2024-05-27.md

2024-05-27 Triage Log

A relatively quiet week, with few large changes, the largest driven by further increasing the scope of unsafe precondition checking.

Triage done by @simulacrum. Revision range: 1d0e4afd..a59072ec

Summary:

(instructions:u)	mean	range	count
Regressions ❌ (primary)	1.6%	[0.2%, 6.3%]	84
Regressions ❌ (secondary)	0.9%	[0.1%, 3.8%]	66
Improvements ✅ (primary)	-0.4%	[-2.3%, -0.2%]	37
Improvements ✅ (secondary)	-1.7%	[-3.8%, -0.2%]	22
All ❌✅ (primary)	1.0%	[-2.3%, 6.3%]	121

2 Regressions, 3 Improvements, 5 Mixed; 3 of them in rollups 51 artifact comparisons made in total

Regressions

Rewrite native thread-local storage #116123 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	0.4%	[0.2%, 1.2%]	10
Regressions ❌ (secondary)	1.0%	[0.7%, 1.6%]	9
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.4%	[0.2%, 1.2%]	10

Regressions deemed worth the overall change's contents (see rust-lang/rust#116123 (comment)). The TLS state is now stored in a single thread-local object rather than two, which should decrease costs of addressing it in general (modulo LLVM difficulties).

Rollup of 6 pull requests #125463 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	1.4%	[1.1%, 2.0%]	8
Regressions ❌ (secondary)	1.0%	[0.4%, 1.8%]	23
Improvements ✅ (primary)	-0.5%	[-0.5%, -0.5%]	1
Improvements ✅ (secondary)	-0.3%	[-0.3%, -0.3%]	1
All ❌✅ (primary)	1.2%	[-0.5%, 2.0%]	9

Likely related to fixing an issue around LLD discovery, see rust-lang/rust#125463 (comment). Bugfix warrants spending a bit more time.

Improvements

Move the checks for Arguments constructors to inline const #125518 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.5%	[-0.7%, -0.3%]	5
Improvements ✅ (secondary)	-0.5%	[-0.9%, -0.4%]	7
All ❌✅ (primary)	-0.5%	[-0.7%, -0.3%]	5

Remove DefId from EarlyParamRegion #125468 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.4%	[-0.9%, -0.2%]	48
Improvements ✅ (secondary)	-0.8%	[-2.0%, -0.3%]	19
All ❌✅ (primary)	-0.4%	[-0.9%, -0.2%]	48

[perf] Delay the construction of early lint diag structs #125410 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.5%	[-0.7%, -0.3%]	12
Improvements ✅ (secondary)	-2.0%	[-2.8%, -1.2%]	13
All ❌✅ (primary)	-0.5%	[-0.7%, -0.3%]	12

Mixed

Follow-up fixes to report_return_mismatched_types #123812 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	1.1%	[0.6%, 1.8%]	3
Regressions ❌ (secondary)	0.2%	[0.1%, 0.2%]	6
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-1.1%	[-1.1%, -1.1%]	2
All ❌✅ (primary)	1.1%	[0.6%, 1.8%]	3

Seems to be limited primarily to one scenario in regex, with lots of new metadata decoding. Asked for follow-up by PR author.

Make early lints translatable #124417 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	1.5%	[0.2%, 6.3%]	66
Regressions ❌ (secondary)	0.4%	[0.3%, 0.5%]	7
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-0.7%	[-1.1%, -0.6%]	12
All ❌✅ (primary)	1.5%	[0.2%, 6.3%]	66

Regressions are partially mitigated by #125410 (see earlier in the report). Further follow-up is expected to investigate closing remaining gap (rust-lang/rust#124417 (comment)).

Panic directly in Arguments::new* instead of recursing #117804 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	0.3%	[0.3%, 0.3%]	2
Regressions ❌ (secondary)	0.4%	[0.1%, 0.9%]	9
Improvements ✅ (primary)	-0.5%	[-0.5%, -0.5%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.0%	[-0.5%, 0.3%]	3

Regressions are likely to be inliner noise, not meaningful changes.

Rollup of 7 pull requests #125456 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	0.3%	[0.2%, 0.4%]	8
Improvements ✅ (primary)	-0.3%	[-0.4%, -0.3%]	2
Improvements ✅ (secondary)	-0.5%	[-0.9%, -0.4%]	7
All ❌✅ (primary)	-0.3%	[-0.4%, -0.3%]	2

Seems plausible that the regression is due to #124389 (since it affected derive macros). But also seems not worth further investigation given it's a secondary benchmark and minimal impact. Marked as triaged.

Add assert_unsafe_precondition to unchecked_{add,sub,neg,mul,shl,shr} methods #121571 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	1.5%	[0.2%, 3.4%]	27
Regressions ❌ (secondary)	1.8%	[0.4%, 3.8%]	5
Improvements ✅ (primary)	-0.9%	[-2.5%, -0.3%]	5
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	1.1%	[-2.5%, 3.4%]	32

Regressions are likely expected as we're lowering more code that's late-removed.