Resolver: Introduce `CmRef` which has a speclative borrow variant for `CmRefCell` by LorrensP-2158466 · Pull Request #160271 · rust-lang/rust

LorrensP-2158466 · 2026-07-31T09:47:18Z

We now introduce CmRef that is returned by CmRefCell::borrow, which does a normal borrow of the underlying T if we are in speculative resolution. We do not alter state during speculative resolution and because it will be run in parallel in the future, updating the RefCell borrow counters is not possible. Thus we just return &T that is "untracked".

Also allows us to remove the CmRefCell wrapper around the external module resolution table.

r? @petrochenkov

petrochenkov · 2026-07-31T10:49:47Z

@bors try @rust-timer queue

Resolver: Introduce `CmRef` which has a speclative borrow variant for `CmRefCell`

rust-bors · 2026-07-31T12:56:34Z

☀️ Try build successful (CI)
Build commit: caf2dd1 (caf2dd11317e37ec790b3746057fae8a36779c87)
Base parent: 922325b (922325bb13bfea5b41454318563f2a65e83c2336)

rust-timer · 2026-07-31T14:08:25Z

Finished benchmarking commit (caf2dd1): comparison URL.

Overall result: ❌ regressions - 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)	0.2%	[0.2%, 0.3%]	34
Regressions ❌ (secondary)	0.4%	[0.1%, 0.8%]	22
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-0.2%	[-0.3%, -0.1%]	3
All ❌✅ (primary)	0.2%	[0.2%, 0.3%]	34

Max RSS (memory usage)

Results (primary 1.5%, secondary 1.7%)

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

	mean	range	count
Regressions ❌ (primary)	3.4%	[3.4%, 3.4%]	1
Regressions ❌ (secondary)	2.8%	[0.8%, 6.2%]	4
Improvements ✅ (primary)	-0.4%	[-0.4%, -0.4%]	1
Improvements ✅ (secondary)	-0.5%	[-0.6%, -0.4%]	2
All ❌✅ (primary)	1.5%	[-0.4%, 3.4%]	2

Cycles

Results (primary 0.3%, secondary -1.2%)

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

	mean	range	count
Regressions ❌ (primary)	1.2%	[0.4%, 2.7%]	7
Regressions ❌ (secondary)	1.7%	[0.5%, 3.5%]	15
Improvements ✅ (primary)	-1.0%	[-2.2%, -0.5%]	5
Improvements ✅ (secondary)	-5.1%	[-10.8%, -0.6%]	11
All ❌✅ (primary)	0.3%	[-2.2%, 2.7%]	12

Binary size

This perf run didn't have relevant results for this metric.

Bootstrap: 490.333s -> 490.941s (0.12%)
Artifact size: 392.58 MiB -> 390.55 MiB (-0.52%)

Kobzol · 2026-07-31T21:47:33Z

@bors try @rust-timer queue

Resolver: Introduce `CmRef` which has a speclative borrow variant for `CmRefCell`

rust-bors · 2026-07-31T23:52:40Z

☀️ Try build successful (CI)
Build commit: 14a8f12 (14a8f12f6d59bdebd04ce9fc8ef7b4f0a25f0a4f)
Base parent: 64af19d (64af19d44c63587455c76f484ce76261a9971e96)

rust-timer · 2026-08-01T00:34:28Z

Finished benchmarking commit (14a8f12): comparison URL.

Overall result: ❌ regressions - 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)	0.3%	[0.2%, 0.4%]	12
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Max RSS (memory usage)

Results (primary 4.1%, secondary 2.7%)

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

	mean	range	count
Regressions ❌ (primary)	5.2%	[2.6%, 9.9%]	4
Regressions ❌ (secondary)	4.3%	[2.2%, 6.4%]	2
Improvements ✅ (primary)	-0.6%	[-0.6%, -0.6%]	1
Improvements ✅ (secondary)	-0.6%	[-0.6%, -0.6%]	1
All ❌✅ (primary)	4.1%	[-0.6%, 9.9%]	5

Cycles

Results (primary 0.4%, secondary 0.2%)

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

	mean	range	count
Regressions ❌ (primary)	0.8%	[0.5%, 1.1%]	10
Regressions ❌ (secondary)	0.9%	[0.5%, 2.6%]	13
Improvements ✅ (primary)	-0.7%	[-0.9%, -0.5%]	3
Improvements ✅ (secondary)	-2.4%	[-4.3%, -0.6%]	3
All ❌✅ (primary)	0.4%	[-0.9%, 1.1%]	13

Binary size

This perf run didn't have relevant results for this metric.

