codegen: skip stores for entirely-uninit constant aggregate fields by glandium · Pull Request #157797 · rust-lang/rust

glandium · 2026-06-12T08:10:46Z

MIR GVN (since #147827) propagates MaybeUninit::uninit() as const <uninit> in aggregate constructions. Without this fix, codegen would emit a memcpy from an [N x i8] undef global for each such field, which LLVM materializes as zero-initialization.

This mirrors the existing all_bytes_uninit skip already present for Rvalue::Use (added in #147827) into the Rvalue::Aggregate field loop.

Fixes: #157743

rustbot · 2026-06-12T08:10:53Z

r? @jackh726

rustbot has assigned @jackh726.
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 73 candidates
Random selection from 20 candidates

jackh726 · 2026-06-25T04:31:57Z

Uhh...not code I'm familiar with. Let's see if a reroll helps.

@rustbot reroll

mu001999 · 2026-06-25T05:04:53Z

@rustbot reroll

chenyukang · 2026-06-26T14:30:00Z

@rustbot reroll

oli-obk

do you have capacity to try this out as a general fix?

View changes since this review

rustbot · 2026-06-26T16:19:31Z

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

rustbot · 2026-06-30T07:30:48Z

rustc_codegen_gcc is developed in its own repository. If possible, consider making this change to rust-lang/rustc_codegen_gcc instead.

cc @antoyo, @GuillaumeGomez

oli-obk

@bors r+

View changes since this review

rust-bors · 2026-07-22T10:00:09Z

📌 Commit ed9de1e has been approved by oli-obk

It is now in the queue for this repository.

oli-obk · 2026-07-22T10:03:20Z

@rustbot label: +perf-regression-triaged

Perf looks clean and slightly improving things across the board. There is one regression (building cranelift with opts) that has a 16% compile-time perf hit entirely in LLVM, so this change tips the optimizations over the edge somewhere and causes lots of changes. I have not checked whether it produces better runtime code, but let's land it regardless

cc @bjorn3

JonathanBrouwer · 2026-07-22T21:22:14Z

@bors p=6
Scheduling

rust-bors · 2026-07-23T02:05:50Z

☀️ Test successful - CI
Approved by: oli-obk
Duration: 3h 14m 57s
Pushing e7795af to main...

github-actions · 2026-07-23T02:08:46Z

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 6f72b5d (parent) -> e7795af (this PR)

Test differences

Show 4 test diffs

Stage 1

[codegen] tests/codegen-llvm/uninit-aggregate-field.rs: [missing] -> pass (J1)

Stage 2

[codegen] tests/codegen-llvm/uninit-aggregate-field.rs: [missing] -> pass (J0)

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

Job group index

Test dashboard

Run

cargo run --manifest-path src/ci/citool/Cargo.toml -- \
    test-dashboard e7795af6d2449fb05a6393c3320ced873a999eb3 --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

i686-gnu-2: 1h 35m -> 54m 40s (-42.8%)
x86_64-mingw-1: 2h 56m -> 1h 56m (-33.8%)
aarch64-apple-macos-26: 2h 24m -> 3h 7m (+29.7%)
dist-ohos-armv7: 56m 11s -> 1h 10m (+24.8%)
dist-various-2: 35m 49s -> 44m 31s (+24.3%)
dist-armhf-linux: 1h 28m -> 1h 7m (-23.2%)
dist-sparcv9-solaris: 1h 32m -> 1h 11m (-22.4%)
dist-x86_64-freebsd: 1h 26m -> 1h 7m (-22.0%)
i686-gnu-1: 2h 2m -> 1h 36m (-21.0%)
dist-x86_64-msvc: 1h 48m -> 2h 10m (+20.4%)

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 · 2026-07-23T02:47:52Z

Finished benchmarking commit (e7795af): comparison URL.

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

Our benchmarks found a performance regression caused by this PR.
This might be an actual regression, but it can also be just noise.

Next Steps:

If the regression was expected or you think it can be justified,
please write a comment with sufficient written justification, and add
@rustbot label: +perf-regression-triaged to it, to mark the regression as triaged.
If you think that you know of a way to resolve the regression, try to create
a new PR with a fix for the regression.
If you do not understand the regression or you think that it is just noise,
you can ask the @rust-lang/wg-compiler-performance working group for help (members of this group
were already notified of this PR).

