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

Ensure floats are returned losslessly by the Rust ABI on 32-bit x86 #123351

Merged
merged 4 commits into from
Jul 12, 2024

Conversation

beetrees
Copy link
Contributor

@beetrees beetrees commented Apr 2, 2024

Solves #115567 for the (default) "Rust" ABI. When compiling for 32-bit x86, this PR changes the "Rust" ABI to return floats indirectly instead of in x87 registers (with the exception of single f32s, which this PR returns in general purpose registers as they are small enough to fit in one). No change is made to the "C" ABI as that ABI requires x87 register usage and therefore will need a different solution.

@rustbot
Copy link
Collaborator

rustbot commented Apr 2, 2024

r? @compiler-errors

rustbot has assigned @compiler-errors.
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. labels Apr 2, 2024
Copy link
Member

@compiler-errors compiler-errors left a comment

Choose a reason for hiding this comment

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

Changes lgtm, but I'm not an ABI expert

@compiler-errors
Copy link
Member

so let's get another look at it

r? compiler

@rustbot rustbot assigned Nadrieril and unassigned compiler-errors Apr 2, 2024
@Nadrieril
Copy link
Member

Not an ABI expert either

r? compiler

@rustbot rustbot assigned oli-obk and unassigned Nadrieril Apr 2, 2024
@Nadrieril

This comment has been minimized.

@rustbot rustbot assigned fmease and unassigned oli-obk Apr 2, 2024
Copy link
Member

@fmease fmease left a comment

Choose a reason for hiding this comment

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

Looks reasonable to me, too. I'm also not an expert on this matter ^^'.
@RalfJung, could you help us out?

compiler/rustc_ty_utils/src/abi.rs Show resolved Hide resolved
@beetrees beetrees force-pushed the x86-ret-snan-rust branch 2 times, most recently from c413b2f to c42c4b5 Compare April 13, 2024 00:59
@beetrees
Copy link
Contributor Author

I've added a commit to update platform-support.md to reflect the current state of #115567 after this PR (I had to rebase to avoid conflicts in platform-support.md). I also fixed the issue number to point to the correct issue.

@RalfJung
Copy link
Member

Looks reasonable to me, too. I'm also not an expert on this matter ^^'. @RalfJung, could you help us out?

My name appears in these issues as I gained just enough knowledge to identify that we have a problem, but I don't feel confident to approve any changes to our ABI, sorry.

With my limited knowledge however, this change does make a lot of sense. And it's a lot simpler than what I thought was necessary to do this. Very nice!

So... who is an ABI expert? @bjorn3, @nikic and @eddyb are the first people coming to my mind.

(Maybe we should have some sort of dedicated ABI expert ping group. It would be really good to have some people in charge of ABI-related questions.)

@beetrees beetrees force-pushed the x86-ret-snan-rust branch from c42c4b5 to 852ffd3 Compare April 13, 2024 19:33
@programmerjake
Copy link
Member

programmerjake commented Apr 23, 2024

I guess it can be left for a followup, but if SSE2 is available, I think f32/f64 should be returned in xmm0 instead of in memory, since the XMM registers are where all the math is done anyway. This can be done by setting the inreg LLVM parameter attribute on the return type:

https://github.com/llvm/llvm-project/blob/575a6483062b8a77b35f48589b2acc1020195ac7/llvm/lib/Target/X86/X86CallingConv.td#L300-L304

@beetrees
Copy link
Contributor Author

@programmerjake See #115919 - currently blocked on #116584. I agree it would be a good future improvement once the target feature issues have been sorted out.

@beetrees
Copy link
Contributor Author

Picking one of the people mentioned by @RalfJung earlier:

r? @bjorn3

@rustbot rustbot assigned bjorn3 and unassigned fmease May 16, 2024
@bjorn3
Copy link
Member

bjorn3 commented May 18, 2024

I don't have time to review this for the next two weeks at least unfortunately.

