[llvm] [AMDGPU] Prefer lower total register usage in regions with spilling (PR #71882)

[Valery Pykhtin via llvm-commits]

================
@@ -88,8 +88,8 @@ void GCNRegPressure::inc(unsigned Reg,
   }
 }
 
-bool GCNRegPressure::less(const GCNSubtarget &ST,
-                          const GCNRegPressure& O,
+bool GCNRegPressure::less(const GCNSubtarget &ST, const MachineFunction &MF,
----------------
vpykhtin wrote:

can you leave only MF parameter and obtain ST from it, like

`const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();`

https://github.com/llvm/llvm-project/pull/71882