[PATCH] D38128: Handle COPYs of physregs better (regalloc hints)

Jonas Paulsson via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Sat Feb 17 02:06:53 PST 2018

jonpa added a comment.

In https://reviews.llvm.org/D38128#1010271, @rampitec wrote:

> Changes look non-essential to me.

Thank you, Stanislav. I committed this even though a last-minute change slipped in (diff from approved changes):

  --- a/test/CodeGen/AMDGPU/ret.ll
  +++ b/test/CodeGen/AMDGPU/ret.ll
  @@ -178,8 +178,8 @@ bb:
   ; GCN-LABEL: {{^}}sgpr:
  -; GCN-DAG: s_add_i32 s0, s3, 2
  -; GCN-DAG: s_mov_b32 s2, s3
  +; GCN: s_mov_b32 s2, s3
  +; GCN: s_add_i32 s0, s2, 2
   ; GCN-NOT: s_endpgm
   define amdgpu_vs { i32, i32, i32 } @sgpr([9 x <16 x i8>] addrspace(4)* byval %arg, i32 inreg %arg1, i32 inreg %arg2, float %arg3) #0 {

This looks harmless to me - the add is using s2 instead of s3, directly after the move of s3 to s2, right? Please take a look.

Committed as r325425.


More information about the llvm-commits mailing list