; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-linux -mattr=+sse2 -mcpu=nehalem | FileCheck %s --check-prefixes=SSE,SSE-LINUX
; RUN: llc < %s -mtriple=x86_64-win32 -mattr=+sse2 -mcpu=nehalem | FileCheck %s --check-prefixes=SSE,SSE-WIN
; RUN: llc < %s -mtriple=x86_64-win32 -mattr=+avx -mcpu=corei7-avx | FileCheck %s --check-prefixes=AVX,AVX1
; RUN: llc < %s -mtriple=x86_64-win32 -mattr=+avx512vl -mcpu=skx | FileCheck %s --check-prefixes=AVX,AVX512VL

define dso_local double @t1(float* nocapture %x) nounwind readonly ssp {
; SSE-LABEL: t1:
; SSE:       # %bb.0: # %entry
; SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; SSE-NEXT:    cvtss2sd %xmm0, %xmm0
; SSE-NEXT:    retq
;
; AVX-LABEL: t1:
; AVX:       # %bb.0: # %entry
; AVX-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; AVX-NEXT:    vcvtss2sd %xmm0, %xmm0, %xmm0
; AVX-NEXT:    retq
entry:

  %0 = load float, float* %x, align 4
  %1 = fpext float %0 to double
  ret double %1
}

define dso_local float @t2(double* nocapture %x) nounwind readonly ssp optsize {
; SSE-LINUX-LABEL: t2:
; SSE-LINUX:       # %bb.0: # %entry
; SSE-LINUX-NEXT:    cvtsd2ss (%rdi), %xmm0
; SSE-LINUX-NEXT:    retq
;
; SSE-WIN-LABEL: t2:
; SSE-WIN:       # %bb.0: # %entry
; SSE-WIN-NEXT:    cvtsd2ss (%rcx), %xmm0
; SSE-WIN-NEXT:    retq
;
; AVX-LABEL: t2:
; AVX:       # %bb.0: # %entry
; AVX-NEXT:    vcvtsd2ss (%rcx), %xmm0, %xmm0
; AVX-NEXT:    retq
entry:
  %0 = load double, double* %x, align 8
  %1 = fptrunc double %0 to float
  ret float %1
}

define dso_local float @squirtf(float* %x) nounwind {
; SSE-LABEL: squirtf:
; SSE:       # %bb.0: # %entry
; SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; SSE-NEXT:    sqrtss %xmm0, %xmm0
; SSE-NEXT:    retq
;
; AVX-LABEL: squirtf:
; AVX:       # %bb.0: # %entry
; AVX-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; AVX-NEXT:    vsqrtss %xmm0, %xmm0, %xmm0
; AVX-NEXT:    retq
entry:
  %z = load float, float* %x
  %t = call float @llvm.sqrt.f32(float %z)
  ret float %t
}

define dso_local double @squirt(double* %x) nounwind {
; SSE-LABEL: squirt:
; SSE:       # %bb.0: # %entry
; SSE-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
; SSE-NEXT:    sqrtsd %xmm0, %xmm0
; SSE-NEXT:    retq
;
; AVX-LABEL: squirt:
; AVX:       # %bb.0: # %entry
; AVX-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
; AVX-NEXT:    vsqrtsd %xmm0, %xmm0, %xmm0
; AVX-NEXT:    retq
entry:
  %z = load double, double* %x
  %t = call double @llvm.sqrt.f64(double %z)
  ret double %t
}

define dso_local float @squirtf_size(float* %x) nounwind optsize {
; SSE-LINUX-LABEL: squirtf_size:
; SSE-LINUX:       # %bb.0: # %entry
; SSE-LINUX-NEXT:    sqrtss (%rdi), %xmm0
; SSE-LINUX-NEXT:    retq
;
; SSE-WIN-LABEL: squirtf_size:
; SSE-WIN:       # %bb.0: # %entry
; SSE-WIN-NEXT:    sqrtss (%rcx), %xmm0
; SSE-WIN-NEXT:    retq
;
; AVX-LABEL: squirtf_size:
; AVX:       # %bb.0: # %entry
; AVX-NEXT:    vsqrtss (%rcx), %xmm0, %xmm0
; AVX-NEXT:    retq
entry:
  %z = load float, float* %x
  %t = call float @llvm.sqrt.f32(float %z)
  ret float %t
}

define dso_local double @squirt_size(double* %x) nounwind optsize {
; SSE-LINUX-LABEL: squirt_size:
; SSE-LINUX:       # %bb.0: # %entry
; SSE-LINUX-NEXT:    sqrtsd (%rdi), %xmm0
; SSE-LINUX-NEXT:    retq
;
; SSE-WIN-LABEL: squirt_size:
; SSE-WIN:       # %bb.0: # %entry
; SSE-WIN-NEXT:    sqrtsd (%rcx), %xmm0
; SSE-WIN-NEXT:    retq
;
; AVX-LABEL: squirt_size:
; AVX:       # %bb.0: # %entry
; AVX-NEXT:    vsqrtsd (%rcx), %xmm0, %xmm0
; AVX-NEXT:    retq
entry:
  %z = load double, double* %x
  %t = call double @llvm.sqrt.f64(double %z)
  ret double %t
}

declare float @llvm.sqrt.f32(float)
declare double @llvm.sqrt.f64(double)

