[AArch64] Add `removeDeadFlags` DAG combine

[Kmeakin]

Abandoned in favour of https://reviews.llvm.org/D124464

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

static SDValue foldOverflowCheck(SDNode *Op, SelectionDAG &DAG) {
 
   SDVTList VTs = SetsFlags
                      ? DAG.getVTList(Op->getValueType(0), Op->getValueType(1))
                      : DAG.getVTList(Op->getValueType(0));
 
   return DAG.getNode(Opcode, SDLoc(Op), VTs, OpLhs, OpRhs, CsetCarry);
 }
 
+// ADDS => ADD
+// SUBS => SUB
+// ADCS => ADC
+// SBSC => SBC
+// if flag is unused
+static SDValue removeDeadFlags(SDNode *Op, SelectionDAG &DAG) {
+  unsigned Opcode;
+  bool CarryIn;
+  switch (Op->getOpcode()) {
+  default:
+    return SDValue();
+  case AArch64ISD::ADDS:
+    Opcode = ISD::ADD;
+    CarryIn = false;
+    break;
+  case AArch64ISD::SUBS:
+    Opcode = ISD::SUB;
+    CarryIn = false;
+    break;
+  case AArch64ISD::ADCS:
+    Opcode = AArch64ISD::ADC;
+    CarryIn = true;
+    break;
+  case AArch64ISD::SBCS:
+    Opcode = AArch64ISD::SBC;
+    CarryIn = true;
+    break;
+  }
+
+  if (Op->hasAnyUseOfValue(1))
+    return SDValue();
+
+  SDLoc DL(Op);
+  EVT VT0 = Op->getValueType(0);
+  EVT VT1 = Op->getValueType(1);
+  SDVTList VTs = DAG.getVTList(VT0, VT1);
+
+  SDValue OpLhs = Op->getOperand(0);
+  SDValue OpRhs = Op->getOperand(1);
+
+  // SelectionDAG combines must return an op that outputs the same number
+  // of values as the original. This means we cannot just return `FlaglessOp`.
+  // Instead we return both `FlaglessOp` and `Flag` and then `Flag` is
+  // removed by SelectionDAG's dead code elimination
+  SDValue FlaglessOp =
+      CarryIn ? DAG.getNode(Opcode, DL, VT0, OpLhs, OpRhs, Op->getOperand(2))
+              : DAG.getNode(Opcode, DL, VT0, OpLhs, OpRhs);
+  SDValue Flag = DAG.getNode(ISD::UNDEF, DL, VT0);
+
+  return DAG.getNode(ISD::MERGE_VALUES, DL, VTs, FlaglessOp, Flag);
+}
+
 static SDValue performAddSubCombine(SDNode *N,
                                     TargetLowering::DAGCombinerInfo &DCI,
                                     SelectionDAG &DAG) {
   if (SDValue Val = foldOverflowCheck(N, DAG))
     return Val;
+  if (SDValue Val = removeDeadFlags(N, DAG))
+    return Val;
   // Try to change sum of two reductions.
   if (SDValue Val = performAddUADDVCombine(N, DAG))
     return Val;
   if (SDValue Val = performAddDotCombine(N, DAG))
     return Val;
   if (SDValue Val = performAddCSelIntoCSinc(N, DAG))
     return Val;
   if (SDValue Val = performNegCSelCombine(N, DAG))

llvm/test/CodeGen/AArch64/adc.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -verify-machineinstrs < %s -mtriple=arm64-apple-ios7.0 | FileCheck --check-prefix=CHECK-LE %s
 ; RUN: llc -verify-machineinstrs < %s -mtriple=aarch64_be-none-linux-gnu | FileCheck --check-prefix=CHECK-BE %s
 
