[llvm] [AMDGPU] Fix GCNRewritePartialRegUses pass: vector regclass is selected instead of scalar. (PR #69957)

[Carl Ritson via llvm-commits]

================
@@ -425,29 +414,41 @@ bool GCNRewritePartialRegUses::rewriteReg(Register Reg) const {
     return false;
 
   for (MachineOperand &MO : Range) {
-    if (MO.getSubReg() == AMDGPU::NoSubRegister) // Whole reg used, quit.
+    if (MO.getSubReg() == AMDGPU::NoSubRegister) // Whole reg used, quit. [1]
       return false;
   }
 
   auto *RC = MRI->getRegClass(Reg);
   LLVM_DEBUG(dbgs() << "Try to rewrite partial reg " << printReg(Reg, TRI)
                     << ':' << TRI->getRegClassName(RC) << '\n');
 
-  // Collect used subregs and constrained reg classes infered from instruction
+  // Collect used subregs and their reg classes infered from instruction
   // operands.
   SubRegMap SubRegs;
   for (MachineOperand &MO : MRI->reg_nodbg_operands(Reg)) {
----------------
perlfu wrote:

Not sure why this is not `Range`?

https://github.com/llvm/llvm-project/pull/69957