[llvm] [AMDGPU][Scheduler] Support for rematerializing SGPRs and AGPRs (PR #140036)

[Lucas Ramirez via llvm-commits]

================
@@ -1705,28 +1707,35 @@ struct ExcessRP {
   bool HasAGPRs = false;
   /// Whether the subtarget has a unified RF.
   bool UnifiedRF;
+  /// Whether we consider that ArchVGPRs can be spilled to AGPRs and the other
+  /// way around.
+  bool AllowVGPRToVGPRSpill;
----------------
lucas-rami wrote:

Thanks for the context, I wasn't clear on how these ArchVGPR/AGPR swaps actually happen. Changed the name to `CombineVGPRSavings` and updated comments where necessary.

https://github.com/llvm/llvm-project/pull/140036