@beetrees
Copy link
Contributor Author

No worries. I'll try someone else.

r? @nikic

@bors
Copy link
Contributor

bors commented Jul 12, 2024

⌛ Testing commit cae9d48 with merge c2437cb...

bors added a commit to rust-lang-ci/rust that referenced this pull request Jul 12, 2024
…workingjubilee

Ensure floats are returned losslessly by the Rust ABI on 32-bit x86

Solves rust-lang#115567 for the (default) `"Rust"` ABI. When compiling for 32-bit x86, this PR changes the `"Rust"` ABI to return floats indirectly instead of in x87 registers (with the exception of single `f32`s, which this PR returns in general purpose registers as they are small enough to fit in one). No change is made to the `"C"` ABI as that ABI requires x87 register usage and therefore will need a different solution.
@rust-log-analyzer
Copy link
Collaborator

The job x86_64-msvc-ext failed! Check out the build log: (web) (plain)

Click to see the possible cause of the failure (guessed by this bot)

@bors
Copy link
Contributor

bors commented Jul 12, 2024

💔 Test failed - checks-actions

@bors 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 Jul 12, 2024
@beetrees
Copy link
Contributor Author

error: failed to remove file `C:\a\rust\rust\build\x86_64-pc-windows-msvc\stage2-tools\x86_64-pc-windows-msvc\release\cargo.exe`

Caused by:
  Access is denied. (os error 5)

Error appears to be spurious.

@Kobzol
Copy link
Contributor

Kobzol commented Jul 12, 2024

@bors r=nikic,workingjubilee

@bors
Copy link
Contributor

bors commented Jul 12, 2024

💡 This pull request was already approved, no need to approve it again.

  • This pull request previously failed. You should add more commits to fix the bug, or use retry to trigger a build again.
  • There's another pull request that is currently being tested, blocking this pull request: [do not merge] Testing Windows CI #127373

@bors
Copy link
Contributor

bors commented Jul 12, 2024

📌 Commit cae9d48 has been approved by nikic,workingjubilee

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 Jul 12, 2024
@bors
Copy link
Contributor

bors commented Jul 12, 2024

⌛ Testing commit cae9d48 with merge c6727fc...

@bors
Copy link
Contributor

bors commented Jul 12, 2024

☀️ Test successful - checks-actions
Approved by: nikic,workingjubilee
Pushing c6727fc to master...

@bors bors added the merged-by-bors This PR was explicitly merged by bors. label Jul 12, 2024
@bors bors merged commit c6727fc into rust-lang:master Jul 12, 2024
7 checks passed
@rustbot rustbot added this to the 1.81.0 milestone Jul 12, 2024
@beetrees beetrees deleted the x86-ret-snan-rust branch July 12, 2024 23:26
@rust-timer
Copy link
Collaborator

Finished benchmarking commit (c6727fc): comparison URL.

Overall result: ✅ improvements - no action needed

@rustbot label: -perf-regression

Instruction count

This is a highly reliable metric that was used to determine the overall result at the top of this comment.

mean range count
Regressions ❌
(primary)
- - 0
Regressions ❌
(secondary)
- - 0
Improvements ✅
(primary)
- - 0
Improvements ✅
(secondary)
-2.6% [-2.6%, -2.6%] 1
All ❌✅ (primary) - - 0

Max RSS (memory usage)

This benchmark run did not return any relevant results for this metric.

Cycles

This benchmark run did not return any relevant results for this metric.

