Remove query function arrays by nnethercote · Pull Request #153114 · rust-lang/rust

nnethercote · 2026-02-26T00:39:50Z

define_queries! produces four arrays of function pointers, which other functions iterate over. These aren't actually necessary.

r? @petrochenkov

nnethercote · 2026-02-26T02:46:29Z

@bors try @rust-timer queue

Remove query function arrays

rust-bors · 2026-02-26T04:56:02Z

☀️ Try build successful (CI)
Build commit: f8df332 (f8df332a248f146a91b1023c5961dff7147fc3f3, parent: 1ed488274bec5bf5cfe6bf7a1cc089abcc4ebd68)

rust-timer · 2026-02-26T05:36:28Z

Finished benchmarking commit (f8df332): comparison URL.

Overall result: no relevant changes - no action needed

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

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

Instruction count

This benchmark run did not return any relevant results for this metric.

Max RSS (memory usage)

Results (primary -4.1%)

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)	-4.1%	[-4.1%, -4.1%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-4.1%	[-4.1%, -4.1%]	1

Cycles

Results (primary 3.7%)

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

	mean	range	count
Regressions ❌ (primary)	3.7%	[3.0%, 4.4%]	2
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	3.7%	[3.0%, 4.4%]	2

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 492.161s -> 478.631s (-2.75%)
Artifact size: 395.78 MiB -> 397.45 MiB (0.42%)

nnethercote · 2026-02-26T06:11:45Z

Here's an explanation. Currently, the pattern is...

We have a hand-written function, v_inner, that does something query-related.
We generate one function per query (each one in its own module) that simply calls v_inner with some query-specific data (e.g. vtable).
We generate an array, V, with elements that point to all of these generated functions.
We have a hand-written top-level function, v_all, that iterates over the array V and calls its elements, one by one.

A simplified code representation:

// Hand-written
fn v_inner(s: &str) { ... }
                                                       
// Generated by the macro                              
mod q1 { fn v() { v_inner("q1"); } }
mod q2 { fn v() { v_inner("q2"); } }
mod q3 { fn v() { v_inner("q3"); } }
                                                       
// Generated by the macro
const V: &[q1::v, q2::v, q3::v];
                                                       
// Hand-written
fn v_all() {
    for v in V.iter() {
        v();
    }
}

After this PR, the pattern is...

We have a hand-written function, v, that does something query-related.
We generate a top-level function, v_all, that calls v for each query.

In code form:

// Hand-written
fn v(s: &str) { ... }
                                                       
// Generated by the macro
fn v_all() {
    v("q1");
    v("q2");
    v("q3");
}

Much nicer.

nnethercote · 2026-02-26T06:19:51Z

Perf effects are neutral for icounts, as I'd expect.

Bootstrap numbers are interesting.

A 13.5s (-2.75%) time reduction(!)
A 1.67MB artifact size increase, but libLLVM.so (which shouldn't be affected) has a 2.01MB increase while librustc_driver.so (which would be affected) has a 360KB decrease.

This PR does eliminate 4 x 320 = 1,280 small functions in the compiler, and also eliminates 4 x 320-element arrays containing pointers to those functions. So I can imagine it could reduce bootstrap times. @Kobzol, do you know how reliable the bootstrap measurements are? I feel like I've seen large variances in the libLLVM.so size lately.

Kobzol · 2026-02-26T06:38:59Z

Bootstrap numbers have been quite noisy in the past week for some reason, yeah :( So hard to say.

panstromek · 2026-02-26T07:23:20Z

Most of the ~13s reduction is in rustc_query_impl, which spiked from 30s to 40s in #153066 (base of these perf results), so that looks like noise for the most part, except maybe those 3.5 additional seconds?

nnethercote · 2026-02-26T08:06:55Z

rustc_query_impl is the crate affected by the change, so maybe at least some of the reduction is real.

nnethercote · 2026-02-26T09:29:19Z

For my local builds librustc_driver.so drops from 699,074,376 bytes to 698,699,856 bytes, a 374,520 byte reduction. This is pretty close to the librustc_driver.so reduction of 359.56 KiB seen on CI (although the before and after sizes are much larger in the local build). I don't want to conclude too much from this measurement, but it is supporting evidence that the compiler's code size has shrunk by some non-trivial amount.

compiler/rustc_query_impl/src/execution.rs

petrochenkov · 2026-02-26T14:00:53Z

compiler/rustc_query_impl/src/plumbing.rs

-            ) {
+            let _prof_timer = tcx.sess.prof.generic_activity("self_profile_alloc_query_strings");
+
+            let mut string_cache = QueryKeyStringCache::new();


It seems like the main drawback is that some pieces of code that previously lived outside of macros now live in a macro. Those pieces are mostly tiny and trivial though.

