llvm/test/CodeGen/X86/xmulo.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -disable-peephole -mtriple=x86_64-linux-unknown < %s | FileCheck %s --check-prefixes=CHECK,LINUX,SDAG
; RUN: llc -disable-peephole -mtriple=x86_64-linux-unknown -fast-isel -fast-isel-abort=1 < %s | FileCheck %s --check-prefixes=CHECK,LINUX,FAST
; RUN: llc -disable-peephole -mtriple=x86_64-linux-unknown -mcpu=knl < %s | FileCheck %s --check-prefixes=CHECK,LINUX,SDAG
; RUN: llc -disable-peephole -mtriple=x86_64-pc-win32 < %s | FileCheck %s --check-prefixes=CHECK,WIN64
; RUN: llc -disable-peephole -mtriple=i386-pc-win32 < %s | FileCheck %s --check-prefix=WIN32

define {i64, i1} @t1() nounwind {
; CHECK-LABEL: t1:
; CHECK:       # %bb.0:
; CHECK-NEXT:    movl $72, %eax
; CHECK-NEXT:    xorl %edx, %edx
; CHECK-NEXT:    retq
;
; WIN32-LABEL: t1:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl $72, %eax
; WIN32-NEXT:    xorl %edx, %edx
; WIN32-NEXT:    xorl %ecx, %ecx
; WIN32-NEXT:    retl
  %1 = call {i64, i1} @llvm.umul.with.overflow.i64(i64 9, i64 8)
  ret {i64, i1} %1
}

define {i64, i1} @t2() nounwind {
; CHECK-LABEL: t2:
; CHECK:       # %bb.0:
; CHECK-NEXT:    xorl %eax, %eax
; CHECK-NEXT:    xorl %edx, %edx
; CHECK-NEXT:    retq
;
; WIN32-LABEL: t2:
; WIN32:       # %bb.0:
; WIN32-NEXT:    xorl %eax, %eax
; WIN32-NEXT:    xorl %edx, %edx
; WIN32-NEXT:    xorl %ecx, %ecx
; WIN32-NEXT:    retl
  %1 = call {i64, i1} @llvm.umul.with.overflow.i64(i64 9, i64 0)
  ret {i64, i1} %1
}

define {i64, i1} @t3() nounwind {
; CHECK-LABEL: t3:
; CHECK:       # %bb.0:
; CHECK-NEXT:    movq $-9, %rax
; CHECK-NEXT:    movb $1, %dl
; CHECK-NEXT:    retq
;
; WIN32-LABEL: t3:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl $-9, %eax
; WIN32-NEXT:    movl $-1, %edx
; WIN32-NEXT:    movb $1, %cl
; WIN32-NEXT:    retl
  %1 = call {i64, i1} @llvm.umul.with.overflow.i64(i64 9, i64 -1)
  ret {i64, i1} %1
}

; SMULO
define zeroext i1 @smuloi8(i8 %v1, i8 %v2, i8* %res) {
; SDAG-LABEL: smuloi8:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movl %edi, %eax
; SDAG-NEXT:    # kill: def $al killed $al killed $eax
; SDAG-NEXT:    imulb %sil
; SDAG-NEXT:    seto %cl
; SDAG-NEXT:    movb %al, (%rdx)
; SDAG-NEXT:    movl %ecx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: smuloi8:
; FAST:       # %bb.0:
; FAST-NEXT:    movl %edi, %eax
; FAST-NEXT:    # kill: def $al killed $al killed $eax
; FAST-NEXT:    imulb %sil
; FAST-NEXT:    seto %cl
; FAST-NEXT:    movb %al, (%rdx)
; FAST-NEXT:    andb $1, %cl
; FAST-NEXT:    movzbl %cl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smuloi8:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    imulb %dl
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movb %al, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloi8:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %edx
; WIN32-NEXT:    movb {{[0-9]+}}(%esp), %al
; WIN32-NEXT:    imulb {{[0-9]+}}(%esp)
; WIN32-NEXT:    seto %cl
; WIN32-NEXT:    movb %al, (%edx)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    retl
  %t = call {i8, i1} @llvm.smul.with.overflow.i8(i8 %v1, i8 %v2)
  %val = extractvalue {i8, i1} %t, 0
  %obit = extractvalue {i8, i1} %t, 1
  store i8 %val, i8* %res
  ret i1 %obit
}

define zeroext i1 @smuloi16(i16 %v1, i16 %v2, i16* %res) {
; SDAG-LABEL: smuloi16:
; SDAG:       # %bb.0:
; SDAG-NEXT:    imulw %si, %di
; SDAG-NEXT:    seto %al
; SDAG-NEXT:    movw %di, (%rdx)
; SDAG-NEXT:    retq
;
; FAST-LABEL: smuloi16:
; FAST:       # %bb.0:
; FAST-NEXT:    imulw %si, %di
; FAST-NEXT:    seto %al
; FAST-NEXT:    movw %di, (%rdx)
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smuloi16:
; WIN64:       # %bb.0:
; WIN64-NEXT:    imulw %dx, %cx
; WIN64-NEXT:    seto %al
; WIN64-NEXT:    movw %cx, (%r8)
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloi16:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movzwl {{[0-9]+}}(%esp), %edx
; WIN32-NEXT:    imulw {{[0-9]+}}(%esp), %dx
; WIN32-NEXT:    seto %al
; WIN32-NEXT:    movw %dx, (%ecx)
; WIN32-NEXT:    retl
  %t = call {i16, i1} @llvm.smul.with.overflow.i16(i16 %v1, i16 %v2)
  %val = extractvalue {i16, i1} %t, 0
  %obit = extractvalue {i16, i1} %t, 1
  store i16 %val, i16* %res
  ret i1 %obit
}

define zeroext i1 @smuloi32(i32 %v1, i32 %v2, i32* %res) {
; SDAG-LABEL: smuloi32:
; SDAG:       # %bb.0:
; SDAG-NEXT:    imull %esi, %edi
; SDAG-NEXT:    seto %al
; SDAG-NEXT:    movl %edi, (%rdx)
; SDAG-NEXT:    retq
;
; FAST-LABEL: smuloi32:
; FAST:       # %bb.0:
; FAST-NEXT:    imull %esi, %edi
; FAST-NEXT:    seto %al
; FAST-NEXT:    movl %edi, (%rdx)
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smuloi32:
; WIN64:       # %bb.0:
; WIN64-NEXT:    imull %edx, %ecx
; WIN64-NEXT:    seto %al
; WIN64-NEXT:    movl %ecx, (%r8)
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloi32:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %edx
; WIN32-NEXT:    imull {{[0-9]+}}(%esp), %edx
; WIN32-NEXT:    seto %al
; WIN32-NEXT:    movl %edx, (%ecx)
; WIN32-NEXT:    retl
  %t = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %v1, i32 %v2)
  %val = extractvalue {i32, i1} %t, 0
  %obit = extractvalue {i32, i1} %t, 1
  store i32 %val, i32* %res
  ret i1 %obit
}

define zeroext i1 @smuloi64(i64 %v1, i64 %v2, i64* %res) {
; SDAG-LABEL: smuloi64:
; SDAG:       # %bb.0:
; SDAG-NEXT:    imulq %rsi, %rdi
; SDAG-NEXT:    seto %al
; SDAG-NEXT:    movq %rdi, (%rdx)
; SDAG-NEXT:    retq
;
; FAST-LABEL: smuloi64:
; FAST:       # %bb.0:
; FAST-NEXT:    imulq %rsi, %rdi
; FAST-NEXT:    seto %al
; FAST-NEXT:    movq %rdi, (%rdx)
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smuloi64:
; WIN64:       # %bb.0:
; WIN64-NEXT:    imulq %rdx, %rcx
; WIN64-NEXT:    seto %al
; WIN64-NEXT:    movq %rcx, (%r8)
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloi64:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebp
; WIN32-NEXT:    pushl %ebx
; WIN32-NEXT:    pushl %edi
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    subl $8, %esp
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ebp
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl %ecx, %edi
; WIN32-NEXT:    sarl $31, %edi
; WIN32-NEXT:    movl %eax, %esi
; WIN32-NEXT:    imull %edi, %esi
; WIN32-NEXT:    mull %edi
; WIN32-NEXT:    movl %eax, %ebx
; WIN32-NEXT:    addl %esi, %edx
; WIN32-NEXT:    movl %ebp, %esi
; WIN32-NEXT:    imull %ebp, %edi
; WIN32-NEXT:    addl %edx, %edi
; WIN32-NEXT:    sarl $31, %esi
; WIN32-NEXT:    movl %esi, %ebp
; WIN32-NEXT:    imull %ecx, %ebp
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl %esi, %eax
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    addl %ebp, %edx
; WIN32-NEXT:    imull %ecx, %esi
; WIN32-NEXT:    addl %edx, %esi
; WIN32-NEXT:    addl %ebx, %eax
; WIN32-NEXT:    movl %eax, (%esp) # 4-byte Spill
; WIN32-NEXT:    adcl %edi, %esi
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    movl %ecx, %edi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    movl %edx, %ebp
; WIN32-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    movl %edx, %ebx
; WIN32-NEXT:    movl %eax, %ecx
; WIN32-NEXT:    addl %ebp, %ecx
; WIN32-NEXT:    adcl $0, %ebx
; WIN32-NEXT:    movl %edi, %eax
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    movl %edx, %edi
; WIN32-NEXT:    movl %eax, %ebp
; WIN32-NEXT:    addl %ecx, %ebp
; WIN32-NEXT:    adcl %ebx, %edi
; WIN32-NEXT:    setb %cl
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    addl %edi, %eax
; WIN32-NEXT:    movzbl %cl, %ecx
; WIN32-NEXT:    adcl %ecx, %edx
; WIN32-NEXT:    addl (%esp), %eax # 4-byte Folded Reload
; WIN32-NEXT:    adcl %esi, %edx
; WIN32-NEXT:    movl %ebp, %ecx
; WIN32-NEXT:    sarl $31, %ecx
; WIN32-NEXT:    xorl %ecx, %edx
; WIN32-NEXT:    xorl %eax, %ecx
; WIN32-NEXT:    orl %edx, %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl %ebp, 4(%eax)
; WIN32-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; WIN32-NEXT:    movl %ecx, (%eax)
; WIN32-NEXT:    setne %al
; WIN32-NEXT:    addl $8, %esp
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    popl %edi
; WIN32-NEXT:    popl %ebx
; WIN32-NEXT:    popl %ebp
; WIN32-NEXT:    retl
  %t = call {i64, i1} @llvm.smul.with.overflow.i64(i64 %v1, i64 %v2)
  %val = extractvalue {i64, i1} %t, 0
  %obit = extractvalue {i64, i1} %t, 1
  store i64 %val, i64* %res
  ret i1 %obit
}

