Use gpu_cpp_unittest for slim CUDA guard tests#19378
Use gpu_cpp_unittest for slim CUDA guard tests#19378rascani wants to merge 1 commit intopytorch:mainfrom
Conversation
Summary: The `test_cuda_guard` and `test_cuda_stream_guard` targets in `executorch/backends/aoti/slim/cuda/test/` skip at runtime via `GTEST_SKIP()` in `SetUp()` when no CUDA device is available. On CPU-only sandcastle hosts they emit SKIPPED on every run, never accumulate a PASS record, and TestInfra flags them with a `CONTINUOUS_STATE` violation, marking them DISABLED_FAILING and contributing 15 phantom failures to the ai_infra_mobile_platform oncall feed. Switch the wrapper macro from `cpp_unittest` to `gpu_cpp_unittest`. The GPU wrapper expands the kwargs through `gpu_common.convert_cpp_args`, which adds CI labels and the appropriate `select_accelerator()` so Buck marks the targets incompatible with non-GPU configurations. TestInfra then omits them from CPU schedules entirely instead of running and reporting SKIPPED. Authored with Claude. Differential Revision: D104297130
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/19378
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New FailureAs of commit a8812b4 with merge base fa857bd ( NEW FAILURE - The following job has failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
@rascani has exported this pull request. If you are a Meta employee, you can view the originating Diff in D104297130. |
meta-cla
Bot
added
the
CLA Signed
This PR needs a
|
1 participant
Summary:
The
test_cuda_guardandtest_cuda_stream_guardtargets inexecutorch/backends/aoti/slim/cuda/test/skip at runtime viaGTEST_SKIP()inSetUp()when no CUDA device is available. On CPU-only sandcastle hosts they emit SKIPPED on every run, never accumulate a PASS record, and TestInfra flags them with aCONTINUOUS_STATEviolation, marking them DISABLED_FAILING and contributing 15 phantom failures to the ai_infra_mobile_platform oncall feed.Switch the wrapper macro from
cpp_unittesttogpu_cpp_unittest. The GPU wrapper expands the kwargs throughgpu_common.convert_cpp_args, which adds CI labels and the appropriateselect_accelerator()so Buck marks the targets incompatible with non-GPU configurations. TestInfra then omits them from CPU schedules entirely instead of running and reporting SKIPPED.Authored with Claude.
Differential Revision: D104297130