[llvm] [AMDGPU] More accurately account for AVGPR pressure (PR #150711)

[Jeffrey Byrnes via llvm-commits]

================
@@ -99,20 +99,22 @@ void GCNRegPressure::inc(unsigned Reg,
 bool GCNRegPressure::less(const MachineFunction &MF, const GCNRegPressure &O,
                           unsigned MaxOccupancy) const {
   const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
+  unsigned ArchVGPRThreshold = ST.getMaxNumVectorRegs(MF.getFunction()).first;
----------------
jrbyrnes wrote:

I think it is better to bury the details in GCNRegPressure to reduce burden on the caller like you say. I can't imagine a case where we would be working with GCNRegPressure and the MF isn't available..

https://github.com/llvm/llvm-project/pull/150711