llvm/test/CodeGen/X86/fpclamptosat_vec.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s

; i32 saturate

define <2 x i32> @stest_f64i32(<2 x double> %x) {
; CHECK-LABEL: stest_f64i32:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttsd2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    cvttsd2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = [4294967295,4294967295]
; CHECK-NEXT:    movdqa %xmm3, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm2, %xmm4
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm5, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm3
; CHECK-NEXT:    pand %xmm3, %xmm1
; CHECK-NEXT:    pandn {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3
; CHECK-NEXT:    por %xmm1, %xmm3
; CHECK-NEXT:    pxor %xmm3, %xmm0
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = [18446744069414584320,18446744069414584320]
; CHECK-NEXT:    movdqa %xmm0, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm1, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
; CHECK-NEXT:    pand %xmm4, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
; CHECK-NEXT:    por %xmm0, %xmm1
; CHECK-NEXT:    pand %xmm1, %xmm3
; CHECK-NEXT:    pandn {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT:    por %xmm3, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x double> %x to <2 x i64>
  %0 = icmp slt <2 x i64> %conv, <i64 2147483647, i64 2147483647>
  %spec.store.select = select <2 x i1> %0, <2 x i64> %conv, <2 x i64> <i64 2147483647, i64 2147483647>
  %1 = icmp sgt <2 x i64> %spec.store.select, <i64 -2147483648, i64 -2147483648>
  %spec.store.select7 = select <2 x i1> %1, <2 x i64> %spec.store.select, <2 x i64> <i64 -2147483648, i64 -2147483648>
  %conv6 = trunc <2 x i64> %spec.store.select7 to <2 x i32>
  ret <2 x i32> %conv6
}

define <2 x i32> @utest_f64i32(<2 x double> %x) {
; CHECK-LABEL: utest_f64i32:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    movsd {{.*#+}} xmm2 = mem[0],zero
; CHECK-NEXT:    movapd %xmm0, %xmm1
; CHECK-NEXT:    subsd %xmm2, %xmm1
; CHECK-NEXT:    cvttsd2si %xmm1, %rax
; CHECK-NEXT:    cvttsd2si %xmm0, %rcx
; CHECK-NEXT:    movq %rcx, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rax, %rdx
; CHECK-NEXT:    orq %rcx, %rdx
; CHECK-NEXT:    movq %rdx, %xmm1
; CHECK-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    cvttsd2si %xmm0, %rax
; CHECK-NEXT:    subsd %xmm2, %xmm0
; CHECK-NEXT:    cvttsd2si %xmm0, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [9223372039002259456,9223372039002259456]
; CHECK-NEXT:    pxor %xmm1, %xmm0
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [9223372039002259455,9223372039002259455]
; CHECK-NEXT:    movdqa %xmm2, %xmm3
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm2, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
; CHECK-NEXT:    pand %xmm4, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,1,3,3]
; CHECK-NEXT:    por %xmm0, %xmm2
; CHECK-NEXT:    pand %xmm2, %xmm1
; CHECK-NEXT:    pandn {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
; CHECK-NEXT:    por %xmm1, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,2,2,3]
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <2 x double> %x to <2 x i64>
  %0 = icmp ult <2 x i64> %conv, <i64 4294967295, i64 4294967295>
  %spec.store.select = select <2 x i1> %0, <2 x i64> %conv, <2 x i64> <i64 4294967295, i64 4294967295>
  %conv6 = trunc <2 x i64> %spec.store.select to <2 x i32>
  ret <2 x i32> %conv6
}

define <2 x i32> @ustest_f64i32(<2 x double> %x) {
; CHECK-LABEL: ustest_f64i32:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttsd2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    cvttsd2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = [2147483647,2147483647]
; CHECK-NEXT:    movdqa %xmm3, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm2, %xmm4
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm5, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm3
; CHECK-NEXT:    pand %xmm3, %xmm1
; CHECK-NEXT:    pandn {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3
; CHECK-NEXT:    por %xmm1, %xmm3
; CHECK-NEXT:    movdqa %xmm3, %xmm1
; CHECK-NEXT:    pxor %xmm0, %xmm1
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm2
; CHECK-NEXT:    pcmpeqd %xmm0, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,3,3]
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
; CHECK-NEXT:    por %xmm0, %xmm1
; CHECK-NEXT:    pand %xmm3, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x double> %x to <2 x i64>
  %0 = icmp slt <2 x i64> %conv, <i64 4294967295, i64 4294967295>
  %spec.store.select = select <2 x i1> %0, <2 x i64> %conv, <2 x i64> <i64 4294967295, i64 4294967295>
  %1 = icmp sgt <2 x i64> %spec.store.select, zeroinitializer
  %spec.store.select7 = select <2 x i1> %1, <2 x i64> %spec.store.select, <2 x i64> zeroinitializer
  %conv6 = trunc <2 x i64> %spec.store.select7 to <2 x i32>
  ret <2 x i32> %conv6
}

define <4 x i32> @stest_f32i32(<4 x float> %x) {
; CHECK-LABEL: stest_f32i32:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    movaps %xmm0, %xmm1
; CHECK-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
; CHECK-NEXT:    cvttss2si %xmm1, %rax
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    movaps %xmm0, %xmm2
; CHECK-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
; CHECK-NEXT:    cvttss2si %xmm2, %rax
; CHECK-NEXT:    movq %rax, %xmm2
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm1[0]
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm4
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm0[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm8 = [2147483647,2147483647]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm4, %xmm1
; CHECK-NEXT:    pxor %xmm0, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm5 = [4294967295,4294967295]
; CHECK-NEXT:    movdqa %xmm5, %xmm6
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm6
; CHECK-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm5, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
; CHECK-NEXT:    pand %xmm7, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm1
; CHECK-NEXT:    pand %xmm1, %xmm4
; CHECK-NEXT:    pandn %xmm8, %xmm1
; CHECK-NEXT:    por %xmm4, %xmm1
; CHECK-NEXT:    movdqa %xmm2, %xmm3
; CHECK-NEXT:    pxor %xmm0, %xmm3
; CHECK-NEXT:    movdqa %xmm5, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm3, %xmm4
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm5, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm4
; CHECK-NEXT:    pand %xmm4, %xmm2
; CHECK-NEXT:    pandn %xmm8, %xmm4
; CHECK-NEXT:    por %xmm2, %xmm4
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [18446744071562067968,18446744071562067968]
; CHECK-NEXT:    movdqa %xmm4, %xmm3
; CHECK-NEXT:    pxor %xmm0, %xmm3
; CHECK-NEXT:    movdqa {{.*#+}} xmm5 = [18446744069414584320,18446744069414584320]
; CHECK-NEXT:    movdqa %xmm3, %xmm6
; CHECK-NEXT:    pcmpgtd %xmm5, %xmm6
; CHECK-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm5, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    pand %xmm7, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm6
; CHECK-NEXT:    pand %xmm6, %xmm4
; CHECK-NEXT:    pandn %xmm2, %xmm6
; CHECK-NEXT:    por %xmm4, %xmm6
; CHECK-NEXT:    pxor %xmm1, %xmm0
; CHECK-NEXT:    movdqa %xmm0, %xmm3
; CHECK-NEXT:    pcmpgtd %xmm5, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm5, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[1,1,3,3]
; CHECK-NEXT:    pand %xmm4, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
; CHECK-NEXT:    por %xmm5, %xmm0
; CHECK-NEXT:    pand %xmm0, %xmm1
; CHECK-NEXT:    pandn %xmm2, %xmm0
; CHECK-NEXT:    por %xmm1, %xmm0
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm6[0,2]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <4 x float> %x to <4 x i64>
  %0 = icmp slt <4 x i64> %conv, <i64 2147483647, i64 2147483647, i64 2147483647, i64 2147483647>
  %spec.store.select = select <4 x i1> %0, <4 x i64> %conv, <4 x i64> <i64 2147483647, i64 2147483647, i64 2147483647, i64 2147483647>
  %1 = icmp sgt <4 x i64> %spec.store.select, <i64 -2147483648, i64 -2147483648, i64 -2147483648, i64 -2147483648>
  %spec.store.select7 = select <4 x i1> %1, <4 x i64> %spec.store.select, <4 x i64> <i64 -2147483648, i64 -2147483648, i64 -2147483648, i64 -2147483648>
  %conv6 = trunc <4 x i64> %spec.store.select7 to <4 x i32>
  ret <4 x i32> %conv6
}

define <4 x i32> @utest_f32i32(<4 x float> %x) {
; CHECK-LABEL: utest_f32i32:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; CHECK-NEXT:    movaps %xmm0, %xmm1
; CHECK-NEXT:    subss %xmm2, %xmm1
; CHECK-NEXT:    cvttss2si %xmm1, %rax
; CHECK-NEXT:    cvttss2si %xmm0, %rcx
; CHECK-NEXT:    movq %rcx, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rax, %rdx
; CHECK-NEXT:    orq %rcx, %rdx
; CHECK-NEXT:    movq %rdx, %xmm1
; CHECK-NEXT:    movaps %xmm0, %xmm3
; CHECK-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm0[1,1]
; CHECK-NEXT:    cvttss2si %xmm3, %rax
; CHECK-NEXT:    subss %xmm2, %xmm3
; CHECK-NEXT:    cvttss2si %xmm3, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm3
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm3[0]
; CHECK-NEXT:    movaps %xmm0, %xmm3
; CHECK-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm0[3,3]
; CHECK-NEXT:    cvttss2si %xmm3, %rax
; CHECK-NEXT:    subss %xmm2, %xmm3
; CHECK-NEXT:    cvttss2si %xmm3, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm3
; CHECK-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    subss %xmm2, %xmm0
; CHECK-NEXT:    cvttss2si %xmm0, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [4294967295,4294967295]
; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = [9223372039002259456,9223372039002259456]
; CHECK-NEXT:    movdqa %xmm0, %xmm4
; CHECK-NEXT:    pxor %xmm3, %xmm4
; CHECK-NEXT:    movdqa {{.*#+}} xmm5 = [9223372039002259455,9223372039002259455]
; CHECK-NEXT:    movdqa %xmm5, %xmm6
; CHECK-NEXT:    pcmpgtd %xmm4, %xmm6
; CHECK-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm5, %xmm4
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
; CHECK-NEXT:    pand %xmm7, %xmm4
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
; CHECK-NEXT:    por %xmm4, %xmm6
; CHECK-NEXT:    pand %xmm6, %xmm0
; CHECK-NEXT:    pandn %xmm2, %xmm6
; CHECK-NEXT:    por %xmm0, %xmm6
; CHECK-NEXT:    pxor %xmm1, %xmm3
; CHECK-NEXT:    movdqa %xmm5, %xmm0
; CHECK-NEXT:    pcmpgtd %xmm3, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm5, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    pand %xmm4, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm0
; CHECK-NEXT:    pand %xmm0, %xmm1
; CHECK-NEXT:    pandn %xmm2, %xmm0
; CHECK-NEXT:    por %xmm1, %xmm0
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm6[0,2]
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <4 x float> %x to <4 x i64>
  %0 = icmp ult <4 x i64> %conv, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
  %spec.store.select = select <4 x i1> %0, <4 x i64> %conv, <4 x i64> <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
  %conv6 = trunc <4 x i64> %spec.store.select to <4 x i32>
  ret <4 x i32> %conv6
}

define <4 x i32> @ustest_f32i32(<4 x float> %x) {
; CHECK-LABEL: ustest_f32i32:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    movaps %xmm0, %xmm1
; CHECK-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
; CHECK-NEXT:    cvttss2si %xmm1, %rax
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    movaps %xmm0, %xmm2
; CHECK-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
; CHECK-NEXT:    cvttss2si %xmm2, %rax
; CHECK-NEXT:    movq %rax, %xmm2
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm1[0]
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm4
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm0[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm8 = [4294967295,4294967295]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm4, %xmm1
; CHECK-NEXT:    pxor %xmm0, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm5 = [2147483647,2147483647]
; CHECK-NEXT:    movdqa %xmm5, %xmm6
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm6
; CHECK-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm5, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
; CHECK-NEXT:    pand %xmm7, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm1
; CHECK-NEXT:    pand %xmm1, %xmm4
; CHECK-NEXT:    pandn %xmm8, %xmm1
; CHECK-NEXT:    por %xmm4, %xmm1
; CHECK-NEXT:    movdqa %xmm2, %xmm3
; CHECK-NEXT:    pxor %xmm0, %xmm3
; CHECK-NEXT:    movdqa %xmm5, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm3, %xmm4
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm5, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm4
; CHECK-NEXT:    pand %xmm4, %xmm2
; CHECK-NEXT:    pandn %xmm8, %xmm4
; CHECK-NEXT:    por %xmm2, %xmm4
; CHECK-NEXT:    movdqa %xmm4, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa %xmm2, %xmm3
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm3
; CHECK-NEXT:    pcmpeqd %xmm0, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm3, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm3
; CHECK-NEXT:    pand %xmm4, %xmm3
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa %xmm2, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm4
; CHECK-NEXT:    pcmpeqd %xmm0, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm4, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm0
; CHECK-NEXT:    pand %xmm1, %xmm0
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm3[0,2]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <4 x float> %x to <4 x i64>
  %0 = icmp slt <4 x i64> %conv, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
  %spec.store.select = select <4 x i1> %0, <4 x i64> %conv, <4 x i64> <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
  %1 = icmp sgt <4 x i64> %spec.store.select, zeroinitializer
  %spec.store.select7 = select <4 x i1> %1, <4 x i64> %spec.store.select, <4 x i64> zeroinitializer
  %conv6 = trunc <4 x i64> %spec.store.select7 to <4 x i32>
  ret <4 x i32> %conv6
}