; This loop contains two cvtsi2ss instructions that update the same xmm
; register.  Verify that the break false dependency fix pass breaks those
; dependencies by inserting xorps instructions.
;
define dso_local float @loopdep1(i32 %m) nounwind uwtable readnone ssp {
; SSE-LINUX-LABEL: loopdep1:
; SSE-LINUX:       # %bb.0: # %entry
; SSE-LINUX-NEXT:    testl %edi, %edi
; SSE-LINUX-NEXT:    je .LBB6_1
; SSE-LINUX-NEXT:  # %bb.2: # %for.body.preheader
; SSE-LINUX-NEXT:    movl $1, %eax
; SSE-LINUX-NEXT:    xorps %xmm0, %xmm0
; SSE-LINUX-NEXT:    xorps %xmm1, %xmm1
; SSE-LINUX-NEXT:    .p2align 4, 0x90
; SSE-LINUX-NEXT:  .LBB6_3: # %for.body
; SSE-LINUX-NEXT:    # =>This Inner Loop Header: Depth=1
; SSE-LINUX-NEXT:    xorps %xmm2, %xmm2
; SSE-LINUX-NEXT:    cvtsi2ss %eax, %xmm2
; SSE-LINUX-NEXT:    xorps %xmm3, %xmm3
; SSE-LINUX-NEXT:    cvtsi2ss %edi, %xmm3
; SSE-LINUX-NEXT:    addss %xmm2, %xmm0
; SSE-LINUX-NEXT:    addss %xmm3, %xmm1
; SSE-LINUX-NEXT:    incl %eax
; SSE-LINUX-NEXT:    decl %edi
; SSE-LINUX-NEXT:    jne .LBB6_3
; SSE-LINUX-NEXT:  # %bb.4: # %for.end
; SSE-LINUX-NEXT:    subss %xmm1, %xmm0
; SSE-LINUX-NEXT:    retq
; SSE-LINUX-NEXT:  .LBB6_1:
; SSE-LINUX-NEXT:    xorps %xmm0, %xmm0
; SSE-LINUX-NEXT:    xorps %xmm1, %xmm1
; SSE-LINUX-NEXT:    subss %xmm1, %xmm0
; SSE-LINUX-NEXT:    retq
;
; SSE-WIN-LABEL: loopdep1:
; SSE-WIN:       # %bb.0: # %entry
; SSE-WIN-NEXT:    testl %ecx, %ecx
; SSE-WIN-NEXT:    je .LBB6_1
; SSE-WIN-NEXT:  # %bb.2: # %for.body.preheader
; SSE-WIN-NEXT:    movl $1, %eax
; SSE-WIN-NEXT:    xorps %xmm0, %xmm0
; SSE-WIN-NEXT:    xorps %xmm1, %xmm1
; SSE-WIN-NEXT:    .p2align 4, 0x90
; SSE-WIN-NEXT:  .LBB6_3: # %for.body
; SSE-WIN-NEXT:    # =>This Inner Loop Header: Depth=1
; SSE-WIN-NEXT:    xorps %xmm2, %xmm2
; SSE-WIN-NEXT:    cvtsi2ss %eax, %xmm2
; SSE-WIN-NEXT:    xorps %xmm3, %xmm3
; SSE-WIN-NEXT:    cvtsi2ss %ecx, %xmm3
; SSE-WIN-NEXT:    addss %xmm2, %xmm0
; SSE-WIN-NEXT:    addss %xmm3, %xmm1
; SSE-WIN-NEXT:    incl %eax
; SSE-WIN-NEXT:    decl %ecx
; SSE-WIN-NEXT:    jne .LBB6_3
; SSE-WIN-NEXT:  # %bb.4: # %for.end
; SSE-WIN-NEXT:    subss %xmm1, %xmm0
; SSE-WIN-NEXT:    retq
; SSE-WIN-NEXT:  .LBB6_1:
; SSE-WIN-NEXT:    xorps %xmm0, %xmm0
; SSE-WIN-NEXT:    xorps %xmm1, %xmm1
; SSE-WIN-NEXT:    subss %xmm1, %xmm0
; SSE-WIN-NEXT:    retq
;
; AVX1-LABEL: loopdep1:
; AVX1:       # %bb.0: # %entry
; AVX1-NEXT:    testl %ecx, %ecx
; AVX1-NEXT:    je .LBB6_1
; AVX1-NEXT:  # %bb.2: # %for.body.preheader
; AVX1-NEXT:    movl $1, %eax
; AVX1-NEXT:    vxorps %xmm0, %xmm0, %xmm0
; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
; AVX1-NEXT:    .p2align 4, 0x90
; AVX1-NEXT:  .LBB6_3: # %for.body
; AVX1-NEXT:    # =>This Inner Loop Header: Depth=1
; AVX1-NEXT:    vcvtsi2ss %eax, %xmm4, %xmm2
; AVX1-NEXT:    vcvtsi2ss %ecx, %xmm4, %xmm3
; AVX1-NEXT:    vaddss %xmm2, %xmm0, %xmm0
; AVX1-NEXT:    vaddss %xmm3, %xmm1, %xmm1
; AVX1-NEXT:    incl %eax
; AVX1-NEXT:    decl %ecx
; AVX1-NEXT:    jne .LBB6_3
; AVX1-NEXT:  # %bb.4: # %for.end
; AVX1-NEXT:    vsubss %xmm1, %xmm0, %xmm0
; AVX1-NEXT:    retq
; AVX1-NEXT:  .LBB6_1:
; AVX1-NEXT:    vxorps %xmm0, %xmm0, %xmm0
; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
; AVX1-NEXT:    vsubss %xmm1, %xmm0, %xmm0
; AVX1-NEXT:    retq
;
; AVX512VL-LABEL: loopdep1:
; AVX512VL:       # %bb.0: # %entry
; AVX512VL-NEXT:    testl %ecx, %ecx
; AVX512VL-NEXT:    je .LBB6_1
; AVX512VL-NEXT:  # %bb.2: # %for.body.preheader
; AVX512VL-NEXT:    movl $1, %eax
; AVX512VL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
; AVX512VL-NEXT:    vxorps %xmm1, %xmm1, %xmm1
; AVX512VL-NEXT:    .p2align 4, 0x90
; AVX512VL-NEXT:  .LBB6_3: # %for.body
; AVX512VL-NEXT:    # =>This Inner Loop Header: Depth=1
; AVX512VL-NEXT:    vcvtsi2ss %eax, %xmm3, %xmm2
; AVX512VL-NEXT:    vaddss %xmm2, %xmm0, %xmm0
; AVX512VL-NEXT:    vcvtsi2ss %ecx, %xmm3, %xmm2
; AVX512VL-NEXT:    vaddss %xmm2, %xmm1, %xmm1
; AVX512VL-NEXT:    incl %eax
; AVX512VL-NEXT:    decl %ecx
; AVX512VL-NEXT:    jne .LBB6_3
; AVX512VL-NEXT:  # %bb.4: # %for.end
; AVX512VL-NEXT:    vsubss %xmm1, %xmm0, %xmm0
; AVX512VL-NEXT:    retq
; AVX512VL-NEXT:  .LBB6_1:
; AVX512VL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
; AVX512VL-NEXT:    vxorps %xmm1, %xmm1, %xmm1
; AVX512VL-NEXT:    vsubss %xmm1, %xmm0, %xmm0
; AVX512VL-NEXT:    retq
entry:
  %tobool3 = icmp eq i32 %m, 0
  br i1 %tobool3, label %for.end, label %for.body

for.body:                                         ; preds = %entry, %for.body
  %m.addr.07 = phi i32 [ %dec, %for.body ], [ %m, %entry ]
  %s1.06 = phi float [ %add, %for.body ], [ 0.000000e+00, %entry ]
  %s2.05 = phi float [ %add2, %for.body ], [ 0.000000e+00, %entry ]
  %n.04 = phi i32 [ %inc, %for.body ], [ 1, %entry ]
  %conv = sitofp i32 %n.04 to float
  %add = fadd float %s1.06, %conv
  %conv1 = sitofp i32 %m.addr.07 to float
  %add2 = fadd float %s2.05, %conv1
  %inc = add nsw i32 %n.04, 1
  %dec = add nsw i32 %m.addr.07, -1
  %tobool = icmp eq i32 %dec, 0
  br i1 %tobool, label %for.end, label %for.body

for.end:                                          ; preds = %for.body, %entry
  %s1.0.lcssa = phi float [ 0.000000e+00, %entry ], [ %add, %for.body ]
  %s2.0.lcssa = phi float [ 0.000000e+00, %entry ], [ %add2, %for.body ]
  %sub = fsub float %s1.0.lcssa, %s2.0.lcssa
  ret float %sub
}

; rdar:15221834 False AVX register dependencies cause 5x slowdown on
; flops-6. Make sure the unused register read by vcvtsi2sd is zeroed
; to avoid cyclic dependence on a write to the same register in a
; previous iteration.

