; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -verify-machineinstrs | FileCheck %s --check-prefix X64 --check-prefix FAST_INC
; RUN: llc < %s -mtriple=i686-unknown-unknown -verify-machineinstrs | FileCheck %s --check-prefix X32
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=slow-incdec -verify-machineinstrs | FileCheck %s --check-prefix X64 --check-prefix SLOW_INC

; This file checks that atomic (non-seq_cst) stores of immediate values are
; done in one mov instruction and not 2. More precisely, it makes sure that the
; immediate is not first copied uselessly into a register.

; Similarily, it checks that a binary operation of an immediate with an atomic
; variable that is stored back in that variable is done as a single instruction.
; For example: x.store(42 + x.load(memory_order_acquire), memory_order_release)
; should be just an add instruction, instead of loading x into a register, doing
; an add and storing the result back.
; The binary operations supported are currently add, and, or, xor.
; sub is not supported because they are translated by an addition of the
; negated immediate.
;
; We also check the same patterns:
; - For inc/dec.
; - For register instead of immediate operands.
; - For floating point operations.

; seq_cst stores are left as (lock) xchgl, but we try to check every other
; attribute at least once.

; Please note that these operations do not require the lock prefix: only
; sequentially consistent stores require this kind of protection on X86.
; And even for seq_cst operations, llvm uses the xchg instruction which has
; an implicit lock prefix, so making it explicit is not required.

define void @store_atomic_imm_8(i8* %p) {
; X64-LABEL: store_atomic_imm_8:
; X64:       # %bb.0:
; X64-NEXT:    movb $42, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: store_atomic_imm_8:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movb $42, (%eax)
; X32-NEXT:    retl
  store atomic i8 42, i8* %p release, align 1
  ret void
}

define void @store_atomic_imm_16(i16* %p) {
; X64-LABEL: store_atomic_imm_16:
; X64:       # %bb.0:
; X64-NEXT:    movw $42, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: store_atomic_imm_16:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movw $42, (%eax)
; X32-NEXT:    retl
  store atomic i16 42, i16* %p monotonic, align 2
  ret void
}

define void @store_atomic_imm_32(i32* %p) {
; X64-LABEL: store_atomic_imm_32:
; X64:       # %bb.0:
; X64-NEXT:    movl $42, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: store_atomic_imm_32:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl $42, (%eax)
; X32-NEXT:    retl
;   On 32 bits, there is an extra movl for each of those functions
;   (probably for alignment reasons).
  store atomic i32 42, i32* %p release, align 4
  ret void
}

define void @store_atomic_imm_64(i64* %p) {
; X64-LABEL: store_atomic_imm_64:
; X64:       # %bb.0:
; X64-NEXT:    movq $42, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: store_atomic_imm_64:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    movl $42, %ebx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB3_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB3_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;   These are implemented with a CAS loop on 32 bit architectures, and thus
;   cannot be optimized in the same way as the others.
  store atomic i64 42, i64* %p release, align 8
  ret void
}

; If an immediate is too big to fit in 32 bits, it cannot be store in one mov,
; even on X64, one must use movabsq that can only target a register.
define void @store_atomic_imm_64_big(i64* %p) {
; X64-LABEL: store_atomic_imm_64_big:
; X64:       # %bb.0:
; X64-NEXT:    movabsq $100000000000, %rax # imm = 0x174876E800
; X64-NEXT:    movq %rax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: store_atomic_imm_64_big:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    movl $23, %ecx
; X32-NEXT:    movl $1215752192, %ebx # imm = 0x4876E800
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB4_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB4_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
  store atomic i64 100000000000, i64* %p monotonic, align 8
  ret void
}

; It would be incorrect to replace a lock xchgl by a movl
define void @store_atomic_imm_32_seq_cst(i32* %p) {
; X64-LABEL: store_atomic_imm_32_seq_cst:
; X64:       # %bb.0:
; X64-NEXT:    movl $42, %eax
; X64-NEXT:    xchgl %eax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: store_atomic_imm_32_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl $42, %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
  store atomic i32 42, i32* %p seq_cst, align 4
  ret void
}

; ----- ADD -----

define void @add_8i(i8* %p) {
; X64-LABEL: add_8i:
; X64:       # %bb.0:
; X64-NEXT:    addb $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: add_8i:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    addb $2, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i8, i8* %p seq_cst, align 1
  %2 = add i8 %1, 2
  store atomic i8 %2, i8* %p release, align 1
  ret void
}

define void @add_8r(i8* %p, i8 %v) {
; X64-LABEL: add_8r:
; X64:       # %bb.0:
; X64-NEXT:    addb %sil, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: add_8r:
; X32:       # %bb.0:
; X32-NEXT:    movb {{[0-9]+}}(%esp), %al
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    addb %al, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i8, i8* %p seq_cst, align 1
  %2 = add i8 %1, %v
  store atomic i8 %2, i8* %p release, align 1
  ret void
}

define void @add_16i(i16* %p) {
; X64-LABEL: add_16i:
; X64:       # %bb.0:
; X64-NEXT:    addw $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: add_16i:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    addw $2, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i16, i16* %p acquire, align 2
  %2 = add i16 %1, 2
  store atomic i16 %2, i16* %p release, align 2
  ret void
}

define void @add_16r(i16* %p, i16 %v) {
; X64-LABEL: add_16r:
; X64:       # %bb.0:
; X64-NEXT:    addw %si, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: add_16r:
; X32:       # %bb.0:
; X32-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    addw %ax, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i16, i16* %p acquire, align 2
  %2 = add i16 %1, %v
  store atomic i16 %2, i16* %p release, align 2
  ret void
}

define void @add_32i(i32* %p) {
; X64-LABEL: add_32i:
; X64:       # %bb.0:
; X64-NEXT:    addl $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: add_32i:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    addl $2, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = add i32 %1, 2
  store atomic i32 %2, i32* %p monotonic, align 4
  ret void
}

define void @add_32r(i32* %p, i32 %v) {
; X64-LABEL: add_32r:
; X64:       # %bb.0:
; X64-NEXT:    addl %esi, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: add_32r:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    addl %eax, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = add i32 %1, %v
  store atomic i32 %2, i32* %p monotonic, align 4
  ret void
}

; The following is a corner case where the load is added to itself. The pattern
; matching should not fold this. We only test with 32-bit add, but the same
; applies to other sizes and operations.
define void @add_32r_self(i32* %p) {
; X64-LABEL: add_32r_self:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    addl %eax, %eax
; X64-NEXT:    movl %eax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: add_32r_self:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    addl %ecx, %ecx
; X32-NEXT:    movl %ecx, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = add i32 %1, %1
  store atomic i32 %2, i32* %p monotonic, align 4
  ret void
}