Bootstrap: 492.106s -> 489.164s (-0.60%)
Artifact size: 390.46 MiB -> 390.45 MiB (-0.00%)

petrochenkov · 2026-08-01T12:14:09Z

I'm surprised it's a regression at all, it's such a tiny thing, checking a flag once per module, not even per declaration.

LorrensP-2158466 · 2026-08-01T12:49:40Z

It is also per Resolution, those are wrapped in a CmRefCell as well.

Deref is also an enum match, niche optimization puts the niche on the Ref.ptr.

petrochenkov · 2026-08-01T13:04:18Z

It is also per Resolution, those are wrapped in a CmRefCell as well.
Deref is also an enum match, niche optimization puts the niche on the Ref.ptr.

I mean in this PR, not in follow ups.
"checking a flag once per module" is the deref on resolutions' returned value.

I've been trying to use a closure-based interface for this for the last half an hour, like

    fn with_resolutions<R>(&self, module: Module<'ra>, f: impl FnOnce(&ResolutionTable<'ra>) -> R) -> R { ... }

but it's getting too unwieldy, maybe not today.

r=me on this PR after squashing commits.
@rustbot author

CmRefCell::borrow with the token optimization will be the next step.

rustbot · 2026-08-01T13:04:20Z

Reminder, once the PR becomes ready for a review, use @rustbot ready.

… the underlying `RefCell`

LorrensP-2158466 · 2026-08-01T13:23:59Z

"checking a flag once per module" is the deref on resolutions' returned value.

right yeah, forgot the revert.

r=me on this PR after squashing commits.

squashed, but can't r=you, @rustbot ready.

LorrensP-2158466 · 2026-08-01T13:24:46Z

CmRefCell::borrow with the token optimization will be the next step.

i'll open a pr based of this pr with the token, rebasing will remove this pr's commit.

Edit: Do you want the new version of borrow as well here? Which I had to revert here.

petrochenkov · 2026-08-01T14:20:01Z

@bors r+ rollup

rust-bors · 2026-08-01T14:20:03Z

📌 Commit 9890f66 has been approved by petrochenkov

It is now in the queue for this repository.

petrochenkov · 2026-08-01T14:21:33Z

Edit: Do you want the new version of borrow as well here? Which I had to revert here.

Yes, the full CmRefCell change.

…uwer Rollup of 8 pull requests Successful merges: - #160262 (Library lock file maintenance) - #158548 (Move `std::io::copy` to `alloc::io`) - #158814 (Produce an error when `#[inline]` and `#[rust_force_inline]` are used together) - #160025 (Fix an edge case with `StepBy::nth` on non-fused iterators) - #160271 (Resolver: Introduce `CmRef` which has a speclative borrow variant for `CmRefCell`) - #160281 (Fix(lib/fs/tests): Avoid permission denials when cleaning up TempDirs in `set_get_permissions_nofollows*`) - #160325 (tidy: Check `proc_macro_deps.rs` by reading it, not by including it) - #160334 (Add regression test for unused_allocation on boxed comparison)

@petrochenkov

Rollup merge of #160271 - LorrensP-2158466:res-cm-refcell, r=petrochenkov Resolver: Introduce `CmRef` which has a speclative borrow variant for `CmRefCell` part of #158845 We now introduce `CmRef` that is returned by `CmRefCell::borrow`, which does a normal borrow of the underlying `T` if we are in speculative resolution. We do not alter state during speculative resolution and because it will be run in parallel in the future, updating the `RefCell` borrow counters is not possible. Thus we just return `&T` that is "untracked". Also allows us to remove the `CmRefCell` wrapper around the external module resolution table. r? @petrochenkov

rustbot assigned petrochenkov Jul 31, 2026

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Jul 31, 2026

LorrensP-2158466 commented Jul 31, 2026

View reviewed changes

Comment thread compiler/rustc_resolve/src/lib.rs Outdated