[PATCH] D115954: [AMDGPU] Re-enabling divergence predicates for min/max

[Jay Foad via Phabricator via llvm-commits]

foad added a comment.

Do you also need to use DivergentBinFrag on the definition of V_MIN_I32 (etc)?

  defm V_MIN_I32 : VOP2Inst <"v_min_i32", VOP_PAT_GEN<VOP_I32_I32_I32>, smin>;

If you don't do that, for a uniform smin, the instruction selector could choose either S_MIN or V_MIN. Does it know that it should prefer S_MIN? How?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D115954/new/

https://reviews.llvm.org/D115954