llvm/llvm/test/CodeGen/X86/bmi.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+cmov,+bmi | FileCheck %s --check-prefixes=X86,X86-SLOW-BEXTR
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+cmov,+bmi,+bmi2 | FileCheck %s --check-prefixes=X86,X86-SLOW-BEXTR
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+bmi | FileCheck %s --check-prefixes=X64,X64-SLOW-BEXTR
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+bmi,+bmi2 | FileCheck %s --check-prefixes=X64,X64-SLOW-BEXTR
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+cmov,+bmi,+fast-bextr | FileCheck %s --check-prefixes=X86,X86-FAST-BEXTR
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+bmi,+fast-bextr | FileCheck %s --check-prefixes=X64,X64-FAST-BEXTR
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+bmi,+fast-bextr,+egpr --show-mc-encoding | FileCheck %s --check-prefix=EGPR

define i32 @andn32(i32 %x, i32 %y)   {
; X86-LABEL: andn32:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    andnl {{[0-9]+}}(%esp), %eax, %eax
; X86-NEXT:    retl
;
; X64-LABEL: andn32:
; X64:       # %bb.0:
; X64-NEXT:    andnl %esi, %edi, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: andn32:
; EGPR:       # %bb.0:
; EGPR-NEXT:    andnl %esi, %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x40,0xf2,0xc6]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp1 = xor i32 %x, -1
  %tmp2 = and i32 %y, %tmp1
  ret i32 %tmp2
}

define i32 @andn32_load(i32 %x, ptr %y)   {
; X86-LABEL: andn32_load:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    andnl (%eax), %ecx, %eax
; X86-NEXT:    retl
;
; X64-LABEL: andn32_load:
; X64:       # %bb.0:
; X64-NEXT:    andnl (%rsi), %edi, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: andn32_load:
; EGPR:       # %bb.0:
; EGPR-NEXT:    andnl (%rsi), %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x40,0xf2,0x06]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %y1 = load i32, ptr %y
  %tmp1 = xor i32 %x, -1
  %tmp2 = and i32 %y1, %tmp1
  ret i32 %tmp2
}

define i64 @andn64(i64 %x, i64 %y)   {
; X86-LABEL: andn64:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    andnl {{[0-9]+}}(%esp), %eax, %eax
; X86-NEXT:    andnl {{[0-9]+}}(%esp), %ecx, %edx
; X86-NEXT:    retl
;
; X64-LABEL: andn64:
; X64:       # %bb.0:
; X64-NEXT:    andnq %rsi, %rdi, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: andn64:
; EGPR:       # %bb.0:
; EGPR-NEXT:    andnq %rsi, %rdi, %rax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xc0,0xf2,0xc6]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp1 = xor i64 %x, -1
  %tmp2 = and i64 %tmp1, %y
  ret i64 %tmp2
}

; Don't choose a 'test' if an 'andn' can be used.
define i1 @andn_cmp(i32 %x, i32 %y) {
; X86-LABEL: andn_cmp:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    andnl {{[0-9]+}}(%esp), %eax, %eax
; X86-NEXT:    sete %al
; X86-NEXT:    retl
;
; X64-LABEL: andn_cmp:
; X64:       # %bb.0:
; X64-NEXT:    andnl %esi, %edi, %eax
; X64-NEXT:    sete %al
; X64-NEXT:    retq
;
; EGPR-LABEL: andn_cmp:
; EGPR:       # %bb.0:
; EGPR-NEXT:    andnl %esi, %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x40,0xf2,0xc6]
; EGPR-NEXT:    testl %eax, %eax # encoding: [0x85,0xc0]
; EGPR-NEXT:    sete %al # encoding: [0x0f,0x94,0xc0]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %notx = xor i32 %x, -1
  %and = and i32 %notx, %y
  %cmp = icmp eq i32 %and, 0
  ret i1 %cmp
}

; Recognize a disguised andn in the following 4 tests.
define i1 @and_cmp1(i32 %x, i32 %y) {
; X86-LABEL: and_cmp1:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    andnl {{[0-9]+}}(%esp), %eax, %eax
; X86-NEXT:    sete %al
; X86-NEXT:    retl
;
; X64-LABEL: and_cmp1:
; X64:       # %bb.0:
; X64-NEXT:    andnl %esi, %edi, %eax
; X64-NEXT:    sete %al
; X64-NEXT:    retq
;
; EGPR-LABEL: and_cmp1:
; EGPR:       # %bb.0:
; EGPR-NEXT:    andnl %esi, %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x40,0xf2,0xc6]
; EGPR-NEXT:    testl %eax, %eax # encoding: [0x85,0xc0]
; EGPR-NEXT:    sete %al # encoding: [0x0f,0x94,0xc0]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %and = and i32 %x, %y
  %cmp = icmp eq i32 %and, %y
  ret i1 %cmp
}

define i1 @and_cmp2(i32 %x, i32 %y) {
; X86-LABEL: and_cmp2:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    andnl {{[0-9]+}}(%esp), %eax, %eax
; X86-NEXT:    setne %al
; X86-NEXT:    retl
;
; X64-LABEL: and_cmp2:
; X64:       # %bb.0:
; X64-NEXT:    andnl %esi, %edi, %eax
; X64-NEXT:    setne %al
; X64-NEXT:    retq
;
; EGPR-LABEL: and_cmp2:
; EGPR:       # %bb.0:
; EGPR-NEXT:    andnl %esi, %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x40,0xf2,0xc6]
; EGPR-NEXT:    testl %eax, %eax # encoding: [0x85,0xc0]
; EGPR-NEXT:    setne %al # encoding: [0x0f,0x95,0xc0]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %and = and i32 %y, %x
  %cmp = icmp ne i32 %and, %y
  ret i1 %cmp
}

define i1 @and_cmp3(i32 %x, i32 %y) {
; X86-LABEL: and_cmp3:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    andnl {{[0-9]+}}(%esp), %eax, %eax
; X86-NEXT:    sete %al
; X86-NEXT:    retl
;
; X64-LABEL: and_cmp3:
; X64:       # %bb.0:
; X64-NEXT:    andnl %esi, %edi, %eax
; X64-NEXT:    sete %al
; X64-NEXT:    retq
;
; EGPR-LABEL: and_cmp3:
; EGPR:       # %bb.0:
; EGPR-NEXT:    andnl %esi, %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x40,0xf2,0xc6]
; EGPR-NEXT:    testl %eax, %eax # encoding: [0x85,0xc0]
; EGPR-NEXT:    sete %al # encoding: [0x0f,0x94,0xc0]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %and = and i32 %x, %y
  %cmp = icmp eq i32 %y, %and
  ret i1 %cmp
}

define i1 @and_cmp4(i32 %x, i32 %y) {
; X86-LABEL: and_cmp4:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    andnl {{[0-9]+}}(%esp), %eax, %eax
; X86-NEXT:    setne %al
; X86-NEXT:    retl
;
; X64-LABEL: and_cmp4:
; X64:       # %bb.0:
; X64-NEXT:    andnl %esi, %edi, %eax
; X64-NEXT:    setne %al
; X64-NEXT:    retq
;
; EGPR-LABEL: and_cmp4:
; EGPR:       # %bb.0:
; EGPR-NEXT:    andnl %esi, %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x40,0xf2,0xc6]
; EGPR-NEXT:    testl %eax, %eax # encoding: [0x85,0xc0]
; EGPR-NEXT:    setne %al # encoding: [0x0f,0x95,0xc0]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %and = and i32 %y, %x
  %cmp = icmp ne i32 %y, %and
  ret i1 %cmp
}

; A mask and compare against constant is ok for an 'andn' too
; even though the BMI instruction doesn't have an immediate form.
define i1 @and_cmp_const(i32 %x) {
; X86-LABEL: and_cmp_const:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    notl %eax
; X86-NEXT:    testb $43, %al
; X86-NEXT:    sete %al
; X86-NEXT:    retl
;
; X64-LABEL: and_cmp_const:
; X64:       # %bb.0:
; X64-NEXT:    notl %edi
; X64-NEXT:    testb $43, %dil
; X64-NEXT:    sete %al
; X64-NEXT:    retq
;
; EGPR-LABEL: and_cmp_const:
; EGPR:       # %bb.0:
; EGPR-NEXT:    notl %edi # encoding: [0xf7,0xd7]
; EGPR-NEXT:    testb $43, %dil # encoding: [0x40,0xf6,0xc7,0x2b]
; EGPR-NEXT:    sete %al # encoding: [0x0f,0x94,0xc0]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %and = and i32 %x, 43
  %cmp = icmp eq i32 %and, 43
  ret i1 %cmp
}

