[AMDGPU] Resolve pseudo registers at encoding uses

[Joe_Nash]

Summary

Pseudo-registers allow different register encodings
between gpu generations. Make sure we resolve the
pseudo regs to real regs whenever we get their
hardware encoding.
Using the correct encodings revealed a register
bank conflict and an unnecessary write dependency.
Tests have been updated to match.

[rampitec]

LGTM

[Joe_Nash]

whitespace change for lint and rerun build

llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp

 
 bool AMDGPUAsmParser::usesConstantBus(const MCInst &Inst, unsigned OpIdx) {
   const MCOperand &MO = Inst.getOperand(OpIdx);
   if (MO.isImm()) {
     return !isInlineConstant(Inst, OpIdx);
   } else if (MO.isReg()) {
     auto Reg = MO.getReg();
     const MCRegisterInfo *TRI = getContext().getRegisterInfo();
-    return isSGPR(mc2PseudoReg(Reg), TRI) && Reg != SGPR_NULL;
+    auto PReg = mc2PseudoReg(Reg);
+    return isSGPR(PReg, TRI) && PReg != SGPR_NULL;
   } else {
     return true;
   }
 }
 
 bool AMDGPUAsmParser::validateConstantBusLimitations(const MCInst &Inst) {
   const unsigned Opcode = Inst.getOpcode();
   const MCInstrDesc &Desc = MII.get(Opcode);

llvm/lib/Target/AMDGPU/GCNRegBankReassign.cpp

 ///
 //===----------------------------------------------------------------------===//
 
 #include "AMDGPU.h"
 #include "AMDGPUSubtarget.h"
 #include "MCTargetDesc/AMDGPUMCTargetDesc.h"
 #include "SIInstrInfo.h"
 #include "SIMachineFunctionInfo.h"
+#include "Utils/AMDGPUBaseInfo.h"
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/CodeGen/LiveInterval.h"
 #include "llvm/CodeGen/LiveIntervals.h"
 #include "llvm/CodeGen/LiveRegMatrix.h"
 #include "llvm/CodeGen/MachineFunctionPass.h"
 #include "llvm/CodeGen/MachineLoopInfo.h"
 #include "llvm/CodeGen/VirtRegMap.h"
else if (Size > 32) {
     }
   }
 
   if (TRI->hasVGPRs(RC)) {
     unsigned RegNo = Reg - AMDGPU::VGPR0;
     return RegNo % NUM_VGPR_BANKS;
   }
 
-  unsigned RegNo = TRI->getEncodingValue(Reg) / 2;
+  unsigned RegNo = TRI->getEncodingValue(AMDGPU::getMCReg(Reg, *ST)) / 2;
   return RegNo % NUM_SGPR_BANKS + SGPR_BANK_OFFSET;
 }
 
 uint32_t GCNRegBankReassign::getRegBankMask(Register Reg, unsigned SubReg,
                                             int Bank) {
   if (Reg.isVirtual()) {
     if (!VRM->isAssignedReg(Reg))
       return 0;
for (unsigned I = 0; I < Size; ++I)
         Used |= 1 << I;
     RegsUsed.set(RegNo, RegNo + Size);
     Mask &= ~Used;
     Mask <<= (Bank == -1) ? RegNo % NUM_VGPR_BANKS : uint32_t(Bank);
     return (Mask | (Mask >> NUM_VGPR_BANKS)) & VGPR_BANK_MASK;
   }
 
   // SGPRs have 8 banks holding 2 consequitive registers each.
-  unsigned RegNo = TRI->getEncodingValue(Reg) / 2;
+  unsigned RegNo = TRI->getEncodingValue(AMDGPU::getMCReg(Reg, *ST)) / 2;
   unsigned StartBit = AMDGPU::VGPR_32RegClass.getNumRegs();
   if (RegNo + StartBit >= RegsUsed.size())
     return 0;
 
   if (Size > 1)
     Size /= 2;
   unsigned Mask = (1 << Size) - 1;
   unsigned Used = 0;
bool GCNRegBankReassign::runOnMachineFunction(MachineFunction &MF) {
   const SIMachineFunctionInfo *MFI = MF.getInfo<SIMachineFunctionInfo>();
   unsigned Occupancy = MFI->getOccupancy();
   MaxNumVGPRs = ST->getMaxNumVGPRs(MF);
   MaxNumSGPRs = ST->getMaxNumSGPRs(MF);
   MaxNumVGPRs = std::min(ST->getMaxNumVGPRs(Occupancy), MaxNumVGPRs);
   MaxNumSGPRs = std::min(ST->getMaxNumSGPRs(Occupancy, true), MaxNumSGPRs);
 
   CSRegs = MRI->getCalleeSavedRegs();
-
-  RegsUsed.resize(AMDGPU::VGPR_32RegClass.getNumRegs() +
-                  TRI->getEncodingValue(AMDGPU::SGPR_NULL) / 2 + 1);
+  unsigned NumRegBanks = AMDGPU::VGPR_32RegClass.getNumRegs() +
+                         // Not a tight bound
+                         AMDGPU::SReg_32RegClass.getNumRegs() / 2 + 1;
+  RegsUsed.resize(NumRegBanks);
 
   LLVM_DEBUG(dbgs() << "=== RegBanks reassign analysis on function " << MF.getName()
                << '\n');
 
   unsigned StallCycles = collectCandidates(MF);
   NumStallsDetected += StallCycles;
 
   LLVM_DEBUG(dbgs() << "=== " << StallCycles << " stall cycles detected in "

llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp

if (!TRI->isInAllocatableClass(Op.getReg()) || TRI->isAGPR(*MRI, Op.getReg()))
     return {-1, -1};
 
   // A use via a PW operand does not need a waitcnt.
   // A partial write is not a WAW.
   assert(!Op.getSubReg() || !Op.isUndef());
 
   RegInterval Result;
 
