src/tools/rustc-perf/triage/2020-07-21.md - toolchain/rustc - Git at Google

 # 2020-07-21 Triage Log

 See the [README](README.md) for details on how this log was gathered.

 This week was a disaster, perf-wise. 28 revisions checked. 7
 regressions, several of them ranging from large to huge, many in rollups. Some
 additional regressions may have occurred in rollups that were masked by other
 regressions/improvements. 3 improvements, one of which was a reversion of a
 regression. Thanks for Mark-Simulacrum and eddyb for follow-up measurements and
 adding new tooling to help investigate regressions in rollups. A follow-up
 thread on Zulip is
 [here](https://rust-lang.zulipchat.com/#narrow/stream/247081-t-compiler.2Fperformance/topic/Follow-up.20to.20the.20bad.20perf.20triage.20of.202020-07-21).

 In better news, rustdoc performance is now being benchmarked, thanks to the
 efforts of [Joshua Nelson](https://github.com/rust-lang/rustc-perf/pull/675).

 Triage done by njn. Revision range: [9d09331e00b02f81c714b0c41ce3a38380dd36a2..71384101ea3b030b80f7def80a37f67e148518b0](https://perf.rust-lang.org/?start=9d09331e00b02f81c714b0c41ce3a38380dd36a2&end=71384101ea3b030b80f7def80a37f67e148518b0&absolute=false&stat=instructions%3Au).

 Regressions
 - [Rollup of 18 pull requests #74461](https://github.com/rust-lang/rust/pull/74461#issuecomment-660492867)
   ([instructions](https://perf.rust-lang.org/compare.html?start=39d5a61f2e4e237123837f5162cc275c2fd7e625&end=d3df8512d2c2afc6d2e7d8b5b951dd7f2ad77b02-v1&stat=instructions:u)):
   Up to 73.0% losses on many benchmarks.
   [#74416](https://github.com/rust-lang/rust/pull/74416#issuecomment-660545238)
   is at fault; it was reverted in
   [#74478](https://github.com/rust-lang/rust/pull/74478), see below.
 - [std: Switch from libbacktrace to gimli #73441](https://github.com/rust-lang/rust/pull/73441#issuecomment-660881380)
   ([instructions](https://perf.rust-lang.org/compare.html?start=7d31ffc1ac9e9ea356e896e63307168a64501b9d&end=1fa54ad9680cc82e7301f8ed4e9b7402dfd6ce0e&stat=instructions:u),
    [max-rss](https://perf.rust-lang.org/compare.html?start=7d31ffc1ac9e9ea356e896e63307168a64501b9d&end=1fa54ad9680cc82e7301f8ed4e9b7402dfd6ce0e&stat=max-rss)):
   Up to 44.8% instruction increases on many benchmarks, and 5-10% max-rss increases on many benchmarks. Being reverted in [#74613](https://github.com/rust-lang/rust/pull/74613).
 - [Support const args in type dependent paths (Take 2) #74113](https://github.com/rust-lang/rust/pull/74113#issuecomment-661470398)
   ([instructions](https://perf.rust-lang.org/compare.html?start=d9e8d6290745a65025a3e082aea72fbe372292c6&end=7e11379f3b4c376fbb9a6c4d44f3286ccc28d149&stat=instructions:u)):
   Up to 1.5% losses on numerous benchmarks, which was expected.
 - [Reduce the amount of interning and `layout_of` calls in const eval. #74202](https://github.com/rust-lang/rust/pull/74202#issuecomment-661475728)
   ([instructions](https://perf.rust-lang.org/compare.html?start=4cd0ee9343da86d9770bf0a514a682d240e0dce8&end=125c58caebc67c32ec45ac6c0581b596fd532082&stat=instructions:u)):
   Up to 2% losses on one benchmark.
 - [Rollup of 9 pull requests #74543](https://github.com/rust-lang/rust/pull/74543#issuecomment-661507295)
   ([instructions](https://perf.rust-lang.org/compare.html?start=2c21a6f3a8b1c75c444b87fde5116853383b3fbd&end=891e6fee572009ff2be4d4057fb33483610c36a7&stat=instructions:u)):
   Up to 12% losses on rustdoc for many benchmarks.
 - [Rollup of 4 pull requests #74518](https://github.com/rust-lang/rust/pull/74518#issuecomment-661498214)
   ([instructions](https://perf.rust-lang.org/compare.html?start=48036804d2bc461b243c5d291b850e44bcca68ef&end=d7f94516345a36ddfcd68cbdf1df835d356795c3&stat=instructions:u)):
   Up to 6.4% losses on rustdoc for two benchmarks.

