2023-11-14 Triage Log

A week dominated by one particular perf improvement that lead to huge performance gains - an avg of 5% improvement across 121 test cases! The perf improvement comes from adding an #[inline] hint to the output from #[derive(Debug)] which presumably allows the compiler to more easily do deadcode elimination reducing the binary size and the amount of code that actually needs to be code-gened.

Triage done by @rylev. Revision range: 7b97a5ca..173b6e68

Summary:

(instructions:u)	mean	range	count
Regressions ❌ (primary)	0.4%	[0.2%, 0.9%]	10
Regressions ❌ (secondary)	1.9%	[0.2%, 3.6%]	12
Improvements ✅ (primary)	-5.6%	[-49.2%, -0.1%]	111
Improvements ✅ (secondary)	-3.5%	[-25.0%, -0.2%]	155
All ❌✅ (primary)	-5.1%	[-49.2%, 0.9%]	121

2 Regressions, 2 Improvements, 3 Mixed; 3 of them in rollups 55 artifact comparisons made in total

Regressions

Bump libc dependency #117617 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	8.3%	[1.1%, 19.0%]	9
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	8.3%	[1.1%, 19.0%]	9

• A curious perf infra glitch - the regression returned to normal after a few runs.

Rollup of 6 pull requests #117736 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	0.4%	[0.2%, 0.5%]	5
Regressions ❌ (secondary)	0.6%	[0.6%, 0.6%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.4%	[0.2%, 0.5%]	5

• The only PR that includes actual changes to the compiler is #117645. Kicked off a test to see if that's the root cause.

Improvements

Rollup of 4 pull requests #117680 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-9.1%	[-16.0%, -2.6%]	7
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-9.1%	[-16.0%, -2.6%]	7

• The flip side of #117617's strange perf regression.

Add std::hash::{DefaultHasher, RandomState} exports (needs FCP) #115694 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.9%	[-1.2%, -0.4%]	6
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-0.9%	[-1.2%, -0.4%]	6

Mixed

Only use normalize_param_env when normalizing predicate in check_item_bounds #117542 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	0.4%	[0.4%, 0.4%]	1
Regressions ❌ (secondary)	1.4%	[0.8%, 2.3%]	12
Improvements ✅ (primary)	-18.4%	[-49.6%, -0.2%]	20
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-17.5%	[-49.6%, 0.4%]	21

• Fixing https://github.com/rust-lang/rust/pull/117131#issuecomment-1791852906 and deemed acceptable for the small regressions.

Emit #[inline] on derive(Debug) #117727 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	3.1%	[0.3%, 13.7%]	5
Regressions ❌ (secondary)	1.8%	[0.2%, 3.4%]	6
Improvements ✅ (primary)	-2.3%	[-10.6%, -0.1%]	110
Improvements ✅ (secondary)	-3.2%	[-25.2%, -0.1%]	177
All ❌✅ (primary)	-2.1%	[-10.6%, 13.7%]	115

• Huge wins outweigh the small losses.

Rollup of 6 pull requests #117769 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	0.3%	[0.3%, 0.3%]	1
Regressions ❌ (secondary)	1.0%	[1.0%, 1.0%]	1
Improvements ✅ (primary)	-6.5%	[-12.3%, -0.7%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-4.2%	[-12.3%, 0.3%]	3

• Noise