-  unsigned Reg = TRI->getEncodingValue(Op.getReg());
+  unsigned Reg = TRI->getEncodingValue(AMDGPU::getMCReg(Op.getReg(), *ST));
 
   if (TRI->isVGPR(*MRI, Op.getReg())) {
     assert(Reg >= RegisterEncoding.VGPR0 && Reg <= RegisterEncoding.VGPRL);
     Result.first = Reg - RegisterEncoding.VGPR0;
     assert(Result.first >= 0 && Result.first < SQ_MAX_PGM_VGPRS);
   } else if (TRI->isSGPRReg(*MRI, Op.getReg())) {
     assert(Reg >= RegisterEncoding.SGPR0 && Reg < SQ_MAX_PGM_SGPRS);
     Result.first = Reg - RegisterEncoding.SGPR0 + NUM_ALL_VGPRS;
if (TII->isDS(Inst) && (Inst.mayStore() || Inst.mayLoad())) {
         // For export the destination registers are really temps that
         // can be used as the actual source after export patching, so
         // we need to treat them like sources and set the EXP_CNT
         // score.
         for (unsigned I = 0, E = Inst.getNumOperands(); I != E; ++I) {
           MachineOperand &DefMO = Inst.getOperand(I);
           if (DefMO.isReg() && DefMO.isDef() &&
               TRI->isVGPR(*MRI, DefMO.getReg())) {
-            setRegScore(TRI->getEncodingValue(DefMO.getReg()), EXP_CNT,
-                        CurrScore);
+            setRegScore(
+                TRI->getEncodingValue(AMDGPU::getMCReg(DefMO.getReg(), *ST)),
+                EXP_CNT, CurrScore);
           }
         }
       }
       for (unsigned I = 0, E = Inst.getNumOperands(); I != E; ++I) {
         MachineOperand &MO = Inst.getOperand(I);
         if (MO.isReg() && !MO.isDef() && TRI->isVGPR(*MRI, MO.getReg())) {
           setExpScore(&Inst, TII, TRI, MRI, I, CurrScore);
         }

llvm/lib/Target/AMDGPU/SIRegisterInfo.h

void eliminateFrameIndex(MachineBasicBlock::iterator MI, int SPAdj,
                            unsigned FIOperandNum,
                            RegScavenger *RS) const override;
 
   bool eliminateSGPRToVGPRSpillFrameIndex(MachineBasicBlock::iterator MI,
                                           int FI, RegScavenger *RS) const;
 
   StringRef getRegAsmName(MCRegister Reg) const override;
 
+  // Pseudo regs are not allowed
   unsigned getHWRegIndex(MCRegister Reg) const {
     return getEncodingValue(Reg) & 0xff;
   }
 
   static const TargetRegisterClass *getVGPRClassForBitWidth(unsigned BitWidth);
   static const TargetRegisterClass *getAGPRClassForBitWidth(unsigned BitWidth);
   static const TargetRegisterClass *getSGPRClassForBitWidth(unsigned BitWidth);
 

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.gather4.a16.dim.ll

 ; GFX9-NEXT:    v_and_or_b32 v0, v0, v2, v1
 ; GFX9-NEXT:    s_and_b64 exec, exec, s[14:15]
 ; GFX9-NEXT:    image_gather4 v[0:3], v0, s[0:7], s[8:11] dmask:0x1 a16
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10NSA-LABEL: gather4_2d:
 ; GFX10NSA:       ; %bb.0: ; %main_body
-; GFX10NSA-NEXT:    s_mov_b32 s14, exec_lo
+; GFX10NSA-NEXT:    s_mov_b32 s28, exec_lo
 ; GFX10NSA-NEXT:    s_mov_b32 s0, s2
 ; GFX10NSA-NEXT:    s_wqm_b32 exec_lo, exec_lo
 ; GFX10NSA-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX10NSA-NEXT:    s_mov_b32 s1, s3
 ; GFX10NSA-NEXT:    s_mov_b32 s2, s4
 ; GFX10NSA-NEXT:    s_mov_b32 s3, s5
 ; GFX10NSA-NEXT:    s_mov_b32 s4, s6
 ; GFX10NSA-NEXT:    s_mov_b32 s5, s7
 ; GFX10NSA-NEXT:    s_mov_b32 s6, s8
 ; GFX10NSA-NEXT:    s_mov_b32 s7, s9
 ; GFX10NSA-NEXT:    s_mov_b32 s8, s10
 ; GFX10NSA-NEXT:    s_mov_b32 s9, s11
 ; GFX10NSA-NEXT:    s_mov_b32 s10, s12
 ; GFX10NSA-NEXT:    s_mov_b32 s11, s13
 ; GFX10NSA-NEXT:    v_and_or_b32 v0, v0, 0xffff, v1
-; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s14
+; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s28
 ; GFX10NSA-NEXT:    image_gather4 v[0:3], v0, s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
 ; GFX10NSA-NEXT:    ; implicit-def: $vcc_hi
 ; GFX10NSA-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10NSA-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.2d.v4f32.f16(i32 1, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }

 ; GFX9-NEXT:    v_and_or_b32 v1, v2, v3, s12
 ; GFX9-NEXT:    s_and_b64 exec, exec, s[14:15]
 ; GFX9-NEXT:    image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16 da
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10NSA-LABEL: gather4_cube:
 ; GFX10NSA:       ; %bb.0: ; %main_body
-; GFX10NSA-NEXT:    s_mov_b32 s14, exec_lo
+; GFX10NSA-NEXT:    s_mov_b32 s28, exec_lo
 ; GFX10NSA-NEXT:    s_mov_b32 s0, s2
 ; GFX10NSA-NEXT:    s_wqm_b32 exec_lo, exec_lo
 ; GFX10NSA-NEXT:    v_mov_b32_e32 v3, 0xffff
 ; GFX10NSA-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX10NSA-NEXT:    s_mov_b32 s2, s4
 ; GFX10NSA-NEXT:    s_mov_b32 s4, s6
 ; GFX10NSA-NEXT:    s_mov_b32 s6, s8
 ; GFX10NSA-NEXT:    s_mov_b32 s8, s10
 ; GFX10NSA-NEXT:    s_mov_b32 s10, s12
 ; GFX10NSA-NEXT:    s_lshl_b32 s12, s0, 16
 ; GFX10NSA-NEXT:    s_mov_b32 s1, s3
 ; GFX10NSA-NEXT:    s_mov_b32 s3, s5
 ; GFX10NSA-NEXT:    s_mov_b32 s5, s7
 ; GFX10NSA-NEXT:    s_mov_b32 s7, s9
 ; GFX10NSA-NEXT:    s_mov_b32 s9, s11
 ; GFX10NSA-NEXT:    v_and_or_b32 v0, v0, v3, v1
 ; GFX10NSA-NEXT:    s_mov_b32 s11, s13
 ; GFX10NSA-NEXT:    v_and_or_b32 v1, v2, v3, s12
-; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s14
+; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s28
 ; GFX10NSA-NEXT:    image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_CUBE a16
 ; GFX10NSA-NEXT:    ; implicit-def: $vcc_hi
 ; GFX10NSA-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10NSA-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.cube.v4f32.f16(i32 1, half %s, half %t, half %face, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }

 ; GFX9-NEXT:    v_and_or_b32 v1, v2, v3, s12
 ; GFX9-NEXT:    s_and_b64 exec, exec, s[14:15]
 ; GFX9-NEXT:    image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16 da
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10NSA-LABEL: gather4_2darray:
 ; GFX10NSA:       ; %bb.0: ; %main_body
-; GFX10NSA-NEXT:    s_mov_b32 s14, exec_lo
+; GFX10NSA-NEXT:    s_mov_b32 s28, exec_lo
 ; GFX10NSA-NEXT:    s_mov_b32 s0, s2
 ; GFX10NSA-NEXT:    s_wqm_b32 exec_lo, exec_lo
 ; GFX10NSA-NEXT:    v_mov_b32_e32 v3, 0xffff
 ; GFX10NSA-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX10NSA-NEXT:    s_mov_b32 s2, s4
 ; GFX10NSA-NEXT:    s_mov_b32 s4, s6
 ; GFX10NSA-NEXT:    s_mov_b32 s6, s8
 ; GFX10NSA-NEXT:    s_mov_b32 s8, s10
 ; GFX10NSA-NEXT:    s_mov_b32 s10, s12
 ; GFX10NSA-NEXT:    s_lshl_b32 s12, s0, 16
 ; GFX10NSA-NEXT:    s_mov_b32 s1, s3
 ; GFX10NSA-NEXT:    s_mov_b32 s3, s5
 ; GFX10NSA-NEXT:    s_mov_b32 s5, s7
 ; GFX10NSA-NEXT:    s_mov_b32 s7, s9
 ; GFX10NSA-NEXT:    s_mov_b32 s9, s11
 ; GFX10NSA-NEXT:    v_and_or_b32 v0, v0, v3, v1
 ; GFX10NSA-NEXT:    s_mov_b32 s11, s13
 ; GFX10NSA-NEXT:    v_and_or_b32 v1, v2, v3, s12
-; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s14
+; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s28
 ; GFX10NSA-NEXT:    image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D_ARRAY a16
 ; GFX10NSA-NEXT:    ; implicit-def: $vcc_hi
 ; GFX10NSA-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10NSA-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.2darray.v4f32.f16(i32 1, half %s, half %t, half %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }

 ; GFX9-NEXT:    v_and_or_b32 v1, v1, v3, v2
 ; GFX9-NEXT:    s_and_b64 exec, exec, s[14:15]
 ; GFX9-NEXT:    image_gather4_c v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10NSA-LABEL: gather4_c_2d:
 ; GFX10NSA:       ; %bb.0: ; %main_body
-; GFX10NSA-NEXT:    s_mov_b32 s14, exec_lo
+; GFX10NSA-NEXT:    s_mov_b32 s28, exec_lo
 ; GFX10NSA-NEXT:    s_mov_b32 s0, s2
 ; GFX10NSA-NEXT:    s_wqm_b32 exec_lo, exec_lo
 ; GFX10NSA-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
 ; GFX10NSA-NEXT:    s_mov_b32 s1, s3
 ; GFX10NSA-NEXT:    s_mov_b32 s2, s4
 ; GFX10NSA-NEXT:    s_mov_b32 s3, s5
 ; GFX10NSA-NEXT:    s_mov_b32 s4, s6
 ; GFX10NSA-NEXT:    s_mov_b32 s5, s7
 ; GFX10NSA-NEXT:    s_mov_b32 s6, s8
 ; GFX10NSA-NEXT:    s_mov_b32 s7, s9
 ; GFX10NSA-NEXT:    s_mov_b32 s8, s10
 ; GFX10NSA-NEXT:    s_mov_b32 s9, s11
 ; GFX10NSA-NEXT:    s_mov_b32 s10, s12
 ; GFX10NSA-NEXT:    s_mov_b32 s11, s13
 ; GFX10NSA-NEXT:    v_and_or_b32 v1, v1, 0xffff, v2
-; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s14
+; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s28
 ; GFX10NSA-NEXT:    image_gather4_c v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
 ; GFX10NSA-NEXT:    ; implicit-def: $vcc_hi
 ; GFX10NSA-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10NSA-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.2d.v4f32.f16(i32 1, float %zcompare, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }

 ; GFX9-NEXT:    v_and_or_b32 v1, v2, v3, s12
 ; GFX9-NEXT:    s_and_b64 exec, exec, s[14:15]
 ; GFX9-NEXT:    image_gather4_cl v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10NSA-LABEL: gather4_cl_2d:
 ; GFX10NSA:       ; %bb.0: ; %main_body
-; GFX10NSA-NEXT:    s_mov_b32 s14, exec_lo
+; GFX10NSA-NEXT:    s_mov_b32 s28, exec_lo
 ; GFX10NSA-NEXT:    s_mov_b32 s0, s2
 ; GFX10NSA-NEXT:    s_wqm_b32 exec_lo, exec_lo
 ; GFX10NSA-NEXT:    v_mov_b32_e32 v3, 0xffff
 ; GFX10NSA-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX10NSA-NEXT:    s_mov_b32 s2, s4
 ; GFX10NSA-NEXT:    s_mov_b32 s4, s6
 ; GFX10NSA-NEXT:    s_mov_b32 s6, s8
 ; GFX10NSA-NEXT:    s_mov_b32 s8, s10
 ; GFX10NSA-NEXT:    s_mov_b32 s10, s12
 ; GFX10NSA-NEXT:    s_lshl_b32 s12, s0, 16
 ; GFX10NSA-NEXT:    s_mov_b32 s1, s3
 ; GFX10NSA-NEXT:    s_mov_b32 s3, s5
 ; GFX10NSA-NEXT:    s_mov_b32 s5, s7
 ; GFX10NSA-NEXT:    s_mov_b32 s7, s9
 ; GFX10NSA-NEXT:    s_mov_b32 s9, s11
 ; GFX10NSA-NEXT:    v_and_or_b32 v0, v0, v3, v1
 ; GFX10NSA-NEXT:    s_mov_b32 s11, s13
 ; GFX10NSA-NEXT:    v_and_or_b32 v1, v2, v3, s12
-; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s14
+; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s28
 ; GFX10NSA-NEXT:    image_gather4_cl v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
 ; GFX10NSA-NEXT:    ; implicit-def: $vcc_hi
 ; GFX10NSA-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10NSA-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.cl.2d.v4f32.f16(i32 1, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }

 ; GFX9-NEXT:    v_and_or_b32 v2, v3, v4, s12
 ; GFX9-NEXT:    s_and_b64 exec, exec, s[14:15]
 ; GFX9-NEXT:    image_gather4_c_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 a16
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10NSA-LABEL: gather4_c_cl_2d:
 ; GFX10NSA:       ; %bb.0: ; %main_body
-; GFX10NSA-NEXT:    s_mov_b32 s14, exec_lo
+; GFX10NSA-NEXT:    s_mov_b32 s28, exec_lo
 ; GFX10NSA-NEXT:    s_mov_b32 s0, s2
 ; GFX10NSA-NEXT:    s_wqm_b32 exec_lo, exec_lo
 ; GFX10NSA-NEXT:    v_mov_b32_e32 v4, 0xffff
 ; GFX10NSA-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
 ; GFX10NSA-NEXT:    s_mov_b32 s2, s4
 ; GFX10NSA-NEXT:    s_mov_b32 s4, s6
 ; GFX10NSA-NEXT:    s_mov_b32 s6, s8
 ; GFX10NSA-NEXT:    s_mov_b32 s8, s10
 ; GFX10NSA-NEXT:    s_mov_b32 s10, s12
 ; GFX10NSA-NEXT:    s_lshl_b32 s12, s0, 16
 ; GFX10NSA-NEXT:    s_mov_b32 s1, s3
 ; GFX10NSA-NEXT:    s_mov_b32 s3, s5
 ; GFX10NSA-NEXT:    s_mov_b32 s5, s7
 ; GFX10NSA-NEXT:    s_mov_b32 s7, s9
 ; GFX10NSA-NEXT:    s_mov_b32 s9, s11
 ; GFX10NSA-NEXT:    v_and_or_b32 v1, v1, v4, v2
 ; GFX10NSA-NEXT:    s_mov_b32 s11, s13
 ; GFX10NSA-NEXT:    v_and_or_b32 v2, v3, v4, s12
-; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s14
+; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s28
 ; GFX10NSA-NEXT:    image_gather4_c_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
 ; GFX10NSA-NEXT:    ; implicit-def: $vcc_hi
 ; GFX10NSA-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10NSA-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.cl.2d.v4f32.f16(i32 1, float %zcompare, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }

 ; GFX9-NEXT:    v_and_or_b32 v1, v1, v3, v2
 ; GFX9-NEXT:    s_and_b64 exec, exec, s[14:15]
 ; GFX9-NEXT:    image_gather4_b v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 a16
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10NSA-LABEL: gather4_b_2d:
 ; GFX10NSA:       ; %bb.0: ; %main_body
-; GFX10NSA-NEXT:    s_mov_b32 s14, exec_lo
+; GFX10NSA-NEXT:    s_mov_b32 s28, exec_lo
 ; GFX10NSA-NEXT:    s_mov_b32 s0, s2
 ; GFX10NSA-NEXT:    s_wqm_b32 exec_lo, exec_lo
 ; GFX10NSA-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
 ; GFX10NSA-NEXT:    s_mov_b32 s1, s3
 ; GFX10NSA-NEXT:    s_mov_b32 s2, s4
 ; GFX10NSA-NEXT:    s_mov_b32 s3, s5
 ; GFX10NSA-NEXT:    s_mov_b32 s4, s6
 ; GFX10NSA-NEXT:    s_mov_b32 s5, s7
 ; GFX10NSA-NEXT:    s_mov_b32 s6, s8
 ; GFX10NSA-NEXT:    s_mov_b32 s7, s9
 ; GFX10NSA-NEXT:    s_mov_b32 s8, s10
 ; GFX10NSA-NEXT:    s_mov_b32 s9, s11
 ; GFX10NSA-NEXT:    s_mov_b32 s10, s12
 ; GFX10NSA-NEXT:    s_mov_b32 s11, s13
 ; GFX10NSA-NEXT:    v_and_or_b32 v1, v1, 0xffff, v2
-; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s14
+; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s28
 ; GFX10NSA-NEXT:    image_gather4_b v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
 ; GFX10NSA-NEXT:    ; implicit-def: $vcc_hi
 ; GFX10NSA-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10NSA-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.b.2d.v4f32.f32.f16(i32 1, float %bias, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }

 ; GFX9-NEXT:    v_and_or_b32 v2, v2, v4, v3
 ; GFX9-NEXT:    s_and_b64 exec, exec, s[14:15]
 ; GFX9-NEXT:    image_gather4_c_b v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 a16
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10NSA-LABEL: gather4_c_b_2d:
 ; GFX10NSA:       ; %bb.0: ; %main_body
-; GFX10NSA-NEXT:    s_mov_b32 s14, exec_lo
+; GFX10NSA-NEXT:    s_mov_b32 s28, exec_lo
 ; GFX10NSA-NEXT:    s_mov_b32 s0, s2
 ; GFX10NSA-NEXT:    s_wqm_b32 exec_lo, exec_lo
 ; GFX10NSA-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
 ; GFX10NSA-NEXT:    s_mov_b32 s1, s3
 ; GFX10NSA-NEXT:    s_mov_b32 s2, s4
 ; GFX10NSA-NEXT:    s_mov_b32 s3, s5
 ; GFX10NSA-NEXT:    s_mov_b32 s4, s6
 ; GFX10NSA-NEXT:    s_mov_b32 s5, s7
 ; GFX10NSA-NEXT:    s_mov_b32 s6, s8
 ; GFX10NSA-NEXT:    s_mov_b32 s7, s9
 ; GFX10NSA-NEXT:    s_mov_b32 s8, s10
 ; GFX10NSA-NEXT:    s_mov_b32 s9, s11
 ; GFX10NSA-NEXT:    s_mov_b32 s10, s12
 ; GFX10NSA-NEXT:    s_mov_b32 s11, s13
 ; GFX10NSA-NEXT:    v_and_or_b32 v2, v2, 0xffff, v3
-; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s14
+; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s28
 ; GFX10NSA-NEXT:    image_gather4_c_b v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
 ; GFX10NSA-NEXT:    ; implicit-def: $vcc_hi
 ; GFX10NSA-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10NSA-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.2d.v4f32.f32.f16(i32 1, float %bias, float %zcompare, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }

 ; GFX9-NEXT:    v_and_or_b32 v2, v3, v4, s12
 ; GFX9-NEXT:    s_and_b64 exec, exec, s[14:15]
 ; GFX9-NEXT:    image_gather4_b_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 a16
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10NSA-LABEL: gather4_b_cl_2d:
 ; GFX10NSA:       ; %bb.0: ; %main_body
-; GFX10NSA-NEXT:    s_mov_b32 s14, exec_lo
+; GFX10NSA-NEXT:    s_mov_b32 s28, exec_lo
 ; GFX10NSA-NEXT:    s_mov_b32 s0, s2
 ; GFX10NSA-NEXT:    s_wqm_b32 exec_lo, exec_lo
 ; GFX10NSA-NEXT:    v_mov_b32_e32 v4, 0xffff
 ; GFX10NSA-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
 ; GFX10NSA-NEXT:    s_mov_b32 s2, s4
 ; GFX10NSA-NEXT:    s_mov_b32 s4, s6
 ; GFX10NSA-NEXT:    s_mov_b32 s6, s8
 ; GFX10NSA-NEXT:    s_mov_b32 s8, s10
 ; GFX10NSA-NEXT:    s_mov_b32 s10, s12
 ; GFX10NSA-NEXT:    s_lshl_b32 s12, s0, 16
 ; GFX10NSA-NEXT:    s_mov_b32 s1, s3
 ; GFX10NSA-NEXT:    s_mov_b32 s3, s5
 ; GFX10NSA-NEXT:    s_mov_b32 s5, s7
 ; GFX10NSA-NEXT:    s_mov_b32 s7, s9
 ; GFX10NSA-NEXT:    s_mov_b32 s9, s11
 ; GFX10NSA-NEXT:    v_and_or_b32 v1, v1, v4, v2
 ; GFX10NSA-NEXT:    s_mov_b32 s11, s13
 ; GFX10NSA-NEXT:    v_and_or_b32 v2, v3, v4, s12
-; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s14
+; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s28
 ; GFX10NSA-NEXT:    image_gather4_b_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
 ; GFX10NSA-NEXT:    ; implicit-def: $vcc_hi
 ; GFX10NSA-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10NSA-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.b.cl.2d.v4f32.f32.f16(i32 1, float %bias, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }

 ; GFX9-NEXT:    v_and_or_b32 v3, v4, v5, s12
 ; GFX9-NEXT:    s_and_b64 exec, exec, s[14:15]
 ; GFX9-NEXT:    image_gather4_c_b_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 a16
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10NSA-LABEL: gather4_c_b_cl_2d:
 ; GFX10NSA:       ; %bb.0: ; %main_body
-; GFX10NSA-NEXT:    s_mov_b32 s14, exec_lo
+; GFX10NSA-NEXT:    s_mov_b32 s28, exec_lo
 ; GFX10NSA-NEXT:    s_mov_b32 s0, s2
 ; GFX10NSA-NEXT:    s_wqm_b32 exec_lo, exec_lo
 ; GFX10NSA-NEXT:    v_mov_b32_e32 v5, 0xffff
 ; GFX10NSA-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
 ; GFX10NSA-NEXT:    s_mov_b32 s2, s4
 ; GFX10NSA-NEXT:    s_mov_b32 s4, s6
 ; GFX10NSA-NEXT:    s_mov_b32 s6, s8
 ; GFX10NSA-NEXT:    s_mov_b32 s8, s10
 ; GFX10NSA-NEXT:    s_mov_b32 s10, s12
 ; GFX10NSA-NEXT:    s_lshl_b32 s12, s0, 16
 ; GFX10NSA-NEXT:    s_mov_b32 s1, s3
 ; GFX10NSA-NEXT:    s_mov_b32 s3, s5
 ; GFX10NSA-NEXT:    s_mov_b32 s5, s7
 ; GFX10NSA-NEXT:    s_mov_b32 s7, s9
 ; GFX10NSA-NEXT:    s_mov_b32 s9, s11
 ; GFX10NSA-NEXT:    v_and_or_b32 v2, v2, v5, v3
 ; GFX10NSA-NEXT:    s_mov_b32 s11, s13
 ; GFX10NSA-NEXT:    v_and_or_b32 v3, v4, v5, s12
-; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s14
+; GFX10NSA-NEXT:    s_and_b32 exec_lo, exec_lo, s28
 ; GFX10NSA-NEXT:    image_gather4_c_b_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
 ; GFX10NSA-NEXT:    ; implicit-def: $vcc_hi
 ; GFX10NSA-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10NSA-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.2d.v4f32.f32.f16(i32 1, float %bias, float %zcompare, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }

llvm/test/CodeGen/AMDGPU/atomic_optimizations_pixelshader.ll

 ; GFX1064-NEXT:    s_and_saveexec_b64 s[8:9], s[10:11]
 ; GFX1064-NEXT:    s_cbranch_execz BB0_4
 ; GFX1064-NEXT:  ; %bb.1:
 ; GFX1064-NEXT:    s_mov_b64 s[12:13], exec
 ; GFX1064-NEXT:    ; implicit-def: $vgpr1
 ; GFX1064-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s12, 0
 ; GFX1064-NEXT:    v_mbcnt_hi_u32_b32_e64 v0, s13, v0
 ; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[30:31], vcc
+; GFX1064-NEXT:    s_and_saveexec_b64 s[28:29], vcc
 ; GFX1064-NEXT:    s_cbranch_execz BB0_3
 ; GFX1064-NEXT:  ; %bb.2:
 ; GFX1064-NEXT:    s_bcnt1_i32_b64 s12, s[12:13]
 ; GFX1064-NEXT:    v_mul_u32_u24_e64 v1, s12, 5
 ; GFX1064-NEXT:    buffer_atomic_add v1, off, s[4:7], 0 glc
 ; GFX1064-NEXT:  BB0_3:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX1064-NEXT:    s_or_b64 exec, exec, s[30:31]
+; GFX1064-NEXT:    s_or_b64 exec, exec, s[28:29]
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0)
 ; GFX1064-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX1064-NEXT:    v_mad_u32_u24 v0, v0, 5, s4
 ; GFX1064-NEXT:  BB0_4: ; %Flow
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; GFX1064-NEXT:    s_wqm_b64 s[4:5], -1
 ; GFX1064-NEXT:    s_andn2_b64 vcc, exec, s[4:5]
 ; GFX1064-NEXT:    s_cbranch_vccnz BB0_6

 ; GFX1064-NEXT:    v_writelane_b32 v3, s13, 32
 ; GFX1064-NEXT:    s_mov_b64 exec, s[10:11]
 ; GFX1064-NEXT:    v_mbcnt_hi_u32_b32_e64 v0, exec_hi, v0
 ; GFX1064-NEXT:    s_or_saveexec_b64 s[10:11], -1
 ; GFX1064-NEXT:    v_writelane_b32 v3, s14, 48
 ; GFX1064-NEXT:    s_mov_b64 exec, s[10:11]
 ; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
 ; GFX1064-NEXT:    ; implicit-def: $vgpr0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[30:31], vcc
+; GFX1064-NEXT:    s_and_saveexec_b64 s[28:29], vcc
 ; GFX1064-NEXT:    s_cbranch_execz BB1_3
 ; GFX1064-NEXT:  ; %bb.2:
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, s12
 ; GFX1064-NEXT:    buffer_atomic_add v0, off, s[4:7], 0 glc
 ; GFX1064-NEXT:  BB1_3:
 ; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
-; GFX1064-NEXT:    s_or_b64 exec, exec, s[30:31]
+; GFX1064-NEXT:    s_or_b64 exec, exec, s[28:29]
 ; GFX1064-NEXT:    s_waitcnt vmcnt(0)
 ; GFX1064-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX1064-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX1064-NEXT:    v_add_nc_u32_e32 v0, s4, v0
 ; GFX1064-NEXT:  BB1_4: ; %Flow
 ; GFX1064-NEXT:    s_or_b64 exec, exec, s[8:9]
 ; GFX1064-NEXT:    s_wqm_b64 s[4:5], -1
 ; GFX1064-NEXT:    s_andn2_b64 vcc, exec, s[4:5]

llvm/test/CodeGen/AMDGPU/flat-scratch.ll

; GFX10-NEXT:    s_setpc_b64 s[30:31]
   call void @llvm.memset.p5i8.i64(i8 addrspace(5)* align 2 dereferenceable(64) %cast, i8 0, i64 64, i1 false)
   ret void
 }
 
 define amdgpu_kernel void @store_load_sindex_kernel(i32 %idx) {
 ; GFX9-LABEL: store_load_sindex_kernel:
 ; GFX9:       ; %bb.0: ; %bb
 ; GFX9-NEXT:    s_load_dword s0, s[0:1], 0x24
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    s_add_u32 flat_scratch_lo, s2, s5
 ; GFX9-NEXT:    s_addc_u32 flat_scratch_hi, s3, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 15
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    s_lshl_b32 s1, s0, 2
 ; GFX9-NEXT:    s_and_b32 s0, s0, 15
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 2
 ; GFX9-NEXT:    s_add_u32 s1, 4, s1
 ; GFX9-NEXT:    scratch_store_dword off, v0, s1
 ; GFX9-NEXT:    s_add_u32 s0, 4, s0
 ; GFX9-NEXT:    scratch_load_dword v0, off, s0
 ; GFX9-NEXT:    s_endpgm
; GFX10-NEXT:    s_setpc_b64 s[30:31]
   call void @llvm.memset.p5i8.i64(i8 addrspace(5)* align 2 dereferenceable(64) %cast, i8 0, i64 64, i1 false)
   ret void
 }
 
