preserve SIMD element type information by folkertdev · Pull Request #155005 · rust-lang/rust

folkertdev · 2026-04-08T20:09:35Z

Preserve the SIMD element type and provide it to LLVM for better optimization.

This is relevant for AArch64 types like int16x4x2_t, see also llvm/llvm-project#181514. Such types are defined like so:

#[repr(simd)]
struct int16x4_t([i16; 4]);

#[repr(C)]
struct int16x4x2_t(pub int16x4_t, pub int16x4_t);

Previously this would be translated to the opaque [2 x <8 x i8>], with this PR it is instead [2 x <4 x i16>]. That change is not relevant for the ABI, but using the correct type prevents bitcasts that can (indeed, do) confuse the LLVM pattern matcher.

This change will make it possible to implement the deinterleaving loads on AArch64 in a portable way (without neon-specific intrinsics), which means that e.g. Miri or the cranelift backend can run them without additional support.

discussion at #t-compiler > loss of vector element type information

rustbot · 2026-04-09T15:26:04Z

Some changes occurred in compiler/rustc_codegen_gcc

cc @antoyo, @GuillaumeGomez

Some changes occurred in compiler/rustc_codegen_cranelift

cc @bjorn3

rustbot · 2026-04-09T15:26:06Z

r? @nnethercote

rustbot has assigned @nnethercote.
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 69 candidates
Random selection from 11 candidates

folkertdev · 2026-04-09T15:32:08Z

+    // CHECK: define [2 x <1 x ptr>] @pair_ptrx1_t([2 x <1 x ptr>] {{.*}} %0)
+    #[unsafe(no_mangle)] extern "C" fn pair_ptrx1_t(x: Pair<Simd<*const (), 1>>) -> Pair<Simd<*const (), 1>> { x }
+
+    // When it fits in a 128-bit register, it's passed directly.


View changes since the review

the types above are actually used by neon intrinsics. Below are a couple that technically work but are unlikely to come up practically.

Then for any type is smaller than 128-bit padding is added which means the type information is lost (but I think that is needed for ABI reasons?). Larger types are passed indirectly, so the type information is not needed there (but we do still technically provide it, maybe it's useful elsewhere).

nnethercote · 2026-04-13T03:53:47Z

+                    Primitive::Pointer(_) => cx.type_ptr(),
+                };
+
+                let len = self.size.bytes() / hint_vector_elem.size(cx).bytes();


View changes since the review

Is it worth asserting that this division is exact?

nnethercote · 2026-04-13T03:58:09Z

+        /// by always picking i8) by codegen backends.
+        ///
+        /// The element kind is used to provide more accurate type information to the backend, which
+        /// helps with optimization (e.g. because it prevents extra bitcasts that obscure a pattern).


View changes since the review

This comment is repetitive ("optimization" appears twice) and I found "The element kind" confusing -- I thought it was referring to something new, before I realized it's just referring back to hint_vector_elem.

A suggested alternative:

The hint_vector_elem is strictly for optimization purposes. E.g. it can be used by a codegen backend to prevent extra bitcasts that obscure a pattern. Alternatively, it can be safely ignored by always picking i8.

and provide it to LLVM for better optimization

rustbot · 2026-04-13T11:27:07Z

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.

folkertdev · 2026-04-13T11:27:58Z

@rustbot ready

nnethercote

LGTM. I'm no expert on this stuff so I will cc @nikic to give them a chance to look if they want.

View changes since this review

nnethercote · 2026-04-14T03:45:06Z

Eh, good enough. Any genuine problems can be fixed in a follow-up.

@bors r+ rollup

rust-bors · 2026-04-14T03:45:10Z

📌 Commit 6f428df has been approved by nnethercote

It is now in the queue for this repository.

