[llvm] AMDGPU: Make rewrite-vgpr-mfma-to-agpr-spill-multi-store.ll less allocator sensitive (PR #226913)

[via llvm-commits]

llvmorg-github-actions[bot] wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-amdgpu

Author: Matt Arsenault (arsenm)

<details>
<summary>Changes</summary>

This test is sensitive to the exact split and spills which occur, and disappeared
under a future upstream improvement. Use basic RA with a fixed occupancy since it more
stably produces the spill pattern.

Also add a codegen reference test for the same kernel, so future codegen improvements are
visible.

Co-Authored-By: Claude Opus 5 <noreply@<!-- -->anthropic.com>

---
Full diff: https://github.com/llvm/llvm-project/pull/226913.diff


2 Files Affected:

- (added) llvm/test/CodeGen/AMDGPU/rewrite-vgpr-mfma-to-agpr-spill-multi-store-codegen.ll (+694) 
- (modified) llvm/test/CodeGen/AMDGPU/rewrite-vgpr-mfma-to-agpr-spill-multi-store.ll (+2-2) 


``````````diff
The server is unavailable at this time. Please wait a few minutes before you try again.
``````````

</details>


https://github.com/llvm/llvm-project/pull/226913