target_features: sse (or at least avx2) is incompatible with soft-float ABI by RalfJung · Pull Request #160302 · rust-lang/rust

RalfJung · 2026-07-31T19:55:27Z

Enabling both the avx2 and soft-float target features is not supported by LLVM and can crash the backend. Let's preempt that with rust-level checks. (I still think there's also an LLVM bug here, it shouldn't just SIGILL on unexpected target feature configurations, but that's a different discussion.)

What is not clear to me is whether this just affects just avx2 or also avx or even sse (we don't support mmx/3dnow separately). @dianqk do you know more about this? To be safe, let's reject "sse" and therefore by implication also all other x86 vector target features.

This PR turns #[target_feature(enable = "sse")] on a softfloat target into an FCW similar to what we do on aarch64 (see #135160). The FCW only affects people building for soft-float targets which is a fairly small percentage of our overall users (and which means we cannot meaningfully crater this). I hence went for "report in deps" immediately so that the people building the actual binaries see these warnings that their upstreams probably will never see.

rustbot · 2026-07-31T19:55:32Z

r? @khyperia

rustbot has assigned @khyperia.
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 75 candidates
Random selection from 16 candidates

RalfJung · 2026-07-31T20:03:03Z

r? @workingjubilee or @dianqk

rustbot · 2026-07-31T20:03:07Z

workingjubilee is currently at their maximum review capacity.
They may take a while to respond.

RalfJung · 2026-07-31T20:09:49Z

Hm unfortunately it seems like LLVM does not crash on all functions with #[target_feature(enable = "avx2")]. This one for example works fine:

#[unsafe(no_mangle)]
#[target_feature(enable = "avx2")]
pub fn foobar(x: __m256i, y: __m256i) -> __m256i {
    _mm256_or_si256(x, y)
}

So we may have to add an FCW for this after all.

@tarcieri do you know which function is causing the trouble in dalek-cryptography/curve25519-dalek#601? All we know is that it's somewhere in poly1305...

RalfJung · 2026-07-31T20:14:09Z

Okay I have a reproducer:

#![no_std]

use core::arch::x86_64::*;

#[unsafe(no_mangle)]
#[target_feature(enable = "avx2")]
pub fn foobar(ptr: *const __m256i) -> __m256i { unsafe {
    let key = _mm256_loadu_si256(ptr);
    _mm256_and_si256(
        _mm256_permutevar8x32_epi32(key, _mm256_set_epi32(3, 7, 2, 6, 1, 5, 0, 4)),
        _mm256_set_epi32(0, -1, 0, -1, 0, -1, 0, -1),
    )
}}

dianqk · 2026-08-01T09:25:04Z

I don't know these features on x86, but the PR seems reasonable to me.

RalfJung · 2026-08-01T09:32:58Z

With Nikita on vacation, who might know which features LLVM supports on x86 in combination with +soft-float?

But I guess we can also just warn about all vector features (as this PR does now) and if we get issues saying sse actually works fine we can always adjust. 🤷

RalfJung · 2026-08-01T09:36:41Z

FWIW the s390x target also has a "soft-float" target feature and there we already mark "vector" as incompatible.

ARM also has "soft-float" and there we don't mark anything. ARM also has much more explicit ABI control so maybe setting FloatABIType to "soft" and enabling neon actually works fine there? No idea.

traviscross · 2026-08-05T07:20:58Z

Makes sense to me. Thanks @RalfJung.

@rfcbot fcp merge lang

rust-rfcbot · 2026-08-05T07:21:00Z

@traviscross has proposed to merge this. The next step is review by the rest of the tagged team members:

No concerns currently listed.