; But don't use 'andn' if the mask is a power-of-two.
define i1 @and_cmp_const_power_of_two(i32 %x, i32 %y) {
; X86-LABEL: and_cmp_const_power_of_two:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    btl %ecx, %eax
; X86-NEXT:    setae %al
; X86-NEXT:    retl
;
; X64-LABEL: and_cmp_const_power_of_two:
; X64:       # %bb.0:
; X64-NEXT:    btl %esi, %edi
; X64-NEXT:    setae %al
; X64-NEXT:    retq
;
; EGPR-LABEL: and_cmp_const_power_of_two:
; EGPR:       # %bb.0:
; EGPR-NEXT:    btl %esi, %edi # encoding: [0x0f,0xa3,0xf7]
; EGPR-NEXT:    setae %al # encoding: [0x0f,0x93,0xc0]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %shl = shl i32 1, %y
  %and = and i32 %x, %shl
  %cmp = icmp ne i32 %and, %shl
  ret i1 %cmp
}

; Don't transform to 'andn' if there's another use of the 'and'.
define i32 @and_cmp_not_one_use(i32 %x) {
; X86-LABEL: and_cmp_not_one_use:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    andl $37, %ecx
; X86-NEXT:    xorl %eax, %eax
; X86-NEXT:    cmpl $37, %ecx
; X86-NEXT:    sete %al
; X86-NEXT:    addl %ecx, %eax
; X86-NEXT:    retl
;
; X64-LABEL: and_cmp_not_one_use:
; X64:       # %bb.0:
; X64-NEXT:    andl $37, %edi
; X64-NEXT:    xorl %eax, %eax
; X64-NEXT:    cmpl $37, %edi
; X64-NEXT:    sete %al
; X64-NEXT:    addl %edi, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: and_cmp_not_one_use:
; EGPR:       # %bb.0:
; EGPR-NEXT:    andl $37, %edi # encoding: [0x83,0xe7,0x25]
; EGPR-NEXT:    xorl %eax, %eax # encoding: [0x31,0xc0]
; EGPR-NEXT:    cmpl $37, %edi # encoding: [0x83,0xff,0x25]
; EGPR-NEXT:    sete %al # encoding: [0x0f,0x94,0xc0]
; EGPR-NEXT:    addl %edi, %eax # encoding: [0x01,0xf8]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %and = and i32 %x, 37
  %cmp = icmp eq i32 %and, 37
  %ext = zext i1 %cmp to i32
  %add = add i32 %and, %ext
  ret i32 %add
}

; Verify that we're not transforming invalid comparison predicates.
define i1 @not_an_andn1(i32 %x, i32 %y) {
; X86-LABEL: not_an_andn1:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    andl %eax, %ecx
; X86-NEXT:    cmpl %ecx, %eax
; X86-NEXT:    setg %al
; X86-NEXT:    retl
;
; X64-LABEL: not_an_andn1:
; X64:       # %bb.0:
; X64-NEXT:    andl %esi, %edi
; X64-NEXT:    cmpl %edi, %esi
; X64-NEXT:    setg %al
; X64-NEXT:    retq
;
; EGPR-LABEL: not_an_andn1:
; EGPR:       # %bb.0:
; EGPR-NEXT:    andl %esi, %edi # encoding: [0x21,0xf7]
; EGPR-NEXT:    cmpl %edi, %esi # encoding: [0x39,0xfe]
; EGPR-NEXT:    setg %al # encoding: [0x0f,0x9f,0xc0]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %and = and i32 %x, %y
  %cmp = icmp sgt i32 %y, %and
  ret i1 %cmp
}

define i1 @not_an_andn2(i32 %x, i32 %y) {
; X86-LABEL: not_an_andn2:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    andl %eax, %ecx
; X86-NEXT:    cmpl %ecx, %eax
; X86-NEXT:    setbe %al
; X86-NEXT:    retl
;
; X64-LABEL: not_an_andn2:
; X64:       # %bb.0:
; X64-NEXT:    andl %esi, %edi
; X64-NEXT:    cmpl %edi, %esi
; X64-NEXT:    setbe %al
; X64-NEXT:    retq
;
; EGPR-LABEL: not_an_andn2:
; EGPR:       # %bb.0:
; EGPR-NEXT:    andl %esi, %edi # encoding: [0x21,0xf7]
; EGPR-NEXT:    cmpl %edi, %esi # encoding: [0x39,0xfe]
; EGPR-NEXT:    setbe %al # encoding: [0x0f,0x96,0xc0]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %and = and i32 %y, %x
  %cmp = icmp ule i32 %y, %and
  ret i1 %cmp
}

; Don't choose a 'test' if an 'andn' can be used.
define i1 @andn_cmp_swap_ops(i64 %x, i64 %y) {
; X86-LABEL: andn_cmp_swap_ops:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    andnl {{[0-9]+}}(%esp), %ecx, %ecx
; X86-NEXT:    andnl {{[0-9]+}}(%esp), %eax, %eax
; X86-NEXT:    orl %ecx, %eax
; X86-NEXT:    sete %al
; X86-NEXT:    retl
;
; X64-LABEL: andn_cmp_swap_ops:
; X64:       # %bb.0:
; X64-NEXT:    andnq %rsi, %rdi, %rax
; X64-NEXT:    sete %al
; X64-NEXT:    retq
;
; EGPR-LABEL: andn_cmp_swap_ops:
; EGPR:       # %bb.0:
; EGPR-NEXT:    andnq %rsi, %rdi, %rax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xc0,0xf2,0xc6]
; EGPR-NEXT:    testq %rax, %rax # encoding: [0x48,0x85,0xc0]
; EGPR-NEXT:    sete %al # encoding: [0x0f,0x94,0xc0]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %notx = xor i64 %x, -1
  %and = and i64 %y, %notx
  %cmp = icmp eq i64 %and, 0
  ret i1 %cmp
}

; Use a 'test' (not an 'and') because 'andn' only works for i32/i64.
define i1 @andn_cmp_i8(i8 %x, i8 %y) {
; X86-LABEL: andn_cmp_i8:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    notb %al
; X86-NEXT:    testb %al, {{[0-9]+}}(%esp)
; X86-NEXT:    sete %al
; X86-NEXT:    retl
;
; X64-LABEL: andn_cmp_i8:
; X64:       # %bb.0:
; X64-NEXT:    notb %sil
; X64-NEXT:    testb %sil, %dil
; X64-NEXT:    sete %al
; X64-NEXT:    retq
;
; EGPR-LABEL: andn_cmp_i8:
; EGPR:       # %bb.0:
; EGPR-NEXT:    notb %sil # encoding: [0x40,0xf6,0xd6]
; EGPR-NEXT:    testb %sil, %dil # encoding: [0x40,0x84,0xf7]
; EGPR-NEXT:    sete %al # encoding: [0x0f,0x94,0xc0]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %noty = xor i8 %y, -1
  %and = and i8 %x, %noty
  %cmp = icmp eq i8 %and, 0
  ret i1 %cmp
}

; PR48768 - 'andn' clears the overflow flag, so we don't need a separate 'test'.
define i1 @andn_cmp_i32_overflow(i32 %x, i32 %y) {
; X86-LABEL: andn_cmp_i32_overflow:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    andnl {{[0-9]+}}(%esp), %eax, %eax
; X86-NEXT:    setle %al
; X86-NEXT:    retl
;
; X64-LABEL: andn_cmp_i32_overflow:
; X64:       # %bb.0:
; X64-NEXT:    andnl %edi, %esi, %eax
; X64-NEXT:    setle %al
; X64-NEXT:    retq
;
; EGPR-LABEL: andn_cmp_i32_overflow:
; EGPR:       # %bb.0:
; EGPR-NEXT:    andnl %edi, %esi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x48,0xf2,0xc7]
; EGPR-NEXT:    testl %eax, %eax # encoding: [0x85,0xc0]
; EGPR-NEXT:    setle %al # encoding: [0x0f,0x9e,0xc0]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %noty = xor i32 %y, -1
  %and = and i32 %x, %noty
  %cmp = icmp slt i32 %and, 1
  ret i1 %cmp
}

declare i32 @llvm.x86.bmi.bextr.32(i32, i32)

define i32 @bextr32(i32 %x, i32 %y)   {
; X86-LABEL: bextr32:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    bextrl %eax, {{[0-9]+}}(%esp), %eax
; X86-NEXT:    retl
;
; X64-LABEL: bextr32:
; X64:       # %bb.0:
; X64-NEXT:    bextrl %esi, %edi, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: bextr32:
; EGPR:       # %bb.0:
; EGPR-NEXT:    bextrl %esi, %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x48,0xf7,0xc7]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp = tail call i32 @llvm.x86.bmi.bextr.32(i32 %x, i32 %y)
  ret i32 %tmp
}

define i32 @bextr32_load(ptr %x, i32 %y)   {
; X86-LABEL: bextr32_load:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    bextrl %eax, (%ecx), %eax
; X86-NEXT:    retl
;
; X64-LABEL: bextr32_load:
; X64:       # %bb.0:
; X64-NEXT:    bextrl %esi, (%rdi), %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: bextr32_load:
; EGPR:       # %bb.0:
; EGPR-NEXT:    bextrl %esi, (%rdi), %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x48,0xf7,0x07]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %x1 = load i32, ptr %x
  %tmp = tail call i32 @llvm.x86.bmi.bextr.32(i32 %x1, i32 %y)
  ret i32 %tmp
}

