Support repr(simd) on ADTs containing a single array field #78863

KodrAus · 2020-11-08T01:00:12Z

This is a squash and rebase of @gnzlbg's #63531

I've never actually written code in the compiler before so just fumbled my way around until it would build 😅

I imagine there'll be some work we need to do in rustc_codegen_cranelift too for this now, but might need some input from @bjorn3 to know what that is.

cc @rust-lang/project-portable-simd

This PR allows using #[repr(simd)] on ADTs containing a single array field:

 #[repr(simd)] struct S0([f32; 4]);
 #[repr(simd)] struct S1<const N: usize>([f32; N]);
 #[repr(simd)] struct S2<T, const N: usize>([T; N]);

This should allow experimenting with portable packed SIMD abstractions on nightly that make use of const generics.

rust-highfive · 2020-11-08T01:00:16Z

r? @oli-obk

(rust_highfive has picked a reviewer for you, use r? to override)

This PR allows using `#[repr(simd)]` on ADTs containing a single array field: ```rust #[repr(simd)] struct S0([f32; 4]); #[repr(simd)] struct S1<const N: usize>([f32; N]); #[repr(simd)] struct S2<T, const N: usize>([T; N]); ``` This should allow experimenting with portable packed SIMD abstractions on nightly that make use of const generics.

src/test/codegen/simd-intrinsic/simd-intrinsic-generic-extract-insert.rs

src/test/codegen/simd-intrinsic/simd-intrinsic-transmute-array.rs

src/test/ui/simd/simd-generics.rs

src/test/ui/simd/simd-intrinsic-generic-arithmetic-saturating.rs

src/test/ui/simd/simd-intrinsic-generic-arithmetic.rs

src/test/codegen/simd-intrinsic/simd-intrinsic-generic-extract-insert.rs

src/test/codegen/simd-intrinsic/simd-intrinsic-transmute-array.rs

src/test/ui/simd/simd-generics.rs

src/test/ui/simd/simd-intrinsic-generic-arithmetic-saturating.rs

src/test/ui/simd/simd-intrinsic-generic-arithmetic.rs

bjorn3 · 2020-11-08T06:45:01Z

I think clif_vector_type will already work fine: https://github.com/bjorn3/rustc_codegen_cranelift/blob/11dd005a88eeb6129c8c19ab02cba608d3b7d682/src/intrinsics/mod.rs#L195

compiler/rustc_middle/src/ty/sty.rs

src/test/codegen/simd-intrinsic/simd-intrinsic-generic-extract-insert.rs

KodrAus · 2020-11-14T23:37:49Z

I've added a compile-fail test for a #[repr(simd)] type that depends on a generic parameter, along with a note inline pointing to here and to that test.

oli-obk · 2020-11-15T12:59:05Z

@bors r+

bors · 2020-11-15T12:59:07Z

📌 Commit e217fc4 has been approved by oli-obk

Support repr(simd) on ADTs containing a single array field This is a squash and rebase of `@gnzlbg's` rust-lang#63531 I've never actually written code in the compiler before so just fumbled my way around until it would build 😅 I imagine there'll be some work we need to do in `rustc_codegen_cranelift` too for this now, but might need some input from `@bjorn3` to know what that is. cc `@rust-lang/project-portable-simd` ----- This PR allows using `#[repr(simd)]` on ADTs containing a single array field: ```rust #[repr(simd)] struct S0([f32; 4]); #[repr(simd)] struct S1<const N: usize>([f32; N]); #[repr(simd)] struct S2<T, const N: usize>([T; N]); ``` This should allow experimenting with portable packed SIMD abstractions on nightly that make use of const generics.

m-ou-se · 2020-11-17T09:03:45Z

This failed in #79122: https://github.com/rust-lang-ci/rust/runs/1410088728

@bors r-

workingjubilee · 2020-11-17T22:41:18Z

src/test/codegen/simd-intrinsic/simd-intrinsic-transmute-array.rs

+    // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* {{.*}} %{{[0-9]+}}, i8* {{.*}} %3, i64 16, i1 false)
+    // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* {{.*}} %{{[0-9]+}}, i8* {{.*}} %6, i64 16, i1 false)


14: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %2, i8* align 4 %3, i32 16, i1 false) check:25'0 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ check:25'1 ? possible intended match

On the i586 platform, it uses i32 here instead, but this test passes on the 64-bit platforms. If this is a pointer size then it needs to be allowed to vary based on platform, I suppose? Or to exclude platforms where it won't work.

Ahh, since these are regex matches we should be able to do something like:

call void @llvm.memcpy.p0i8.p0i8.{{i32|i64}}(i8* {{.*}} %{{[0-9]+}}, i8* {{.*}} %3, {{i32|i64}} 16, i1 false)

I’ll give it a try!

I'd guess you also want the i16 version for 16-bit targets such as msp430 or avr

KodrAus · 2020-11-23T19:28:04Z

@bors try

bors · 2020-11-23T22:05:22Z

💔 Test failed - checks-actions

KodrAus · 2020-11-29T00:18:56Z

@bors r=oli-obk rollup=never

bors · 2020-11-29T00:18:57Z

📌 Commit 90255c8 has been approved by oli-obk

bors · 2020-11-29T01:04:51Z

⌛ Testing commit 90255c8 with merge 592a105933bbca87806f71c11640a71131c7e949...

bors · 2020-11-29T01:41:12Z

💔 Test failed - checks-actions

KodrAus · 2020-11-29T08:37:24Z

@bors r=oli-obk rollup=never

bors · 2020-11-29T08:37:25Z

📌 Commit 354c7d0 has been approved by oli-obk

bors · 2020-11-29T09:28:16Z

⌛ Testing commit 354c7d0 with merge 760430e...

bors · 2020-11-29T11:37:36Z

☀️ Test successful - checks-actions
Approved by: oli-obk
Pushing 760430e to master...