Once a majority of reviewers approve (and at most 2 approvals are outstanding), this will enter its final comment period. If you spot a major issue that hasn't been raised at any point in this process, please speak up!

cc @rust-lang/lang-advisors: FCP proposed for lang, please feel free to register concerns.
See this document for info about what commands tagged team members can give me.

rust-bors · 2026-08-23T13:03:06Z

☀️ Try build successful (CI)
Build commit: 58729dd (58729dd2855f18195fd476b08899b00a57ff537c)
Base parent: be0ea33 (be0ea33bba5000ac134e58acde780af83d908e44)

RalfJung · 2026-08-23T13:08:44Z

@bors r=workingjubilee

rust-bors · 2026-08-23T13:08:46Z

📌 Commit 68c66e8 has been approved by workingjubilee

It is now in the queue for this repository.

@dianqk

…kingjubilee target_features: sse (or at least avx2) is incompatible with soft-float ABI Fixes rust-lang#117938 Enabling both the avx2 and soft-float target features is [not supported by LLVM](rust-lang#117938 (comment)) and can crash the backend. Let's preempt that with rust-level checks. (I still think there's also an LLVM bug here, it shouldn't just SIGILL on unexpected target feature configurations, but that's a different discussion.) What is not clear to me is whether this just affects just avx2 or also avx or even sse (we don't support mmx/3dnow separately). @dianqk do you know more about this? To be safe, let's reject "sse" and therefore by implication also all other x86 vector target features. This PR turns `#[target_feature(enable = "sse")]` on a softfloat target into an FCW similar to what we do on aarch64 (see rust-lang#135160). The FCW only affects people building for soft-float targets which is a fairly small percentage of our overall users (and which means we cannot meaningfully crater this). I hence went for "report in deps" immediately so that the people building the actual binaries see these warnings that their upstreams probably will never see.

…uwer Rollup of 9 pull requests Successful merges: - #160302 (target_features: sse (or at least avx2) is incompatible with soft-float ABI) - #161586 (miri subtree update) - #155254 (Recover on attribute in use tree) - #161166 (add crashtests [5/N], remove unused aux files ) - #161294 (add crashtests [6/N]) - #161569 (Bump cfg_aliases to 0.2.2) - #161573 (re-bless `pretty-std` on windows) - #161588 (update eyre) - #161598 ([Bootstrap] Pass exact CI `llvm-config` executable path)

@dianqk

…kingjubilee target_features: sse (or at least avx2) is incompatible with soft-float ABI Fixes rust-lang#117938 Enabling both the avx2 and soft-float target features is [not supported by LLVM](rust-lang#117938 (comment)) and can crash the backend. Let's preempt that with rust-level checks. (I still think there's also an LLVM bug here, it shouldn't just SIGILL on unexpected target feature configurations, but that's a different discussion.) What is not clear to me is whether this just affects just avx2 or also avx or even sse (we don't support mmx/3dnow separately). @dianqk do you know more about this? To be safe, let's reject "sse" and therefore by implication also all other x86 vector target features. This PR turns `#[target_feature(enable = "sse")]` on a softfloat target into an FCW similar to what we do on aarch64 (see rust-lang#135160). The FCW only affects people building for soft-float targets which is a fairly small percentage of our overall users (and which means we cannot meaningfully crater this). I hence went for "report in deps" immediately so that the people building the actual binaries see these warnings that their upstreams probably will never see.

@dianqk

target_features: sse (or at least avx2) is incompatible with soft-float ABI Fixes #117938 Enabling both the avx2 and soft-float target features is [not supported by LLVM](#117938 (comment)) and can crash the backend. Let's preempt that with rust-level checks. (I still think there's also an LLVM bug here, it shouldn't just SIGILL on unexpected target feature configurations, but that's a different discussion.) What is not clear to me is whether this just affects just avx2 or also avx or even sse (we don't support mmx/3dnow separately). @dianqk do you know more about this? To be safe, let's reject "sse" and therefore by implication also all other x86 vector target features. This PR turns `#[target_feature(enable = "sse")]` on a softfloat target into an FCW similar to what we do on aarch64 (see #135160). The FCW only affects people building for soft-float targets which is a fairly small percentage of our overall users (and which means we cannot meaningfully crater this). I hence went for "report in deps" immediately so that the people building the actual binaries see these warnings that their upstreams probably will never see.

JonathanBrouwer · 2026-08-23T15:02:36Z

@bors yield
Yielding to enclosing rollup

rust-bors · 2026-08-23T15:02:39Z

Auto build was cancelled. Cancelled workflows:

https://github.com/rust-lang/rust/actions/runs/32647296368

The next pull request likely to be tested is #161605.

…uwer Rollup of 14 pull requests Successful merges: - #160302 (target_features: sse (or at least avx2) is incompatible with soft-float ABI) - #161187 (add `Complex<T>` layout tests for straightforward targets) - #161524 (Put back `tests/rustdoc-gui/search-result-display.goml`) - #161586 (miri subtree update) - #155254 (Recover on attribute in use tree) - #158695 (Replace `CrateDump` with the Debug impl from `CStore`) - #160452 (-Ctarget-feature is not unsafe (any more)) - #161007 (tests/ui/union/union-nodrop.rs: fix typo "expressios") - #161166 (add crashtests [5/N], remove unused aux files ) - #161569 (Bump cfg_aliases to 0.2.2) - #161573 (re-bless `pretty-std` on windows) - #161588 (update eyre) - #161589 (std: reduce visibility of some internal OsStr related types) - #161598 ([Bootstrap] Pass exact CI `llvm-config` executable path)

RalfJung · 2026-08-23T15:13:40Z

You're brave, rolling this up a 2nd time.^^

JonathanBrouwer · 2026-08-23T15:41:16Z

My idea there is that if the rollup fails because of this PR, we don't waste any queue time because this PR would've failed regardless and used the same amount of time

@dianqk

target_features: sse (or at least avx2) is incompatible with soft-float ABI Fixes #117938 Enabling both the avx2 and soft-float target features is [not supported by LLVM](#117938 (comment)) and can crash the backend. Let's preempt that with rust-level checks. (I still think there's also an LLVM bug here, it shouldn't just SIGILL on unexpected target feature configurations, but that's a different discussion.) What is not clear to me is whether this just affects just avx2 or also avx or even sse (we don't support mmx/3dnow separately). @dianqk do you know more about this? To be safe, let's reject "sse" and therefore by implication also all other x86 vector target features. This PR turns `#[target_feature(enable = "sse")]` on a softfloat target into an FCW similar to what we do on aarch64 (see #135160). The FCW only affects people building for soft-float targets which is a fairly small percentage of our overall users (and which means we cannot meaningfully crater this). I hence went for "report in deps" immediately so that the people building the actual binaries see these warnings that their upstreams probably will never see.

JonathanBrouwer · 2026-08-23T16:09:13Z

@bors yield

rust-bors · 2026-08-23T16:09:16Z

Auto build was cancelled. Cancelled workflows:

https://github.com/rust-lang/rust/actions/runs/32650693547

The next pull request likely to be tested is #161605.

…uwer Rollup of 14 pull requests Successful merges: - #160302 (target_features: sse (or at least avx2) is incompatible with soft-float ABI) - #161187 (add `Complex<T>` layout tests for straightforward targets) - #161524 (Put back `tests/rustdoc-gui/search-result-display.goml`) - #161586 (miri subtree update) - #155254 (Recover on attribute in use tree) - #158695 (Replace `CrateDump` with the Debug impl from `CStore`) - #160452 (-Ctarget-feature is not unsafe (any more)) - #161007 (tests/ui/union/union-nodrop.rs: fix typo "expressios") - #161166 (add crashtests [5/N], remove unused aux files ) - #161569 (Bump cfg_aliases to 0.2.2) - #161573 (re-bless `pretty-std` on windows) - #161588 (update eyre) - #161589 (std: reduce visibility of some internal OsStr related types) - #161598 ([Bootstrap] Pass exact CI `llvm-config` executable path)

rust-bors · 2026-08-23T21:33:03Z

☀️ Test successful - CI
Approved by: workingjubilee
Duration: 3h 6m 47s
Pushing fb6531d to main...