define <4 x i32> @stest_f16i32(<4 x half> %x) {
; CHECK-LABEL: stest_f16i32:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    subq $32, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 64
; CHECK-NEXT:    .cfi_offset %rbx, -32
; CHECK-NEXT:    .cfi_offset %r14, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %edx, %ebp
; CHECK-NEXT:    movl %esi, %ebx
; CHECK-NEXT:    movl %edi, %r14d
; CHECK-NEXT:    movzwl %cx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq (%rsp), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movzwl %bx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r14w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm3
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm3 = xmm3[0],mem[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [2147483647,2147483647]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm3, %xmm1
; CHECK-NEXT:    pxor %xmm0, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm4 = [4294967295,4294967295]
; CHECK-NEXT:    movdqa %xmm4, %xmm5
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm7 = xmm1[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm7
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,3,3]
; CHECK-NEXT:    por %xmm7, %xmm1
; CHECK-NEXT:    pand %xmm1, %xmm3
; CHECK-NEXT:    pandn %xmm2, %xmm1
; CHECK-NEXT:    por %xmm3, %xmm1
; CHECK-NEXT:    movdqa (%rsp), %xmm7 # 16-byte Reload
; CHECK-NEXT:    movdqa %xmm7, %xmm3
; CHECK-NEXT:    pxor %xmm0, %xmm3
; CHECK-NEXT:    movdqa %xmm4, %xmm5
; CHECK-NEXT:    pcmpgtd %xmm3, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm4
; CHECK-NEXT:    movdqa %xmm7, %xmm3
; CHECK-NEXT:    pand %xmm4, %xmm3
; CHECK-NEXT:    pandn %xmm2, %xmm4
; CHECK-NEXT:    por %xmm3, %xmm4
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [18446744071562067968,18446744071562067968]
; CHECK-NEXT:    movdqa %xmm4, %xmm3
; CHECK-NEXT:    pxor %xmm0, %xmm3
; CHECK-NEXT:    movdqa {{.*#+}} xmm5 = [18446744069414584320,18446744069414584320]
; CHECK-NEXT:    movdqa %xmm3, %xmm6
; CHECK-NEXT:    pcmpgtd %xmm5, %xmm6
; CHECK-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm5, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    pand %xmm7, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm6
; CHECK-NEXT:    pand %xmm6, %xmm4
; CHECK-NEXT:    pandn %xmm2, %xmm6
; CHECK-NEXT:    por %xmm4, %xmm6
; CHECK-NEXT:    pxor %xmm1, %xmm0
; CHECK-NEXT:    movdqa %xmm0, %xmm3
; CHECK-NEXT:    pcmpgtd %xmm5, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm5, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[1,1,3,3]
; CHECK-NEXT:    pand %xmm4, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
; CHECK-NEXT:    por %xmm5, %xmm0
; CHECK-NEXT:    pand %xmm0, %xmm1
; CHECK-NEXT:    pandn %xmm2, %xmm0
; CHECK-NEXT:    por %xmm1, %xmm0
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm6[0,2]
; CHECK-NEXT:    addq $32, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <4 x half> %x to <4 x i64>
  %0 = icmp slt <4 x i64> %conv, <i64 2147483647, i64 2147483647, i64 2147483647, i64 2147483647>
  %spec.store.select = select <4 x i1> %0, <4 x i64> %conv, <4 x i64> <i64 2147483647, i64 2147483647, i64 2147483647, i64 2147483647>
  %1 = icmp sgt <4 x i64> %spec.store.select, <i64 -2147483648, i64 -2147483648, i64 -2147483648, i64 -2147483648>
  %spec.store.select7 = select <4 x i1> %1, <4 x i64> %spec.store.select, <4 x i64> <i64 -2147483648, i64 -2147483648, i64 -2147483648, i64 -2147483648>
  %conv6 = trunc <4 x i64> %spec.store.select7 to <4 x i32>
  ret <4 x i32> %conv6
}

define <4 x i32> @utesth_f16i32(<4 x half> %x) {
; CHECK-LABEL: utesth_f16i32:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    subq $32, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 64
; CHECK-NEXT:    .cfi_offset %rbx, -32
; CHECK-NEXT:    .cfi_offset %r14, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %ecx, %ebp
; CHECK-NEXT:    movl %edx, %r14d
; CHECK-NEXT:    movl %edi, %ebx
; CHECK-NEXT:    movzwl %si, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT:    cvttss2si %xmm0, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm0
; CHECK-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movzwl %bx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT:    cvttss2si %xmm0, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm0
; CHECK-NEXT:    punpcklqdq (%rsp), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT:    cvttss2si %xmm0, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r14w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT:    cvttss2si %xmm0, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm0
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = [4294967295,4294967295]
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
; CHECK-NEXT:    movdqa %xmm0, %xmm3
; CHECK-NEXT:    pxor %xmm2, %xmm3
; CHECK-NEXT:    movdqa {{.*#+}} xmm4 = [9223372039002259455,9223372039002259455]
; CHECK-NEXT:    movdqa %xmm4, %xmm5
; CHECK-NEXT:    pcmpgtd %xmm3, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm5
; CHECK-NEXT:    pand %xmm5, %xmm0
; CHECK-NEXT:    pandn %xmm1, %xmm5
; CHECK-NEXT:    por %xmm0, %xmm5
; CHECK-NEXT:    movdqa (%rsp), %xmm6 # 16-byte Reload
; CHECK-NEXT:    pxor %xmm6, %xmm2
; CHECK-NEXT:    movdqa %xmm4, %xmm0
; CHECK-NEXT:    pcmpgtd %xmm2, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm3, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm0
; CHECK-NEXT:    pand %xmm0, %xmm6
; CHECK-NEXT:    pandn %xmm1, %xmm0
; CHECK-NEXT:    por %xmm6, %xmm0
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
; CHECK-NEXT:    addq $32, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <4 x half> %x to <4 x i64>
  %0 = icmp ult <4 x i64> %conv, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
  %spec.store.select = select <4 x i1> %0, <4 x i64> %conv, <4 x i64> <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
  %conv6 = trunc <4 x i64> %spec.store.select to <4 x i32>
  ret <4 x i32> %conv6
}

define <4 x i32> @ustest_f16i32(<4 x half> %x) {
; CHECK-LABEL: ustest_f16i32:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    subq $32, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 64
; CHECK-NEXT:    .cfi_offset %rbx, -32
; CHECK-NEXT:    .cfi_offset %r14, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %edx, %ebp
; CHECK-NEXT:    movl %esi, %ebx
; CHECK-NEXT:    movl %edi, %r14d
; CHECK-NEXT:    movzwl %cx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq (%rsp), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movzwl %bx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r14w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm3
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm3 = xmm3[0],mem[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [4294967295,4294967295]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm3, %xmm1
; CHECK-NEXT:    pxor %xmm0, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm4 = [2147483647,2147483647]
; CHECK-NEXT:    movdqa %xmm4, %xmm5
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm7 = xmm1[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm7
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,3,3]
; CHECK-NEXT:    por %xmm7, %xmm1
; CHECK-NEXT:    pand %xmm1, %xmm3
; CHECK-NEXT:    pandn %xmm2, %xmm1
; CHECK-NEXT:    por %xmm3, %xmm1
; CHECK-NEXT:    movdqa (%rsp), %xmm7 # 16-byte Reload
; CHECK-NEXT:    movdqa %xmm7, %xmm3
; CHECK-NEXT:    pxor %xmm0, %xmm3
; CHECK-NEXT:    movdqa %xmm4, %xmm5
; CHECK-NEXT:    pcmpgtd %xmm3, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm4
; CHECK-NEXT:    movdqa %xmm7, %xmm3
; CHECK-NEXT:    pand %xmm4, %xmm3
; CHECK-NEXT:    pandn %xmm2, %xmm4
; CHECK-NEXT:    por %xmm3, %xmm4
; CHECK-NEXT:    movdqa %xmm4, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa %xmm2, %xmm3
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm3
; CHECK-NEXT:    pcmpeqd %xmm0, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm3, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm3
; CHECK-NEXT:    pand %xmm4, %xmm3
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa %xmm2, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm4
; CHECK-NEXT:    pcmpeqd %xmm0, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm4, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm0
; CHECK-NEXT:    pand %xmm1, %xmm0
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm3[0,2]
; CHECK-NEXT:    addq $32, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <4 x half> %x to <4 x i64>
  %0 = icmp slt <4 x i64> %conv, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
  %spec.store.select = select <4 x i1> %0, <4 x i64> %conv, <4 x i64> <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
  %1 = icmp sgt <4 x i64> %spec.store.select, zeroinitializer
  %spec.store.select7 = select <4 x i1> %1, <4 x i64> %spec.store.select, <4 x i64> zeroinitializer
  %conv6 = trunc <4 x i64> %spec.store.select7 to <4 x i32>
  ret <4 x i32> %conv6
}

; i16 saturate

define <2 x i16> @stest_f64i16(<2 x double> %x) {
; CHECK-LABEL: stest_f64i16:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttpd2dq %xmm0, %xmm0
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = <32767,32767,u,u>
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm2
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    pandn %xmm1, %xmm2
; CHECK-NEXT:    por %xmm0, %xmm2
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = <4294934528,4294934528,u,u>
; CHECK-NEXT:    movdqa %xmm2, %xmm1
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm1
; CHECK-NEXT:    pand %xmm1, %xmm2
; CHECK-NEXT:    pandn %xmm0, %xmm1
; CHECK-NEXT:    por %xmm2, %xmm1
; CHECK-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x double> %x to <2 x i32>
  %0 = icmp slt <2 x i32> %conv, <i32 32767, i32 32767>
  %spec.store.select = select <2 x i1> %0, <2 x i32> %conv, <2 x i32> <i32 32767, i32 32767>
  %1 = icmp sgt <2 x i32> %spec.store.select, <i32 -32768, i32 -32768>
  %spec.store.select7 = select <2 x i1> %1, <2 x i32> %spec.store.select, <2 x i32> <i32 -32768, i32 -32768>
  %conv6 = trunc <2 x i32> %spec.store.select7 to <2 x i16>
  ret <2 x i16> %conv6
}

define <2 x i16> @utest_f64i16(<2 x double> %x) {
; CHECK-LABEL: utest_f64i16:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttpd2dq %xmm0, %xmm1
; CHECK-NEXT:    movapd %xmm1, %xmm2
; CHECK-NEXT:    psrad $31, %xmm2
; CHECK-NEXT:    addpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT:    cvttpd2dq %xmm0, %xmm0
; CHECK-NEXT:    andpd %xmm2, %xmm0
; CHECK-NEXT:    orpd %xmm1, %xmm0
; CHECK-NEXT:    movapd {{.*#+}} xmm1 = [2147483648,2147483648,2147483648,2147483648]
; CHECK-NEXT:    xorpd %xmm0, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = <2147549183,2147549183,u,u>
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm2
; CHECK-NEXT:    andpd %xmm2, %xmm0
; CHECK-NEXT:    andnpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
; CHECK-NEXT:    orpd %xmm0, %xmm2
; CHECK-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,2,2,3,4,5,6,7]
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <2 x double> %x to <2 x i32>
  %0 = icmp ult <2 x i32> %conv, <i32 65535, i32 65535>
  %spec.store.select = select <2 x i1> %0, <2 x i32> %conv, <2 x i32> <i32 65535, i32 65535>
  %conv6 = trunc <2 x i32> %spec.store.select to <2 x i16>
  ret <2 x i16> %conv6
}

define <2 x i16> @ustest_f64i16(<2 x double> %x) {
; CHECK-LABEL: ustest_f64i16:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttpd2dq %xmm0, %xmm0
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = <65535,65535,u,u>
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm2
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    pandn %xmm1, %xmm2
; CHECK-NEXT:    por %xmm0, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm0
; CHECK-NEXT:    movdqa %xmm2, %xmm1
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm1
; CHECK-NEXT:    pand %xmm2, %xmm1
; CHECK-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x double> %x to <2 x i32>
  %0 = icmp slt <2 x i32> %conv, <i32 65535, i32 65535>
  %spec.store.select = select <2 x i1> %0, <2 x i32> %conv, <2 x i32> <i32 65535, i32 65535>
  %1 = icmp sgt <2 x i32> %spec.store.select, zeroinitializer
  %spec.store.select7 = select <2 x i1> %1, <2 x i32> %spec.store.select, <2 x i32> zeroinitializer
  %conv6 = trunc <2 x i32> %spec.store.select7 to <2 x i16>
  ret <2 x i16> %conv6
}

define <4 x i16> @stest_f32i16(<4 x float> %x) {
; CHECK-LABEL: stest_f32i16:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttps2dq %xmm0, %xmm0
; CHECK-NEXT:    packssdw %xmm0, %xmm0
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <4 x float> %x to <4 x i32>
  %0 = icmp slt <4 x i32> %conv, <i32 32767, i32 32767, i32 32767, i32 32767>
  %spec.store.select = select <4 x i1> %0, <4 x i32> %conv, <4 x i32> <i32 32767, i32 32767, i32 32767, i32 32767>
  %1 = icmp sgt <4 x i32> %spec.store.select, <i32 -32768, i32 -32768, i32 -32768, i32 -32768>
  %spec.store.select7 = select <4 x i1> %1, <4 x i32> %spec.store.select, <4 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768>
  %conv6 = trunc <4 x i32> %spec.store.select7 to <4 x i16>
  ret <4 x i16> %conv6
}