; UMULO
define zeroext i1 @umuloi8(i8 %v1, i8 %v2, i8* %res) {
; SDAG-LABEL: umuloi8:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movl %edi, %eax
; SDAG-NEXT:    # kill: def $al killed $al killed $eax
; SDAG-NEXT:    mulb %sil
; SDAG-NEXT:    seto %cl
; SDAG-NEXT:    movb %al, (%rdx)
; SDAG-NEXT:    movl %ecx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umuloi8:
; FAST:       # %bb.0:
; FAST-NEXT:    movl %edi, %eax
; FAST-NEXT:    # kill: def $al killed $al killed $eax
; FAST-NEXT:    mulb %sil
; FAST-NEXT:    seto %cl
; FAST-NEXT:    movb %al, (%rdx)
; FAST-NEXT:    andb $1, %cl
; FAST-NEXT:    movzbl %cl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umuloi8:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    mulb %dl
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movb %al, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloi8:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %edx
; WIN32-NEXT:    movb {{[0-9]+}}(%esp), %al
; WIN32-NEXT:    mulb {{[0-9]+}}(%esp)
; WIN32-NEXT:    seto %cl
; WIN32-NEXT:    movb %al, (%edx)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    retl
  %t = call {i8, i1} @llvm.umul.with.overflow.i8(i8 %v1, i8 %v2)
  %val = extractvalue {i8, i1} %t, 0
  %obit = extractvalue {i8, i1} %t, 1
  store i8 %val, i8* %res
  ret i1 %obit
}

define zeroext i1 @umuloi16(i16 %v1, i16 %v2, i16* %res) {
; SDAG-LABEL: umuloi16:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movq %rdx, %rcx
; SDAG-NEXT:    movl %edi, %eax
; SDAG-NEXT:    # kill: def $ax killed $ax killed $eax
; SDAG-NEXT:    mulw %si
; SDAG-NEXT:    seto %dl
; SDAG-NEXT:    movw %ax, (%rcx)
; SDAG-NEXT:    movl %edx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umuloi16:
; FAST:       # %bb.0:
; FAST-NEXT:    movq %rdx, %rcx
; FAST-NEXT:    movl %edi, %eax
; FAST-NEXT:    # kill: def $ax killed $ax killed $eax
; FAST-NEXT:    mulw %si
; FAST-NEXT:    seto %dl
; FAST-NEXT:    movw %ax, (%rcx)
; FAST-NEXT:    andb $1, %dl
; FAST-NEXT:    movzbl %dl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umuloi16:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    mulw %dx
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movw %ax, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloi16:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    mulw {{[0-9]+}}(%esp)
; WIN32-NEXT:    seto %cl
; WIN32-NEXT:    movw %ax, (%esi)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    retl
  %t = call {i16, i1} @llvm.umul.with.overflow.i16(i16 %v1, i16 %v2)
  %val = extractvalue {i16, i1} %t, 0
  %obit = extractvalue {i16, i1} %t, 1
  store i16 %val, i16* %res
  ret i1 %obit
}

define zeroext i1 @umuloi32(i32 %v1, i32 %v2, i32* %res) {
; SDAG-LABEL: umuloi32:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movq %rdx, %rcx
; SDAG-NEXT:    movl %edi, %eax
; SDAG-NEXT:    mull %esi
; SDAG-NEXT:    seto %dl
; SDAG-NEXT:    movl %eax, (%rcx)
; SDAG-NEXT:    movl %edx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umuloi32:
; FAST:       # %bb.0:
; FAST-NEXT:    movq %rdx, %rcx
; FAST-NEXT:    movl %edi, %eax
; FAST-NEXT:    mull %esi
; FAST-NEXT:    seto %dl
; FAST-NEXT:    movl %eax, (%rcx)
; FAST-NEXT:    andb $1, %dl
; FAST-NEXT:    movzbl %dl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umuloi32:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    mull %edx
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movl %eax, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloi32:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    seto %cl
; WIN32-NEXT:    movl %eax, (%esi)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    retl
  %t = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %v1, i32 %v2)
  %val = extractvalue {i32, i1} %t, 0
  %obit = extractvalue {i32, i1} %t, 1
  store i32 %val, i32* %res
  ret i1 %obit
}

define zeroext i1 @umuloi64(i64 %v1, i64 %v2, i64* %res) {
; SDAG-LABEL: umuloi64:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movq %rdx, %rcx
; SDAG-NEXT:    movq %rdi, %rax
; SDAG-NEXT:    mulq %rsi
; SDAG-NEXT:    seto %dl
; SDAG-NEXT:    movq %rax, (%rcx)
; SDAG-NEXT:    movl %edx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umuloi64:
; FAST:       # %bb.0:
; FAST-NEXT:    movq %rdx, %rcx
; FAST-NEXT:    movq %rdi, %rax
; FAST-NEXT:    mulq %rsi
; FAST-NEXT:    seto %dl
; FAST-NEXT:    movq %rax, (%rcx)
; FAST-NEXT:    andb $1, %dl
; FAST-NEXT:    movzbl %dl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umuloi64:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movq %rcx, %rax
; WIN64-NEXT:    mulq %rdx
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movq %rax, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloi64:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebp
; WIN32-NEXT:    pushl %ebx
; WIN32-NEXT:    pushl %edi
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ebp
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    testl %esi, %esi
; WIN32-NEXT:    setne %dl
; WIN32-NEXT:    testl %eax, %eax
; WIN32-NEXT:    setne %bl
; WIN32-NEXT:    andb %dl, %bl
; WIN32-NEXT:    mull %ebp
; WIN32-NEXT:    movl %eax, %edi
; WIN32-NEXT:    seto %bh
; WIN32-NEXT:    movl %esi, %eax
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    movl %ecx, %edx
; WIN32-NEXT:    movl %eax, %esi
; WIN32-NEXT:    seto %ch
; WIN32-NEXT:    orb %bh, %ch
; WIN32-NEXT:    addl %edi, %esi
; WIN32-NEXT:    movl %edx, %eax
; WIN32-NEXT:    mull %ebp
; WIN32-NEXT:    addl %esi, %edx
; WIN32-NEXT:    setb %cl
; WIN32-NEXT:    orb %ch, %cl
; WIN32-NEXT:    orb %bl, %cl
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    movl %eax, (%esi)
; WIN32-NEXT:    movl %edx, 4(%esi)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    popl %edi
; WIN32-NEXT:    popl %ebx
; WIN32-NEXT:    popl %ebp
; WIN32-NEXT:    retl
  %t = call {i64, i1} @llvm.umul.with.overflow.i64(i64 %v1, i64 %v2)
  %val = extractvalue {i64, i1} %t, 0
  %obit = extractvalue {i64, i1} %t, 1
  store i64 %val, i64* %res
  ret i1 %obit
}

;
; Check the use of the overflow bit in combination with a select instruction.
;
define i32 @smuloselecti32(i32 %v1, i32 %v2) {
; LINUX-LABEL: smuloselecti32:
; LINUX:       # %bb.0:
; LINUX-NEXT:    movl %esi, %eax
; LINUX-NEXT:    movl %edi, %ecx
; LINUX-NEXT:    imull %esi, %ecx
; LINUX-NEXT:    cmovol %edi, %eax
; LINUX-NEXT:    retq
;
; WIN64-LABEL: smuloselecti32:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %edx, %eax
; WIN64-NEXT:    movl %ecx, %edx
; WIN64-NEXT:    imull %eax, %edx
; WIN64-NEXT:    cmovol %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloselecti32:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl %eax, %edx
; WIN32-NEXT:    imull %ecx, %edx
; WIN32-NEXT:    jo LBB11_2
; WIN32-NEXT:  # %bb.1:
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:  LBB11_2:
; WIN32-NEXT:    retl
  %t = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %v1, i32 %v2)
  %obit = extractvalue {i32, i1} %t, 1
  %ret = select i1 %obit, i32 %v1, i32 %v2
  ret i32 %ret
}

define i64 @smuloselecti64(i64 %v1, i64 %v2) {
; LINUX-LABEL: smuloselecti64:
; LINUX:       # %bb.0:
; LINUX-NEXT:    movq %rsi, %rax
; LINUX-NEXT:    movq %rdi, %rcx
; LINUX-NEXT:    imulq %rsi, %rcx
; LINUX-NEXT:    cmovoq %rdi, %rax
; LINUX-NEXT:    retq
;
; WIN64-LABEL: smuloselecti64:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movq %rdx, %rax
; WIN64-NEXT:    movq %rcx, %rdx
; WIN64-NEXT:    imulq %rax, %rdx
; WIN64-NEXT:    cmovoq %rcx, %rax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloselecti64:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebp
; WIN32-NEXT:    pushl %ebx
; WIN32-NEXT:    pushl %edi
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    pushl %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ebp
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl %eax, %ecx
; WIN32-NEXT:    movl %eax, %esi
; WIN32-NEXT:    sarl $31, %ecx
; WIN32-NEXT:    movl %ebp, %edi
; WIN32-NEXT:    imull %ecx, %edi
; WIN32-NEXT:    movl %ebp, %eax
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    movl %eax, (%esp) # 4-byte Spill
; WIN32-NEXT:    addl %edi, %edx
; WIN32-NEXT:    imull %ebx, %ecx
; WIN32-NEXT:    addl %edx, %ecx
; WIN32-NEXT:    sarl $31, %ebx
; WIN32-NEXT:    movl %ebx, %edi
; WIN32-NEXT:    imull %esi, %edi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    movl %ebx, %eax
; WIN32-NEXT:    mull %esi
; WIN32-NEXT:    addl %edi, %edx
; WIN32-NEXT:    movl %esi, %edi
; WIN32-NEXT:    imull %esi, %ebx
; WIN32-NEXT:    addl %edx, %ebx
; WIN32-NEXT:    addl (%esp), %eax # 4-byte Folded Reload
; WIN32-NEXT:    movl %eax, (%esp) # 4-byte Spill
; WIN32-NEXT:    adcl %ecx, %ebx
; WIN32-NEXT:    movl %edi, %eax
; WIN32-NEXT:    mull %ebp
; WIN32-NEXT:    movl %edx, %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    mull %ebp
; WIN32-NEXT:    movl %edx, %ecx
; WIN32-NEXT:    movl %eax, %ebp
; WIN32-NEXT:    addl %esi, %ebp
; WIN32-NEXT:    adcl $0, %ecx
; WIN32-NEXT:    movl %edi, %eax
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    movl %edx, %edi
; WIN32-NEXT:    movl %eax, %esi
; WIN32-NEXT:    addl %ebp, %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ebp
; WIN32-NEXT:    adcl %ecx, %edi
; WIN32-NEXT:    setb %cl
; WIN32-NEXT:    movl %ebp, %eax
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    addl %edi, %eax
; WIN32-NEXT:    movzbl %cl, %ecx
; WIN32-NEXT:    adcl %ecx, %edx
; WIN32-NEXT:    addl (%esp), %eax # 4-byte Folded Reload
; WIN32-NEXT:    adcl %ebx, %edx
; WIN32-NEXT:    sarl $31, %esi
; WIN32-NEXT:    xorl %esi, %edx
; WIN32-NEXT:    xorl %eax, %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    orl %edx, %esi
; WIN32-NEXT:    jne LBB12_2
; WIN32-NEXT:  # %bb.1:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ebp
; WIN32-NEXT:  LBB12_2:
; WIN32-NEXT:    movl %ebp, %edx
; WIN32-NEXT:    addl $4, %esp
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    popl %edi
; WIN32-NEXT:    popl %ebx
; WIN32-NEXT:    popl %ebp
; WIN32-NEXT:    retl
  %t = call {i64, i1} @llvm.smul.with.overflow.i64(i64 %v1, i64 %v2)
  %obit = extractvalue {i64, i1} %t, 1
  %ret = select i1 %obit, i64 %v1, i64 %v2
  ret i64 %ret
}

define i32 @umuloselecti32(i32 %v1, i32 %v2) {
; LINUX-LABEL: umuloselecti32:
; LINUX:       # %bb.0:
; LINUX-NEXT:    movl %edi, %eax
; LINUX-NEXT:    mull %esi
; LINUX-NEXT:    cmovol %edi, %esi
; LINUX-NEXT:    movl %esi, %eax
; LINUX-NEXT:    retq
;
; WIN64-LABEL: umuloselecti32:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %edx, %r8d
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    mull %edx
; WIN64-NEXT:    cmovol %ecx, %r8d
; WIN64-NEXT:    movl %r8d, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloselecti32:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    mull %esi
; WIN32-NEXT:    jo LBB13_2
; WIN32-NEXT:  # %bb.1:
; WIN32-NEXT:    movl %esi, %ecx
; WIN32-NEXT:  LBB13_2:
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    retl
  %t = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %v1, i32 %v2)
  %obit = extractvalue {i32, i1} %t, 1
  %ret = select i1 %obit, i32 %v1, i32 %v2
  ret i32 %ret
}

define i64 @umuloselecti64(i64 %v1, i64 %v2) {
; LINUX-LABEL: umuloselecti64:
; LINUX:       # %bb.0:
; LINUX-NEXT:    movq %rdi, %rax
; LINUX-NEXT:    mulq %rsi
; LINUX-NEXT:    cmovoq %rdi, %rsi
; LINUX-NEXT:    movq %rsi, %rax
; LINUX-NEXT:    retq
;
; WIN64-LABEL: umuloselecti64:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movq %rdx, %r8
; WIN64-NEXT:    movq %rcx, %rax
; WIN64-NEXT:    mulq %rdx
; WIN64-NEXT:    cmovoq %rcx, %r8
; WIN64-NEXT:    movq %r8, %rax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloselecti64:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebp
; WIN32-NEXT:    pushl %ebx
; WIN32-NEXT:    pushl %edi
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    pushl %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ebp
; WIN32-NEXT:    testl %ebp, %ebp
; WIN32-NEXT:    setne %al
; WIN32-NEXT:    testl %esi, %esi
; WIN32-NEXT:    setne %bl
; WIN32-NEXT:    andb %al, %bl
; WIN32-NEXT:    movl %esi, %eax
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    movl %eax, %edi
; WIN32-NEXT:    seto {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Folded Spill
; WIN32-NEXT:    movl %ebp, %eax
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    movl %eax, %ebp
; WIN32-NEXT:    seto %bh
; WIN32-NEXT:    orb {{[-0-9]+}}(%e{{[sb]}}p), %bh # 1-byte Folded Reload
; WIN32-NEXT:    addl %edi, %ebp
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %edi
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    mull %edi
; WIN32-NEXT:    addl %ebp, %edx
; WIN32-NEXT:    setb %al
; WIN32-NEXT:    orb %bh, %al
; WIN32-NEXT:    orb %bl, %al
; WIN32-NEXT:    testb %al, %al
; WIN32-NEXT:    jne LBB14_2
; WIN32-NEXT:  # %bb.1:
; WIN32-NEXT:    movl %edi, %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:  LBB14_2:
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    movl %esi, %edx
; WIN32-NEXT:    addl $4, %esp
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    popl %edi
; WIN32-NEXT:    popl %ebx
; WIN32-NEXT:    popl %ebp
; WIN32-NEXT:    retl
  %t = call {i64, i1} @llvm.umul.with.overflow.i64(i64 %v1, i64 %v2)
  %obit = extractvalue {i64, i1} %t, 1
  %ret = select i1 %obit, i64 %v1, i64 %v2
  ret i64 %ret
}

;
; Check the use of the overflow bit in combination with a branch instruction.
;
define zeroext i1 @smulobri8(i8 %v1, i8 %v2) {
; SDAG-LABEL: smulobri8:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movl %edi, %eax
; SDAG-NEXT:    # kill: def $al killed $al killed $eax
; SDAG-NEXT:    imulb %sil
; SDAG-NEXT:    jo .LBB15_1
; SDAG-NEXT:  # %bb.2: # %continue
; SDAG-NEXT:    movb $1, %al
; SDAG-NEXT:    retq
; SDAG-NEXT:  .LBB15_1: # %overflow
; SDAG-NEXT:    xorl %eax, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: smulobri8:
; FAST:       # %bb.0:
; FAST-NEXT:    movl %edi, %eax
; FAST-NEXT:    # kill: def $al killed $al killed $eax
; FAST-NEXT:    imulb %sil
; FAST-NEXT:    seto %al
; FAST-NEXT:    testb $1, %al
; FAST-NEXT:    jne .LBB15_1
; FAST-NEXT:  # %bb.2: # %continue
; FAST-NEXT:    movb $1, %al
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
; FAST-NEXT:  .LBB15_1: # %overflow
; FAST-NEXT:    xorl %eax, %eax
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smulobri8:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    imulb %dl
; WIN64-NEXT:    jo .LBB15_1
; WIN64-NEXT:  # %bb.2: # %continue
; WIN64-NEXT:    movb $1, %al
; WIN64-NEXT:    retq
; WIN64-NEXT:  .LBB15_1: # %overflow
; WIN64-NEXT:    xorl %eax, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smulobri8:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movb {{[0-9]+}}(%esp), %al
; WIN32-NEXT:    imulb {{[0-9]+}}(%esp)
; WIN32-NEXT:    jo LBB15_1
; WIN32-NEXT:  # %bb.2: # %continue
; WIN32-NEXT:    movb $1, %al
; WIN32-NEXT:    retl
; WIN32-NEXT:  LBB15_1: # %overflow
; WIN32-NEXT:    xorl %eax, %eax
; WIN32-NEXT:    retl
  %t = call {i8, i1} @llvm.smul.with.overflow.i8(i8 %v1, i8 %v2)
  %val = extractvalue {i8, i1} %t, 0
  %obit = extractvalue {i8, i1} %t, 1
  br i1 %obit, label %overflow, label %continue, !prof !0

overflow:
  ret i1 false

continue:
  ret i1 true
}

define zeroext i1 @smulobri16(i16 %v1, i16 %v2) {
; SDAG-LABEL: smulobri16:
; SDAG:       # %bb.0:
; SDAG-NEXT:    imulw %si, %di
; SDAG-NEXT:    jo .LBB16_1
; SDAG-NEXT:  # %bb.2: # %continue
; SDAG-NEXT:    movb $1, %al
; SDAG-NEXT:    retq
; SDAG-NEXT:  .LBB16_1: # %overflow
; SDAG-NEXT:    xorl %eax, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: smulobri16:
; FAST:       # %bb.0:
; FAST-NEXT:    imulw %si, %di
; FAST-NEXT:    seto %al
; FAST-NEXT:    testb $1, %al
; FAST-NEXT:    jne .LBB16_1
; FAST-NEXT:  # %bb.2: # %continue
; FAST-NEXT:    movb $1, %al
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
; FAST-NEXT:  .LBB16_1: # %overflow
; FAST-NEXT:    xorl %eax, %eax
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smulobri16:
; WIN64:       # %bb.0:
; WIN64-NEXT:    imulw %dx, %cx
; WIN64-NEXT:    jo .LBB16_1
; WIN64-NEXT:  # %bb.2: # %continue
; WIN64-NEXT:    movb $1, %al
; WIN64-NEXT:    retq
; WIN64-NEXT:  .LBB16_1: # %overflow
; WIN64-NEXT:    xorl %eax, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smulobri16:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    imulw {{[0-9]+}}(%esp), %ax
; WIN32-NEXT:    jo LBB16_1
; WIN32-NEXT:  # %bb.2: # %continue
; WIN32-NEXT:    movb $1, %al
; WIN32-NEXT:    retl
; WIN32-NEXT:  LBB16_1: # %overflow
; WIN32-NEXT:    xorl %eax, %eax
; WIN32-NEXT:    retl
  %t = call {i16, i1} @llvm.smul.with.overflow.i16(i16 %v1, i16 %v2)
  %val = extractvalue {i16, i1} %t, 0
  %obit = extractvalue {i16, i1} %t, 1
  br i1 %obit, label %overflow, label %continue, !prof !0

