2021-02-02 Triage Log

Another week dominated by rollups, most of which had relatively small changes with unclear causes embedded. Overall no major changes in performance this week.

Triage done by @simulacrum. Revision range: 1483e67addd37d9bd20ba3b4613b678ee9ad4d68..f6cb45ad01a4518f615926f39801996622f46179

2 Regressions, 1 Improvements, 1 Mixed 3 of them in rollups

Regressions

Rollup of 12 pull requests #81625

• Moderate regression in instruction counts (up to 3.3% on incr-unchanged builds of match-stress-enum-check)
• Unclear responsible PR. Nothing in this rollup stands out, and self-profile information does not reveal obvious causes.

Add visitors for checking #[inline] #80641

• Moderate regression in instruction counts (up to 1.7% on incr-unchanged builds of unicode_normalization-check)
• Primarily limited to incremental benchmarks; we are simply visiting more nodes for this check, so it is unsurprising that we take a performance hit.

Improvements

Rollup of 13 pull requests #81461

• Large improvement in instruction counts (up to -5.7% on incr-unchanged builds of deeply-nested-async-check)
• Unknown causes; no major investigation.

Mixed

Rollup of 10 pull requests #81493

• Moderate regression in instruction counts (up to 1.0% on incr-patched: b9b3e592dd cherry picked builds of style-servo-debug)
• Also a moderate improvement in instruction counts, around 0.3%.
• Not clear what is responsible, but the regression is very limited to a few crates and incremental, so not spending too much time investigating.

Nags requiring follow up

From last week:

• Deprecate in future regression #80958
  • Some commentary in response but seems likely this will not be addressed one way or another.
• Waiting on response to inquiry.
  • Some investigation in #81476, potentially regression was down to a bugfix.
• Regression in prevent query cycles in the MIR inliner - expected? #68828
  • No updates yet from @oli-obk on their investigation.