 Improvements
 - [Rollup of 7 pull requests #74493](https://github.com/rust-lang/rust/pull/74493#issuecomment-661521298)
   ([instructions](https://perf.rust-lang.org/compare.html?start=1fa54ad9680cc82e7301f8ed4e9b7402dfd6ce0e&end=0701419e96d94e5493c7ebfcecb66511ab0aa778&stat=instructions:u)):
   Up to 33.5% improvements, at least partly due to the backout of
   [#74416](https://github.com/rust-lang/rust/pull/74416) in
   [#74478](https://github.com/rust-lang/rust/pull/74478), but it's also
   possible that another PR in that rollup [caused a regression that was masked
   by the improvement from the backout](https://github.com/rust-lang/rust/pull/74493#issuecomment-661521298).
 - [Change `SymbolName::name` to a `&str`. #74214](https://github.com/rust-lang/rust/pull/74214#issuecomment-661459141)
   ([instructions](https://perf.rust-lang.org/compare.html?start=c714eae0e3b4ba263c193d54f6e46bb9a1feb2cc&end=d9e8d6290745a65025a3e082aea72fbe372292c6&stat=instructions:u)):
   Up to 2.5% wins on numerous benchmarks.
 - [Don't run `everybody_loops` for rustdoc; instead ignore resolution errors #73566](https://github.com/rust-lang/rust/pull/73566#issuecomment-661484787)
   ([instructions](https://perf.rust-lang.org/compare.html?start=6ee1b62c811a6eb68d6db6dfb91f66a49956749b&end=5c9e5df3a097e094641f16dab501ab1c4da10e9f&stat=instructions:u)):
   Wins of up to 62.6% and losses of up to 8.5%, all on rustdoc builds. Overall,
   the improvements greatly outweigh the losses. (Landed in rollup
   [#74408](https://github.com/rust-lang/rust/pull/74408).)

 ## Notes

 [Rollup of 11 pull requests #74468](https://github.com/rust-lang/rust/pull/74468#issuecomment-661371815)
  * ([instructions](https://perf.rust-lang.org/compare.html?start=d3df8512d2c2afc6d2e7d8b5b951dd7f2ad77b02-v1&end=7d31ffc1ac9e9ea356e896e63307168a64501b9d&stat=instructions:u))

 This rollup was originally judged as responsible for a 10% regression in
 instrutions:u. However, since then, it has been determined that the likely cause
 of this regression is actually perf's
 [move](https://github.com/rust-lang/rustc-perf/commit/048360c77a74244ba6e70d9b3f2bcd7779b8129f) to using `x.py dist`-shipped std's
 rather than building one locally. Investigation into *why* this move caused a
 regression is as yet not done, but is being tracked in [rustc-perf#724](https://github.com/rust-lang/rustc-perf/issues/724).

 Initially [#74069](https://github.com/rust-lang/rust/pull/74069) and/or
 [#72414](https://github.com/rust-lang/rust/pull/72414) were suspected
 as the cause of the regression, but further testing showed that to not be the
 case.

  * The original rollup, #74468, landed with a [5-10% performance regression](https://perf.rust-lang.org/compare.html?start=d3df8512d2c2afc6d2e7d8b5b951dd7f2ad77b02-v1&end=7d31ffc1ac9e9ea356e896e63307168a64501b9d&stat=instructions:u)
  * In #74611, we tested as-if a rollup of #74069 and #72414 landed. ([This](https://gist.github.com/Mark-Simulacrum/6893dff239d116947bca647f4f8128c7) is the diff between d3df85 and cfade73) That yielded [identical results to the rollup](https://perf.rust-lang.org/compare.html?start=d3df8512d2c2afc6d2e7d8b5b951dd7f2ad77b02-v1&end=cfade73820883adf654fe34fd6b0b03a99458a51).
  * We reverted #74069 in #74611, this yielded [neutral performance results](https://perf.rust-lang.org/compare.html?start=e8b55a4ad230ebec762fdfc4f241ba98a98560af&end=fcac11993ca055bbdc7683a2f6ed7b88a838fb0f&stat=instructions:u).
  * #74716 then tried a revert of #72414 but this also yielded [neutral performance results](https://perf.rust-lang.org/compare.html?start=900869371e13cead086f4f9809419daa6a63cfaf&end=193b2f77c9463ed7378c20bad843a9031489e215)

 We have since opened a PR to re-land #74069, as well: #74802.
	# 2020-07-21 Triage Log

	See the [README](README.md) for details on how this log was gathered.

	This week was a disaster, perf-wise. 28 revisions checked. 7
	regressions, several of them ranging from large to huge, many in rollups. Some
	additional regressions may have occurred in rollups that were masked by other
	regressions/improvements. 3 improvements, one of which was a reversion of a
	regression. Thanks for Mark-Simulacrum and eddyb for follow-up measurements and
	adding new tooling to help investigate regressions in rollups. A follow-up
	thread on Zulip is
	[here](https://rust-lang.zulipchat.com/#narrow/stream/247081-t-compiler.2Fperformance/topic/Follow-up.20to.20the.20bad.20perf.20triage.20of.202020-07-21).

	In better news, rustdoc performance is now being benchmarked, thanks to the
	efforts of [Joshua Nelson](https://github.com/rust-lang/rustc-perf/pull/675).

	Triage done by njn. Revision range: [9d09331e00b02f81c714b0c41ce3a38380dd36a2..71384101ea3b030b80f7def80a37f67e148518b0](https://perf.rust-lang.org/?start=9d09331e00b02f81c714b0c41ce3a38380dd36a2&end=71384101ea3b030b80f7def80a37f67e148518b0&absolute=false&stat=instructions%3Au).

	Regressions
	- [Rollup of 18 pull requests #74461](https://github.com/rust-lang/rust/pull/74461#issuecomment-660492867)
	([instructions](https://perf.rust-lang.org/compare.html?start=39d5a61f2e4e237123837f5162cc275c2fd7e625&end=d3df8512d2c2afc6d2e7d8b5b951dd7f2ad77b02-v1&stat=instructions:u)):
	Up to 73.0% losses on many benchmarks.
	[#74416](https://github.com/rust-lang/rust/pull/74416#issuecomment-660545238)
	is at fault; it was reverted in
	[#74478](https://github.com/rust-lang/rust/pull/74478), see below.
	- [std: Switch from libbacktrace to gimli #73441](https://github.com/rust-lang/rust/pull/73441#issuecomment-660881380)
	([instructions](https://perf.rust-lang.org/compare.html?start=7d31ffc1ac9e9ea356e896e63307168a64501b9d&end=1fa54ad9680cc82e7301f8ed4e9b7402dfd6ce0e&stat=instructions:u),
	[max-rss](https://perf.rust-lang.org/compare.html?start=7d31ffc1ac9e9ea356e896e63307168a64501b9d&end=1fa54ad9680cc82e7301f8ed4e9b7402dfd6ce0e&stat=max-rss)):
	Up to 44.8% instruction increases on many benchmarks, and 5-10% max-rss increases on many benchmarks. Being reverted in [#74613](https://github.com/rust-lang/rust/pull/74613).
	- [Support const args in type dependent paths (Take 2) #74113](https://github.com/rust-lang/rust/pull/74113#issuecomment-661470398)
	([instructions](https://perf.rust-lang.org/compare.html?start=d9e8d6290745a65025a3e082aea72fbe372292c6&end=7e11379f3b4c376fbb9a6c4d44f3286ccc28d149&stat=instructions:u)):
	Up to 1.5% losses on numerous benchmarks, which was expected.
	- [Reduce the amount of interning and `layout_of` calls in const eval. #74202](https://github.com/rust-lang/rust/pull/74202#issuecomment-661475728)
	([instructions](https://perf.rust-lang.org/compare.html?start=4cd0ee9343da86d9770bf0a514a682d240e0dce8&end=125c58caebc67c32ec45ac6c0581b596fd532082&stat=instructions:u)):
	Up to 2% losses on one benchmark.
	- [Rollup of 9 pull requests #74543](https://github.com/rust-lang/rust/pull/74543#issuecomment-661507295)
	([instructions](https://perf.rust-lang.org/compare.html?start=2c21a6f3a8b1c75c444b87fde5116853383b3fbd&end=891e6fee572009ff2be4d4057fb33483610c36a7&stat=instructions:u)):
	Up to 12% losses on rustdoc for many benchmarks.
	- [Rollup of 4 pull requests #74518](https://github.com/rust-lang/rust/pull/74518#issuecomment-661498214)
	([instructions](https://perf.rust-lang.org/compare.html?start=48036804d2bc461b243c5d291b850e44bcca68ef&end=d7f94516345a36ddfcd68cbdf1df835d356795c3&stat=instructions:u)):
	Up to 6.4% losses on rustdoc for two benchmarks.

