[CI][Bench] Implement exponentially weighted moving average for SYCL nightly regression CI #18766
Conversation
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
uditagarwal97
pushed a commit
that referenced
this pull request
Jul 4, 2025
Performance of SYCL has regressed a lot recently, causing nightly to fail consistently. However, if nightly is constantly failing, people may start to ignore nightly. #18766 should help with the amount of failures, but I am still figuring out the best parameters to use here, and additionally often times regressions are caused by driver changes, which is out of intel/llvm's control. Thus, for future's sake, it might be best to move benchmarking out of sycl-nightly.yml. This PR moves nightly benchmarking into its own workflow/nightly job, 1hr before sycl-nightly.yml, in order to reduce the amount of failures in sycl nightly.
|
@intel/llvm-gatekeepers PR is ready for merge, thanks! |
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Outdated suggestions cannot be applied.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.
Median does not respond very fast to changes in performance, and thus is not a suitable metric to be used for regression checking. This PR implements an option to use exponentially weighted moving average instead.
The hope is that this implementation could also be used on CPU instructions retired metric when SYCL compute-benchmark tests start reporting instructions retired; this would create a far more robust metric to use in order to spot regressions.
Note to llvm-reviewers-benchmarking: Observe that changes here are removed from the core benchmarking scripts -- This change should be functionally NFC for the core benchmarking scripts.