; The following is a corner case where the load's result is returned. The
; optimizer isn't allowed to duplicate the load because it's atomic.
define i32 @add_32r_ret_load(i32* %p, i32 %v) {
; X64-LABEL: add_32r_ret_load:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    addl %eax, %esi
; X64-NEXT:    movl %esi, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: add_32r_ret_load:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    movl (%ecx), %eax
; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
; X32-NEXT:    addl %eax, %edx
; X32-NEXT:    movl %edx, (%ecx)
; X32-NEXT:    retl
; More code here, we just don't want it to load from P.
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = add i32 %1, %v
  store atomic i32 %2, i32* %p monotonic, align 4
  ret i32 %1
}

define void @add_64i(i64* %p) {
; X64-LABEL: add_64i:
; X64:       # %bb.0:
; X64-NEXT:    addq $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: add_64i:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    movl %edx, %ecx
; X32-NEXT:    movl %eax, %ebx
; X32-NEXT:    addl $2, %ebx
; X32-NEXT:    adcl $0, %ecx
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB14_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB14_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;   We do not check X86-32 as it cannot do 'addq'.
  %1 = load atomic i64, i64* %p acquire, align 8
  %2 = add i64 %1, 2
  store atomic i64 %2, i64* %p release, align 8
  ret void
}

define void @add_64r(i64* %p, i64 %v) {
; X64-LABEL: add_64r:
; X64:       # %bb.0:
; X64-NEXT:    addq %rsi, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: add_64r:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    movl %edx, %ecx
; X32-NEXT:    movl %eax, %ebx
; X32-NEXT:    addl {{[0-9]+}}(%esp), %ebx
; X32-NEXT:    adcl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB15_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB15_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;   We do not check X86-32 as it cannot do 'addq'.
  %1 = load atomic i64, i64* %p acquire, align 8
  %2 = add i64 %1, %v
  store atomic i64 %2, i64* %p release, align 8
  ret void
}

define void @add_32i_seq_cst(i32* %p) {
; X64-LABEL: add_32i_seq_cst:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    addl $2, %eax
; X64-NEXT:    xchgl %eax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: add_32i_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    addl $2, %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p monotonic, align 4
  %2 = add i32 %1, 2
  store atomic i32 %2, i32* %p seq_cst, align 4
  ret void
}

define void @add_32r_seq_cst(i32* %p, i32 %v) {
; X64-LABEL: add_32r_seq_cst:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    addl %esi, %eax
; X64-NEXT:    xchgl %eax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: add_32r_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    addl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p monotonic, align 4
  %2 = add i32 %1, %v
  store atomic i32 %2, i32* %p seq_cst, align 4
  ret void
}

; ----- SUB -----

define void @sub_8r(i8* %p, i8 %v) {
; X64-LABEL: sub_8r:
; X64:       # %bb.0:
; X64-NEXT:    subb %sil, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: sub_8r:
; X32:       # %bb.0:
; X32-NEXT:    movb {{[0-9]+}}(%esp), %al
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    subb %al, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i8, i8* %p seq_cst, align 1
  %2 = sub i8 %1, %v
  store atomic i8 %2, i8* %p release, align 1
  ret void
}

define void @sub_16r(i16* %p, i16 %v) {
; X64-LABEL: sub_16r:
; X64:       # %bb.0:
; X64-NEXT:    subw %si, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: sub_16r:
; X32:       # %bb.0:
; X32-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    subw %ax, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i16, i16* %p acquire, align 2
  %2 = sub i16 %1, %v
  store atomic i16 %2, i16* %p release, align 2
  ret void
}

define void @sub_32r(i32* %p, i32 %v) {
; X64-LABEL: sub_32r:
; X64:       # %bb.0:
; X64-NEXT:    subl %esi, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: sub_32r:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    subl %eax, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = sub i32 %1, %v
  store atomic i32 %2, i32* %p monotonic, align 4
  ret void
}

; The following is a corner case where the load is subed to itself. The pattern
; matching should not fold this. We only test with 32-bit sub, but the same
; applies to other sizes and operations.
define void @sub_32r_self(i32* %p) {
; X64-LABEL: sub_32r_self:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    movl $0, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: sub_32r_self:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    movl $0, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = sub i32 %1, %1
  store atomic i32 %2, i32* %p monotonic, align 4
  ret void
}

; The following is a corner case where the load's result is returned. The
; optimizer isn't allowed to duplicate the load because it's atomic.
define i32 @sub_32r_ret_load(i32* %p, i32 %v) {
; X64-LABEL: sub_32r_ret_load:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    movl %eax, %ecx
; X64-NEXT:    subl %esi, %ecx
; X64-NEXT:    movl %ecx, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: sub_32r_ret_load:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    movl (%ecx), %eax
; X32-NEXT:    movl %eax, %edx
; X32-NEXT:    subl {{[0-9]+}}(%esp), %edx
; X32-NEXT:    movl %edx, (%ecx)
; X32-NEXT:    retl
; More code here, we just don't want it to load from P.
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = sub i32 %1, %v
  store atomic i32 %2, i32* %p monotonic, align 4
  ret i32 %1
}

define void @sub_64r(i64* %p, i64 %v) {
; X64-LABEL: sub_64r:
; X64:       # %bb.0:
; X64-NEXT:    subq %rsi, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: sub_64r:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    movl %edx, %ecx
; X32-NEXT:    movl %eax, %ebx
; X32-NEXT:    subl {{[0-9]+}}(%esp), %ebx
; X32-NEXT:    sbbl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB23_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB23_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;   We do not check X86-32 as it cannot do 'subq'.
  %1 = load atomic i64, i64* %p acquire, align 8
  %2 = sub i64 %1, %v
  store atomic i64 %2, i64* %p release, align 8
  ret void
}

define void @sub_32r_seq_cst(i32* %p, i32 %v) {
; X64-LABEL: sub_32r_seq_cst:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    subl %esi, %eax
; X64-NEXT:    xchgl %eax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: sub_32r_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    subl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p monotonic, align 4
  %2 = sub i32 %1, %v
  store atomic i32 %2, i32* %p seq_cst, align 4
  ret void
}