define i32 @bextr32b(i32 %x)  uwtable  ssp {
; X86-SLOW-BEXTR-LABEL: bextr32b:
; X86-SLOW-BEXTR:       # %bb.0:
; X86-SLOW-BEXTR-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-SLOW-BEXTR-NEXT:    shrl $4, %eax
; X86-SLOW-BEXTR-NEXT:    andl $4095, %eax # imm = 0xFFF
; X86-SLOW-BEXTR-NEXT:    retl
;
; X64-SLOW-BEXTR-LABEL: bextr32b:
; X64-SLOW-BEXTR:       # %bb.0:
; X64-SLOW-BEXTR-NEXT:    movl %edi, %eax
; X64-SLOW-BEXTR-NEXT:    shrl $4, %eax
; X64-SLOW-BEXTR-NEXT:    andl $4095, %eax # imm = 0xFFF
; X64-SLOW-BEXTR-NEXT:    retq
;
; X86-FAST-BEXTR-LABEL: bextr32b:
; X86-FAST-BEXTR:       # %bb.0:
; X86-FAST-BEXTR-NEXT:    movl $3076, %eax # imm = 0xC04
; X86-FAST-BEXTR-NEXT:    bextrl %eax, {{[0-9]+}}(%esp), %eax
; X86-FAST-BEXTR-NEXT:    retl
;
; X64-FAST-BEXTR-LABEL: bextr32b:
; X64-FAST-BEXTR:       # %bb.0:
; X64-FAST-BEXTR-NEXT:    movl $3076, %eax # imm = 0xC04
; X64-FAST-BEXTR-NEXT:    bextrl %eax, %edi, %eax
; X64-FAST-BEXTR-NEXT:    retq
;
; EGPR-LABEL: bextr32b:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movl $3076, %eax # encoding: [0xb8,0x04,0x0c,0x00,0x00]
; EGPR-NEXT:    # imm = 0xC04
; EGPR-NEXT:    bextrl %eax, %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf7,0xc7]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %1 = lshr i32 %x, 4
  %2 = and i32 %1, 4095
  ret i32 %2
}

; Make sure we still use AH subreg trick to extract 15:8
define i32 @bextr32_subreg(i32 %x)  uwtable  ssp {
; X86-LABEL: bextr32_subreg:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    retl
;
; X64-LABEL: bextr32_subreg:
; X64:       # %bb.0:
; X64-NEXT:    movl %edi, %eax
; X64-NEXT:    movzbl %ah, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: bextr32_subreg:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movl %edi, %eax # encoding: [0x89,0xf8]
; EGPR-NEXT:    movzbl %ah, %eax # encoding: [0x0f,0xb6,0xc4]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %1 = lshr i32 %x, 8
  %2 = and i32 %1, 255
  ret i32 %2
}

define i32 @bextr32b_load(ptr %x)  uwtable  ssp {
; X86-SLOW-BEXTR-LABEL: bextr32b_load:
; X86-SLOW-BEXTR:       # %bb.0:
; X86-SLOW-BEXTR-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-SLOW-BEXTR-NEXT:    movl (%eax), %eax
; X86-SLOW-BEXTR-NEXT:    shrl $4, %eax
; X86-SLOW-BEXTR-NEXT:    andl $4095, %eax # imm = 0xFFF
; X86-SLOW-BEXTR-NEXT:    retl
;
; X64-SLOW-BEXTR-LABEL: bextr32b_load:
; X64-SLOW-BEXTR:       # %bb.0:
; X64-SLOW-BEXTR-NEXT:    movl (%rdi), %eax
; X64-SLOW-BEXTR-NEXT:    shrl $4, %eax
; X64-SLOW-BEXTR-NEXT:    andl $4095, %eax # imm = 0xFFF
; X64-SLOW-BEXTR-NEXT:    retq
;
; X86-FAST-BEXTR-LABEL: bextr32b_load:
; X86-FAST-BEXTR:       # %bb.0:
; X86-FAST-BEXTR-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-FAST-BEXTR-NEXT:    movl $3076, %ecx # imm = 0xC04
; X86-FAST-BEXTR-NEXT:    bextrl %ecx, (%eax), %eax
; X86-FAST-BEXTR-NEXT:    retl
;
; X64-FAST-BEXTR-LABEL: bextr32b_load:
; X64-FAST-BEXTR:       # %bb.0:
; X64-FAST-BEXTR-NEXT:    movl $3076, %eax # imm = 0xC04
; X64-FAST-BEXTR-NEXT:    bextrl %eax, (%rdi), %eax
; X64-FAST-BEXTR-NEXT:    retq
;
; EGPR-LABEL: bextr32b_load:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movl $3076, %eax # encoding: [0xb8,0x04,0x0c,0x00,0x00]
; EGPR-NEXT:    # imm = 0xC04
; EGPR-NEXT:    bextrl %eax, (%rdi), %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf7,0x07]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %1 = load i32, ptr %x
  %2 = lshr i32 %1, 4
  %3 = and i32 %2, 4095
  ret i32 %3
}

; PR34042
define i32 @bextr32c(i32 %x, i16 zeroext %y) {
; X86-LABEL: bextr32c:
; X86:       # %bb.0:
; X86-NEXT:    movswl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    bextrl %eax, {{[0-9]+}}(%esp), %eax
; X86-NEXT:    retl
;
; X64-LABEL: bextr32c:
; X64:       # %bb.0:
; X64-NEXT:    bextrl %esi, %edi, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: bextr32c:
; EGPR:       # %bb.0:
; EGPR-NEXT:    bextrl %esi, %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x48,0xf7,0xc7]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp0 = sext i16 %y to i32
  %tmp1 = tail call i32 @llvm.x86.bmi.bextr.32(i32 %x, i32 %tmp0)
  ret i32 %tmp1
}

define i32 @non_bextr32(i32 %x) {
; X86-LABEL: non_bextr32:
; X86:       # %bb.0: # %entry
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    shrl $2, %eax
; X86-NEXT:    andl $111, %eax
; X86-NEXT:    retl
;
; X64-LABEL: non_bextr32:
; X64:       # %bb.0: # %entry
; X64-NEXT:    movl %edi, %eax
; X64-NEXT:    shrl $2, %eax
; X64-NEXT:    andl $111, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: non_bextr32:
; EGPR:       # %bb.0: # %entry
; EGPR-NEXT:    movl %edi, %eax # encoding: [0x89,0xf8]
; EGPR-NEXT:    shrl $2, %eax # encoding: [0xc1,0xe8,0x02]
; EGPR-NEXT:    andl $111, %eax # encoding: [0x83,0xe0,0x6f]
; EGPR-NEXT:    retq # encoding: [0xc3]
entry:
  %shr = lshr i32 %x, 2
  %and = and i32 %shr, 111
  ret i32 %and
}

define i32 @blsi32(i32 %x)   {
; X86-LABEL: blsi32:
; X86:       # %bb.0:
; X86-NEXT:    blsil {{[0-9]+}}(%esp), %eax
; X86-NEXT:    retl
;
; X64-LABEL: blsi32:
; X64:       # %bb.0:
; X64-NEXT:    blsil %edi, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsi32:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsil %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf3,0xdf]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp = sub i32 0, %x
  %tmp2 = and i32 %x, %tmp
  ret i32 %tmp2
}

define i32 @blsi32_load(ptr %x)   {
; X86-LABEL: blsi32_load:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    blsil (%eax), %eax
; X86-NEXT:    retl
;
; X64-LABEL: blsi32_load:
; X64:       # %bb.0:
; X64-NEXT:    blsil (%rdi), %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsi32_load:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsil (%rdi), %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf3,0x1f]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %x1 = load i32, ptr %x
  %tmp = sub i32 0, %x1
  %tmp2 = and i32 %x1, %tmp
  ret i32 %tmp2
}

define i32 @blsi32_z(i32 %a, i32 %b) nounwind {
; X86-LABEL: blsi32_z:
; X86:       # %bb.0:
; X86-NEXT:    blsil {{[0-9]+}}(%esp), %eax
; X86-NEXT:    jne .LBB25_2
; X86-NEXT:  # %bb.1:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:  .LBB25_2:
; X86-NEXT:    retl
;
; X64-LABEL: blsi32_z:
; X64:       # %bb.0:
; X64-NEXT:    blsil %edi, %eax
; X64-NEXT:    cmovel %esi, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsi32_z:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsil %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf3,0xdf]
; EGPR-NEXT:    testl %eax, %eax # encoding: [0x85,0xc0]
; EGPR-NEXT:    cmovel %esi, %eax # encoding: [0x0f,0x44,0xc6]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i32 0, %a
  %t1 = and i32 %t0, %a
  %t2 = icmp eq i32 %t1, 0
  %t3 = select i1 %t2, i32 %b, i32 %t1
  ret i32 %t3
}

define i32 @blsi32_z2(i32 %a, i32 %b, i32 %c) nounwind {
; X86-LABEL: blsi32_z2:
; X86:       # %bb.0:
; X86-NEXT:    blsil {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    cmovel %eax, %ecx
; X86-NEXT:    movl (%ecx), %eax
; X86-NEXT:    retl
;
; X64-LABEL: blsi32_z2:
; X64:       # %bb.0:
; X64-NEXT:    movl %esi, %eax
; X64-NEXT:    blsil %edi, %ecx
; X64-NEXT:    cmovnel %edx, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsi32_z2:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movl %esi, %eax # encoding: [0x89,0xf0]
; EGPR-NEXT:    blsil %edi, %ecx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x70,0xf3,0xdf]
; EGPR-NEXT:    testl %ecx, %ecx # encoding: [0x85,0xc9]
; EGPR-NEXT:    cmovnel %edx, %eax # encoding: [0x0f,0x45,0xc2]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i32 0, %a
  %t1 = and i32 %t0, %a
  %t2 = icmp eq i32 %t1, 0
  %t3 = select i1 %t2, i32 %b, i32 %c
  ret i32 %t3
}