define i64 @loopdep2(i64* nocapture %x, double* nocapture %y) nounwind {
; SSE-LINUX-LABEL: loopdep2:
; SSE-LINUX:       # %bb.0: # %entry
; SSE-LINUX-NEXT:    movq (%rdi), %rax
; SSE-LINUX-NEXT:    movl $1, %ecx
; SSE-LINUX-NEXT:    .p2align 4, 0x90
; SSE-LINUX-NEXT:  .LBB7_1: # %loop
; SSE-LINUX-NEXT:    # =>This Inner Loop Header: Depth=1
; SSE-LINUX-NEXT:    xorps %xmm0, %xmm0
; SSE-LINUX-NEXT:    cvtsi2sd %rcx, %xmm0
; SSE-LINUX-NEXT:    movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; SSE-LINUX-NEXT:    # xmm0 = mem[0],zero
; SSE-LINUX-NEXT:    addsd (%rsi), %xmm0
; SSE-LINUX-NEXT:    cvttsd2si %xmm0, %rdx
; SSE-LINUX-NEXT:    addq %rdx, %rax
; SSE-LINUX-NEXT:    incq %rcx
; SSE-LINUX-NEXT:    cmpq $156250000, %rcx # imm = 0x9502F90
; SSE-LINUX-NEXT:    jne .LBB7_1
; SSE-LINUX-NEXT:  # %bb.2: # %ret
; SSE-LINUX-NEXT:    retq
;
; SSE-WIN-LABEL: loopdep2:
; SSE-WIN:       # %bb.0: # %entry
; SSE-WIN-NEXT:    subq $184, %rsp
; SSE-WIN-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movq (%rcx), %rax
; SSE-WIN-NEXT:    movl $1, %r8d
; SSE-WIN-NEXT:    .p2align 4, 0x90
; SSE-WIN-NEXT:  .LBB7_1: # %loop
; SSE-WIN-NEXT:    # =>This Inner Loop Header: Depth=1
; SSE-WIN-NEXT:    xorps %xmm0, %xmm0
; SSE-WIN-NEXT:    cvtsi2sd %r8, %xmm0
; SSE-WIN-NEXT:    movsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    movsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; SSE-WIN-NEXT:    # xmm0 = mem[0],zero
; SSE-WIN-NEXT:    addsd (%rdx), %xmm0
; SSE-WIN-NEXT:    cvttsd2si %xmm0, %rcx
; SSE-WIN-NEXT:    addq %rcx, %rax
; SSE-WIN-NEXT:    incq %r8
; SSE-WIN-NEXT:    cmpq $156250000, %r8 # imm = 0x9502F90
; SSE-WIN-NEXT:    jne .LBB7_1
; SSE-WIN-NEXT:  # %bb.2: # %ret
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-WIN-NEXT:    addq $184, %rsp
; SSE-WIN-NEXT:    retq
;
; AVX-LABEL: loopdep2:
; AVX:       # %bb.0: # %entry
; AVX-NEXT:    subq $184, %rsp
; AVX-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    movq (%rcx), %rax
; AVX-NEXT:    movl $1, %r8d
; AVX-NEXT:    .p2align 4, 0x90
; AVX-NEXT:  .LBB7_1: # %loop
; AVX-NEXT:    # =>This Inner Loop Header: Depth=1
; AVX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
; AVX-NEXT:    vcvtsi2sd %r8, %xmm1, %xmm0
; AVX-NEXT:    vmovsd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    vmovsd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 8-byte Reload
; AVX-NEXT:    # xmm0 = mem[0],zero
; AVX-NEXT:    vaddsd (%rdx), %xmm0, %xmm0
; AVX-NEXT:    vcvttsd2si %xmm0, %rcx
; AVX-NEXT:    addq %rcx, %rax
; AVX-NEXT:    incq %r8
; AVX-NEXT:    cmpq $156250000, %r8 # imm = 0x9502F90
; AVX-NEXT:    jne .LBB7_1
; AVX-NEXT:  # %bb.2: # %ret
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    addq $184, %rsp
; AVX-NEXT:    retq
entry:
  %vx = load i64, i64* %x
  br label %loop
loop:
  %i = phi i64 [ 1, %entry ], [ %inc, %loop ]
  %s1 = phi i64 [ %vx, %entry ], [ %s2, %loop ]
  %fi = sitofp i64 %i to double
  tail call void asm sideeffect "", "~{xmm0},~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{dirflag},~{fpsr},~{flags}"()
  %vy = load double, double* %y
  %fipy = fadd double %fi, %vy
  %iipy = fptosi double %fipy to i64
  %s2 = add i64 %s1, %iipy
  %inc = add nsw i64 %i, 1
  %exitcond = icmp eq i64 %inc, 156250000
  br i1 %exitcond, label %ret, label %loop
ret:
  ret i64 %s2
}

; This loop contains a cvtsi2sd instruction that has a loop-carried
; false dependency on an xmm that is modified by other scalar instructions
; that follow it in the loop. Additionally, the source of convert is a
; memory operand. Verify the break false dependency fix pass breaks this
; dependency by inserting a xor before the convert.
@x = common dso_local global [1024 x double] zeroinitializer, align 16
@y = common dso_local global [1024 x double] zeroinitializer, align 16
@z = common dso_local global [1024 x double] zeroinitializer, align 16
@w = common dso_local global [1024 x double] zeroinitializer, align 16
@v = common dso_local global [1024 x i32] zeroinitializer, align 16