I thought about absolutely minimizing this by moving all the code outside the $( ...$name... )* repetition into a separate function outside the macro. But in each case that extra code is so small (at most 6 lines) that I figured it wasn't worth it. (Except for gather_active_jobs, which has the 019e247 precursor.)

petrochenkov · 2026-02-26T14:05:24Z

compiler/rustc_query_impl/src/plumbing.rs

+            let mut string_cache = QueryKeyStringCache::new();
+
+            $(
                $crate::profiling_support::alloc_self_profile_query_strings_for_query_cache(


Another potential drawback is that all these hundreds of function calls can now potentially be inlined and bloat rustc_query_impl, but if the benchmarks don't show anything, then it's not an issue in practice.

It would be very strange for these very large functions to be inlined. We could add inline(never) but I don't think it's necessary. The benchmarks show, if anything, the compiler's generated code getting smaller.

petrochenkov · 2026-02-26T14:06:25Z

r=me with nits addressed.
@rustbot author

rustbot · 2026-02-26T14:06:29Z

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

And also `query_key_hash_verify` for each query. This is done by generating `query_key_hash_verify_all` and having it do things more directly.

Currently `gather_active_jobs` and `gather_active_jobs_inner` do some of the work each. This commit changes things so that `gather_active_jobs` is just a thin wrapper around `gather_active_jobs_inner`. This paves the way for removing `gather_active_jobs` in the next commit.

And also `gather_active_jobs` for each query. This is done by generating `collect_active_jobs_from_all_queries` and having it do things more directly.

And also `alloc_self_profile_query_strings` for each query. This is done by generating the top-level `alloc_self_profile_query_strings` and having it do things more directly.

And also `encode_query_results` for each cacheable query. This is done by generating `encode_all_query_results` and having it do things more directly.

rustbot · 2026-02-26T21:07:26Z

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

nnethercote · 2026-02-26T21:09:30Z

@bors r=petrochenkov

I will leave this as rollup=never because of the effects on bootstrap measurements.

rust-bors · 2026-02-26T21:09:33Z

📌 Commit 90abede has been approved by petrochenkov

It is now in the queue for this repository.

rust-bors · 2026-03-01T08:31:15Z

☀️ Test successful - CI
Approved by: petrochenkov
Duration: 3h 8m 26s
Pushing 765fd2d to main...

github-actions · 2026-03-01T08:34:27Z

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 c2c6f74 (parent) -> 765fd2d (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 765fd2d8c77a570e7069d9f30bb6d3d8fe437f9e --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

dist-aarch64-llvm-mingw: 1h 30m -> 1h 48m (+20.8%)
pr-check-1: 33m 22s -> 27m 32s (-17.4%)
dist-aarch64-apple: 1h 52m -> 2h 12m (+17.4%)
x86_64-gnu-debug: 2h 3m -> 1h 44m (-15.6%)
aarch64-apple: 3h 24m -> 2h 52m (-15.3%)
i686-gnu-2: 1h 44m -> 1h 29m (-14.4%)
dist-aarch64-msvc: 1h 41m -> 1h 56m (+14.0%)
i686-gnu-1: 2h 18m -> 2h (-13.0%)
x86_64-gnu: 2h 19m -> 2h 2m (-12.2%)
x86_64-rust-for-linux: 51m 30s -> 45m 34s (-11.5%)

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 · 2026-03-01T09:11:47Z

Finished benchmarking commit (765fd2d): comparison URL.

Overall result: no relevant changes - no action needed

@rustbot label: -perf-regression

Instruction count

This benchmark run did not return any relevant results for this metric.

Max RSS (memory usage)

Results (primary 3.5%)

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

	mean	range	count
Regressions ❌ (primary)	3.5%	[3.5%, 3.5%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	3.5%	[3.5%, 3.5%]	1

Cycles

Results (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
Regressions ❌ (secondary)	7.0%	[6.8%, 7.2%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-13.2%	[-13.2%, -13.2%]	1
All ❌✅ (primary)	-	-	0

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 479.752s -> 477.793s (-0.41%)
Artifact size: 397.58 MiB -> 397.19 MiB (-0.10%)

nnethercote · 2026-03-01T09:23:45Z

Post-merge perf result shows

a 389KiB size reduction for librustc_driver.so
a 2 second bootstrap reduction, of which 1.5 seconds are in rustc_query_impl

I think both of these measurements are somewhere close to the truth.

