If you add more command line options like are added for the [trigger] step for webgl_conformance on the bots, e.g.:

https://build.chromium.org/p/chromium.gpu/builders/Linux%20Release%20%28NVIDIA%29/builds/74875
https://build.chromium.org/p/chromium.gpu/builders/Linux%20Release%20%28NVIDIA%29/builds/74875/steps/%5Btrigger%5D%20webgl_conformance_tests%20on%20NVIDIA%20GPU%20on%20Linux%20on%20Linux/logs/stdio

python -u /b/build/nested/0/build/slave/Linux_Release__NVIDIA_/build/src/tools/swarming_client/swarming.py trigger --swarming https://chromium-swarm.appspot.com --isolate-server https://isolateserver.appspot.com --priority 25 --shards 1 --task-name 'webgl_conformance_tests on NVIDIA GPU on Linux/Linux/e1d410cd15/Linux Release (NVIDIA)/74875' --dump-json /tmp/tmppSpi4r.json --expiration 3600 --io-timeout 1200 --hard-timeout 3600 --dimension cpu x86-64 --dimension gpu 10de:104a --dimension os Linux --dimension pool Chrome --tag 'buildername:Linux Release (NVIDIA)' --tag buildnumber:74875 --tag data:e1d410cd15e6c9f40bf640f704f0d4af421c2714 --tag master:chromium.gpu --tag name:webgl_conformance_tests --tag os:Linux --tag project:chromium --tag purpose:CI --tag purpose:post-commit --tag 'slavename:slave65-c1#0' --tag 'stepname:webgl_conformance_tests on NVIDIA GPU on Linux on Linux' e1d410cd15e6c9f40bf640f704f0d4af421c2714 -- webgl_conformance --show-stdout '--browser=release' -v '--extra-browser-args=--enable-logging=stderr --js-flags=--expose-gc' '--isolated-script-test-output=${ISOLATED_OUTDIR}/output.json'

(documented on https://www.chromium.org/developers/testing/gpu-testing#TOC-Running-the-GPU-Tests-Locally )

then passing e.g.:

-v --extra-browser-args="--enable-logging=stderr --js-flags=--expose-gc"

should give you the GPU info that's dumped in the bots' logs and tell you what Chromium thinks the primary GPU is on your system.

Thanks. I already found it's in fact thinking I got an nvidia in pdb. Now just to find which code does the gpu discovery and try to make it work similarly to chromium's discovery...(Btw, do you by chance know where Chromium does its GPU discovery?)

It's in src/gpu/config/gpu_info_collector*.{cc,h}.

If you provide about:gpu information and the exact model of your machine perhaps we can help.

Can you provide the about:gpu page again?

We ever investigated this bug, and it did hang or crash with old Mesa driver. But with latest Mesa code (version >= 11.3.0), the case could pass but with performance issue (It takes 7 mins to finish with Mesa, while just 1 second with NV driver on same machine). I reported bug at https://bugs.freedesktop.org/show_bug.cgi?id=94302, and Intel Mesa developer already confirmed the issue.

Yes, I already reported that in https://bugs.chromium.org/p/chromium/issues/detail?id=593680 and https://bugs.freedesktop.org/show_bug.cgi?id=94477, though I only tried it with Mesa 11.2. Maybe you could add that info to the bug report.

Anyway, I found out why it's not being skipped - it's a regression of  issue 547025 .

Deleted: gpu.html 50.9 KB

gpu.html 50.9 KB View Download

Hi, please try with current master (make sure it contains this fix https://codereview.chromium.org/1874643003/) and let us know. Thx, Julien

Sorry I forgot to close this, this has been fixed a while ago.