[llvm] [WIP][AMDGPU] Improve the handling of `inreg` arguments (PR #133614)

[Shilei Tian via llvm-commits]

================
@@ -2841,6 +2841,86 @@ void SITargetLowering::insertCopiesSplitCSR(
   }
 }
 
+class InregVPGRSpiller {
+  CCState &State;
+  const unsigned WaveFrontSize;
+
+  Register CurReg;
+  unsigned CurLane = 0;
+
+protected:
+  SelectionDAG &DAG;
+  MachineFunction &MF;
+
+  Register getCurReg() const { return CurReg; }
+  unsigned getCurLane() const { return CurLane; }
----------------
shiltian wrote:

This should be `CurLane % WaveFrontSize`.

https://github.com/llvm/llvm-project/pull/133614