rustdoc: optimize impl sorting by nnethercote · Pull Request #157179 · rust-lang/rust

nnethercote · 2026-05-31T04:54:18Z

Currently rustdoc sorts impls in a couple of places using long HTML string representations of the impls. Using plain text representations instead speeds things up. Details in individual commits.

Currently impls printed in alternate mode use plain text for the first part and HTML for the where clause. This commit fixes the where clause so it's also plain text.

These sorts are hot, so it's significantly faster to use concise plain text rather than verbose HTML.

We don't need to because the char ordering works out naturally.

nnethercote · 2026-05-31T04:56:21Z

LLM disclosure: I wrote the three commits by hand after viewing profiling data. I used an LLM to review the commits before submitting. The LLM identified that ImplString::is_negative was no longer used, so I removed that field by hand.

nnethercote · 2026-05-31T04:56:30Z

@bors try @rust-timer queue

rustdoc: optimize impl sorting

rust-bors · 2026-05-31T07:21:38Z

☀️ Try build successful (CI)
Build commit: e444e2a (e444e2a215db55a372a3012da35a7c1b15bc6b4d, parent: e7a21fa2562f67cf9cac407c29287a16862b4463)

rust-timer · 2026-05-31T08:58:23Z

Finished benchmarking commit (e444e2a): comparison URL.

Overall result: ✅ improvements - no action needed

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.

@bors rollup=never
@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
Improvements ✅ (primary)	-1.5%	[-6.7%, -0.4%]	13
Improvements ✅ (secondary)	-1.1%	[-1.9%, -0.3%]	2
All ❌✅ (primary)	-1.5%	[-6.7%, -0.4%]	13

Max RSS (memory usage)

Results (secondary -3.8%)

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

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.8%	[-5.3%, -2.2%]	2
All ❌✅ (primary)	-	-	0

Cycles

Results (primary -2.2%, secondary -4.4%)

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

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-2.2%	[-2.7%, -1.8%]	2
Improvements ✅ (secondary)	-4.4%	[-4.4%, -4.4%]	1
All ❌✅ (primary)	-2.2%	[-2.7%, -1.8%]	2

Binary size

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

Bootstrap: 513.315s -> 511.874s (-0.28%)
Artifact size: 400.75 MiB -> 400.75 MiB (-0.00%)

Kobzol · 2026-05-31T09:34:52Z

Nice wins! Have you tried replacing compare_names with just <, to see what is its effect on perf.?

Btw I noticed at https://doc.rust-lang.org/std/cell/struct.Cell.html#trait-implementations, then in the bar on the left, !Sync is sorted as the first item, but in the actual list of trait impls in the middle, it is shown as last. I wonder if that is an existing bug in the sorting logic, or if it is supposed to be like that.

nnethercote · 2026-05-31T10:24:29Z

Nice wins! Have you tried replacing compare_names with just <, to see what is its effect on perf.?

Yes... this approach gets about 90-95% of the improvement of replacing compare_names with Ord::cmp.

Btw I noticed at https://doc.rust-lang.org/std/cell/struct.Cell.html#trait-implementations, then in the bar on the left, !Sync is sorted as the first item, but in the actual list of trait impls in the middle, it is shown as last. I wonder if that is an existing bug in the sorting logic, or if it is supposed to be like that.

Good question, I'm not sure. CoerceUnsize and DispatchFromDyn are also in different places; those two both have impl<T, U> instead of impl<T>, which might be a factor? And the AsRef impls are also in different orders in the two places. In general the order in the left sidebar seems better.

I just checked with my local builds. The orders are unchanged by this PR, so I think I've successfully preserved the existing behaviour. Which seems reasonable for an optimization PR. Any bugs should probably be fixed in a separate PR.

nnethercote · 2026-05-31T10:24:44Z

r? @GuillaumeGomez

GuillaumeGomez · 2026-05-31T10:28:32Z

Looks great, thanks!

@bors r+ rollup

rust-bors · 2026-05-31T10:28:35Z

📌 Commit f7c8bc2 has been approved by GuillaumeGomez

It is now in the queue for this repository.

GuillaumeGomez · 2026-05-31T10:28:44Z

Woups, perf sensitive.

@bors rollup=never

Kobzol · 2026-05-31T14:39:02Z

Nice wins! Have you tried replacing compare_names with just <, to see what is its effect on perf.?

Yes... this approach gets about 90-95% of the improvement of replacing compare_names with Ord::cmp.