define <4 x i16> @utest_f32i16(<4 x float> %x) {
; CHECK-LABEL: utest_f32i16:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttps2dq %xmm0, %xmm1
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    psrad $31, %xmm2
; CHECK-NEXT:    subps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT:    cvttps2dq %xmm0, %xmm0
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    por %xmm1, %xmm0
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = [2147483648,2147483648,2147483648,2147483648]
; CHECK-NEXT:    pxor %xmm0, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [2147549183,2147549183,2147549183,2147549183]
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm2
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    pandn {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
; CHECK-NEXT:    por %xmm0, %xmm2
; CHECK-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,2,2,3,4,5,6,7]
; CHECK-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <4 x float> %x to <4 x i32>
  %0 = icmp ult <4 x i32> %conv, <i32 65535, i32 65535, i32 65535, i32 65535>
  %spec.store.select = select <4 x i1> %0, <4 x i32> %conv, <4 x i32> <i32 65535, i32 65535, i32 65535, i32 65535>
  %conv6 = trunc <4 x i32> %spec.store.select to <4 x i16>
  ret <4 x i16> %conv6
}

define <4 x i16> @ustest_f32i16(<4 x float> %x) {
; CHECK-LABEL: ustest_f32i16:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttps2dq %xmm0, %xmm0
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535]
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm2
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    pandn %xmm1, %xmm2
; CHECK-NEXT:    por %xmm0, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm0
; CHECK-NEXT:    movdqa %xmm2, %xmm1
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm1
; CHECK-NEXT:    pand %xmm2, %xmm1
; CHECK-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
; CHECK-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <4 x float> %x to <4 x i32>
  %0 = icmp slt <4 x i32> %conv, <i32 65535, i32 65535, i32 65535, i32 65535>
  %spec.store.select = select <4 x i1> %0, <4 x i32> %conv, <4 x i32> <i32 65535, i32 65535, i32 65535, i32 65535>
  %1 = icmp sgt <4 x i32> %spec.store.select, zeroinitializer
  %spec.store.select7 = select <4 x i1> %1, <4 x i32> %spec.store.select, <4 x i32> zeroinitializer
  %conv6 = trunc <4 x i32> %spec.store.select7 to <4 x i16>
  ret <4 x i16> %conv6
}

define <8 x i16> @stest_f16i16(<8 x half> %x) {
; CHECK-LABEL: stest_f16i16:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r15
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    pushq %r13
; CHECK-NEXT:    .cfi_def_cfa_offset 40
; CHECK-NEXT:    pushq %r12
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 56
; CHECK-NEXT:    subq $72, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 128
; CHECK-NEXT:    .cfi_offset %rbx, -56
; CHECK-NEXT:    .cfi_offset %r12, -48
; CHECK-NEXT:    .cfi_offset %r13, -40
; CHECK-NEXT:    .cfi_offset %r14, -32
; CHECK-NEXT:    .cfi_offset %r15, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %r9d, %ebp
; CHECK-NEXT:    movl %r8d, %ebx
; CHECK-NEXT:    movl %ecx, %r13d
; CHECK-NEXT:    movl %edx, %r12d
; CHECK-NEXT:    movl %esi, %r15d
; CHECK-NEXT:    movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT:    movzwl {{[0-9]+}}(%rsp), %r14d
; CHECK-NEXT:    movzwl {{[0-9]+}}(%rsp), %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movl %r14d, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %bx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r13w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r12w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r15w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl {{[-0-9]+}}(%r{{[sb]}}p), %edi # 2-byte Folded Reload
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    packssdw {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    addq $72, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 56
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    popq %r12
; CHECK-NEXT:    .cfi_def_cfa_offset 40
; CHECK-NEXT:    popq %r13
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r15
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <8 x half> %x to <8 x i32>
  %0 = icmp slt <8 x i32> %conv, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
  %spec.store.select = select <8 x i1> %0, <8 x i32> %conv, <8 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
  %1 = icmp sgt <8 x i32> %spec.store.select, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
  %spec.store.select7 = select <8 x i1> %1, <8 x i32> %spec.store.select, <8 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
  %conv6 = trunc <8 x i32> %spec.store.select7 to <8 x i16>
  ret <8 x i16> %conv6
}

define <8 x i16> @utesth_f16i16(<8 x half> %x) {
; CHECK-LABEL: utesth_f16i16:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r15
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    pushq %r13
; CHECK-NEXT:    .cfi_def_cfa_offset 40
; CHECK-NEXT:    pushq %r12
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 56
; CHECK-NEXT:    subq $72, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 128
; CHECK-NEXT:    .cfi_offset %rbx, -56
; CHECK-NEXT:    .cfi_offset %r12, -48
; CHECK-NEXT:    .cfi_offset %r13, -40
; CHECK-NEXT:    .cfi_offset %r14, -32
; CHECK-NEXT:    .cfi_offset %r15, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %r9d, %ebp
; CHECK-NEXT:    movl %r8d, %ebx
; CHECK-NEXT:    movl %ecx, %r13d
; CHECK-NEXT:    movl %edx, %r12d
; CHECK-NEXT:    movl %esi, %r15d
; CHECK-NEXT:    movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT:    movzwl {{[0-9]+}}(%rsp), %r14d
; CHECK-NEXT:    movzwl {{[0-9]+}}(%rsp), %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movl %r14d, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %bx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r13w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r12w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r15w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl {{[-0-9]+}}(%r{{[sb]}}p), %edi # 2-byte Folded Reload
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm1
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm1 = xmm1[0],mem[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648,2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm1, %xmm3
; CHECK-NEXT:    pxor %xmm2, %xmm3
; CHECK-NEXT:    movdqa {{.*#+}} xmm4 = [2147549183,2147549183,2147549183,2147549183]
; CHECK-NEXT:    movdqa %xmm4, %xmm0
; CHECK-NEXT:    pcmpgtd %xmm3, %xmm0
; CHECK-NEXT:    pand %xmm0, %xmm1
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm3
; CHECK-NEXT:    pxor %xmm3, %xmm0
; CHECK-NEXT:    por %xmm1, %xmm0
; CHECK-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; CHECK-NEXT:    pxor %xmm1, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm2, %xmm4
; CHECK-NEXT:    pand %xmm4, %xmm1
; CHECK-NEXT:    pxor %xmm3, %xmm4
; CHECK-NEXT:    por %xmm1, %xmm4
; CHECK-NEXT:    pslld $16, %xmm4
; CHECK-NEXT:    psrad $16, %xmm4
; CHECK-NEXT:    pslld $16, %xmm0
; CHECK-NEXT:    psrad $16, %xmm0
; CHECK-NEXT:    packssdw %xmm4, %xmm0
; CHECK-NEXT:    addq $72, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 56
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    popq %r12
; CHECK-NEXT:    .cfi_def_cfa_offset 40
; CHECK-NEXT:    popq %r13
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r15
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <8 x half> %x to <8 x i32>
  %0 = icmp ult <8 x i32> %conv, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
  %spec.store.select = select <8 x i1> %0, <8 x i32> %conv, <8 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
  %conv6 = trunc <8 x i32> %spec.store.select to <8 x i16>
  ret <8 x i16> %conv6
}

define <8 x i16> @ustest_f16i16(<8 x half> %x) {
; CHECK-LABEL: ustest_f16i16:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r15
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    pushq %r13
; CHECK-NEXT:    .cfi_def_cfa_offset 40
; CHECK-NEXT:    pushq %r12
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 56
; CHECK-NEXT:    subq $72, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 128
; CHECK-NEXT:    .cfi_offset %rbx, -56
; CHECK-NEXT:    .cfi_offset %r12, -48
; CHECK-NEXT:    .cfi_offset %r13, -40
; CHECK-NEXT:    .cfi_offset %r14, -32
; CHECK-NEXT:    .cfi_offset %r15, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %r9d, %r15d
; CHECK-NEXT:    movl %r8d, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT:    movl %edx, %ebp
; CHECK-NEXT:    movl %esi, %r14d
; CHECK-NEXT:    movl %edi, %ebx
; CHECK-NEXT:    movzwl {{[0-9]+}}(%rsp), %r12d
; CHECK-NEXT:    movzwl {{[0-9]+}}(%rsp), %r13d
; CHECK-NEXT:    movzwl %cx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r14w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %bx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movl %r13d, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movl %r12d, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r15w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl {{[-0-9]+}}(%r{{[sb]}}p), %edi # 2-byte Folded Reload
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535]
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm2
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    pandn %xmm1, %xmm2
; CHECK-NEXT:    por %xmm0, %xmm2
; CHECK-NEXT:    movdqa %xmm1, %xmm3
; CHECK-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm3
; CHECK-NEXT:    pand %xmm3, %xmm0
; CHECK-NEXT:    pandn %xmm1, %xmm3
; CHECK-NEXT:    por %xmm0, %xmm3
; CHECK-NEXT:    pxor %xmm1, %xmm1
; CHECK-NEXT:    movdqa %xmm3, %xmm0
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm0
; CHECK-NEXT:    pand %xmm3, %xmm0
; CHECK-NEXT:    movdqa %xmm2, %xmm3
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm3
; CHECK-NEXT:    pand %xmm2, %xmm3
; CHECK-NEXT:    pslld $16, %xmm3
; CHECK-NEXT:    psrad $16, %xmm3
; CHECK-NEXT:    pslld $16, %xmm0
; CHECK-NEXT:    psrad $16, %xmm0
; CHECK-NEXT:    packssdw %xmm3, %xmm0
; CHECK-NEXT:    addq $72, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 56
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    popq %r12
; CHECK-NEXT:    .cfi_def_cfa_offset 40
; CHECK-NEXT:    popq %r13
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r15
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <8 x half> %x to <8 x i32>
  %0 = icmp slt <8 x i32> %conv, <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
  %spec.store.select = select <8 x i1> %0, <8 x i32> %conv, <8 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>
  %1 = icmp sgt <8 x i32> %spec.store.select, zeroinitializer
  %spec.store.select7 = select <8 x i1> %1, <8 x i32> %spec.store.select, <8 x i32> zeroinitializer
  %conv6 = trunc <8 x i32> %spec.store.select7 to <8 x i16>
  ret <8 x i16> %conv6
}

