Avoid allocations when canonicalizing - #161077

nnethercote · 2026-08-13T23:36:22Z

Canonicalization is hot in the new solver. This commit avoids some allocations while doing it. Details in individual commits.

r? @lcnr

There are many short lived `Canonicalizer` instances. These each involve some vecs and hashmaps, and the cost of the allocations for these adds up. This commit introduces `CanonicalizerState` which holds the vecs and hashmaps, and puts a single instance in `InferCtxt` to be cleared and reused for each canonicalizer.

It's a small performance win. This requires adding `DelayedMap::clear` (and `DelayedSet::clear` is also added for symmetry).

nnethercote · 2026-08-13T23:36:39Z

@bors try @rust-timer queue

Avoid allocations when canonicalizing

rust-bors · 2026-08-14T00:35:17Z

☀️ Try build successful (CI)
Build commit: 42e1aad (42e1aad072317b0ddfbfc70731cbb92bf50ea796)
Base parent: ba28ff7 (ba28ff76f353a722f31c4f3dd2ac4e437d36411b)

rust-timer · 2026-08-14T01:18:19Z

Finished benchmarking commit (42e1aad): 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)	0.1%	[0.1%, 0.1%]	3
Regressions ❌ (secondary)	0.1%	[0.1%, 0.2%]	6
Improvements ✅ (primary)	-0.2%	[-0.2%, -0.2%]	2
Improvements ✅ (secondary)	-1.4%	[-3.4%, -0.4%]	22
All ❌✅ (primary)	-0.0%	[-0.2%, 0.1%]	5

Max RSS (memory usage)

Results (primary 6.2%)

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

	mean	range	count
Regressions ❌ (primary)	6.2%	[6.2%, 6.2%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	6.2%	[6.2%, 6.2%]	1

Cycles

Results (primary -0.7%, 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)	2.1%	[2.1%, 2.1%]	1
Regressions ❌ (secondary)	4.7%	[2.8%, 6.7%]	9
Improvements ✅ (primary)	-2.1%	[-2.1%, -2.0%]	2
Improvements ✅ (secondary)	-3.5%	[-4.3%, -2.9%]	3
All ❌✅ (primary)	-0.7%	[-2.1%, 2.1%]	3

Binary size

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

Bootstrap: 457.997s -> 456.358s (-0.36%)
Artifact size: 396.43 MiB -> 397.06 MiB (0.16%)

rustbot · 2026-08-14T02:57:04Z

Some changes occurred to the core trait solver

cc @rust-lang/initiative-trait-system-refactor

nnethercote · 2026-08-14T02:57:47Z

LLM disclosure: the idea for this came from an LLM after analyzing a Cachegrind profile. I wrote the code and text myself.

jdonszelmann · 2026-08-14T06:37:28Z

+
+    fn release_canonicalizer_state(&self, mut state: CanonicalizerState<Self::Interner>) {
+        // Clear (don't deallocate) the state for later reuse.
+        state.clear();


could be worth adding a debug assertion here that the cleared state == a default state. To make sure we never forget a field...

View changes since the review

jdonszelmann · 2026-08-14T06:38:48Z

+    pub fn clear(&mut self) {
+        // Deconstruct to ensure no fields are missed.
+        let Self { variables, var_kinds, variable_lookup_table, sub_root_lookup_table, cache } =
+            self;


I guess deconstructing does help as well

View changes since the review

jdonszelmann

Consider my comment, unsure whether its worth it, other than that I think this PR looks good to me

View changes since this review

nnethercote · 2026-08-14T07:00:25Z

I think the deconstruction is good enough; it's a well-established idiom. Thanks!

@bors r=jdonszelmann

rust-bors · 2026-08-14T07:00:28Z

📌 Commit 445d379 has been approved by jdonszelmann

It is now in the queue for this repository.

rust-bors · 2026-08-14T20:31:20Z

☀️ Test successful - CI
Approved by: jdonszelmann
Duration: 3h 22m 9s
Pushing d453bdd to main...

github-actions · 2026-08-14T20:34:22Z

What is this?

This is an experimental post-merge analysis report that shows differences in test outcomes between the merged PR and its parent PR.

