[PATCH] D136946: [AMDGPU] Enable `update/mov.dpp` selection with `+dpp,+gfx10-insts`

[Pierre van Houtryve via Phabricator via llvm-commits]

Pierre-vh created this revision.
Pierre-vh added reviewers: arsenm, foad, Joe_Nash.
Herald added subscribers: kosarev, kerbowa, hiraditya, t-tye, tpr, dstuttard, yaxunl, jvesely, kzhuravl.
Herald added a project: All.
Pierre-vh requested review of this revision.
Herald added subscribers: llvm-commits, wdng.
Herald added a project: LLVM.

ROCm device libs can emit those intrinsics w/ the +dpp attribute, and it counts on the optimizer to remove the call if the GPU is too old.
When built at O0 it caused codegen issues as Clang allowed this intrinsic to go through with just +dpp, but the backend wanted the GPU to be >=GFX8 as well.

This patch allows selecting that intrinsic when just minimum required attributes are present. That is, +gfx8-insts & +dpp.

Depends on D136945 <https://reviews.llvm.org/D136945>


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D136946

Files:
  llvm/lib/Target/AMDGPU/AMDGPU.td
  llvm/lib/Target/AMDGPU/GCNSubtarget.h
  llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
  llvm/lib/Target/AMDGPU/VOP1Instructions.td
  llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mov.dpp.ll
  llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.update.dpp.ll
  llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mov.dpp.ll
  llvm/test/CodeGen/AMDGPU/llvm.amdgcn.update.dpp.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D136946.471510.patch
Type: text/x-patch
Size: 12105 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20221028/aa2cd4a9/attachment.bin>