; i64 saturate

define <2 x i64> @stest_f64i64(<2 x double> %x) {
; CHECK-LABEL: stest_f64i64:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    subq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset %rbx, -24
; CHECK-NEXT:    .cfi_offset %r14, -16
; CHECK-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    callq __fixdfti@PLT
; CHECK-NEXT:    movq %rax, %r14
; CHECK-NEXT:    movq %rdx, %rbx
; CHECK-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT:    callq __fixdfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    movabsq $9223372036854775807, %rsi # imm = 0x7FFFFFFFFFFFFFFF
; CHECK-NEXT:    cmpq %rsi, %rax
; CHECK-NEXT:    movq %rdx, %rdi
; CHECK-NEXT:    sbbq $0, %rdi
; CHECK-NEXT:    cmovgeq %rcx, %rdx
; CHECK-NEXT:    cmovgeq %rsi, %rax
; CHECK-NEXT:    cmpq %rsi, %r14
; CHECK-NEXT:    movq %rbx, %rdi
; CHECK-NEXT:    sbbq $0, %rdi
; CHECK-NEXT:    cmovlq %rbx, %rcx
; CHECK-NEXT:    cmovlq %r14, %rsi
; CHECK-NEXT:    movabsq $-9223372036854775808, %r8 # imm = 0x8000000000000000
; CHECK-NEXT:    cmpq %rsi, %r8
; CHECK-NEXT:    movq $-1, %rbx
; CHECK-NEXT:    movq $-1, %rdi
; CHECK-NEXT:    sbbq %rcx, %rdi
; CHECK-NEXT:    cmovgeq %r8, %rsi
; CHECK-NEXT:    cmpq %rax, %r8
; CHECK-NEXT:    sbbq %rdx, %rbx
; CHECK-NEXT:    cmovgeq %r8, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movq %rsi, %xmm1
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    addq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x double> %x to <2 x i128>
  %0 = icmp slt <2 x i128> %conv, <i128 9223372036854775807, i128 9223372036854775807>
  %spec.store.select = select <2 x i1> %0, <2 x i128> %conv, <2 x i128> <i128 9223372036854775807, i128 9223372036854775807>
  %1 = icmp sgt <2 x i128> %spec.store.select, <i128 -9223372036854775808, i128 -9223372036854775808>
  %spec.store.select7 = select <2 x i1> %1, <2 x i128> %spec.store.select, <2 x i128> <i128 -9223372036854775808, i128 -9223372036854775808>
  %conv6 = trunc <2 x i128> %spec.store.select7 to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @utest_f64i64(<2 x double> %x) {
; CHECK-LABEL: utest_f64i64:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    subq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset %rbx, -24
; CHECK-NEXT:    .cfi_offset %r14, -16
; CHECK-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    callq __fixunsdfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %r14
; CHECK-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    callq __fixunsdfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    cmovneq %rcx, %rax
; CHECK-NEXT:    testq %r14, %r14
; CHECK-NEXT:    cmovneq %rcx, %rbx
; CHECK-NEXT:    movq %rbx, %xmm0
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    addq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <2 x double> %x to <2 x i128>
  %0 = icmp ult <2 x i128> %conv, <i128 18446744073709551616, i128 18446744073709551616>
  %spec.store.select = select <2 x i1> %0, <2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>
  %conv6 = trunc <2 x i128> %spec.store.select to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @ustest_f64i64(<2 x double> %x) {
; CHECK-LABEL: ustest_f64i64:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    subq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset %rbx, -24
; CHECK-NEXT:    .cfi_offset %r14, -16
; CHECK-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    callq __fixdfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %r14
; CHECK-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT:    callq __fixdfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    movl $1, %esi
; CHECK-NEXT:    cmovgq %rsi, %rdx
; CHECK-NEXT:    cmovgq %rcx, %rax
; CHECK-NEXT:    testq %r14, %r14
; CHECK-NEXT:    cmovleq %r14, %rsi
; CHECK-NEXT:    cmovgq %rcx, %rbx
; CHECK-NEXT:    movq %rbx, %rdi
; CHECK-NEXT:    negq %rdi
; CHECK-NEXT:    movl $0, %edi
; CHECK-NEXT:    sbbq %rsi, %rdi
; CHECK-NEXT:    cmovgeq %rcx, %rbx
; CHECK-NEXT:    movq %rax, %rsi
; CHECK-NEXT:    negq %rsi
; CHECK-NEXT:    movl $0, %esi
; CHECK-NEXT:    sbbq %rdx, %rsi
; CHECK-NEXT:    cmovgeq %rcx, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movq %rbx, %xmm1
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    addq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x double> %x to <2 x i128>
  %0 = icmp slt <2 x i128> %conv, <i128 18446744073709551616, i128 18446744073709551616>
  %spec.store.select = select <2 x i1> %0, <2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>
  %1 = icmp sgt <2 x i128> %spec.store.select, zeroinitializer
  %spec.store.select7 = select <2 x i1> %1, <2 x i128> %spec.store.select, <2 x i128> zeroinitializer
  %conv6 = trunc <2 x i128> %spec.store.select7 to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @stest_f32i64(<2 x float> %x) {
; CHECK-LABEL: stest_f32i64:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    subq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset %rbx, -24
; CHECK-NEXT:    .cfi_offset %r14, -16
; CHECK-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    movq %rax, %r14
; CHECK-NEXT:    movq %rdx, %rbx
; CHECK-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    movabsq $9223372036854775807, %rsi # imm = 0x7FFFFFFFFFFFFFFF
; CHECK-NEXT:    cmpq %rsi, %rax
; CHECK-NEXT:    movq %rdx, %rdi
; CHECK-NEXT:    sbbq $0, %rdi
; CHECK-NEXT:    cmovgeq %rcx, %rdx
; CHECK-NEXT:    cmovgeq %rsi, %rax
; CHECK-NEXT:    cmpq %rsi, %r14
; CHECK-NEXT:    movq %rbx, %rdi
; CHECK-NEXT:    sbbq $0, %rdi
; CHECK-NEXT:    cmovlq %rbx, %rcx
; CHECK-NEXT:    cmovlq %r14, %rsi
; CHECK-NEXT:    movabsq $-9223372036854775808, %r8 # imm = 0x8000000000000000
; CHECK-NEXT:    cmpq %rsi, %r8
; CHECK-NEXT:    movq $-1, %rbx
; CHECK-NEXT:    movq $-1, %rdi
; CHECK-NEXT:    sbbq %rcx, %rdi
; CHECK-NEXT:    cmovgeq %r8, %rsi
; CHECK-NEXT:    cmpq %rax, %r8
; CHECK-NEXT:    sbbq %rdx, %rbx
; CHECK-NEXT:    cmovgeq %r8, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movq %rsi, %xmm1
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    addq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x float> %x to <2 x i128>
  %0 = icmp slt <2 x i128> %conv, <i128 9223372036854775807, i128 9223372036854775807>
  %spec.store.select = select <2 x i1> %0, <2 x i128> %conv, <2 x i128> <i128 9223372036854775807, i128 9223372036854775807>
  %1 = icmp sgt <2 x i128> %spec.store.select, <i128 -9223372036854775808, i128 -9223372036854775808>
  %spec.store.select7 = select <2 x i1> %1, <2 x i128> %spec.store.select, <2 x i128> <i128 -9223372036854775808, i128 -9223372036854775808>
  %conv6 = trunc <2 x i128> %spec.store.select7 to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @utest_f32i64(<2 x float> %x) {
; CHECK-LABEL: utest_f32i64:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    subq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset %rbx, -24
; CHECK-NEXT:    .cfi_offset %r14, -16
; CHECK-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    callq __fixunssfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %r14
; CHECK-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT:    callq __fixunssfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    cmovneq %rcx, %rax
; CHECK-NEXT:    testq %r14, %r14
; CHECK-NEXT:    cmovneq %rcx, %rbx
; CHECK-NEXT:    movq %rbx, %xmm0
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    addq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <2 x float> %x to <2 x i128>
  %0 = icmp ult <2 x i128> %conv, <i128 18446744073709551616, i128 18446744073709551616>
  %spec.store.select = select <2 x i1> %0, <2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>
  %conv6 = trunc <2 x i128> %spec.store.select to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @ustest_f32i64(<2 x float> %x) {
; CHECK-LABEL: ustest_f32i64:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    subq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset %rbx, -24
; CHECK-NEXT:    .cfi_offset %r14, -16
; CHECK-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %r14
; CHECK-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    movl $1, %esi
; CHECK-NEXT:    cmovgq %rsi, %rdx
; CHECK-NEXT:    cmovgq %rcx, %rax
; CHECK-NEXT:    testq %r14, %r14
; CHECK-NEXT:    cmovleq %r14, %rsi
; CHECK-NEXT:    cmovgq %rcx, %rbx
; CHECK-NEXT:    movq %rbx, %rdi
; CHECK-NEXT:    negq %rdi
; CHECK-NEXT:    movl $0, %edi
; CHECK-NEXT:    sbbq %rsi, %rdi
; CHECK-NEXT:    cmovgeq %rcx, %rbx
; CHECK-NEXT:    movq %rax, %rsi
; CHECK-NEXT:    negq %rsi
; CHECK-NEXT:    movl $0, %esi
; CHECK-NEXT:    sbbq %rdx, %rsi
; CHECK-NEXT:    cmovgeq %rcx, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movq %rbx, %xmm1
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    addq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x float> %x to <2 x i128>
  %0 = icmp slt <2 x i128> %conv, <i128 18446744073709551616, i128 18446744073709551616>
  %spec.store.select = select <2 x i1> %0, <2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>
  %1 = icmp sgt <2 x i128> %spec.store.select, zeroinitializer
  %spec.store.select7 = select <2 x i1> %1, <2 x i128> %spec.store.select, <2 x i128> zeroinitializer
  %conv6 = trunc <2 x i128> %spec.store.select7 to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @stest_f16i64(<2 x half> %x) {
; CHECK-LABEL: stest_f16i64:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    .cfi_offset %rbx, -32
; CHECK-NEXT:    .cfi_offset %r14, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %esi, %r14d
; CHECK-NEXT:    movzwl %di, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %rbp
; CHECK-NEXT:    movzwl %r14w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    movabsq $9223372036854775807, %rsi # imm = 0x7FFFFFFFFFFFFFFF
; CHECK-NEXT:    cmpq %rsi, %rax
; CHECK-NEXT:    movq %rdx, %rdi
; CHECK-NEXT:    sbbq $0, %rdi
; CHECK-NEXT:    cmovgeq %rcx, %rdx
; CHECK-NEXT:    cmovgeq %rsi, %rax
; CHECK-NEXT:    cmpq %rsi, %rbx
; CHECK-NEXT:    movq %rbp, %rdi
; CHECK-NEXT:    sbbq $0, %rdi
; CHECK-NEXT:    cmovlq %rbp, %rcx
; CHECK-NEXT:    cmovlq %rbx, %rsi
; CHECK-NEXT:    movabsq $-9223372036854775808, %rdi # imm = 0x8000000000000000
; CHECK-NEXT:    cmpq %rsi, %rdi
; CHECK-NEXT:    movq $-1, %rbp
; CHECK-NEXT:    movq $-1, %rbx
; CHECK-NEXT:    sbbq %rcx, %rbx
; CHECK-NEXT:    cmovgeq %rdi, %rsi
; CHECK-NEXT:    cmpq %rax, %rdi
; CHECK-NEXT:    sbbq %rdx, %rbp
; CHECK-NEXT:    cmovgeq %rdi, %rax
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    movq %rsi, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x half> %x to <2 x i128>
  %0 = icmp slt <2 x i128> %conv, <i128 9223372036854775807, i128 9223372036854775807>
  %spec.store.select = select <2 x i1> %0, <2 x i128> %conv, <2 x i128> <i128 9223372036854775807, i128 9223372036854775807>
  %1 = icmp sgt <2 x i128> %spec.store.select, <i128 -9223372036854775808, i128 -9223372036854775808>
  %spec.store.select7 = select <2 x i1> %1, <2 x i128> %spec.store.select, <2 x i128> <i128 -9223372036854775808, i128 -9223372036854775808>
  %conv6 = trunc <2 x i128> %spec.store.select7 to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @utesth_f16i64(<2 x half> %x) {
; CHECK-LABEL: utesth_f16i64:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    .cfi_offset %rbx, -32
; CHECK-NEXT:    .cfi_offset %r14, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %edi, %ebp
; CHECK-NEXT:    movzwl %si, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    callq __fixunssfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %r14
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    callq __fixunssfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    cmovneq %rcx, %rax
; CHECK-NEXT:    testq %r14, %r14
; CHECK-NEXT:    cmovneq %rcx, %rbx
; CHECK-NEXT:    movq %rbx, %xmm1
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <2 x half> %x to <2 x i128>
  %0 = icmp ult <2 x i128> %conv, <i128 18446744073709551616, i128 18446744073709551616>
  %spec.store.select = select <2 x i1> %0, <2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>
  %conv6 = trunc <2 x i128> %spec.store.select to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @ustest_f16i64(<2 x half> %x) {
; CHECK-LABEL: ustest_f16i64:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    .cfi_offset %rbx, -32
; CHECK-NEXT:    .cfi_offset %r14, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %esi, %r14d
; CHECK-NEXT:    movzwl %di, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %rbp
; CHECK-NEXT:    movzwl %r14w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    movl $1, %esi
; CHECK-NEXT:    cmovgq %rsi, %rdx
; CHECK-NEXT:    cmovgq %rcx, %rax
; CHECK-NEXT:    testq %rbp, %rbp
; CHECK-NEXT:    cmovleq %rbp, %rsi
; CHECK-NEXT:    cmovgq %rcx, %rbx
; CHECK-NEXT:    movq %rbx, %rdi
; CHECK-NEXT:    negq %rdi
; CHECK-NEXT:    movl $0, %edi
; CHECK-NEXT:    sbbq %rsi, %rdi
; CHECK-NEXT:    cmovgeq %rcx, %rbx
; CHECK-NEXT:    movq %rax, %rsi
; CHECK-NEXT:    negq %rsi
; CHECK-NEXT:    movl $0, %esi
; CHECK-NEXT:    sbbq %rdx, %rsi
; CHECK-NEXT:    cmovgeq %rcx, %rax
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    movq %rbx, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x half> %x to <2 x i128>
  %0 = icmp slt <2 x i128> %conv, <i128 18446744073709551616, i128 18446744073709551616>
  %spec.store.select = select <2 x i1> %0, <2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>
  %1 = icmp sgt <2 x i128> %spec.store.select, zeroinitializer
  %spec.store.select7 = select <2 x i1> %1, <2 x i128> %spec.store.select, <2 x i128> zeroinitializer
  %conv6 = trunc <2 x i128> %spec.store.select7 to <2 x i64>
  ret <2 x i64> %conv6
}



