[all-commits] [llvm/llvm-project] 1912ac: AMDGPU: Move handling of AGPR copies to a separate...

[Matt Arsenault via All-commits]

  Branch: refs/heads/master
  Home:   https://github.com/llvm/llvm-project
  Commit: 1912ace968766dfeae8f40425b1931ff5371b725
      https://github.com/llvm/llvm-project/commit/1912ace968766dfeae8f40425b1931ff5371b725
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2020-07-16 (Thu, 16 Jul 2020)

  Changed paths:
    M llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

  Log Message:
  -----------
  AMDGPU: Move handling of AGPR copies to a separate function

This is in preparation for fixing multiple problems with the way AGPR
copies are handled, but this change is NFC itself. First, it's relying
on recursively calling copyPhysReg, which is losing information
necessary to get correct super register handling.

Second, it's constructing a new RegScavenger and doing a O(N^2) walk
on every single sub-spill for every AGPR tuple copy. Third, it's using
the forward form of the scavenger, and not using the preferred
backwards scan.