; ----- AND -----

define void @and_8i(i8* %p) {
; X64-LABEL: and_8i:
; X64:       # %bb.0:
; X64-NEXT:    andb $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: and_8i:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    andb $2, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i8, i8* %p monotonic, align 1
  %2 = and i8 %1, 2
  store atomic i8 %2, i8* %p release, align 1
  ret void
}

define void @and_8r(i8* %p, i8 %v) {
; X64-LABEL: and_8r:
; X64:       # %bb.0:
; X64-NEXT:    andb %sil, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: and_8r:
; X32:       # %bb.0:
; X32-NEXT:    movb {{[0-9]+}}(%esp), %al
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    andb %al, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i8, i8* %p monotonic, align 1
  %2 = and i8 %1, %v
  store atomic i8 %2, i8* %p release, align 1
  ret void
}

define void @and_16i(i16* %p) {
; X64-LABEL: and_16i:
; X64:       # %bb.0:
; X64-NEXT:    andw $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: and_16i:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    andw $2, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i16, i16* %p acquire, align 2
  %2 = and i16 %1, 2
  store atomic i16 %2, i16* %p release, align 2
  ret void
}

define void @and_16r(i16* %p, i16 %v) {
; X64-LABEL: and_16r:
; X64:       # %bb.0:
; X64-NEXT:    andw %si, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: and_16r:
; X32:       # %bb.0:
; X32-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    andw %ax, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i16, i16* %p acquire, align 2
  %2 = and i16 %1, %v
  store atomic i16 %2, i16* %p release, align 2
  ret void
}

define void @and_32i(i32* %p) {
; X64-LABEL: and_32i:
; X64:       # %bb.0:
; X64-NEXT:    andl $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: and_32i:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    andl $2, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = and i32 %1, 2
  store atomic i32 %2, i32* %p release, align 4
  ret void
}

define void @and_32r(i32* %p, i32 %v) {
; X64-LABEL: and_32r:
; X64:       # %bb.0:
; X64-NEXT:    andl %esi, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: and_32r:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    andl %eax, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = and i32 %1, %v
  store atomic i32 %2, i32* %p release, align 4
  ret void
}

define void @and_64i(i64* %p) {
; X64-LABEL: and_64i:
; X64:       # %bb.0:
; X64-NEXT:    andq $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: and_64i:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    movl %eax, %ebx
; X32-NEXT:    andl $2, %ebx
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB31_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB31_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;   We do not check X86-32 as it cannot do 'andq'.
  %1 = load atomic i64, i64* %p acquire, align 8
  %2 = and i64 %1, 2
  store atomic i64 %2, i64* %p release, align 8
  ret void
}

define void @and_64r(i64* %p, i64 %v) {
; X64-LABEL: and_64r:
; X64:       # %bb.0:
; X64-NEXT:    andq %rsi, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: and_64r:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    movl %eax, %ebx
; X32-NEXT:    movl %edx, %ecx
; X32-NEXT:    andl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    andl {{[0-9]+}}(%esp), %ebx
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB32_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB32_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;   We do not check X86-32 as it cannot do 'andq'.
  %1 = load atomic i64, i64* %p acquire, align 8
  %2 = and i64 %1, %v
  store atomic i64 %2, i64* %p release, align 8
  ret void
}

define void @and_32i_seq_cst(i32* %p) {
; X64-LABEL: and_32i_seq_cst:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    andl $2, %eax
; X64-NEXT:    xchgl %eax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: and_32i_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    andl $2, %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p monotonic, align 4
  %2 = and i32 %1, 2
  store atomic i32 %2, i32* %p seq_cst, align 4
  ret void
}

define void @and_32r_seq_cst(i32* %p, i32 %v) {
; X64-LABEL: and_32r_seq_cst:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    andl %esi, %eax
; X64-NEXT:    xchgl %eax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: and_32r_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    andl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p monotonic, align 4
  %2 = and i32 %1, %v
  store atomic i32 %2, i32* %p seq_cst, align 4
  ret void
}

; ----- OR -----

define void @or_8i(i8* %p) {
; X64-LABEL: or_8i:
; X64:       # %bb.0:
; X64-NEXT:    orb $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: or_8i:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    orb $2, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i8, i8* %p acquire, align 1
  %2 = or i8 %1, 2
  store atomic i8 %2, i8* %p release, align 1
  ret void
}

define void @or_8r(i8* %p, i8 %v) {
; X64-LABEL: or_8r:
; X64:       # %bb.0:
; X64-NEXT:    orb %sil, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: or_8r:
; X32:       # %bb.0:
; X32-NEXT:    movb {{[0-9]+}}(%esp), %al
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    orb %al, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i8, i8* %p acquire, align 1
  %2 = or i8 %1, %v
  store atomic i8 %2, i8* %p release, align 1
  ret void
}

define void @or_16i(i16* %p) {
; X64-LABEL: or_16i:
; X64:       # %bb.0:
; X64-NEXT:    orw $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: or_16i:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    orw $2, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i16, i16* %p acquire, align 2
  %2 = or i16 %1, 2
  store atomic i16 %2, i16* %p release, align 2
  ret void
}

define void @or_16r(i16* %p, i16 %v) {
; X64-LABEL: or_16r:
; X64:       # %bb.0:
; X64-NEXT:    orw %si, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: or_16r:
; X32:       # %bb.0:
; X32-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    orw %ax, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i16, i16* %p acquire, align 2
  %2 = or i16 %1, %v
  store atomic i16 %2, i16* %p release, align 2
  ret void
}

define void @or_32i(i32* %p) {
; X64-LABEL: or_32i:
; X64:       # %bb.0:
; X64-NEXT:    orl $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: or_32i:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    orl $2, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = or i32 %1, 2
  store atomic i32 %2, i32* %p release, align 4
  ret void
}

define void @or_32r(i32* %p, i32 %v) {
; X64-LABEL: or_32r:
; X64:       # %bb.0:
; X64-NEXT:    orl %esi, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: or_32r:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    orl %eax, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = or i32 %1, %v
  store atomic i32 %2, i32* %p release, align 4
  ret void
}

define void @or_64i(i64* %p) {
; X64-LABEL: or_64i:
; X64:       # %bb.0:
; X64-NEXT:    orq $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: or_64i:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    movl %edx, %ecx
; X32-NEXT:    movl %eax, %ebx
; X32-NEXT:    orl $2, %ebx
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB41_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB41_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;   We do not check X86-32 as it cannot do 'orq'.
  %1 = load atomic i64, i64* %p acquire, align 8
  %2 = or i64 %1, 2
  store atomic i64 %2, i64* %p release, align 8
  ret void
}

define void @or_64r(i64* %p, i64 %v) {
; X64-LABEL: or_64r:
; X64:       # %bb.0:
; X64-NEXT:    orq %rsi, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: or_64r:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    movl %eax, %ebx
; X32-NEXT:    movl %edx, %ecx
; X32-NEXT:    orl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    orl {{[0-9]+}}(%esp), %ebx
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB42_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB42_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;   We do not check X86-32 as it cannot do 'orq'.
  %1 = load atomic i64, i64* %p acquire, align 8
  %2 = or i64 %1, %v
  store atomic i64 %2, i64* %p release, align 8
  ret void
}

define void @or_32i_seq_cst(i32* %p) {
; X64-LABEL: or_32i_seq_cst:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    orl $2, %eax
; X64-NEXT:    xchgl %eax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: or_32i_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    orl $2, %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p monotonic, align 4
  %2 = or i32 %1, 2
  store atomic i32 %2, i32* %p seq_cst, align 4
  ret void
}

define void @or_32r_seq_cst(i32* %p, i32 %v) {
; X64-LABEL: or_32r_seq_cst:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    orl %esi, %eax
; X64-NEXT:    xchgl %eax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: or_32r_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    orl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p monotonic, align 4
  %2 = or i32 %1, %v
  store atomic i32 %2, i32* %p seq_cst, align 4
  ret void
}