overflow:
  ret i1 false

continue:
  ret i1 true
}

define zeroext i1 @smulobri32(i32 %v1, i32 %v2) {
; SDAG-LABEL: smulobri32:
; SDAG:       # %bb.0:
; SDAG-NEXT:    imull %esi, %edi
; SDAG-NEXT:    jo .LBB17_1
; SDAG-NEXT:  # %bb.2: # %continue
; SDAG-NEXT:    movb $1, %al
; SDAG-NEXT:    retq
; SDAG-NEXT:  .LBB17_1: # %overflow
; SDAG-NEXT:    xorl %eax, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: smulobri32:
; FAST:       # %bb.0:
; FAST-NEXT:    imull %esi, %edi
; FAST-NEXT:    jo .LBB17_1
; FAST-NEXT:  # %bb.2: # %continue
; FAST-NEXT:    movb $1, %al
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
; FAST-NEXT:  .LBB17_1: # %overflow
; FAST-NEXT:    xorl %eax, %eax
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smulobri32:
; WIN64:       # %bb.0:
; WIN64-NEXT:    imull %edx, %ecx
; WIN64-NEXT:    jo .LBB17_1
; WIN64-NEXT:  # %bb.2: # %continue
; WIN64-NEXT:    movb $1, %al
; WIN64-NEXT:    retq
; WIN64-NEXT:  .LBB17_1: # %overflow
; WIN64-NEXT:    xorl %eax, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smulobri32:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    imull {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    jo LBB17_1
; WIN32-NEXT:  # %bb.2: # %continue
; WIN32-NEXT:    movb $1, %al
; WIN32-NEXT:    retl
; WIN32-NEXT:  LBB17_1: # %overflow
; WIN32-NEXT:    xorl %eax, %eax
; WIN32-NEXT:    retl
  %t = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %v1, i32 %v2)
  %val = extractvalue {i32, i1} %t, 0
  %obit = extractvalue {i32, i1} %t, 1
  br i1 %obit, label %overflow, label %continue, !prof !0

overflow:
  ret i1 false

continue:
  ret i1 true
}

define zeroext i1 @smulobri64(i64 %v1, i64 %v2) {
; SDAG-LABEL: smulobri64:
; SDAG:       # %bb.0:
; SDAG-NEXT:    imulq %rsi, %rdi
; SDAG-NEXT:    jo .LBB18_1
; SDAG-NEXT:  # %bb.2: # %continue
; SDAG-NEXT:    movb $1, %al
; SDAG-NEXT:    retq
; SDAG-NEXT:  .LBB18_1: # %overflow
; SDAG-NEXT:    xorl %eax, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: smulobri64:
; FAST:       # %bb.0:
; FAST-NEXT:    imulq %rsi, %rdi
; FAST-NEXT:    jo .LBB18_1
; FAST-NEXT:  # %bb.2: # %continue
; FAST-NEXT:    movb $1, %al
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
; FAST-NEXT:  .LBB18_1: # %overflow
; FAST-NEXT:    xorl %eax, %eax
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smulobri64:
; WIN64:       # %bb.0:
; WIN64-NEXT:    imulq %rdx, %rcx
; WIN64-NEXT:    jo .LBB18_1
; WIN64-NEXT:  # %bb.2: # %continue
; WIN64-NEXT:    movb $1, %al
; WIN64-NEXT:    retq
; WIN64-NEXT:  .LBB18_1: # %overflow
; WIN64-NEXT:    xorl %eax, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smulobri64:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebp
; WIN32-NEXT:    pushl %ebx
; WIN32-NEXT:    pushl %edi
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    pushl %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl %ecx, %edi
; WIN32-NEXT:    sarl $31, %edi
; WIN32-NEXT:    movl %eax, %esi
; WIN32-NEXT:    imull %edi, %esi
; WIN32-NEXT:    mull %edi
; WIN32-NEXT:    movl %eax, %ebp
; WIN32-NEXT:    addl %esi, %edx
; WIN32-NEXT:    movl %ebx, %esi
; WIN32-NEXT:    imull %ebx, %edi
; WIN32-NEXT:    addl %edx, %edi
; WIN32-NEXT:    sarl $31, %esi
; WIN32-NEXT:    movl %esi, %ebx
; WIN32-NEXT:    imull %ecx, %ebx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl %esi, %eax
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    addl %ebx, %edx
; WIN32-NEXT:    imull %ecx, %esi
; WIN32-NEXT:    addl %edx, %esi
; WIN32-NEXT:    addl %ebp, %eax
; WIN32-NEXT:    movl %eax, (%esp) # 4-byte Spill
; WIN32-NEXT:    adcl %edi, %esi
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    movl %ecx, %edi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    movl %edx, %ebx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    movl %edx, %ebp
; WIN32-NEXT:    movl %eax, %ecx
; WIN32-NEXT:    addl %ebx, %ecx
; WIN32-NEXT:    adcl $0, %ebp
; WIN32-NEXT:    movl %edi, %eax
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    movl %edx, %edi
; WIN32-NEXT:    movl %eax, %ebx
; WIN32-NEXT:    addl %ecx, %ebx
; WIN32-NEXT:    adcl %ebp, %edi
; WIN32-NEXT:    setb %cl
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    addl %edi, %eax
; WIN32-NEXT:    movzbl %cl, %ecx
; WIN32-NEXT:    adcl %ecx, %edx
; WIN32-NEXT:    addl (%esp), %eax # 4-byte Folded Reload
; WIN32-NEXT:    adcl %esi, %edx
; WIN32-NEXT:    sarl $31, %ebx
; WIN32-NEXT:    xorl %ebx, %edx
; WIN32-NEXT:    xorl %eax, %ebx
; WIN32-NEXT:    orl %edx, %ebx
; WIN32-NEXT:    jne LBB18_1
; WIN32-NEXT:  # %bb.3: # %continue
; WIN32-NEXT:    movb $1, %al
; WIN32-NEXT:  LBB18_2: # %overflow
; WIN32-NEXT:    addl $4, %esp
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    popl %edi
; WIN32-NEXT:    popl %ebx
; WIN32-NEXT:    popl %ebp
; WIN32-NEXT:    retl
; WIN32-NEXT:  LBB18_1: # %overflow
; WIN32-NEXT:    xorl %eax, %eax
; WIN32-NEXT:    jmp LBB18_2
  %t = call {i64, i1} @llvm.smul.with.overflow.i64(i64 %v1, i64 %v2)
  %val = extractvalue {i64, i1} %t, 0
  %obit = extractvalue {i64, i1} %t, 1
  br i1 %obit, label %overflow, label %continue, !prof !0

overflow:
  ret i1 false

continue:
  ret i1 true
}

define zeroext i1 @umulobri8(i8 %v1, i8 %v2) {
; SDAG-LABEL: umulobri8:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movl %edi, %eax
; SDAG-NEXT:    # kill: def $al killed $al killed $eax
; SDAG-NEXT:    mulb %sil
; SDAG-NEXT:    jo .LBB19_1
; SDAG-NEXT:  # %bb.2: # %continue
; SDAG-NEXT:    movb $1, %al
; SDAG-NEXT:    retq
; SDAG-NEXT:  .LBB19_1: # %overflow
; SDAG-NEXT:    xorl %eax, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umulobri8:
; FAST:       # %bb.0:
; FAST-NEXT:    movl %edi, %eax
; FAST-NEXT:    # kill: def $al killed $al killed $eax
; FAST-NEXT:    mulb %sil
; FAST-NEXT:    seto %al
; FAST-NEXT:    testb $1, %al
; FAST-NEXT:    jne .LBB19_1
; FAST-NEXT:  # %bb.2: # %continue
; FAST-NEXT:    movb $1, %al
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
; FAST-NEXT:  .LBB19_1: # %overflow
; FAST-NEXT:    xorl %eax, %eax
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umulobri8:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    mulb %dl
; WIN64-NEXT:    jo .LBB19_1
; WIN64-NEXT:  # %bb.2: # %continue
; WIN64-NEXT:    movb $1, %al
; WIN64-NEXT:    retq
; WIN64-NEXT:  .LBB19_1: # %overflow
; WIN64-NEXT:    xorl %eax, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umulobri8:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movb {{[0-9]+}}(%esp), %al
; WIN32-NEXT:    mulb {{[0-9]+}}(%esp)
; WIN32-NEXT:    jo LBB19_1
; WIN32-NEXT:  # %bb.2: # %continue
; WIN32-NEXT:    movb $1, %al
; WIN32-NEXT:    retl
; WIN32-NEXT:  LBB19_1: # %overflow
; WIN32-NEXT:    xorl %eax, %eax
; WIN32-NEXT:    retl
  %t = call {i8, i1} @llvm.umul.with.overflow.i8(i8 %v1, i8 %v2)
  %val = extractvalue {i8, i1} %t, 0
  %obit = extractvalue {i8, i1} %t, 1
  br i1 %obit, label %overflow, label %continue, !prof !0

overflow:
  ret i1 false

continue:
  ret i1 true
}

define zeroext i1 @umulobri16(i16 %v1, i16 %v2) {
; SDAG-LABEL: umulobri16:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movl %edi, %eax
; SDAG-NEXT:    # kill: def $ax killed $ax killed $eax
; SDAG-NEXT:    mulw %si
; SDAG-NEXT:    jo .LBB20_1
; SDAG-NEXT:  # %bb.2: # %continue
; SDAG-NEXT:    movb $1, %al
; SDAG-NEXT:    retq
; SDAG-NEXT:  .LBB20_1: # %overflow
; SDAG-NEXT:    xorl %eax, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umulobri16:
; FAST:       # %bb.0:
; FAST-NEXT:    movl %edi, %eax
; FAST-NEXT:    # kill: def $ax killed $ax killed $eax
; FAST-NEXT:    mulw %si
; FAST-NEXT:    seto %al
; FAST-NEXT:    testb $1, %al
; FAST-NEXT:    jne .LBB20_1
; FAST-NEXT:  # %bb.2: # %continue
; FAST-NEXT:    movb $1, %al
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
; FAST-NEXT:  .LBB20_1: # %overflow
; FAST-NEXT:    xorl %eax, %eax
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umulobri16:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    mulw %dx
; WIN64-NEXT:    jo .LBB20_1
; WIN64-NEXT:  # %bb.2: # %continue
; WIN64-NEXT:    movb $1, %al
; WIN64-NEXT:    retq
; WIN64-NEXT:  .LBB20_1: # %overflow
; WIN64-NEXT:    xorl %eax, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umulobri16:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    mulw {{[0-9]+}}(%esp)
; WIN32-NEXT:    jo LBB20_1
; WIN32-NEXT:  # %bb.2: # %continue
; WIN32-NEXT:    movb $1, %al
; WIN32-NEXT:    retl
; WIN32-NEXT:  LBB20_1: # %overflow
; WIN32-NEXT:    xorl %eax, %eax
; WIN32-NEXT:    retl
  %t = call {i16, i1} @llvm.umul.with.overflow.i16(i16 %v1, i16 %v2)
  %val = extractvalue {i16, i1} %t, 0
  %obit = extractvalue {i16, i1} %t, 1
  br i1 %obit, label %overflow, label %continue, !prof !0

overflow:
  ret i1 false

continue:
  ret i1 true
}

define zeroext i1 @umulobri32(i32 %v1, i32 %v2) {
; SDAG-LABEL: umulobri32:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movl %edi, %eax
; SDAG-NEXT:    mull %esi
; SDAG-NEXT:    jo .LBB21_1
; SDAG-NEXT:  # %bb.2: # %continue
; SDAG-NEXT:    movb $1, %al
; SDAG-NEXT:    retq
; SDAG-NEXT:  .LBB21_1: # %overflow
; SDAG-NEXT:    xorl %eax, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umulobri32:
; FAST:       # %bb.0:
; FAST-NEXT:    movl %edi, %eax
; FAST-NEXT:    mull %esi
; FAST-NEXT:    jo .LBB21_1
; FAST-NEXT:  # %bb.2: # %continue
; FAST-NEXT:    movb $1, %al
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
; FAST-NEXT:  .LBB21_1: # %overflow
; FAST-NEXT:    xorl %eax, %eax
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umulobri32:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    mull %edx
; WIN64-NEXT:    jo .LBB21_1
; WIN64-NEXT:  # %bb.2: # %continue
; WIN64-NEXT:    movb $1, %al
; WIN64-NEXT:    retq
; WIN64-NEXT:  .LBB21_1: # %overflow
; WIN64-NEXT:    xorl %eax, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umulobri32:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    jo LBB21_1
; WIN32-NEXT:  # %bb.2: # %continue
; WIN32-NEXT:    movb $1, %al
; WIN32-NEXT:    retl
; WIN32-NEXT:  LBB21_1: # %overflow
; WIN32-NEXT:    xorl %eax, %eax
; WIN32-NEXT:    retl
  %t = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %v1, i32 %v2)
  %val = extractvalue {i32, i1} %t, 0
  %obit = extractvalue {i32, i1} %t, 1
  br i1 %obit, label %overflow, label %continue, !prof !0

overflow:
  ret i1 false

continue:
  ret i1 true
}

define zeroext i1 @umulobri64(i64 %v1, i64 %v2) {
; SDAG-LABEL: umulobri64:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movq %rdi, %rax
; SDAG-NEXT:    mulq %rsi
; SDAG-NEXT:    jo .LBB22_1
; SDAG-NEXT:  # %bb.2: # %continue
; SDAG-NEXT:    movb $1, %al
; SDAG-NEXT:    retq
; SDAG-NEXT:  .LBB22_1: # %overflow
; SDAG-NEXT:    xorl %eax, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umulobri64:
; FAST:       # %bb.0:
; FAST-NEXT:    movq %rdi, %rax
; FAST-NEXT:    mulq %rsi
; FAST-NEXT:    jo .LBB22_1
; FAST-NEXT:  # %bb.2: # %continue
; FAST-NEXT:    movb $1, %al
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
; FAST-NEXT:  .LBB22_1: # %overflow
; FAST-NEXT:    xorl %eax, %eax
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umulobri64:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movq %rcx, %rax
; WIN64-NEXT:    mulq %rdx
; WIN64-NEXT:    jo .LBB22_1
; WIN64-NEXT:  # %bb.2: # %continue
; WIN64-NEXT:    movb $1, %al
; WIN64-NEXT:    retq
; WIN64-NEXT:  .LBB22_1: # %overflow
; WIN64-NEXT:    xorl %eax, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umulobri64:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebp
; WIN32-NEXT:    pushl %ebx
; WIN32-NEXT:    pushl %edi
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ebp
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    testl %esi, %esi
; WIN32-NEXT:    setne %dl
; WIN32-NEXT:    testl %eax, %eax
; WIN32-NEXT:    setne %bl
; WIN32-NEXT:    andb %dl, %bl
; WIN32-NEXT:    mull %ebp
; WIN32-NEXT:    movl %eax, %edi
; WIN32-NEXT:    seto %bh
; WIN32-NEXT:    movl %esi, %eax
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    movl %ecx, %edx
; WIN32-NEXT:    movl %eax, %esi
; WIN32-NEXT:    seto %cl
; WIN32-NEXT:    orb %bh, %cl
; WIN32-NEXT:    addl %edi, %esi
; WIN32-NEXT:    movl %edx, %eax
; WIN32-NEXT:    mull %ebp
; WIN32-NEXT:    addl %esi, %edx
; WIN32-NEXT:    setb %al
; WIN32-NEXT:    orb %cl, %al
; WIN32-NEXT:    orb %bl, %al
; WIN32-NEXT:    subb $1, %al
; WIN32-NEXT:    je LBB22_1
; WIN32-NEXT:  # %bb.3: # %continue
; WIN32-NEXT:    movb $1, %al
; WIN32-NEXT:  LBB22_2: # %overflow
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    popl %edi
; WIN32-NEXT:    popl %ebx
; WIN32-NEXT:    popl %ebp
; WIN32-NEXT:    retl
; WIN32-NEXT:  LBB22_1: # %overflow
; WIN32-NEXT:    xorl %eax, %eax
; WIN32-NEXT:    jmp LBB22_2
  %t = call {i64, i1} @llvm.umul.with.overflow.i64(i64 %v1, i64 %v2)
  %val = extractvalue {i64, i1} %t, 0
  %obit = extractvalue {i64, i1} %t, 1
  br i1 %obit, label %overflow, label %continue, !prof !0

overflow:
  ret i1 false

continue:
  ret i1 true
}

define i1 @bug27873(i64 %c1, i1 %c2) {
; LINUX-LABEL: bug27873:
; LINUX:       # %bb.0:
; LINUX-NEXT:    movq %rdi, %rax
; LINUX-NEXT:    movl $160, %ecx
; LINUX-NEXT:    mulq %rcx
; LINUX-NEXT:    seto %al
; LINUX-NEXT:    orb %sil, %al
; LINUX-NEXT:    retq
;
; WIN64-LABEL: bug27873:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %edx, %r8d
; WIN64-NEXT:    movq %rcx, %rax
; WIN64-NEXT:    movl $160, %ecx
; WIN64-NEXT:    mulq %rcx
; WIN64-NEXT:    seto %al
; WIN64-NEXT:    orb %r8b, %al
; WIN64-NEXT:    retq
;
; WIN32-LABEL: bug27873:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebx
; WIN32-NEXT:    movl $160, %eax
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    movl %eax, %ecx
; WIN32-NEXT:    seto %bl
; WIN32-NEXT:    movl $160, %eax
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    addl %ecx, %edx
; WIN32-NEXT:    setb %al
; WIN32-NEXT:    orb %bl, %al
; WIN32-NEXT:    orb {{[0-9]+}}(%esp), %al
; WIN32-NEXT:    popl %ebx
; WIN32-NEXT:    retl
  %mul = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 %c1, i64 160)
  %mul.overflow = extractvalue { i64, i1 } %mul, 1
  %x1 = or i1 %c2, %mul.overflow
  ret i1 %x1
}

define zeroext i1 @smuloi8_load(i8* %ptr1, i8 %v2, i8* %res) {
; SDAG-LABEL: smuloi8_load:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movl %esi, %eax
; SDAG-NEXT:    # kill: def $al killed $al killed $eax
; SDAG-NEXT:    imulb (%rdi)
; SDAG-NEXT:    seto %cl
; SDAG-NEXT:    movb %al, (%rdx)
; SDAG-NEXT:    movl %ecx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: smuloi8_load:
; FAST:       # %bb.0:
; FAST-NEXT:    movb (%rdi), %al
; FAST-NEXT:    imulb %sil
; FAST-NEXT:    seto %cl
; FAST-NEXT:    movb %al, (%rdx)
; FAST-NEXT:    andb $1, %cl
; FAST-NEXT:    movzbl %cl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smuloi8_load:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %edx, %eax
; WIN64-NEXT:    imulb (%rcx)
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movb %al, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloi8_load:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %edx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movb (%eax), %al
; WIN32-NEXT:    imulb {{[0-9]+}}(%esp)
; WIN32-NEXT:    seto %cl
; WIN32-NEXT:    movb %al, (%edx)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    retl
  %v1 = load i8, i8* %ptr1
  %t = call {i8, i1} @llvm.smul.with.overflow.i8(i8 %v1, i8 %v2)
  %val = extractvalue {i8, i1} %t, 0
  %obit = extractvalue {i8, i1} %t, 1
  store i8 %val, i8* %res
  ret i1 %obit
}

define zeroext i1 @smuloi8_load2(i8 %v1, i8* %ptr2, i8* %res) {
; SDAG-LABEL: smuloi8_load2:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movl %edi, %eax
; SDAG-NEXT:    # kill: def $al killed $al killed $eax
; SDAG-NEXT:    imulb (%rsi)
; SDAG-NEXT:    seto %cl
; SDAG-NEXT:    movb %al, (%rdx)
; SDAG-NEXT:    movl %ecx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: smuloi8_load2:
; FAST:       # %bb.0:
; FAST-NEXT:    movl %edi, %eax
; FAST-NEXT:    # kill: def $al killed $al killed $eax
; FAST-NEXT:    imulb (%rsi)
; FAST-NEXT:    seto %cl
; FAST-NEXT:    movb %al, (%rdx)
; FAST-NEXT:    andb $1, %cl
; FAST-NEXT:    movzbl %cl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smuloi8_load2:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    imulb (%rdx)
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movb %al, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloi8_load2:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %edx
; WIN32-NEXT:    movb {{[0-9]+}}(%esp), %al
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    imulb (%ecx)
; WIN32-NEXT:    seto %cl
; WIN32-NEXT:    movb %al, (%edx)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    retl
  %v2 = load i8, i8* %ptr2
  %t = call {i8, i1} @llvm.smul.with.overflow.i8(i8 %v1, i8 %v2)
  %val = extractvalue {i8, i1} %t, 0
  %obit = extractvalue {i8, i1} %t, 1
  store i8 %val, i8* %res
  ret i1 %obit
}

define zeroext i1 @smuloi16_load(i16* %ptr1, i16 %v2, i16* %res) {
; SDAG-LABEL: smuloi16_load:
; SDAG:       # %bb.0:
; SDAG-NEXT:    imulw (%rdi), %si
; SDAG-NEXT:    seto %al
; SDAG-NEXT:    movw %si, (%rdx)
; SDAG-NEXT:    retq
;
; FAST-LABEL: smuloi16_load:
; FAST:       # %bb.0:
; FAST-NEXT:    imulw (%rdi), %si
; FAST-NEXT:    seto %al
; FAST-NEXT:    movw %si, (%rdx)
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smuloi16_load:
; WIN64:       # %bb.0:
; WIN64-NEXT:    imulw (%rcx), %dx
; WIN64-NEXT:    seto %al
; WIN64-NEXT:    movw %dx, (%r8)
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloi16_load:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movzwl (%eax), %edx
; WIN32-NEXT:    imulw {{[0-9]+}}(%esp), %dx
; WIN32-NEXT:    seto %al
; WIN32-NEXT:    movw %dx, (%ecx)
; WIN32-NEXT:    retl
  %v1 = load i16, i16* %ptr1
  %t = call {i16, i1} @llvm.smul.with.overflow.i16(i16 %v1, i16 %v2)
  %val = extractvalue {i16, i1} %t, 0
  %obit = extractvalue {i16, i1} %t, 1
  store i16 %val, i16* %res
  ret i1 %obit
}

define zeroext i1 @smuloi16_load2(i16 %v1, i16* %ptr2, i16* %res) {
; SDAG-LABEL: smuloi16_load2:
; SDAG:       # %bb.0:
; SDAG-NEXT:    imulw (%rsi), %di
; SDAG-NEXT:    seto %al
; SDAG-NEXT:    movw %di, (%rdx)
; SDAG-NEXT:    retq
;
; FAST-LABEL: smuloi16_load2:
; FAST:       # %bb.0:
; FAST-NEXT:    imulw (%rsi), %di
; FAST-NEXT:    seto %al
; FAST-NEXT:    movw %di, (%rdx)
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smuloi16_load2:
; WIN64:       # %bb.0:
; WIN64-NEXT:    imulw (%rdx), %cx
; WIN64-NEXT:    seto %al
; WIN64-NEXT:    movw %cx, (%r8)
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloi16_load2:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movzwl {{[0-9]+}}(%esp), %edx
; WIN32-NEXT:    imulw (%eax), %dx
; WIN32-NEXT:    seto %al
; WIN32-NEXT:    movw %dx, (%ecx)
; WIN32-NEXT:    retl
  %v2 = load i16, i16* %ptr2
  %t = call {i16, i1} @llvm.smul.with.overflow.i16(i16 %v1, i16 %v2)
  %val = extractvalue {i16, i1} %t, 0
  %obit = extractvalue {i16, i1} %t, 1
  store i16 %val, i16* %res
  ret i1 %obit
}

define zeroext i1 @smuloi32_load(i32* %ptr1, i32 %v2, i32* %res) {
; SDAG-LABEL: smuloi32_load:
; SDAG:       # %bb.0:
; SDAG-NEXT:    imull (%rdi), %esi
; SDAG-NEXT:    seto %al
; SDAG-NEXT:    movl %esi, (%rdx)
; SDAG-NEXT:    retq
;
; FAST-LABEL: smuloi32_load:
; FAST:       # %bb.0:
; FAST-NEXT:    imull (%rdi), %esi
; FAST-NEXT:    seto %al
; FAST-NEXT:    movl %esi, (%rdx)
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smuloi32_load:
; WIN64:       # %bb.0:
; WIN64-NEXT:    imull (%rcx), %edx
; WIN64-NEXT:    seto %al
; WIN64-NEXT:    movl %edx, (%r8)
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloi32_load:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl (%eax), %edx
; WIN32-NEXT:    imull {{[0-9]+}}(%esp), %edx
; WIN32-NEXT:    seto %al
; WIN32-NEXT:    movl %edx, (%ecx)
; WIN32-NEXT:    retl
  %v1 = load i32, i32* %ptr1
  %t = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %v1, i32 %v2)
  %val = extractvalue {i32, i1} %t, 0
  %obit = extractvalue {i32, i1} %t, 1
  store i32 %val, i32* %res
  ret i1 %obit
}

define zeroext i1 @smuloi32_load2(i32 %v1, i32* %ptr2, i32* %res) {
; SDAG-LABEL: smuloi32_load2:
; SDAG:       # %bb.0:
; SDAG-NEXT:    imull (%rsi), %edi
; SDAG-NEXT:    seto %al
; SDAG-NEXT:    movl %edi, (%rdx)
; SDAG-NEXT:    retq
;
; FAST-LABEL: smuloi32_load2:
; FAST:       # %bb.0:
; FAST-NEXT:    imull (%rsi), %edi
; FAST-NEXT:    seto %al
; FAST-NEXT:    movl %edi, (%rdx)
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smuloi32_load2:
; WIN64:       # %bb.0:
; WIN64-NEXT:    imull (%rdx), %ecx
; WIN64-NEXT:    seto %al
; WIN64-NEXT:    movl %ecx, (%r8)
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloi32_load2:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %edx
; WIN32-NEXT:    imull (%eax), %edx
; WIN32-NEXT:    seto %al
; WIN32-NEXT:    movl %edx, (%ecx)
; WIN32-NEXT:    retl
  %v2 = load i32, i32* %ptr2
  %t = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %v1, i32 %v2)
  %val = extractvalue {i32, i1} %t, 0
  %obit = extractvalue {i32, i1} %t, 1
  store i32 %val, i32* %res
  ret i1 %obit
}

define zeroext i1 @smuloi64_load(i64* %ptr1, i64 %v2, i64* %res) {
; SDAG-LABEL: smuloi64_load:
; SDAG:       # %bb.0:
; SDAG-NEXT:    imulq (%rdi), %rsi
; SDAG-NEXT:    seto %al
; SDAG-NEXT:    movq %rsi, (%rdx)
; SDAG-NEXT:    retq
;
; FAST-LABEL: smuloi64_load:
; FAST:       # %bb.0:
; FAST-NEXT:    imulq (%rdi), %rsi
; FAST-NEXT:    seto %al
; FAST-NEXT:    movq %rsi, (%rdx)
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smuloi64_load:
; WIN64:       # %bb.0:
; WIN64-NEXT:    imulq (%rcx), %rdx
; WIN64-NEXT:    seto %al
; WIN64-NEXT:    movq %rdx, (%r8)
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloi64_load:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebp
; WIN32-NEXT:    pushl %ebx
; WIN32-NEXT:    pushl %edi
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    subl $16, %esp
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl (%eax), %esi
; WIN32-NEXT:    movl 4(%eax), %ebp
; WIN32-NEXT:    sarl $31, %ebx
; WIN32-NEXT:    movl %ebx, %ecx
; WIN32-NEXT:    imull %ebp, %ecx
; WIN32-NEXT:    movl %ebx, %eax
; WIN32-NEXT:    mull %esi
; WIN32-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; WIN32-NEXT:    addl %ecx, %edx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    imull %esi, %ebx
; WIN32-NEXT:    addl %edx, %ebx
; WIN32-NEXT:    movl %ebp, %ecx
; WIN32-NEXT:    movl %ebp, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; WIN32-NEXT:    sarl $31, %ecx
; WIN32-NEXT:    movl %eax, %edi
; WIN32-NEXT:    imull %ecx, %edi
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    addl %edi, %edx
; WIN32-NEXT:    imull {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    addl %edx, %ecx
; WIN32-NEXT:    addl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
; WIN32-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; WIN32-NEXT:    adcl %ebx, %ecx
; WIN32-NEXT:    movl %esi, %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %edi
; WIN32-NEXT:    mull %edi
; WIN32-NEXT:    movl %edx, %ebx
; WIN32-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; WIN32-NEXT:    movl %ebp, %eax
; WIN32-NEXT:    mull %edi
; WIN32-NEXT:    movl %edx, %ebp
; WIN32-NEXT:    movl %eax, %edi
; WIN32-NEXT:    addl %ebx, %edi
; WIN32-NEXT:    adcl $0, %ebp
; WIN32-NEXT:    movl %esi, %eax
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    movl %edx, %ebx
; WIN32-NEXT:    movl %eax, %esi
; WIN32-NEXT:    addl %edi, %esi
; WIN32-NEXT:    adcl %ebp, %ebx
; WIN32-NEXT:    setb {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Folded Spill
; WIN32-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    addl %ebx, %eax
; WIN32-NEXT:    movzbl {{[-0-9]+}}(%e{{[sb]}}p), %edi # 1-byte Folded Reload
; WIN32-NEXT:    adcl %edi, %edx
; WIN32-NEXT:    addl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
; WIN32-NEXT:    adcl %ecx, %edx
; WIN32-NEXT:    movl %esi, %ecx
; WIN32-NEXT:    sarl $31, %ecx
; WIN32-NEXT:    xorl %ecx, %edx
; WIN32-NEXT:    xorl %eax, %ecx
; WIN32-NEXT:    orl %edx, %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl %esi, 4(%eax)
; WIN32-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; WIN32-NEXT:    movl %ecx, (%eax)
; WIN32-NEXT:    setne %al
; WIN32-NEXT:    addl $16, %esp
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    popl %edi
; WIN32-NEXT:    popl %ebx
; WIN32-NEXT:    popl %ebp
; WIN32-NEXT:    retl
  %v1 = load i64, i64* %ptr1
  %t = call {i64, i1} @llvm.smul.with.overflow.i64(i64 %v1, i64 %v2)
  %val = extractvalue {i64, i1} %t, 0
  %obit = extractvalue {i64, i1} %t, 1
  store i64 %val, i64* %res
  ret i1 %obit
}