define dso_local void @loopdep3() {
; SSE-LINUX-LABEL: loopdep3:
; SSE-LINUX:       # %bb.0: # %entry
; SSE-LINUX-NEXT:    xorl %eax, %eax
; SSE-LINUX-NEXT:    .p2align 4, 0x90
; SSE-LINUX-NEXT:  .LBB8_1: # %for.cond1.preheader
; SSE-LINUX-NEXT:    # =>This Loop Header: Depth=1
; SSE-LINUX-NEXT:    # Child Loop BB8_2 Depth 2
; SSE-LINUX-NEXT:    movq $-4096, %rcx # imm = 0xF000
; SSE-LINUX-NEXT:    .p2align 4, 0x90
; SSE-LINUX-NEXT:  .LBB8_2: # %for.body3
; SSE-LINUX-NEXT:    # Parent Loop BB8_1 Depth=1
; SSE-LINUX-NEXT:    # => This Inner Loop Header: Depth=2
; SSE-LINUX-NEXT:    xorps %xmm0, %xmm0
; SSE-LINUX-NEXT:    cvtsi2sdl v+4096(%rcx), %xmm0
; SSE-LINUX-NEXT:    mulsd x+8192(%rcx,%rcx), %xmm0
; SSE-LINUX-NEXT:    mulsd y+8192(%rcx,%rcx), %xmm0
; SSE-LINUX-NEXT:    mulsd z+8192(%rcx,%rcx), %xmm0
; SSE-LINUX-NEXT:    movsd %xmm0, w+8192(%rcx,%rcx)
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    addq $4, %rcx
; SSE-LINUX-NEXT:    jne .LBB8_2
; SSE-LINUX-NEXT:  # %bb.3: # %for.inc14
; SSE-LINUX-NEXT:    # in Loop: Header=BB8_1 Depth=1
; SSE-LINUX-NEXT:    incl %eax
; SSE-LINUX-NEXT:    cmpl $100000, %eax # imm = 0x186A0
; SSE-LINUX-NEXT:    jne .LBB8_1
; SSE-LINUX-NEXT:  # %bb.4: # %for.end16
; SSE-LINUX-NEXT:    retq
;
; SSE-WIN-LABEL: loopdep3:
; SSE-WIN:       # %bb.0: # %entry
; SSE-WIN-NEXT:    pushq %rsi
; SSE-WIN-NEXT:    .seh_pushreg %rsi
; SSE-WIN-NEXT:    subq $160, %rsp
; SSE-WIN-NEXT:    .seh_stackalloc 160
; SSE-WIN-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm15, 144
; SSE-WIN-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm14, 128
; SSE-WIN-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm13, 112
; SSE-WIN-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm12, 96
; SSE-WIN-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm11, 80
; SSE-WIN-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm10, 64
; SSE-WIN-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm9, 48
; SSE-WIN-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm8, 32
; SSE-WIN-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm7, 16
; SSE-WIN-NEXT:    movaps %xmm6, (%rsp) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm6, 0
; SSE-WIN-NEXT:    .seh_endprologue
; SSE-WIN-NEXT:    xorl %r9d, %r9d
; SSE-WIN-NEXT:    leaq {{.*}}(%rip), %r8
; SSE-WIN-NEXT:    leaq {{.*}}(%rip), %r10
; SSE-WIN-NEXT:    leaq {{.*}}(%rip), %r11
; SSE-WIN-NEXT:    leaq {{.*}}(%rip), %rax
; SSE-WIN-NEXT:    leaq {{.*}}(%rip), %rdx
; SSE-WIN-NEXT:    .p2align 4, 0x90
; SSE-WIN-NEXT:  .LBB8_1: # %for.cond1.preheader
; SSE-WIN-NEXT:    # =>This Loop Header: Depth=1
; SSE-WIN-NEXT:    # Child Loop BB8_2 Depth 2
; SSE-WIN-NEXT:    movq %r8, %rcx
; SSE-WIN-NEXT:    xorl %esi, %esi
; SSE-WIN-NEXT:    .p2align 4, 0x90
; SSE-WIN-NEXT:  .LBB8_2: # %for.body3
; SSE-WIN-NEXT:    # Parent Loop BB8_1 Depth=1
; SSE-WIN-NEXT:    # => This Inner Loop Header: Depth=2
; SSE-WIN-NEXT:    xorps %xmm0, %xmm0
; SSE-WIN-NEXT:    cvtsi2sdl (%rcx), %xmm0
; SSE-WIN-NEXT:    mulsd (%rsi,%r10), %xmm0
; SSE-WIN-NEXT:    mulsd (%rsi,%r11), %xmm0
; SSE-WIN-NEXT:    mulsd (%rsi,%rax), %xmm0
; SSE-WIN-NEXT:    movsd %xmm0, (%rsi,%rdx)
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    addq $8, %rsi
; SSE-WIN-NEXT:    addq $4, %rcx
; SSE-WIN-NEXT:    cmpq $8192, %rsi # imm = 0x2000
; SSE-WIN-NEXT:    jne .LBB8_2
; SSE-WIN-NEXT:  # %bb.3: # %for.inc14
; SSE-WIN-NEXT:    # in Loop: Header=BB8_1 Depth=1
; SSE-WIN-NEXT:    incl %r9d
; SSE-WIN-NEXT:    cmpl $100000, %r9d # imm = 0x186A0
; SSE-WIN-NEXT:    jne .LBB8_1
; SSE-WIN-NEXT:  # %bb.4: # %for.end16
; SSE-WIN-NEXT:    movaps (%rsp), %xmm6 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-WIN-NEXT:    addq $160, %rsp
; SSE-WIN-NEXT:    popq %rsi
; SSE-WIN-NEXT:    retq
; SSE-WIN-NEXT:    .seh_endproc
;
; AVX-LABEL: loopdep3:
; AVX:       # %bb.0: # %entry
; AVX-NEXT:    pushq %rsi
; AVX-NEXT:    .seh_pushreg %rsi
; AVX-NEXT:    subq $160, %rsp
; AVX-NEXT:    .seh_stackalloc 160
; AVX-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm15, 144
; AVX-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm14, 128
; AVX-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm13, 112
; AVX-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm12, 96
; AVX-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm11, 80
; AVX-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm10, 64
; AVX-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm9, 48
; AVX-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm8, 32
; AVX-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm7, 16
; AVX-NEXT:    vmovaps %xmm6, (%rsp) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm6, 0
; AVX-NEXT:    .seh_endprologue
; AVX-NEXT:    xorl %r9d, %r9d
; AVX-NEXT:    leaq {{.*}}(%rip), %r8
; AVX-NEXT:    leaq {{.*}}(%rip), %r10
; AVX-NEXT:    leaq {{.*}}(%rip), %r11
; AVX-NEXT:    leaq {{.*}}(%rip), %rax
; AVX-NEXT:    leaq {{.*}}(%rip), %rdx
; AVX-NEXT:    .p2align 4, 0x90
; AVX-NEXT:  .LBB8_1: # %for.cond1.preheader
; AVX-NEXT:    # =>This Loop Header: Depth=1
; AVX-NEXT:    # Child Loop BB8_2 Depth 2
; AVX-NEXT:    movq %r8, %rcx
; AVX-NEXT:    xorl %esi, %esi
; AVX-NEXT:    .p2align 4, 0x90
; AVX-NEXT:  .LBB8_2: # %for.body3
; AVX-NEXT:    # Parent Loop BB8_1 Depth=1
; AVX-NEXT:    # => This Inner Loop Header: Depth=2
; AVX-NEXT:    vxorps %xmm0, %xmm0, %xmm0
; AVX-NEXT:    vcvtsi2sdl (%rcx), %xmm0, %xmm0
; AVX-NEXT:    vmulsd (%rsi,%r10), %xmm0, %xmm0
; AVX-NEXT:    vmulsd (%rsi,%r11), %xmm0, %xmm0
; AVX-NEXT:    vmulsd (%rsi,%rax), %xmm0, %xmm0
; AVX-NEXT:    vmovsd %xmm0, (%rsi,%rdx)
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    addq $8, %rsi
; AVX-NEXT:    addq $4, %rcx
; AVX-NEXT:    cmpq $8192, %rsi # imm = 0x2000
; AVX-NEXT:    jne .LBB8_2
; AVX-NEXT:  # %bb.3: # %for.inc14
; AVX-NEXT:    # in Loop: Header=BB8_1 Depth=1
; AVX-NEXT:    incl %r9d
; AVX-NEXT:    cmpl $100000, %r9d # imm = 0x186A0
; AVX-NEXT:    jne .LBB8_1
; AVX-NEXT:  # %bb.4: # %for.end16
; AVX-NEXT:    vmovaps (%rsp), %xmm6 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    addq $160, %rsp
; AVX-NEXT:    popq %rsi
; AVX-NEXT:    retq
; AVX-NEXT:    .seh_endproc
entry:
  br label %for.cond1.preheader

for.cond1.preheader:                              ; preds = %for.inc14, %entry
  %i.025 = phi i32 [ 0, %entry ], [ %inc15, %for.inc14 ]
  br label %for.body3

for.body3:
  %indvars.iv = phi i64 [ 0, %for.cond1.preheader ], [ %indvars.iv.next, %for.body3 ]
  %arrayidx = getelementptr inbounds [1024 x i32], [1024 x i32]* @v, i64 0, i64 %indvars.iv
  %0 = load i32, i32* %arrayidx, align 4
  %conv = sitofp i32 %0 to double
  %arrayidx5 = getelementptr inbounds [1024 x double], [1024 x double]* @x, i64 0, i64 %indvars.iv
  %1 = load double, double* %arrayidx5, align 8
  %mul = fmul double %conv, %1
  %arrayidx7 = getelementptr inbounds [1024 x double], [1024 x double]* @y, i64 0, i64 %indvars.iv
  %2 = load double, double* %arrayidx7, align 8
  %mul8 = fmul double %mul, %2
  %arrayidx10 = getelementptr inbounds [1024 x double], [1024 x double]* @z, i64 0, i64 %indvars.iv
  %3 = load double, double* %arrayidx10, align 8
  %mul11 = fmul double %mul8, %3
  %arrayidx13 = getelementptr inbounds [1024 x double], [1024 x double]* @w, i64 0, i64 %indvars.iv
  store double %mul11, double* %arrayidx13, align 8
  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
  %exitcond = icmp eq i64 %indvars.iv.next, 1024
  tail call void asm sideeffect "", "~{xmm0},~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{dirflag},~{fpsr},~{flags}"()
  br i1 %exitcond, label %for.inc14, label %for.body3

for.inc14:                                        ; preds = %for.body3
  %inc15 = add nsw i32 %i.025, 1
  %exitcond26 = icmp eq i32 %inc15, 100000
  br i1 %exitcond26, label %for.end16, label %for.cond1.preheader

for.end16:                                        ; preds = %for.inc14
  ret void