 define i128 @test_simple(i128 %a, i128 %b, i128 %c) {
 ; CHECK-LE-LABEL: test_simple:
 ; CHECK-LE:       ; %bb.0:
 ; CHECK-LE-NEXT:    adds x8, x0, x2
-; CHECK-LE-NEXT:    adcs x9, x1, x3
+; CHECK-LE-NEXT:    adc x9, x1, x3
 ; CHECK-LE-NEXT:    subs x0, x8, x4
-; CHECK-LE-NEXT:    sbcs x1, x9, x5
+; CHECK-LE-NEXT:    sbc x1, x9, x5
 ; CHECK-LE-NEXT:    ret
 ;
 ; CHECK-BE-LABEL: test_simple:
 ; CHECK-BE:       // %bb.0:
 ; CHECK-BE-NEXT:    adds x8, x1, x3
-; CHECK-BE-NEXT:    adcs x9, x0, x2
+; CHECK-BE-NEXT:    adc x9, x0, x2
 ; CHECK-BE-NEXT:    subs x1, x8, x5
-; CHECK-BE-NEXT:    sbcs x0, x9, x4
+; CHECK-BE-NEXT:    sbc x0, x9, x4
 ; CHECK-BE-NEXT:    ret
 
   %valadd = add i128 %a, %b
 
   %valsub = sub i128 %valadd, %c
 
   ret i128 %valsub
 }
 
 define i128 @test_imm(i128 %a) {
 ; CHECK-LE-LABEL: test_imm:
 ; CHECK-LE:       ; %bb.0:
 ; CHECK-LE-NEXT:    adds x0, x0, #12
-; CHECK-LE-NEXT:    adcs x1, x1, xzr
+; CHECK-LE-NEXT:    adc x1, x1, xzr
 ; CHECK-LE-NEXT:    ret
 ;
 ; CHECK-BE-LABEL: test_imm:
 ; CHECK-BE:       // %bb.0:
 ; CHECK-BE-NEXT:    adds x1, x1, #12
-; CHECK-BE-NEXT:    adcs x0, x0, xzr
+; CHECK-BE-NEXT:    adc x0, x0, xzr
 ; CHECK-BE-NEXT:    ret
 
   %val = add i128 %a, 12
 
   ret i128 %val
 }
 
 define i128 @test_shifted(i128 %a, i128 %b) {
 ; CHECK-LE-LABEL: test_shifted:
 ; CHECK-LE:       ; %bb.0:
 ; CHECK-LE-NEXT:    extr x8, x3, x2, #19
 ; CHECK-LE-NEXT:    adds x0, x0, x2, lsl #45
-; CHECK-LE-NEXT:    adcs x1, x1, x8
+; CHECK-LE-NEXT:    adc x1, x1, x8
 ; CHECK-LE-NEXT:    ret
 ;
 ; CHECK-BE-LABEL: test_shifted:
 ; CHECK-BE:       // %bb.0:
 ; CHECK-BE-NEXT:    extr x8, x2, x3, #19
 ; CHECK-BE-NEXT:    adds x1, x1, x3, lsl #45
-; CHECK-BE-NEXT:    adcs x0, x0, x8
+; CHECK-BE-NEXT:    adc x0, x0, x8
 ; CHECK-BE-NEXT:    ret
 
   %rhs = shl i128 %b, 45
 
   %val = add i128 %a, %rhs
 
   ret i128 %val
 }
 
 define i128 @test_extended(i128 %a, i16 %b) {
 ; CHECK-LE-LABEL: test_extended:
 ; CHECK-LE:       ; %bb.0:
 ; CHECK-LE-NEXT:    ; kill: def $w2 killed $w2 def $x2
 ; CHECK-LE-NEXT:    sxth x8, w2
 ; CHECK-LE-NEXT:    adds x0, x0, w2, sxth #3
 ; CHECK-LE-NEXT:    asr x9, x8, #63
 ; CHECK-LE-NEXT:    extr x8, x9, x8, #61
-; CHECK-LE-NEXT:    adcs x1, x1, x8
+; CHECK-LE-NEXT:    adc x1, x1, x8
 ; CHECK-LE-NEXT:    ret
 ;
 ; CHECK-BE-LABEL: test_extended:
 ; CHECK-BE:       // %bb.0:
 ; CHECK-BE-NEXT:    // kill: def $w2 killed $w2 def $x2
 ; CHECK-BE-NEXT:    sxth x8, w2
 ; CHECK-BE-NEXT:    adds x1, x1, w2, sxth #3
 ; CHECK-BE-NEXT:    asr x9, x8, #63
 ; CHECK-BE-NEXT:    extr x8, x9, x8, #61
-; CHECK-BE-NEXT:    adcs x0, x0, x8
+; CHECK-BE-NEXT:    adc x0, x0, x8
 ; CHECK-BE-NEXT:    ret
 
