Perf triage, 2020-07-07.

Author: nnethercote

triage/2020.md

@@ -3,6 +3,40 @@
 See the [README](README.md) for details on how these logs are collated. They
 are in reverse chronological order.
 
+## 2020-07-07
+
+Triage done by njn. Latest revision: 0c03aee8b81185d65b5821518661c30ecdb42de5.
+One unimportant regression, on a rollup; six improvements, two on rollups.
+
+Regressions
+- [Rollup of 12 pull requests #74073](https://github.com/rust-lang/rust/pull/74073)
+  ([instructions](https://perf.rust-lang.org/compare.html?start=2753fab7ce3647033146b07c8b6c9f4856a910b0&end=0c03aee8b81185d65b5821518661c30ecdb42de5&stat=instructions:u)):
+  Up to 1.9% losses, but only on the synthetic `wf-projection-stress-65510`
+  benchmark, which doesn't matter that much.
+
+Improvements
+- [Serialize all foreign `SourceFile`s into proc-macro crate metadata #73706](https://github.com/rust-lang/rust/pull/73706)
+  ([instructions](https://perf.rust-lang.org/compare.html?start=16957bd4d3a5377263f76ed74c572aad8e4b7e59&end=d462551a8600e57d8b6f87e71ea56868bc5da6cf&stat=instructions:u)):
+  Up to 7.5% wins across numerous benchmarks, mostly the shorter-running ones.
+- [Handle inactive enum variants in `MaybeUninitializedPlaces` #73879](https://github.com/rust-lang/rust/pull/73879)
+  ([instructions](https://perf.rust-lang.org/compare.html?start=0cd7ff7ddfb75a38dca81ad3e76b1e984129e939&end=2753fab7ce3647033146b07c8b6c9f4856a910b0&stat=instructions:u)):
+  Up to 7.1% wins across numerous benchmarks, mostly on opt builds. (Plus one
+  21% improvement, but that benchmark has high variability.)
+- [Rollup of 16 pull requests #73950](https://github.com/rust-lang/rust/pull/73950)
+  ([instructions](https://perf.rust-lang.org/compare.html?start=9491f18c5de3ff1c4bf9c3fdacf52d9859e26f7c&end=b7856f695d65a8ebc846754f97d15814bcb1c244&stat=instructions:u)):
+  Up to 1.9% wins across numerous benchmarks.
+- [Avoid `unwrap_or_else` in `RawVec::allocate_in`. #73882](https://github.com/rust-lang/rust/pull/73882)
+  ([instructions](https://perf.rust-lang.org/compare.html?start=5f4abc16e1b2cb035eee6a5079ce45ce924c1f33&end=cd1a46d644791c79433db934ad4e6131c577efcc&stat=instructions:u)):
+  Up to 1.2% wins across a few benchmarks, mostly on debug builds. Maybe
+  [#73345](https://github.com/rust-lang/rust/pull/73345) or
+  [#73569](https://github.com/rust-lang/rust/pull/73569) is responsible?
+- [Remove `TypeckTables::empty(None)` and make hir_owner non-optional. #73751](https://github.com/rust-lang/rust/pull/73751)
+  ([instructions](https://perf.rust-lang.org/compare.html?start=8a6d4342be6a6acbade8e7ef65e73d27ee8c9144&end=3503f565e1fb7296983757d2716346f48a4a262b&stat=instructions:u)):
+  Up to 3.8% wins, mostly on `unused-warnings`.
+- [Rollup of 17 pull requests #73924](https://github.com/rust-lang/rust/pull/73924)
+  ([instructions](https://perf.rust-lang.org/compare.html?start=d462551a8600e57d8b6f87e71ea56868bc5da6cf&end=f781babf87dea29c44f93842b7ac9eb809549d29&stat=instructions:u)):
+  Up to 1.9% wins, mostly on `clap-rs`. Unclear which PR is responsible.
+
 ## 2020-06-30
 
 Triage done by njn. Latest revision: 0ca7f74dbd23a3e8ec491cd3438f490a3ac22741.

triage/README.md

@@ -32,6 +32,8 @@ Look for significant changes (regressions or improvements) in the graphs.
 - Click on a data point to open the "compare" page for that merge.
 - Click on the "compare" link at the top of the measurements on that page to
   open the page of commits in the merge.
+- A change isn't significant unless one or more of the benchmarks changed by at
+  least 1%.
 
 Easy cases: there is only a single PR in the merge.
 - Add a comment to the PR pointing to the "compare" page (unless someone else
@@ -57,10 +59,9 @@ Difficult cases: the merge was a rollup of multiple PRs.
   that are likely to affect performance.
 - Add an entry to the triage log, as for the easy cases.
 
-Repeat with the [`max-rss`
-graphs](https://perf.rust-lang.org/?start=&end=&absolute=true&stat=max-rss).
-These measurements are much noisier than `instructions:u`, so only larger
-changes will be clear.
+Repeat with the `max-rss` graphs. These measurements are much noisier than
+`instructions:u`, so only larger changes will be clear; recheck the "Absolute
+data" checkbox otherwise the noise becomes unmanageable.
 
 Once finished, file a PR adding a link to the log entry in [This Week in
 Rust](https://github.com/emberian/this-week-in-rust/).