[llvm] [AMDGPU] Teach iterative schedulers about IGLP (PR #134953)

[Jeffrey Byrnes via llvm-commits]

================
@@ -118,21 +119,41 @@ void GCNIterativeScheduler::printSchedRP(raw_ostream &OS,
 }
 #endif
 
+void GCNIterativeScheduler::swapIGLPMutations(const Region &R, bool IsReentry) {
----------------
jrbyrnes wrote:

Yeah, that was my thought. 

For the multi-stage scheduling approach it makes sense to cache them for facilitating different strategies based on mutation selection. But for these schedulers, it seems we should either apply them or not.

https://github.com/llvm/llvm-project/pull/134953