[all-commits] [llvm/llvm-project] d159b4: [AMDGPU] Enable divergence predicates for negative...

[alex-t via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: d159b4444c880b8f42c4d507a49107535657be94
      https://github.com/llvm/llvm-project/commit/d159b4444c880b8f42c4d507a49107535657be94
  Author: alex-t <alexander.timofeev at amd.com>
  Date:   2022-03-10 (Thu, 10 Mar 2022)

  Changed paths:
    M llvm/lib/Target/AMDGPU/SIInstructions.td
    A llvm/test/CodeGen/AMDGPU/divergence-driven-negsubinlineconst.ll
    M llvm/test/CodeGen/AMDGPU/shift-i128.ll

  Log Message:
  -----------
  [AMDGPU] Enable divergence predicates for negative inline constant subtraction

We have a pattern that undo sub x, c -> add x, -c canonicalization since c is more likely
 an inline immediate than -c. This patch enables it to select scalar or vector subtracion by the input node divergence.

Reviewed By: foad

Differential Revision: https://reviews.llvm.org/D121360