[llvm] [AMDGPU] Enable reordering of VMEM loads during clustering (PR #107986)

[Matt Arsenault via llvm-commits]

================
@@ -1668,6 +1668,16 @@ class LLVM_ABI TargetInstrInfo : public MCInstrInfo {
     llvm_unreachable("target did not implement shouldClusterMemOps()");
   }
 
+  /// Returns true if the two given memory operations can be reordered
+  /// while clustering.
+  /// Will only be queried if ReorderWhileClustering is enabled and
+  /// shouldClusterMemOps already returned true for the same operation pair.
+  virtual bool
----------------
arsenm wrote:

This is essentially a different flavor of areMemAccessesTriviallyDisjoint. You can implement a generic helper somewhere else in terms of that

https://github.com/llvm/llvm-project/pull/107986