; i32 saturate

define <2 x i32> @stest_f64i32_mm(<2 x double> %x) {
; CHECK-LABEL: stest_f64i32_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttsd2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    cvttsd2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = [4294967295,4294967295]
; CHECK-NEXT:    movdqa %xmm3, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm2, %xmm4
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm5, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm3
; CHECK-NEXT:    pand %xmm3, %xmm1
; CHECK-NEXT:    pandn {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3
; CHECK-NEXT:    por %xmm1, %xmm3
; CHECK-NEXT:    pxor %xmm3, %xmm0
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = [18446744069414584320,18446744069414584320]
; CHECK-NEXT:    movdqa %xmm0, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm1, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
; CHECK-NEXT:    pand %xmm4, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
; CHECK-NEXT:    por %xmm0, %xmm1
; CHECK-NEXT:    pand %xmm1, %xmm3
; CHECK-NEXT:    pandn {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT:    por %xmm3, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x double> %x to <2 x i64>
  %spec.store.select = call <2 x i64> @llvm.smin.v2i64(<2 x i64> %conv, <2 x i64> <i64 2147483647, i64 2147483647>)
  %spec.store.select7 = call <2 x i64> @llvm.smax.v2i64(<2 x i64> %spec.store.select, <2 x i64> <i64 -2147483648, i64 -2147483648>)
  %conv6 = trunc <2 x i64> %spec.store.select7 to <2 x i32>
  ret <2 x i32> %conv6
}

define <2 x i32> @utest_f64i32_mm(<2 x double> %x) {
; CHECK-LABEL: utest_f64i32_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    movsd {{.*#+}} xmm2 = mem[0],zero
; CHECK-NEXT:    movapd %xmm0, %xmm1
; CHECK-NEXT:    subsd %xmm2, %xmm1
; CHECK-NEXT:    cvttsd2si %xmm1, %rax
; CHECK-NEXT:    cvttsd2si %xmm0, %rcx
; CHECK-NEXT:    movq %rcx, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rax, %rdx
; CHECK-NEXT:    orq %rcx, %rdx
; CHECK-NEXT:    movq %rdx, %xmm1
; CHECK-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    cvttsd2si %xmm0, %rax
; CHECK-NEXT:    subsd %xmm2, %xmm0
; CHECK-NEXT:    cvttsd2si %xmm0, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [9223372039002259456,9223372039002259456]
; CHECK-NEXT:    pxor %xmm1, %xmm0
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [9223372039002259455,9223372039002259455]
; CHECK-NEXT:    movdqa %xmm2, %xmm3
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm2, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
; CHECK-NEXT:    pand %xmm4, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,1,3,3]
; CHECK-NEXT:    por %xmm0, %xmm2
; CHECK-NEXT:    pand %xmm2, %xmm1
; CHECK-NEXT:    pandn {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
; CHECK-NEXT:    por %xmm1, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,2,2,3]
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <2 x double> %x to <2 x i64>
  %spec.store.select = call <2 x i64> @llvm.umin.v2i64(<2 x i64> %conv, <2 x i64> <i64 4294967295, i64 4294967295>)
  %conv6 = trunc <2 x i64> %spec.store.select to <2 x i32>
  ret <2 x i32> %conv6
}

define <2 x i32> @ustest_f64i32_mm(<2 x double> %x) {
; CHECK-LABEL: ustest_f64i32_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttsd2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    cvttsd2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = [2147483647,2147483647]
; CHECK-NEXT:    movdqa %xmm3, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm2, %xmm4
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm5, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm3
; CHECK-NEXT:    pand %xmm3, %xmm1
; CHECK-NEXT:    pandn {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3
; CHECK-NEXT:    por %xmm1, %xmm3
; CHECK-NEXT:    movdqa %xmm3, %xmm1
; CHECK-NEXT:    pxor %xmm0, %xmm1
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm2
; CHECK-NEXT:    pcmpeqd %xmm0, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,3,3]
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
; CHECK-NEXT:    por %xmm0, %xmm1
; CHECK-NEXT:    pand %xmm3, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x double> %x to <2 x i64>
  %spec.store.select = call <2 x i64> @llvm.smin.v2i64(<2 x i64> %conv, <2 x i64> <i64 4294967295, i64 4294967295>)
  %spec.store.select7 = call <2 x i64> @llvm.smax.v2i64(<2 x i64> %spec.store.select, <2 x i64> zeroinitializer)
  %conv6 = trunc <2 x i64> %spec.store.select7 to <2 x i32>
  ret <2 x i32> %conv6
}

define <4 x i32> @stest_f32i32_mm(<4 x float> %x) {
; CHECK-LABEL: stest_f32i32_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    movaps %xmm0, %xmm1
; CHECK-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
; CHECK-NEXT:    cvttss2si %xmm1, %rax
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    movaps %xmm0, %xmm2
; CHECK-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
; CHECK-NEXT:    cvttss2si %xmm2, %rax
; CHECK-NEXT:    movq %rax, %xmm2
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm1[0]
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm3
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm3, %xmm1
; CHECK-NEXT:    pxor %xmm0, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm4 = [4294967295,4294967295]
; CHECK-NEXT:    movdqa %xmm4, %xmm5
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm7 = xmm1[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm7
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,3,3]
; CHECK-NEXT:    por %xmm7, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm5 = [2147483647,2147483647]
; CHECK-NEXT:    pand %xmm1, %xmm3
; CHECK-NEXT:    pandn %xmm5, %xmm1
; CHECK-NEXT:    por %xmm3, %xmm1
; CHECK-NEXT:    movdqa %xmm2, %xmm3
; CHECK-NEXT:    pxor %xmm0, %xmm3
; CHECK-NEXT:    movdqa %xmm4, %xmm6
; CHECK-NEXT:    pcmpgtd %xmm3, %xmm6
; CHECK-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    pand %xmm7, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm4
; CHECK-NEXT:    pand %xmm4, %xmm2
; CHECK-NEXT:    pandn %xmm5, %xmm4
; CHECK-NEXT:    por %xmm2, %xmm4
; CHECK-NEXT:    movdqa %xmm4, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = [18446744069414584320,18446744069414584320]
; CHECK-NEXT:    movdqa %xmm2, %xmm5
; CHECK-NEXT:    pcmpgtd %xmm3, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm5
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [18446744071562067968,18446744071562067968]
; CHECK-NEXT:    pand %xmm5, %xmm4
; CHECK-NEXT:    pandn %xmm2, %xmm5
; CHECK-NEXT:    por %xmm4, %xmm5
; CHECK-NEXT:    pxor %xmm1, %xmm0
; CHECK-NEXT:    movdqa %xmm0, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm3, %xmm4
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm0
; CHECK-NEXT:    pand %xmm0, %xmm1
; CHECK-NEXT:    pandn %xmm2, %xmm0
; CHECK-NEXT:    por %xmm1, %xmm0
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <4 x float> %x to <4 x i64>
  %spec.store.select = call <4 x i64> @llvm.smin.v4i64(<4 x i64> %conv, <4 x i64> <i64 2147483647, i64 2147483647, i64 2147483647, i64 2147483647>)
  %spec.store.select7 = call <4 x i64> @llvm.smax.v4i64(<4 x i64> %spec.store.select, <4 x i64> <i64 -2147483648, i64 -2147483648, i64 -2147483648, i64 -2147483648>)
  %conv6 = trunc <4 x i64> %spec.store.select7 to <4 x i32>
  ret <4 x i32> %conv6
}

define <4 x i32> @utest_f32i32_mm(<4 x float> %x) {
; CHECK-LABEL: utest_f32i32_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; CHECK-NEXT:    movaps %xmm0, %xmm1
; CHECK-NEXT:    subss %xmm2, %xmm1
; CHECK-NEXT:    cvttss2si %xmm1, %rax
; CHECK-NEXT:    cvttss2si %xmm0, %rcx
; CHECK-NEXT:    movq %rcx, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rax, %rdx
; CHECK-NEXT:    orq %rcx, %rdx
; CHECK-NEXT:    movq %rdx, %xmm1
; CHECK-NEXT:    movaps %xmm0, %xmm3
; CHECK-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm0[1,1]
; CHECK-NEXT:    cvttss2si %xmm3, %rax
; CHECK-NEXT:    subss %xmm2, %xmm3
; CHECK-NEXT:    cvttss2si %xmm3, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm3
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm3[0]
; CHECK-NEXT:    movaps %xmm0, %xmm3
; CHECK-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm0[3,3]
; CHECK-NEXT:    cvttss2si %xmm3, %rax
; CHECK-NEXT:    subss %xmm2, %xmm3
; CHECK-NEXT:    cvttss2si %xmm3, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm3
; CHECK-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    subss %xmm2, %xmm0
; CHECK-NEXT:    cvttss2si %xmm0, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
; CHECK-NEXT:    movdqa %xmm0, %xmm3
; CHECK-NEXT:    pxor %xmm2, %xmm3
; CHECK-NEXT:    movdqa {{.*#+}} xmm4 = [9223372039002259455,9223372039002259455]
; CHECK-NEXT:    movdqa %xmm4, %xmm5
; CHECK-NEXT:    pcmpgtd %xmm3, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm5
; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = [4294967295,4294967295]
; CHECK-NEXT:    pand %xmm5, %xmm0
; CHECK-NEXT:    pandn %xmm3, %xmm5
; CHECK-NEXT:    por %xmm0, %xmm5
; CHECK-NEXT:    pxor %xmm1, %xmm2
; CHECK-NEXT:    movdqa %xmm4, %xmm0
; CHECK-NEXT:    pcmpgtd %xmm2, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm0
; CHECK-NEXT:    pand %xmm0, %xmm1
; CHECK-NEXT:    pandn %xmm3, %xmm0
; CHECK-NEXT:    por %xmm1, %xmm0
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <4 x float> %x to <4 x i64>
  %spec.store.select = call <4 x i64> @llvm.umin.v4i64(<4 x i64> %conv, <4 x i64> <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>)
  %conv6 = trunc <4 x i64> %spec.store.select to <4 x i32>
  ret <4 x i32> %conv6
}

define <4 x i32> @ustest_f32i32_mm(<4 x float> %x) {
; CHECK-LABEL: ustest_f32i32_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    movaps %xmm0, %xmm1
; CHECK-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
; CHECK-NEXT:    cvttss2si %xmm1, %rax
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    movaps %xmm0, %xmm2
; CHECK-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
; CHECK-NEXT:    cvttss2si %xmm2, %rax
; CHECK-NEXT:    movq %rax, %xmm2
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm1[0]
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm3
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm3, %xmm1
; CHECK-NEXT:    pxor %xmm0, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm4 = [2147483647,2147483647]
; CHECK-NEXT:    movdqa %xmm4, %xmm5
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm7 = xmm1[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm7
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,3,3]
; CHECK-NEXT:    por %xmm7, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm5 = [4294967295,4294967295]
; CHECK-NEXT:    pand %xmm1, %xmm3
; CHECK-NEXT:    pandn %xmm5, %xmm1
; CHECK-NEXT:    por %xmm3, %xmm1
; CHECK-NEXT:    movdqa %xmm2, %xmm3
; CHECK-NEXT:    pxor %xmm0, %xmm3
; CHECK-NEXT:    movdqa %xmm4, %xmm6
; CHECK-NEXT:    pcmpgtd %xmm3, %xmm6
; CHECK-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    pand %xmm7, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[1,1,3,3]
; CHECK-NEXT:    por %xmm3, %xmm4
; CHECK-NEXT:    pand %xmm4, %xmm2
; CHECK-NEXT:    pandn %xmm5, %xmm4
; CHECK-NEXT:    por %xmm2, %xmm4
; CHECK-NEXT:    movdqa %xmm4, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa %xmm2, %xmm3
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm3
; CHECK-NEXT:    pcmpeqd %xmm0, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm3, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm3
; CHECK-NEXT:    pand %xmm4, %xmm3
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa %xmm2, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm4
; CHECK-NEXT:    pcmpeqd %xmm0, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm4, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm0
; CHECK-NEXT:    pand %xmm1, %xmm0
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm3[0,2]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <4 x float> %x to <4 x i64>
  %spec.store.select = call <4 x i64> @llvm.smin.v4i64(<4 x i64> %conv, <4 x i64> <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>)
  %spec.store.select7 = call <4 x i64> @llvm.smax.v4i64(<4 x i64> %spec.store.select, <4 x i64> zeroinitializer)
  %conv6 = trunc <4 x i64> %spec.store.select7 to <4 x i32>
  ret <4 x i32> %conv6
}

define <4 x i32> @stest_f16i32_mm(<4 x half> %x) {
; CHECK-LABEL: stest_f16i32_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    subq $32, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 64
; CHECK-NEXT:    .cfi_offset %rbx, -32
; CHECK-NEXT:    .cfi_offset %r14, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %edx, %ebp
; CHECK-NEXT:    movl %esi, %ebx
; CHECK-NEXT:    movl %edi, %r14d
; CHECK-NEXT:    movzwl %cx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq (%rsp), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movzwl %bx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r14w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm2
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm2 = xmm2[0],mem[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm2, %xmm1
; CHECK-NEXT:    pxor %xmm0, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = [4294967295,4294967295]
; CHECK-NEXT:    movdqa %xmm3, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm4
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[1,1,3,3]
; CHECK-NEXT:    pand %xmm5, %xmm6
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm6, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm4 = [2147483647,2147483647]
; CHECK-NEXT:    pand %xmm1, %xmm2
; CHECK-NEXT:    pandn %xmm4, %xmm1
; CHECK-NEXT:    por %xmm2, %xmm1
; CHECK-NEXT:    movdqa (%rsp), %xmm7 # 16-byte Reload
; CHECK-NEXT:    movdqa %xmm7, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa %xmm3, %xmm5
; CHECK-NEXT:    pcmpgtd %xmm2, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm3
; CHECK-NEXT:    movdqa %xmm7, %xmm2
; CHECK-NEXT:    pand %xmm3, %xmm2
; CHECK-NEXT:    pandn %xmm4, %xmm3
; CHECK-NEXT:    por %xmm2, %xmm3
; CHECK-NEXT:    movdqa %xmm3, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa {{.*#+}} xmm4 = [18446744069414584320,18446744069414584320]
; CHECK-NEXT:    movdqa %xmm2, %xmm5
; CHECK-NEXT:    pcmpgtd %xmm4, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm5
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [18446744071562067968,18446744071562067968]
; CHECK-NEXT:    pand %xmm5, %xmm3
; CHECK-NEXT:    pandn %xmm2, %xmm5
; CHECK-NEXT:    por %xmm3, %xmm5
; CHECK-NEXT:    pxor %xmm1, %xmm0
; CHECK-NEXT:    movdqa %xmm0, %xmm3
; CHECK-NEXT:    pcmpgtd %xmm4, %xmm3
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm4, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm4
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
; CHECK-NEXT:    por %xmm4, %xmm0
; CHECK-NEXT:    pand %xmm0, %xmm1
; CHECK-NEXT:    pandn %xmm2, %xmm0
; CHECK-NEXT:    por %xmm1, %xmm0
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
; CHECK-NEXT:    addq $32, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <4 x half> %x to <4 x i64>
  %spec.store.select = call <4 x i64> @llvm.smin.v4i64(<4 x i64> %conv, <4 x i64> <i64 2147483647, i64 2147483647, i64 2147483647, i64 2147483647>)
  %spec.store.select7 = call <4 x i64> @llvm.smax.v4i64(<4 x i64> %spec.store.select, <4 x i64> <i64 -2147483648, i64 -2147483648, i64 -2147483648, i64 -2147483648>)
  %conv6 = trunc <4 x i64> %spec.store.select7 to <4 x i32>
  ret <4 x i32> %conv6
}