I meant combining this PR with replacing compare_names with just cmp.

rust-bors · 2026-05-31T16:36:36Z

☀️ Test successful - CI
Approved by: GuillaumeGomez
Duration: 3h 15m 1s
Pushing 1ce45a0 to main...

github-actions · 2026-05-31T16:39:36Z

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 9e293ae (parent) -> 1ce45a0 (this PR)

Test differences

Show 4 test diffs

4 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 1ce45a011a3fe99ab08643d8eb7229bfc3b40bda --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-msvc-ext2: 2h 2m -> 1h 28m (-27.8%)
dist-powerpc-linux: 1h 15m -> 1h 31m (+21.6%)
x86_64-msvc-ext3: 1h 44m -> 1h 24m (-19.5%)
dist-s390x-linux: 1h 30m -> 1h 12m (-19.4%)
dist-x86_64-solaris: 1h 21m -> 1h 35m (+17.9%)
x86_64-gnu-tools: 55m 8s -> 45m 41s (-17.2%)
x86_64-gnu-llvm-21-2: 1h 26m -> 1h 39m (+15.2%)
armhf-gnu: 1h 18m -> 1h 27m (+11.3%)
dist-apple-various: 1h 39m -> 1h 49m (+11.0%)
pr-check-1: 25m 14s -> 27m 59s (+10.9%)

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.

rust-timer · 2026-05-31T17:18:52Z

Finished benchmarking commit (1ce45a0): comparison URL.

Overall result: ✅ improvements - no action needed

@rustbot label: -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
Improvements ✅ (primary)	-1.5%	[-6.7%, -0.3%]	13
Improvements ✅ (secondary)	-1.1%	[-1.9%, -0.4%]	2
All ❌✅ (primary)	-1.5%	[-6.7%, -0.3%]	13

Max RSS (memory usage)

Results (primary -3.2%, secondary -5.8%)

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

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-3.2%	[-4.7%, -1.3%]	3
Improvements ✅ (secondary)	-5.8%	[-5.8%, -5.8%]	1
All ❌✅ (primary)	-3.2%	[-4.7%, -1.3%]	3

Cycles

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

Binary size

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

Bootstrap: 508.947s -> 513.507s (0.90%)
Artifact size: 400.66 MiB -> 400.69 MiB (0.01%)

nnethercote · 2026-06-01T05:59:58Z

Btw I noticed at https://doc.rust-lang.org/std/cell/struct.Cell.html#trait-implementations, then in the bar on the left, !Sync is sorted as the first item, but in the actual list of trait impls in the middle, it is shown as last.

I have a fix for this in #157233.

nnethercote added 3 commits May 31, 2026 11:29

Add a missing alternate case to print_where_clause.

043c6db

Currently impls printed in alternate mode use plain text for the first part and HTML for the where clause. This commit fixes the where clause so it's also plain text.

Sort impls using plain text, not HTML.

60b4da5

These sorts are hot, so it's significantly faster to use concise plain text rather than verbose HTML.

Don't explicitly sort negative impls first.

f7c8bc2

We don't need to because the char ordering works out naturally.

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 May 31, 2026

This comment has been minimized.

Sign in to view

rust-bors Bot pushed a commit that referenced this pull request May 31, 2026

Auto merge of #157179 - nnethercote:compare_names, r=<try>

e444e2a

rustdoc: optimize impl sorting

This comment has been minimized.

Sign in to view

rustbot removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label May 31, 2026

nnethercote marked this pull request as ready for review May 31, 2026 10:24

rustbot assigned GuillaumeGomez May 31, 2026

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 May 31, 2026

GuillaumeGomez approved these changes May 31, 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 May 31, 2026

This comment has been minimized.

Sign in to view

rust-bors Bot added merged-by-bors This PR was explicitly merged by bors. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels May 31, 2026

rust-bors Bot merged commit 1ce45a0 into rust-lang:main May 31, 2026
14 checks passed

rustbot added this to the 1.98.0 milestone May 31, 2026

nnethercote deleted the compare_names branch June 3, 2026 00:27

Uh oh!

Conversation

nnethercote commented May 31, 2026

Uh oh!

nnethercote commented May 31, 2026

Uh oh!

nnethercote commented May 31, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

rust-bors Bot commented May 31, 2026

Uh oh!

This comment has been minimized.

rust-timer commented May 31, 2026

Overall result: ✅ improvements - no action needed

Uh oh!

Kobzol commented May 31, 2026

Uh oh!

nnethercote commented May 31, 2026

Uh oh!

nnethercote commented May 31, 2026

Uh oh!

GuillaumeGomez commented May 31, 2026

Uh oh!

rust-bors Bot commented May 31, 2026

Uh oh!

GuillaumeGomez commented May 31, 2026

Uh oh!

This comment has been minimized.

Kobzol commented May 31, 2026

Uh oh!

rust-bors Bot commented May 31, 2026

Uh oh!

Uh oh!

github-actions Bot commented May 31, 2026

Test differences

Job duration changes

Uh oh!

rust-timer commented May 31, 2026

Overall result: ✅ improvements - no action needed

Uh oh!

nnethercote commented Jun 1, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants