Drop cold-start sample from import-time profiling#4552
Merged
alejoe91 merged 3 commits intoSpikeInterface:mainfrom Apr 22, 2026
Merged
Drop cold-start sample from import-time profiling#4552alejoe91 merged 3 commits intoSpikeInterface:mainfrom
alejoe91 merged 3 commits intoSpikeInterface:mainfrom
Conversation
alejoe91
approved these changes
Apr 22, 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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The previous PR (#4532) raised the per-sample threshold to 6.0 s as a temporary measure and added the per-sample table to the step summary so we could see the actual distribution. After looking at the data of some rounds I think it is just that the first sample is dominated by OS side-effects on running python for the first time. In Windows this must be worse, as that cost is dominated by OS-side effects (filesystem cache priming, DLL loading, Defender scanning a freshly-installed package). This is not a property of SpikeInterface and should not be measured on the import weight.
This PR adds a discarded warm-up subprocess before the timed loop for each import statement, so the table only contains steady-state samples. With the cold-start outlier removed, the per-sample threshold goes back to 3.0 s and the average gate stays at 2.0 s. The gate is now more sensitive to real regressions (a heavy import added to a top-level
__init__.pyshows up on every sample, not just the cold one), and the noise floor on Windows drops to the ~0.4 s steady-state we already see on the other runners.I have also bumped
actions/checkout,actions/cache,actions/cache/restore, andcodecov/codecov-actionto their current Node.js 24 majors across the workflows to clear the Node.js 20 deprecation warnings.