 define amdgpu_kernel void @store_load_sindex_small_offset_kernel(i32 %idx) {
 ; GFX9-LABEL: store_load_sindex_small_offset_kernel:
 ; GFX9:       ; %bb.0: ; %bb
 ; GFX9-NEXT:    s_load_dword s0, s[0:1], 0x24
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    s_add_u32 flat_scratch_lo, s2, s5
 ; GFX9-NEXT:    s_addc_u32 flat_scratch_hi, s3, 0
 ; GFX9-NEXT:    s_mov_b32 vcc_hi, 0
 ; GFX9-NEXT:    scratch_load_dword v0, off, vcc_hi offset:4
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    s_lshl_b32 s1, s0, 2
 ; GFX9-NEXT:    s_and_b32 s0, s0, 15
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 2
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 15
 ; GFX9-NEXT:    s_add_u32 s1, 0x104, s1
 ; GFX9-NEXT:    scratch_store_dword off, v0, s1
 ; GFX9-NEXT:    s_add_u32 s0, 0x104, s0
 ; GFX9-NEXT:    scratch_load_dword v0, off, s0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: store_load_sindex_small_offset_kernel:
 ; GFX10:       ; %bb.0: ; %bb
 ; GFX10-NEXT:    s_add_u32 s2, s2, s5
 ; GFX10-NEXT:    s_addc_u32 s3, s3, 0
 ; GFX10-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_LO), s2
 ; GFX10-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s3
 ; GFX10-NEXT:    s_load_dword s0, s[0:1], 0x24
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    scratch_load_dword v0, off, off offset:4
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    v_mov_b32_e32 v0, 15
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_and_b32 s1, s0, 15
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, 2
 ; GFX10-NEXT:    s_lshl_b32 s1, s1, 2
 ; GFX10-NEXT:    s_add_u32 s0, 0x104, s0
 ; GFX10-NEXT:    s_add_u32 s1, 0x104, s1
 ; GFX10-NEXT:    scratch_store_dword off, v0, s0
 ; GFX10-NEXT:    scratch_load_dword v0, off, s1
 ; GFX10-NEXT:    s_endpgm
; GFX10-NEXT:    s_setpc_b64 s[30:31]
   call void @llvm.memset.p5i8.i64(i8 addrspace(5)* align 2 dereferenceable(64) %cast, i8 0, i64 64, i1 false)
   ret void
 }
 
 define amdgpu_kernel void @store_load_sindex_large_offset_kernel(i32 %idx) {
 ; GFX9-LABEL: store_load_sindex_large_offset_kernel:
 ; GFX9:       ; %bb.0: ; %bb
 ; GFX9-NEXT:    s_load_dword s0, s[0:1], 0x24
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    s_add_u32 flat_scratch_lo, s2, s5
 ; GFX9-NEXT:    s_addc_u32 flat_scratch_hi, s3, 0
 ; GFX9-NEXT:    s_mov_b32 vcc_hi, 0
 ; GFX9-NEXT:    scratch_load_dword v0, off, vcc_hi offset:4
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    s_lshl_b32 s1, s0, 2
 ; GFX9-NEXT:    s_and_b32 s0, s0, 15
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 2
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 15
 ; GFX9-NEXT:    s_add_u32 s1, 0x4004, s1
 ; GFX9-NEXT:    scratch_store_dword off, v0, s1
 ; GFX9-NEXT:    s_add_u32 s0, 0x4004, s0
 ; GFX9-NEXT:    scratch_load_dword v0, off, s0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: store_load_sindex_large_offset_kernel:
 ; GFX10:       ; %bb.0: ; %bb
 ; GFX10-NEXT:    s_add_u32 s2, s2, s5
 ; GFX10-NEXT:    s_addc_u32 s3, s3, 0
 ; GFX10-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_LO), s2
 ; GFX10-NEXT:    s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s3
 ; GFX10-NEXT:    s_load_dword s0, s[0:1], 0x24
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    scratch_load_dword v0, off, off offset:4
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    v_mov_b32_e32 v0, 15
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_and_b32 s1, s0, 15
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, 2
 ; GFX10-NEXT:    s_lshl_b32 s1, s1, 2
 ; GFX10-NEXT:    s_add_u32 s0, 0x4004, s0
 ; GFX10-NEXT:    s_add_u32 s1, 0x4004, s1
 ; GFX10-NEXT:    scratch_store_dword off, v0, s0
 ; GFX10-NEXT:    scratch_load_dword v0, off, s1
 ; GFX10-NEXT:    s_endpgm
bb:
   %i12 = load volatile i32, i32 addrspace(5)* %i10, align 4
   ret void
 }
 
 define amdgpu_kernel void @store_load_vidx_sidx_offset(i32 %sidx) {
 ; GFX9-LABEL: store_load_vidx_sidx_offset:
 ; GFX9:       ; %bb.0: ; %bb
 ; GFX9-NEXT:    s_load_dword s0, s[0:1], 0x24
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    s_add_u32 flat_scratch_lo, s2, s5
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 4
 ; GFX9-NEXT:    s_addc_u32 flat_scratch_hi, s3, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    v_add_u32_e32 v0, s0, v0
 ; GFX9-NEXT:    v_lshl_add_u32 v0, v0, 2, v1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 15
 ; GFX9-NEXT:    scratch_store_dword v0, v1, off offset:1024
 ; GFX9-NEXT:    scratch_load_dword v0, v0, off offset:1024
 ; GFX9-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: store_load_vidx_sidx_offset:

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.d16.dim.ll

 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    global_store_dword v[0:1], v3, off
 ; GFX9-NEXT:    v_mov_b32_e32 v0, v2
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: image_sample_2d_f16_tfe:
 ; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s14, exec_lo
+; GFX10-NEXT:    s_mov_b32 s28, exec_lo
 ; GFX10-NEXT:    s_wqm_b32 exec_lo, exec_lo
 ; GFX10-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-NEXT:    v_mov_b32_e32 v3, v2
-; GFX10-NEXT:    s_and_b32 exec_lo, exec_lo, s14
+; GFX10-NEXT:    s_and_b32 exec_lo, exec_lo, s28
 ; GFX10-NEXT:    image_sample v[2:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D tfe d16
 ; GFX10-NEXT:    v_mov_b32_e32 v0, s12
 ; GFX10-NEXT:    v_mov_b32_e32 v1, s13
 ; GFX10-NEXT:    ; implicit-def: $vcc_hi
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    global_store_dword v[0:1], v3, off
 ; GFX10-NEXT:    v_mov_b32_e32 v0, v2
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.dim.ll

 ; GFX6789-NEXT:    v_mov_b32_e32 v6, s13
 ; GFX6789-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6789-NEXT:    global_store_dword v[5:6], v4, off
 ; GFX6789-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6789-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: sample_1d_tfe:
 ; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s14, exec_lo ; encoding: [0x7e,0x03,0x8e,0xbe]
+; GFX10-NEXT:    s_mov_b32 s28, exec_lo ; encoding: [0x7e,0x03,0x9c,0xbe]
 ; GFX10-NEXT:    s_wqm_b32 exec_lo, exec_lo ; encoding: [0x7e,0x09,0xfe,0xbe]
 ; GFX10-NEXT:    v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e]
 ; GFX10-NEXT:    v_mov_b32_e32 v0, 0 ; encoding: [0x80,0x02,0x00,0x7e]
 ; GFX10-NEXT:    v_mov_b32_e32 v1, v0 ; encoding: [0x00,0x03,0x02,0x7e]
 ; GFX10-NEXT:    v_mov_b32_e32 v2, v0 ; encoding: [0x00,0x03,0x04,0x7e]
 ; GFX10-NEXT:    v_mov_b32_e32 v3, v0 ; encoding: [0x00,0x03,0x06,0x7e]
 ; GFX10-NEXT:    v_mov_b32_e32 v4, v0 ; encoding: [0x00,0x03,0x08,0x7e]
-; GFX10-NEXT:    s_and_b32 exec_lo, exec_lo, s14 ; encoding: [0x7e,0x0e,0x7e,0x87]
+; GFX10-NEXT:    s_and_b32 exec_lo, exec_lo, s28 ; encoding: [0x7e,0x1c,0x7e,0x87]
 ; GFX10-NEXT:    image_sample v[0:4], v5, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D tfe ; encoding: [0x00,0x0f,0x81,0xf0,0x05,0x00,0x40,0x00]
 ; GFX10-NEXT:    v_mov_b32_e32 v5, s12 ; encoding: [0x0c,0x02,0x0a,0x7e]
 ; GFX10-NEXT:    v_mov_b32_e32 v6, s13 ; encoding: [0x0d,0x02,0x0c,0x7e]
 ; GFX10-NEXT:    ; implicit-def: $vcc_hi
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
 ; GFX10-NEXT:    global_store_dword v[5:6], v4, off ; encoding: [0x00,0x80,0x70,0xdc,0x05,0x04,0x7d,0x00]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb]
 ; GFX10-NEXT:    ; return to shader part epilog

 ; GFX6789-NEXT:    v_mov_b32_e32 v6, s13
 ; GFX6789-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6789-NEXT:    global_store_dword v[5:6], v4, off
 ; GFX6789-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6789-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: sample_1d_lwe:
 ; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s14, exec_lo ; encoding: [0x7e,0x03,0x8e,0xbe]
+; GFX10-NEXT:    s_mov_b32 s28, exec_lo ; encoding: [0x7e,0x03,0x9c,0xbe]
 ; GFX10-NEXT:    s_wqm_b32 exec_lo, exec_lo ; encoding: [0x7e,0x09,0xfe,0xbe]
 ; GFX10-NEXT:    v_mov_b32_e32 v5, v0 ; encoding: [0x00,0x03,0x0a,0x7e]
 ; GFX10-NEXT:    v_mov_b32_e32 v0, 0 ; encoding: [0x80,0x02,0x00,0x7e]
 ; GFX10-NEXT:    v_mov_b32_e32 v1, v0 ; encoding: [0x00,0x03,0x02,0x7e]
 ; GFX10-NEXT:    v_mov_b32_e32 v2, v0 ; encoding: [0x00,0x03,0x04,0x7e]
 ; GFX10-NEXT:    v_mov_b32_e32 v3, v0 ; encoding: [0x00,0x03,0x06,0x7e]
 ; GFX10-NEXT:    v_mov_b32_e32 v4, v0 ; encoding: [0x00,0x03,0x08,0x7e]
-; GFX10-NEXT:    s_and_b32 exec_lo, exec_lo, s14 ; encoding: [0x7e,0x0e,0x7e,0x87]
+; GFX10-NEXT:    s_and_b32 exec_lo, exec_lo, s28 ; encoding: [0x7e,0x1c,0x7e,0x87]
 ; GFX10-NEXT:    image_sample v[0:4], v5, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D lwe ; encoding: [0x00,0x0f,0x82,0xf0,0x05,0x00,0x40,0x00]
 ; GFX10-NEXT:    v_mov_b32_e32 v5, s12 ; encoding: [0x0c,0x02,0x0a,0x7e]
 ; GFX10-NEXT:    v_mov_b32_e32 v6, s13 ; encoding: [0x0d,0x02,0x0c,0x7e]
 ; GFX10-NEXT:    ; implicit-def: $vcc_hi
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
 ; GFX10-NEXT:    global_store_dword v[5:6], v4, off ; encoding: [0x00,0x80,0x70,0xdc,0x05,0x04,0x7d,0x00]
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0 ; encoding: [0x00,0x00,0xfd,0xbb]
 ; GFX10-NEXT:    ; return to shader part epilog