; ----- XOR -----

define void @xor_8i(i8* %p) {
; X64-LABEL: xor_8i:
; X64:       # %bb.0:
; X64-NEXT:    xorb $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: xor_8i:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    xorb $2, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i8, i8* %p acquire, align 1
  %2 = xor i8 %1, 2
  store atomic i8 %2, i8* %p release, align 1
  ret void
}

define void @xor_8r(i8* %p, i8 %v) {
; X64-LABEL: xor_8r:
; X64:       # %bb.0:
; X64-NEXT:    xorb %sil, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: xor_8r:
; X32:       # %bb.0:
; X32-NEXT:    movb {{[0-9]+}}(%esp), %al
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    xorb %al, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i8, i8* %p acquire, align 1
  %2 = xor i8 %1, %v
  store atomic i8 %2, i8* %p release, align 1
  ret void
}

define void @xor_16i(i16* %p) {
; X64-LABEL: xor_16i:
; X64:       # %bb.0:
; X64-NEXT:    xorw $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: xor_16i:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    xorw $2, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i16, i16* %p acquire, align 2
  %2 = xor i16 %1, 2
  store atomic i16 %2, i16* %p release, align 2
  ret void
}

define void @xor_16r(i16* %p, i16 %v) {
; X64-LABEL: xor_16r:
; X64:       # %bb.0:
; X64-NEXT:    xorw %si, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: xor_16r:
; X32:       # %bb.0:
; X32-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    xorw %ax, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i16, i16* %p acquire, align 2
  %2 = xor i16 %1, %v
  store atomic i16 %2, i16* %p release, align 2
  ret void
}

define void @xor_32i(i32* %p) {
; X64-LABEL: xor_32i:
; X64:       # %bb.0:
; X64-NEXT:    xorl $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: xor_32i:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    xorl $2, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = xor i32 %1, 2
  store atomic i32 %2, i32* %p release, align 4
  ret void
}

define void @xor_32r(i32* %p, i32 %v) {
; X64-LABEL: xor_32r:
; X64:       # %bb.0:
; X64-NEXT:    xorl %esi, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: xor_32r:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    xorl %eax, (%ecx)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = xor i32 %1, %v
  store atomic i32 %2, i32* %p release, align 4
  ret void
}

define void @xor_64i(i64* %p) {
; X64-LABEL: xor_64i:
; X64:       # %bb.0:
; X64-NEXT:    xorq $2, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: xor_64i:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    movl %edx, %ecx
; X32-NEXT:    movl %eax, %ebx
; X32-NEXT:    xorl $2, %ebx
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB51_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB51_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;   We do not check X86-32 as it cannot do 'xorq'.
  %1 = load atomic i64, i64* %p acquire, align 8
  %2 = xor i64 %1, 2
  store atomic i64 %2, i64* %p release, align 8
  ret void
}

define void @xor_64r(i64* %p, i64 %v) {
; X64-LABEL: xor_64r:
; X64:       # %bb.0:
; X64-NEXT:    xorq %rsi, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: xor_64r:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    movl %eax, %ebx
; X32-NEXT:    movl %edx, %ecx
; X32-NEXT:    xorl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    xorl {{[0-9]+}}(%esp), %ebx
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB52_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB52_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;   We do not check X86-32 as it cannot do 'xorq'.
  %1 = load atomic i64, i64* %p acquire, align 8
  %2 = xor i64 %1, %v
  store atomic i64 %2, i64* %p release, align 8
  ret void
}

define void @xor_32i_seq_cst(i32* %p) {
; X64-LABEL: xor_32i_seq_cst:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    xorl $2, %eax
; X64-NEXT:    xchgl %eax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: xor_32i_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    xorl $2, %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p monotonic, align 4
  %2 = xor i32 %1, 2
  store atomic i32 %2, i32* %p seq_cst, align 4
  ret void
}

define void @xor_32r_seq_cst(i32* %p, i32 %v) {
; X64-LABEL: xor_32r_seq_cst:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    xorl %esi, %eax
; X64-NEXT:    xchgl %eax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: xor_32r_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    xorl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p monotonic, align 4
  %2 = xor i32 %1, %v
  store atomic i32 %2, i32* %p seq_cst, align 4
  ret void
}

; ----- INC -----

define void @inc_8(i8* %p) {
; FAST_INC-LABEL: inc_8:
; FAST_INC:       # %bb.0:
; FAST_INC-NEXT:    incb (%rdi)
; FAST_INC-NEXT:    retq
;
; X32-LABEL: inc_8:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    incb (%eax)
; X32-NEXT:    retl
;
; SLOW_INC-LABEL: inc_8:
; SLOW_INC:       # %bb.0:
; SLOW_INC-NEXT:    addb $1, (%rdi)
; SLOW_INC-NEXT:    retq
  %1 = load atomic i8, i8* %p seq_cst, align 1
  %2 = add i8 %1, 1
  store atomic i8 %2, i8* %p release, align 1
  ret void
}

define void @inc_16(i16* %p) {
; FAST_INC-LABEL: inc_16:
; FAST_INC:       # %bb.0:
; FAST_INC-NEXT:    incw (%rdi)
; FAST_INC-NEXT:    retq
;
; X32-LABEL: inc_16:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    incw (%eax)
; X32-NEXT:    retl
;
; SLOW_INC-LABEL: inc_16:
; SLOW_INC:       # %bb.0:
; SLOW_INC-NEXT:    addw $1, (%rdi)
; SLOW_INC-NEXT:    retq
  %1 = load atomic i16, i16* %p acquire, align 2
  %2 = add i16 %1, 1
  store atomic i16 %2, i16* %p release, align 2
  ret void
}

define void @inc_32(i32* %p) {
; FAST_INC-LABEL: inc_32:
; FAST_INC:       # %bb.0:
; FAST_INC-NEXT:    incl (%rdi)
; FAST_INC-NEXT:    retq
;
; X32-LABEL: inc_32:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    incl (%eax)
; X32-NEXT:    retl
;
; SLOW_INC-LABEL: inc_32:
; SLOW_INC:       # %bb.0:
; SLOW_INC-NEXT:    addl $1, (%rdi)
; SLOW_INC-NEXT:    retq
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = add i32 %1, 1
  store atomic i32 %2, i32* %p monotonic, align 4
  ret void
}

