[llvm] [AMDGPU] Compiler should synthesize private buffer resource descriptor from flat_scratch_init (PR #79586)

[Scott Linder via llvm-commits]

================
@@ -407,7 +408,7 @@ void SIFrameLowering::emitEntryFunctionFlatScratchInit(
 
     // Find unused reg to load flat scratch init into
     MachineRegisterInfo &MRI = MF.getRegInfo();
-    Register FlatScrInit = AMDGPU::NoRegister;
+    Register FlatScratchInitReg = AMDGPU::NoRegister;
----------------
slinder1 wrote:

I'm not sure I follow the different cases completely, but regardless of whether the current code is correct or not I don't think this inner variable should shadow the new function scope one.

Maybe call this one `PalFlatScratchInitReg` or something?

https://github.com/llvm/llvm-project/pull/79586