define <4 x i32> @utesth_f16i32_mm(<4 x half> %x) {
; CHECK-LABEL: utesth_f16i32_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    subq $32, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 64
; CHECK-NEXT:    .cfi_offset %rbx, -32
; CHECK-NEXT:    .cfi_offset %r14, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %ecx, %ebp
; CHECK-NEXT:    movl %edx, %r14d
; CHECK-NEXT:    movl %edi, %ebx
; CHECK-NEXT:    movzwl %si, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT:    cvttss2si %xmm0, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm0
; CHECK-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movzwl %bx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT:    cvttss2si %xmm0, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm0
; CHECK-NEXT:    punpcklqdq (%rsp), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT:    cvttss2si %xmm0, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r14w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT:    cvttss2si %xmm0, %rcx
; CHECK-NEXT:    movq %rax, %rdx
; CHECK-NEXT:    sarq $63, %rdx
; CHECK-NEXT:    andq %rcx, %rdx
; CHECK-NEXT:    orq %rax, %rdx
; CHECK-NEXT:    movq %rdx, %xmm0
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = [9223372039002259456,9223372039002259456]
; CHECK-NEXT:    movdqa %xmm0, %xmm2
; CHECK-NEXT:    pxor %xmm1, %xmm2
; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = [9223372039002259455,9223372039002259455]
; CHECK-NEXT:    movdqa %xmm3, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm2, %xmm4
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm5, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm4
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [4294967295,4294967295]
; CHECK-NEXT:    pand %xmm4, %xmm0
; CHECK-NEXT:    pandn %xmm2, %xmm4
; CHECK-NEXT:    por %xmm0, %xmm4
; CHECK-NEXT:    movdqa (%rsp), %xmm6 # 16-byte Reload
; CHECK-NEXT:    pxor %xmm6, %xmm1
; CHECK-NEXT:    movdqa %xmm3, %xmm0
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm0
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; CHECK-NEXT:    pand %xmm5, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
; CHECK-NEXT:    por %xmm1, %xmm0
; CHECK-NEXT:    pand %xmm0, %xmm6
; CHECK-NEXT:    pandn %xmm2, %xmm0
; CHECK-NEXT:    por %xmm6, %xmm0
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm4[0,2]
; CHECK-NEXT:    addq $32, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <4 x half> %x to <4 x i64>
  %spec.store.select = call <4 x i64> @llvm.umin.v4i64(<4 x i64> %conv, <4 x i64> <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>)
  %conv6 = trunc <4 x i64> %spec.store.select to <4 x i32>
  ret <4 x i32> %conv6
}

define <4 x i32> @ustest_f16i32_mm(<4 x half> %x) {
; CHECK-LABEL: ustest_f16i32_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    subq $32, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 64
; CHECK-NEXT:    .cfi_offset %rbx, -32
; CHECK-NEXT:    .cfi_offset %r14, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %edx, %ebp
; CHECK-NEXT:    movl %esi, %ebx
; CHECK-NEXT:    movl %edi, %r14d
; CHECK-NEXT:    movzwl %cx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq (%rsp), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movzwl %bx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r14w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movq %rax, %xmm2
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm2 = xmm2[0],mem[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm2, %xmm1
; CHECK-NEXT:    pxor %xmm0, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = [2147483647,2147483647]
; CHECK-NEXT:    movdqa %xmm3, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm4
; CHECK-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm1
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[1,1,3,3]
; CHECK-NEXT:    pand %xmm5, %xmm6
; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm6, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm4 = [4294967295,4294967295]
; CHECK-NEXT:    pand %xmm1, %xmm2
; CHECK-NEXT:    pandn %xmm4, %xmm1
; CHECK-NEXT:    por %xmm2, %xmm1
; CHECK-NEXT:    movdqa (%rsp), %xmm7 # 16-byte Reload
; CHECK-NEXT:    movdqa %xmm7, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa %xmm3, %xmm5
; CHECK-NEXT:    pcmpgtd %xmm2, %xmm5
; CHECK-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm6, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm3
; CHECK-NEXT:    movdqa %xmm7, %xmm2
; CHECK-NEXT:    pand %xmm3, %xmm2
; CHECK-NEXT:    pandn %xmm4, %xmm3
; CHECK-NEXT:    por %xmm2, %xmm3
; CHECK-NEXT:    movdqa %xmm3, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa %xmm2, %xmm4
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm4
; CHECK-NEXT:    pcmpeqd %xmm0, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm4, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm4
; CHECK-NEXT:    pand %xmm3, %xmm4
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm2
; CHECK-NEXT:    movdqa %xmm2, %xmm3
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm3
; CHECK-NEXT:    pcmpeqd %xmm0, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; CHECK-NEXT:    pand %xmm3, %xmm2
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
; CHECK-NEXT:    por %xmm2, %xmm0
; CHECK-NEXT:    pand %xmm1, %xmm0
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm4[0,2]
; CHECK-NEXT:    addq $32, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <4 x half> %x to <4 x i64>
  %spec.store.select = call <4 x i64> @llvm.smin.v4i64(<4 x i64> %conv, <4 x i64> <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>)
  %spec.store.select7 = call <4 x i64> @llvm.smax.v4i64(<4 x i64> %spec.store.select, <4 x i64> zeroinitializer)
  %conv6 = trunc <4 x i64> %spec.store.select7 to <4 x i32>
  ret <4 x i32> %conv6
}

; i16 saturate

define <2 x i16> @stest_f64i16_mm(<2 x double> %x) {
; CHECK-LABEL: stest_f64i16_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttpd2dq %xmm0, %xmm0
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = <32767,32767,u,u>
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm2
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    pandn %xmm1, %xmm2
; CHECK-NEXT:    por %xmm0, %xmm2
; CHECK-NEXT:    movdqa {{.*#+}} xmm0 = <4294934528,4294934528,u,u>
; CHECK-NEXT:    movdqa %xmm2, %xmm1
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm1
; CHECK-NEXT:    pand %xmm1, %xmm2
; CHECK-NEXT:    pandn %xmm0, %xmm1
; CHECK-NEXT:    por %xmm2, %xmm1
; CHECK-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x double> %x to <2 x i32>
  %spec.store.select = call <2 x i32> @llvm.smin.v2i32(<2 x i32> %conv, <2 x i32> <i32 32767, i32 32767>)
  %spec.store.select7 = call <2 x i32> @llvm.smax.v2i32(<2 x i32> %spec.store.select, <2 x i32> <i32 -32768, i32 -32768>)
  %conv6 = trunc <2 x i32> %spec.store.select7 to <2 x i16>
  ret <2 x i16> %conv6
}

define <2 x i16> @utest_f64i16_mm(<2 x double> %x) {
; CHECK-LABEL: utest_f64i16_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttpd2dq %xmm0, %xmm1
; CHECK-NEXT:    movapd %xmm1, %xmm2
; CHECK-NEXT:    psrad $31, %xmm2
; CHECK-NEXT:    addpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT:    cvttpd2dq %xmm0, %xmm0
; CHECK-NEXT:    andpd %xmm2, %xmm0
; CHECK-NEXT:    orpd %xmm1, %xmm0
; CHECK-NEXT:    movapd {{.*#+}} xmm1 = [2147483648,2147483648,2147483648,2147483648]
; CHECK-NEXT:    xorpd %xmm0, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = <2147549183,2147549183,u,u>
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm2
; CHECK-NEXT:    andpd %xmm2, %xmm0
; CHECK-NEXT:    andnpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
; CHECK-NEXT:    orpd %xmm0, %xmm2
; CHECK-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,2,2,3,4,5,6,7]
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <2 x double> %x to <2 x i32>
  %spec.store.select = call <2 x i32> @llvm.umin.v2i32(<2 x i32> %conv, <2 x i32> <i32 65535, i32 65535>)
  %conv6 = trunc <2 x i32> %spec.store.select to <2 x i16>
  ret <2 x i16> %conv6
}

define <2 x i16> @ustest_f64i16_mm(<2 x double> %x) {
; CHECK-LABEL: ustest_f64i16_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttpd2dq %xmm0, %xmm0
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = <65535,65535,u,u>
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm2
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    pandn %xmm1, %xmm2
; CHECK-NEXT:    por %xmm0, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm0
; CHECK-NEXT:    movdqa %xmm2, %xmm1
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm1
; CHECK-NEXT:    pand %xmm2, %xmm1
; CHECK-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x double> %x to <2 x i32>
  %spec.store.select = call <2 x i32> @llvm.smin.v2i32(<2 x i32> %conv, <2 x i32> <i32 65535, i32 65535>)
  %spec.store.select7 = call <2 x i32> @llvm.smax.v2i32(<2 x i32> %spec.store.select, <2 x i32> zeroinitializer)
  %conv6 = trunc <2 x i32> %spec.store.select7 to <2 x i16>
  ret <2 x i16> %conv6
}

define <4 x i16> @stest_f32i16_mm(<4 x float> %x) {
; CHECK-LABEL: stest_f32i16_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttps2dq %xmm0, %xmm0
; CHECK-NEXT:    packssdw %xmm0, %xmm0
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <4 x float> %x to <4 x i32>
  %spec.store.select = call <4 x i32> @llvm.smin.v4i32(<4 x i32> %conv, <4 x i32> <i32 32767, i32 32767, i32 32767, i32 32767>)
  %spec.store.select7 = call <4 x i32> @llvm.smax.v4i32(<4 x i32> %spec.store.select, <4 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768>)
  %conv6 = trunc <4 x i32> %spec.store.select7 to <4 x i16>
  ret <4 x i16> %conv6
}

define <4 x i16> @utest_f32i16_mm(<4 x float> %x) {
; CHECK-LABEL: utest_f32i16_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttps2dq %xmm0, %xmm1
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    psrad $31, %xmm2
; CHECK-NEXT:    subps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT:    cvttps2dq %xmm0, %xmm0
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    por %xmm1, %xmm0
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = [2147483648,2147483648,2147483648,2147483648]
; CHECK-NEXT:    pxor %xmm0, %xmm1
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [2147549183,2147549183,2147549183,2147549183]
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm2
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    pandn {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
; CHECK-NEXT:    por %xmm0, %xmm2
; CHECK-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,2,2,3,4,5,6,7]
; CHECK-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <4 x float> %x to <4 x i32>
  %spec.store.select = call <4 x i32> @llvm.umin.v4i32(<4 x i32> %conv, <4 x i32> <i32 65535, i32 65535, i32 65535, i32 65535>)
  %conv6 = trunc <4 x i32> %spec.store.select to <4 x i16>
  ret <4 x i16> %conv6
}

define <4 x i16> @ustest_f32i16_mm(<4 x float> %x) {
; CHECK-LABEL: ustest_f32i16_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    cvttps2dq %xmm0, %xmm0
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535]
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm2
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    pandn %xmm1, %xmm2
; CHECK-NEXT:    por %xmm0, %xmm2
; CHECK-NEXT:    pxor %xmm0, %xmm0
; CHECK-NEXT:    movdqa %xmm2, %xmm1
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm1
; CHECK-NEXT:    pand %xmm2, %xmm1
; CHECK-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
; CHECK-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
; CHECK-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <4 x float> %x to <4 x i32>
  %spec.store.select = call <4 x i32> @llvm.smin.v4i32(<4 x i32> %conv, <4 x i32> <i32 65535, i32 65535, i32 65535, i32 65535>)
  %spec.store.select7 = call <4 x i32> @llvm.smax.v4i32(<4 x i32> %spec.store.select, <4 x i32> zeroinitializer)
  %conv6 = trunc <4 x i32> %spec.store.select7 to <4 x i16>
  ret <4 x i16> %conv6
}

