[PATCH] D31095: AMDGPU: Remove unnecessary ands when f16 is legal

[Matt Arsenault via Phabricator via llvm-commits]

arsenm created this revision.
Herald added subscribers: tpr, dstuttard, tony-tye, yaxunl, nhaehnle, wdng, kzhuravl.

Add a new node to act as a fancy bitcast from f16 operations to
i32 that implicitly zero the high 16-bits of the result.

      

Alternatively could try making v2f16 legal and canonicalizing
on build_vectors.


https://reviews.llvm.org/D31095

Files:
  lib/Target/AMDGPU/AMDGPUISelLowering.cpp
  lib/Target/AMDGPU/AMDGPUISelLowering.h
  lib/Target/AMDGPU/AMDGPUInstrInfo.td
  lib/Target/AMDGPU/SIISelLowering.cpp
  lib/Target/AMDGPU/SIISelLowering.h
  lib/Target/AMDGPU/SIInstructions.td
  test/CodeGen/AMDGPU/fabs.f16.ll
  test/CodeGen/AMDGPU/fadd.f16.ll
  test/CodeGen/AMDGPU/fcanonicalize.f16.ll
  test/CodeGen/AMDGPU/fmul.f16.ll
  test/CodeGen/AMDGPU/fptrunc.f16.ll
  test/CodeGen/AMDGPU/fsub.f16.ll
  test/CodeGen/AMDGPU/llvm.ceil.f16.ll
  test/CodeGen/AMDGPU/llvm.cos.f16.ll
  test/CodeGen/AMDGPU/llvm.exp2.f16.ll
  test/CodeGen/AMDGPU/llvm.floor.f16.ll
  test/CodeGen/AMDGPU/llvm.fma.f16.ll
  test/CodeGen/AMDGPU/llvm.fmuladd.f16.ll
  test/CodeGen/AMDGPU/llvm.log2.f16.ll
  test/CodeGen/AMDGPU/llvm.maxnum.f16.ll
  test/CodeGen/AMDGPU/llvm.minnum.f16.ll
  test/CodeGen/AMDGPU/llvm.rint.f16.ll
  test/CodeGen/AMDGPU/llvm.sin.f16.ll
  test/CodeGen/AMDGPU/llvm.sqrt.f16.ll
  test/CodeGen/AMDGPU/llvm.trunc.f16.ll
  test/CodeGen/AMDGPU/v_mac_f16.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D31095.92182.patch
Type: text/x-patch
Size: 40292 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170317/7648e001/attachment.bin>