[PATCH] R600/SI: Allow commuting some 3 op instructions

Matt Arsenault Matthew.Arsenault at amd.com
Fri Oct 17 18:57:11 PDT 2014


e.g. v_mad_f32 a, b, c -> v_mad_f32 b, a, c
    
This simplifies matching v_madmk_f32.
    
This looks somewhat surprising, but it appears to be
OK to do this. We can commute src0 and src1 in all
of these instructions, and that's all that appears
to matter.

http://reviews.llvm.org/D5854

Files:
  lib/Target/R600/SIInstructions.td
  test/CodeGen/R600/fma.ll
  test/CodeGen/R600/fmuladd.ll
  test/CodeGen/R600/llvm.AMDGPU.umad24.ll
  test/CodeGen/R600/use-sgpr-multiple-times.ll
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D5854.15111.patch
Type: text/x-patch
Size: 11515 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20141018/6687b0ce/attachment.bin>


More information about the llvm-commits mailing list