[llvm] [WIP] AMDGPU: VGPR bank conflict avoidance hints (PR #192140)

[Matt Arsenault via llvm-commits]

================
@@ -3865,6 +3891,7 @@ bool SIRegisterInfo::getRegAllocationHints(Register VirtReg,
 
   std::pair<unsigned, Register> Hint = MRI.getRegAllocationHint(VirtReg);
 
+  bool BaseImplRetVal = false;
----------------
arsenm wrote:

Avoid assign and break pattern, the implementation is big enough to move into a utility function 

https://github.com/llvm/llvm-project/pull/192140