; Inspired by PR48768, but using cmovcc instead of setcc. There should be
; no test instruction.
define i32 @blsi32_sle(i32 %a, i32 %b, i32 %c) nounwind {
; X86-LABEL: blsi32_sle:
; X86:       # %bb.0:
; X86-NEXT:    blsil {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    cmovlel %eax, %ecx
; X86-NEXT:    movl (%ecx), %eax
; X86-NEXT:    retl
;
; X64-LABEL: blsi32_sle:
; X64:       # %bb.0:
; X64-NEXT:    movl %esi, %eax
; X64-NEXT:    blsil %edi, %ecx
; X64-NEXT:    cmovgl %edx, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsi32_sle:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movl %esi, %eax # encoding: [0x89,0xf0]
; EGPR-NEXT:    blsil %edi, %ecx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x70,0xf3,0xdf]
; EGPR-NEXT:    testl %ecx, %ecx # encoding: [0x85,0xc9]
; EGPR-NEXT:    cmovgl %edx, %eax # encoding: [0x0f,0x4f,0xc2]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i32 0, %a
  %t1 = and i32 %t0, %a
  %t2 = icmp sle i32 %t1, 0
  %t3 = select i1 %t2, i32 %b, i32 %c
  ret i32 %t3
}

define i64 @blsi64(i64 %x)   {
; X86-LABEL: blsi64:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    .cfi_offset %esi, -8
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X86-NEXT:    xorl %edx, %edx
; X86-NEXT:    movl %ecx, %eax
; X86-NEXT:    negl %eax
; X86-NEXT:    sbbl %esi, %edx
; X86-NEXT:    andl %esi, %edx
; X86-NEXT:    andl %ecx, %eax
; X86-NEXT:    popl %esi
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl
;
; X64-LABEL: blsi64:
; X64:       # %bb.0:
; X64-NEXT:    blsiq %rdi, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsi64:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsiq %rdi, %rax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf8,0xf3,0xdf]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp = sub i64 0, %x
  %tmp2 = and i64 %tmp, %x
  ret i64 %tmp2
}

define i64 @blsi64_z(i64 %a, i64 %b) nounwind {
; X86-LABEL: blsi64_z:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X86-NEXT:    xorl %edx, %edx
; X86-NEXT:    movl %ecx, %eax
; X86-NEXT:    negl %eax
; X86-NEXT:    sbbl %esi, %edx
; X86-NEXT:    andl %esi, %edx
; X86-NEXT:    andl %ecx, %eax
; X86-NEXT:    movl %eax, %ecx
; X86-NEXT:    orl %edx, %ecx
; X86-NEXT:    jne .LBB29_2
; X86-NEXT:  # %bb.1:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:  .LBB29_2:
; X86-NEXT:    popl %esi
; X86-NEXT:    retl
;
; X64-LABEL: blsi64_z:
; X64:       # %bb.0:
; X64-NEXT:    blsiq %rdi, %rax
; X64-NEXT:    cmoveq %rsi, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsi64_z:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsiq %rdi, %rax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf8,0xf3,0xdf]
; EGPR-NEXT:    testq %rax, %rax # encoding: [0x48,0x85,0xc0]
; EGPR-NEXT:    cmoveq %rsi, %rax # encoding: [0x48,0x0f,0x44,0xc6]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i64 0, %a
  %t1 = and i64 %t0, %a
  %t2 = icmp eq i64 %t1, 0
  %t3 = select i1 %t2, i64 %b, i64 %t1
  ret i64 %t3
}

define i64 @blsi64_z2(i64 %a, i64 %b, i64 %c) nounwind {
; X86-LABEL: blsi64_z2:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    xorl %edx, %edx
; X86-NEXT:    movl %eax, %esi
; X86-NEXT:    negl %esi
; X86-NEXT:    sbbl %ecx, %edx
; X86-NEXT:    andl %ecx, %edx
; X86-NEXT:    andl %eax, %esi
; X86-NEXT:    orl %edx, %esi
; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    cmovel %eax, %ecx
; X86-NEXT:    movl (%ecx), %eax
; X86-NEXT:    movl 4(%ecx), %edx
; X86-NEXT:    popl %esi
; X86-NEXT:    retl
;
; X64-LABEL: blsi64_z2:
; X64:       # %bb.0:
; X64-NEXT:    movq %rsi, %rax
; X64-NEXT:    blsiq %rdi, %rcx
; X64-NEXT:    cmovneq %rdx, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsi64_z2:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movq %rsi, %rax # encoding: [0x48,0x89,0xf0]
; EGPR-NEXT:    blsiq %rdi, %rcx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf0,0xf3,0xdf]
; EGPR-NEXT:    testq %rcx, %rcx # encoding: [0x48,0x85,0xc9]
; EGPR-NEXT:    cmovneq %rdx, %rax # encoding: [0x48,0x0f,0x45,0xc2]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i64 0, %a
  %t1 = and i64 %t0, %a
  %t2 = icmp eq i64 %t1, 0
  %t3 = select i1 %t2, i64 %b, i64 %c
  ret i64 %t3
}

define i64 @blsi64_sle(i64 %a, i64 %b, i64 %c) nounwind {
; X86-LABEL: blsi64_sle:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    xorl %edx, %edx
; X86-NEXT:    movl %eax, %esi
; X86-NEXT:    negl %esi
; X86-NEXT:    sbbl %ecx, %edx
; X86-NEXT:    andl %ecx, %edx
; X86-NEXT:    andl %eax, %esi
; X86-NEXT:    cmpl $1, %esi
; X86-NEXT:    sbbl $0, %edx
; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    cmovll %eax, %ecx
; X86-NEXT:    movl (%ecx), %eax
; X86-NEXT:    movl 4(%ecx), %edx
; X86-NEXT:    popl %esi
; X86-NEXT:    retl
;
; X64-LABEL: blsi64_sle:
; X64:       # %bb.0:
; X64-NEXT:    movq %rsi, %rax
; X64-NEXT:    blsiq %rdi, %rcx
; X64-NEXT:    cmovgq %rdx, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsi64_sle:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movq %rsi, %rax # encoding: [0x48,0x89,0xf0]
; EGPR-NEXT:    blsiq %rdi, %rcx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf0,0xf3,0xdf]
; EGPR-NEXT:    testq %rcx, %rcx # encoding: [0x48,0x85,0xc9]
; EGPR-NEXT:    cmovgq %rdx, %rax # encoding: [0x48,0x0f,0x4f,0xc2]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i64 0, %a
  %t1 = and i64 %t0, %a
  %t2 = icmp sle i64 %t1, 0
  %t3 = select i1 %t2, i64 %b, i64 %c
  ret i64 %t3
}

define i32 @blsmsk32(i32 %x)   {
; X86-LABEL: blsmsk32:
; X86:       # %bb.0:
; X86-NEXT:    blsmskl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    retl
;
; X64-LABEL: blsmsk32:
; X64:       # %bb.0:
; X64-NEXT:    blsmskl %edi, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsmsk32:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsmskl %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf3,0xd7]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp = sub i32 %x, 1
  %tmp2 = xor i32 %x, %tmp
  ret i32 %tmp2
}

define i32 @blsmsk32_load(ptr %x)   {
; X86-LABEL: blsmsk32_load:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    blsmskl (%eax), %eax
; X86-NEXT:    retl
;
; X64-LABEL: blsmsk32_load:
; X64:       # %bb.0:
; X64-NEXT:    blsmskl (%rdi), %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsmsk32_load:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsmskl (%rdi), %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf3,0x17]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %x1 = load i32, ptr %x
  %tmp = sub i32 %x1, 1
  %tmp2 = xor i32 %x1, %tmp
  ret i32 %tmp2
}