define void @inc_64(i64* %p) {
; FAST_INC-LABEL: inc_64:
; FAST_INC:       # %bb.0:
; FAST_INC-NEXT:    incq (%rdi)
; FAST_INC-NEXT:    retq
;
; X32-LABEL: inc_64:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    movl %edx, %ecx
; X32-NEXT:    movl %eax, %ebx
; X32-NEXT:    addl $1, %ebx
; X32-NEXT:    adcl $0, %ecx
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB58_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB58_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;
; SLOW_INC-LABEL: inc_64:
; SLOW_INC:       # %bb.0:
; SLOW_INC-NEXT:    addq $1, (%rdi)
; SLOW_INC-NEXT:    retq
;   We do not check X86-32 as it cannot do 'incq'.
  %1 = load atomic i64, i64* %p acquire, align 8
  %2 = add i64 %1, 1
  store atomic i64 %2, i64* %p release, align 8
  ret void
}

define void @inc_32_seq_cst(i32* %p) {
; FAST_INC-LABEL: inc_32_seq_cst:
; FAST_INC:       # %bb.0:
; FAST_INC-NEXT:    movl (%rdi), %eax
; FAST_INC-NEXT:    incl %eax
; FAST_INC-NEXT:    xchgl %eax, (%rdi)
; FAST_INC-NEXT:    retq
;
; X32-LABEL: inc_32_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    incl %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
;
; SLOW_INC-LABEL: inc_32_seq_cst:
; SLOW_INC:       # %bb.0:
; SLOW_INC-NEXT:    movl (%rdi), %eax
; SLOW_INC-NEXT:    addl $1, %eax
; SLOW_INC-NEXT:    xchgl %eax, (%rdi)
; SLOW_INC-NEXT:    retq
  %1 = load atomic i32, i32* %p monotonic, align 4
  %2 = add i32 %1, 1
  store atomic i32 %2, i32* %p seq_cst, align 4
  ret void
}

; ----- DEC -----

define void @dec_8(i8* %p) {
; FAST_INC-LABEL: dec_8:
; FAST_INC:       # %bb.0:
; FAST_INC-NEXT:    decb (%rdi)
; FAST_INC-NEXT:    retq
;
; X32-LABEL: dec_8:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    decb (%eax)
; X32-NEXT:    retl
;
; SLOW_INC-LABEL: dec_8:
; SLOW_INC:       # %bb.0:
; SLOW_INC-NEXT:    addb $-1, (%rdi)
; SLOW_INC-NEXT:    retq
  %1 = load atomic i8, i8* %p seq_cst, align 1
  %2 = sub i8 %1, 1
  store atomic i8 %2, i8* %p release, align 1
  ret void
}

define void @dec_16(i16* %p) {
; FAST_INC-LABEL: dec_16:
; FAST_INC:       # %bb.0:
; FAST_INC-NEXT:    decw (%rdi)
; FAST_INC-NEXT:    retq
;
; X32-LABEL: dec_16:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    decw (%eax)
; X32-NEXT:    retl
;
; SLOW_INC-LABEL: dec_16:
; SLOW_INC:       # %bb.0:
; SLOW_INC-NEXT:    addw $-1, (%rdi)
; SLOW_INC-NEXT:    retq
  %1 = load atomic i16, i16* %p acquire, align 2
  %2 = sub i16 %1, 1
  store atomic i16 %2, i16* %p release, align 2
  ret void
}

define void @dec_32(i32* %p) {
; FAST_INC-LABEL: dec_32:
; FAST_INC:       # %bb.0:
; FAST_INC-NEXT:    decl (%rdi)
; FAST_INC-NEXT:    retq
;
; X32-LABEL: dec_32:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    decl (%eax)
; X32-NEXT:    retl
;
; SLOW_INC-LABEL: dec_32:
; SLOW_INC:       # %bb.0:
; SLOW_INC-NEXT:    addl $-1, (%rdi)
; SLOW_INC-NEXT:    retq
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = sub i32 %1, 1
  store atomic i32 %2, i32* %p monotonic, align 4
  ret void
}

define void @dec_64(i64* %p) {
; FAST_INC-LABEL: dec_64:
; FAST_INC:       # %bb.0:
; FAST_INC-NEXT:    decq (%rdi)
; FAST_INC-NEXT:    retq
;
; X32-LABEL: dec_64:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    movl %edx, %ecx
; X32-NEXT:    movl %eax, %ebx
; X32-NEXT:    addl $-1, %ebx
; X32-NEXT:    adcl $-1, %ecx
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB63_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB63_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;
; SLOW_INC-LABEL: dec_64:
; SLOW_INC:       # %bb.0:
; SLOW_INC-NEXT:    addq $-1, (%rdi)
; SLOW_INC-NEXT:    retq
;   We do not check X86-32 as it cannot do 'decq'.
  %1 = load atomic i64, i64* %p acquire, align 8
  %2 = sub i64 %1, 1
  store atomic i64 %2, i64* %p release, align 8
  ret void
}

define void @dec_32_seq_cst(i32* %p) {
; FAST_INC-LABEL: dec_32_seq_cst:
; FAST_INC:       # %bb.0:
; FAST_INC-NEXT:    movl (%rdi), %eax
; FAST_INC-NEXT:    decl %eax
; FAST_INC-NEXT:    xchgl %eax, (%rdi)
; FAST_INC-NEXT:    retq
;
; X32-LABEL: dec_32_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    decl %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
;
; SLOW_INC-LABEL: dec_32_seq_cst:
; SLOW_INC:       # %bb.0:
; SLOW_INC-NEXT:    movl (%rdi), %eax
; SLOW_INC-NEXT:    addl $-1, %eax
; SLOW_INC-NEXT:    xchgl %eax, (%rdi)
; SLOW_INC-NEXT:    retq
  %1 = load atomic i32, i32* %p monotonic, align 4
  %2 = sub i32 %1, 1
  store atomic i32 %2, i32* %p seq_cst, align 4
  ret void
}