}

define dso_local double @inlineasmdep(i64 %arg) {
; SSE-LINUX-LABEL: inlineasmdep:
; SSE-LINUX:       # %bb.0: # %top
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    xorps %xmm0, %xmm0
; SSE-LINUX-NEXT:    cvtsi2sd %rdi, %xmm0
; SSE-LINUX-NEXT:    retq
;
; SSE-WIN-LABEL: inlineasmdep:
; SSE-WIN:       # %bb.0: # %top
; SSE-WIN-NEXT:    subq $168, %rsp
; SSE-WIN-NEXT:    .seh_stackalloc 168
; SSE-WIN-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm15, 144
; SSE-WIN-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm14, 128
; SSE-WIN-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm13, 112
; SSE-WIN-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm12, 96
; SSE-WIN-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm11, 80
; SSE-WIN-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm10, 64
; SSE-WIN-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm9, 48
; SSE-WIN-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm8, 32
; SSE-WIN-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm7, 16
; SSE-WIN-NEXT:    movaps %xmm6, (%rsp) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm6, 0
; SSE-WIN-NEXT:    .seh_endprologue
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    xorps %xmm0, %xmm0
; SSE-WIN-NEXT:    cvtsi2sd %rcx, %xmm0
; SSE-WIN-NEXT:    movaps (%rsp), %xmm6 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-WIN-NEXT:    addq $168, %rsp
; SSE-WIN-NEXT:    retq
; SSE-WIN-NEXT:    .seh_endproc
;
; AVX-LABEL: inlineasmdep:
; AVX:       # %bb.0: # %top
; AVX-NEXT:    subq $168, %rsp
; AVX-NEXT:    .seh_stackalloc 168
; AVX-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm15, 144
; AVX-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm14, 128
; AVX-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm13, 112
; AVX-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm12, 96
; AVX-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm11, 80
; AVX-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm10, 64
; AVX-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm9, 48
; AVX-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm8, 32
; AVX-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm7, 16
; AVX-NEXT:    vmovaps %xmm6, (%rsp) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm6, 0
; AVX-NEXT:    .seh_endprologue
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    vxorps %xmm0, %xmm0, %xmm0
; AVX-NEXT:    vcvtsi2sd %rcx, %xmm0, %xmm0
; AVX-NEXT:    vmovaps (%rsp), %xmm6 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    addq $168, %rsp
; AVX-NEXT:    retq
; AVX-NEXT:    .seh_endproc
top:
  tail call void asm sideeffect "", "~{xmm0},~{xmm1},~{xmm2},~{xmm3},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{dirflag},~{fpsr},~{flags}"()
  %tmp1 = sitofp i64 %arg to double
  ret double %tmp1
}

; Make sure we are making a smart choice regarding undef registers and
; hiding the false dependency behind a true dependency
define dso_local double @truedeps(float %arg) {
; SSE-LINUX-LABEL: truedeps:
; SSE-LINUX:       # %bb.0: # %top
; SSE-LINUX-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
; SSE-LINUX-NEXT:    # xmm0 = mem[0],zero,zero,zero
; SSE-LINUX-NEXT:    cvtss2sd %xmm0, %xmm0
; SSE-LINUX-NEXT:    retq
;
; SSE-WIN-LABEL: truedeps:
; SSE-WIN:       # %bb.0: # %top
; SSE-WIN-NEXT:    subq $184, %rsp
; SSE-WIN-NEXT:    .seh_stackalloc 184
; SSE-WIN-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm15, 160
; SSE-WIN-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm14, 144
; SSE-WIN-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm13, 128
; SSE-WIN-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm12, 112
; SSE-WIN-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm11, 96
; SSE-WIN-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm10, 80
; SSE-WIN-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm9, 64
; SSE-WIN-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm8, 48
; SSE-WIN-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm7, 32
; SSE-WIN-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm6, 16
; SSE-WIN-NEXT:    .seh_endprologue
; SSE-WIN-NEXT:    movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
; SSE-WIN-NEXT:    # xmm0 = mem[0],zero,zero,zero
; SSE-WIN-NEXT:    cvtss2sd %xmm0, %xmm0
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-WIN-NEXT:    addq $184, %rsp
; SSE-WIN-NEXT:    retq
; SSE-WIN-NEXT:    .seh_endproc
;
; AVX-LABEL: truedeps:
; AVX:       # %bb.0: # %top
; AVX-NEXT:    subq $184, %rsp
; AVX-NEXT:    .seh_stackalloc 184
; AVX-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm15, 160
; AVX-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm14, 144
; AVX-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm13, 128
; AVX-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm12, 112
; AVX-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm11, 96
; AVX-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm10, 80
; AVX-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm9, 64
; AVX-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm8, 48
; AVX-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm7, 32
; AVX-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm6, 16
; AVX-NEXT:    .seh_endprologue
; AVX-NEXT:    vmovss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    vmovss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
; AVX-NEXT:    # xmm0 = mem[0],zero,zero,zero
; AVX-NEXT:    vcvtss2sd %xmm0, %xmm0, %xmm0
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    addq $184, %rsp
; AVX-NEXT:    retq
; AVX-NEXT:    .seh_endproc
top:
  tail call void asm sideeffect "", "~{xmm6},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm0},~{xmm1},~{xmm2},~{xmm3},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm4},~{xmm5},~{xmm7},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{dirflag},~{fpsr},~{flags}"()
  %tmp1 = fpext float %arg to double
  ret double %tmp1
}

; Make sure we are making a smart choice regarding undef registers and
; choosing the register with the highest clearence
define dso_local double @clearence(i64 %arg) {
; SSE-LINUX-LABEL: clearence:
; SSE-LINUX:       # %bb.0: # %top
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    xorps %xmm0, %xmm0
; SSE-LINUX-NEXT:    cvtsi2sd %rdi, %xmm0
; SSE-LINUX-NEXT:    retq
;
; SSE-WIN-LABEL: clearence:
; SSE-WIN:       # %bb.0: # %top
; SSE-WIN-NEXT:    subq $168, %rsp
; SSE-WIN-NEXT:    .seh_stackalloc 168
; SSE-WIN-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm15, 144
; SSE-WIN-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm14, 128
; SSE-WIN-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm13, 112
; SSE-WIN-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm12, 96
; SSE-WIN-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm11, 80
; SSE-WIN-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm10, 64
; SSE-WIN-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm9, 48
; SSE-WIN-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm8, 32
; SSE-WIN-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm7, 16
; SSE-WIN-NEXT:    movaps %xmm6, (%rsp) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm6, 0
; SSE-WIN-NEXT:    .seh_endprologue
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    xorps %xmm0, %xmm0
; SSE-WIN-NEXT:    cvtsi2sd %rcx, %xmm0
; SSE-WIN-NEXT:    movaps (%rsp), %xmm6 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-WIN-NEXT:    addq $168, %rsp
; SSE-WIN-NEXT:    retq
; SSE-WIN-NEXT:    .seh_endproc
;
; AVX-LABEL: clearence:
; AVX:       # %bb.0: # %top
; AVX-NEXT:    subq $168, %rsp
; AVX-NEXT:    .seh_stackalloc 168
; AVX-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm15, 144
; AVX-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm14, 128
; AVX-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm13, 112
; AVX-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm12, 96
; AVX-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm11, 80
; AVX-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm10, 64
; AVX-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm9, 48
; AVX-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm8, 32
; AVX-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm7, 16
; AVX-NEXT:    vmovaps %xmm6, (%rsp) # 16-byte Spill
; AVX-NEXT:    .seh_savexmm %xmm6, 0
; AVX-NEXT:    .seh_endprologue
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    vxorps %xmm6, %xmm6, %xmm6
; AVX-NEXT:    vcvtsi2sd %rcx, %xmm6, %xmm0
; AVX-NEXT:    vmovaps (%rsp), %xmm6 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    addq $168, %rsp
; AVX-NEXT:    retq
; AVX-NEXT:    .seh_endproc
top:
  tail call void asm sideeffect "", "~{xmm6},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm0},~{xmm1},~{xmm2},~{xmm3},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm4},~{xmm5},~{xmm7},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{dirflag},~{fpsr},~{flags}"()
  %tmp1 = sitofp i64 %arg to double
  ret double %tmp1
}