Rollup of 3 pull requests Successful merges: - #152530 (Use the term struct-like variant instead of anonymous structs for data of struct-like enum variants) - #155005 (preserve SIMD element type information) - #155230 (Avoid linting `doc_cfg` as unused in rustc)

Rollup merge of #155005 - folkertdev:simd-element-type-llvm, r=nnethercote preserve SIMD element type information Preserve the SIMD element type and provide it to LLVM for better optimization. This is relevant for AArch64 types like `int16x4x2_t`, see also llvm/llvm-project#181514. Such types are defined like so: ```rust #[repr(simd)] struct int16x4_t([i16; 4]); #[repr(C)] struct int16x4x2_t(pub int16x4_t, pub int16x4_t); ``` Previously this would be translated to the opaque `[2 x <8 x i8>]`, with this PR it is instead `[2 x <4 x i16>]`. That change is not relevant for the ABI, but using the correct type prevents bitcasts that can (indeed, do) confuse the LLVM pattern matcher. This change will make it possible to implement the deinterleaving loads on AArch64 in a portable way (without neon-specific intrinsics), which means that e.g. Miri or the cranelift backend can run them without additional support. discussion at [#t-compiler > loss of vector element type information](https://rust-lang.zulipchat.com/#narrow/channel/131828-t-compiler/topic/loss.20of.20vector.20element.20type.20information/with/584483611)

This comment has been minimized.

Sign in to view

folkertdev force-pushed the simd-element-type-llvm branch from c1444a0 to 05b0368 Compare April 8, 2026 20:40

This comment has been minimized.

Sign in to view

folkertdev force-pushed the simd-element-type-llvm branch 3 times, most recently from c9dd7e6 to 68ed562 Compare April 9, 2026 09:51

folkertdev marked this pull request as ready for review April 9, 2026 15:26

rustbot added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Apr 9, 2026

rustbot removed the S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. label Apr 9, 2026

rustbot assigned nnethercote Apr 9, 2026

folkertdev commented Apr 9, 2026

View reviewed changes

folkertdev changed the title ~~preseve SIMD element type information~~ preserve SIMD element type information Apr 9, 2026

programmerjake reviewed Apr 9, 2026

View reviewed changes

Comment thread compiler/rustc_codegen_llvm/src/abi.rs Outdated

programmerjake reviewed Apr 9, 2026

View reviewed changes

Comment thread tests/assembly-llvm/aarch64-vld2-s16.rs

folkertdev force-pushed the simd-element-type-llvm branch from 68ed562 to fbab8b2 Compare April 9, 2026 22:58

nnethercote 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 Apr 13, 2026

nnethercote reviewed Apr 13, 2026

View reviewed changes

preseve SIMD element type information

6f428df

and provide it to LLVM for better optimization

folkertdev force-pushed the simd-element-type-llvm branch from fbab8b2 to 6f428df Compare April 13, 2026 11:27

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 Apr 13, 2026

nnethercote approved these changes Apr 13, 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 Apr 14, 2026

jhpratt mentioned this pull request Apr 14, 2026

Rollup of 3 pull requests #155270

Merged

rust-bors bot merged commit 803a722 into rust-lang:main Apr 14, 2026
11 checks passed

rustbot added this to the 1.97.0 milestone Apr 14, 2026

folkertdev mentioned this pull request Apr 17, 2026

use deinterleaving_load! for vld2 implementations rust-lang/stdarch#2088

Draft

Uh oh!

Conversation

folkertdev commented Apr 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment has been minimized.

This comment has been minimized.

rustbot commented Apr 9, 2026

Uh oh!

rustbot commented Apr 9, 2026

Uh oh!

folkertdev Apr 9, 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!

Uh oh!

Uh oh!

nnethercote Apr 13, 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!

nnethercote Apr 13, 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!

rustbot commented Apr 13, 2026

Uh oh!

folkertdev commented Apr 13, 2026

Uh oh!

nnethercote 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 Apr 14, 2026

Uh oh!

rust-bors bot commented Apr 14, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

folkertdev commented Apr 8, 2026 •

edited

Loading

folkertdev Apr 9, 2026 •

edited by rustbot

Loading

nnethercote Apr 13, 2026 •

edited by rustbot

Loading

nnethercote Apr 13, 2026 •

edited by rustbot

Loading

nnethercote left a comment •

edited by rustbot

Loading