[all-commits] [llvm/llvm-project] b7b4f4: [RISCV] Make scalable vector FMA commutable for re...

[Craig Topper via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: b7b4f4cbc3a6d2a3ea49aa47ef600271dc4fb19f
      https://github.com/llvm/llvm-project/commit/b7b4f4cbc3a6d2a3ea49aa47ef600271dc4fb19f
  Author: Craig Topper <craig.topper at sifive.com>
  Date:   2021-02-08 (Mon, 08 Feb 2021)

  Changed paths:
    M llvm/lib/Target/RISCV/MCTargetDesc/RISCVBaseInfo.h
    M llvm/lib/Target/RISCV/RISCVISelLowering.cpp
    M llvm/lib/Target/RISCV/RISCVInstrFormats.td
    M llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
    M llvm/lib/Target/RISCV/RISCVInstrInfo.h
    M llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
    M llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
    M llvm/test/CodeGen/RISCV/rvv/vfmadd-sdnode.ll
    M llvm/test/CodeGen/RISCV/rvv/vfmsub-sdnode.ll
    M llvm/test/CodeGen/RISCV/rvv/vfnmadd-sdnode.ll
    M llvm/test/CodeGen/RISCV/rvv/vfnmsub-sdnode.ll

  Log Message:
  -----------
  [RISCV] Make scalable vector FMA commutable for register allocation.

This adds support for commuting operands and converting between
vfmadd and vfmacc to avoid register copies.

To avoid messing up intrinsic behavior, I've added new pseudo
instructions that have the isCommutable flag set. These pseudos also
force a tail agnostic policy. The intrinsic version still use
the tail undisturbed policy.

For best results it looks like we need to start with fmadd and only
pick fmacc if its beneficial. MachineCSE commutes without contraining
the operands and then commutes back if it didn't help with CSE. So
I've made sure that when the operand choice isn't constrained, we
will keep fmadd for MachineCSE and when it does the second commute,
we get back the original instruction.

Reviewed By: frasercrmck

Differential Revision: https://reviews.llvm.org/D95800