Allow constants using an Abi::Vector layout to be passed to the backend

[GeorgeWort]

This allows constant vectors using a repr(simd) type to be propagated through to the backend. This fixes a few issues with LLVM simd intrinsics where a constant vector is expected but cannot be produced by rust at opt-level=0.

[rustbot]

Thanks for the pull request, and welcome! The Rust team is excited to review your changes, and you should hear from @b-naber (or someone else) soon.

Please see the contribution instructions for more information. Namely, in order to ensure the minimum review times lag, PR authors and assigned reviewers should ensure that the review label (S-waiting-on-review and S-waiting-on-author) stays updated, invoking these commands when appropriate:

• @rustbot author: the review is finished, PR author should check the comments and take action accordingly
• @rustbot review: the author is ready for a review, this PR will be queued again in the reviewer's queue

[rustbot]

Some changes occurred to the CTFE / Miri engine

cc @rust-lang/miri

Some changes occurred in compiler/rustc_codegen_gcc

cc @antoyo

[rust-log-analyzer]

The job mingw-check failed! Check out the build log: (web) (plain)

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

GITHUB_ACTION=__run_7
GITHUB_ACTIONS=true
GITHUB_ACTION_REF=
GITHUB_ACTION_REPOSITORY=
GITHUB_ACTOR=GeorgeWort
GITHUB_API_URL=https://api.github.com
GITHUB_BASE_REF=master
GITHUB_ENV=/home/runner/work/_temp/_runner_file_commands/set_env_8f9e498c-8b21-4b41-bd1c-b87b454913f6
GITHUB_EVENT_NAME=pull_request
---
GITHUB_SERVER_URL=https://github.com
GITHUB_SHA=3ba119b511935de08f96807f9817939ff272a412
GITHUB_STATE=/home/runner/work/_temp/_runner_file_commands/save_state_8f9e498c-8b21-4b41-bd1c-b87b454913f6
GITHUB_STEP_SUMMARY=/home/runner/work/_temp/_runner_file_commands/step_summary_8f9e498c-8b21-4b41-bd1c-b87b454913f6
GITHUB_TRIGGERING_ACTOR=GeorgeWort
GITHUB_WORKFLOW_REF=rust-lang/rust/.github/workflows/ci.yml@refs/pull/118171/merge
GITHUB_WORKFLOW_SHA=3ba119b511935de08f96807f9817939ff272a412
GITHUB_WORKSPACE=/home/runner/work/rust/rust
GOROOT_1_19_X64=/opt/hostedtoolcache/go/1.19.13/x64
---
    Checking gccjit v1.0.0 (https://github.com/antoyo/gccjit.rs#6e290f25)
    Checking object v0.30.4
    Checking tempfile v3.7.1
    Checking rustc_codegen_gcc v0.1.0 (/checkout/compiler/rustc_codegen_gcc)
error[E0599]: no method named `type_ix` found for mutable reference `&mut builder::Builder<'a, 'gcc, 'tcx>` in the current scope
    |
    |
323 |                     self.sext(cmp, self.type_ix(32))
    |
    = help: items from traits can only be used if the trait is in scope
help: the following trait is implemented but not in scope; perhaps add a `use` for it:
    |
    |
1   + use crate::rustc_codegen_ssa::traits::BaseTypeMethods;
help: there is a method with a similar name
    |
    |
323 |                     self.sext(cmp, self.type_i8p_ext(32))

For more information about this error, try `rustc --explain E0599`.
error: could not compile `rustc_codegen_gcc` (lib) due to previous error
Build completed unsuccessfully in 0:02:26

[RalfJung]

Thanks for the PR! However, conceptually what you are suggesting here is a huge change and has to be done extremely carefully. I don't think we have sufficient motivation for such a fundamental refactor of our immediate handling. This needs a compiler MCP for sure.

The discussion here is also relevant: you are looking for a way to give vectors SSA variable treatment similar to Scalar and ScalarPair.

If we do want to go with this, we definitely want a new variant in the interpreter's Immediate, these should align with the variants of the Abi type. I'm not sure I even understand what the current representation is, but Immediate::Scalar is for scalar values only and a vector is not a scalar value.

[RalfJung]

You say this fixes some issue. Where's that issue tracked, where can I find the details? I sure hope there's an easier way to resolve that. Generally whenever LLVM requires something to be a constant we want to represent it as a const generic in Rust.

If you haven't done so already, please file an issue explaining the problem and your investigation of it.

[GeorgeWort]

Thanks for the feedback! That makes sense. There may be alternative options to fix this issue, I would be happy to go with any better alternative that you can think of.

I have created a bug report here.

[bors]

☔ The latest upstream changes (presumably #118408) made this pull request unmergeable. Please resolve the merge conflicts.

[workingjubilee]

#120690

r? compiler

[Dylan-DPC]

@GeorgeWort any updates on this? thanks

[GeorgeWort]

I should have closed this PR sorry, all continuing related work is linked to from the issue #118209.