[llvm-branch-commits] [llvm] [AMDGPU][Scheduler] Simplify scheduling revert logic (PR #177203)

[Lucas Ramirez via llvm-branch-commits]

lucas-rami wrote:

Further work based on this shows some very significant regressions in some cases due to slot index differences. While I can also see improvements in some cases, the change may not be desirable. #177205 and #177206 depend on the change however, so I am looking for a different approach that does not rely on this change.

https://github.com/llvm/llvm-project/pull/177203