Implement rustc_datastructures::RwLock as union over RefCell and parking_lot::RwLock.

[LorrensP-2158466]

Revive of #136772.

cc "Parallel Rustc Front-end" #113349

Instead of explicit enums I decided to see what would happen if we would use the ModeUnion with Mode. Nothing changes API wise, but perf may or may not improve.

Second commit shows what actually changes, first commit just moves RwLock to a separate module.

[rustbot]

r? @nnethercote

rustbot has assigned @nnethercote.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: compiler
• compiler expanded to 74 candidates
• Random selection from 18 candidates

[petrochenkov]

#158845 replaces more cells with rwlocks, so the results for this PR should look more contrast if we benchmark it after landing that PR.

[petrochenkov]

Can't look at the diff rn, need to go, but @bors try @rust-timer queue

[LorrensP-2158466]

#113349 replaces more cells with rwlocks, so the results for this PR should look more contrast if we benchmark it after landing that PR.

Understood, but you linked the tracking issue. Do you happen to have the link to the pr in question?

[rust-bors]

☀️ Try build successful (CI)
Build commit: 57bfc2b (57bfc2bb07a49fc7e0c070fc155c156857526ca6)
Base parent: 4a9d536 (4a9d5368df6450bbd2fc8dde4508c3e5d83bb19d)

[rust-timer]

Finished benchmarking commit (57bfc2b): comparison URL.

Overall result: ❌✅ regressions and improvements - please read:

Benchmarking means the PR may be perf-sensitive. It's automatically marked not fit for rolling up. Overriding is possible but disadvised: it risks changing compiler perf.

Next, please: If you can, justify the regressions found in this try perf run in writing along with @rustbot label: +perf-regression-triaged. If not, fix the regressions and do another perf run. Neutral or positive results will clear the label automatically.

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

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	0.1%	[0.1%, 0.2%]	2
Improvements ✅ (primary)	-0.2%	[-0.2%, -0.2%]	5
Improvements ✅ (secondary)	-0.4%	[-0.7%, -0.2%]	8
All ❌✅ (primary)	-0.2%	[-0.2%, -0.2%]	5

Max RSS (memory usage)

Results (primary 2.0%, secondary 9.6%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	2.0%	[2.0%, 2.0%]	1
Regressions ❌ (secondary)	9.6%	[9.6%, 9.6%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	2.0%	[2.0%, 2.0%]	1

Cycles

Results (primary 2.2%, secondary -3.6%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	2.2%	[2.2%, 2.2%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.6%	[-3.6%, -3.6%]	1
All ❌✅ (primary)	2.2%	[2.2%, 2.2%]	1

Binary size

Results (primary 0.1%, secondary 0.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	0.1%	[0.0%, 0.1%]	55
Regressions ❌ (secondary)	0.1%	[0.0%, 0.1%]	23
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.1%	[0.0%, 0.1%]	55

Bootstrap: 491.028s -> 491.544s (0.11%)
Artifact size: 389.42 MiB -> 389.67 MiB (0.06%)

[LorrensP-2158466]

Definitely better results than #136772 but not a lot of significant improvements. Though the ~10% RSS increase for deep_vector looks concerning, no idea if its signifcant.

I did notice that I have no inline(always) on try_read.

And a lot less hint::cold_path() when branching on self.mode. I mostly followed the impl of Lock when deciding where to place these inlines or cold_path statements, maybe they could improve a bit more.

[LorrensP-2158466]

we do not have any Clone impls as they are not needed atm. We could in theory use borrow.replace(UNUSED) here. A very micro optimization.

View changes since the review

[LorrensP-2158466]

Followed the parking_lot::MappedRwLockReadGuard design:

pub struct MappedRwLockReadGuard<'a, R: RawRwLock, T: ?Sized> {
    raw: &'a R, // the mode union in our case
    data: *const T,
    marker: PhantomData<&'a T>,
}

View changes since the review