define i32 @blsmsk32_z(i32 %a, i32 %b) nounwind {
; X86-LABEL: blsmsk32_z:
; X86:       # %bb.0:
; X86-NEXT:    blsmskl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    jne .LBB34_2
; X86-NEXT:  # %bb.1:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:  .LBB34_2:
; X86-NEXT:    retl
;
; X64-LABEL: blsmsk32_z:
; X64:       # %bb.0:
; X64-NEXT:    blsmskl %edi, %eax
; X64-NEXT:    cmovel %esi, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsmsk32_z:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsmskl %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf3,0xd7]
; EGPR-NEXT:    testl %eax, %eax # encoding: [0x85,0xc0]
; EGPR-NEXT:    cmovel %esi, %eax # encoding: [0x0f,0x44,0xc6]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i32 %a, 1
  %t1 = xor i32 %t0, %a
  %t2 = icmp eq i32 %t1, 0
  %t3 = select i1 %t2, i32 %b, i32 %t1
  ret i32 %t3
}

define i32 @blsmsk32_z2(i32 %a, i32 %b, i32 %c) nounwind {
; X86-LABEL: blsmsk32_z2:
; X86:       # %bb.0:
; X86-NEXT:    blsmskl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    cmovel %eax, %ecx
; X86-NEXT:    movl (%ecx), %eax
; X86-NEXT:    retl
;
; X64-LABEL: blsmsk32_z2:
; X64:       # %bb.0:
; X64-NEXT:    movl %esi, %eax
; X64-NEXT:    blsmskl %edi, %ecx
; X64-NEXT:    cmovnel %edx, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsmsk32_z2:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movl %esi, %eax # encoding: [0x89,0xf0]
; EGPR-NEXT:    blsmskl %edi, %ecx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x70,0xf3,0xd7]
; EGPR-NEXT:    cmovnel %edx, %eax # encoding: [0x0f,0x45,0xc2]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i32 %a, 1
  %t1 = xor i32 %t0, %a
  %t2 = icmp eq i32 %t1, 0
  %t3 = select i1 %t2, i32 %b, i32 %c
  ret i32 %t3
}

define i32 @blsmsk32_sle(i32 %a, i32 %b, i32 %c) nounwind {
; X86-LABEL: blsmsk32_sle:
; X86:       # %bb.0:
; X86-NEXT:    blsmskl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    cmovlel %eax, %ecx
; X86-NEXT:    movl (%ecx), %eax
; X86-NEXT:    retl
;
; X64-LABEL: blsmsk32_sle:
; X64:       # %bb.0:
; X64-NEXT:    movl %esi, %eax
; X64-NEXT:    blsmskl %edi, %ecx
; X64-NEXT:    cmovgl %edx, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsmsk32_sle:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movl %esi, %eax # encoding: [0x89,0xf0]
; EGPR-NEXT:    blsmskl %edi, %ecx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x70,0xf3,0xd7]
; EGPR-NEXT:    testl %ecx, %ecx # encoding: [0x85,0xc9]
; EGPR-NEXT:    cmovgl %edx, %eax # encoding: [0x0f,0x4f,0xc2]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i32 %a, 1
  %t1 = xor i32 %t0, %a
  %t2 = icmp sle i32 %t1, 0
  %t3 = select i1 %t2, i32 %b, i32 %c
  ret i32 %t3
}

define i64 @blsmsk64(i64 %x)   {
; X86-LABEL: blsmsk64:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    .cfi_offset %esi, -8
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X86-NEXT:    movl %ecx, %eax
; X86-NEXT:    addl $-1, %eax
; X86-NEXT:    movl %esi, %edx
; X86-NEXT:    adcl $-1, %edx
; X86-NEXT:    xorl %ecx, %eax
; X86-NEXT:    xorl %esi, %edx
; X86-NEXT:    popl %esi
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl
;
; X64-LABEL: blsmsk64:
; X64:       # %bb.0:
; X64-NEXT:    blsmskq %rdi, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsmsk64:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsmskq %rdi, %rax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf8,0xf3,0xd7]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp = sub i64 %x, 1
  %tmp2 = xor i64 %tmp, %x
  ret i64 %tmp2
}

define i64 @blsmsk64_z(i64 %a, i64 %b) nounwind {
; X86-LABEL: blsmsk64_z:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X86-NEXT:    movl %ecx, %eax
; X86-NEXT:    addl $-1, %eax
; X86-NEXT:    movl %esi, %edx
; X86-NEXT:    adcl $-1, %edx
; X86-NEXT:    xorl %ecx, %eax
; X86-NEXT:    xorl %esi, %edx
; X86-NEXT:    movl %eax, %ecx
; X86-NEXT:    orl %edx, %ecx
; X86-NEXT:    jne .LBB38_2
; X86-NEXT:  # %bb.1:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
; X86-NEXT:  .LBB38_2:
; X86-NEXT:    popl %esi
; X86-NEXT:    retl
;
; X64-LABEL: blsmsk64_z:
; X64:       # %bb.0:
; X64-NEXT:    blsmskq %rdi, %rax
; X64-NEXT:    cmoveq %rsi, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsmsk64_z:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsmskq %rdi, %rax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf8,0xf3,0xd7]
; EGPR-NEXT:    testq %rax, %rax # encoding: [0x48,0x85,0xc0]
; EGPR-NEXT:    cmoveq %rsi, %rax # encoding: [0x48,0x0f,0x44,0xc6]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i64 %a, 1
  %t1 = xor i64 %t0, %a
  %t2 = icmp eq i64 %t1, 0
  %t3 = select i1 %t2, i64 %b, i64 %t1
  ret i64 %t3
}

define i64 @blsmsk64_z2(i64 %a, i64 %b, i64 %c) nounwind {
; X86-LABEL: blsmsk64_z2:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl %eax, %edx
; X86-NEXT:    addl $-1, %edx
; X86-NEXT:    movl %ecx, %esi
; X86-NEXT:    adcl $-1, %esi
; X86-NEXT:    xorl %eax, %edx
; X86-NEXT:    xorl %ecx, %esi
; X86-NEXT:    orl %edx, %esi
; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    cmovel %eax, %ecx
; X86-NEXT:    movl (%ecx), %eax
; X86-NEXT:    movl 4(%ecx), %edx
; X86-NEXT:    popl %esi
; X86-NEXT:    retl
;
; X64-LABEL: blsmsk64_z2:
; X64:       # %bb.0:
; X64-NEXT:    movq %rsi, %rax
; X64-NEXT:    blsmskq %rdi, %rcx
; X64-NEXT:    cmovneq %rdx, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsmsk64_z2:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movq %rsi, %rax # encoding: [0x48,0x89,0xf0]
; EGPR-NEXT:    blsmskq %rdi, %rcx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf0,0xf3,0xd7]
; EGPR-NEXT:    cmovneq %rdx, %rax # encoding: [0x48,0x0f,0x45,0xc2]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i64 %a, 1
  %t1 = xor i64 %t0, %a
  %t2 = icmp eq i64 %t1, 0
  %t3 = select i1 %t2, i64 %b, i64 %c
  ret i64 %t3
}

define i64 @blsmsk64_sle(i64 %a, i64 %b, i64 %c) nounwind {
; X86-LABEL: blsmsk64_sle:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl %eax, %edx
; X86-NEXT:    addl $-1, %edx
; X86-NEXT:    movl %ecx, %esi
; X86-NEXT:    adcl $-1, %esi
; X86-NEXT:    xorl %ecx, %esi
; X86-NEXT:    xorl %eax, %edx
; X86-NEXT:    cmpl $1, %edx
; X86-NEXT:    sbbl $0, %esi
; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    cmovll %eax, %ecx
; X86-NEXT:    movl (%ecx), %eax
; X86-NEXT:    movl 4(%ecx), %edx
; X86-NEXT:    popl %esi
; X86-NEXT:    retl
;
; X64-LABEL: blsmsk64_sle:
; X64:       # %bb.0:
; X64-NEXT:    movq %rsi, %rax
; X64-NEXT:    blsmskq %rdi, %rcx
; X64-NEXT:    cmovgq %rdx, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsmsk64_sle:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movq %rsi, %rax # encoding: [0x48,0x89,0xf0]
; EGPR-NEXT:    blsmskq %rdi, %rcx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf0,0xf3,0xd7]
; EGPR-NEXT:    testq %rcx, %rcx # encoding: [0x48,0x85,0xc9]
; EGPR-NEXT:    cmovgq %rdx, %rax # encoding: [0x48,0x0f,0x4f,0xc2]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i64 %a, 1
  %t1 = xor i64 %t0, %a
  %t2 = icmp sle i64 %t1, 0
  %t3 = select i1 %t2, i64 %b, i64 %c
  ret i64 %t3
}

