Compare distributed revisions with one benchmark harness - #24
Open
jeregrine wants to merge 1 commit into
Open
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
The distributed comparison runs each revision's own benchmark implementation. Existing release and candidate harnesses already use different completion barriers and timing boundaries, so the reports mix library performance changes with changes in what is measured.
Fix
Run the candidate's distributed harness for both revisions, changing only its Group path dependency. Give each run a fresh, separate build directory so baseline code cannot leak into the candidate measurement. Keep the per-revision deadline, separate reports, benchmark options, and failure propagation through
tee.The same comparison launcher is available outside CI.
Supporting information
Historical baselines must satisfy the common workload rather than silently switching to their older measurement logic. In particular,
v0.2.1cannot complete the current Group-restart/discovery barrier; this remains a failed baseline scenario rather than an incomparable timing result.