[PATCH] D136944: [AMDGPU] Enable `s_sendmsg_rtn` selection with `+gfx11-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 a s_sendmsg_rtn w/ the +gfx11-insts attribute, and it counts on the optimizer to remove the call if the GPU is <GFX11.
When built at O0 it caused codegen issues as Clang allowed this intrinsic to go through with just +gfx11-insts, but the backend wanted the GPU to be >=GFX11 as well.

This patch allows selecting that intrinsic when just the +gfx11-insts attribute is present.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D136944

Files:
  llvm/lib/Target/AMDGPU/AMDGPU.td
  llvm/lib/Target/AMDGPU/GCNSubtarget.h
  llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
  llvm/lib/Target/AMDGPU/SOPInstructions.td
  llvm/lib/Target/AMDGPU/Utils/AMDGPUAsmUtils.cpp
  llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
  llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
  llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sendmsg.rtn.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D136944.471507.patch
Type: text/x-patch
Size: 17454 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20221028/07b1cc4c/attachment.bin>