; ----- NOT -----

define void @not_8(i8* %p) {
; X64-LABEL: not_8:
; X64:       # %bb.0:
; X64-NEXT:    notb (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: not_8:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    notb (%eax)
; X32-NEXT:    retl
  %1 = load atomic i8, i8* %p seq_cst, align 1
  %2 = xor i8 %1, -1
  store atomic i8 %2, i8* %p release, align 1
  ret void
}

define void @not_16(i16* %p) {
; X64-LABEL: not_16:
; X64:       # %bb.0:
; X64-NEXT:    notw (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: not_16:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    notw (%eax)
; X32-NEXT:    retl
  %1 = load atomic i16, i16* %p acquire, align 2
  %2 = xor i16 %1, -1
  store atomic i16 %2, i16* %p release, align 2
  ret void
}

define void @not_32(i32* %p) {
; X64-LABEL: not_32:
; X64:       # %bb.0:
; X64-NEXT:    notl (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: not_32:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    notl (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = xor i32 %1, -1
  store atomic i32 %2, i32* %p monotonic, align 4
  ret void
}

define void @not_64(i64* %p) {
; X64-LABEL: not_64:
; X64:       # %bb.0:
; X64-NEXT:    notq (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: not_64:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    .cfi_offset %esi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    movl %eax, %ebx
; X32-NEXT:    movl %edx, %ecx
; X32-NEXT:    notl %ecx
; X32-NEXT:    notl %ebx
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB68_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB68_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;   We do not check X86-32 as it cannot do 'notq'.
  %1 = load atomic i64, i64* %p acquire, align 8
  %2 = xor i64 %1, -1
  store atomic i64 %2, i64* %p release, align 8
  ret void
}

define void @not_32_seq_cst(i32* %p) {
; X64-LABEL: not_32_seq_cst:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    notl %eax
; X64-NEXT:    xchgl %eax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: not_32_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    notl %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p monotonic, align 4
  %2 = xor i32 %1, -1
  store atomic i32 %2, i32* %p seq_cst, align 4
  ret void
}

; ----- NEG -----

define void @neg_8(i8* %p) {
; X64-LABEL: neg_8:
; X64:       # %bb.0:
; X64-NEXT:    negb (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: neg_8:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    negb (%eax)
; X32-NEXT:    retl
  %1 = load atomic i8, i8* %p seq_cst, align 1
  %2 = sub i8 0, %1
  store atomic i8 %2, i8* %p release, align 1
  ret void
}

define void @neg_16(i16* %p) {
; X64-LABEL: neg_16:
; X64:       # %bb.0:
; X64-NEXT:    movzwl (%rdi), %eax
; X64-NEXT:    negl %eax
; X64-NEXT:    movw %ax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: neg_16:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movzwl (%eax), %ecx
; X32-NEXT:    negl %ecx
; X32-NEXT:    movw %cx, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i16, i16* %p acquire, align 2
  %2 = sub i16 0, %1
  store atomic i16 %2, i16* %p release, align 2
  ret void
}

define void @neg_32(i32* %p) {
; X64-LABEL: neg_32:
; X64:       # %bb.0:
; X64-NEXT:    negl (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: neg_32:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    negl (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p acquire, align 4
  %2 = sub i32 0, %1
  store atomic i32 %2, i32* %p monotonic, align 4
  ret void
}

define void @neg_64(i64* %p) {
; X64-LABEL: neg_64:
; X64:       # %bb.0:
; X64-NEXT:    negq (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: neg_64:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    pushl %edi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    pushl %esi
; X32-NEXT:    .cfi_def_cfa_offset 16
; X32-NEXT:    .cfi_offset %esi, -16
; X32-NEXT:    .cfi_offset %edi, -12
; X32-NEXT:    .cfi_offset %ebx, -8
; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi
; X32-NEXT:    xorl %esi, %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%edi)
; X32-NEXT:    movl %eax, %ebx
; X32-NEXT:    negl %ebx
; X32-NEXT:    sbbl %edx, %esi
; X32-NEXT:    movl (%edi), %eax
; X32-NEXT:    movl 4(%edi), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB73_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    movl %esi, %ecx
; X32-NEXT:    lock cmpxchg8b (%edi)
; X32-NEXT:    jne .LBB73_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    popl %esi
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    popl %edi
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    popl %ebx
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
;   We do neg check X86-32 as it canneg do 'negq'.
  %1 = load atomic i64, i64* %p acquire, align 8
  %2 = sub i64 0, %1
  store atomic i64 %2, i64* %p release, align 8
  ret void
}

define void @neg_32_seq_cst(i32* %p) {
; X64-LABEL: neg_32_seq_cst:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    negl %eax
; X64-NEXT:    xchgl %eax, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: neg_32_seq_cst:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    negl %ecx
; X32-NEXT:    xchgl %ecx, (%eax)
; X32-NEXT:    retl
  %1 = load atomic i32, i32* %p monotonic, align 4
  %2 = sub i32 0, %1
  store atomic i32 %2, i32* %p seq_cst, align 4
  ret void
}

; ----- FADD -----

define void @fadd_32r(float* %loc, float %val) {
; X64-LABEL: fadd_32r:
; X64:       # %bb.0:
; X64-NEXT:    addss (%rdi), %xmm0
; X64-NEXT:    movss %xmm0, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: fadd_32r:
; X32:       # %bb.0:
; X32-NEXT:    subl $8, %esp
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl (%eax), %ecx
; X32-NEXT:    movl %ecx, (%esp)
; X32-NEXT:    flds (%esp)
; X32-NEXT:    fadds {{[0-9]+}}(%esp)
; X32-NEXT:    fstps {{[0-9]+}}(%esp)
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    movl %ecx, (%eax)
; X32-NEXT:    addl $8, %esp
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
; Don't check x86-32.
; LLVM's SSE handling is conservative on x86-32 even without using atomics.
  %floc = bitcast float* %loc to i32*
  %1 = load atomic i32, i32* %floc seq_cst, align 4
  %2 = bitcast i32 %1 to float
  %add = fadd float %2, %val
  %3 = bitcast float %add to i32
  store atomic i32 %3, i32* %floc release, align 4
  ret void
}

define void @fadd_64r(double* %loc, double %val) {
; X64-LABEL: fadd_64r:
; X64:       # %bb.0:
; X64-NEXT:    addsd (%rdi), %xmm0
; X64-NEXT:    movsd %xmm0, (%rdi)
; X64-NEXT:    retq
;
; X32-LABEL: fadd_64r:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebp
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    .cfi_offset %ebp, -8
; X32-NEXT:    movl %esp, %ebp
; X32-NEXT:    .cfi_def_cfa_register %ebp
; X32-NEXT:    pushl %ebx
; X32-NEXT:    pushl %esi
; X32-NEXT:    andl $-8, %esp
; X32-NEXT:    subl $16, %esp
; X32-NEXT:    .cfi_offset %esi, -16
; X32-NEXT:    .cfi_offset %ebx, -12
; X32-NEXT:    movl 8(%ebp), %esi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp)
; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp)
; X32-NEXT:    fldl {{[0-9]+}}(%esp)
; X32-NEXT:    faddl 12(%ebp)
; X32-NEXT:    fstpl (%esp)
; X32-NEXT:    movl (%esp), %ebx
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    movl (%esi), %eax
; X32-NEXT:    movl 4(%esi), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB76_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esi)
; X32-NEXT:    jne .LBB76_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    leal -8(%ebp), %esp
; X32-NEXT:    popl %esi
; X32-NEXT:    popl %ebx
; X32-NEXT:    popl %ebp
; X32-NEXT:    .cfi_def_cfa %esp, 4
; X32-NEXT:    retl
; Don't check x86-32 (see comment above).
  %floc = bitcast double* %loc to i64*
  %1 = load atomic i64, i64* %floc seq_cst, align 8
  %2 = bitcast i64 %1 to double
  %add = fadd double %2, %val
  %3 = bitcast double %add to i64
  store atomic i64 %3, i64* %floc release, align 8
  ret void
}

