[PATCH] D94645: [AMDGPU] Fix llvm.amdgcn.init.exec and frame materialization

[Carl Ritson via Phabricator via llvm-commits]

critson created this revision.
critson added reviewers: mareko, arsenm, ruiling.
Herald added subscribers: kerbowa, hiraditya, t-tye, tpr, dstuttard, yaxunl, nhaehnle, jvesely, kzhuravl.
critson requested review of this revision.
Herald added subscribers: llvm-commits, wdng.
Herald added a project: LLVM.

Frame-base materialization we may insert vector instructions
before EXEC is initialised.  Fix this by moving lowering of
llvm.amdgcn.init.exec post-RA.
Also remove SI_INIT_EXEC_LO pseudo as this is not necessary.

Based on code by Ruiling Song.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D94645

Files:
  llvm/lib/Target/AMDGPU/SIISelLowering.cpp
  llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
  llvm/lib/Target/AMDGPU/SIInstructions.td
  llvm/test/CodeGen/AMDGPU/llvm.amdgcn.init.exec.ll
  llvm/test/CodeGen/AMDGPU/llvm.amdgcn.init.exec.wave32.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D94645.316551.patch
Type: text/x-patch
Size: 13882 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210114/7b2b0555/attachment.bin>