Select `G_SADDE`, `G_UADDE`, `G_SSUBE` and `G_USUBE` for AArch64

[niwinanto]

Summary

Instruction select G_SADDE, G_UADDE, G_SSUBE and G_USUBE for AArch64.

Fix the issue: https://github.com/llvm/llvm-project/issues/59407

[tobias-stadler]

I fixed these same issues here https://reviews.llvm.org/D153164 2 minutes earlier. I do generate better code though, because I generate less instructions for carry-in setting and optimize these instructions away whenever possible. For >O0 I generate identical code to SelectionDag. I also have higher test coverage.

[niwinanto]

In D153165#4429338, @tobias-stadler wrote:

I fixed these same issues here https://reviews.llvm.org/D153164 2 minutes earlier. I do generate better code though, because I generate less instructions for carry-in setting and optimize these instructions away whenever possible. For >O0 I generate identical code to SelectionDag. I also have higher test coverage.

Ohh cool. Insane coincidence. My plan was to support missing instructions, other wise it can cause internal compiler crashes. I did not touch the optimization, hence less optimized code is expected.

[aemerson]

In D153165#4430161, @niwinanto wrote:

In D153165#4429338, @tobias-stadler wrote:

I fixed these same issues here https://reviews.llvm.org/D153164 2 minutes earlier. I do generate better code though, because I generate less instructions for carry-in setting and optimize these instructions away whenever possible. For >O0 I generate identical code to SelectionDag. I also have higher test coverage.

Ohh cool. Insane coincidence. My plan was to support missing instructions, other wise it can cause internal compiler crashes. I did not touch the optimization, hence less optimized code is expected.

Insane indeed! By internal compiler crashes do you mean fallbacks? If so I generally don't class them as crashes (although certainly it's great to fix them).

Thanks for this work anyway, I'll review the other patch from @tobias-stadler

[arsenm]

Can this be abandoned?

[niwinanto]

In D153165#4612362, @arsenm wrote:

Can this be abandoned?

Yes, Its done https://reviews.llvm.org/D153164 here.

llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp

MachineInstr *emitAddSub(
       MachineIRBuilder &MIRBuilder) const;
   MachineInstr *emitADD(Register DefReg, MachineOperand &LHS,
                         MachineOperand &RHS,
                         MachineIRBuilder &MIRBuilder) const;
   MachineInstr *emitADDS(Register Dst, MachineOperand &LHS, MachineOperand &RHS,
                          MachineIRBuilder &MIRBuilder) const;
   MachineInstr *emitSUBS(Register Dst, MachineOperand &LHS, MachineOperand &RHS,
                          MachineIRBuilder &MIRBuilder) const;
+  MachineInstr *emitADCS(Register Dst, MachineOperand &LHS, MachineOperand &RHS,
+                         MachineIRBuilder &MIRBuilder) const;
+  MachineInstr *emitSetCarryFlag(MachineOperand &CarryIn,
+                                 MachineIRBuilder &MIRBuilder) const;
+  MachineInstr *emitSBCS(Register Dst, MachineOperand &LHS, MachineOperand &RHS,
+                         MachineIRBuilder &MIRBuilder) const;
   MachineInstr *emitCMN(MachineOperand &LHS, MachineOperand &RHS,
                         MachineIRBuilder &MIRBuilder) const;
   MachineInstr *emitTST(MachineOperand &LHS, MachineOperand &RHS,
                         MachineIRBuilder &MIRBuilder) const;
   MachineInstr *emitSelect(Register Dst, Register LHS, Register RHS,
                            AArch64CC::CondCode CC,
                            MachineIRBuilder &MIRBuilder) const;
   MachineInstr *emitExtractVectorElt(std::optional<Register> DstReg,
case TargetOpcode::G_OR: {
     return constrainSelectedInstRegOperands(I, TII, TRI, RBI);
   }
 