   %ext = sext i16 %b to i128
   %rhs = shl i128 %ext, 3
 
   %val = add i128 %a, %rhs
 
   ret i128 %val
 }

llvm/test/CodeGen/AArch64/addcarry-crash.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s | FileCheck %s
 target triple = "arm64-apple-ios7.0"
 
 define i64 @foo(i64* nocapture readonly %ptr, i64 %a, i64 %b, i64 %c) local_unnamed_addr #0 {
 ; CHECK-LABEL: foo:
 ; CHECK:       ; %bb.0: ; %entry
 ; CHECK-NEXT:    ldr w8, [x0, #4]
 ; CHECK-NEXT:    lsr x9, x1, #32
 ; CHECK-NEXT:    cmn x3, x2
 ; CHECK-NEXT:    mul x8, x8, x9
-; CHECK-NEXT:    adcs x0, x8, xzr
+; CHECK-NEXT:    adc x0, x8, xzr
 ; CHECK-NEXT:    ret
 entry:
   %0 = lshr i64 %a, 32
   %1 = load i64, i64* %ptr, align 8
   %2 = lshr i64 %1, 32
   %3 = mul nuw i64 %2, %0
   %4 = add i64 %c, %b
   %5 = icmp ult i64 %4, %c
   %6 = zext i1 %5 to i64
   %7 = add i64 %3, %6
   ret i64 %7
 }
 
 attributes #0 = { norecurse nounwind readonly }

llvm/test/CodeGen/AArch64/arm64-atomic-128.ll

 
 define void @fetch_and_add(i128* %p, i128 %bits) {
 ; CHECK-LABEL: fetch_and_add:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:  .LBB6_1: // %atomicrmw.start
 ; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    ldaxp x9, x8, [x0]
 ; CHECK-NEXT:    adds x10, x9, x2
-; CHECK-NEXT:    adcs x11, x8, x3
+; CHECK-NEXT:    adc x11, x8, x3
 ; CHECK-NEXT:    stlxp w12, x10, x11, [x0]
 ; CHECK-NEXT:    cbnz w12, .LBB6_1
 ; CHECK-NEXT:  // %bb.2: // %atomicrmw.end
 ; CHECK-NEXT:    adrp x10, :got:var
 ; CHECK-NEXT:    ldr x10, [x10, :got_lo12:var]
 ; CHECK-NEXT:    stp x9, x8, [x10]
 ; CHECK-NEXT:    ret
 
   %val = atomicrmw add i128* %p, i128 %bits seq_cst
   store i128 %val, i128* @var, align 16
   ret void
 }
 
 define void @fetch_and_sub(i128* %p, i128 %bits) {
 ; CHECK-LABEL: fetch_and_sub:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:  .LBB7_1: // %atomicrmw.start
 ; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    ldaxp x9, x8, [x0]
 ; CHECK-NEXT:    subs x10, x9, x2
-; CHECK-NEXT:    sbcs x11, x8, x3
+; CHECK-NEXT:    sbc x11, x8, x3
 ; CHECK-NEXT:    stlxp w12, x10, x11, [x0]
 ; CHECK-NEXT:    cbnz w12, .LBB7_1
 ; CHECK-NEXT:  // %bb.2: // %atomicrmw.end
 ; CHECK-NEXT:    adrp x10, :got:var
 ; CHECK-NEXT:    ldr x10, [x10, :got_lo12:var]
 ; CHECK-NEXT:    stp x9, x8, [x10]
 ; CHECK-NEXT:    ret
 