define <8 x i16> @stest_f16i16_mm(<8 x half> %x) {
; CHECK-LABEL: stest_f16i16_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r15
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    pushq %r13
; CHECK-NEXT:    .cfi_def_cfa_offset 40
; CHECK-NEXT:    pushq %r12
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 56
; CHECK-NEXT:    subq $72, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 128
; CHECK-NEXT:    .cfi_offset %rbx, -56
; CHECK-NEXT:    .cfi_offset %r12, -48
; CHECK-NEXT:    .cfi_offset %r13, -40
; CHECK-NEXT:    .cfi_offset %r14, -32
; CHECK-NEXT:    .cfi_offset %r15, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %r9d, %ebp
; CHECK-NEXT:    movl %r8d, %ebx
; CHECK-NEXT:    movl %ecx, %r13d
; CHECK-NEXT:    movl %edx, %r12d
; CHECK-NEXT:    movl %esi, %r15d
; CHECK-NEXT:    movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT:    movzwl {{[0-9]+}}(%rsp), %r14d
; CHECK-NEXT:    movzwl {{[0-9]+}}(%rsp), %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movl %r14d, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %bx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r13w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r12w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r15w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl {{[-0-9]+}}(%r{{[sb]}}p), %edi # 2-byte Folded Reload
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    packssdw {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    addq $72, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 56
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    popq %r12
; CHECK-NEXT:    .cfi_def_cfa_offset 40
; CHECK-NEXT:    popq %r13
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r15
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <8 x half> %x to <8 x i32>
  %spec.store.select = call <8 x i32> @llvm.smin.v8i32(<8 x i32> %conv, <8 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>)
  %spec.store.select7 = call <8 x i32> @llvm.smax.v8i32(<8 x i32> %spec.store.select, <8 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>)
  %conv6 = trunc <8 x i32> %spec.store.select7 to <8 x i16>
  ret <8 x i16> %conv6
}

define <8 x i16> @utesth_f16i16_mm(<8 x half> %x) {
; CHECK-LABEL: utesth_f16i16_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r15
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    pushq %r13
; CHECK-NEXT:    .cfi_def_cfa_offset 40
; CHECK-NEXT:    pushq %r12
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 56
; CHECK-NEXT:    subq $72, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 128
; CHECK-NEXT:    .cfi_offset %rbx, -56
; CHECK-NEXT:    .cfi_offset %r12, -48
; CHECK-NEXT:    .cfi_offset %r13, -40
; CHECK-NEXT:    .cfi_offset %r14, -32
; CHECK-NEXT:    .cfi_offset %r15, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %r9d, %ebp
; CHECK-NEXT:    movl %r8d, %ebx
; CHECK-NEXT:    movl %ecx, %r13d
; CHECK-NEXT:    movl %edx, %r12d
; CHECK-NEXT:    movl %esi, %r15d
; CHECK-NEXT:    movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT:    movzwl {{[0-9]+}}(%rsp), %r14d
; CHECK-NEXT:    movzwl {{[0-9]+}}(%rsp), %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movl %r14d, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %bx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r13w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r12w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r15w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl {{[-0-9]+}}(%r{{[sb]}}p), %edi # 2-byte Folded Reload
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %rax
; CHECK-NEXT:    movd %eax, %xmm1
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm1 = xmm1[0],mem[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648,2147483648,2147483648]
; CHECK-NEXT:    movdqa %xmm1, %xmm3
; CHECK-NEXT:    pxor %xmm2, %xmm3
; CHECK-NEXT:    movdqa {{.*#+}} xmm4 = [2147549183,2147549183,2147549183,2147549183]
; CHECK-NEXT:    movdqa %xmm4, %xmm0
; CHECK-NEXT:    pcmpgtd %xmm3, %xmm0
; CHECK-NEXT:    pand %xmm0, %xmm1
; CHECK-NEXT:    pcmpeqd %xmm3, %xmm3
; CHECK-NEXT:    pxor %xmm3, %xmm0
; CHECK-NEXT:    por %xmm1, %xmm0
; CHECK-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; CHECK-NEXT:    pxor %xmm1, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm2, %xmm4
; CHECK-NEXT:    pand %xmm4, %xmm1
; CHECK-NEXT:    pxor %xmm3, %xmm4
; CHECK-NEXT:    por %xmm1, %xmm4
; CHECK-NEXT:    pslld $16, %xmm4
; CHECK-NEXT:    psrad $16, %xmm4
; CHECK-NEXT:    pslld $16, %xmm0
; CHECK-NEXT:    psrad $16, %xmm0
; CHECK-NEXT:    packssdw %xmm4, %xmm0
; CHECK-NEXT:    addq $72, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 56
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    popq %r12
; CHECK-NEXT:    .cfi_def_cfa_offset 40
; CHECK-NEXT:    popq %r13
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r15
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <8 x half> %x to <8 x i32>
  %spec.store.select = call <8 x i32> @llvm.umin.v8i32(<8 x i32> %conv, <8 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>)
  %conv6 = trunc <8 x i32> %spec.store.select to <8 x i16>
  ret <8 x i16> %conv6
}

define <8 x i16> @ustest_f16i16_mm(<8 x half> %x) {
; CHECK-LABEL: ustest_f16i16_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r15
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    pushq %r13
; CHECK-NEXT:    .cfi_def_cfa_offset 40
; CHECK-NEXT:    pushq %r12
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 56
; CHECK-NEXT:    subq $72, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 128
; CHECK-NEXT:    .cfi_offset %rbx, -56
; CHECK-NEXT:    .cfi_offset %r12, -48
; CHECK-NEXT:    .cfi_offset %r13, -40
; CHECK-NEXT:    .cfi_offset %r14, -32
; CHECK-NEXT:    .cfi_offset %r15, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %r9d, %r15d
; CHECK-NEXT:    movl %r8d, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT:    movl %edx, %ebp
; CHECK-NEXT:    movl %esi, %r14d
; CHECK-NEXT:    movl %edi, %ebx
; CHECK-NEXT:    movzwl {{[0-9]+}}(%rsp), %r12d
; CHECK-NEXT:    movzwl {{[0-9]+}}(%rsp), %r13d
; CHECK-NEXT:    movzwl %cx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r14w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %bx, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movl %r13d, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movl %r12d, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl %r15w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT:    movzwl {{[-0-9]+}}(%r{{[sb]}}p), %edi # 2-byte Folded Reload
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    cvttss2si %xmm0, %eax
; CHECK-NEXT:    movd %eax, %xmm0
; CHECK-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT:    # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535]
; CHECK-NEXT:    movdqa %xmm1, %xmm2
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm2
; CHECK-NEXT:    pand %xmm2, %xmm0
; CHECK-NEXT:    pandn %xmm1, %xmm2
; CHECK-NEXT:    por %xmm0, %xmm2
; CHECK-NEXT:    movdqa %xmm1, %xmm3
; CHECK-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT:    pcmpgtd %xmm0, %xmm3
; CHECK-NEXT:    pand %xmm3, %xmm0
; CHECK-NEXT:    pandn %xmm1, %xmm3
; CHECK-NEXT:    por %xmm0, %xmm3
; CHECK-NEXT:    pxor %xmm1, %xmm1
; CHECK-NEXT:    movdqa %xmm3, %xmm0
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm0
; CHECK-NEXT:    pand %xmm3, %xmm0
; CHECK-NEXT:    movdqa %xmm2, %xmm3
; CHECK-NEXT:    pcmpgtd %xmm1, %xmm3
; CHECK-NEXT:    pand %xmm2, %xmm3
; CHECK-NEXT:    pslld $16, %xmm3
; CHECK-NEXT:    psrad $16, %xmm3
; CHECK-NEXT:    pslld $16, %xmm0
; CHECK-NEXT:    psrad $16, %xmm0
; CHECK-NEXT:    packssdw %xmm3, %xmm0
; CHECK-NEXT:    addq $72, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 56
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    popq %r12
; CHECK-NEXT:    .cfi_def_cfa_offset 40
; CHECK-NEXT:    popq %r13
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r15
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <8 x half> %x to <8 x i32>
  %spec.store.select = call <8 x i32> @llvm.smin.v8i32(<8 x i32> %conv, <8 x i32> <i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535, i32 65535>)
  %spec.store.select7 = call <8 x i32> @llvm.smax.v8i32(<8 x i32> %spec.store.select, <8 x i32> zeroinitializer)
  %conv6 = trunc <8 x i32> %spec.store.select7 to <8 x i16>
  ret <8 x i16> %conv6
}

