Open Bug 1850537 (webgpu-narrow-long-timeouts) Opened 1 year ago Updated 1 year ago

Narrow long timeouts to only where it's needed in WebGPU CTS in CI

Categories

(Core :: Graphics: WebGPU, task, P3)

task

Tracking

()

People

(Reporter: ErichDonGubler, Unassigned, Mentored, NeedInfo)

References

(Depends on 1 open bug)

Details

Some tests in WebGPU CTS are big, because they are composed of large matrices of test parameters. Texture tests, for instance, almost all have large set of texture formats to work with, easily generating hundreds or thousands of test cases with a few parameters.

In bug 1850356, long timeouts were added to WebGPU CTS runs in CI, in order to allow more of these tests to finish executing before timing out. The blanket application of long timeouts was decided to be worth the risk of extra time cost that we may incurring from tests that legitimately time out, which let us skip applying long timeouts in a more fine-grained way.

This bug represents the skipped effort from above; we want to narrow long timeouts to only the tests that need it. In order to do this, we'll need to have a means of manually grouping tests (from bug 1822630). With this, we should be able to isolate long-running tests to their own group(s) with long timeouts, and remove long timeouts from the rest of the test suite.

:jimb: I think this can be prioritized for even after phase 2; I don't see this being important enough to do for a long while yet (CC :nical). NI'ing for confirmation of this prioritization.

Flags: needinfo?(jimb)
You need to log in before you can comment on or make changes to this bug.