llvm/test/CodeGen/AArch64/arm64-vabs.ll

; GISEL-NEXT:    ret
   %absel = select <2 x i1> %abcmp, <2 x i64> %ababs, <2 x i64> %abdiff
   ret <2 x i64> %absel
 }
 
 
 define <2 x i128> @uabd_i64(<2 x i64> %a, <2 x i64> %b) {
 ; CHECK-LABEL: uabd_i64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    fmov x9, d0
-; CHECK-NEXT:    fmov x11, d1
 ; CHECK-NEXT:    mov.d x8, v0[1]
+; CHECK-NEXT:    fmov x9, d0
 ; CHECK-NEXT:    mov.d x10, v1[1]
+; CHECK-NEXT:    fmov x11, d1
 ; CHECK-NEXT:    asr x12, x9, #63
 ; CHECK-NEXT:    asr x13, x11, #63
 ; CHECK-NEXT:    subs x9, x9, x11
-; CHECK-NEXT:    sbcs x11, x12, x13
+; CHECK-NEXT:    sbc x11, x12, x13
 ; CHECK-NEXT:    asr x12, x8, #63
 ; CHECK-NEXT:    asr x13, x10, #63
 ; CHECK-NEXT:    subs x8, x8, x10
-; CHECK-NEXT:    sbcs x10, x12, x13
-; CHECK-NEXT:    asr x13, x11, #63
-; CHECK-NEXT:    asr x12, x10, #63
-; CHECK-NEXT:    eor x8, x8, x12
-; CHECK-NEXT:    eor x10, x10, x12
-; CHECK-NEXT:    subs x2, x8, x12
-; CHECK-NEXT:    eor x8, x9, x13
-; CHECK-NEXT:    sbcs x3, x10, x12
-; CHECK-NEXT:    eor x9, x11, x13
-; CHECK-NEXT:    subs x8, x8, x13
-; CHECK-NEXT:    sbcs x1, x9, x13
+; CHECK-NEXT:    sbc x10, x12, x13
+; CHECK-NEXT:    asr x12, x11, #63
+; CHECK-NEXT:    asr x13, x10, #63
+; CHECK-NEXT:    eor x9, x9, x12
+; CHECK-NEXT:    eor x8, x8, x13
+; CHECK-NEXT:    eor x10, x10, x13
+; CHECK-NEXT:    subs x2, x8, x13
+; CHECK-NEXT:    sbc x3, x10, x13
+; CHECK-NEXT:    subs x8, x9, x12
+; CHECK-NEXT:    eor x9, x11, x12
+; CHECK-NEXT:    sbc x1, x9, x12
 ; CHECK-NEXT:    fmov d0, x8
 ; CHECK-NEXT:    mov.d v0[1], x1
 ; CHECK-NEXT:    fmov x0, d0
 ; CHECK-NEXT:    ret
   %aext = sext <2 x i64> %a to <2 x i128>
   %bext = sext <2 x i64> %b to <2 x i128>
   %abdiff = sub nsw <2 x i128> %aext, %bext
   %abcmp = icmp slt <2 x i128> %abdiff, zeroinitializer
   %ababs = sub nsw <2 x i128> zeroinitializer, %abdiff
   %absel = select <2 x i1> %abcmp, <2 x i128> %ababs, <2 x i128> %abdiff
   ret <2 x i128> %absel
 }

