[all-commits] [llvm/llvm-project] 3091bd: [AMDGPU] Do not release VGPRs at -O0

[Jay Foad via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 3091bdb86d55e404866823b64d21fd87c247d893
      https://github.com/llvm/llvm-project/commit/3091bdb86d55e404866823b64d21fd87c247d893
  Author: Jay Foad <jay.foad at amd.com>
  Date:   2023-08-10 (Thu, 10 Aug 2023)

  Changed paths:
    M llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
    A llvm/test/CodeGen/AMDGPU/release-vgprs.mir

  Log Message:
  -----------
  [AMDGPU] Do not release VGPRs at -O0

This was an oversight when the GFX11 early release VGPRs optimization
was reimplemented in D153279.

Sending the DEALLOC_VGPRS message is a performance optimization so there
is no need to do it at -O0. In addition it makes some kinds of post
mortem debugging hard or impossible, since VGPR values are no longer
available to inspect at the s_endpgm instruction.

Differential Revision: https://reviews.llvm.org/D157599