Skip to content

[Docs] Point the README quick-check at the GEMM correctness test#647

Merged
coderfeli merged 1 commit into
ROCm:mainfrom
Arist12:docs/readme-quick-env-check
Jun 4, 2026
Merged

[Docs] Point the README quick-check at the GEMM correctness test#647
coderfeli merged 1 commit into
ROCm:mainfrom
Arist12:docs/readme-quick-env-check

Conversation

@Arist12
Copy link
Copy Markdown
Contributor

@Arist12 Arist12 commented Jun 3, 2026

Summary

The README "Run tests" section labels bash scripts/run_tests.sh as a quick post-install check:

# Run GEMM correctness tests (fast, ~15s)
bash scripts/run_tests.sh

But run_tests.sh runs the entire test suite, which takes many minutes — the "~15s GEMM correctness" label is inaccurate and misleads contributors validating a fresh build. This PR points the line at the actual GEMM correctness test, which runs the real GPU path in ~10s and matches the original intent of the comment.

Closes #646

Findings

  • run_tests.sh runs the full suite, in three phases: pytest over tests/kernels|unit|system|python/examples (3118 tests selected with the default -m "not large_shape"), then all standalone examples/
    scripts, then every tests/mlir/**/*.mlir FileCheck test.
  • Timings rule out ~15s for the full suite: a full run did not complete within a 300s cap; tests/kernels/test_allreduce.py alone (8-GPU accuracy tests, @pytest.mark.multi_gpu) did not finish 6 parametrizations in 150s on an 8-GPU host. Those multi-GPU tests auto-skip on <4 GPUs but run (and dominate) on larger hosts.
  • The label is stale, but ~15s was once right for GEMM: git blame traces the comment to the v0.1 PR (Pr/v0.1 #164), when the suite was small. The targeted GEMM correctness test still runs in ~10s today, so keeping the comment and swapping the command restores its accuracy.

Change (one line)

Replace bash scripts/run_tests.sh with the GEMM correctness test:

# Run GEMM correctness tests (fast, ~15s)
python -m pytest tests/kernels/test_preshuffle_gemm.py -m "not large_shape"

Verification

$ python -m pytest tests/kernels/test_preshuffle_gemm.py -m "not large_shape" 43 passed, 4 skipped, 76 deselected in 9.19s

@Arist12 Arist12 marked this pull request as ready for review June 3, 2026 12:47
The Run tests section attached a "~15s GEMM correctness" comment to
`bash scripts/run_tests.sh`, but that script runs the entire suite (3000+
tests plus example scripts and MLIR FileCheck) and takes many minutes, so
the label is misleading for a post-install sanity check.

Point the line at the actual GEMM correctness test instead. It runs the
real GPU MMA path in ~10s (`not large_shape`), matches the original intent
of the comment, and avoids the full suite (including the 8-GPU allreduce
tests). See ROCm#646.
@Arist12 Arist12 force-pushed the docs/readme-quick-env-check branch from a927e3c to 3800ee0 Compare June 4, 2026 06:37
@coderfeli coderfeli merged commit 56a579e into ROCm:main Jun 4, 2026
@coderfeli
Copy link
Copy Markdown
Collaborator

Thanks @Arist12

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[Documentation]: run_tests.sh is labeled a ~15s GEMM check but runs the full suite

2 participants