2022-09-06 Triage Log

A relatively quiet week where regressions unfortunately outweighed improvements. What's more, many of the regressions that were found seemed somewhat mysterious requiring some deeper investigations.

Triage done by @rylev. Revision range: 0631ea5d73..09fb0bc6e

Summary:

(instructions:u)	mean	range	count
Regressions ❌ (primary)	0.7%	[0.2%, 4.5%]	85
Regressions ❌ (secondary)	1.0%	[0.3%, 5.4%]	87
Improvements ✅ (primary)	-0.7%	[-1.0%, -0.5%]	9
Improvements ✅ (secondary)	-1.4%	[-2.7%, -0.5%]	22
All ❌✅ (primary)	0.5%	[-1.0%, 4.5%]	94

2 Regressions, 3 Improvements, 2 Mixed; 3 of them in rollups 40 artifact comparisons made in total

Regressions

Replace rustc_data_structures::thin_vec::ThinVec with thin_vec::ThinVec #100869 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	0.3%	[0.2%, 0.5%]	37
Regressions ❌ (secondary)	0.5%	[0.2%, 0.9%]	12
Improvements ✅ (primary)	-0.5%	[-0.6%, -0.4%]	4
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.2%	[-0.6%, 0.5%]	41

• Pre-merge results looked much more performance neutral than this ended up being.
• The author is looking into whether there are any obvious follow-ups.

Rollup of 14 pull requests #101295 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	1.5%	[0.4%, 4.2%]	6
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	1.5%	[0.4%, 4.2%]	6

• #101279 is responsible for these regressions.
• This was a bug fix and the performance impact was not too surprising.

Improvements

Rollup of 7 pull requests #101238 (Comparison Link)

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

interpret: fix unnecessary allocation in validation visitor #101154 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.5%	[-0.5%, -0.5%]	1
Improvements ✅ (secondary)	-1.6%	[-3.3%, -0.6%]	13
All ❌✅ (primary)	-0.5%	[-0.5%, -0.5%]	1

Revert “Remove deferred sized checks” #100966 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	0.9%	[0.6%, 2.4%]	7
Improvements ✅ (primary)	-1.2%	[-1.3%, -1.1%]	2
Improvements ✅ (secondary)	-1.7%	[-1.7%, -1.6%]	2
All ❌✅ (primary)	-1.2%	[-1.3%, -1.1%]	2

• The regressions listed here are just noise.

Mixed

Rework definition of MIR phases to more closely reflect semantic concerns #99102 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	0.7%	[0.3%, 1.7%]	33
Regressions ❌ (secondary)	1.4%	[0.5%, 2.0%]	11
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-0.5%	[-0.5%, -0.5%]	4
All ❌✅ (primary)	0.7%	[0.3%, 1.7%]	33

• A late addition to the PR to simplify some passes has caused the regression.
• Pinged the author to see if there is any additional follow-up.

Rollup of 9 pull requests #101318 (Comparison Link)

(instructions:u)	mean	range	count
Regressions ❌ (primary)	0.2%	[0.2%, 0.2%]	3
Regressions ❌ (secondary)	1.0%	[0.3%, 3.9%]	5
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-0.8%	[-0.8%, -0.8%]	1
All ❌✅ (primary)	0.2%	[0.2%, 0.2%]	3

• There doesn't seem to be a clear cause so running a few of the PRs through performance runs to see what happens.