-
Notifications
You must be signed in to change notification settings - Fork 12.7k
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
Refactor powerpc64
call ABI handling
#128643
Conversation
r? @bjorn3 (feel free to reroll, or, I can open a T-compiler thread asking for a suitable reviewer) |
cc @ecnelises and @bzEq since you were listed as powerpc64-ibm-aix target maintainers, and this affects your target AFAICT. |
Is a codegen regression test possible here? |
An assembly test would be nice. |
Thank you! This looks right to me. I'll do a test run on my PPC64 musl workstation soon (it is currently doing some LLVM work) and will let you know the result. |
Thanks for heads up. However, AIX has its own ABI and is not using ELFv1/v2. Anyway, I'll test this PR on AIX. |
Further, I don't think AIX ABI is a blocker for this PR. If it breaks AIX in our buildbot, we will upstream AIX change. |
I've added an assembly test for the PowerPC64 struct ABI. |
This comment has been minimized.
This comment has been minimized.
I've just queued the latest version of this patch (dd2fc07) on my powerpc64-musl build system, and will let you know the result in about two hours when it finishes. |
This comment has been minimized.
This comment has been minimized.
All tests pass here on powerpc64-musl (BE)! Including the new tests that the CI bot seems to believe are failing:
This system is using LLVM 18.1.8, so perhaps it is an LLVM 17 issue. |
1c55e73
to
6308d13
Compare
I've fixed the test so that it works on LLVM 17. |
By the way did you try abi-cafe with these changes? It is pretty good at finding abi divergences between compilers. |
r=me if abi-cafe passes. |
I've ran |
I see the same on powerpc64-musl here; i128/u128 failures but otherwise |
Thanls for testing. The u128/i128 issue can be fixed later. @bors r+ |
Refactor `powerpc64` call ABI handling As the [specification](https://openpowerfoundation.org/specifications/64bitelfabi/) for the ELFv2 ABI states that returned aggregates are returned like arguments as long as they are at most two doublewords, I've merged the `classify_arg` and `classify_ret` functions to reduce code duplication. The only functional change is to fix rust-lang#128579: the `classify_ret` function was incorrectly handling aggregates where `bits > 64 && bits < 128`. I've used the aggregate handling implementation from `classify_arg` which doesn't have this issue. `@awilfox` could you test this on `powerpc64-unknown-linux-musl`? I'm only able to cross-test on `powerpc64-unknown-linux-gnu` and `powerpc64le-unknown-linux-gnu` locally at the moment, and as a tier 3 target `powerpc64-unknown-linux-musl` has zero CI coverage. Fixes: rust-lang#128579
Refactor `powerpc64` call ABI handling As the [specification](https://openpowerfoundation.org/specifications/64bitelfabi/) for the ELFv2 ABI states that returned aggregates are returned like arguments as long as they are at most two doublewords, I've merged the `classify_arg` and `classify_ret` functions to reduce code duplication. The only functional change is to fix rust-lang#128579: the `classify_ret` function was incorrectly handling aggregates where `bits > 64 && bits < 128`. I've used the aggregate handling implementation from `classify_arg` which doesn't have this issue. ``@awilfox`` could you test this on `powerpc64-unknown-linux-musl`? I'm only able to cross-test on `powerpc64-unknown-linux-gnu` and `powerpc64le-unknown-linux-gnu` locally at the moment, and as a tier 3 target `powerpc64-unknown-linux-musl` has zero CI coverage. Fixes: rust-lang#128579
Rollup of 5 pull requests Successful merges: - rust-lang#128643 (Refactor `powerpc64` call ABI handling) - rust-lang#128873 (Add windows-targets crate to std's sysroot) - rust-lang#128916 (Tidy up `dump-ice-to-disk` and make assertion failures dump ICE messages) - rust-lang#128929 (Fix codegen-units tests that were disabled 8 years ago) - rust-lang#128937 (Fix warnings in rmake tests on `x86_64-unknown-linux-gnu`) r? `@ghost` `@rustbot` modify labels: rollup
Failed in #128968 (comment).
|
6308d13
to
715728f
Compare
I've fixed the test. |
@bors r+ |
…iaskrgr Rollup of 7 pull requests Successful merges: - rust-lang#128643 (Refactor `powerpc64` call ABI handling) - rust-lang#128655 (std: refactor UNIX random data generation) - rust-lang#128745 (Remove unused lifetime parameter from spawn_unchecked) - rust-lang#128841 (bootstrap: don't use rustflags for `--rustc-args`) - rust-lang#128983 (Slightly refactor `TargetSelection` in bootstrap) - rust-lang#129026 (CFI: Move CFI ui tests to cfi directory) - rust-lang#129040 (Fix blessing of rmake tests) r? `@ghost` `@rustbot` modify labels: rollup
Rollup merge of rust-lang#128643 - beetrees:ppc64-abi-fix, r=bjorn3 Refactor `powerpc64` call ABI handling As the [specification](https://openpowerfoundation.org/specifications/64bitelfabi/) for the ELFv2 ABI states that returned aggregates are returned like arguments as long as they are at most two doublewords, I've merged the `classify_arg` and `classify_ret` functions to reduce code duplication. The only functional change is to fix rust-lang#128579: the `classify_ret` function was incorrectly handling aggregates where `bits > 64 && bits < 128`. I've used the aggregate handling implementation from `classify_arg` which doesn't have this issue. `@awilfox` could you test this on `powerpc64-unknown-linux-musl`? I'm only able to cross-test on `powerpc64-unknown-linux-gnu` and `powerpc64le-unknown-linux-gnu` locally at the moment, and as a tier 3 target `powerpc64-unknown-linux-musl` has zero CI coverage. Fixes: rust-lang#128579
As the specification for the ELFv2 ABI states that returned aggregates are returned like arguments as long as they are at most two doublewords, I've merged the
classify_arg
andclassify_ret
functions to reduce code duplication. The only functional change is to fix #128579: theclassify_ret
function was incorrectly handling aggregates wherebits > 64 && bits < 128
. I've used the aggregate handling implementation fromclassify_arg
which doesn't have this issue.@awilfox could you test this on
powerpc64-unknown-linux-musl
? I'm only able to cross-test onpowerpc64-unknown-linux-gnu
andpowerpc64le-unknown-linux-gnu
locally at the moment, and as a tier 3 targetpowerpc64-unknown-linux-musl
has zero CI coverage.Fixes: #128579