2022-03-09 Triage Log

A pretty smooth week. All three regressions were small, and two were isolated to rustdoc alone.

Triage done by @pnkfelix. Revision range: f0c4da49983aa699f715caf681e3154b445fb60b..10dccdc7fcbdc64ee9efe2c1ed975ab8c1d61287

3 Regressions, 3 Improvements, 1 Mixed; 2 of them in rollups 53 comparisons made in total

Regressions

Rollup of 9 pull requests #94541

  • Arithmetic mean of relevant regressions: 1.5%
  • Largest regression in instruction counts: 3.9% on full builds of deeply-nested-async check
  • After looking at the performance data, it seems like the regressions to deeply-nested-async are isolated to some slowdown in check_mod_privacy.
  • But AFAICT nothing in this rollup should have affected check_mod_privacy.
  • Chalking this up to optimization noise; marked as triaged.

Support GATs in Rustdoc #94009

  • Arithmetic mean of relevant regressions: 0.8%
  • Largest regression in instruction counts: 1.1% on full builds of ctfe-stress-4 doc
  • Already marked as triaged by rustdoc team; this change makes rustdoc track more data, which inevitably leads to perf regressions.

rustdoc: Stop textually replacing Self in doc links before resolving them #93805

  • Arithmetic mean of relevant regressions: 0.3%
  • Arithmetic mean of all relevant changes: 0.3%
  • Largest regression in instruction counts: 1.0% on full builds of serde doc
  • rustdoc team did not expect a regression here.
  • it looks like collect-intra-doc-links got slower. not clear why, though.
  • slowdown was solely observed in serde doc. posted some theories and marked as triaged.

Improvements

remove obligation dedup from impl_or_trait_obligations #84944

  • Arithmetic mean of relevant improvements: -1.4%
  • Largest improvement in instruction counts: -4.0% on full builds of deeply-nested check

Introduce ConstAllocation. #94597

  • Arithmetic mean of relevant improvements: -0.9%
  • Arithmetic mean of all relevant changes: -0.8%
  • Largest improvement in instruction counts: -1.4% on full builds of keccak check

Rollup of 6 pull requests #94761

  • Arithmetic mean of relevant improvements: -1.1%
  • Largest improvement in instruction counts: -1.2% on full builds of token-stream-stress check

Mixed

Reopen 91719 #94570

  • Arithmetic mean of relevant regressions: 0.5%
  • Arithmetic mean of relevant improvements: -1.4%
  • Arithmetic mean of all relevant changes: -1.0%
  • Largest improvement in instruction counts: -6.9% on full builds of deeply-nested-async check
  • Largest regression in instruction counts: 1.3% on incr-full builds of piston-image opt
  • Already triaged by workingjubilee when they r+'ed it.

Untriaged Pull Requests

Nags requiring follow up

TODO: Nags