refactor Lock for parallel compiler

[SparrowLii]

part of #101566
This PR refactor Lock for parallel compiler, facilitating code review and perf test.

ps. refactored Lock is not Send or Sync. It depends on #107586 to get thread safety.

[rustbot]

r? @WaffleLapkin

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

[SparrowLii]

@bors try @rust-timer queue

[bors]

⌛ Trying commit 51cfd3e with merge 529b96fb784d8a2ac0a170a5e566f13095770326...

[bors]

☀️ Try build successful - checks-actions
Build commit: 529b96fb784d8a2ac0a170a5e566f13095770326 (529b96fb784d8a2ac0a170a5e566f13095770326)

[joboet]

Couldn't these two fields be merged into a single AtomicBool? The single-threaded version would become borrow.as_ptr().replace(true), while the multi-threaded version would use parking_lot_core::park to handle waiting.

[SparrowLii]

Thanks! It makes sense, I'm trying it soon

[joboet]

Shouldn't loading PARALLEL with relaxed ordering (or maybe a non-atomic read) be just as efficient? That avoids having to copy around the bool in every lock.

[SparrowLii]

Thanks!

[rust-timer]

Finished benchmarking commit (529b96fb784d8a2ac0a170a5e566f13095770326): 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)	2.0%	[0.4%, 4.3%]	190
Regressions ❌ (secondary)	1.8%	[0.6%, 6.7%]	190
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	2.0%	[0.4%, 4.3%]	190

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)	0.8%	[0.8%, 0.8%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-0.7%	[-1.0%, -0.5%]	7
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)	1.8%	[1.0%, 3.2%]	45
Regressions ❌ (secondary)	1.9%	[0.7%, 4.7%]	29
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	1.8%	[1.0%, 3.2%]	45

[SparrowLii]

In my local tests, the drop of LockGuard is the main cause of performance loss. I am trying to replace lock() with with_lock() without LockGuard to see the impact on performance

[Dylan-DPC]

r? @cjgillot

assigning them as they are the assigned reviewer on the other pr as well

[Dylan-DPC]

@SparrowLii any updates on this?

[SparrowLii]

@Zoxc has a more completed implementation in #111713. I will first do the work of query cache next. If they still have no time to update #111713 then, I will finish this PR instead.

[Dylan-DPC]

Closing this as #111713 is now merged