwebgpu objects if you think that's appropriate.

emscripten/embuilder.py

Lines 103 to 104 in 3b22cd6

'libwebgpu',

'libwebgpu_cpp',

I think it comes from this? I don't think it'll be that easy to add Emdawnwebgpu here though.

Oh, it would be added via this

emscripten/embuilder.py

Line 139 in 3b22cd6

PORTS = sorted(list(ports.ports_by_name.keys()) + list(ports.port_variants.keys()))

But the old code and the new code here both use --use-port=emdawnwebgpu.

Oh, sorry, I missed which test this was. I actually have no idea. If I remove emdawnwebgpu from my local cache and run the old test (on main), it still fails with FROZEN_CACHE is set, but cache file is missing.

Could the @flaky be suppressing it?

Oh, CI has an explicit separate Embuilder step. Let me figure out how to test that locally and then see where I need to skip tests.

Okay, I figured out which configurations don't get precached and skipped only those.

... Although I did not figure out exactly why it was passing before and isn't now. @also_with_wasm64 isn't newly-added.

test/webgpu_required_limits.c

test/webgpu_basic_rendering.cpp

juj · 2025-10-03T23:01:34Z

@juj might have some opinions?

LGTM with the test changes - looks like they're effectively refactoring callbacks to lambdas.

kainino0x · 2025-10-03T23:03:25Z

LGTM with the test changes - looks like they're effectively refactoring callbacks to lambdas.

Mainly yes. But also taking advantage of keepalives to let a bunch of async stuff finish before the test completes (without using asyncify).

kainino0x mentioned this pull request Sep 26, 2025

Remove -sUSE_WEBGPU in favor of --use-port=emdawnwebgpu #25398

Draft

lokokung approved these changes Sep 26, 2025

View reviewed changes

test/webgpu_required_limits.c Show resolved Hide resolved

kainino0x force-pushed the roll-and-repoint-tests branch 2 times, most recently from a88588a to ffb124d Compare September 26, 2025 18:51

stronger test that keepalives work correctly

d12517b

kainino0x force-pushed the roll-and-repoint-tests branch from ffb124d to d12517b Compare September 26, 2025 18:55

lokokung approved these changes Sep 26, 2025

View reviewed changes

Roll Emdawnwebgpu

ef5a6da

kainino0x force-pushed the roll-and-repoint-tests branch from 3b2d73d to 4c7bf8a Compare September 27, 2025 06:01

Skip with FROZEN_CACHE

4f93d5f

kainino0x force-pushed the roll-and-repoint-tests branch from 4c7bf8a to 4f93d5f Compare September 27, 2025 06:53

kainino0x requested a review from sbc100 September 27, 2025 08:58

kainino0x marked this pull request as ready for review September 27, 2025 08:58

sbc100 reviewed Sep 29, 2025

View reviewed changes

test/test_browser.py Outdated Show resolved Hide resolved

sbc100 reviewed Sep 29, 2025

View reviewed changes

tools/ports/emdawnwebgpu.py Show resolved Hide resolved

kainino0x added 2 commits September 29, 2025 16:26

Remove unneeded NO_DEFAULT_TO_CXX

3d62df6

Merge branch 'main' into roll-and-repoint-tests

e817b60

kainino0x added 3 commits October 2, 2025 23:40

roll dawn

e070e8e

Remove flaky check for keepRuntimeAlive (emscripten_set_main_loop doe…

b3c5761

…sn't release until after the callback)

Merge branch 'main' into roll-and-repoint-tests

c674a2c

kainino0x requested a review from sbc100 October 3, 2025 19:50

sbc100 reviewed Oct 3, 2025

View reviewed changes

test/test_other.py Show resolved Hide resolved

kainino0x added 2 commits October 3, 2025 15:31

add crossplatform

abd2580

Merge branch 'main' into roll-and-repoint-tests

1dd4a1c

sbc100 reviewed Oct 3, 2025

View reviewed changes

address nits

9523fef

juj reviewed Oct 3, 2025

View reviewed changes

test/webgpu_basic_rendering.cpp Outdated Show resolved Hide resolved

kainino0x added 3 commits October 3, 2025 16:08

Don't exit from atexit

1946c55

fix typo

1714f1c

Re-enable most tests on FROZEN_CACHE

8ebf5ef

kainino0x force-pushed the roll-and-repoint-tests branch from 2b69827 to 8ebf5ef Compare October 4, 2025 00:58

autoformat

74c3ec6

Roll Emdawnwebgpu and repoint USE_WEBGPU tests at Emdawnwebgpu #25397

Are you sure you want to change the base?

Roll Emdawnwebgpu and repoint USE_WEBGPU tests at Emdawnwebgpu #25397

Conversation

kainino0x commented Sep 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

kainino0x commented Sep 26, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

sbc100 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

kainino0x Oct 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

juj commented Oct 3, 2025

Uh oh!

kainino0x commented Oct 3, 2025

Uh oh!

Uh oh!

kainino0x commented Sep 26, 2025 •

edited

Loading

kainino0x Oct 3, 2025 •

edited

Loading