[all-commits] [llvm/llvm-project] 8bed52: [AMDGPU] Make more use of madmk/fmamk instructions

[Jay Foad via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 8bed52c9eb8b871aaee3e082b71e60c5dbf2fc46
      https://github.com/llvm/llvm-project/commit/8bed52c9eb8b871aaee3e082b71e60c5dbf2fc46
  Author: Jay Foad <jay.foad at amd.com>
  Date:   2022-03-02 (Wed, 02 Mar 2022)

  Changed paths:
    M llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
    M llvm/test/CodeGen/AMDGPU/GlobalISel/urem.i64.ll
    M llvm/test/CodeGen/AMDGPU/twoaddr-fma.mir
    M llvm/test/CodeGen/AMDGPU/twoaddr-mad.mir

  Log Message:
  -----------
  [AMDGPU] Make more use of madmk/fmamk instructions

In convertToThreeAddress handle VOP2 mac/fmac instructions with a
literal src0 operand, since these are prime candidates for
converting to madmk/fmamk.

Previously this would only happen if src0 (or src1) was a register
defined by a move-immediate instruction, but in many cases these
operands have already been folded because SIFoldOperands runs
before TwoAddressInstructionPass.

Differential Revision: https://reviews.llvm.org/D120736