[all-commits] [llvm/llvm-project] ce16b6: AMDGPU: Don't kill super-register with overlapping...

[Matt Arsenault via All-commits]

  Branch: refs/heads/master
  Home:   https://github.com/llvm/llvm-project
  Commit: ce16b6835bce18989e1dc0796305fe703e59ca4d
      https://github.com/llvm/llvm-project/commit/ce16b6835bce18989e1dc0796305fe703e59ca4d
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2020-10-16 (Fri, 16 Oct 2020)

  Changed paths:
    M llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
    M llvm/test/CodeGen/AMDGPU/accvgpr-copy.mir
    A llvm/test/CodeGen/AMDGPU/copy-overlap-vgpr-kill.mir

  Log Message:
  -----------
  AMDGPU: Don't kill super-register with overlapping copy

This would end up killing part of the result super-register, resulting
in a verifier error on a later use of the overlapping registers.  We
could add kills of any non-aliasing registers, but we should be moving
away from relying on kill flags.