; Make sure we are making a smart choice regarding undef registers in order to
; avoid a cyclic dependence on a write to the same register in a previous
; iteration, especially when we cannot zero out the undef register because it
; is alive.
define i64 @loopclearence(i64* nocapture %x, double* nocapture %y) nounwind {
; SSE-LINUX-LABEL: loopclearence:
; SSE-LINUX:       # %bb.0: # %entry
; SSE-LINUX-NEXT:    movq (%rdi), %rax
; SSE-LINUX-NEXT:    movl $1, %ecx
; SSE-LINUX-NEXT:    .p2align 4, 0x90
; SSE-LINUX-NEXT:  .LBB12_1: # %loop
; SSE-LINUX-NEXT:    # =>This Inner Loop Header: Depth=1
; SSE-LINUX-NEXT:    xorps %xmm4, %xmm4
; SSE-LINUX-NEXT:    cvtsi2sd %rcx, %xmm4
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    addsd (%rsi), %xmm4
; SSE-LINUX-NEXT:    cvttsd2si %xmm4, %rdx
; SSE-LINUX-NEXT:    addq %rdx, %rax
; SSE-LINUX-NEXT:    incq %rcx
; SSE-LINUX-NEXT:    cmpq $156250000, %rcx # imm = 0x9502F90
; SSE-LINUX-NEXT:    jne .LBB12_1
; SSE-LINUX-NEXT:  # %bb.2: # %ret
; SSE-LINUX-NEXT:    retq
;
; SSE-WIN-LABEL: loopclearence:
; SSE-WIN:       # %bb.0: # %entry
; SSE-WIN-NEXT:    subq $136, %rsp
; SSE-WIN-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    movaps %xmm8, (%rsp) # 16-byte Spill
; SSE-WIN-NEXT:    movq (%rcx), %rax
; SSE-WIN-NEXT:    movl $1, %r8d
; SSE-WIN-NEXT:    .p2align 4, 0x90
; SSE-WIN-NEXT:  .LBB12_1: # %loop
; SSE-WIN-NEXT:    # =>This Inner Loop Header: Depth=1
; SSE-WIN-NEXT:    xorps %xmm4, %xmm4
; SSE-WIN-NEXT:    cvtsi2sd %r8, %xmm4
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    addsd (%rdx), %xmm4
; SSE-WIN-NEXT:    cvttsd2si %xmm4, %rcx
; SSE-WIN-NEXT:    addq %rcx, %rax
; SSE-WIN-NEXT:    incq %r8
; SSE-WIN-NEXT:    cmpq $156250000, %r8 # imm = 0x9502F90
; SSE-WIN-NEXT:    jne .LBB12_1
; SSE-WIN-NEXT:  # %bb.2: # %ret
; SSE-WIN-NEXT:    movaps (%rsp), %xmm8 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-WIN-NEXT:    addq $136, %rsp
; SSE-WIN-NEXT:    retq
;
; AVX-LABEL: loopclearence:
; AVX:       # %bb.0: # %entry
; AVX-NEXT:    subq $136, %rsp
; AVX-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps %xmm8, (%rsp) # 16-byte Spill
; AVX-NEXT:    movq (%rcx), %rax
; AVX-NEXT:    movl $1, %r8d
; AVX-NEXT:    .p2align 4, 0x90
; AVX-NEXT:  .LBB12_1: # %loop
; AVX-NEXT:    # =>This Inner Loop Header: Depth=1
; AVX-NEXT:    vcvtsi2sd %r8, %xmm5, %xmm4
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    #APP
; AVX-NEXT:    #NO_APP
; AVX-NEXT:    vaddsd (%rdx), %xmm4, %xmm0
; AVX-NEXT:    vcvttsd2si %xmm0, %rcx
; AVX-NEXT:    addq %rcx, %rax
; AVX-NEXT:    incq %r8
; AVX-NEXT:    cmpq $156250000, %r8 # imm = 0x9502F90
; AVX-NEXT:    jne .LBB12_1
; AVX-NEXT:  # %bb.2: # %ret
; AVX-NEXT:    vmovaps (%rsp), %xmm8 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    addq $136, %rsp
; AVX-NEXT:    retq
entry:
  %vx = load i64, i64* %x
  br label %loop
loop:
  %i = phi i64 [ 1, %entry ], [ %inc, %loop ]
  %s1 = phi i64 [ %vx, %entry ], [ %s2, %loop ]
  %fi = sitofp i64 %i to double
  tail call void asm sideeffect "", "~{xmm0},~{xmm1},~{xmm2},~{xmm3},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{dirflag},~{fpsr},~{flags}"()
  %vy = load double, double* %y
  %fipy = fadd double %fi, %vy
  %iipy = fptosi double %fipy to i64
  %s2 = add i64 %s1, %iipy
  %inc = add nsw i64 %i, 1
  %exitcond = icmp eq i64 %inc, 156250000
  br i1 %exitcond, label %ret, label %loop
ret:
  ret i64 %s2
}

