Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add f16 inline ASM support for 32-bit ARM #126555

Merged
merged 1 commit into from
Jun 22, 2024

Conversation

beetrees
Copy link
Contributor

Adds f16 inline ASM support for 32-bit ARM. SIMD vector types are taken from here.

Relevant issue: #125398
Tracking issue: #116909

@rustbot label +F-f16_and_f128

@rustbot
Copy link
Collaborator

rustbot commented Jun 16, 2024

r? @cjgillot

rustbot has assigned @cjgillot.
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

@rustbot 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. F-f16_and_f128 `#![feature(f16)]`, `#![feature(f128)]` labels Jun 16, 2024
@tgross35
Copy link
Contributor

r? @Amanieu probably, based on the others

),
Abi::Vector { element, count: 4 },
) if element.primitive() == Primitive::Float(Float::F16) => {
bx.bitcast(value, bx.type_f64())
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you transmute it to i16x4 instead? I'm concerned about potential issues on big-endian targets.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I've checked on the big-endian armebv7r-none-eabi and bitcasting to i16x4 and f64 generate exactly the same assembly. Unfortunately bitcasting to i16x4 causes LLVM compilation failures when only the vfp2 (or d32) target features are enabled (but works fine when the neon target feature is enabled).

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In that case we should move all vector types to be under the neon feature rather than vfp2, otherwise the exact same problem will happen for i16x4.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done. I've left a note on the relevant tracking issue (#111800). I also found a bug in the handling of explicit register classes (#126797) but I'll leave that for a separate PR as fixing it will affect the behaviour of stable (by allowing more code to compile).

| ArmInlineAsmRegClass::dreg_low16,
),
Abi::Vector { element, count: 4 },
) if element.primitive() == Primitive::Float(Float::F16) => cx.type_f64(),
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same here.

@beetrees beetrees force-pushed the f16-inline-asm-arm branch from 0a6a421 to 4b492a1 Compare June 21, 2024 17:18
@Amanieu
Copy link
Member

Amanieu commented Jun 22, 2024

@bors r+

@bors
Copy link
Contributor

bors commented Jun 22, 2024

📌 Commit 753fb07 has been approved by Amanieu

It is now in the queue for this repository.

@bors 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 Jun 22, 2024
matthiaskrgr added a commit to matthiaskrgr/rust that referenced this pull request Jun 22, 2024
…nieu

Add `f16` inline ASM support for 32-bit ARM

Adds `f16` inline ASM support for 32-bit ARM. SIMD vector types are taken from [here](https://developer.arm.com/architectures/instruction-sets/intrinsics/#f:``@navigationhierarchiesreturnbasetype=[float]&f:@navigationhierarchieselementbitsize=[16]&f:@navigationhierarchiesarchitectures=[A32]).``

Relevant issue: rust-lang#125398
Tracking issue: rust-lang#116909

``@rustbot`` label +F-f16_and_f128
bors added a commit to rust-lang-ci/rust that referenced this pull request Jun 22, 2024
…llaumeGomez

Rollup of 5 pull requests

Successful merges:

 - rust-lang#126555 (Add `f16` inline ASM support for 32-bit ARM)
 - rust-lang#126686 (Add `#[rustc_dump_{predicates,item_bounds}]`)
 - rust-lang#126723 (Fix `...` in multline code-skips in suggestions)
 - rust-lang#126731 (Bootstrap command refactoring: refactor `BootstrapCommand` (step 1))
 - rust-lang#126823 (Migrate `run-make/inline-always-many-cgu` to `rmake.rs`)

r? `@ghost`
`@rustbot` modify labels: rollup
@bors bors merged commit 07e8b3a into rust-lang:master Jun 22, 2024
6 checks passed
@rustbot rustbot added this to the 1.81.0 milestone Jun 22, 2024
rust-timer added a commit to rust-lang-ci/rust that referenced this pull request Jun 22, 2024
Rollup merge of rust-lang#126555 - beetrees:f16-inline-asm-arm, r=Amanieu