@rustbot label: +perf-regression
cc @rust-lang/wg-compiler-performance

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)	16.2%	[16.2%, 16.2%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-0.3%	[-0.5%, -0.2%]	8
All ❌✅ (primary)	16.2%	[16.2%, 16.2%]	1

Max RSS (memory usage)

Results (primary 5.5%, secondary 5.7%)

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

	mean	range	count
Regressions ❌ (primary)	5.5%	[3.2%, 7.6%]	3
Regressions ❌ (secondary)	5.7%	[4.2%, 7.2%]	2
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	5.5%	[3.2%, 7.6%]	3

Cycles

Results (primary 12.5%, secondary -10.6%)

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

	mean	range	count
Regressions ❌ (primary)	12.5%	[12.5%, 12.5%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-10.6%	[-18.5%, -2.8%]	2
All ❌✅ (primary)	12.5%	[12.5%, 12.5%]	1

Binary size

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

Bootstrap: 486.811s -> 486.654s (-0.03%)
Artifact size: 387.67 MiB -> 387.58 MiB (-0.02%)

tmiasko · 2026-07-24T21:55:43Z

        for arg in args {
            match arg.val {
                OperandValue::ZeroSized => {}
+                OperandValue::Uninit => {}


Skipping uninit arguments looks wrong (similar for gcc implementation):

#![feature(link_llvm_intrinsics)] use std::mem::*; extern "C" { #[link_name = "llvm.sqrt.f32"] fn sqrt(x: f32) -> f32; } fn main() { std::hint::black_box(unsafe { sqrt(MaybeUninit::uninit().assume_init()) }); }

$ env RUSTC_BOOTSTRAP=1 rustc +beta -O i.rs $ env RUSTC_BOOTSTRAP=1 rustc +nightly -O i.rs error: failed to parse bitcode for LTO module: Invalid record (Producer: 'LLVM22.1.8-rust-1.99.0-nightly' Reader: 'LLVM 22.1.8-rust-1.99.0-nightly') error: aborting due to 1 previous error; 2 warnings emitted

codegen: handle OperandValue::Uninit in codegen_return_terminator Fixes rust-lang#159815. rust-lang#157797 added `OperandValue::Uninit` to skip stores for entirely-uninit constants, but missed the `PassMode::Direct | PassMode::Pair` branch of `codegen_return_terminator`, which called `immediate_or_packed_pair` unconditionally. A function directly returning an all-uninit value (e.g. `MaybeUninit::uninit()`) would hit that branch and ICE in `OperandRef::immediate`, as reported against tokio and reduced by tmiasko.

codegen: handle OperandValue::Uninit in codegen_return_terminator Fixes #159815. #157797 added `OperandValue::Uninit` to skip stores for entirely-uninit constants, but missed the `PassMode::Direct | PassMode::Pair` branch of `codegen_return_terminator`, which called `immediate_or_packed_pair` unconditionally. A function directly returning an all-uninit value (e.g. `MaybeUninit::uninit()`) would hit that branch and ICE in `OperandRef::immediate`, as reported against tokio and reduced by tmiasko.

codegen: handle OperandValue::Uninit in codegen_return_terminator Fixes rust-lang#159815. rust-lang#157797 added `OperandValue::Uninit` to skip stores for entirely-uninit constants, but missed the `PassMode::Direct | PassMode::Pair` branch of `codegen_return_terminator`, which called `immediate_or_packed_pair` unconditionally. A function directly returning an all-uninit value (e.g. `MaybeUninit::uninit()`) would hit that branch and ICE in `OperandRef::immediate`, as reported against tokio and reduced by tmiasko.

codegen: handle OperandValue::Uninit in codegen_return_terminator Fixes #159815. #157797 added `OperandValue::Uninit` to skip stores for entirely-uninit constants, but missed the `PassMode::Direct | PassMode::Pair` branch of `codegen_return_terminator`, which called `immediate_or_packed_pair` unconditionally. A function directly returning an all-uninit value (e.g. `MaybeUninit::uninit()`) would hit that branch and ICE in `OperandRef::immediate`, as reported against tokio and reduced by tmiasko.

scottmcm · 2026-07-28T05:33:02Z

    #[must_use]
    pub(crate) fn is_expected_variant_for_type<'tcx>(&self, ty: TyAndLayout<'tcx>) -> bool {
        match (self, ty.backend_repr) {
+            (OperandValue::Uninit, _) => true,


This is unfortunate because it means that there's no longer a fixed variant if you know the layout. It risks every https://doc.rust-lang.org/nightly/nightly-rustc/rustc_codegen_ssa/mir/operand/struct.OperandRef.html#method.immediate being a potential ICE now when it wasn't before :/

(Or other similar problems -- just fixing that method wouldn't be enough.)