Comparing 2fb4ed8 (parent) -> d453bdd (this PR)

Test differences

Show 2 test diffs

2 doctest diffs were found. These are ignored, as they are noisy.

Test dashboard

Run

cargo run --manifest-path src/ci/citool/Cargo.toml -- \
    test-dashboard d453bdd8f092d099bc336f0bda4163f809ad18e0 --output-dir test-dashboard

And then open test-dashboard/index.html in your browser to see an overview of all executed tests.

Job duration changes

x86_64-gnu-debug: 1h 21m -> 1h 56m (+44.1%)
dist-x86_64-netbsd: 59m 37s -> 1h 24m (+41.9%)
x86_64-gnu-aux: 1h 53m -> 2h 37m (+39.3%)
dist-x86_64-illumos: 1h 18m -> 1h 48m (+38.7%)
aarch64-apple-1: 1h 33m -> 2h 5m (+35.0%)
aarch64-apple-macos-26-2: 2h 31m -> 1h 40m (-34.1%)
i686-gnu-2: 1h 37m -> 1h 5m (-32.9%)
x86_64-gnu-next-trait-solver-polonius: 43m 49s -> 56m 25s (+28.7%)
dist-x86_64-musl: 1h 44m -> 2h 14m (+28.6%)
x86_64-gnu: 2h 37m -> 1h 53m (-28.2%)

How to interpret the job duration changes?

Job durations can vary a lot, based on the actual runner instance
that executed the job, system noise, invalidated caches, etc. The table above is provided
mostly for t-infra members, for simpler debugging of potential CI slow-downs.

nnethercote added 2 commits August 14, 2026 09:00

Use DelayedMap for CanonicalizerState::cache

445d379

It's a small performance win. This requires adding `DelayedMap::clear` (and `DelayedSet::clear` is also added for symmetry).

This comment has been minimized.

Sign in to view

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Aug 13, 2026

This comment has been minimized.

Sign in to view

rust-bors Bot pushed a commit that referenced this pull request Aug 13, 2026

Auto merge of #161077 - nnethercote:CanonicalizerState, r=<try>

42e1aad

Avoid allocations when canonicalizing

This comment has been minimized.

Sign in to view

rustbot added perf-regression Performance regression. and removed S-waiting-on-perf Status: Waiting on a perf run to be completed. labels Aug 14, 2026

rustbot assigned lcnr Aug 14, 2026

nnethercote marked this pull request as ready for review August 14, 2026 02:57

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Aug 14, 2026

jdonszelmann reviewed Aug 14, 2026

View reviewed changes

rust-bors Bot added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Aug 14, 2026

This comment has been minimized.

Sign in to view

rust-bors Bot added the merged-by-bors This PR was explicitly merged by bors. label Aug 14, 2026

rust-bors Bot removed the S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. label Aug 14, 2026

rust-bors Bot merged commit d453bdd into rust-lang:main Aug 14, 2026
15 checks passed

rustbot added this to the 1.99.0 milestone Aug 14, 2026

rust-bors Bot mentioned this pull request Aug 14, 2026

Make const eval return normalized ty #160526

Open

Uh oh!

Uh oh!

Conversation

nnethercote commented Aug 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

nnethercote commented Aug 13, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

rust-bors Bot commented Aug 14, 2026

Uh oh!

This comment has been minimized.

rust-timer commented Aug 14, 2026

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

Instruction count

Max RSS (memory usage)

Cycles

Binary size

Uh oh!

rustbot commented Aug 14, 2026

Uh oh!

nnethercote commented Aug 14, 2026

Uh oh!

jdonszelmann Aug 14, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jdonszelmann Aug 14, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jdonszelmann left a comment • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

nnethercote commented Aug 14, 2026

Uh oh!

rust-bors Bot commented Aug 14, 2026

Uh oh!

This comment has been minimized.

rust-bors Bot commented Aug 14, 2026

Uh oh!

Uh oh!

github-actions Bot commented Aug 14, 2026

Test differences

Job duration changes

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

nnethercote commented Aug 13, 2026 •

edited

Loading

jdonszelmann Aug 14, 2026 •

edited by rustbot

Loading

jdonszelmann Aug 14, 2026 •

edited by rustbot

Loading

jdonszelmann left a comment •

edited by rustbot

Loading