llvm/test/CodeGen/AArch64/atomicrmw-O0.ll

 ; NOLSE-NEXT:  .LBB4_1: // %atomicrmw.start
 ; NOLSE-NEXT:    // =>This Loop Header: Depth=1
 ; NOLSE-NEXT:    // Child Loop BB4_2 Depth 2
 ; NOLSE-NEXT:    ldr x11, [sp, #40] // 8-byte Folded Reload
 ; NOLSE-NEXT:    ldr x8, [sp, #32] // 8-byte Folded Reload
 ; NOLSE-NEXT:    ldr x13, [sp, #24] // 8-byte Folded Reload
 ; NOLSE-NEXT:    adds x14, x8, #1
 ; NOLSE-NEXT:    mov x9, xzr
-; NOLSE-NEXT:    adcs x15, x11, x9
+; NOLSE-NEXT:    adc x15, x11, x9
 ; NOLSE-NEXT:  .LBB4_2: // %atomicrmw.start
 ; NOLSE-NEXT:    // Parent Loop BB4_1 Depth=1
 ; NOLSE-NEXT:    // => This Inner Loop Header: Depth=2
 ; NOLSE-NEXT:    ldaxp x10, x9, [x13]
 ; NOLSE-NEXT:    cmp x10, x8
 ; NOLSE-NEXT:    cset w12, ne
 ; NOLSE-NEXT:    cmp x9, x11
 ; NOLSE-NEXT:    cinc w12, w12, ne

 ; LSE-NEXT:    b .LBB4_1
 ; LSE-NEXT:  .LBB4_1: // %atomicrmw.start
 ; LSE-NEXT:    // =>This Inner Loop Header: Depth=1
 ; LSE-NEXT:    ldr x10, [sp, #72] // 8-byte Folded Reload
 ; LSE-NEXT:    ldr x8, [sp, #64] // 8-byte Folded Reload
 ; LSE-NEXT:    ldr x9, [sp, #56] // 8-byte Folded Reload
 ; LSE-NEXT:    adds x2, x8, #1
 ; LSE-NEXT:    mov x11, xzr
-; LSE-NEXT:    adcs x11, x10, x11
+; LSE-NEXT:    adc x11, x10, x11
 ; LSE-NEXT:    // kill: def $x2 killed $x2 def $x2_x3
 ; LSE-NEXT:    mov x3, x11
 ; LSE-NEXT:    mov x0, x8
 ; LSE-NEXT:    mov x1, x10
 ; LSE-NEXT:    stp x0, x1, [sp, #8] // 16-byte Folded Spill
 ; LSE-NEXT:    caspal x0, x1, x2, x3, [x9]
 ; LSE-NEXT:    stp x0, x1, [sp, #24] // 16-byte Folded Spill
 ; LSE-NEXT:    mov x9, x1

llvm/test/CodeGen/AArch64/i128-math.ll

 
 declare { i128, i1 } @llvm.smul.with.overflow.i128(i128, i128)
 declare   i128       @llvm.smul.sat.i128(i128, i128)
 
 define i128 @u128_add(i128 %x, i128 %y) {
 ; CHECK-LABEL: u128_add:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    adds x0, x0, x2
-; CHECK-NEXT:    adcs x1, x1, x3
+; CHECK-NEXT:    adc x1, x1, x3
 ; CHECK-NEXT:    ret
   %1 = add i128 %x, %y
   ret i128 %1
 }
 
 define { i128, i8 } @u128_checked_add(i128 %x, i128 %y) {
 ; CHECK-LABEL: u128_checked_add:
 ; CHECK:       // %bb.0:
; CHECK-NEXT:    ret
   %1 = tail call i128 @llvm.uadd.sat.i128(i128 %x, i128 %y)
   ret i128 %1
 }
 
 define i128 @u128_sub(i128 %x, i128 %y) {
 ; CHECK-LABEL: u128_sub:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    subs x0, x0, x2
-; CHECK-NEXT:    sbcs x1, x1, x3
+; CHECK-NEXT:    sbc x1, x1, x3
 ; CHECK-NEXT:    ret
   %1 = sub i128 %x, %y
   ret i128 %1
 }
 
 define { i128, i8 } @u128_checked_sub(i128 %x, i128 %y) {
 ; CHECK-LABEL: u128_checked_sub:
 ; CHECK:       // %bb.0:
; CHECK-NEXT:    ret
   %1 = tail call i128 @llvm.usub.sat.i128(i128 %x, i128 %y)
   ret i128 %1
 }
 