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 703.434s -> 705.609s (0.31%)
Artifact size: 328.58 MiB -> 328.61 MiB (0.01%)

@workingjubilee workingjubilee added the O-x86_32 Target: x86 processors, 32 bit (like i686-*) label Jul 18, 2024
jaisnan pushed a commit to jaisnan/rust-dev that referenced this pull request Jul 29, 2024
Update Rust toolchain from nightly-2024-07-12 to nightly-2024-07-13
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@5315cbe
up to
rust-lang@c6727fc.
The log for this commit range is:
rust-lang@c6727fc9b5 Auto merge of
rust-lang#123351 - beetrees:x86-ret-snan-rust, r=nikic,workingjubilee
rust-lang@62c068feea Auto merge of
rust-lang#127636 - nnethercote:fix-Parser-look_ahead, r=oli-obk
rust-lang@5d76a13bbe Auto merge of
rust-lang#127653 - matthiaskrgr:rollup-72bqgvp, r=matthiaskrgr
rust-lang@f11c2c8e95 Rollup merge of
rust-lang#127648 - Kobzol:ci-lower-timeout, r=pietroalbini
rust-lang@526da2366a Rollup merge of
rust-lang#127627 - lcnr:rustc_search_graph, r=compiler-errors
rust-lang@f5fa6fb602 Rollup merge of
rust-lang#127613 - nikic:riscv-update, r=cuviper
rust-lang@b4f002d2e5 Rollup merge of
rust-lang#127552 - onur-ozkan:unnecessary-git-usage, r=Kobzol
rust-lang@8ceb4e49ff Rollup merge of
rust-lang#127433 - dtolnay:conststrlen, r=workingjubilee
rust-lang@f9b3e8b387 Rollup merge of
rust-lang#126827 - the8472:pidfd-spawn, r=workingjubilee
rust-lang@18152d72a4 Rollup merge of
rust-lang#126639 - sayantn:amx, r=Amanieu
rust-lang@65ea92d4a1 Rollup merge of
rust-lang#124980 - zachs18:rc-allocator, r=Amanieu
rust-lang@05eac57ef6 Auto merge of
rust-lang#127479 - Urgau:rustc-stable-hash, r=michaelwoerister
rust-lang@15f770b143 enable fuzzing of
`SearchGraph`
rust-lang@cae9d480bf Adjust tests for x86
"Rust" ABI changes
rust-lang@3f4b9dd463 Lower timeout of CI
jobs to 4 hours
rust-lang@7f1518bddd Add instability
attribute on private const_strlen function
rust-lang@b286722878 Auto merge of
rust-lang#127635 - matthiaskrgr:rollup-foopajr, r=matthiaskrgr
rust-lang@100f3fd133 Add a new special
case to `Parser::look_ahead`.
rust-lang@ebe1305b1e Remove the bogus
special case from `Parser::look_ahead`.
rust-lang@dad95578b0 Add unit tests for
`Parser::look_ahead`.
rust-lang@ec05c4ea3f Add the feature gate
and target-features
rust-lang@c2b7842555 Rollup merge of
rust-lang#127625 - SkiFire13:revert-comment-deletion, r=workingjubilee
rust-lang@ca576eae4e Rollup merge of
rust-lang#127622 - compiler-errors:builtin-internal, r=lqd
rust-lang@fe564c10ab Rollup merge of
rust-lang#127607 - Zalathar:normalize-hint, r=wesleywiser
rust-lang@83d1a1b252 Rollup merge of
rust-lang#127596 - tesuji:help-unwrap-or, r=compiler-errors
rust-lang@1e7ad4c3ed Rollup merge of
rust-lang#127422 - greaka:master, r=workingjubilee
rust-lang@58fe37f2c3 Rollup merge of
rust-lang#127164 - Nadrieril:clean-lowering-loop, r=matthewjasper
rust-lang@4a31a6c32a Auto merge of
rust-lang#127382 - estebank:const-let, r=compiler-errors
rust-lang@5e311f933d Auto merge of
rust-lang#127614 - matthiaskrgr:rollup-8geziwi, r=matthiaskrgr
rust-lang@a776e5f922 Add doc for
deconstruct_option_or_result
rust-lang@872d7b82e1 Add suggestion for
`Option<&Vec<T>> -> Option<&[T]`
rust-lang@d9170dc666 Add regression test
for issue 127545
rust-lang@4df75140dd Fix aarch64 test
rust-lang@cbe75486f7 Account for `let foo
= expr`; to suggest `const foo: Ty = expr;`