define zeroext i1 @smuloi64_load2(i64 %v1, i64* %ptr2, i64* %res) {
; SDAG-LABEL: smuloi64_load2:
; SDAG:       # %bb.0:
; SDAG-NEXT:    imulq (%rsi), %rdi
; SDAG-NEXT:    seto %al
; SDAG-NEXT:    movq %rdi, (%rdx)
; SDAG-NEXT:    retq
;
; FAST-LABEL: smuloi64_load2:
; FAST:       # %bb.0:
; FAST-NEXT:    imulq (%rsi), %rdi
; FAST-NEXT:    seto %al
; FAST-NEXT:    movq %rdi, (%rdx)
; FAST-NEXT:    andb $1, %al
; FAST-NEXT:    movzbl %al, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: smuloi64_load2:
; WIN64:       # %bb.0:
; WIN64-NEXT:    imulq (%rdx), %rcx
; WIN64-NEXT:    seto %al
; WIN64-NEXT:    movq %rcx, (%r8)
; WIN64-NEXT:    retq
;
; WIN32-LABEL: smuloi64_load2:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebp
; WIN32-NEXT:    pushl %ebx
; WIN32-NEXT:    pushl %edi
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    subl $12, %esp
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl (%eax), %ebp
; WIN32-NEXT:    movl 4(%eax), %ebx
; WIN32-NEXT:    movl %ecx, %edi
; WIN32-NEXT:    sarl $31, %edi
; WIN32-NEXT:    movl %ebp, %esi
; WIN32-NEXT:    imull %edi, %esi
; WIN32-NEXT:    movl %ebp, %eax
; WIN32-NEXT:    mull %edi
; WIN32-NEXT:    movl %eax, (%esp) # 4-byte Spill
; WIN32-NEXT:    addl %esi, %edx
; WIN32-NEXT:    movl %ebx, %esi
; WIN32-NEXT:    movl %ebx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; WIN32-NEXT:    imull %ebx, %edi
; WIN32-NEXT:    addl %edx, %edi
; WIN32-NEXT:    sarl $31, %esi
; WIN32-NEXT:    movl %esi, %ebx
; WIN32-NEXT:    imull %ecx, %ebx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    movl %esi, %eax
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    addl %ebx, %edx
; WIN32-NEXT:    imull %ecx, %esi
; WIN32-NEXT:    addl %edx, %esi
; WIN32-NEXT:    addl (%esp), %eax # 4-byte Folded Reload
; WIN32-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; WIN32-NEXT:    adcl %edi, %esi
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    mull %ebp
; WIN32-NEXT:    movl %edx, %edi
; WIN32-NEXT:    movl %eax, (%esp) # 4-byte Spill
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    mull %ebp
; WIN32-NEXT:    movl %edx, %ebx
; WIN32-NEXT:    movl %eax, %ecx
; WIN32-NEXT:    addl %edi, %ecx
; WIN32-NEXT:    adcl $0, %ebx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    mull {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
; WIN32-NEXT:    movl %edx, %edi
; WIN32-NEXT:    movl %eax, %ebp
; WIN32-NEXT:    addl %ecx, %ebp
; WIN32-NEXT:    adcl %ebx, %edi
; WIN32-NEXT:    setb %cl
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    mull {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Folded Reload
; WIN32-NEXT:    addl %edi, %eax
; WIN32-NEXT:    movzbl %cl, %ecx
; WIN32-NEXT:    adcl %ecx, %edx
; WIN32-NEXT:    addl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Folded Reload
; WIN32-NEXT:    adcl %esi, %edx
; WIN32-NEXT:    movl %ebp, %ecx
; WIN32-NEXT:    sarl $31, %ecx
; WIN32-NEXT:    xorl %ecx, %edx
; WIN32-NEXT:    xorl %eax, %ecx
; WIN32-NEXT:    orl %edx, %ecx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl %ebp, 4(%eax)
; WIN32-NEXT:    movl (%esp), %ecx # 4-byte Reload
; WIN32-NEXT:    movl %ecx, (%eax)
; WIN32-NEXT:    setne %al
; WIN32-NEXT:    addl $12, %esp
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    popl %edi
; WIN32-NEXT:    popl %ebx
; WIN32-NEXT:    popl %ebp
; WIN32-NEXT:    retl
  %v2 = load i64, i64* %ptr2
  %t = call {i64, i1} @llvm.smul.with.overflow.i64(i64 %v1, i64 %v2)
  %val = extractvalue {i64, i1} %t, 0
  %obit = extractvalue {i64, i1} %t, 1
  store i64 %val, i64* %res
  ret i1 %obit
}

define zeroext i1 @umuloi8_load(i8* %ptr1, i8 %v2, i8* %res) {
; SDAG-LABEL: umuloi8_load:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movl %esi, %eax
; SDAG-NEXT:    # kill: def $al killed $al killed $eax
; SDAG-NEXT:    mulb (%rdi)
; SDAG-NEXT:    seto %cl
; SDAG-NEXT:    movb %al, (%rdx)
; SDAG-NEXT:    movl %ecx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umuloi8_load:
; FAST:       # %bb.0:
; FAST-NEXT:    movb (%rdi), %al
; FAST-NEXT:    mulb %sil
; FAST-NEXT:    seto %cl
; FAST-NEXT:    movb %al, (%rdx)
; FAST-NEXT:    andb $1, %cl
; FAST-NEXT:    movzbl %cl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umuloi8_load:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %edx, %eax
; WIN64-NEXT:    mulb (%rcx)
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movb %al, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloi8_load:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %edx
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movb (%eax), %al
; WIN32-NEXT:    mulb {{[0-9]+}}(%esp)
; WIN32-NEXT:    seto %cl
; WIN32-NEXT:    movb %al, (%edx)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    retl
  %v1 = load i8, i8* %ptr1
  %t = call {i8, i1} @llvm.umul.with.overflow.i8(i8 %v1, i8 %v2)
  %val = extractvalue {i8, i1} %t, 0
  %obit = extractvalue {i8, i1} %t, 1
  store i8 %val, i8* %res
  ret i1 %obit
}

define zeroext i1 @umuloi8_load2(i8 %v1, i8* %ptr2, i8* %res) {
; SDAG-LABEL: umuloi8_load2:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movl %edi, %eax
; SDAG-NEXT:    # kill: def $al killed $al killed $eax
; SDAG-NEXT:    mulb (%rsi)
; SDAG-NEXT:    seto %cl
; SDAG-NEXT:    movb %al, (%rdx)
; SDAG-NEXT:    movl %ecx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umuloi8_load2:
; FAST:       # %bb.0:
; FAST-NEXT:    movl %edi, %eax
; FAST-NEXT:    # kill: def $al killed $al killed $eax
; FAST-NEXT:    mulb (%rsi)
; FAST-NEXT:    seto %cl
; FAST-NEXT:    movb %al, (%rdx)
; FAST-NEXT:    andb $1, %cl
; FAST-NEXT:    movzbl %cl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umuloi8_load2:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    mulb (%rdx)
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movb %al, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloi8_load2:
; WIN32:       # %bb.0:
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %edx
; WIN32-NEXT:    movb {{[0-9]+}}(%esp), %al
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    mulb (%ecx)
; WIN32-NEXT:    seto %cl
; WIN32-NEXT:    movb %al, (%edx)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    retl
  %v2 = load i8, i8* %ptr2
  %t = call {i8, i1} @llvm.umul.with.overflow.i8(i8 %v1, i8 %v2)
  %val = extractvalue {i8, i1} %t, 0
  %obit = extractvalue {i8, i1} %t, 1
  store i8 %val, i8* %res
  ret i1 %obit
}

define zeroext i1 @umuloi16_load(i16* %ptr1, i16 %v2, i16* %res) {
; SDAG-LABEL: umuloi16_load:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movq %rdx, %rcx
; SDAG-NEXT:    movl %esi, %eax
; SDAG-NEXT:    # kill: def $ax killed $ax killed $eax
; SDAG-NEXT:    mulw (%rdi)
; SDAG-NEXT:    seto %dl
; SDAG-NEXT:    movw %ax, (%rcx)
; SDAG-NEXT:    movl %edx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umuloi16_load:
; FAST:       # %bb.0:
; FAST-NEXT:    movq %rdx, %rcx
; FAST-NEXT:    movzwl (%rdi), %eax
; FAST-NEXT:    mulw %si
; FAST-NEXT:    seto %dl
; FAST-NEXT:    movw %ax, (%rcx)
; FAST-NEXT:    andb $1, %dl
; FAST-NEXT:    movzbl %dl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umuloi16_load:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %edx, %eax
; WIN64-NEXT:    mulw (%rcx)
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movw %ax, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloi16_load:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movzwl (%eax), %eax
; WIN32-NEXT:    mulw {{[0-9]+}}(%esp)
; WIN32-NEXT:    seto %cl
; WIN32-NEXT:    movw %ax, (%esi)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    retl
  %v1 = load i16, i16* %ptr1
  %t = call {i16, i1} @llvm.umul.with.overflow.i16(i16 %v1, i16 %v2)
  %val = extractvalue {i16, i1} %t, 0
  %obit = extractvalue {i16, i1} %t, 1
  store i16 %val, i16* %res
  ret i1 %obit
}

define zeroext i1 @umuloi16_load2(i16 %v1, i16* %ptr2, i16* %res) {
; SDAG-LABEL: umuloi16_load2:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movq %rdx, %rcx
; SDAG-NEXT:    movl %edi, %eax
; SDAG-NEXT:    # kill: def $ax killed $ax killed $eax
; SDAG-NEXT:    mulw (%rsi)
; SDAG-NEXT:    seto %dl
; SDAG-NEXT:    movw %ax, (%rcx)
; SDAG-NEXT:    movl %edx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umuloi16_load2:
; FAST:       # %bb.0:
; FAST-NEXT:    movq %rdx, %rcx
; FAST-NEXT:    movl %edi, %eax
; FAST-NEXT:    # kill: def $ax killed $ax killed $eax
; FAST-NEXT:    mulw (%rsi)
; FAST-NEXT:    seto %dl
; FAST-NEXT:    movw %ax, (%rcx)
; FAST-NEXT:    andb $1, %dl
; FAST-NEXT:    movzbl %dl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umuloi16_load2:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    mulw (%rdx)
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movw %ax, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloi16_load2:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    mulw (%ecx)
; WIN32-NEXT:    seto %cl
; WIN32-NEXT:    movw %ax, (%esi)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    retl
  %v2 = load i16, i16* %ptr2
  %t = call {i16, i1} @llvm.umul.with.overflow.i16(i16 %v1, i16 %v2)
  %val = extractvalue {i16, i1} %t, 0
  %obit = extractvalue {i16, i1} %t, 1
  store i16 %val, i16* %res
  ret i1 %obit
}

