Turn `stdarch` into a Josh subtree #141899

Kobzol · 2025-06-02T11:55:55Z

In a similar vein as #141229, this PR makes the stdarch repository a Josh subtree (it was previously a submodule). The initial commit of stdarch upon this is based is b6e2249e388f520627544812649b77b0944e1a2e, which is the previous commit SHA of the stdarch submodule. The sync was performed according to https://hackmd.io/7pOuxnkdQDaL1Y1FQr65xg.

This was decided in rust-lang/stdarch#1655.

Test pull PR on my fork: Kobzol/stdarch#1
Test push PR on my fork: Kobzol#59

I plan to use the same Rust (miri-inspired) tooling that we use for rustc-dev-guide to enable pulls/pushes on stdarch.

Note that this repository currently doesn't have any stdarch-specific tests, so before that, the subtree should only be modified through this repository only when dealing with changes that contain "cyclical dependencies" between stdarch and rustc. The long term vision is to integrate stdarch into rust-lang/rust completely.

CC @Amanieu

r? @RalfJung (feel free to reroll)

on s390x I had some problems with the old implementation not combining with other instructions well, and using the purpose-built intrinsic is just clearer.

these implementations work with just the vector target feature, but they only get a dedicated instruction in vector-enhancements-1

Upstream Rust currently does not support printing f16s in decimal. For the intrinsics tests to work, make C++ print float16_t in the same format. Can be droppen once rust-lang#127013 is merged.

Mark all SSE SIMD-computing intrinsics as safe, except for those involving memory operations.

…EX variants of avx512

RalfJung · 2025-06-03T06:33:53Z

Thanks a lot! However, I feel like this should be approved by t-libs... maybe r? @Amanieu ?

Also it my be a good idea to wait for compiler-builtins having their first successful syncs in both directions to ensure there's no hidden pitfalls in the setup.

How hard would it be to add a smoke test that at least checks that the crates build by themselves (i.e., outside of libcore)? That should be feasible wrt CI time and should catch some of the obvious issues -- the lib.rs file is otherwise entirely unused on the rustc side.

Kobzol · 2025-06-03T06:50:53Z

Also it my be a good idea to wait for compiler-builtins having their first successful syncs in both directions to ensure there's no hidden pitfalls in the setup.

Definitely agreed.

How hard would it be to add a smoke test that at least checks that the crates build by themselves (i.e., outside of libcore)? That should be feasible wrt CI time and should catch some of the obvious issues -- the lib.rs file is otherwise entirely unused on the rustc side.

Yeah, I think we should do at least some smoke test. I guess that we could do it in bootstrap, I'll take a look.

bors · 2025-06-04T00:13:16Z

☔ The latest upstream changes (presumably #141229) made this pull request unmergeable. Please resolve the merge conflicts.

Amanieu · 2025-06-04T10:30:28Z

This is going to conflict with #141964 which updates the submodule.

folkertdev and others added 30 commits February 23, 2025 23:21

use simd_ctlz instead of an extern

1fab09a

use simd_shuffle in the implementation of vec_splat

2dbc1e9

use simd_neg in the impl crate::ops::Neg

b75bc77

on s390x I had some problems with the old implementation not combining with other instructions well, and using the purpose-built intrinsic is just clearer.

implement vec_round and friends

cefc61d

impl VectorMax/Min for vector_float/double

bed1521

these implementations work with just the vector target feature, but they only get a dedicated instruction in vector-enhancements-1

