AMDGPU/GlobalISel: Fix processing new phi in waterfall loop

[arsenm]

Summary

The adjusted iterator range included the last we just inserted, and
don't want to process. Figure out the new iterator range before
inserting phis. This was a harmless problem, but added an unnecessary
complication for a future patch.

[nhaehnle]

LGTM

[arsenm]

307e0d5490a5f2a069a00e4273784d1a2e005ab9

llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp

const unsigned MovTermOpc = Subtarget.isWave32() ?
     AMDGPU::S_MOV_B32_term : AMDGPU::S_MOV_B64_term;
   const unsigned XorTermOpc = Subtarget.isWave32() ?
     AMDGPU::S_XOR_B32_term : AMDGPU::S_XOR_B64_term;
   const unsigned AndSaveExecOpc =  Subtarget.isWave32() ?
     AMDGPU::S_AND_SAVEEXEC_B32 : AMDGPU::S_AND_SAVEEXEC_B64;
   const unsigned ExecReg =  Subtarget.isWave32() ?
     AMDGPU::EXEC_LO : AMDGPU::EXEC;
 
+#ifndef NDEBUG
+  const int OrigRangeSize = std::distance(Range.begin(), Range.end());
+#endif
+
   for (MachineInstr &MI : Range) {
     for (MachineOperand &Def : MI.defs()) {
       LLT ResTy = MRI.getType(Def.getReg());
       const RegisterBank *DefBank = getRegBank(Def.getReg(), MRI, *TRI);
       ResultRegs.push_back(Def.getReg());
       Register InitReg = B.buildUndef(ResTy).getReg(0);
       Register PhiReg = MRI.createGenericVirtualRegister(ResTy);
       InitResultRegs.push_back(InitReg);
B.buildInstr(TargetOpcode::G_PHI)
       .addReg(std::get<0>(Result)) // Initial value / implicit_def
       .addMBB(&MBB)
       .addReg(std::get<1>(Result)) // Mid-loop value.
       .addMBB(LoopBB);
   }
 
   const DebugLoc &DL = B.getDL();
 
-  // Figure out the iterator range after splicing the instructions.
-  auto NewBegin = std::prev(LoopBB->end());
+  MachineInstr &FirstInst = *Range.begin();
 
   // Move the instruction into the loop. Note we moved everything after
   // Range.end() already into a new block, so Range.end() is no longer valid.
   LoopBB->splice(LoopBB->end(), &MBB, Range.begin(), MBB.end());
 
+  // Figure out the iterator range after splicing the instructions.
+  MachineBasicBlock::iterator NewBegin = FirstInst.getIterator();
   auto NewEnd = LoopBB->end();
 
   MachineBasicBlock::iterator I = Range.begin();
   B.setInsertPt(*LoopBB, I);
 
   Register CondReg;
 
+  assert(std::distance(NewBegin, NewEnd) == OrigRangeSize);
+
   for (MachineInstr &MI : make_range(NewBegin, NewEnd)) {
     for (MachineOperand &Op : MI.uses()) {
       if (!Op.isReg() || Op.isDef())
         continue;
 
       if (!SGPROperandRegs.count(Op.getReg()))
         continue;