@oli-obk

codegen: skip stores for entirely-uninit constant aggregate fields, attempt rust-lang#2 This reverts both rust-lang#159825 and rust-lang#157797, and goes back to the original fix from rust-lang#157797, albeit with a small refactor to share code between the now three branches that do the same check. This preserves the original issue (rust-lang#157743) in a fixed state, but avoids the whack-a-mole the initially landed attempt created. Fixes: rust-lang#159867 r? @oli-obk

@oli-obk

Rollup merge of #160060 - glandium:issue159867, r=oli-obk codegen: skip stores for entirely-uninit constant aggregate fields, attempt #2 This reverts both #159825 and #157797, and goes back to the original fix from #157797, albeit with a small refactor to share code between the now three branches that do the same check. This preserves the original issue (#157743) in a fixed state, but avoids the whack-a-mole the initially landed attempt created. Fixes: #159867 r? @oli-obk

rustbot assigned jackh726 Jun 12, 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 Jun 12, 2026

This comment has been minimized.

Sign in to view

glandium force-pushed the uninit branch from 779bd92 to 18d4354 Compare June 12, 2026 13:06

This comment has been minimized.

Sign in to view

rustbot assigned mu001999 and unassigned jackh726 Jun 25, 2026

rustbot assigned chenyukang and unassigned mu001999 Jun 25, 2026

rustbot assigned oli-obk and unassigned chenyukang Jun 26, 2026

oli-obk requested changes Jun 26, 2026

View reviewed changes

Comment thread compiler/rustc_codegen_ssa/src/mir/rvalue.rs Outdated

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 Jun 26, 2026

glandium force-pushed the uninit branch from 18d4354 to 6179425 Compare June 30, 2026 07:30

rustbot added the A-LLVM Area: Code generation parts specific to LLVM. Both correctness bugs and optimization-related issues. label Jun 30, 2026

This comment has been minimized.

Sign in to view

glandium requested a review from oli-obk June 30, 2026 07:31

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 Jun 30, 2026

This comment has been minimized.

Sign in to view

glandium force-pushed the uninit branch from 6179425 to 4b16700 Compare June 30, 2026 14:57

rustbot removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Jul 21, 2026

oli-obk approved these changes Jul 22, 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 Jul 22, 2026

rustbot added the perf-regression-triaged The performance regression has been triaged. label Jul 22, 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 Jul 23, 2026

rust-bors Bot merged commit e7795af into rust-lang:main Jul 23, 2026
15 checks passed

rustbot added this to the 1.99.0 milestone Jul 23, 2026

tmiasko mentioned this pull request Jul 24, 2026

[ICE]: thread 'rustc' (2584) panicked at ... compiler/rustc_codegen_ssa/src/mir/operand.rs:291:18: #159815

Closed

glandium mentioned this pull request Jul 24, 2026

codegen: handle OperandValue::Uninit in codegen_return_terminator #159825

Merged

tmiasko mentioned this pull request Jul 24, 2026

[ICE]: operand.rs:114:32: OperandValue cannot be a pointer: Uninit #159867

Closed

tmiasko reviewed Jul 24, 2026

View reviewed changes

scottmcm reviewed Jul 28, 2026

View reviewed changes

glandium mentioned this pull request Jul 28, 2026

codegen: skip stores for entirely-uninit constant aggregate fields, attempt #2 #160060

Merged

Uh oh!

Uh oh!

Conversation

glandium commented Jun 12, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented Jun 12, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

jackh726 commented Jun 25, 2026

Uh oh!

mu001999 commented Jun 25, 2026

Uh oh!

chenyukang commented Jun 26, 2026

Uh oh!

oli-obk 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!

Uh oh!

rustbot commented Jun 26, 2026

Uh oh!

rustbot commented Jun 30, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

oli-obk 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!

rust-bors Bot commented Jul 22, 2026

Uh oh!

oli-obk commented Jul 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

JonathanBrouwer commented Jul 22, 2026

Uh oh!

This comment has been minimized.

rust-bors Bot commented Jul 23, 2026

Uh oh!

Uh oh!

github-actions Bot commented Jul 23, 2026

Test differences

Stage 1

Stage 2

Job duration changes

Uh oh!

rust-timer commented Jul 23, 2026

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

Instruction count

Max RSS (memory usage)

Cycles

Binary size

Uh oh!

tmiasko Jul 24, 2026

Choose a reason for hiding this comment

Uh oh!

scottmcm Jul 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

12 participants

glandium commented Jun 12, 2026 •

edited by rustbot

Loading

oli-obk left a comment •

edited by rustbot

Loading

oli-obk left a comment •

edited by rustbot

Loading

oli-obk commented Jul 22, 2026 •

edited

Loading

scottmcm Jul 28, 2026 •

edited

Loading