define i32 @blsr32(i32 %x)   {
; X86-LABEL: blsr32:
; X86:       # %bb.0:
; X86-NEXT:    blsrl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    retl
;
; X64-LABEL: blsr32:
; X64:       # %bb.0:
; X64-NEXT:    blsrl %edi, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsr32:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsrl %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf3,0xcf]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp = sub i32 %x, 1
  %tmp2 = and i32 %x, %tmp
  ret i32 %tmp2
}

define i32 @blsr32_load(ptr %x)   {
; X86-LABEL: blsr32_load:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    blsrl (%eax), %eax
; X86-NEXT:    retl
;
; X64-LABEL: blsr32_load:
; X64:       # %bb.0:
; X64-NEXT:    blsrl (%rdi), %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsr32_load:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsrl (%rdi), %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf3,0x0f]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %x1 = load i32, ptr %x
  %tmp = sub i32 %x1, 1
  %tmp2 = and i32 %x1, %tmp
  ret i32 %tmp2
}

define i32 @blsr32_z(i32 %a, i32 %b) nounwind {
; X86-LABEL: blsr32_z:
; X86:       # %bb.0:
; X86-NEXT:    blsrl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    jne .LBB43_2
; X86-NEXT:  # %bb.1:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:  .LBB43_2:
; X86-NEXT:    retl
;
; X64-LABEL: blsr32_z:
; X64:       # %bb.0:
; X64-NEXT:    blsrl %edi, %eax
; X64-NEXT:    cmovel %esi, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsr32_z:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsrl %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf3,0xcf]
; EGPR-NEXT:    testl %eax, %eax # encoding: [0x85,0xc0]
; EGPR-NEXT:    cmovel %esi, %eax # encoding: [0x0f,0x44,0xc6]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i32 %a, 1
  %t1 = and i32 %t0, %a
  %t2 = icmp eq i32 %t1, 0
  %t3 = select i1 %t2, i32 %b, i32 %t1
  ret i32 %t3
}

define i32 @blsr32_z2(i32 %a, i32 %b, i32 %c) nounwind {
; X86-LABEL: blsr32_z2:
; X86:       # %bb.0:
; X86-NEXT:    blsrl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    cmovel %eax, %ecx
; X86-NEXT:    movl (%ecx), %eax
; X86-NEXT:    retl
;
; X64-LABEL: blsr32_z2:
; X64:       # %bb.0:
; X64-NEXT:    movl %esi, %eax
; X64-NEXT:    blsrl %edi, %ecx
; X64-NEXT:    cmovnel %edx, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsr32_z2:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movl %esi, %eax # encoding: [0x89,0xf0]
; EGPR-NEXT:    blsrl %edi, %ecx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x70,0xf3,0xcf]
; EGPR-NEXT:    testl %ecx, %ecx # encoding: [0x85,0xc9]
; EGPR-NEXT:    cmovnel %edx, %eax # encoding: [0x0f,0x45,0xc2]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i32 %a, 1
  %t1 = and i32 %t0, %a
  %t2 = icmp eq i32 %t1, 0
  %t3 = select i1 %t2, i32 %b, i32 %c
  ret i32 %t3
}

define i32 @blsr32_sle(i32 %a, i32 %b, i32 %c) nounwind {
; X86-LABEL: blsr32_sle:
; X86:       # %bb.0:
; X86-NEXT:    blsrl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    cmovlel %eax, %ecx
; X86-NEXT:    movl (%ecx), %eax
; X86-NEXT:    retl
;
; X64-LABEL: blsr32_sle:
; X64:       # %bb.0:
; X64-NEXT:    movl %esi, %eax
; X64-NEXT:    blsrl %edi, %ecx
; X64-NEXT:    cmovgl %edx, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsr32_sle:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movl %esi, %eax # encoding: [0x89,0xf0]
; EGPR-NEXT:    blsrl %edi, %ecx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x70,0xf3,0xcf]
; EGPR-NEXT:    testl %ecx, %ecx # encoding: [0x85,0xc9]
; EGPR-NEXT:    cmovgl %edx, %eax # encoding: [0x0f,0x4f,0xc2]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i32 %a, 1
  %t1 = and i32 %t0, %a
  %t2 = icmp sle i32 %t1, 0
  %t3 = select i1 %t2, i32 %b, i32 %c
  ret i32 %t3
}

define i64 @blsr64(i64 %x)   {
; X86-LABEL: blsr64:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    .cfi_offset %esi, -8
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X86-NEXT:    movl %ecx, %eax
; X86-NEXT:    addl $-1, %eax
; X86-NEXT:    movl %esi, %edx
; X86-NEXT:    adcl $-1, %edx
; X86-NEXT:    andl %ecx, %eax
; X86-NEXT:    andl %esi, %edx
; X86-NEXT:    popl %esi
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl
;
; X64-LABEL: blsr64:
; X64:       # %bb.0:
; X64-NEXT:    blsrq %rdi, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsr64:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsrq %rdi, %rax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf8,0xf3,0xcf]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp = sub i64 %x, 1
  %tmp2 = and i64 %tmp, %x
  ret i64 %tmp2
}

define i64 @blsr64_z(i64 %a, i64 %b) nounwind {
; X86-LABEL: blsr64_z:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X86-NEXT:    movl %ecx, %eax
; X86-NEXT:    addl $-1, %eax
; X86-NEXT:    movl %esi, %edx
; X86-NEXT:    adcl $-1, %edx
; X86-NEXT:    andl %ecx, %eax
; X86-NEXT:    andl %esi, %edx
; X86-NEXT:    movl %eax, %ecx
; X86-NEXT:    orl %edx, %ecx
; X86-NEXT:    jne .LBB47_2
; X86-NEXT:  # %bb.1:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
; X86-NEXT:  .LBB47_2:
; X86-NEXT:    popl %esi
; X86-NEXT:    retl
;
; X64-LABEL: blsr64_z:
; X64:       # %bb.0:
; X64-NEXT:    blsrq %rdi, %rax
; X64-NEXT:    cmoveq %rsi, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsr64_z:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsrq %rdi, %rax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf8,0xf3,0xcf]
; EGPR-NEXT:    testq %rax, %rax # encoding: [0x48,0x85,0xc0]
; EGPR-NEXT:    cmoveq %rsi, %rax # encoding: [0x48,0x0f,0x44,0xc6]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i64 %a, 1
  %t1 = and i64 %t0, %a
  %t2 = icmp eq i64 %t1, 0
  %t3 = select i1 %t2, i64 %b, i64 %t1
  ret i64 %t3
}

define i64 @blsr64_z2(i64 %a, i64 %b, i64 %c) nounwind {
; X86-LABEL: blsr64_z2:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl %eax, %edx
; X86-NEXT:    addl $-1, %edx
; X86-NEXT:    movl %ecx, %esi
; X86-NEXT:    adcl $-1, %esi
; X86-NEXT:    andl %eax, %edx
; X86-NEXT:    andl %ecx, %esi
; X86-NEXT:    orl %edx, %esi
; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    cmovel %eax, %ecx
; X86-NEXT:    movl (%ecx), %eax
; X86-NEXT:    movl 4(%ecx), %edx
; X86-NEXT:    popl %esi
; X86-NEXT:    retl
;
; X64-LABEL: blsr64_z2:
; X64:       # %bb.0:
; X64-NEXT:    movq %rsi, %rax
; X64-NEXT:    blsrq %rdi, %rcx
; X64-NEXT:    cmovneq %rdx, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsr64_z2:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movq %rsi, %rax # encoding: [0x48,0x89,0xf0]
; EGPR-NEXT:    blsrq %rdi, %rcx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf0,0xf3,0xcf]
; EGPR-NEXT:    testq %rcx, %rcx # encoding: [0x48,0x85,0xc9]
; EGPR-NEXT:    cmovneq %rdx, %rax # encoding: [0x48,0x0f,0x45,0xc2]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i64 %a, 1
  %t1 = and i64 %t0, %a
  %t2 = icmp eq i64 %t1, 0
  %t3 = select i1 %t2, i64 %b, i64 %c
  ret i64 %t3
}

define i64 @blsr64_sle(i64 %a, i64 %b, i64 %c) nounwind {
; X86-LABEL: blsr64_sle:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl %eax, %edx
; X86-NEXT:    addl $-1, %edx
; X86-NEXT:    movl %ecx, %esi
; X86-NEXT:    adcl $-1, %esi
; X86-NEXT:    andl %ecx, %esi
; X86-NEXT:    andl %eax, %edx
; X86-NEXT:    cmpl $1, %edx
; X86-NEXT:    sbbl $0, %esi
; X86-NEXT:    leal {{[0-9]+}}(%esp), %eax
; X86-NEXT:    leal {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    cmovll %eax, %ecx
; X86-NEXT:    movl (%ecx), %eax
; X86-NEXT:    movl 4(%ecx), %edx
; X86-NEXT:    popl %esi
; X86-NEXT:    retl
;
; X64-LABEL: blsr64_sle:
; X64:       # %bb.0:
; X64-NEXT:    movq %rsi, %rax
; X64-NEXT:    blsrq %rdi, %rcx
; X64-NEXT:    cmovgq %rdx, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsr64_sle:
; EGPR:       # %bb.0:
; EGPR-NEXT:    movq %rsi, %rax # encoding: [0x48,0x89,0xf0]
; EGPR-NEXT:    blsrq %rdi, %rcx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf0,0xf3,0xcf]
; EGPR-NEXT:    testq %rcx, %rcx # encoding: [0x48,0x85,0xc9]
; EGPR-NEXT:    cmovgq %rdx, %rax # encoding: [0x48,0x0f,0x4f,0xc2]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %t0 = sub i64 %a, 1
  %t1 = and i64 %t0, %a
  %t2 = icmp sle i64 %t1, 0
  %t3 = select i1 %t2, i64 %b, i64 %c
  ret i64 %t3
}

