Benchmark regression tracker

[cplieger]

Benchmark regression tracker

Weekly medians for plex-language-sync, compared against the median of the previous 6 runs.
Updated 2026-09-02 01:26 UTC by this run.

Full series and charts: benchmarks branch.

Trend

Run	Commit	Benchmarks	Regressed	Improved
2026-09-02 00:25	0fe13335	18	0	0
2026-08-25 23:04	1fdf3206	18	0	0

Findings

No benchmark moved past its threshold this week. 🎉

How to read this

Each run measures every func Benchmark in the repo, takes 10 samples, and
reduces them to one median per benchmark per metric. A finding compares that
median against the median of the previous 6 runs.

Metric	Flagged when	Why this threshold
ns/op	1.5x slower	Wall clock on a shared runner varies 10-20% by itself, so only a multiple is evidence. The target is an algorithmic regression.
B/op, allocs/op	1.05x higher	Counted, not timed, so they do not move with runner load. A real increase is a real change.
any	zero to non-zero	An allocation-free path that starts allocating. The ratio is undefined and the change is the point.

Nothing here gates a merge or fails a build. A flagged wall-clock finding is
evidence to investigate, not proof the code got slower; check the ± spread
before believing a single run.

Free-form notes

Add anything below — won't be touched by the auto-updater.