   case TargetOpcode::G_PTR_ADD: {
     emitADD(I.getOperand(0).getReg(), I.getOperand(1), I.getOperand(2), MIB);
     I.eraseFromParent();
     return true;
   }
+  case TargetOpcode::G_SADDE:
+  case TargetOpcode::G_UADDE:
+  case TargetOpcode::G_SSUBE:
+  case TargetOpcode::G_USUBE: {
+    // Add 0xFFFFFFFF and carryIn to update PSTATE.C flag
+    // FIXME : Try some other optimal operation 
+    emitSetCarryFlag(I.getOperand(4), MIB);
+  }
+    [[fallthrough]];
   case TargetOpcode::G_SADDO:
   case TargetOpcode::G_UADDO:
   case TargetOpcode::G_SSUBO:
   case TargetOpcode::G_USUBO: {
     // Emit the operation and get the correct condition code.
     auto OpAndCC = emitOverflowOp(Opcode, I.getOperand(0).getReg(),
                                   I.getOperand(2), I.getOperand(3), MIB);
 
const std::array<std::array<unsigned, 2>, 5> OpcTable{
        {AArch64::ADDSXrs, AArch64::ADDSWrs},
        {AArch64::ADDSXrr, AArch64::ADDSWrr},
        {AArch64::SUBSXri, AArch64::SUBSWri},
        {AArch64::ADDSXrx, AArch64::ADDSWrx}}};
   return emitAddSub(OpcTable, Dst, LHS, RHS, MIRBuilder);
 }
 
 MachineInstr *
+AArch64InstructionSelector::emitADCS(Register Dst, MachineOperand &LHS,
+                                     MachineOperand &RHS,
+                                     MachineIRBuilder &MIRBuilder) const {
+  MachineRegisterInfo &MRI = MIRBuilder.getMF().getRegInfo();
+  assert(LHS.isReg() && RHS.isReg() && "Expected register operands?");
+  auto Ty = MRI.getType(LHS.getReg());
+  assert(!Ty.isVector() && "Expected a scalar or pointer?");
+  unsigned Size = Ty.getSizeInBits();
+  assert((Size == 32 || Size == 64) && "Expected a 32-bit or 64-bit type only");
+  bool Is32Bit = Size == 32;
+  const std::array<unsigned, 2> OpcTable{AArch64::ADCSXr, AArch64::ADCSWr};
+  return emitInstr(OpcTable[Is32Bit], {Dst}, {LHS, RHS}, MIRBuilder);
+}
+
+MachineInstr *AArch64InstructionSelector::emitSetCarryFlag(
+    MachineOperand &CarryIn, MachineIRBuilder &MIRBuilder) const {
+  MachineRegisterInfo &MRI = MIRBuilder.getMF().getRegInfo();
+  assert(CarryIn.isReg() && "Expected register operands?");
+  auto Ty = MRI.getType(CarryIn.getReg());
+  unsigned Size = Ty.getSizeInBits();
+  assert(Size == 32 && "Expected 32-bit type only");
+  auto DestReg = MRI.createVirtualRegister(&AArch64::GPR32RegClass);
+  auto CopyReg = MRI.createVirtualRegister(&AArch64::GPR32RegClass);
+  MIRBuilder.buildInstr(AArch64::MOVi32imm, {CopyReg}, {}).addImm(0xFFFFFFFF);
+  AArch64GenRegisterBankInfo::constrainGenericRegister(
+      CopyReg, AArch64::GPR32RegClass, MRI);
+  auto MI =
+      MIRBuilder.buildInstr(AArch64::ADDSWrr, {DestReg}, {CarryIn, CopyReg});
+  constrainSelectedInstRegOperands(*MI, TII, TRI, RBI);
+  return &*MI;
+}
+
+MachineInstr *
+AArch64InstructionSelector::emitSBCS(Register Dst, MachineOperand &LHS,
+                                     MachineOperand &RHS,
+                                     MachineIRBuilder &MIRBuilder) const {
+  MachineRegisterInfo &MRI = MIRBuilder.getMF().getRegInfo();
+  assert(LHS.isReg() && RHS.isReg() && "Expected register operands?");
+  auto Ty = MRI.getType(LHS.getReg());
+  assert(!Ty.isVector() && "Expected a scalar or pointer?");
+  unsigned Size = Ty.getSizeInBits();
+  assert((Size == 32 || Size == 64) && "Expected a 32-bit or 64-bit type only");
+  bool Is32Bit = Size == 32;
+  const std::array<unsigned, 2> OpcTable{AArch64::SBCSXr, AArch64::SBCSWr};
+  return emitInstr(OpcTable[Is32Bit], {Dst}, {LHS, RHS}, MIRBuilder);
+}
+
+MachineInstr *
 AArch64InstructionSelector::emitSUBS(Register Dst, MachineOperand &LHS,
                                      MachineOperand &RHS,
                                      MachineIRBuilder &MIRBuilder) const {
   const std::array<std::array<unsigned, 2>, 5> OpcTable{
       {{AArch64::SUBSXri, AArch64::SUBSWri},
        {AArch64::SUBSXrs, AArch64::SUBSWrs},
        {AArch64::SUBSXrr, AArch64::SUBSWrr},
        {AArch64::ADDSXri, AArch64::ADDSWri},
AArch64InstructionSelector::emitOverflowOp(unsigned Opcode, Register Dst,
   case TargetOpcode::G_SADDO:
     return std::make_pair(emitADDS(Dst, LHS, RHS, MIRBuilder), AArch64CC::VS);
   case TargetOpcode::G_UADDO:
     return std::make_pair(emitADDS(Dst, LHS, RHS, MIRBuilder), AArch64CC::HS);
   case TargetOpcode::G_SSUBO:
     return std::make_pair(emitSUBS(Dst, LHS, RHS, MIRBuilder), AArch64CC::VS);
   case TargetOpcode::G_USUBO:
     return std::make_pair(emitSUBS(Dst, LHS, RHS, MIRBuilder), AArch64CC::LO);
+  case TargetOpcode::G_SADDE:
+    return std::make_pair(emitADCS(Dst, LHS, RHS, MIRBuilder), AArch64CC::VS);
+  case TargetOpcode::G_UADDE:
+    return std::make_pair(emitADCS(Dst, LHS, RHS, MIRBuilder), AArch64CC::HS);
+  case TargetOpcode::G_SSUBE:
+    return std::make_pair(emitSBCS(Dst, LHS, RHS, MIRBuilder), AArch64CC::VS);
+  case TargetOpcode::G_USUBE:
+    return std::make_pair(emitSBCS(Dst, LHS, RHS, MIRBuilder), AArch64CC::LO);
   }
 }
 
 /// Returns true if @p Val is a tree of AND/OR/CMP operations that can be
 /// expressed as a conjunction.
 /// \param CanNegate    Set to true if we can negate the whole sub-tree just by
 ///                     changing the conditions on the CMP tests.
 ///                     (this means we can call emitConjunctionRec() with

llvm/test/CodeGen/AArch64/Atomics/aarch64-atomicrmw-lse2.ll

 ; -O1:    add x9, x0, x1
 ; -O1:    stlxr w10, x9, [x8]
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_monotonic:
-; -O0:    adds x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_monotonic:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acquire:
-; -O0:    adds x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acquire:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_release:
-; -O0:    adds x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_release:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acq_rel:
-; -O0:    adds x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acq_rel:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_seq_cst:
-; -O0:    adds x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_seq_cst:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 16
     ret i128 %r
 }

 ; -O1:    add x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_monotonic:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acquire:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acquire:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_release:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_release:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value release, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 1
     ret i128 %r

 ; -O1:    sub x9, x0, x1
 ; -O1:    stlxr w10, x9, [x8]
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_monotonic:
-; -O0:    subs x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_monotonic:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acquire:
-; -O0:    subs x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acquire:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_release:
-; -O0:    subs x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_release:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
-; -O0:    subs x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
-; -O0:    subs x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value seq_cst, align 16
     ret i128 %r
 }

 ; -O1:    sub x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_acquire:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_acquire:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value acquire, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_release:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_release:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value release, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_acq_rel:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_acq_rel:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value acq_rel, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_seq_cst:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_seq_cst:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value seq_cst, align 1
     ret i128 %r

llvm/test/CodeGen/AArch64/Atomics/aarch64-atomicrmw-lse2_lse128.ll

 ; CHECK-LABEL: atomicrmw_add_i64_aligned_seq_cst:
 ; CHECK:    ldaddal x1, x0, [x0]
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_monotonic:
-; -O0:    adds x2, x10, x12
+; -O0:    adds x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    casp x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_monotonic:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    adds x8, x4, x2
 ; -O1:    casp x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acquire:
-; -O0:    adds x2, x10, x12
+; -O0:    adds x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspa x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acquire:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    adds x8, x4, x2
 ; -O1:    caspa x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_release:
-; -O0:    adds x2, x10, x12
+; -O0:    adds x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspl x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_release:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    adds x8, x4, x2
 ; -O1:    caspl x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw add ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acq_rel:
-; -O0:    adds x2, x10, x12
+; -O0:    adds x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspal x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acq_rel:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    adds x8, x4, x2
 ; -O1:    caspal x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_seq_cst:
-; -O0:    adds x2, x10, x12
+; -O0:    adds x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspal x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_seq_cst:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    adds x8, x4, x2
 ; -O1:    caspal x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 16

 ; -O1:    add x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_monotonic:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acquire:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acquire:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_release:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_release:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value release, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 1
     ret i128 %r

 ; CHECK-LABEL: atomicrmw_sub_i64_aligned_seq_cst:
 ; CHECK:    ldaddal x8, x0, [x0]
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_monotonic:
-; -O0:    subs x2, x10, x12
+; -O0:    subs x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    casp x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_monotonic:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    subs x8, x4, x2
 ; -O1:    casp x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acquire:
-; -O0:    subs x2, x10, x12
+; -O0:    subs x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspa x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acquire:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    subs x8, x4, x2
 ; -O1:    caspa x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw sub ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_release:
-; -O0:    subs x2, x10, x12
+; -O0:    subs x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspl x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_release:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    subs x8, x4, x2
 ; -O1:    caspl x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw sub ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
-; -O0:    subs x2, x10, x12
+; -O0:    subs x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspal x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    subs x8, x4, x2
 ; -O1:    caspal x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw sub ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
-; -O0:    subs x2, x10, x12
+; -O0:    subs x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspal x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    subs x8, x4, x2
 ; -O1:    caspal x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw sub ptr %ptr, i128 %value seq_cst, align 16

 ; -O1:    sub x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_acquire:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_acquire:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value acquire, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_release:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_release:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value release, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_acq_rel:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_acq_rel:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value acq_rel, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_seq_cst:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_seq_cst:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value seq_cst, align 1
     ret i128 %r

llvm/test/CodeGen/AArch64/Atomics/aarch64-atomicrmw-outline_atomics.ll

 ; CHECK-LABEL: atomicrmw_add_i64_aligned_seq_cst:
 ; CHECK:    bl __aarch64_ldadd8_acq_rel
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_monotonic:
-; -O0:    adds x2, x0, x9
-; -O0:    bl __aarch64_cas16_relax
-; -O0:    subs x10, x10, x11
-; -O0:    ccmp x8, x9, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_monotonic:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acquire:
-; -O0:    adds x2, x0, x9
-; -O0:    bl __aarch64_cas16_acq
-; -O0:    subs x10, x10, x11
-; -O0:    ccmp x8, x9, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acquire:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_release:
-; -O0:    adds x2, x0, x9
-; -O0:    bl __aarch64_cas16_rel
-; -O0:    subs x10, x10, x11
-; -O0:    ccmp x8, x9, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_release:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acq_rel:
-; -O0:    adds x2, x0, x9
-; -O0:    bl __aarch64_cas16_acq_rel
-; -O0:    subs x10, x10, x11
-; -O0:    ccmp x8, x9, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acq_rel:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_seq_cst:
-; -O0:    adds x2, x0, x9
-; -O0:    bl __aarch64_cas16_acq_rel
-; -O0:    subs x10, x10, x11
-; -O0:    ccmp x8, x9, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_seq_cst:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 16
     ret i128 %r
 }

 ; -O1:    add x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_monotonic:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acquire:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acquire:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_release:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_release:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value release, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 1
     ret i128 %r

 ; -O1-LABEL: atomicrmw_sub_i64_aligned_seq_cst:
 ; -O1:    bl __aarch64_ldadd8_acq_rel
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_monotonic:
-; -O0:    subs x2, x0, x9
-; -O0:    bl __aarch64_cas16_relax
-; -O0:    subs x10, x10, x11
-; -O0:    ccmp x8, x9, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_monotonic:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acquire:
-; -O0:    subs x2, x0, x9
-; -O0:    bl __aarch64_cas16_acq
-; -O0:    subs x10, x10, x11
-; -O0:    ccmp x8, x9, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acquire:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_release:
-; -O0:    subs x2, x0, x9
-; -O0:    bl __aarch64_cas16_rel
-; -O0:    subs x10, x10, x11
-; -O0:    ccmp x8, x9, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_release:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
-; -O0:    subs x2, x0, x9
-; -O0:    bl __aarch64_cas16_acq_rel
-; -O0:    subs x10, x10, x11
-; -O0:    ccmp x8, x9, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
-; -O0:    subs x2, x0, x9
-; -O0:    bl __aarch64_cas16_acq_rel
-; -O0:    subs x10, x10, x11
-; -O0:    ccmp x8, x9, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value seq_cst, align 16
     ret i128 %r
 }

 ; -O1:    sub x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_acquire:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_acquire:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value acquire, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_release:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_release:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value release, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_acq_rel:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_acq_rel:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value acq_rel, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_seq_cst:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_seq_cst:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value seq_cst, align 1
     ret i128 %r

llvm/test/CodeGen/AArch64/Atomics/aarch64-atomicrmw-rcpc.ll

 ; -O1:    add x9, x0, x1
 ; -O1:    stlxr w10, x9, [x8]
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_monotonic:
-; -O0:    adds x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_monotonic:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acquire:
-; -O0:    adds x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acquire:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_release:
-; -O0:    adds x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_release:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acq_rel:
-; -O0:    adds x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acq_rel:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_seq_cst:
-; -O0:    adds x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_seq_cst:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 16
     ret i128 %r
 }

 ; -O1:    add x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_monotonic:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acquire:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acquire:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_release:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_release:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value release, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 1
     ret i128 %r

 ; -O1:    sub x9, x0, x1
 ; -O1:    stlxr w10, x9, [x8]
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_monotonic:
-; -O0:    subs x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_monotonic:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acquire:
-; -O0:    subs x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acquire:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_release:
-; -O0:    subs x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_release:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
-; -O0:    subs x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
-; -O0:    subs x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value seq_cst, align 16
     ret i128 %r
 }

 ; -O1:    sub x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_acquire:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_acquire:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value acquire, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_release:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_release:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value release, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_acq_rel:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_acq_rel:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value acq_rel, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_seq_cst:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_seq_cst:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value seq_cst, align 1
     ret i128 %r

llvm/test/CodeGen/AArch64/Atomics/aarch64-atomicrmw-rcpc3.ll

 ; -O1:    add x9, x0, x1
 ; -O1:    stlxr w10, x9, [x8]
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_monotonic:
-; -O0:    adds x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_monotonic:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acquire:
-; -O0:    adds x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acquire:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_release:
-; -O0:    adds x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_release:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acq_rel:
-; -O0:    adds x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acq_rel:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_seq_cst:
-; -O0:    adds x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_seq_cst:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 16
     ret i128 %r
 }

 ; -O1:    add x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_monotonic:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acquire:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acquire:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_release:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_release:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value release, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 1
     ret i128 %r

 ; -O1:    sub x9, x0, x1
 ; -O1:    stlxr w10, x9, [x8]
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_monotonic:
-; -O0:    subs x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_monotonic:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acquire:
-; -O0:    subs x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acquire:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_release:
-; -O0:    subs x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_release:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
-; -O0:    subs x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
-; -O0:    subs x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value seq_cst, align 16
     ret i128 %r
 }

 ; -O1:    sub x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_acquire:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_acquire:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value acquire, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_release:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_release:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value release, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_acq_rel:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_acq_rel:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value acq_rel, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_seq_cst:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_seq_cst:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value seq_cst, align 1
     ret i128 %r

llvm/test/CodeGen/AArch64/Atomics/aarch64-atomicrmw-v8_1a.ll

 ; CHECK-LABEL: atomicrmw_add_i64_aligned_seq_cst:
 ; CHECK:    ldaddal x1, x0, [x0]
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_monotonic:
-; -O0:    adds x2, x10, x12
+; -O0:    adds x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    casp x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_monotonic:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    adds x8, x4, x2
 ; -O1:    casp x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acquire:
-; -O0:    adds x2, x10, x12
+; -O0:    adds x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspa x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acquire:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    adds x8, x4, x2
 ; -O1:    caspa x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_release:
-; -O0:    adds x2, x10, x12
+; -O0:    adds x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspl x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_release:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    adds x8, x4, x2
 ; -O1:    caspl x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw add ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acq_rel:
-; -O0:    adds x2, x10, x12
+; -O0:    adds x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspal x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acq_rel:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    adds x8, x4, x2
 ; -O1:    caspal x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_seq_cst:
-; -O0:    adds x2, x10, x12
+; -O0:    adds x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspal x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_seq_cst:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    adds x8, x4, x2
 ; -O1:    caspal x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 16

 ; -O1:    add x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_monotonic:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acquire:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acquire:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_release:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_release:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value release, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 1
     ret i128 %r

 ; CHECK-LABEL: atomicrmw_sub_i64_aligned_seq_cst:
 ; CHECK:    ldaddal x8, x0, [x0]
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_monotonic:
-; -O0:    subs x2, x10, x12
+; -O0:    subs x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    casp x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_monotonic:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    subs x8, x4, x2
 ; -O1:    casp x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acquire:
-; -O0:    subs x2, x10, x12
+; -O0:    subs x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspa x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acquire:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    subs x8, x4, x2
 ; -O1:    caspa x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw sub ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_release:
-; -O0:    subs x2, x10, x12
+; -O0:    subs x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspl x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_release:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    subs x8, x4, x2
 ; -O1:    caspl x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw sub ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
-; -O0:    subs x2, x10, x12
+; -O0:    subs x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspal x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    subs x8, x4, x2
 ; -O1:    caspal x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw sub ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
-; -O0:    subs x2, x10, x12
+; -O0:    subs x2, x9, x11
+; -O0:    and w11, w9, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    caspal x0, x1, x2, x3, [x8]
-; -O0:    subs x11, x8, x11
-; -O0:    ccmp x9, x10, #0, eq
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
 ; -O1:    ldp x4, x5, [x0]
 ; -O1:    subs x8, x4, x2
 ; -O1:    caspal x4, x5, x8, x9, [x0]
 ; -O1:    cmp x5, x7
 ; -O1:    ccmp x4, x6, #0, eq
     %r = atomicrmw sub ptr %ptr, i128 %value seq_cst, align 16

 ; -O1:    sub x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_acquire:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_acquire:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value acquire, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_release:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_release:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value release, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_acq_rel:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_acq_rel:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value acq_rel, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_seq_cst:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_seq_cst:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value seq_cst, align 1
     ret i128 %r

llvm/test/CodeGen/AArch64/Atomics/aarch64-atomicrmw-v8a.ll

 ; -O1:    add x9, x0, x1
 ; -O1:    stlxr w10, x9, [x8]
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_monotonic:
-; -O0:    adds x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_monotonic:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acquire:
-; -O0:    adds x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acquire:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_release:
-; -O0:    adds x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_release:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_acq_rel:
-; -O0:    adds x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_acq_rel:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_aligned_seq_cst:
-; -O0:    adds x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    adds x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_add_i128_aligned_seq_cst:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    adds x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 16
     ret i128 %r
 }

 ; -O1:    add x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_monotonic:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acquire:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acquire:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acquire, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_release:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_release:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value release, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_acq_rel:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value acq_rel, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_add_i128_unaligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
-; -O0:    adds x8, x11, x8
+; -O0:    adds x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_add_i128_unaligned_seq_cst:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    adds x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw add ptr %ptr, i128 %value seq_cst, align 1
     ret i128 %r

 ; -O1:    sub x9, x0, x1
 ; -O1:    stlxr w10, x9, [x8]
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 8
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_monotonic:
-; -O0:    subs x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_monotonic:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acquire:
-; -O0:    subs x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stxp w8, x14, x15, [x9]
-; -O0:    stxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stxp w8, x14, x15, [x11]
+; -O0:    stxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acquire:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value acquire, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_release(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_release:
-; -O0:    subs x14, x11, x10
-; -O0:    ldxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_release:
 ; -O1:    ldxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value release, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_acq_rel(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
-; -O0:    subs x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_acq_rel:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value acq_rel, align 16
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_aligned_seq_cst(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
-; -O0:    subs x14, x11, x10
-; -O0:    ldaxp x10, x12, [x9]
-; -O0:    cmp x10, x11
-; -O0:    cmp x12, x13
-; -O0:    stlxp w8, x14, x15, [x9]
-; -O0:    stlxp w8, x10, x12, [x9]
-; -O0:    subs x12, x12, x13
-; -O0:    ccmp x10, x11, #0, eq
+; -O0:    subs x14, x8, x10
+; -O0:    and w10, w8, #0x1
+; -O0:    adds w10, w10, w12
+; -O0:    ldaxp x10, x9, [x11]
+; -O0:    cmp x10, x12
+; -O0:    cmp x9, x13
+; -O0:    stlxp w8, x14, x15, [x11]
+; -O0:    stlxp w8, x10, x9, [x11]
+; -O0:    eor x8, x10, x8
+; -O0:    eor x11, x9, x11
+; -O0:    orr x8, x8, x11
+; -O0:    subs x8, x8, #0
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_aligned_seq_cst:
 ; -O1:    ldaxp x0, x1, [x8]
 ; -O1:    subs x9, x0, x2
 ; -O1:    stlxp w11, x9, x10, [x8]
     %r = atomicrmw sub ptr %ptr, i128 %value seq_cst, align 16
     ret i128 %r
 }

 ; -O1:    sub x8, x0, x19
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i64 %value seq_cst, align 1
     ret i64 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_monotonic(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
-; -O0:    subs x8, x11, x8
+; -O0:    subs x9, x8, x9
+; -O0:    and w11, w8, #0x1
+; -O0:    adds w11, w11, w12
 ; -O0:    bl __atomic_compare_exchange
 ;
 ; -O1-LABEL: atomicrmw_sub_i128_unaligned_monotonic:
 ; -O1:    ldp x0, x1, [x0]
 ; -O1:    subs x8, x0, x20
 ; -O1:    bl __atomic_compare_exchange
     %r = atomicrmw sub ptr %ptr, i128 %value monotonic, align 1
     ret i128 %r
 }
 
 define dso_local i128 @atomicrmw_sub_i128_unaligned_acquire(ptr %ptr, i128 %value) {
 ; -O0-LABEL: atomicrmw_sub_i128_unaligned_acquire: