[llvm] [AMDGPU][Scheduler] Fix non-monotonic SlotIndex after schedule revert (PR #192039)

[via llvm-commits]

================
@@ -717,11 +720,14 @@ class PreRARematStage : public GCNSchedStage {
     std::vector<MachineInstr *> OrigMIOrder;
     /// Maximum pressure recorded in the region.
     GCNRegPressure MaxPressure;
+    /// Whether the region was already reverted by per-region checkScheduling.
+    bool AlreadyReverted = false;
----------------
michaelselehov wrote:

Done — agreed, with the recent changes this is no longer necessary. Verified: when `TargetOcc` is set, `shouldRevertScheduling` always returns false, and when it's not, `finalizeGCNSchedStage` returns early. No path to double-revert.

https://github.com/llvm/llvm-project/pull/192039