; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=arm64-eabi -mcpu=cyclone | FileCheck %s

define i32 @qadds(<4 x i32> %b, <4 x i32> %c) nounwind readnone optsize ssp {
; CHECK-LABEL: qadds:
; CHECK:       // %bb.0:
; CHECK-NEXT:    sqadd s0, s0, s1
; CHECK-NEXT:    fmov w0, s0
; CHECK-NEXT:    ret
  %vecext = extractelement <4 x i32> %b, i32 0
  %vecext1 = extractelement <4 x i32> %c, i32 0
  %vqadd.i = tail call i32 @llvm.aarch64.neon.sqadd.i32(i32 %vecext, i32 %vecext1) nounwind
  ret i32 %vqadd.i
}

define i64 @qaddd(<2 x i64> %b, <2 x i64> %c) nounwind readnone optsize ssp {
; CHECK-LABEL: qaddd:
; CHECK:       // %bb.0:
; CHECK-NEXT:    sqadd d0, d0, d1
; CHECK-NEXT:    fmov x0, d0
; CHECK-NEXT:    ret
  %vecext = extractelement <2 x i64> %b, i32 0
  %vecext1 = extractelement <2 x i64> %c, i32 0
  %vqadd.i = tail call i64 @llvm.aarch64.neon.sqadd.i64(i64 %vecext, i64 %vecext1) nounwind
  ret i64 %vqadd.i
}

define i32 @uqadds(<4 x i32> %b, <4 x i32> %c) nounwind readnone optsize ssp {
; CHECK-LABEL: uqadds:
; CHECK:       // %bb.0:
; CHECK-NEXT:    uqadd s0, s0, s1
; CHECK-NEXT:    fmov w0, s0
; CHECK-NEXT:    ret
  %vecext = extractelement <4 x i32> %b, i32 0
  %vecext1 = extractelement <4 x i32> %c, i32 0
  %vqadd.i = tail call i32 @llvm.aarch64.neon.uqadd.i32(i32 %vecext, i32 %vecext1) nounwind
  ret i32 %vqadd.i
}

define i64 @uqaddd(<2 x i64> %b, <2 x i64> %c) nounwind readnone optsize ssp {
; CHECK-LABEL: uqaddd:
; CHECK:       // %bb.0:
; CHECK-NEXT:    uqadd d0, d0, d1
; CHECK-NEXT:    fmov x0, d0
; CHECK-NEXT:    ret
  %vecext = extractelement <2 x i64> %b, i32 0
  %vecext1 = extractelement <2 x i64> %c, i32 0
  %vqadd.i = tail call i64 @llvm.aarch64.neon.uqadd.i64(i64 %vecext, i64 %vecext1) nounwind
  ret i64 %vqadd.i
}

declare i64 @llvm.aarch64.neon.uqadd.i64(i64, i64) nounwind readnone
declare i32 @llvm.aarch64.neon.uqadd.i32(i32, i32) nounwind readnone
declare i64 @llvm.aarch64.neon.sqadd.i64(i64, i64) nounwind readnone
declare i32 @llvm.aarch64.neon.sqadd.i32(i32, i32) nounwind readnone

define i32 @qsubs(<4 x i32> %b, <4 x i32> %c) nounwind readnone optsize ssp {
; CHECK-LABEL: qsubs:
; CHECK:       // %bb.0:
; CHECK-NEXT:    sqsub s0, s0, s1
; CHECK-NEXT:    fmov w0, s0
; CHECK-NEXT:    ret
  %vecext = extractelement <4 x i32> %b, i32 0
  %vecext1 = extractelement <4 x i32> %c, i32 0
  %vqsub.i = tail call i32 @llvm.aarch64.neon.sqsub.i32(i32 %vecext, i32 %vecext1) nounwind
  ret i32 %vqsub.i
}

define i64 @qsubd(<2 x i64> %b, <2 x i64> %c) nounwind readnone optsize ssp {
; CHECK-LABEL: qsubd:
; CHECK:       // %bb.0:
; CHECK-NEXT:    sqsub d0, d0, d1
; CHECK-NEXT:    fmov x0, d0
; CHECK-NEXT:    ret
  %vecext = extractelement <2 x i64> %b, i32 0
  %vecext1 = extractelement <2 x i64> %c, i32 0
  %vqsub.i = tail call i64 @llvm.aarch64.neon.sqsub.i64(i64 %vecext, i64 %vecext1) nounwind
  ret i64 %vqsub.i
}