github-actions · 2026-08-23T21:36:07Z

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 9a4ad59 (parent) -> fb6531d (this PR)

Test differences

Show 662 test diffs

Stage 1

[ui] tests/ui/target-feature/abi-incompatible-target-feature-attribute-fcw.rs: pass -> [missing] (J1)
[ui] tests/ui/target-feature/abi-incompatible-target-feature-attribute-fcw.rs#aarch64: [missing] -> pass (J1)
[ui] tests/ui/target-feature/abi-incompatible-target-feature-attribute-fcw.rs#x86_64: [missing] -> pass (J1)
[ui (polonius)] tests/ui/target-feature/abi-incompatible-target-feature-attribute-fcw.rs: pass -> [missing] (J3)
[ui (polonius)] tests/ui/target-feature/abi-incompatible-target-feature-attribute-fcw.rs#aarch64: [missing] -> pass (J3)
[ui (polonius)] tests/ui/target-feature/abi-incompatible-target-feature-attribute-fcw.rs#x86_64: [missing] -> pass (J3)

Stage 2

[ui] tests/ui/target-feature/abi-incompatible-target-feature-attribute-fcw.rs: ignore (gcc backend is marked as ignore) -> [missing] (J0)
[ui] tests/ui/target-feature/abi-incompatible-target-feature-attribute-fcw.rs#aarch64: [missing] -> ignore (gcc backend is marked as ignore) (J0)
[ui] tests/ui/target-feature/abi-incompatible-target-feature-attribute-fcw.rs#x86_64: [missing] -> ignore (gcc backend is marked as ignore) (J0)
[ui] tests/ui/target-feature/abi-incompatible-target-feature-attribute-fcw.rs: pass -> [missing] (J2)
[ui] tests/ui/target-feature/abi-incompatible-target-feature-attribute-fcw.rs#aarch64: [missing] -> pass (J2)
[ui] tests/ui/target-feature/abi-incompatible-target-feature-attribute-fcw.rs#x86_64: [missing] -> pass (J2)

Additionally, 650 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 fb6531d550e0075b9eb9a51464f404805eec87d9 --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 30m -> 2h 10m (+43.4%)
i686-msvc-1: 3h 7m -> 1h 47m (-42.4%)
x86_64-msvc-ext3: 1h 58m -> 1h 12m (-38.4%)
dist-x86_64-solaris: 1h 14m -> 1h 42m (+38.1%)
pr-check-2: 33m -> 45m 30s (+37.9%)
x86_64-msvc-ext1: 2h 16m -> 1h 26m (-36.5%)
x86_64-gnu-llvm-21-1: 43m 2s -> 58m 21s (+35.6%)
aarch64-apple-macos-26-2: 1h 40m -> 2h 14m (+34.7%)
arm-android: 1h 22m -> 1h 48m (+31.9%)
x86_64-gnu-gcc: 57m 27s -> 1h 15m (+31.0%)

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 · 2026-08-23T22:15:59Z

Finished benchmarking commit (fb6531d): comparison URL.

Overall result: no relevant changes - no action needed

@rustbot label: -perf-regression

Instruction count

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

Max RSS (memory usage)

Results (secondary 3.9%)

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)	3.9%	[1.2%, 6.2%]	3
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

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: 468.884s -> 469.882s (0.21%)
Artifact size: 400.29 MiB -> 400.35 MiB (0.02%)

rustbot assigned khyperia 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

RalfJung mentioned this pull request Jul 31, 2026

LLVM produces SIGILL when enabling avx2 target feature on x86_64-unknown-none #117938

