[PATCH] D75504: [AMDGPU] moving vcc branch optimization into peephole

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added a comment.

If this can't work in SSA, then it shouldn't be done in PeepholeOptimizer.

I'm also noticing a few defects in the existing handling. If I disable the optimization in test/CodeGen/AMDGPU/multilevel-break.ll, the dead and is actually left behind. I'm assuming this is because if the check for dead SCC, but this should be using LivePhysRegs to make sure SCC is not live out


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D75504/new/

https://reviews.llvm.org/D75504