define i32 @uqsubs(<4 x i32> %b, <4 x i32> %c) nounwind readnone optsize ssp {
; CHECK-LABEL: uqsubs:
; CHECK:       // %bb.0:
; CHECK-NEXT:    uqsub s0, s0, s1
; CHECK-NEXT:    fmov w0, s0
; CHECK-NEXT:    ret
  %vecext = extractelement <4 x i32> %b, i32 0
  %vecext1 = extractelement <4 x i32> %c, i32 0
  %vqsub.i = tail call i32 @llvm.aarch64.neon.uqsub.i32(i32 %vecext, i32 %vecext1) nounwind
  ret i32 %vqsub.i
}

define i64 @uqsubd(<2 x i64> %b, <2 x i64> %c) nounwind readnone optsize ssp {
; CHECK-LABEL: uqsubd:
; CHECK:       // %bb.0:
; CHECK-NEXT:    uqsub d0, d0, d1
; CHECK-NEXT:    fmov x0, d0
; CHECK-NEXT:    ret
  %vecext = extractelement <2 x i64> %b, i32 0
  %vecext1 = extractelement <2 x i64> %c, i32 0
  %vqsub.i = tail call i64 @llvm.aarch64.neon.uqsub.i64(i64 %vecext, i64 %vecext1) nounwind
  ret i64 %vqsub.i
}

declare i64 @llvm.aarch64.neon.uqsub.i64(i64, i64) nounwind readnone
declare i32 @llvm.aarch64.neon.uqsub.i32(i32, i32) nounwind readnone
declare i64 @llvm.aarch64.neon.sqsub.i64(i64, i64) nounwind readnone
declare i32 @llvm.aarch64.neon.sqsub.i32(i32, i32) nounwind readnone

define i32 @qabss(<4 x i32> %b, <4 x i32> %c) nounwind readnone {
; CHECK-LABEL: qabss:
; CHECK:       // %bb.0:
; CHECK-NEXT:    sqabs s0, s0
; CHECK-NEXT:    fmov w0, s0
; CHECK-NEXT:    ret
  %vecext = extractelement <4 x i32> %b, i32 0
  %vqabs.i = tail call i32 @llvm.aarch64.neon.sqabs.i32(i32 %vecext) nounwind
  ret i32 %vqabs.i
}

define i64 @qabsd(<2 x i64> %b, <2 x i64> %c) nounwind readnone {
; CHECK-LABEL: qabsd:
; CHECK:       // %bb.0:
; CHECK-NEXT:    sqabs d0, d0
; CHECK-NEXT:    fmov x0, d0
; CHECK-NEXT:    ret
  %vecext = extractelement <2 x i64> %b, i32 0
  %vqabs.i = tail call i64 @llvm.aarch64.neon.sqabs.i64(i64 %vecext) nounwind
  ret i64 %vqabs.i
}

define i32 @qnegs(<4 x i32> %b, <4 x i32> %c) nounwind readnone {
; CHECK-LABEL: qnegs:
; CHECK:       // %bb.0:
; CHECK-NEXT:    sqneg s0, s0
; CHECK-NEXT:    fmov w0, s0
; CHECK-NEXT:    ret
  %vecext = extractelement <4 x i32> %b, i32 0
  %vqneg.i = tail call i32 @llvm.aarch64.neon.sqneg.i32(i32 %vecext) nounwind
  ret i32 %vqneg.i
}

define i64 @qnegd(<2 x i64> %b, <2 x i64> %c) nounwind readnone {
; CHECK-LABEL: qnegd:
; CHECK:       // %bb.0:
; CHECK-NEXT:    sqneg d0, d0
; CHECK-NEXT:    fmov x0, d0
; CHECK-NEXT:    ret
  %vecext = extractelement <2 x i64> %b, i32 0
  %vqneg.i = tail call i64 @llvm.aarch64.neon.sqneg.i64(i64 %vecext) nounwind
  ret i64 %vqneg.i
}

declare i64 @llvm.aarch64.neon.sqneg.i64(i64) nounwind readnone
declare i32 @llvm.aarch64.neon.sqneg.i32(i32) nounwind readnone
declare i64 @llvm.aarch64.neon.sqabs.i64(i64) nounwind readnone
declare i32 @llvm.aarch64.neon.sqabs.i32(i32) nounwind readnone


define i32 @vqmovund(<2 x i64> %b) nounwind readnone {
; CHECK-LABEL: vqmovund:
; CHECK:       // %bb.0:
; CHECK-NEXT:    sqxtun s0, d0
; CHECK-NEXT:    fmov w0, s0
; CHECK-NEXT:    ret
  %vecext = extractelement <2 x i64> %b, i32 0
  %vqmovun.i = tail call i32 @llvm.aarch64.neon.scalar.sqxtun.i32.i64(i64 %vecext) nounwind
  ret i32 %vqmovun.i
}

