2020-12-08 Triage Log

Quiet week.

Triage done by @simulacrum. Revision range: c7cff213e937c1bb301be807ce04fcf6092b9163..4fd4a98d4788bc987d7f7add9df5f5ead6a1c15e

0 Regressions, 2 Improvements, 1 Mixed 0 of them in rollups

Regressions

Improvements

Moderate improvement in instruction counts (up to -1.7% on incr-full builds of ctfe-stress-4-debug)
Likely due to the stress test not reflecting benefits of memoizing CTFE, which this partially removed (due to the addition of heap allocation, which should not be deduplicated).

Very large improvement in instruction counts (up to -20.6% on full builds of match-stress-enum-check)
Fixes a regression from last week by adding a #[inline] attribute on some hot code. Improvements are likely not significant outside stress tests.

Mixed

Very large improvement in instruction counts (up to -28.4% on incr-patched: println builds of clap-rs-debug)
Large regression in instruction counts (up to 9.9% on incr-patched: println builds of regression-31157-opt)
Soundness fix (“Don't leak return value after panic in drop”) and the perf results are mixed, largest ones mostly in incremental and generally likely just “generating more (necessary) LLVM IR”.

Nags requiring follow up

stdarch expansion causing a 40% libcore compile time regression is still not resolved, and resolution is unclear. It seems likely that this could be one of our key elements for improving std compile times at least, though.
No new nags.