rust-lang@b56dc8ee90 Use verbose style
when suggesting changing `const` with `let`
rust-lang@d9021791eb Revert accidental
comment deletion
rust-lang@b77d3ef7c4 Mark builtin syntax
as internal
rust-lang@fa3ce50f0b Rollup merge of
rust-lang#127605 - nikic:remove-extern-wasm, r=oli-obk
rust-lang@d433f176ef Rollup merge of
rust-lang#127601 - trevyn:issue-127600, r=compiler-errors
rust-lang@47ab86653e Rollup merge of
rust-lang#127599 - tgross35:lazy_cell_consume-rename, r=workingjubilee
rust-lang@a10b4d1463 Rollup merge of
rust-lang#127598 - weiznich:diagnostic_do_not_recommend_also_skips_help,
r=compiler-errors
rust-lang@73c500b3a7 Rollup merge of
rust-lang#127591 - compiler-errors:label-after-primary, r=lcnr
rust-lang@380c78741e Rollup merge of
rust-lang#127588 - uweigand:s390x-f16-doctests, r=tgross35
rust-lang@6fd955549a Rollup merge of
rust-lang#127572 - tbu-:pr_debug_event_nonpacked, r=jhpratt
rust-lang@8de487fdbd Rollup merge of
rust-lang#124599 - estebank:issue-41708, r=wesleywiser
rust-lang@55256c5a18 Update
dist-riscv64-linux to binutils 2.40
rust-lang@977439d9b8 Use uplifted
`rustc-stable-hash` crate in `rustc_data_structures`
rust-lang@f56b2074c6 solve -> solve/mod
rust-lang@08a2992d6b compiletest: Better
error message for bad `normalize-*` headers
rust-lang@8a50bcbdce Remove extern "wasm"
ABI
rust-lang@a01f49e7f3 check is_ident
before parse_ident
rust-lang@ab56fe2053 Rename
`lazy_cell_consume` to `lazy_cell_into_inner`
rust-lang@27d5db166e Allows
`#[diagnostic::do_not_recommend]` to supress trait impls in suggestions
as well
rust-lang@12ae282987 Fix diagnostic and
add a test for it
rust-lang@df72e478b0 Make sure that
labels are defined after the primary span in diagnostics
rust-lang@0065763950 core: Limit
remaining f16 doctests to x86_64 linux
rust-lang@45ad522e87 Don't mark
`DEBUG_EVENT` struct as `repr(packed)`
rust-lang@0134bd2e67 remove unnecessary
`git` usages
rust-lang@42772e98e0 Address review
comments
rust-lang@3e030b38ef Return the
`otherwise_block` instead of passing it as argument
rust-lang@fc40247c6b Factor out the
"process remaining candidates" cases
rust-lang@8a222ffd6b Don't try to save an
extra block
rust-lang@c5062f7318 Move or-pattern
expansion inside the main part of the algorithm
rust-lang@bff4d213fa Factor out the
special handling of or-patterns
rust-lang@5bf50e66f9 Move a function
rust-lang@53d3e6217b Stabilize
const_cstr_from_ptr (CStr::from_ptr, CStr::count_bytes)
rust-lang@585ca16e0b as_simd: fix comment
to be in line with 507583a (rust-lang#121201)
rust-lang@0f643c449a Ensure tests don't
fail on i586 in CI
rust-lang@ec0c755704 Check that we get
somewhat sane PIDs when spawning with pidfds
rust-lang@3e4e31b7bf more fine-grained
feature-detection for pidfd spawning
rust-lang@0ce361938e document safety
properties of the internal Process::new constructor
rust-lang@6687a3f7da use pidfd_spawn for
faster process creation when pidfds are requested
rust-lang@5c46acac04 document the cvt
methods
rust-lang@0e1c832dbd Update
`platform-support.md` to reflect improvements in returning floats on
32-bit x86
rust-lang@952becc0bd Ensure floats are
returned losslessly by the Rust ABI on 32-bit x86
rust-lang@a1ad6346d6 Add fn allocator
method to rc/sync::Weak. Relax Rc<T>/Arc<T>::allocator to allow unsized
T.
rust-lang@2df4f7dd8c Suggest borrowing on
fn argument that is `impl AsRef`

