[PATCH] D134272: RFD: AMDGPU: Stop validating earlyclobber operands in assembler

Wed Sep 21 05:29:03 PDT 2022

foad added a comment.

In D134272#3803400 <https://reviews.llvm.org/D134272#3803400>, @rampitec wrote:

> In D134272#3803307 <https://reviews.llvm.org/D134272#3803307>, @arsenm wrote:
>
>> In D134272#3803303 <https://reviews.llvm.org/D134272#3803303>, @rampitec wrote:
>>
>>> Why not keep this validation and limit it to specific opcodes? For example multiplass DGEMM has the same restriction for the same reason.
>>
>> Not only opcodes, but the specifically invalid overlaps
>
> Right, it affects specific operands only, we just have no way to express it in the td.

Adding checks for specific operands of specific (DGEMM?) opcodes sounds great, if the restrictions are clearly documented, but I don't think it belongs in this patch.

For mad/qsad/mqsad I don't think there are any clearly documented restrictions, just some reports of hardware bugs that we try to work around in codegen, so I don't think the assembler should be enforcing any restrictions.

So I'd still like to proceed with this patch.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D134272/new/

https://reviews.llvm.org/D134272