[llvm] [MachineCSE][NVPTX] Add MachineCSE optimization to CSE inverted predicates (PR #185755)

[via llvm-commits]

================

----------------
modiking wrote:

I don't think that's needed for this case. We're strictly looking for CBranch/CBranchOther from the predicate instruction that feeds into them and replacing it with a semantically equivalent inverted instruction. The destination blocks are unchanged logically and in IR.

https://github.com/llvm/llvm-project/pull/185755