 define i128 @i128_add(i128 %x, i128 %y) {
 ; CHECK-LABEL: i128_add:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    adds x0, x0, x2
-; CHECK-NEXT:    adcs x1, x1, x3
+; CHECK-NEXT:    adc x1, x1, x3
 ; CHECK-NEXT:    ret
   %1 = add i128 %x, %y
   ret i128 %1
 }
 
 define { i128, i8 } @i128_checked_add(i128 %x, i128 %y) {
 ; CHECK-LABEL: i128_checked_add:
 ; CHECK:       // %bb.0:
; CHECK-NEXT:    ret
   %1 = tail call i128 @llvm.sadd.sat.i128(i128 %x, i128 %y)
   ret i128 %1
 }
 
 define i128 @i128_sub(i128 %x, i128 %y) {
 ; CHECK-LABEL: i128_sub:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    subs x0, x0, x2
-; CHECK-NEXT:    sbcs x1, x1, x3
+; CHECK-NEXT:    sbc x1, x1, x3
 ; CHECK-NEXT:    ret
   %1 = sub i128 %x, %y
   ret i128 %1
 }
 
 define { i128, i8 } @i128_checked_sub(i128 %x, i128 %y) {
 ; CHECK-LABEL: i128_checked_sub:
 ; CHECK:       // %bb.0:

llvm/test/CodeGen/AArch64/icmp-shift-opt.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=arm64-eabi | FileCheck %s
 
 ; Optimize expanded SRL/SHL used as an input of
 ; SETCC comparing it with zero by removing rotation.
 ;
 ; See https://bugs.llvm.org/show_bug.cgi?id=50197
 define i128 @opt_setcc_lt_power_of_2(i128 %a) nounwind {
 ; CHECK-LABEL: opt_setcc_lt_power_of_2:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:  .LBB0_1: // %loop
 ; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    adds x0, x0, #1
-; CHECK-NEXT:    adcs x1, x1, xzr
+; CHECK-NEXT:    adc x1, x1, xzr
 ; CHECK-NEXT:    orr x8, x1, x0, lsr #60
 ; CHECK-NEXT:    cbnz x8, .LBB0_1
 ; CHECK-NEXT:  // %bb.2: // %exit
 ; CHECK-NEXT:    ret
   br label %loop
 
 loop:
   %phi.a = phi i128 [ %a, %0 ], [ %inc, %loop ]

llvm/test/CodeGen/AArch64/neg-abs.ll

 
 define i128 @neg_abs128(i128 %x) {
 ; CHECK-LABEL: neg_abs128:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    asr x8, x1, #63
 ; CHECK-NEXT:    eor x9, x0, x8
 ; CHECK-NEXT:    eor x10, x1, x8
 ; CHECK-NEXT:    subs x0, x8, x9
-; CHECK-NEXT:    sbcs x1, x8, x10
+; CHECK-NEXT:    sbc x1, x8, x10
 ; CHECK-NEXT:    ret
   %abs = tail call i128 @llvm.abs.i128(i128 %x, i1 true)
   %neg = sub nsw i128 0, %abs
   ret i128 %neg
 }
 
 
 

 
 define i128 @abs128(i128 %x) {
 ; CHECK-LABEL: abs128:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    asr x8, x1, #63
 ; CHECK-NEXT:    eor x9, x0, x8
 ; CHECK-NEXT:    eor x10, x1, x8
 ; CHECK-NEXT:    subs x0, x9, x8
-; CHECK-NEXT:    sbcs x1, x10, x8
+; CHECK-NEXT:    sbc x1, x10, x8
 ; CHECK-NEXT:    ret
   %abs = tail call i128 @llvm.abs.i128(i128 %x, i1 true)
   ret i128 %abs
 }
 

