[PATCH] D106758: [AMDGPU] We would need FP if there is call and caller save VGPR spills

[Venkata Ramanaiah Nalamothu via Phabricator via llvm-commits]

RamNalamothu added a comment.

In D106758#2907507 <https://reviews.llvm.org/D106758#2907507>, @arsenm wrote:

> In D106758#2907086 <https://reviews.llvm.org/D106758#2907086>, @RamNalamothu wrote:
>
>> In D106758#2904887 <https://reviews.llvm.org/D106758#2904887>, @arsenm wrote:
>>
>>> LGTM except the test is using dead code. I also think we ought to be able to remove VGPRReservedForSGPRSpill now that SGPR and VGPR allocation are split
>>
>> I think that's out of the scope for this review.
>
> Yes, but would be a good follow up change.
>
> This test looks nearly identical to test/CodeGen/AMDGPU/need-fp-from-csr-vgpr-spill.ll. Can the two be merged? I don't actually see why the existing test didn't catch this and isn't changed.

Because `test/CodeGen/AMDGPU/need-fp-from-csr-vgpr-spill.ll` covers the path where we have only CSR VGPR spills and `llvm/test/CodeGen/AMDGPU/need-fp-from-caller-save-vgpr-spill.ll` covers the other scenario where we might reserve a caller save VGPR for SGPR spill.

Will merge both tests/files into `test/CodeGen/AMDGPU/need-fp-from-vgpr-spills.ll`.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D106758/new/

https://reviews.llvm.org/D106758