Repository navigation
Tighten the benchmark regression threshold to 10% and gate more cases - #3978
Merged
Merged
Conversation
Three runs on the GNR benchmark runner (two on one commit, one a day later after frontend-only changes) differ by at most 3% in every gated case, so a 25% threshold would miss most real regressions. Lower it to 10%. 09_shuffle on all targets and 13_rotate/14_shift on avx2 targets differ by at most 1.5%, as little as the gated cases. The stack alignment modes that kept the avx2 shifts out were seen on the Ice Lake machine and don't show up with setarch -R and env -i. Gate them. 05_packed_load_store (up to 13.5%) and 17_pdep_pext (up to 8.7% on sse4) stay not gated. The threshold is only applied when BENCHMARKS_FAIL_ON_REGRESSION is true: with both fail-on-alert and comment-on-alert off, github-action-benchmark doesn't check the cases. Say so in the workflow. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
azwolski
approved these changes
Oct 26, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Three runs on the GNR benchmark runner differ by at most 3% in every gated case, so a 25% threshold would miss most real regressions. Lower it to 10%.
09_shuffle on all targets and 13_rotate/14_shift on avx2 targets differ by at most 1.5%, as little as the gated cases. Gate them. 05_packed_load_store (up to 13.5%) and 17_pdep_pext (up to 8.7% on sse4) stay not gated.
The threshold is only applied when BENCHMARKS_FAIL_ON_REGRESSION is true: with both fail-on-alert and comment-on-alert off, github-action-benchmark doesn't check the cases.
Related Issue
Checklist
clang-format(e.g.,clang-format -i src/ispc.cpp)