; Make sure we are making a smart choice regarding undef registers even for more
; complicated loop structures. This example is the inner loop from
; julia> a = falses(10000); a[1:4:end] = true
; julia> linspace(1.0,2.0,10000)[a]
define dso_local void @loopclearance2(double* nocapture %y, i64* %x, double %c1, double %c2, double %c3, double %c4, i64 %size) {
; SSE-LINUX-LABEL: loopclearance2:
; SSE-LINUX:       # %bb.0: # %entry
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    #APP
; SSE-LINUX-NEXT:    #NO_APP
; SSE-LINUX-NEXT:    movl $1, %r8d
; SSE-LINUX-NEXT:    xorl %ecx, %ecx
; SSE-LINUX-NEXT:    .p2align 4, 0x90
; SSE-LINUX-NEXT:  .LBB13_1: # %inner_loop
; SSE-LINUX-NEXT:    # =>This Inner Loop Header: Depth=1
; SSE-LINUX-NEXT:    movq %rcx, %rax
; SSE-LINUX-NEXT:    shrq $6, %rcx
; SSE-LINUX-NEXT:    movq (%rsi,%rcx,8), %rcx
; SSE-LINUX-NEXT:    btq %rax, %rcx
; SSE-LINUX-NEXT:    leaq 1(%rax), %rcx
; SSE-LINUX-NEXT:    jae .LBB13_1
; SSE-LINUX-NEXT:  # %bb.2: # %loop_end
; SSE-LINUX-NEXT:    # in Loop: Header=BB13_1 Depth=1
; SSE-LINUX-NEXT:    leaq 1(%r8), %r9
; SSE-LINUX-NEXT:    xorps %xmm4, %xmm4
; SSE-LINUX-NEXT:    cvtsi2sd %r9, %xmm4
; SSE-LINUX-NEXT:    movapd %xmm0, %xmm5
; SSE-LINUX-NEXT:    subsd %xmm4, %xmm5
; SSE-LINUX-NEXT:    mulsd %xmm1, %xmm5
; SSE-LINUX-NEXT:    leaq -1(%rcx), %rax
; SSE-LINUX-NEXT:    xorps %xmm4, %xmm4
; SSE-LINUX-NEXT:    cvtsi2sd %rax, %xmm4
; SSE-LINUX-NEXT:    mulsd %xmm2, %xmm4
; SSE-LINUX-NEXT:    addsd %xmm5, %xmm4
; SSE-LINUX-NEXT:    divsd %xmm3, %xmm4
; SSE-LINUX-NEXT:    movsd %xmm4, -8(%rdi,%r8,8)
; SSE-LINUX-NEXT:    movq %r9, %r8
; SSE-LINUX-NEXT:    cmpq %r9, %rdx
; SSE-LINUX-NEXT:    jge .LBB13_1
; SSE-LINUX-NEXT:  # %bb.3: # %loopdone
; SSE-LINUX-NEXT:    retq
;
; SSE-WIN-LABEL: loopclearance2:
; SSE-WIN:       # %bb.0: # %entry
; SSE-WIN-NEXT:    subq $152, %rsp
; SSE-WIN-NEXT:    .seh_stackalloc 152
; SSE-WIN-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm15, 128
; SSE-WIN-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm14, 112
; SSE-WIN-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm13, 96
; SSE-WIN-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm12, 80
; SSE-WIN-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm11, 64
; SSE-WIN-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm10, 48
; SSE-WIN-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm9, 32
; SSE-WIN-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm8, 16
; SSE-WIN-NEXT:    movaps %xmm7, (%rsp) # 16-byte Spill
; SSE-WIN-NEXT:    .seh_savexmm %xmm7, 0
; SSE-WIN-NEXT:    .seh_endprologue
; SSE-WIN-NEXT:    movq {{[0-9]+}}(%rsp), %r8
; SSE-WIN-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
; SSE-WIN-NEXT:    movsd {{.*#+}} xmm1 = mem[0],zero
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    #APP
; SSE-WIN-NEXT:    #NO_APP
; SSE-WIN-NEXT:    movl $1, %r9d
; SSE-WIN-NEXT:    xorl %r11d, %r11d
; SSE-WIN-NEXT:    .p2align 4, 0x90
; SSE-WIN-NEXT:  .LBB13_1: # %inner_loop
; SSE-WIN-NEXT:    # =>This Inner Loop Header: Depth=1
; SSE-WIN-NEXT:    movq %r11, %r10
; SSE-WIN-NEXT:    movq %r11, %rax
; SSE-WIN-NEXT:    shrq $6, %rax
; SSE-WIN-NEXT:    movq (%rdx,%rax,8), %rax
; SSE-WIN-NEXT:    btq %r11, %rax
; SSE-WIN-NEXT:    leaq 1(%r11), %r11
; SSE-WIN-NEXT:    jae .LBB13_1
; SSE-WIN-NEXT:  # %bb.2: # %loop_end
; SSE-WIN-NEXT:    # in Loop: Header=BB13_1 Depth=1
; SSE-WIN-NEXT:    leaq 1(%r9), %r10
; SSE-WIN-NEXT:    xorps %xmm4, %xmm4
; SSE-WIN-NEXT:    cvtsi2sd %r10, %xmm4
; SSE-WIN-NEXT:    movapd %xmm2, %xmm5
; SSE-WIN-NEXT:    subsd %xmm4, %xmm5
; SSE-WIN-NEXT:    mulsd %xmm3, %xmm5
; SSE-WIN-NEXT:    leaq -1(%r11), %rax
; SSE-WIN-NEXT:    xorps %xmm4, %xmm4
; SSE-WIN-NEXT:    cvtsi2sd %rax, %xmm4
; SSE-WIN-NEXT:    mulsd %xmm1, %xmm4
; SSE-WIN-NEXT:    addsd %xmm5, %xmm4
; SSE-WIN-NEXT:    divsd %xmm0, %xmm4
; SSE-WIN-NEXT:    movsd %xmm4, -8(%rcx,%r9,8)
; SSE-WIN-NEXT:    movq %r10, %r9
; SSE-WIN-NEXT:    cmpq %r10, %r8
; SSE-WIN-NEXT:    jge .LBB13_1
; SSE-WIN-NEXT:  # %bb.3: # %loopdone
; SSE-WIN-NEXT:    movaps (%rsp), %xmm7 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-WIN-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-WIN-NEXT:    addq $152, %rsp
; SSE-WIN-NEXT:    retq
; SSE-WIN-NEXT:    .seh_endproc
;
; AVX1-LABEL: loopclearance2:
; AVX1:       # %bb.0: # %entry
; AVX1-NEXT:    subq $152, %rsp
; AVX1-NEXT:    .seh_stackalloc 152
; AVX1-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    .seh_savexmm %xmm15, 128
; AVX1-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    .seh_savexmm %xmm14, 112
; AVX1-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    .seh_savexmm %xmm13, 96
; AVX1-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    .seh_savexmm %xmm12, 80
; AVX1-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    .seh_savexmm %xmm11, 64
; AVX1-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    .seh_savexmm %xmm10, 48
; AVX1-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    .seh_savexmm %xmm9, 32
; AVX1-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    .seh_savexmm %xmm8, 16
; AVX1-NEXT:    vmovaps %xmm7, (%rsp) # 16-byte Spill
; AVX1-NEXT:    .seh_savexmm %xmm7, 0
; AVX1-NEXT:    .seh_endprologue
; AVX1-NEXT:    movq {{[0-9]+}}(%rsp), %r8
; AVX1-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
; AVX1-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
; AVX1-NEXT:    #APP
; AVX1-NEXT:    #NO_APP
; AVX1-NEXT:    #APP
; AVX1-NEXT:    #NO_APP
; AVX1-NEXT:    #APP
; AVX1-NEXT:    #NO_APP
; AVX1-NEXT:    #APP
; AVX1-NEXT:    #NO_APP
; AVX1-NEXT:    #APP
; AVX1-NEXT:    #NO_APP
; AVX1-NEXT:    #APP
; AVX1-NEXT:    #NO_APP
; AVX1-NEXT:    #APP
; AVX1-NEXT:    #NO_APP
; AVX1-NEXT:    movl $1, %r9d
; AVX1-NEXT:    xorl %r11d, %r11d
; AVX1-NEXT:    .p2align 4, 0x90
; AVX1-NEXT:  .LBB13_1: # %inner_loop
; AVX1-NEXT:    # =>This Inner Loop Header: Depth=1
; AVX1-NEXT:    movq %r11, %r10
; AVX1-NEXT:    movq %r11, %rax
; AVX1-NEXT:    shrq $6, %rax
; AVX1-NEXT:    movq (%rdx,%rax,8), %rax
; AVX1-NEXT:    btq %r11, %rax
; AVX1-NEXT:    leaq 1(%r11), %r11
; AVX1-NEXT:    jae .LBB13_1
; AVX1-NEXT:  # %bb.2: # %loop_end
; AVX1-NEXT:    # in Loop: Header=BB13_1 Depth=1
; AVX1-NEXT:    leaq 1(%r9), %r10
; AVX1-NEXT:    vcvtsi2sd %r10, %xmm6, %xmm4
; AVX1-NEXT:    vsubsd %xmm4, %xmm2, %xmm4
; AVX1-NEXT:    vmulsd %xmm3, %xmm4, %xmm4
; AVX1-NEXT:    leaq -1(%r11), %rax
; AVX1-NEXT:    vcvtsi2sd %rax, %xmm6, %xmm5
; AVX1-NEXT:    vmulsd %xmm1, %xmm5, %xmm5
; AVX1-NEXT:    vaddsd %xmm5, %xmm4, %xmm4
; AVX1-NEXT:    vdivsd %xmm0, %xmm4, %xmm4
; AVX1-NEXT:    vmovsd %xmm4, -8(%rcx,%r9,8)
; AVX1-NEXT:    movq %r10, %r9
; AVX1-NEXT:    cmpq %r10, %r8
; AVX1-NEXT:    jge .LBB13_1
; AVX1-NEXT:  # %bb.3: # %loopdone
; AVX1-NEXT:    vmovaps (%rsp), %xmm7 # 16-byte Reload
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX1-NEXT:    addq $152, %rsp
; AVX1-NEXT:    retq
; AVX1-NEXT:    .seh_endproc
;
; AVX512VL-LABEL: loopclearance2:
; AVX512VL:       # %bb.0: # %entry
; AVX512VL-NEXT:    subq $152, %rsp
; AVX512VL-NEXT:    .seh_stackalloc 152
; AVX512VL-NEXT:    vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512VL-NEXT:    .seh_savexmm %xmm15, 128
; AVX512VL-NEXT:    vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512VL-NEXT:    .seh_savexmm %xmm14, 112
; AVX512VL-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512VL-NEXT:    .seh_savexmm %xmm13, 96
; AVX512VL-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512VL-NEXT:    .seh_savexmm %xmm12, 80
; AVX512VL-NEXT:    vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512VL-NEXT:    .seh_savexmm %xmm11, 64
; AVX512VL-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512VL-NEXT:    .seh_savexmm %xmm10, 48
; AVX512VL-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512VL-NEXT:    .seh_savexmm %xmm9, 32
; AVX512VL-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512VL-NEXT:    .seh_savexmm %xmm8, 16
; AVX512VL-NEXT:    vmovaps %xmm7, (%rsp) # 16-byte Spill
; AVX512VL-NEXT:    .seh_savexmm %xmm7, 0
; AVX512VL-NEXT:    .seh_endprologue
; AVX512VL-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
; AVX512VL-NEXT:    movq {{[0-9]+}}(%rsp), %r8
; AVX512VL-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
; AVX512VL-NEXT:    #APP
; AVX512VL-NEXT:    #NO_APP
; AVX512VL-NEXT:    #APP
; AVX512VL-NEXT:    #NO_APP
; AVX512VL-NEXT:    #APP
; AVX512VL-NEXT:    #NO_APP
; AVX512VL-NEXT:    #APP
; AVX512VL-NEXT:    #NO_APP
; AVX512VL-NEXT:    #APP
; AVX512VL-NEXT:    #NO_APP
; AVX512VL-NEXT:    #APP
; AVX512VL-NEXT:    #NO_APP
; AVX512VL-NEXT:    #APP
; AVX512VL-NEXT:    #NO_APP
; AVX512VL-NEXT:    movl $1, %r9d
; AVX512VL-NEXT:    xorl %r11d, %r11d
; AVX512VL-NEXT:    .p2align 4, 0x90
; AVX512VL-NEXT:  .LBB13_1: # %inner_loop
; AVX512VL-NEXT:    # =>This Inner Loop Header: Depth=1
; AVX512VL-NEXT:    movq %r11, %r10
; AVX512VL-NEXT:    movq %r11, %rax
; AVX512VL-NEXT:    shrq $6, %rax
; AVX512VL-NEXT:    movq (%rdx,%rax,8), %rax
; AVX512VL-NEXT:    btq %r11, %rax
; AVX512VL-NEXT:    leaq 1(%r11), %r11
; AVX512VL-NEXT:    jae .LBB13_1
; AVX512VL-NEXT:  # %bb.2: # %loop_end
; AVX512VL-NEXT:    # in Loop: Header=BB13_1 Depth=1
; AVX512VL-NEXT:    leaq 1(%r9), %r10
; AVX512VL-NEXT:    vcvtsi2sd %r10, %xmm6, %xmm4
; AVX512VL-NEXT:    vsubsd %xmm4, %xmm2, %xmm4
; AVX512VL-NEXT:    vmulsd %xmm3, %xmm4, %xmm4
; AVX512VL-NEXT:    leaq -1(%r11), %rax
; AVX512VL-NEXT:    vcvtsi2sd %rax, %xmm6, %xmm5
; AVX512VL-NEXT:    vmulsd %xmm1, %xmm5, %xmm5
; AVX512VL-NEXT:    vaddsd %xmm5, %xmm4, %xmm4
; AVX512VL-NEXT:    vdivsd %xmm0, %xmm4, %xmm4
; AVX512VL-NEXT:    vmovsd %xmm4, -8(%rcx,%r9,8)
; AVX512VL-NEXT:    movq %r10, %r9
; AVX512VL-NEXT:    cmpq %r10, %r8
; AVX512VL-NEXT:    jge .LBB13_1
; AVX512VL-NEXT:  # %bb.3: # %loopdone
; AVX512VL-NEXT:    vmovaps (%rsp), %xmm7 # 16-byte Reload
; AVX512VL-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX512VL-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX512VL-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX512VL-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX512VL-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX512VL-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX512VL-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX512VL-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX512VL-NEXT:    addq $152, %rsp
; AVX512VL-NEXT:    retq
; AVX512VL-NEXT:    .seh_endproc
entry:
  tail call void asm sideeffect "", "~{xmm7},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{dirflag},~{fpsr},~{flags}"()
  tail call void asm sideeffect "", "~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{dirflag},~{fpsr},~{flags}"()
  br label %loop