simplify the vec_abs implementation

3e3c523

use generic llvm intrinsics for rounding

d845304

start using the newly added target features for testing

055daf7

mark riscv intrinsics as safe

3ac35a7

add newly-added s390x features to is_s390x_feature_detected

67468b2

Update generator to facilitate big endian

f64b610

Update YAML & generated files

3b93df8

Remove hand crafted intrinsics that are now generated

d120278

Update test runner to support big endian

7294081

Update CI to facilitate big endian at some point

95d5d61

ensure correct linker gets chosen

f4974ae

fix non-working intrinsics

443d1cd

fix test runner for armv7

bec322f

remove print statment and correct target name

2ab4624

fix up yaml & re-generate

b95fc44

update clang to clang-19

5c10dc3

tidy up proposed Dockerfile & run.sh

1d166d4

fix - only use simd_shuffle! on intrinsics that require it

014fe0b

intrinsics_data: Add Neon fp16 intrinsics to arm_intrinsics.json

e2ce7e1

stdarch-verify: Support f16 vector types

78e6b3e

intrinsic-test: Support testing f16 intrinsics

ed7bf87

core-arch: Add Arm f16 vector types

7cdc915

core-arch: Add NEON fp16 intrinsics

f4a31fd

intrinsic-test: Print C++ float16_t in hex

3b58340

Upstream Rust currently does not support printing f16s in decimal. For the intrinsics tests to work, make C++ print float16_t in the same format. Can be droppen once rust-lang#127013 is merged.

mark SSE SIMD intrinsics as safe

74ff875

Mark all SSE SIMD-computing intrinsics as safe, except for those involving memory operations.

Jamesbarford and others added 19 commits May 3, 2025 05:06

pr feedback - take the instruction count bump out of Dockerfile

45ee41c

retry adding aarch64_be to the pipeline

af3bd7b

add printouts for debugging

052293b

change how the test is run

206c35a

Add riscv32 CI run

b1ad50f

Fix segfault in cirrus CI

f88c990

Edit macro_trailing_commas to enable tests in all architectures

949fd56

Run aarch64-pc-windows-msvc runs on the new windows-11-arm runners

546e265

Partially stabilize LoongArch target features

63235a6

Stabilize runtime detection of VEX variants of avx512

98c4ba9

Correct rustc version for the stabilization of runtime detection of V…

806848f

…EX variants of avx512

avx512_target_feature is now stable on nightly

a850e87

use the right load instruction

42e6ad3

in intrinsic-test, format f16 like C

4f35c02

use a tuple to return the condition code

7b99525

allow aarch64_softfloat_neon for backwards compatibility

b58ce62

Removed library/stdarch submodule

f282add

Merge stdarch as a Josh subtree

64131ba

Add autolabel and mentions entries to triagebot.toml

66f684f

rustbot assigned RalfJung Jun 2, 2025

rustbot added A-meta Area: Issues & PRs about the rust-lang/rust repository itself T-libs Relevant to the library team, which will review and decide on the PR/issue. labels Jun 2, 2025

Kobzol mentioned this pull request Jun 2, 2025

Turn stdarch into a subtree rust-lang/stdarch#1655

Open

rustbot assigned Amanieu and unassigned RalfJung Jun 3, 2025

bors added the S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. label Jun 4, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Turn `stdarch` into a Josh subtree #141899

Turn `stdarch` into a Josh subtree #141899

Kobzol commented Jun 2, 2025 •

edited

Loading

Uh oh!

RalfJung commented Jun 3, 2025

Uh oh!

Kobzol commented Jun 3, 2025

Uh oh!

bors commented Jun 4, 2025

Uh oh!

Amanieu commented Jun 4, 2025

Uh oh!

Uh oh!

Turn stdarch into a Josh subtree #141899

Are you sure you want to change the base?

Turn stdarch into a Josh subtree #141899

Conversation

Kobzol commented Jun 2, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

RalfJung commented Jun 3, 2025

Uh oh!

Kobzol commented Jun 3, 2025

Uh oh!

bors commented Jun 4, 2025

Uh oh!

Amanieu commented Jun 4, 2025

Uh oh!

Uh oh!

Turn `stdarch` into a Josh subtree #141899

Turn `stdarch` into a Josh subtree #141899

Kobzol commented Jun 2, 2025 •

edited

Loading