This comment has been minimized.

Sign in to view

rust-bors bot pushed a commit that referenced this pull request Feb 26, 2026

Auto merge of #153114 - nnethercote:rm-query-arrays, r=<try>

f8df332

Remove query function arrays

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Feb 26, 2026

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 Feb 26, 2026

rustbot assigned petrochenkov Feb 26, 2026

nnethercote marked this pull request as ready for review February 26, 2026 06:20

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 Feb 26, 2026

nnethercote force-pushed the rm-query-arrays branch from c9101b5 to 0e1d4e7 Compare February 26, 2026 09:48

petrochenkov reviewed Feb 26, 2026

View reviewed changes

rustbot added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Feb 26, 2026

nnethercote added 2 commits February 27, 2026 07:19

Remove QUERY_KEY_HASH_VERIFY.

8f0ca1d

And also `query_key_hash_verify` for each query. This is done by generating `query_key_hash_verify_all` and having it do things more directly.

nnethercote added 3 commits February 27, 2026 07:55

Remove PER_QUERY_GATHER_ACTIVE_JOBS_FNS.

6b0beec

And also `gather_active_jobs` for each query. This is done by generating `collect_active_jobs_from_all_queries` and having it do things more directly.

Remove ALLOC_SELF_PROFILE_QUERY_STRINGS.

7323750

And also `alloc_self_profile_query_strings` for each query. This is done by generating the top-level `alloc_self_profile_query_strings` and having it do things more directly.

Remove ENCODE_QUERY_RESULTS.

90abede

And also `encode_query_results` for each cacheable query. This is done by generating `encode_all_query_results` and having it do things more directly.

nnethercote force-pushed the rm-query-arrays branch from 0e1d4e7 to 90abede Compare February 26, 2026 21:07

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-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Feb 26, 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 Mar 1, 2026

rust-bors bot merged commit 765fd2d into rust-lang:main Mar 1, 2026
12 checks passed

rustbot added this to the 1.96.0 milestone Mar 1, 2026

This was referenced Mar 1, 2026

Improve the forcing/promotion functions in DepKindVTable #153122

Merged

Various small query cleanups #153169

Merged

Rollup of 5 pull requests #153245

Closed

rust-bors bot mentioned this pull request Mar 1, 2026

Rejig rustc_with_all_queries! #153161

Merged

nnethercote deleted the rm-query-arrays branch March 1, 2026 09:20

This was referenced Mar 1, 2026

Show recent PRs when a user is banned rust-lang/triagebot#2313

Merged

Rename the comments command to user-info and extend it rust-lang/triagebot#2317

Merged

Uh oh!

Conversation

nnethercote commented Feb 26, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

nnethercote commented Feb 26, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

rust-bors bot commented Feb 26, 2026

Uh oh!

This comment has been minimized.

rust-timer commented Feb 26, 2026

Overall result: no relevant changes - no action needed

Uh oh!

nnethercote commented Feb 26, 2026

Uh oh!

nnethercote commented Feb 26, 2026

Uh oh!

Kobzol commented Feb 26, 2026

Uh oh!

panstromek commented Feb 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

nnethercote commented Feb 26, 2026

Uh oh!

nnethercote commented Feb 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

petrochenkov commented Feb 26, 2026

Uh oh!

rustbot commented Feb 26, 2026

Uh oh!

rustbot commented Feb 26, 2026

Uh oh!

nnethercote commented Feb 26, 2026

Uh oh!

rust-bors bot commented Feb 26, 2026

Uh oh!

This comment has been minimized.

rust-bors bot commented Mar 1, 2026

Uh oh!

Uh oh!

github-actions bot commented Mar 1, 2026

Test differences

Job duration changes

Uh oh!

rust-timer commented Mar 1, 2026

Overall result: no relevant changes - no action needed

Uh oh!

nnethercote commented Mar 1, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

nnethercote commented Feb 26, 2026 •

edited by rustbot

Loading

panstromek commented Feb 26, 2026 •

edited

Loading

nnethercote commented Feb 26, 2026 •

edited

Loading