Closed

RalfJung changed the title ~~target_featurs: avx2 is incompatible with soft-float ABI~~ target_features: avx2 is incompatible with soft-float ABI Jul 31, 2026

rustbot assigned workingjubilee Jul 31, 2026

rustbot unassigned khyperia Jul 31, 2026

RalfJung force-pushed the soft-float-no-avx2 branch 2 times, most recently from 58062d1 to 21a23e9 Compare July 31, 2026 20:47

RalfJung added the I-lang-nominated Nominated for discussion during a lang team meeting. label Jul 31, 2026

RalfJung changed the title ~~target_features: avx2 is incompatible with soft-float ABI~~ target_features: sse (or at least avx2) is incompatible with soft-float ABI Jul 31, 2026

This comment has been minimized.

Sign in to view

RalfJung force-pushed the soft-float-no-avx2 branch from 21a23e9 to 5594310 Compare August 5, 2026 06:30

This comment has been minimized.

Sign in to view

traviscross added I-lang-radar Items that are on lang's radar and will need eventual work or consideration. P-lang-drag-1 Lang team prioritization drag level 1. https://rust-lang.zulipchat.com/#narrow/channel/410516-t-lang labels Aug 5, 2026

traviscross reviewed Aug 5, 2026

View reviewed changes

Comment thread compiler/rustc_lint_defs/src/builtin.rs

traviscross added the T-lang Relevant to the language team label Aug 5, 2026

rust-rfcbot added proposed-final-comment-period Proposed to merge/close by relevant subteam, see T-<team> label. Will enter FCP once signed off. disposition-merge This issue / PR is in PFCP or FCP with a disposition to merge it. labels Aug 5, 2026

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 Aug 23, 2026

JonathanBrouwer mentioned this pull request Aug 23, 2026

Rollup of 9 pull requests #161599

Closed

JonathanBrouwer mentioned this pull request Aug 23, 2026

Rollup of 14 pull requests #161605

Closed

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 Aug 23, 2026

rust-bors Bot merged commit fb6531d into rust-lang:main Aug 23, 2026
15 checks passed

rustbot added this to the 1.100.0 milestone Aug 23, 2026

Uh oh!

Uh oh!

Conversation

RalfJung commented Jul 31, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented Jul 31, 2026

Uh oh!

RalfJung commented Jul 31, 2026

Uh oh!

rustbot commented Jul 31, 2026

Uh oh!

RalfJung commented Jul 31, 2026

Uh oh!

RalfJung commented Jul 31, 2026

Uh oh!

dianqk commented Aug 1, 2026

Uh oh!

RalfJung commented Aug 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

RalfJung commented Aug 1, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

Uh oh!

traviscross commented Aug 5, 2026

Uh oh!

rust-rfcbot commented Aug 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rust-bors Bot commented Aug 23, 2026

Uh oh!

RalfJung commented Aug 23, 2026

Uh oh!

rust-bors Bot commented Aug 23, 2026

Uh oh!

This comment has been minimized.

JonathanBrouwer commented Aug 23, 2026

Uh oh!

rust-bors Bot commented Aug 23, 2026

Uh oh!

RalfJung commented Aug 23, 2026

Uh oh!

JonathanBrouwer commented Aug 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment has been minimized.

JonathanBrouwer commented Aug 23, 2026

Uh oh!

rust-bors Bot commented Aug 23, 2026

Uh oh!

This comment has been minimized.

rust-bors Bot commented Aug 23, 2026

Uh oh!

Uh oh!

github-actions Bot commented Aug 23, 2026

Test differences

Stage 1

Stage 2

Job duration changes

Uh oh!

rust-timer commented Aug 23, 2026

Overall result: no relevant changes - no action needed

Instruction count

Max RSS (memory usage)

Cycles

Binary size

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

13 participants

RalfJung commented Jul 31, 2026 •

edited by rustbot

Loading

RalfJung commented Aug 1, 2026 •

edited

Loading

rust-rfcbot commented Aug 5, 2026 •

edited

Loading

JonathanBrouwer commented Aug 23, 2026 •

edited

Loading