	Improvements
	- [Rollup of 7 pull requests #74493](https://github.com/rust-lang/rust/pull/74493#issuecomment-661521298)
	([instructions](https://perf.rust-lang.org/compare.html?start=1fa54ad9680cc82e7301f8ed4e9b7402dfd6ce0e&end=0701419e96d94e5493c7ebfcecb66511ab0aa778&stat=instructions:u)):
	Up to 33.5% improvements, at least partly due to the backout of
	[#74416](https://github.com/rust-lang/rust/pull/74416) in
	[#74478](https://github.com/rust-lang/rust/pull/74478), but it's also
	possible that another PR in that rollup [caused a regression that was masked
	by the improvement from the backout](https://github.com/rust-lang/rust/pull/74493#issuecomment-661521298).
	- [Change `SymbolName::name` to a `&str`. #74214](https://github.com/rust-lang/rust/pull/74214#issuecomment-661459141)
	([instructions](https://perf.rust-lang.org/compare.html?start=c714eae0e3b4ba263c193d54f6e46bb9a1feb2cc&end=d9e8d6290745a65025a3e082aea72fbe372292c6&stat=instructions:u)):
	Up to 2.5% wins on numerous benchmarks.
	- [Don't run `everybody_loops` for rustdoc; instead ignore resolution errors #73566](https://github.com/rust-lang/rust/pull/73566#issuecomment-661484787)
	([instructions](https://perf.rust-lang.org/compare.html?start=6ee1b62c811a6eb68d6db6dfb91f66a49956749b&end=5c9e5df3a097e094641f16dab501ab1c4da10e9f&stat=instructions:u)):
	Wins of up to 62.6% and losses of up to 8.5%, all on rustdoc builds. Overall,
	the improvements greatly outweigh the losses. (Landed in rollup
	[#74408](https://github.com/rust-lang/rust/pull/74408).)

	## Notes

	[Rollup of 11 pull requests #74468](https://github.com/rust-lang/rust/pull/74468#issuecomment-661371815)
	* ([instructions](https://perf.rust-lang.org/compare.html?start=d3df8512d2c2afc6d2e7d8b5b951dd7f2ad77b02-v1&end=7d31ffc1ac9e9ea356e896e63307168a64501b9d&stat=instructions:u))

	This rollup was originally judged as responsible for a 10% regression in
	instrutions:u. However, since then, it has been determined that the likely cause
	of this regression is actually perf's
	[move](https://github.com/rust-lang/rustc-perf/commit/048360c77a74244ba6e70d9b3f2bcd7779b8129f) to using `x.py dist`-shipped std's
	rather than building one locally. Investigation into why this move caused a
	regression is as yet not done, but is being tracked in [rustc-perf#724](https://github.com/rust-lang/rustc-perf/issues/724).

	Initially [#74069](https://github.com/rust-lang/rust/pull/74069) and/or
	[#72414](https://github.com/rust-lang/rust/pull/72414) were suspected
	as the cause of the regression, but further testing showed that to not be the
	case.

	* The original rollup, #74468, landed with a [5-10% performance regression](https://perf.rust-lang.org/compare.html?start=d3df8512d2c2afc6d2e7d8b5b951dd7f2ad77b02-v1&end=7d31ffc1ac9e9ea356e896e63307168a64501b9d&stat=instructions:u)
	* In #74611, we tested as-if a rollup of #74069 and #72414 landed. ([This](https://gist.github.com/Mark-Simulacrum/6893dff239d116947bca647f4f8128c7) is the diff between d3df85 and cfade73) That yielded [identical results to the rollup](https://perf.rust-lang.org/compare.html?start=d3df8512d2c2afc6d2e7d8b5b951dd7f2ad77b02-v1&end=cfade73820883adf654fe34fd6b0b03a99458a51).
	* We reverted #74069 in #74611, this yielded [neutral performance results](https://perf.rust-lang.org/compare.html?start=e8b55a4ad230ebec762fdfc4f241ba98a98560af&end=fcac11993ca055bbdc7683a2f6ed7b88a838fb0f&stat=instructions:u).
	* #74716 then tried a revert of #72414 but this also yielded [neutral performance results](https://perf.rust-lang.org/compare.html?start=900869371e13cead086f4f9809419daa6a63cfaf&end=193b2f77c9463ed7378c20bad843a9031489e215)

	We have since opened a PR to re-land #74069, as well: #74802.