Complement to rust-lang/rust#78863

rust-highfive assigned oli-obk Nov 8, 2020

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Nov 8, 2020

KodrAus commented Nov 8, 2020

View reviewed changes

update ui tests

9bb4202

oli-obk reviewed Nov 9, 2020

View reviewed changes

compiler/rustc_middle/src/ty/sty.rs Show resolved Hide resolved

KodrAus mentioned this pull request Nov 9, 2020

Portable SIMD Project Group rust-lang/libs-team#4

Open

KodrAus added 2 commits November 13, 2020 14:43

remove const_generics from codegen tests

540b5db

remove internal simd_size_and_ty from llvm backend

045105b

KodrAus commented Nov 13, 2020

View reviewed changes

compiler/rustc_middle/src/ty/sty.rs Outdated Show resolved Hide resolved

oli-obk reviewed Nov 13, 2020

View reviewed changes

src/test/codegen/simd-intrinsic/simd-intrinsic-generic-extract-insert.rs Show resolved Hide resolved

KodrAus added 2 commits November 14, 2020 21:12

tweak new codegen test to work on local

74f2941

add a canary test for complex repr(simd)

7565809

fix up tidy

e217fc4

bors 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 Nov 15, 2020

Dylan-DPC-zz mentioned this pull request Nov 17, 2020

Rollup of 9 pull requests #79122

Closed

bors 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-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Nov 17, 2020

workingjubilee reviewed Nov 17, 2020

View reviewed changes

generalize codegen to non 64bit platforms

b5b35a9

bors added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Nov 23, 2020

looser regex on local args

90255c8

bors 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 Nov 29, 2020

bors added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Nov 29, 2020

args may be passed by value

354c7d0

bors 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 Nov 29, 2020

bors added the merged-by-bors This PR was explicitly merged by bors. label Nov 29, 2020

bors merged commit 760430e into rust-lang:master Nov 29, 2020

rustbot added this to the 1.50.0 milestone Nov 29, 2020

bors mentioned this pull request Nov 29, 2020

Remove extra call to upvar_tys #78725

Merged

KodrAus deleted the feat/simd-array branch November 29, 2020 23:45

bjorn3 added a commit to rust-lang/rustc_codegen_cranelift that referenced this pull request Dec 3, 2020

Support #[repr(simd)] on array wrappers

d95d03a

Complement to rust-lang/rust#78863

workingjubilee mentioned this pull request Dec 4, 2020

Explore implementing SIMD types as arrays in the compiler rust-lang/portable-simd#25

Closed

workingjubilee mentioned this pull request Oct 5, 2021

Remove repr(simd) attribute and use a lang-item instead #63633

Open

RalfJung mentioned this pull request Feb 20, 2024

require simd_insert, simd_extract indices to be constants #121225

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Support repr(simd) on ADTs containing a single array field #78863

Support repr(simd) on ADTs containing a single array field #78863

KodrAus commented Nov 8, 2020

rust-highfive commented Nov 8, 2020

bjorn3 commented Nov 8, 2020

KodrAus commented Nov 14, 2020

oli-obk commented Nov 15, 2020

bors commented Nov 15, 2020

m-ou-se commented Nov 17, 2020

workingjubilee Nov 17, 2020 •

edited

Loading

KodrAus Nov 18, 2020 •

edited

Loading

programmerjake Nov 18, 2020

KodrAus commented Nov 23, 2020

bors commented Nov 23, 2020

KodrAus commented Nov 29, 2020

bors commented Nov 29, 2020

bors commented Nov 29, 2020

bors commented Nov 29, 2020

KodrAus commented Nov 29, 2020

bors commented Nov 29, 2020

bors commented Nov 29, 2020

bors commented Nov 29, 2020

		// CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* {{.}} %{{[0-9]+}}, i8 {{.*}} %3, i64 16, i1 false)
		// CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* {{.}} %{{[0-9]+}}, i8 {{.*}} %6, i64 16, i1 false)

Support repr(simd) on ADTs containing a single array field #78863

Support repr(simd) on ADTs containing a single array field #78863

Conversation

KodrAus commented Nov 8, 2020

rust-highfive commented Nov 8, 2020

bjorn3 commented Nov 8, 2020

KodrAus commented Nov 14, 2020

oli-obk commented Nov 15, 2020

bors commented Nov 15, 2020

m-ou-se commented Nov 17, 2020

workingjubilee Nov 17, 2020 • edited Loading

Choose a reason for hiding this comment

KodrAus Nov 18, 2020 • edited Loading

Choose a reason for hiding this comment

programmerjake Nov 18, 2020

Choose a reason for hiding this comment

KodrAus commented Nov 23, 2020

bors commented Nov 23, 2020

KodrAus commented Nov 29, 2020

bors commented Nov 29, 2020

bors commented Nov 29, 2020

bors commented Nov 29, 2020

KodrAus commented Nov 29, 2020

bors commented Nov 29, 2020

bors commented Nov 29, 2020

bors commented Nov 29, 2020

workingjubilee Nov 17, 2020 •

edited

Loading

KodrAus Nov 18, 2020 •

edited

Loading