Add `f16` inline ASM support for 32-bit ARM

Adds `f16` inline ASM support for 32-bit ARM. SIMD vector types are taken from [here](https://developer.arm.com/architectures/instruction-sets/intrinsics/#f:`@navigationhierarchiesreturnbasetype=[float]&f:@navigationhierarchieselementbitsize=[16]&f:@navigationhierarchiesarchitectures=[A32]).`

Relevant issue: rust-lang#125398
Tracking issue: rust-lang#116909

`@rustbot` label +F-f16_and_f128
@beetrees beetrees deleted the f16-inline-asm-arm branch June 22, 2024 20:40
jaisnan pushed a commit to jaisnan/rust-dev that referenced this pull request Jul 29, 2024
Update Rust toolchain from nightly-2024-06-22 to nightly-2024-06-23
without any other source changes.
This is an automatically generated pull request. If any of the CI checks
fail, manual intervention is required. In such a case, review the
changes at https://github.com/rust-lang/rust from
rust-lang@c1b336c
up to
rust-lang@3cb521a.
The log for this commit range is:
rust-lang@3cb521a434 Auto merge of
rust-lang#126761 - GuillaumeGomez:unsafe_extern_blocks, r=spastorino
rust-lang@a0f01c3c10 Auto merge of
rust-lang#126838 - matthiaskrgr:rollup-qkop22o, r=matthiaskrgr
rust-lang@dc9a08f535 Rollup merge of
rust-lang#126552 - fee1-dead-contrib:rmfx, r=compiler-errors
rust-lang@162120b4fa Rollup merge of
rust-lang#126318 - Kobzol:bootstrap-perf, r=onur-ozkan
rust-lang@f3ced9d540 Rollup merge of
rust-lang#126140 - eduardosm:stabilize-fs_try_exists, r=Amanieu
rust-lang@f944afe380 Auto merge of
rust-lang#116113 - kpreid:arcmut, r=dtolnay
rust-lang@88c3db57e4 Generalize
`{Rc,Arc}::make_mut()` to unsized types.
rust-lang@a9a4830d25 Replace
`WriteCloneIntoRaw` with `CloneToUninit`.
rust-lang@ec201b8650 Add
`core::clone::CloneToUninit`.
rust-lang@81da6a6d40 Make `effects` an
incomplete feature
rust-lang@ac47dbad50 Auto merge of
rust-lang#126824 - GuillaumeGomez:rollup-sybv8o7, r=GuillaumeGomez
rust-lang@d265538016 Rollup merge of
rust-lang#126823 - GuillaumeGomez:migrate-run-make-inline-always-many-cgu,
r=Kobzol
rust-lang@25bcc7d130 Rollup merge of
rust-lang#126731 - Kobzol:bootstrap-cmd-refactor, r=onur-ozkan
rust-lang@399c5cabdd Rollup merge of
rust-lang#126723 - estebank:dot-dot-dot, r=Nadrieril
rust-lang@3ed2cd74b5 Rollup merge of
rust-lang#126686 - fmease:dump-preds-n-item-bounds, r=compiler-errors
rust-lang@07e8b3ac01 Rollup merge of
rust-lang#126555 - beetrees:f16-inline-asm-arm, r=Amanieu
rust-lang@d03d6c0fea Auto merge of
rust-lang#126750 - scottmcm:less-unlikely, r=jhpratt
rust-lang@e7dfd4a913 Migrate
`run-make/inline-always-many-cgu` to `rmake.rs`
rust-lang@d9962bb4d8 Make `read_dir`
method take a mutable callback
rust-lang@f1b0d54ca9 Auto merge of
rust-lang#126816 - weihanglo:update-cargo, r=weihanglo
rust-lang@0bd58d8122 Apply review
comments.
rust-lang@250586cb2e Wrap std `Output` in
`CommandOutput`
rust-lang@f0aceed540 Auto merge of
rust-lang#126817 - workingjubilee:rollup-0rg0k55, r=workingjubilee
rust-lang@38bd7a0fcb Add
`#[rustc_dump_{predicates,item_bounds}]`
rust-lang@1916b3d57f Rollup merge of
rust-lang#126811 - compiler-errors:tidy-ftl, r=estebank
rust-lang@539090e5cd Rollup merge of
rust-lang#126809 - estebank:wording-tweak, r=oli-obk
rust-lang@b9ab6c3501 Rollup merge of
rust-lang#126798 - miguelfrde:master, r=tmandry
rust-lang@9498d5cf2f Rollup merge of
rust-lang#126787 - Strophox:get-bytes, r=RalfJung
rust-lang@1f9793f1aa Rollup merge of
rust-lang#126722 - adwinwhite:ptr_fn_abi, r=celinval
rust-lang@84b0922565 Rollup merge of
rust-lang#126712 - Oneirical:bootest-constestllation, r=jieyouxu
rust-lang@e7956cd994 Rollup merge of
rust-lang#126530 - beetrees:f16-inline-asm-riscv, r=Amanieu
rust-lang@10e1f5d212 Auto merge of
rust-lang#124101 - the8472:pidfd-methods, r=cuviper
rust-lang@2c65a24b8c Update cargo
rust-lang@fcae62649e Auto merge of
rust-lang#126758 - spastorino:avoid-safe-outside-unsafe-blocks, r=compiler-errors
rust-lang@ffd72b1700 Fix remaining cases
rust-lang@ea681ef281 Add a tidy rule to
make sure that diagnostics don't end in periods
rust-lang@8abf149bde to extract a pidfd
we must consume the child
rust-lang@0787c7308c Add PidFd::{kill,
wait, try_wait}
rust-lang@5d5892e966 Remove stray `.`
from error message
rust-lang@d94a40516e
[fuchsia-test-runner] Remove usage of kw_only
rust-lang@771e44ebd3 Add `f16` inline ASM
support for RISC-V
rust-lang@753fb070bb Add `f16` inline ASM
support for 32-bit ARM
rust-lang@22831ed117 Do not allow safe
usafe on static and fn items
rust-lang@a6a83d3d4e bless tests
rust-lang@b512bf6f77 add as_ptr to trait
AllocBytes, fix 2 impls; add pub fn get_bytes_unchecked_raw in
allocation.rs; add pub fn get_alloc_bytes_unchecked_raw[_mut] in
memory.rs
rust-lang@02aaea1803 update intrinsic
const param counting
rust-lang@3b14b756d8 Remove
`feature(effects)` from the standard library
rust-lang@a314f7363a Stop using
`unlikely` in `strict_*` methods
rust-lang@225796a2df Add method to get
`FnAbi` of function pointer
rust-lang@630c3adb14 Add regression test
for `unsafe_extern_blocks`
rust-lang@bb9a3ef90c Implement
`unsafe_extern_blocks` feature in rustdoc
rust-lang@3c0a4bc915 rewrite
crate-name-priority to rmake
rust-lang@bc12972bcd Slightly refactor
the dumping of HIR analysis data
rust-lang@3fe4d134dd Appease `clippy`
rust-lang@c15293407f Remove unused import
rust-lang@5c4318d02c Implement `run_cmd`
in terms of `run_tracked`
rust-lang@0de7b92cc6 Remove
`run_delaying_failure`
rust-lang@e933cfb13c Remove
`run_quiet_delaying_failure`
rust-lang@949e667d3f Remove `run_quiet`
rust-lang@a12f541a18 Implement new
command execution logic
rust-lang@9fd7784b97 Fix `...` in
multline code-skips in suggestions
rust-lang@f22b5afa6a rewrite
error-writing-dependencies to rmake
rust-lang@75ee1d74a9 rewrite
relocation-model to rmake
rust-lang@87d2e61428 Add `x perf` command
for profiling the compiler using `rustc-perf`
rust-lang@fd44aca2aa Copy `rustc-fake`
binary when building the `rustc-perf` tool
rust-lang@9e0b76201b Add `RustcPerf`
bootstrap tool
rust-lang@9ec178df0b Add `cargo_args` to
`ToolBuild`
rust-lang@6a04dfe78c Rename
`std::fs::try_exists` to `std::fs::exists` and stabilize fs_try_exists

Co-authored-by: celinval <35149715+celinval@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
F-f16_and_f128 `#![feature(f16)]`, `#![feature(f128)]` S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue.
Projects
None yet
Development

Successfully merging this pull request may close these issues.

6 participants