Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add NVRTC_SKIP_KERNEL_RUN tag to compile, but skip running NVRTC test #434

Merged
merged 3 commits into from
Sep 18, 2023

Conversation

ahendriksen
Copy link
Contributor

Description

closes #433

This PR adds the NVRTC_SKIP_KERNEL_RUN tag that, if present in the source code of a test, changes how the test is executed under NVRTC. Specifically:

  • it is still compiled under NVRTC
  • but running device code is skipped

Checklist

  • New or existing tests cover these changes.
  • The documentation is up to date with these changes. (Perhaps we should have some documentation of the test framework?)

Some kernels using TMA need some host code to run as well. In addition,
they need the device symbols to be accessible from host code. This turns
out to be quite difficult to achieve. Therefore, since the device code
is already executed in the non-nvrtc tests, it is reasonable to allow a
test to skip running device code when NVRTC is enabled. This still tests
whether the code actually compiles.
@ahendriksen ahendriksen requested review from a team as code owners September 12, 2023 11:54
@ahendriksen ahendriksen requested review from ericniebler and gevtushenko and removed request for a team September 12, 2023 11:54
@ahendriksen
Copy link
Contributor Author

ahendriksen commented Sep 12, 2023

TODO: Once #379 is merged. The bulk tensor tests should be enabled for NVRTC again.

libcudacxx/.upstream-tests/test/cuda/barrier/cp_async_bulk_tensor_{1,2..,5}d.pass.cpp

@miscco miscco merged commit 6be81eb into NVIDIA:main Sep 18, 2023
463 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
Archived in project
Development

Successfully merging this pull request may close these issues.

[FEA]: Allow tests to skip kernel launch when run with NVRTC
3 participants