; PR35792 - https://bugs.llvm.org/show_bug.cgi?id=35792

define i64 @blsr_disguised_constant(i64 %x) {
; X86-LABEL: blsr_disguised_constant:
; X86:       # %bb.0:
; X86-NEXT:    blsrl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movzwl %ax, %eax
; X86-NEXT:    xorl %edx, %edx
; X86-NEXT:    retl
;
; X64-LABEL: blsr_disguised_constant:
; X64:       # %bb.0:
; X64-NEXT:    blsrl %edi, %eax
; X64-NEXT:    movzwl %ax, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsr_disguised_constant:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsrl %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf3,0xcf]
; EGPR-NEXT:    movzwl %ax, %eax # encoding: [0x0f,0xb7,0xc0]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %a1 = and i64 %x, 65535
  %a2 = add i64 %x, 65535
  %r = and i64 %a1, %a2
  ret i64 %r
}

; The add here used to get shrunk, but the and did not thus hiding the blsr pattern.
; We now use the knowledge that upper bits of the shift guarantee the and result has 0s in the upper bits to reduce it too.
define i64 @blsr_disguised_shrunk_add(i64 %x) {
; X86-LABEL: blsr_disguised_shrunk_add:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    shrl $16, %eax
; X86-NEXT:    blsrl %eax, %eax
; X86-NEXT:    xorl %edx, %edx
; X86-NEXT:    retl
;
; X64-LABEL: blsr_disguised_shrunk_add:
; X64:       # %bb.0:
; X64-NEXT:    shrq $48, %rdi
; X64-NEXT:    blsrl %edi, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsr_disguised_shrunk_add:
; EGPR:       # %bb.0:
; EGPR-NEXT:    shrq $48, %rdi # encoding: [0x48,0xc1,0xef,0x30]
; EGPR-NEXT:    blsrl %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf3,0xcf]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %a = lshr i64 %x, 48
  %b = add i64 %a, -1
  %c = and i64 %b, %a
  ret i64 %c
}

define void @pr40060(i32, i32) {
; X86-LABEL: pr40060:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    bextrl %eax, {{[0-9]+}}(%esp), %eax
; X86-NEXT:    testl %eax, %eax
; X86-NEXT:    jns bar # TAILCALL
; X86-NEXT:  # %bb.1:
; X86-NEXT:    retl
;
; X64-LABEL: pr40060:
; X64:       # %bb.0:
; X64-NEXT:    bextrl %esi, %edi, %eax
; X64-NEXT:    testl %eax, %eax
; X64-NEXT:    jns bar # TAILCALL
; X64-NEXT:  # %bb.1:
; X64-NEXT:    retq
;
; EGPR-LABEL: pr40060:
; EGPR:       # %bb.0:
; EGPR-NEXT:    bextrl %esi, %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x48,0xf7,0xc7]
; EGPR-NEXT:    testl %eax, %eax # encoding: [0x85,0xc0]
; EGPR-NEXT:    jns bar # TAILCALL
; EGPR-NEXT:    # encoding: [0x79,A]
; EGPR-NEXT:    # fixup A - offset: 1, value: bar-1, kind: FK_PCRel_1
; EGPR-NEXT:  # %bb.1:
; EGPR-NEXT:    retq # encoding: [0xc3]
  %3 = tail call i32 @llvm.x86.bmi.bextr.32(i32 %0, i32 %1)
  %4 = icmp sgt i32 %3, -1
  br i1 %4, label %5, label %6

  tail call void @bar()
  br label %6

  ret void
}

define i32 @blsr32_branch(i32 %x) {
; X86-LABEL: blsr32_branch:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    .cfi_offset %esi, -8
; X86-NEXT:    blsrl {{[0-9]+}}(%esp), %esi
; X86-NEXT:    jne .LBB53_2
; X86-NEXT:  # %bb.1:
; X86-NEXT:    calll bar
; X86-NEXT:  .LBB53_2:
; X86-NEXT:    movl %esi, %eax
; X86-NEXT:    popl %esi
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl
;
; X64-LABEL: blsr32_branch:
; X64:       # %bb.0:
; X64-NEXT:    pushq %rbx
; X64-NEXT:    .cfi_def_cfa_offset 16
; X64-NEXT:    .cfi_offset %rbx, -16
; X64-NEXT:    blsrl %edi, %ebx
; X64-NEXT:    jne .LBB53_2
; X64-NEXT:  # %bb.1:
; X64-NEXT:    callq bar
; X64-NEXT:  .LBB53_2:
; X64-NEXT:    movl %ebx, %eax
; X64-NEXT:    popq %rbx
; X64-NEXT:    .cfi_def_cfa_offset 8
; X64-NEXT:    retq
;
; EGPR-LABEL: blsr32_branch:
; EGPR:       # %bb.0:
; EGPR-NEXT:    pushq %rbx # encoding: [0x53]
; EGPR-NEXT:    .cfi_def_cfa_offset 16
; EGPR-NEXT:    .cfi_offset %rbx, -16
; EGPR-NEXT:    blsrl %edi, %ebx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x60,0xf3,0xcf]
; EGPR-NEXT:    jne .LBB53_2 # encoding: [0x75,A]
; EGPR-NEXT:    # fixup A - offset: 1, value: .LBB53_2-1, kind: FK_PCRel_1
; EGPR-NEXT:  # %bb.1:
; EGPR-NEXT:    callq bar # encoding: [0xe8,A,A,A,A]
; EGPR-NEXT:    # fixup A - offset: 1, value: bar-4, kind: reloc_branch_4byte_pcrel
; EGPR-NEXT:  .LBB53_2:
; EGPR-NEXT:    movl %ebx, %eax # encoding: [0x89,0xd8]
; EGPR-NEXT:    popq %rbx # encoding: [0x5b]
; EGPR-NEXT:    .cfi_def_cfa_offset 8
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp = sub i32 %x, 1
  %tmp2 = and i32 %x, %tmp
  %cmp = icmp eq i32 %tmp2, 0
  br i1 %cmp, label %1, label %2

  tail call void @bar()
  br label %2
  ret i32 %tmp2
}

define i64 @blsr64_branch(i64 %x) {
; X86-LABEL: blsr64_branch:
; X86:       # %bb.0:
; X86-NEXT:    pushl %edi
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    pushl %esi
; X86-NEXT:    .cfi_def_cfa_offset 12
; X86-NEXT:    .cfi_offset %esi, -12
; X86-NEXT:    .cfi_offset %edi, -8
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl %eax, %esi
; X86-NEXT:    addl $-1, %esi
; X86-NEXT:    movl %ecx, %edi
; X86-NEXT:    adcl $-1, %edi
; X86-NEXT:    andl %eax, %esi
; X86-NEXT:    andl %ecx, %edi
; X86-NEXT:    movl %esi, %eax
; X86-NEXT:    orl %edi, %eax
; X86-NEXT:    jne .LBB54_2
; X86-NEXT:  # %bb.1:
; X86-NEXT:    calll bar
; X86-NEXT:  .LBB54_2:
; X86-NEXT:    movl %esi, %eax
; X86-NEXT:    movl %edi, %edx
; X86-NEXT:    popl %esi
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    popl %edi
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl
;
; X64-LABEL: blsr64_branch:
; X64:       # %bb.0:
; X64-NEXT:    pushq %rbx
; X64-NEXT:    .cfi_def_cfa_offset 16
; X64-NEXT:    .cfi_offset %rbx, -16
; X64-NEXT:    blsrq %rdi, %rbx
; X64-NEXT:    jne .LBB54_2
; X64-NEXT:  # %bb.1:
; X64-NEXT:    callq bar
; X64-NEXT:  .LBB54_2:
; X64-NEXT:    movq %rbx, %rax
; X64-NEXT:    popq %rbx
; X64-NEXT:    .cfi_def_cfa_offset 8
; X64-NEXT:    retq
;
; EGPR-LABEL: blsr64_branch:
; EGPR:       # %bb.0:
; EGPR-NEXT:    pushq %rbx # encoding: [0x53]
; EGPR-NEXT:    .cfi_def_cfa_offset 16
; EGPR-NEXT:    .cfi_offset %rbx, -16
; EGPR-NEXT:    blsrq %rdi, %rbx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xe0,0xf3,0xcf]
; EGPR-NEXT:    jne .LBB54_2 # encoding: [0x75,A]
; EGPR-NEXT:    # fixup A - offset: 1, value: .LBB54_2-1, kind: FK_PCRel_1
; EGPR-NEXT:  # %bb.1:
; EGPR-NEXT:    callq bar # encoding: [0xe8,A,A,A,A]
; EGPR-NEXT:    # fixup A - offset: 1, value: bar-4, kind: reloc_branch_4byte_pcrel
; EGPR-NEXT:  .LBB54_2:
; EGPR-NEXT:    movq %rbx, %rax # encoding: [0x48,0x89,0xd8]
; EGPR-NEXT:    popq %rbx # encoding: [0x5b]
; EGPR-NEXT:    .cfi_def_cfa_offset 8
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp = sub i64 %x, 1
  %tmp2 = and i64 %x, %tmp
  %cmp = icmp eq i64 %tmp2, 0
  br i1 %cmp, label %1, label %2

