[PATCH] D84639: AMDGPU: Add type mangling to llvm.amdgcn.readfirstlane

[Matt Arsenault via Phabricator via llvm-commits]

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

I've thought this should at minimum work for all the legal 32-bit
types for a long time. Arguably we should also legalize other types,
or at least multiples of 32.

      

This fixes selection for non-s32 types when readfirstlane is inserted
for SGPR return values.

      

This will require API users to add the type mangling when getting the
declaration. I think the way mesa constructs intrinsics will happen to
get away without needing to change anything.


https://reviews.llvm.org/D84639

Files:
  clang/lib/CodeGen/CGBuiltin.cpp
  clang/test/CodeGenOpenCL/builtins-amdgcn.cl
  llvm/include/llvm/IR/IntrinsicsAMDGPU.td
  llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
  llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp
  llvm/lib/Target/AMDGPU/SIISelLowering.cpp
  llvm/lib/Target/AMDGPU/SIInstructions.td
  llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-amdgcn.readfirstlane.mir
  llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
  llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D84639.280860.patch
Type: text/x-patch
Size: 22293 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20200727/c8f1f3f9/attachment.bin>