; i64 saturate

define <2 x i64> @stest_f64i64_mm(<2 x double> %x) {
; CHECK-LABEL: stest_f64i64_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    subq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset %rbx, -24
; CHECK-NEXT:    .cfi_offset %r14, -16
; CHECK-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    callq __fixdfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %r14
; CHECK-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT:    callq __fixdfti@PLT
; CHECK-NEXT:    movabsq $9223372036854775807, %rcx # imm = 0x7FFFFFFFFFFFFFFF
; CHECK-NEXT:    cmpq %rcx, %rax
; CHECK-NEXT:    movq %rcx, %rsi
; CHECK-NEXT:    cmovbq %rax, %rsi
; CHECK-NEXT:    xorl %edi, %edi
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    cmovnsq %rcx, %rax
; CHECK-NEXT:    cmoveq %rsi, %rax
; CHECK-NEXT:    cmovnsq %rdi, %rdx
; CHECK-NEXT:    cmpq %rcx, %rbx
; CHECK-NEXT:    movq %rcx, %rsi
; CHECK-NEXT:    cmovbq %rbx, %rsi
; CHECK-NEXT:    testq %r14, %r14
; CHECK-NEXT:    cmovsq %rbx, %rcx
; CHECK-NEXT:    cmoveq %rsi, %rcx
; CHECK-NEXT:    cmovsq %r14, %rdi
; CHECK-NEXT:    testq %rdi, %rdi
; CHECK-NEXT:    movabsq $-9223372036854775808, %rbx # imm = 0x8000000000000000
; CHECK-NEXT:    movq %rbx, %rsi
; CHECK-NEXT:    cmovnsq %rcx, %rsi
; CHECK-NEXT:    cmpq %rbx, %rcx
; CHECK-NEXT:    cmovbeq %rbx, %rcx
; CHECK-NEXT:    cmpq $-1, %rdi
; CHECK-NEXT:    cmovneq %rsi, %rcx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    movq %rbx, %rsi
; CHECK-NEXT:    cmovnsq %rax, %rsi
; CHECK-NEXT:    cmpq %rbx, %rax
; CHECK-NEXT:    cmovbeq %rbx, %rax
; CHECK-NEXT:    cmpq $-1, %rdx
; CHECK-NEXT:    cmovneq %rsi, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movq %rcx, %xmm1
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    addq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x double> %x to <2 x i128>
  %spec.store.select = call <2 x i128> @llvm.smin.v2i128(<2 x i128> %conv, <2 x i128> <i128 9223372036854775807, i128 9223372036854775807>)
  %spec.store.select7 = call <2 x i128> @llvm.smax.v2i128(<2 x i128> %spec.store.select, <2 x i128> <i128 -9223372036854775808, i128 -9223372036854775808>)
  %conv6 = trunc <2 x i128> %spec.store.select7 to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @utest_f64i64_mm(<2 x double> %x) {
; CHECK-LABEL: utest_f64i64_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    subq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset %rbx, -24
; CHECK-NEXT:    .cfi_offset %r14, -16
; CHECK-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    callq __fixunsdfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %r14
; CHECK-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    callq __fixunsdfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    cmovneq %rcx, %rax
; CHECK-NEXT:    cmpq $1, %rdx
; CHECK-NEXT:    cmoveq %rcx, %rax
; CHECK-NEXT:    testq %r14, %r14
; CHECK-NEXT:    cmovneq %rcx, %rbx
; CHECK-NEXT:    cmpq $1, %r14
; CHECK-NEXT:    cmoveq %rcx, %rbx
; CHECK-NEXT:    movq %rbx, %xmm0
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    addq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <2 x double> %x to <2 x i128>
  %spec.store.select = call <2 x i128> @llvm.umin.v2i128(<2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>)
  %conv6 = trunc <2 x i128> %spec.store.select to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @ustest_f64i64_mm(<2 x double> %x) {
; CHECK-LABEL: ustest_f64i64_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    subq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset %rbx, -24
; CHECK-NEXT:    .cfi_offset %r14, -16
; CHECK-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT:    callq __fixdfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %r14
; CHECK-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT:    callq __fixdfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    movl $1, %esi
; CHECK-NEXT:    movl $1, %edi
; CHECK-NEXT:    cmovleq %rdx, %rdi
; CHECK-NEXT:    cmovgq %rcx, %rax
; CHECK-NEXT:    cmpq $1, %rdx
; CHECK-NEXT:    cmoveq %rcx, %rax
; CHECK-NEXT:    testq %r14, %r14
; CHECK-NEXT:    cmovleq %r14, %rsi
; CHECK-NEXT:    cmovgq %rcx, %rbx
; CHECK-NEXT:    cmpq $1, %r14
; CHECK-NEXT:    cmoveq %rcx, %rbx
; CHECK-NEXT:    testq %rsi, %rsi
; CHECK-NEXT:    cmovsq %rcx, %rbx
; CHECK-NEXT:    testq %rdi, %rdi
; CHECK-NEXT:    cmovsq %rcx, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movq %rbx, %xmm1
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    addq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x double> %x to <2 x i128>
  %spec.store.select = call <2 x i128> @llvm.smin.v2i128(<2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>)
  %spec.store.select7 = call <2 x i128> @llvm.smax.v2i128(<2 x i128> %spec.store.select, <2 x i128> zeroinitializer)
  %conv6 = trunc <2 x i128> %spec.store.select7 to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @stest_f32i64_mm(<2 x float> %x) {
; CHECK-LABEL: stest_f32i64_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    subq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset %rbx, -24
; CHECK-NEXT:    .cfi_offset %r14, -16
; CHECK-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %r14
; CHECK-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    movabsq $9223372036854775807, %rcx # imm = 0x7FFFFFFFFFFFFFFF
; CHECK-NEXT:    cmpq %rcx, %rax
; CHECK-NEXT:    movq %rcx, %rsi
; CHECK-NEXT:    cmovbq %rax, %rsi
; CHECK-NEXT:    xorl %edi, %edi
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    cmovnsq %rcx, %rax
; CHECK-NEXT:    cmoveq %rsi, %rax
; CHECK-NEXT:    cmovnsq %rdi, %rdx
; CHECK-NEXT:    cmpq %rcx, %rbx
; CHECK-NEXT:    movq %rcx, %rsi
; CHECK-NEXT:    cmovbq %rbx, %rsi
; CHECK-NEXT:    testq %r14, %r14
; CHECK-NEXT:    cmovsq %rbx, %rcx
; CHECK-NEXT:    cmoveq %rsi, %rcx
; CHECK-NEXT:    cmovsq %r14, %rdi
; CHECK-NEXT:    testq %rdi, %rdi
; CHECK-NEXT:    movabsq $-9223372036854775808, %rbx # imm = 0x8000000000000000
; CHECK-NEXT:    movq %rbx, %rsi
; CHECK-NEXT:    cmovnsq %rcx, %rsi
; CHECK-NEXT:    cmpq %rbx, %rcx
; CHECK-NEXT:    cmovbeq %rbx, %rcx
; CHECK-NEXT:    cmpq $-1, %rdi
; CHECK-NEXT:    cmovneq %rsi, %rcx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    movq %rbx, %rsi
; CHECK-NEXT:    cmovnsq %rax, %rsi
; CHECK-NEXT:    cmpq %rbx, %rax
; CHECK-NEXT:    cmovbeq %rbx, %rax
; CHECK-NEXT:    cmpq $-1, %rdx
; CHECK-NEXT:    cmovneq %rsi, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movq %rcx, %xmm1
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    addq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x float> %x to <2 x i128>
  %spec.store.select = call <2 x i128> @llvm.smin.v2i128(<2 x i128> %conv, <2 x i128> <i128 9223372036854775807, i128 9223372036854775807>)
  %spec.store.select7 = call <2 x i128> @llvm.smax.v2i128(<2 x i128> %spec.store.select, <2 x i128> <i128 -9223372036854775808, i128 -9223372036854775808>)
  %conv6 = trunc <2 x i128> %spec.store.select7 to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @utest_f32i64_mm(<2 x float> %x) {
; CHECK-LABEL: utest_f32i64_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    subq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset %rbx, -24
; CHECK-NEXT:    .cfi_offset %r14, -16
; CHECK-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    callq __fixunssfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %r14
; CHECK-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT:    callq __fixunssfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    cmovneq %rcx, %rax
; CHECK-NEXT:    cmpq $1, %rdx
; CHECK-NEXT:    cmoveq %rcx, %rax
; CHECK-NEXT:    testq %r14, %r14
; CHECK-NEXT:    cmovneq %rcx, %rbx
; CHECK-NEXT:    cmpq $1, %r14
; CHECK-NEXT:    cmoveq %rcx, %rbx
; CHECK-NEXT:    movq %rbx, %xmm0
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    addq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <2 x float> %x to <2 x i128>
  %spec.store.select = call <2 x i128> @llvm.umin.v2i128(<2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>)
  %conv6 = trunc <2 x i128> %spec.store.select to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @ustest_f32i64_mm(<2 x float> %x) {
; CHECK-LABEL: ustest_f32i64_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    subq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 48
; CHECK-NEXT:    .cfi_offset %rbx, -24
; CHECK-NEXT:    .cfi_offset %r14, -16
; CHECK-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %r14
; CHECK-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    movl $1, %esi
; CHECK-NEXT:    movl $1, %edi
; CHECK-NEXT:    cmovleq %rdx, %rdi
; CHECK-NEXT:    cmovgq %rcx, %rax
; CHECK-NEXT:    cmpq $1, %rdx
; CHECK-NEXT:    cmoveq %rcx, %rax
; CHECK-NEXT:    testq %r14, %r14
; CHECK-NEXT:    cmovleq %r14, %rsi
; CHECK-NEXT:    cmovgq %rcx, %rbx
; CHECK-NEXT:    cmpq $1, %r14
; CHECK-NEXT:    cmoveq %rcx, %rbx
; CHECK-NEXT:    testq %rsi, %rsi
; CHECK-NEXT:    cmovsq %rcx, %rbx
; CHECK-NEXT:    testq %rdi, %rdi
; CHECK-NEXT:    cmovsq %rcx, %rax
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    movq %rbx, %xmm1
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    addq $24, %rsp
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x float> %x to <2 x i128>
  %spec.store.select = call <2 x i128> @llvm.smin.v2i128(<2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>)
  %spec.store.select7 = call <2 x i128> @llvm.smax.v2i128(<2 x i128> %spec.store.select, <2 x i128> zeroinitializer)
  %conv6 = trunc <2 x i128> %spec.store.select7 to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @stest_f16i64_mm(<2 x half> %x) {
; CHECK-LABEL: stest_f16i64_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    .cfi_offset %rbx, -32
; CHECK-NEXT:    .cfi_offset %r14, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %esi, %ebp
; CHECK-NEXT:    movzwl %di, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %r14
; CHECK-NEXT:    movzwl %bp, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    movabsq $9223372036854775807, %rcx # imm = 0x7FFFFFFFFFFFFFFF
; CHECK-NEXT:    cmpq %rcx, %rax
; CHECK-NEXT:    movq %rcx, %rsi
; CHECK-NEXT:    cmovbq %rax, %rsi
; CHECK-NEXT:    xorl %edi, %edi
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    cmovnsq %rcx, %rax
; CHECK-NEXT:    cmoveq %rsi, %rax
; CHECK-NEXT:    cmovnsq %rdi, %rdx
; CHECK-NEXT:    cmpq %rcx, %rbx
; CHECK-NEXT:    movq %rcx, %rsi
; CHECK-NEXT:    cmovbq %rbx, %rsi
; CHECK-NEXT:    testq %r14, %r14
; CHECK-NEXT:    cmovsq %rbx, %rcx
; CHECK-NEXT:    cmoveq %rsi, %rcx
; CHECK-NEXT:    cmovsq %r14, %rdi
; CHECK-NEXT:    testq %rdi, %rdi
; CHECK-NEXT:    movabsq $-9223372036854775808, %rbp # imm = 0x8000000000000000
; CHECK-NEXT:    movq %rbp, %rsi
; CHECK-NEXT:    cmovnsq %rcx, %rsi
; CHECK-NEXT:    cmpq %rbp, %rcx
; CHECK-NEXT:    cmovbeq %rbp, %rcx
; CHECK-NEXT:    cmpq $-1, %rdi
; CHECK-NEXT:    cmovneq %rsi, %rcx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    movq %rbp, %rsi
; CHECK-NEXT:    cmovnsq %rax, %rsi
; CHECK-NEXT:    cmpq %rbp, %rax
; CHECK-NEXT:    cmovbeq %rbp, %rax
; CHECK-NEXT:    cmpq $-1, %rdx
; CHECK-NEXT:    cmovneq %rsi, %rax
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    movq %rcx, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x half> %x to <2 x i128>
  %spec.store.select = call <2 x i128> @llvm.smin.v2i128(<2 x i128> %conv, <2 x i128> <i128 9223372036854775807, i128 9223372036854775807>)
  %spec.store.select7 = call <2 x i128> @llvm.smax.v2i128(<2 x i128> %spec.store.select, <2 x i128> <i128 -9223372036854775808, i128 -9223372036854775808>)
  %conv6 = trunc <2 x i128> %spec.store.select7 to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @utesth_f16i64_mm(<2 x half> %x) {
; CHECK-LABEL: utesth_f16i64_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    .cfi_offset %rbx, -32
; CHECK-NEXT:    .cfi_offset %r14, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %edi, %r14d
; CHECK-NEXT:    movzwl %si, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    callq __fixunssfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %rbp
; CHECK-NEXT:    movzwl %r14w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    callq __fixunssfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    cmovneq %rcx, %rax
; CHECK-NEXT:    cmpq $1, %rdx
; CHECK-NEXT:    cmoveq %rcx, %rax
; CHECK-NEXT:    testq %rbp, %rbp
; CHECK-NEXT:    cmovneq %rcx, %rbx
; CHECK-NEXT:    cmpq $1, %rbp
; CHECK-NEXT:    cmoveq %rcx, %rbx
; CHECK-NEXT:    movq %rbx, %xmm1
; CHECK-NEXT:    movq %rax, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptoui <2 x half> %x to <2 x i128>
  %spec.store.select = call <2 x i128> @llvm.umin.v2i128(<2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>)
  %conv6 = trunc <2 x i128> %spec.store.select to <2 x i64>
  ret <2 x i64> %conv6
}

define <2 x i64> @ustest_f16i64_mm(<2 x half> %x) {
; CHECK-LABEL: ustest_f16i64_mm:
; CHECK:       # %bb.0: # %entry
; CHECK-NEXT:    pushq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    pushq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    pushq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 32
; CHECK-NEXT:    .cfi_offset %rbx, -32
; CHECK-NEXT:    .cfi_offset %r14, -24
; CHECK-NEXT:    .cfi_offset %rbp, -16
; CHECK-NEXT:    movl %esi, %r14d
; CHECK-NEXT:    movzwl %di, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    movq %rax, %rbx
; CHECK-NEXT:    movq %rdx, %rbp
; CHECK-NEXT:    movzwl %r14w, %edi
; CHECK-NEXT:    callq __gnu_h2f_ieee@PLT
; CHECK-NEXT:    callq __fixsfti@PLT
; CHECK-NEXT:    xorl %ecx, %ecx
; CHECK-NEXT:    testq %rdx, %rdx
; CHECK-NEXT:    movl $1, %esi
; CHECK-NEXT:    movl $1, %edi
; CHECK-NEXT:    cmovleq %rdx, %rdi
; CHECK-NEXT:    cmovgq %rcx, %rax
; CHECK-NEXT:    cmpq $1, %rdx
; CHECK-NEXT:    cmoveq %rcx, %rax
; CHECK-NEXT:    testq %rbp, %rbp
; CHECK-NEXT:    cmovleq %rbp, %rsi
; CHECK-NEXT:    cmovgq %rcx, %rbx
; CHECK-NEXT:    cmpq $1, %rbp
; CHECK-NEXT:    cmoveq %rcx, %rbx
; CHECK-NEXT:    testq %rsi, %rsi
; CHECK-NEXT:    cmovsq %rcx, %rbx
; CHECK-NEXT:    testq %rdi, %rdi
; CHECK-NEXT:    cmovsq %rcx, %rax
; CHECK-NEXT:    movq %rax, %xmm1
; CHECK-NEXT:    movq %rbx, %xmm0
; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-NEXT:    popq %rbx
; CHECK-NEXT:    .cfi_def_cfa_offset 24
; CHECK-NEXT:    popq %r14
; CHECK-NEXT:    .cfi_def_cfa_offset 16
; CHECK-NEXT:    popq %rbp
; CHECK-NEXT:    .cfi_def_cfa_offset 8
; CHECK-NEXT:    retq
entry:
  %conv = fptosi <2 x half> %x to <2 x i128>
  %spec.store.select = call <2 x i128> @llvm.smin.v2i128(<2 x i128> %conv, <2 x i128> <i128 18446744073709551616, i128 18446744073709551616>)
  %spec.store.select7 = call <2 x i128> @llvm.smax.v2i128(<2 x i128> %spec.store.select, <2 x i128> zeroinitializer)
  %conv6 = trunc <2 x i128> %spec.store.select7 to <2 x i64>
  ret <2 x i64> %conv6
}

declare <2 x i32> @llvm.smin.v2i32(<2 x i32>, <2 x i32>)
declare <2 x i32> @llvm.smax.v2i32(<2 x i32>, <2 x i32>)
declare <2 x i32> @llvm.umin.v2i32(<2 x i32>, <2 x i32>)
declare <4 x i32> @llvm.smin.v4i32(<4 x i32>, <4 x i32>)
declare <4 x i32> @llvm.smax.v4i32(<4 x i32>, <4 x i32>)
declare <4 x i32> @llvm.umin.v4i32(<4 x i32>, <4 x i32>)
declare <8 x i32> @llvm.smin.v8i32(<8 x i32>, <8 x i32>)
declare <8 x i32> @llvm.smax.v8i32(<8 x i32>, <8 x i32>)
declare <8 x i32> @llvm.umin.v8i32(<8 x i32>, <8 x i32>)
declare <2 x i64> @llvm.smin.v2i64(<2 x i64>, <2 x i64>)
declare <2 x i64> @llvm.smax.v2i64(<2 x i64>, <2 x i64>)
declare <2 x i64> @llvm.umin.v2i64(<2 x i64>, <2 x i64>)
declare <4 x i64> @llvm.smin.v4i64(<4 x i64>, <4 x i64>)
declare <4 x i64> @llvm.smax.v4i64(<4 x i64>, <4 x i64>)
declare <4 x i64> @llvm.umin.v4i64(<4 x i64>, <4 x i64>)
declare <2 x i128> @llvm.smin.v2i128(<2 x i128>, <2 x i128>)
declare <2 x i128> @llvm.smax.v2i128(<2 x i128>, <2 x i128>)
declare <2 x i128> @llvm.umin.v2i128(<2 x i128>, <2 x i128>)