make query caches use Sharded only for multiple threads

[SparrowLii]

This PR make query caches use Sharded only for multiple threads. This can improve spatial locality in single-threaded scenarios to improve performance

cc @Zoxc
r? @cjgillot

[SparrowLii]

@bors try @rust-timer queue

[bors]

⌛ Trying commit 16b9941 with merge c5169d918670de7b50199e90a38e96c2aa1d4bc2...

[bors]

☀️ Try build successful - checks-actions
Build commit: c5169d918670de7b50199e90a38e96c2aa1d4bc2 (c5169d918670de7b50199e90a38e96c2aa1d4bc2)

[Zoxc]

This change is probably a bit costly as is doubles the size of the query accessor.

[SparrowLii]

Yea, I added likely hint here, maybe helps a lillte

[Zoxc]

That doesn't really affect the code size.

[Zoxc]

I did a benchmark run for this PR with cfg(parallel_compiler) and 1 thread with #111713 and a local branch with optimized lock_shard methods applied:

Benchmark	Before	After
	Time	Time	%
🟣 clap:check	1.6604s	1.6786s	💔 1.09%
🟣 hyper:check	0.2419s	0.2442s	0.96%
🟣 regex:check	0.9228s	0.9339s	💔 1.20%
🟣 syn:check	1.5121s	1.5293s	💔 1.14%
🟣 syntex_syntax:check	5.9729s	6.0290s	0.94%
Total	10.3102s	10.4150s	💔 1.02%
Summary	1.0000s	1.0107s	💔 1.07%

[SparrowLii]

Yea I don't know why rustc-perf makes this PR starves. I gonna try use RefCell directly.

[SparrowLii]

@Zoxc Could you help test the perf again? If the performace not as good as your local test, it is very good. It means we can fetch better performace than previous work

[rust-timer]

Finished benchmarking commit (c5169d918670de7b50199e90a38e96c2aa1d4bc2): comparison URL.

Overall result: ❌ regressions - no 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.

@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
Regressions ❌ (secondary)	0.2%	[0.2%, 0.2%]	3
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

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)	2.1%	[1.6%, 3.2%]	3
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

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)	3.7%	[3.6%, 3.8%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 631.241s -> 631.856s (0.10%)
Artifact size: 346.61 MiB -> 346.57 MiB (-0.01%)

[bors]

☔ The latest upstream changes (presumably #115198) made this pull request unmergeable. Please resolve the merge conflicts.

[apiraino]

Switching to waiting on author to resolve conflicts. I'm reading this comment: do the latest perf. run results align with your expectations? Feel free to request a review with @rustbot ready if you need another look at this, thanks!

@rustbot author

[SparrowLii]

Close this since we already had better solutions