define zeroext i1 @umuloi32_load(i32* %ptr1, i32 %v2, i32* %res) {
; SDAG-LABEL: umuloi32_load:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movq %rdx, %rcx
; SDAG-NEXT:    movl %esi, %eax
; SDAG-NEXT:    mull (%rdi)
; SDAG-NEXT:    seto %dl
; SDAG-NEXT:    movl %eax, (%rcx)
; SDAG-NEXT:    movl %edx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umuloi32_load:
; FAST:       # %bb.0:
; FAST-NEXT:    movq %rdx, %rcx
; FAST-NEXT:    movl (%rdi), %eax
; FAST-NEXT:    mull %esi
; FAST-NEXT:    seto %dl
; FAST-NEXT:    movl %eax, (%rcx)
; FAST-NEXT:    andb $1, %dl
; FAST-NEXT:    movzbl %dl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umuloi32_load:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %edx, %eax
; WIN64-NEXT:    mull (%rcx)
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movl %eax, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloi32_load:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl (%eax), %eax
; WIN32-NEXT:    mull {{[0-9]+}}(%esp)
; WIN32-NEXT:    seto %cl
; WIN32-NEXT:    movl %eax, (%esi)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    retl
  %v1 = load i32, i32* %ptr1
  %t = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %v1, i32 %v2)
  %val = extractvalue {i32, i1} %t, 0
  %obit = extractvalue {i32, i1} %t, 1
  store i32 %val, i32* %res
  ret i1 %obit
}

define zeroext i1 @umuloi32_load2(i32 %v1, i32* %ptr2, i32* %res) {
; SDAG-LABEL: umuloi32_load2:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movq %rdx, %rcx
; SDAG-NEXT:    movl %edi, %eax
; SDAG-NEXT:    mull (%rsi)
; SDAG-NEXT:    seto %dl
; SDAG-NEXT:    movl %eax, (%rcx)
; SDAG-NEXT:    movl %edx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umuloi32_load2:
; FAST:       # %bb.0:
; FAST-NEXT:    movq %rdx, %rcx
; FAST-NEXT:    movl %edi, %eax
; FAST-NEXT:    mull (%rsi)
; FAST-NEXT:    seto %dl
; FAST-NEXT:    movl %eax, (%rcx)
; FAST-NEXT:    andb $1, %dl
; FAST-NEXT:    movzbl %dl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umuloi32_load2:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    mull (%rdx)
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movl %eax, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloi32_load2:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; WIN32-NEXT:    mull (%ecx)
; WIN32-NEXT:    seto %cl
; WIN32-NEXT:    movl %eax, (%esi)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    retl
  %v2 = load i32, i32* %ptr2
  %t = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %v1, i32 %v2)
  %val = extractvalue {i32, i1} %t, 0
  %obit = extractvalue {i32, i1} %t, 1
  store i32 %val, i32* %res
  ret i1 %obit
}

define zeroext i1 @umuloi64_load(i64* %ptr1, i64 %v2, i64* %res) {
; SDAG-LABEL: umuloi64_load:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movq %rdx, %rcx
; SDAG-NEXT:    movq %rsi, %rax
; SDAG-NEXT:    mulq (%rdi)
; SDAG-NEXT:    seto %dl
; SDAG-NEXT:    movq %rax, (%rcx)
; SDAG-NEXT:    movl %edx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umuloi64_load:
; FAST:       # %bb.0:
; FAST-NEXT:    movq %rdx, %rcx
; FAST-NEXT:    movq (%rdi), %rax
; FAST-NEXT:    mulq %rsi
; FAST-NEXT:    seto %dl
; FAST-NEXT:    movq %rax, (%rcx)
; FAST-NEXT:    andb $1, %dl
; FAST-NEXT:    movzbl %dl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umuloi64_load:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movq %rdx, %rax
; WIN64-NEXT:    mulq (%rcx)
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movq %rax, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloi64_load:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebp
; WIN32-NEXT:    pushl %ebx
; WIN32-NEXT:    pushl %edi
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    pushl %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ebp
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl (%eax), %ecx
; WIN32-NEXT:    movl 4(%eax), %eax
; WIN32-NEXT:    testl %esi, %esi
; WIN32-NEXT:    setne %dl
; WIN32-NEXT:    testl %eax, %eax
; WIN32-NEXT:    setne %bl
; WIN32-NEXT:    andb %dl, %bl
; WIN32-NEXT:    mull %ebp
; WIN32-NEXT:    movl %eax, %edi
; WIN32-NEXT:    seto {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Folded Spill
; WIN32-NEXT:    movl %esi, %eax
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    movl %eax, %esi
; WIN32-NEXT:    seto %bh
; WIN32-NEXT:    orb {{[-0-9]+}}(%e{{[sb]}}p), %bh # 1-byte Folded Reload
; WIN32-NEXT:    addl %edi, %esi
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    mull %ebp
; WIN32-NEXT:    addl %esi, %edx
; WIN32-NEXT:    setb %cl
; WIN32-NEXT:    orb %bh, %cl
; WIN32-NEXT:    orb %bl, %cl
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    movl %eax, (%esi)
; WIN32-NEXT:    movl %edx, 4(%esi)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    addl $4, %esp
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    popl %edi
; WIN32-NEXT:    popl %ebx
; WIN32-NEXT:    popl %ebp
; WIN32-NEXT:    retl
  %v1 = load i64, i64* %ptr1
  %t = call {i64, i1} @llvm.umul.with.overflow.i64(i64 %v1, i64 %v2)
  %val = extractvalue {i64, i1} %t, 0
  %obit = extractvalue {i64, i1} %t, 1
  store i64 %val, i64* %res
  ret i1 %obit
}

define zeroext i1 @umuloi64_load2(i64 %v1, i64* %ptr2, i64* %res) {
; SDAG-LABEL: umuloi64_load2:
; SDAG:       # %bb.0:
; SDAG-NEXT:    movq %rdx, %rcx
; SDAG-NEXT:    movq %rdi, %rax
; SDAG-NEXT:    mulq (%rsi)
; SDAG-NEXT:    seto %dl
; SDAG-NEXT:    movq %rax, (%rcx)
; SDAG-NEXT:    movl %edx, %eax
; SDAG-NEXT:    retq
;
; FAST-LABEL: umuloi64_load2:
; FAST:       # %bb.0:
; FAST-NEXT:    movq %rdx, %rcx
; FAST-NEXT:    movq %rdi, %rax
; FAST-NEXT:    mulq (%rsi)
; FAST-NEXT:    seto %dl
; FAST-NEXT:    movq %rax, (%rcx)
; FAST-NEXT:    andb $1, %dl
; FAST-NEXT:    movzbl %dl, %eax
; FAST-NEXT:    retq
;
; WIN64-LABEL: umuloi64_load2:
; WIN64:       # %bb.0:
; WIN64-NEXT:    movq %rcx, %rax
; WIN64-NEXT:    mulq (%rdx)
; WIN64-NEXT:    seto %cl
; WIN64-NEXT:    movq %rax, (%r8)
; WIN64-NEXT:    movl %ecx, %eax
; WIN64-NEXT:    retq
;
; WIN32-LABEL: umuloi64_load2:
; WIN32:       # %bb.0:
; WIN32-NEXT:    pushl %ebp
; WIN32-NEXT:    pushl %ebx
; WIN32-NEXT:    pushl %edi
; WIN32-NEXT:    pushl %esi
; WIN32-NEXT:    pushl %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %ebp
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %edx
; WIN32-NEXT:    movl (%edx), %ecx
; WIN32-NEXT:    movl 4(%edx), %esi
; WIN32-NEXT:    testl %eax, %eax
; WIN32-NEXT:    setne %dl
; WIN32-NEXT:    testl %esi, %esi
; WIN32-NEXT:    setne %bl
; WIN32-NEXT:    andb %dl, %bl
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    movl %eax, %edi
; WIN32-NEXT:    seto {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Folded Spill
; WIN32-NEXT:    movl %esi, %eax
; WIN32-NEXT:    mull %ebp
; WIN32-NEXT:    movl %eax, %esi
; WIN32-NEXT:    seto %bh
; WIN32-NEXT:    orb {{[-0-9]+}}(%e{{[sb]}}p), %bh # 1-byte Folded Reload
; WIN32-NEXT:    addl %edi, %esi
; WIN32-NEXT:    movl %ebp, %eax
; WIN32-NEXT:    mull %ecx
; WIN32-NEXT:    addl %esi, %edx
; WIN32-NEXT:    setb %cl
; WIN32-NEXT:    orb %bh, %cl
; WIN32-NEXT:    orb %bl, %cl
; WIN32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; WIN32-NEXT:    movl %eax, (%esi)
; WIN32-NEXT:    movl %edx, 4(%esi)
; WIN32-NEXT:    movl %ecx, %eax
; WIN32-NEXT:    addl $4, %esp
; WIN32-NEXT:    popl %esi
; WIN32-NEXT:    popl %edi
; WIN32-NEXT:    popl %ebx
; WIN32-NEXT:    popl %ebp
; WIN32-NEXT:    retl
  %v2 = load i64, i64* %ptr2
  %t = call {i64, i1} @llvm.umul.with.overflow.i64(i64 %v1, i64 %v2)
  %val = extractvalue {i64, i1} %t, 0
  %obit = extractvalue {i64, i1} %t, 1
  store i64 %val, i64* %res
  ret i1 %obit
}

declare {i8,  i1} @llvm.smul.with.overflow.i8 (i8,  i8 ) nounwind readnone
declare {i16, i1} @llvm.smul.with.overflow.i16(i16, i16) nounwind readnone
declare {i32, i1} @llvm.smul.with.overflow.i32(i32, i32) nounwind readnone
declare {i64, i1} @llvm.smul.with.overflow.i64(i64, i64) nounwind readnone
declare {i8,  i1} @llvm.umul.with.overflow.i8 (i8,  i8 ) nounwind readnone
declare {i16, i1} @llvm.umul.with.overflow.i16(i16, i16) nounwind readnone
declare {i32, i1} @llvm.umul.with.overflow.i32(i32, i32) nounwind readnone
declare {i64, i1} @llvm.umul.with.overflow.i64(i64, i64) nounwind readnone

!0 = !{!"branch_weights", i32 0, i32 2147483647}