loop:
  %phi_i = phi i64 [ 1, %entry ], [ %nexti, %loop_end ]
  %phi_j = phi i64 [ 1, %entry ], [ %nextj, %loop_end ]
  %phi_k = phi i64 [ 0, %entry ], [ %nextk, %loop_end ]
  br label %inner_loop

inner_loop:
  %phi = phi i64 [ %phi_k, %loop ], [ %nextk, %inner_loop ]
  %idx = lshr i64 %phi, 6
  %inputptr = getelementptr i64, i64* %x, i64 %idx
  %input = load i64, i64* %inputptr, align 8
  %masked = and i64 %phi, 63
  %shiftedmasked = shl i64 1, %masked
  %maskedinput = and i64 %input, %shiftedmasked
  %cmp = icmp eq i64 %maskedinput, 0
  %nextk = add i64 %phi, 1
  br i1 %cmp, label %inner_loop, label %loop_end

loop_end:
  %nexti = add i64 %phi_i, 1
  %nextj = add i64 %phi_j, 1
  ; Register use, plus us clobbering 7-15 above, basically forces xmm6 here as
  ; the only reasonable choice. The primary thing we care about is that it's
  ; not one of the registers used in the loop (e.g. not the output reg here)
  %nexti_f = sitofp i64 %nexti to double
  %sub = fsub double %c1, %nexti_f
  %mul = fmul double %sub, %c2
  %phi_f = sitofp i64 %phi to double
  %mul2 = fmul double %phi_f, %c3
  %add2 = fadd double %mul, %mul2
  %div = fdiv double %add2, %c4
  %prev_j = add i64 %phi_j, -1
  %outptr = getelementptr double, double* %y, i64 %prev_j
  store double %div, double* %outptr, align 8
  %done = icmp slt i64 %size, %nexti
  br i1 %done, label %loopdone, label %loop

loopdone:
  ret void
}