[all-commits] [llvm/llvm-project] 763871: [DAGCombiner] Require nsz for aggressive fma fold

[Qiu Chaofan via All-commits]

  Branch: refs/heads/master
  Home:   https://github.com/llvm/llvm-project
  Commit: 763871053cc0c209e14f07f7f74bb8f1c16872d9
      https://github.com/llvm/llvm-project/commit/763871053cc0c209e14f07f7f74bb8f1c16872d9
  Author: Qiu Chaofan <qiucofan at cn.ibm.com>
  Date:   2020-03-22 (Sun, 22 Mar 2020)

  Changed paths:
    M llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
    M llvm/test/CodeGen/AMDGPU/fma-combine.ll
    M llvm/test/CodeGen/AMDGPU/mad-combine.ll
    M llvm/test/CodeGen/PowerPC/fma-assoc.ll

  Log Message:
  -----------
  [DAGCombiner] Require nsz for aggressive fma fold

For folding pattern `x-(fma y,z,u*v) -> (fma -y,z,(fma -u,v,x))`, if
`yz` is 1, `uv` is -1 and `x` is -0, sign of result would be changed.

Differential Revision: https://reviews.llvm.org/D76419