Create a query cache for DefId

[Zoxc]

This adds a specialized cache for DefId which stores a IndexVec of hashmaps reducing the size of the stored hashmap key.

Benchmark	Before	After
	Time	Time	%
🟣 clap:check	1.6924s	1.6879s	-0.26%
🟣 hyper:check	0.2522s	0.2527s	0.21%
🟣 regex:check	0.9573s	0.9542s	-0.32%
🟣 syn:check	1.5345s	1.5265s	-0.52%
🟣 syntex_syntax:check	5.8742s	5.8500s	-0.41%
Total	10.3105s	10.2714s	-0.38%
Summary	1.0000s	0.9974s	-0.26%

[rustbot]

r? @wesleywiser

(rustbot has picked a reviewer for you, use r? to override)

[WaffleLapkin]

@bors try @rust-timer queue

[bors]

⌛ Trying commit 18a3ca059d4a99fa0430a44d3d0b48547e038032 with merge 8775a3d31421b553c1b7985b0ef0bb31e2ab121e...

[bors]

☀️ Try build successful - checks-actions
Build commit: 8775a3d31421b553c1b7985b0ef0bb31e2ab121e (8775a3d31421b553c1b7985b0ef0bb31e2ab121e)

[rust-timer]

Finished benchmarking commit (8775a3d31421b553c1b7985b0ef0bb31e2ab121e): comparison URL.

Overall result: ❌ regressions - ACTION NEEDED

Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. While you can manually mark this PR as fit for rollup, we strongly recommend not doing so since this PR may lead to changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please indicate this with @rustbot label: +perf-regression-triaged along with sufficient written justification. If you cannot justify the regressions please fix the regressions and do another perf run. If the next run shows neutral or positive results, the label will be automatically removed.

@bors rollup=never
@rustbot label: -S-waiting-on-perf +perf-regression

Instruction count

This is a highly reliable metric that was used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	0.4%	[0.2%, 0.9%]	97
Regressions ❌ (secondary)	0.5%	[0.2%, 1.0%]	48
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.4%	[0.2%, 0.9%]	97

Max RSS (memory usage)

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	2.7%	[1.9%, 4.9%]	6
Improvements ✅ (primary)	-0.8%	[-1.4%, -0.4%]	8
Improvements ✅ (secondary)	-3.1%	[-3.1%, -3.1%]	1
All ❌✅ (primary)	-0.8%	[-1.4%, -0.4%]	8

Cycles

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-2.1%	[-2.8%, -1.2%]	8
Improvements ✅ (secondary)	-2.5%	[-2.5%, -2.5%]	1
All ❌✅ (primary)	-2.1%	[-2.8%, -1.2%]	8

[Zoxc]

The bootstrap regressions are a bit high, so it probably make sense to re-evaluate this if #108638 lands.

I also ran unchanged incremental benchmarks locally to see if any of the instruction regressions turned into wall time regressions. They seem to be reasonably neutral.

Benchmark	Before	After
	Time	Time	%
🟣 clap:check:unchanged	0.4674s	0.4676s	0.04%
🟣 hyper:check:unchanged	0.1323s	0.1324s	0.10%
🟣 regex:check:unchanged	0.3281s	0.3289s	0.25%
🟣 syn:check:unchanged	0.6412s	0.6369s	-0.66%
🟣 syntex_syntax:check:unchanged	1.8809s	1.8759s	-0.27%
Total	3.4499s	3.4418s	-0.23%
Summary	1.0000s	0.9989s	-0.11%

[oli-obk]

@rustbot blocked #108638

edit: hmm... i was sure rustbot knew that command

[cjgillot]

Do you have a measurement of the data pattern in this cache? Which proportion of local/non-local def-ids? Could you explain why you chose to shard along CrateNum, instead of a IndexVec for local def-ids and a FxHashMap for non-local ones?

[Zoxc]

This no longer seems to be an improvement:

Benchmark	Before	After
	Time	Time	%
🟣 clap:check	1.6371s	1.6392s	0.13%
🟣 hyper:check	0.2519s	0.2520s	0.02%
🟣 regex:check	0.9170s	0.9186s	0.18%
🟣 syn:check	1.4787s	1.4804s	0.11%
🟣 syntex_syntax:check	5.6790s	5.6898s	0.19%
Total	9.9637s	9.9800s	0.16%
Summary	1.0000s	1.0013s	0.13%

I don't have data on non-local vs. local def-ids, but I did previously try specializing only the storage for local def-ids, which wasn't an improvement.