[PATCH] D73815: AMDGPU: Stop scanning for control flow intrinsic users

[Matt Arsenault via Phabricator via llvm-commits]

arsenm created this revision.
arsenm added reviewers: alex-t, nhaehnle, rampitec.
Herald added subscribers: kerbowa, hiraditya, t-tye, tpr, dstuttard, yaxunl, wdng, jvesely, kzhuravl.
Herald added a project: LLVM.

The divergence analysis should properly report these return uniform
results. There should really be no need for requiresUniformRegister if
DivergenceAnalysis did its job properly. This requires fixing inline
asm handling, which is a big enough task to split into a separate
change.

      

This saves compile time and also seems to improve code.


https://reviews.llvm.org/D73815

Files:
  llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
  llvm/lib/Target/AMDGPU/SIISelLowering.cpp
  llvm/test/CodeGen/AMDGPU/divergent-branch-uniform-condition.ll
  llvm/test/CodeGen/AMDGPU/i1-copy-from-loop.ll
  llvm/test/CodeGen/AMDGPU/multilevel-break.ll
  llvm/test/CodeGen/AMDGPU/sgpr-control-flow.ll
  llvm/test/CodeGen/AMDGPU/smrd.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D73815.241834.patch
Type: text/x-patch
Size: 14061 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20200131/d9492f7c/attachment.bin>