llvm/test/CodeGen/AArch64/neon-abd.ll

 }
 
 define <2 x i64> @sabd_2d(<2 x i64> %a, <2 x i64> %b) #0 {
 ; CHECK-LABEL: sabd_2d:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov x8, v0.d[1]
 ; CHECK-NEXT:    fmov x10, d0
 ; CHECK-NEXT:    mov x9, v1.d[1]
-; CHECK-NEXT:    asr x11, x10, #63
-; CHECK-NEXT:    asr x12, x8, #63
-; CHECK-NEXT:    asr x13, x9, #63
+; CHECK-NEXT:    fmov x11, d1
+; CHECK-NEXT:    asr x12, x10, #63
+; CHECK-NEXT:    asr x14, x8, #63
+; CHECK-NEXT:    asr x15, x9, #63
 ; CHECK-NEXT:    subs x8, x8, x9
-; CHECK-NEXT:    fmov x9, d1
-; CHECK-NEXT:    sbcs x12, x12, x13
-; CHECK-NEXT:    asr x13, x9, #63
-; CHECK-NEXT:    subs x9, x10, x9
-; CHECK-NEXT:    sbcs x10, x11, x13
-; CHECK-NEXT:    asr x11, x12, #63
-; CHECK-NEXT:    asr x10, x10, #63
-; CHECK-NEXT:    eor x8, x8, x11
-; CHECK-NEXT:    eor x9, x9, x10
-; CHECK-NEXT:    sub x8, x8, x11
-; CHECK-NEXT:    sub x9, x9, x10
+; CHECK-NEXT:    asr x13, x11, #63
+; CHECK-NEXT:    sbc x9, x14, x15
+; CHECK-NEXT:    subs x10, x10, x11
+; CHECK-NEXT:    asr x9, x9, #63
+; CHECK-NEXT:    sbc x11, x12, x13
+; CHECK-NEXT:    eor x8, x8, x9
+; CHECK-NEXT:    asr x11, x11, #63
+; CHECK-NEXT:    sub x8, x8, x9
+; CHECK-NEXT:    eor x10, x10, x11
+; CHECK-NEXT:    sub x10, x10, x11
 ; CHECK-NEXT:    fmov d1, x8
-; CHECK-NEXT:    fmov d0, x9
+; CHECK-NEXT:    fmov d0, x10
 ; CHECK-NEXT:    mov v0.d[1], v1.d[0]
 ; CHECK-NEXT:    ret
   %a.sext = sext <2 x i64> %a to <2 x i128>
   %b.sext = sext <2 x i64> %b to <2 x i128>
   %sub = sub <2 x i128> %a.sext, %b.sext
   %abs = call <2 x i128> @llvm.abs.v2i128(<2 x i128> %sub, i1 true)
   %trunc = trunc <2 x i128> %abs to <2 x i64>
   ret <2 x i64> %trunc

 }
 
 define <2 x i64> @uabd_2d(<2 x i64> %a, <2 x i64> %b) #0 {
 ; CHECK-LABEL: uabd_2d:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov x8, v0.d[1]
 ; CHECK-NEXT:    fmov x10, d0
 ; CHECK-NEXT:    mov x9, v1.d[1]
+; CHECK-NEXT:    fmov x11, d1
 ; CHECK-NEXT:    subs x8, x8, x9
-; CHECK-NEXT:    fmov x9, d1
-; CHECK-NEXT:    ngcs x11, xzr
+; CHECK-NEXT:    ngc x9, xzr
+; CHECK-NEXT:    subs x10, x10, x11
+; CHECK-NEXT:    ngc x11, xzr
+; CHECK-NEXT:    asr x9, x9, #63
 ; CHECK-NEXT:    asr x11, x11, #63
-; CHECK-NEXT:    subs x9, x10, x9
-; CHECK-NEXT:    eor x8, x8, x11
-; CHECK-NEXT:    ngcs x10, xzr
-; CHECK-NEXT:    sub x8, x8, x11
-; CHECK-NEXT:    asr x10, x10, #63
-; CHECK-NEXT:    eor x9, x9, x10
-; CHECK-NEXT:    sub x9, x9, x10
+; CHECK-NEXT:    eor x8, x8, x9
+; CHECK-NEXT:    eor x10, x10, x11
+; CHECK-NEXT:    sub x8, x8, x9
+; CHECK-NEXT:    sub x10, x10, x11
 ; CHECK-NEXT:    fmov d1, x8
-; CHECK-NEXT:    fmov d0, x9
+; CHECK-NEXT:    fmov d0, x10
 ; CHECK-NEXT:    mov v0.d[1], v1.d[0]
 ; CHECK-NEXT:    ret
   %a.zext = zext <2 x i64> %a to <2 x i128>
   %b.zext = zext <2 x i64> %b to <2 x i128>
   %sub = sub <2 x i128> %a.zext, %b.zext
   %abs = call <2 x i128> @llvm.abs.v2i128(<2 x i128> %sub, i1 true)
   %trunc = trunc <2 x i128> %abs to <2 x i64>
   ret <2 x i64> %trunc