define i32 @vqmovnd_s(<2 x i64> %b) nounwind readnone {
; CHECK-LABEL: vqmovnd_s:
; CHECK:       // %bb.0:
; CHECK-NEXT:    sqxtn s0, d0
; CHECK-NEXT:    fmov w0, s0
; CHECK-NEXT:    ret
  %vecext = extractelement <2 x i64> %b, i32 0
  %vqmovn.i = tail call i32 @llvm.aarch64.neon.scalar.sqxtn.i32.i64(i64 %vecext) nounwind
  ret i32 %vqmovn.i
}

define i32 @vqmovnd_u(<2 x i64> %b) nounwind readnone {
; CHECK-LABEL: vqmovnd_u:
; CHECK:       // %bb.0:
; CHECK-NEXT:    uqxtn s0, d0
; CHECK-NEXT:    fmov w0, s0
; CHECK-NEXT:    ret
  %vecext = extractelement <2 x i64> %b, i32 0
  %vqmovn.i = tail call i32 @llvm.aarch64.neon.scalar.uqxtn.i32.i64(i64 %vecext) nounwind
  ret i32 %vqmovn.i
}

define i32 @uqxtn_ext(<4 x i32> noundef %a, <4 x i32> noundef %b, i32 %c, float %d, <2 x i64> %e) {
; CHECK-LABEL: uqxtn_ext:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    mov v0.d[0], v3.d[1]
; CHECK-NEXT:    uqxtn s0, d0
; CHECK-NEXT:    fmov w0, s0
; CHECK-NEXT:    ret
entry:
  %e1 = extractelement <2 x i64> %e, i64 1
  %r = tail call i32 @llvm.aarch64.neon.scalar.uqxtn.i32.i64(i64 %e1)
  ret i32 %r
}

define <4 x i32> @sqxtn_ins(<4 x i32> noundef %a, i64 %c) {
; CHECK-LABEL: sqxtn_ins:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    fmov d1, x0
; CHECK-NEXT:    sqxtn s1, d1
; CHECK-NEXT:    mov v0.s[3], v1.s[0]
; CHECK-NEXT:    ret
entry:
  %vqmovnd_s64.i = tail call i32 @llvm.aarch64.neon.scalar.sqxtn.i32.i64(i64 %c)
  %vecins = insertelement <4 x i32> %a, i32 %vqmovnd_s64.i, i64 3
  ret <4 x i32> %vecins
}

define <4 x i32> @sqxtun_insext(<4 x i32> noundef %a, <2 x i64> %e) {
; CHECK-LABEL: sqxtun_insext:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    mov v1.d[0], v1.d[1]
; CHECK-NEXT:    sqxtun s1, d1
; CHECK-NEXT:    mov v0.s[3], v1.s[0]
; CHECK-NEXT:    ret
entry:
  %c = extractelement <2 x i64> %e, i64 1
  %vqmovnd_s64.i = tail call i32 @llvm.aarch64.neon.scalar.sqxtun.i32.i64(i64 %c)
  %vecins = insertelement <4 x i32> %a, i32 %vqmovnd_s64.i, i64 3
  ret <4 x i32> %vecins
}

define <4 x i32> @saddluse(<4 x i32> noundef %a, <4 x i32> noundef %b, i32 %c, float %d, <2 x i64> %e) {
; CHECK-LABEL: saddluse:
; CHECK:       // %bb.0: // %entry
; CHECK-NEXT:    saddlv d1, v1.4s
; CHECK-NEXT:    sqxtn s1, d1
; CHECK-NEXT:    mov v0.s[1], v1.s[0]
; CHECK-NEXT:    ret
entry:
  %vaddlvq_s32.i = tail call i64 @llvm.aarch64.neon.saddlv.i64.v4i32(<4 x i32> %b)
  %vqmovnd_s64.i = tail call i32 @llvm.aarch64.neon.scalar.sqxtn.i32.i64(i64 %vaddlvq_s32.i)
  %vecins = insertelement <4 x i32> %a, i32 %vqmovnd_s64.i, i64 1
  ret <4 x i32> %vecins
}

declare i32 @llvm.aarch64.neon.scalar.uqxtn.i32.i64(i64) nounwind readnone
declare i32 @llvm.aarch64.neon.scalar.sqxtn.i32.i64(i64) nounwind readnone
declare i32 @llvm.aarch64.neon.scalar.sqxtun.i32.i64(i64) nounwind readnone
declare i64 @llvm.aarch64.neon.saddlv.i64.v4i32(<4 x i32>)