Co-authored-by: celinval <35149715+celinval@users.noreply.github.com>
GuillaumeGomez added a commit to GuillaumeGomez/rust that referenced this pull request Sep 6, 2024
…ilee

enable const-float-classify test, and test_next_up/down on 32bit x86

The  test_next_up/down tests have been disabled on all 32bit x86 targets, which goes too far -- they should definitely work on our (tier 1) i686 target, it is only without SSE that we might run into trouble due to rust-lang#114479. However, I cannot reproduce that trouble any more -- maybe that got fixed by rust-lang#123351?

The  const-float-classify test relied on const traits "because we can", and got disabled when const traits got removed. That's an unfortunate reduction in test coverage of our float functionality, so let's restore the test in a way that does not rely on const traits.

The const-float tests are actually testing runtime behavior as well, and I don't think that runtime behavior is covered anywhere else. Probably they shouldn't be called "const-float", but we don't have a `tests/ui/float` folder... should I create one and move them there? Are there any other ui tests that should be moved there?

I also removed some FIXME referring to not use x87 for Rust-to-Rust-calls -- that has happened in rust-lang#123351 so this got fixed indeed. Does that mean we can simplify all that float code again? I am not sure how to test it. Is running the test suite with an i586 target enough?

Cc `@tgross35` `@workingjubilee`
workingjubilee added a commit to workingjubilee/rustc that referenced this pull request Sep 11, 2024
…ilee

enable const-float-classify test, and test_next_up/down on 32bit x86

The  test_next_up/down tests have been disabled on all 32bit x86 targets, which goes too far -- they should definitely work on our (tier 1) i686 target, it is only without SSE that we might run into trouble due to rust-lang#114479. However, I cannot reproduce that trouble any more -- maybe that got fixed by rust-lang#123351?

The  const-float-classify test relied on const traits "because we can", and got disabled when const traits got removed. That's an unfortunate reduction in test coverage of our float functionality, so let's restore the test in a way that does not rely on const traits.

The const-float tests are actually testing runtime behavior as well, and I don't think that runtime behavior is covered anywhere else. Probably they shouldn't be called "const-float", but we don't have a `tests/ui/float` folder... should I create one and move them there? Are there any other ui tests that should be moved there?

I also removed some FIXME referring to not use x87 for Rust-to-Rust-calls -- that has happened in rust-lang#123351 so this got fixed indeed. Does that mean we can simplify all that float code again? I am not sure how to test it. Is running the test suite with an i586 target enough?

Cc `@tgross35` `@workingjubilee`
workingjubilee added a commit to workingjubilee/rustc that referenced this pull request Sep 11, 2024
…ilee

enable const-float-classify test, and test_next_up/down on 32bit x86

The  test_next_up/down tests have been disabled on all 32bit x86 targets, which goes too far -- they should definitely work on our (tier 1) i686 target, it is only without SSE that we might run into trouble due to rust-lang#114479. However, I cannot reproduce that trouble any more -- maybe that got fixed by rust-lang#123351?

The  const-float-classify test relied on const traits "because we can", and got disabled when const traits got removed. That's an unfortunate reduction in test coverage of our float functionality, so let's restore the test in a way that does not rely on const traits.

The const-float tests are actually testing runtime behavior as well, and I don't think that runtime behavior is covered anywhere else. Probably they shouldn't be called "const-float", but we don't have a `tests/ui/float` folder... should I create one and move them there? Are there any other ui tests that should be moved there?

I also removed some FIXME referring to not use x87 for Rust-to-Rust-calls -- that has happened in rust-lang#123351 so this got fixed indeed. Does that mean we can simplify all that float code again? I am not sure how to test it. Is running the test suite with an i586 target enough?

Cc ``@tgross35`` ``@workingjubilee``
workingjubilee added a commit to workingjubilee/rustc that referenced this pull request Sep 11, 2024
…ilee

enable const-float-classify test, and test_next_up/down on 32bit x86

The  test_next_up/down tests have been disabled on all 32bit x86 targets, which goes too far -- they should definitely work on our (tier 1) i686 target, it is only without SSE that we might run into trouble due to rust-lang#114479. However, I cannot reproduce that trouble any more -- maybe that got fixed by rust-lang#123351?

The  const-float-classify test relied on const traits "because we can", and got disabled when const traits got removed. That's an unfortunate reduction in test coverage of our float functionality, so let's restore the test in a way that does not rely on const traits.

The const-float tests are actually testing runtime behavior as well, and I don't think that runtime behavior is covered anywhere else. Probably they shouldn't be called "const-float", but we don't have a `tests/ui/float` folder... should I create one and move them there? Are there any other ui tests that should be moved there?

I also removed some FIXME referring to not use x87 for Rust-to-Rust-calls -- that has happened in rust-lang#123351 so this got fixed indeed. Does that mean we can simplify all that float code again? I am not sure how to test it. Is running the test suite with an i586 target enough?

Cc ```@tgross35``` ```@workingjubilee```
rust-timer added a commit to rust-lang-ci/rust that referenced this pull request Sep 12, 2024
Rollup merge of rust-lang#129835 - RalfJung:float-tests, r=workingjubilee

enable const-float-classify test, and test_next_up/down on 32bit x86

The  test_next_up/down tests have been disabled on all 32bit x86 targets, which goes too far -- they should definitely work on our (tier 1) i686 target, it is only without SSE that we might run into trouble due to rust-lang#114479. However, I cannot reproduce that trouble any more -- maybe that got fixed by rust-lang#123351?

The  const-float-classify test relied on const traits "because we can", and got disabled when const traits got removed. That's an unfortunate reduction in test coverage of our float functionality, so let's restore the test in a way that does not rely on const traits.

The const-float tests are actually testing runtime behavior as well, and I don't think that runtime behavior is covered anywhere else. Probably they shouldn't be called "const-float", but we don't have a `tests/ui/float` folder... should I create one and move them there? Are there any other ui tests that should be moved there?

I also removed some FIXME referring to not use x87 for Rust-to-Rust-calls -- that has happened in rust-lang#123351 so this got fixed indeed. Does that mean we can simplify all that float code again? I am not sure how to test it. Is running the test suite with an i586 target enough?

Cc ```@tgross35``` ```@workingjubilee```
github-actions bot pushed a commit to rust-lang/miri that referenced this pull request Sep 12, 2024
enable const-float-classify test, and test_next_up/down on 32bit x86

The  test_next_up/down tests have been disabled on all 32bit x86 targets, which goes too far -- they should definitely work on our (tier 1) i686 target, it is only without SSE that we might run into trouble due to rust-lang/rust#114479. However, I cannot reproduce that trouble any more -- maybe that got fixed by rust-lang/rust#123351?

The  const-float-classify test relied on const traits "because we can", and got disabled when const traits got removed. That's an unfortunate reduction in test coverage of our float functionality, so let's restore the test in a way that does not rely on const traits.

The const-float tests are actually testing runtime behavior as well, and I don't think that runtime behavior is covered anywhere else. Probably they shouldn't be called "const-float", but we don't have a `tests/ui/float` folder... should I create one and move them there? Are there any other ui tests that should be moved there?

I also removed some FIXME referring to not use x87 for Rust-to-Rust-calls -- that has happened in #123351 so this got fixed indeed. Does that mean we can simplify all that float code again? I am not sure how to test it. Is running the test suite with an i586 target enough?

Cc ```@tgross35``` ```@workingjubilee```
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
merged-by-bors This PR was explicitly merged by bors. O-x86_32 Target: x86 processors, 32 bit (like i686-*) 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.