@glob32 = global float 0.000000e+00, align 4
@glob64 = global double 0.000000e+00, align 8

; Floating-point add to a global using an immediate.
define void @fadd_32g() {
; X64-LABEL: fadd_32g:
; X64:       # %bb.0:
; X64-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X64-NEXT:    addss {{.*}}(%rip), %xmm0
; X64-NEXT:    movss %xmm0, {{.*}}(%rip)
; X64-NEXT:    retq
;
; X32-LABEL: fadd_32g:
; X32:       # %bb.0:
; X32-NEXT:    subl $8, %esp
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    movl glob32, %eax
; X32-NEXT:    movl %eax, (%esp)
; X32-NEXT:    fld1
; X32-NEXT:    fadds (%esp)
; X32-NEXT:    fstps {{[0-9]+}}(%esp)
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl %eax, glob32
; X32-NEXT:    addl $8, %esp
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
; Don't check x86-32 (see comment above).
  %i = load atomic i32, i32* bitcast (float* @glob32 to i32*) monotonic, align 4
  %f = bitcast i32 %i to float
  %add = fadd float %f, 1.000000e+00
  %s = bitcast float %add to i32
  store atomic i32 %s, i32* bitcast (float* @glob32 to i32*) monotonic, align 4
  ret void
}

define void @fadd_64g() {
; X64-LABEL: fadd_64g:
; X64:       # %bb.0:
; X64-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
; X64-NEXT:    addsd {{.*}}(%rip), %xmm0
; X64-NEXT:    movsd %xmm0, {{.*}}(%rip)
; X64-NEXT:    retq
;
; X32-LABEL: fadd_64g:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebp
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    .cfi_offset %ebp, -8
; X32-NEXT:    movl %esp, %ebp
; X32-NEXT:    .cfi_def_cfa_register %ebp
; X32-NEXT:    pushl %ebx
; X32-NEXT:    andl $-8, %esp
; X32-NEXT:    subl $24, %esp
; X32-NEXT:    .cfi_offset %ebx, -12
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b glob64
; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp)
; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp)
; X32-NEXT:    fld1
; X32-NEXT:    faddl {{[0-9]+}}(%esp)
; X32-NEXT:    fstpl (%esp)
; X32-NEXT:    movl (%esp), %ebx
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    movl glob64+4, %edx
; X32-NEXT:    movl glob64, %eax
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB78_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b glob64
; X32-NEXT:    jne .LBB78_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    leal -4(%ebp), %esp
; X32-NEXT:    popl %ebx
; X32-NEXT:    popl %ebp
; X32-NEXT:    .cfi_def_cfa %esp, 4
; X32-NEXT:    retl
; Don't check x86-32 (see comment above).
  %i = load atomic i64, i64* bitcast (double* @glob64 to i64*) monotonic, align 8
  %f = bitcast i64 %i to double
  %add = fadd double %f, 1.000000e+00
  %s = bitcast double %add to i64
  store atomic i64 %s, i64* bitcast (double* @glob64 to i64*) monotonic, align 8
  ret void
}

; Floating-point add to a hard-coded immediate location using an immediate.
define void @fadd_32imm() {
; X64-LABEL: fadd_32imm:
; X64:       # %bb.0:
; X64-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X64-NEXT:    movl $3735928559, %eax # imm = 0xDEADBEEF
; X64-NEXT:    addss (%rax), %xmm0
; X64-NEXT:    movss %xmm0, (%rax)
; X64-NEXT:    retq
;
; X32-LABEL: fadd_32imm:
; X32:       # %bb.0:
; X32-NEXT:    subl $8, %esp
; X32-NEXT:    .cfi_def_cfa_offset 12
; X32-NEXT:    movl -559038737, %eax
; X32-NEXT:    movl %eax, (%esp)
; X32-NEXT:    fld1
; X32-NEXT:    fadds (%esp)
; X32-NEXT:    fstps {{[0-9]+}}(%esp)
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl %eax, -559038737
; X32-NEXT:    addl $8, %esp
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
; Don't check x86-32 (see comment above).
  %i = load atomic i32, i32* inttoptr (i32 3735928559 to i32*) monotonic, align 4
  %f = bitcast i32 %i to float
  %add = fadd float %f, 1.000000e+00
  %s = bitcast float %add to i32
  store atomic i32 %s, i32* inttoptr (i32 3735928559 to i32*) monotonic, align 4
  ret void
}

define void @fadd_64imm() {
; X64-LABEL: fadd_64imm:
; X64:       # %bb.0:
; X64-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
; X64-NEXT:    movl $3735928559, %eax # imm = 0xDEADBEEF
; X64-NEXT:    addsd (%rax), %xmm0
; X64-NEXT:    movsd %xmm0, (%rax)
; X64-NEXT:    retq
;
; X32-LABEL: fadd_64imm:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebp
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    .cfi_offset %ebp, -8
; X32-NEXT:    movl %esp, %ebp
; X32-NEXT:    .cfi_def_cfa_register %ebp
; X32-NEXT:    pushl %ebx
; X32-NEXT:    andl $-8, %esp
; X32-NEXT:    subl $24, %esp
; X32-NEXT:    .cfi_offset %ebx, -12
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b -559038737
; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp)
; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp)
; X32-NEXT:    fld1
; X32-NEXT:    faddl {{[0-9]+}}(%esp)
; X32-NEXT:    fstpl (%esp)
; X32-NEXT:    movl (%esp), %ebx
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    movl -559038737, %eax
; X32-NEXT:    movl -559038733, %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB80_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b -559038737
; X32-NEXT:    jne .LBB80_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    leal -4(%ebp), %esp
; X32-NEXT:    popl %ebx
; X32-NEXT:    popl %ebp
; X32-NEXT:    .cfi_def_cfa %esp, 4
; X32-NEXT:    retl
; Don't check x86-32 (see comment above).
  %i = load atomic i64, i64* inttoptr (i64 3735928559 to i64*) monotonic, align 8
  %f = bitcast i64 %i to double
  %add = fadd double %f, 1.000000e+00
  %s = bitcast double %add to i64
  store atomic i64 %s, i64* inttoptr (i64 3735928559 to i64*) monotonic, align 8
  ret void
}

; Floating-point add to a stack location.
define void @fadd_32stack() {
; X64-LABEL: fadd_32stack:
; X64:       # %bb.0:
; X64-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X64-NEXT:    addss -{{[0-9]+}}(%rsp), %xmm0
; X64-NEXT:    movss %xmm0, -{{[0-9]+}}(%rsp)
; X64-NEXT:    retq
;
; X32-LABEL: fadd_32stack:
; X32:       # %bb.0:
; X32-NEXT:    subl $12, %esp
; X32-NEXT:    .cfi_def_cfa_offset 16
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl %eax, (%esp)
; X32-NEXT:    fld1
; X32-NEXT:    fadds (%esp)
; X32-NEXT:    fstps {{[0-9]+}}(%esp)
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp)
; X32-NEXT:    addl $12, %esp
; X32-NEXT:    .cfi_def_cfa_offset 4
; X32-NEXT:    retl
; Don't check x86-32 (see comment above).
  %ptr = alloca i32, align 4
  %bc3 = bitcast i32* %ptr to float*
  %load = load atomic i32, i32* %ptr acquire, align 4
  %bc0 = bitcast i32 %load to float
  %fadd = fadd float 1.000000e+00, %bc0
  %bc1 = bitcast float %fadd to i32
  store atomic i32 %bc1, i32* %ptr release, align 4
  ret void
}

define void @fadd_64stack() {
; X64-LABEL: fadd_64stack:
; X64:       # %bb.0:
; X64-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
; X64-NEXT:    addsd -{{[0-9]+}}(%rsp), %xmm0
; X64-NEXT:    movsd %xmm0, -{{[0-9]+}}(%rsp)
; X64-NEXT:    retq
;
; X32-LABEL: fadd_64stack:
; X32:       # %bb.0:
; X32-NEXT:    pushl %ebp
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    .cfi_offset %ebp, -8
; X32-NEXT:    movl %esp, %ebp
; X32-NEXT:    .cfi_def_cfa_register %ebp
; X32-NEXT:    pushl %ebx
; X32-NEXT:    andl $-8, %esp
; X32-NEXT:    subl $32, %esp
; X32-NEXT:    .cfi_offset %ebx, -12
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%esp)
; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp)
; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp)
; X32-NEXT:    fld1
; X32-NEXT:    faddl {{[0-9]+}}(%esp)
; X32-NEXT:    fstpl {{[0-9]+}}(%esp)
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    movl (%esp), %eax
; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB82_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%esp)
; X32-NEXT:    jne .LBB82_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    leal -4(%ebp), %esp
; X32-NEXT:    popl %ebx
; X32-NEXT:    popl %ebp
; X32-NEXT:    .cfi_def_cfa %esp, 4
; X32-NEXT:    retl
; Don't check x86-32 (see comment above).
  %ptr = alloca i64, align 8
  %bc3 = bitcast i64* %ptr to double*
  %load = load atomic i64, i64* %ptr acquire, align 8
  %bc0 = bitcast i64 %load to double
  %fadd = fadd double 1.000000e+00, %bc0
  %bc1 = bitcast double %fadd to i64
  store atomic i64 %bc1, i64* %ptr release, align 8
  ret void
}

define void @fadd_array(i64* %arg, double %arg1, i64 %arg2) {
; X64-LABEL: fadd_array:
; X64:       # %bb.0: # %bb
; X64-NEXT:    addsd (%rdi,%rsi,8), %xmm0
; X64-NEXT:    movsd %xmm0, (%rdi,%rsi,8)
; X64-NEXT:    retq
;
; X32-LABEL: fadd_array:
; X32:       # %bb.0: # %bb
; X32-NEXT:    pushl %ebp
; X32-NEXT:    .cfi_def_cfa_offset 8
; X32-NEXT:    .cfi_offset %ebp, -8
; X32-NEXT:    movl %esp, %ebp
; X32-NEXT:    .cfi_def_cfa_register %ebp
; X32-NEXT:    pushl %ebx
; X32-NEXT:    pushl %edi
; X32-NEXT:    pushl %esi
; X32-NEXT:    andl $-8, %esp
; X32-NEXT:    subl $24, %esp
; X32-NEXT:    .cfi_offset %esi, -20
; X32-NEXT:    .cfi_offset %edi, -16
; X32-NEXT:    .cfi_offset %ebx, -12
; X32-NEXT:    movl 20(%ebp), %esi
; X32-NEXT:    movl 8(%ebp), %edi
; X32-NEXT:    xorl %eax, %eax
; X32-NEXT:    xorl %edx, %edx
; X32-NEXT:    xorl %ecx, %ecx
; X32-NEXT:    xorl %ebx, %ebx
; X32-NEXT:    lock cmpxchg8b (%edi,%esi,8)
; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp)
; X32-NEXT:    movl %eax, {{[0-9]+}}(%esp)
; X32-NEXT:    fldl {{[0-9]+}}(%esp)
; X32-NEXT:    faddl 12(%ebp)
; X32-NEXT:    fstpl (%esp)
; X32-NEXT:    movl (%esp), %ebx
; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X32-NEXT:    movl (%edi,%esi,8), %eax
; X32-NEXT:    movl 4(%edi,%esi,8), %edx
; X32-NEXT:    .p2align 4, 0x90
; X32-NEXT:  .LBB83_1: # %atomicrmw.start
; X32-NEXT:    # =>This Inner Loop Header: Depth=1
; X32-NEXT:    lock cmpxchg8b (%edi,%esi,8)
; X32-NEXT:    jne .LBB83_1
; X32-NEXT:  # %bb.2: # %atomicrmw.end
; X32-NEXT:    leal -12(%ebp), %esp
; X32-NEXT:    popl %esi
; X32-NEXT:    popl %edi
; X32-NEXT:    popl %ebx
; X32-NEXT:    popl %ebp
; X32-NEXT:    .cfi_def_cfa %esp, 4
; X32-NEXT:    retl
; Don't check x86-32 (see comment above).
bb:
  %tmp4 = getelementptr inbounds i64, i64* %arg, i64 %arg2
  %tmp6 = load atomic i64, i64* %tmp4 monotonic, align 8
  %tmp7 = bitcast i64 %tmp6 to double
  %tmp8 = fadd double %tmp7, %arg1
  %tmp9 = bitcast double %tmp8 to i64
  store atomic i64 %tmp9, i64* %tmp4 monotonic, align 8
  ret void
}