llvm/test/CodeGen/AArch64/nzcv-save.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -verify-machineinstrs -mtriple=aarch64-eabi | FileCheck %s
 
 ; DAG ends up with two uses for the flags from an ADCS node, which means they
 ; must be saved for later.
 define void @f(i256* nocapture %a, i256* nocapture %b, i256* nocapture %cc, i256* nocapture %dd) nounwind uwtable noinline ssp {
 ; CHECK-LABEL: f:
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    ldp x9, x8, [x2]
 ; CHECK-NEXT:    ldp x11, x10, [x3]
+; CHECK-NEXT:    ldp x13, x12, [x2, #16]
+; CHECK-NEXT:    ldp x14, x15, [x3, #16]
 ; CHECK-NEXT:    adds x9, x9, x11
-; CHECK-NEXT:    ldp x12, x11, [x2, #16]
 ; CHECK-NEXT:    adcs x8, x8, x10
-; CHECK-NEXT:    ldp x13, x10, [x3, #16]
-; CHECK-NEXT:    adcs x12, x12, x13
-; CHECK-NEXT:    mrs x13, NZCV
-; CHECK-NEXT:    adcs x14, x11, x10
-; CHECK-NEXT:    orr x11, x11, #0x100
-; CHECK-NEXT:    msr NZCV, x13
+; CHECK-NEXT:    adcs x10, x13, x14
+; CHECK-NEXT:    adc x11, x12, x15
+; CHECK-NEXT:    orr x12, x12, #0x100
+; CHECK-NEXT:    adc x12, x12, x15
 ; CHECK-NEXT:    stp x9, x8, [x0]
-; CHECK-NEXT:    adcs x10, x11, x10
-; CHECK-NEXT:    stp x12, x14, [x0, #16]
+; CHECK-NEXT:    stp x10, x11, [x0, #16]
+; CHECK-NEXT:    stp x10, x12, [x1, #16]
 ; CHECK-NEXT:    stp x9, x8, [x1]
-; CHECK-NEXT:    stp x12, x10, [x1, #16]
 ; CHECK-NEXT:    ret
 entry:
   %c = load i256, i256* %cc
   %d = load i256, i256* %dd
   %add = add nsw i256 %c, %d
   store i256 %add, i256* %a, align 8
   %or = or i256 %c, 1606938044258990275541962092341162602522202993782792835301376
   %add6 = add nsw i256 %or, %d
   store i256 %add6, i256* %b, align 8
   ret void
 }

llvm/test/CodeGen/AArch64/vecreduce-add-legalization.ll

; CHECK-NEXT:    ret
   %b = call i24 @llvm.vector.reduce.add.v4i24(<4 x i24> %a)
   ret i24 %b
 }
 
 define i128 @test_v2i128(<2 x i128> %a) nounwind {
 ; CHECK-LABEL: test_v2i128:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    adds x0, x0, x2
-; CHECK-NEXT:    adcs x1, x1, x3
+; CHECK-NEXT:    adc x1, x1, x3
 ; CHECK-NEXT:    ret
   %b = call i128 @llvm.vector.reduce.add.v2i128(<2 x i128> %a)
   ret i128 %b
 }
 
 define i32 @test_v16i32(<16 x i32> %a) nounwind {
 ; CHECK-LABEL: test_v16i32:
 ; CHECK:       // %bb.0: