2021-09-07 Triage Log

A busy week, with lots of mixed changes, though in the end only a few were deemed significant enough to report here.

Triage done by @pnkfelix. Revision range: fe37929e4cba2c5c21e6805805769630c736bc3d..69c4aa2901ffadf69deaf91b2f90604bcbc2eb36

3 Regressions, 1 Improvements, 3 Mixed; 0 of them in rollups 57 comparisons made in total

Regressions

Shrink Session a bit #88530

Concrete regions can show up in mir borrowck if the originated from there #88533

  • Very small regression in instruction counts (up to 1.0% on incr-patched: add static arr item builds of coercions)
  • This is a small regression to coercions and that may be noise. However, there are many others that are over 0.4% regression to instruction counts. The combination of those two factors led pnkfelix to think that we may want to take a second look at the effects of this PR.

Introduce let...else #87688

  • Small improvement in instruction counts (up to -0.3% on incr-unchanged builds of deeply-nested-closures)
  • Small regression in instruction counts (up to 0.4% on full builds of issue-46449)
  • Small regression (approximately 1%) in instruction counts on several of the *-doc tests, found via human eye.

Improvements

Move global analyses from lowering to resolution #88597

  • Moderate improvement in instruction counts (up to -1.6% on incr-unchanged builds of keccak)

Mixed

Preserve most sub-obligations in the projection cache #85868

  • Very large improvement in instruction counts (up to -6.9% on incr-unchanged builds of deeply-nested)
  • Very large regression in instruction counts (up to 9.4% on full builds of deeply-nested)

BTreeMap/BTreeSet::from_iter: use bulk building to improve the performance #88448

Avoid invoking the hir_crate query to traverse the HIR #88435

  • Small improvement in instruction counts (up to -1.2% on incr-patched: add static arr item builds of coercions)
  • Small regression in instruction counts (up to 0.5% on incr-unchanged builds of helloworld)
  • pnkfelix included this one as notable largely because the expected perf changes from the actual PR largely showed slight improvements across the board, while after it landed on nightly, it is best categorized as “Mixed”

Untriaged Pull Requests