  tail call void @bar()
  br label %2
  ret i64 %tmp2
}

define i32 @blsi32_branch(i32 %x) {
; X86-LABEL: blsi32_branch:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    .cfi_offset %esi, -8
; X86-NEXT:    blsil {{[0-9]+}}(%esp), %esi
; X86-NEXT:    jne .LBB55_2
; X86-NEXT:  # %bb.1:
; X86-NEXT:    calll bar
; X86-NEXT:  .LBB55_2:
; X86-NEXT:    movl %esi, %eax
; X86-NEXT:    popl %esi
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl
;
; X64-LABEL: blsi32_branch:
; X64:       # %bb.0:
; X64-NEXT:    pushq %rbx
; X64-NEXT:    .cfi_def_cfa_offset 16
; X64-NEXT:    .cfi_offset %rbx, -16
; X64-NEXT:    blsil %edi, %ebx
; X64-NEXT:    jne .LBB55_2
; X64-NEXT:  # %bb.1:
; X64-NEXT:    callq bar
; X64-NEXT:  .LBB55_2:
; X64-NEXT:    movl %ebx, %eax
; X64-NEXT:    popq %rbx
; X64-NEXT:    .cfi_def_cfa_offset 8
; X64-NEXT:    retq
;
; EGPR-LABEL: blsi32_branch:
; EGPR:       # %bb.0:
; EGPR-NEXT:    pushq %rbx # encoding: [0x53]
; EGPR-NEXT:    .cfi_def_cfa_offset 16
; EGPR-NEXT:    .cfi_offset %rbx, -16
; EGPR-NEXT:    blsil %edi, %ebx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x60,0xf3,0xdf]
; EGPR-NEXT:    jne .LBB55_2 # encoding: [0x75,A]
; EGPR-NEXT:    # fixup A - offset: 1, value: .LBB55_2-1, kind: FK_PCRel_1
; EGPR-NEXT:  # %bb.1:
; EGPR-NEXT:    callq bar # encoding: [0xe8,A,A,A,A]
; EGPR-NEXT:    # fixup A - offset: 1, value: bar-4, kind: reloc_branch_4byte_pcrel
; EGPR-NEXT:  .LBB55_2:
; EGPR-NEXT:    movl %ebx, %eax # encoding: [0x89,0xd8]
; EGPR-NEXT:    popq %rbx # encoding: [0x5b]
; EGPR-NEXT:    .cfi_def_cfa_offset 8
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp = sub i32 0, %x
  %tmp2 = and i32 %x, %tmp
  %cmp = icmp eq i32 %tmp2, 0
  br i1 %cmp, label %1, label %2

  tail call void @bar()
  br label %2
  ret i32 %tmp2
}

define i64 @blsi64_branch(i64 %x) {
; X86-LABEL: blsi64_branch:
; X86:       # %bb.0:
; X86-NEXT:    pushl %edi
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    pushl %esi
; X86-NEXT:    .cfi_def_cfa_offset 12
; X86-NEXT:    .cfi_offset %esi, -12
; X86-NEXT:    .cfi_offset %edi, -8
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    xorl %esi, %esi
; X86-NEXT:    movl %eax, %edi
; X86-NEXT:    negl %edi
; X86-NEXT:    sbbl %ecx, %esi
; X86-NEXT:    andl %ecx, %esi
; X86-NEXT:    andl %eax, %edi
; X86-NEXT:    movl %edi, %eax
; X86-NEXT:    orl %esi, %eax
; X86-NEXT:    jne .LBB56_2
; X86-NEXT:  # %bb.1:
; X86-NEXT:    calll bar
; X86-NEXT:  .LBB56_2:
; X86-NEXT:    movl %edi, %eax
; X86-NEXT:    movl %esi, %edx
; X86-NEXT:    popl %esi
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    popl %edi
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl
;
; X64-LABEL: blsi64_branch:
; X64:       # %bb.0:
; X64-NEXT:    pushq %rbx
; X64-NEXT:    .cfi_def_cfa_offset 16
; X64-NEXT:    .cfi_offset %rbx, -16
; X64-NEXT:    blsiq %rdi, %rbx
; X64-NEXT:    jne .LBB56_2
; X64-NEXT:  # %bb.1:
; X64-NEXT:    callq bar
; X64-NEXT:  .LBB56_2:
; X64-NEXT:    movq %rbx, %rax
; X64-NEXT:    popq %rbx
; X64-NEXT:    .cfi_def_cfa_offset 8
; X64-NEXT:    retq
;
; EGPR-LABEL: blsi64_branch:
; EGPR:       # %bb.0:
; EGPR-NEXT:    pushq %rbx # encoding: [0x53]
; EGPR-NEXT:    .cfi_def_cfa_offset 16
; EGPR-NEXT:    .cfi_offset %rbx, -16
; EGPR-NEXT:    blsiq %rdi, %rbx # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xe0,0xf3,0xdf]
; EGPR-NEXT:    jne .LBB56_2 # encoding: [0x75,A]
; EGPR-NEXT:    # fixup A - offset: 1, value: .LBB56_2-1, kind: FK_PCRel_1
; EGPR-NEXT:  # %bb.1:
; EGPR-NEXT:    callq bar # encoding: [0xe8,A,A,A,A]
; EGPR-NEXT:    # fixup A - offset: 1, value: bar-4, kind: reloc_branch_4byte_pcrel
; EGPR-NEXT:  .LBB56_2:
; EGPR-NEXT:    movq %rbx, %rax # encoding: [0x48,0x89,0xd8]
; EGPR-NEXT:    popq %rbx # encoding: [0x5b]
; EGPR-NEXT:    .cfi_def_cfa_offset 8
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp = sub i64 0, %x
  %tmp2 = and i64 %x, %tmp
  %cmp = icmp eq i64 %tmp2, 0
  br i1 %cmp, label %1, label %2

  tail call void @bar()
  br label %2
  ret i64 %tmp2
}

declare dso_local void @bar()

define void @pr42118_i32(i32 %x) {
; X86-LABEL: pr42118_i32:
; X86:       # %bb.0:
; X86-NEXT:    blsrl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    je bar # TAILCALL
; X86-NEXT:  # %bb.1:
; X86-NEXT:    retl
;
; X64-LABEL: pr42118_i32:
; X64:       # %bb.0:
; X64-NEXT:    blsrl %edi, %eax
; X64-NEXT:    je bar # TAILCALL
; X64-NEXT:  # %bb.1:
; X64-NEXT:    retq
;
; EGPR-LABEL: pr42118_i32:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsrl %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf3,0xcf]
; EGPR-NEXT:    testl %eax, %eax # encoding: [0x85,0xc0]
; EGPR-NEXT:    je bar # TAILCALL
; EGPR-NEXT:    # encoding: [0x74,A]
; EGPR-NEXT:    # fixup A - offset: 1, value: bar-1, kind: FK_PCRel_1
; EGPR-NEXT:  # %bb.1:
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp = sub i32 0, %x
  %tmp1 = and i32 %tmp, %x
  %cmp = icmp eq i32 %tmp1, %x
  br i1 %cmp, label %1, label %2

  tail call void @bar()
  br label %2

  ret void
}

define void @pr42118_i64(i64 %x) {
; X86-LABEL: pr42118_i64:
; X86:       # %bb.0:
; X86-NEXT:    pushl %esi
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    .cfi_offset %esi, -8
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl %eax, %edx
; X86-NEXT:    addl $-1, %edx
; X86-NEXT:    movl %ecx, %esi
; X86-NEXT:    adcl $-1, %esi
; X86-NEXT:    andl %eax, %edx
; X86-NEXT:    andl %ecx, %esi
; X86-NEXT:    orl %edx, %esi
; X86-NEXT:    jne .LBB58_1
; X86-NEXT:  # %bb.2:
; X86-NEXT:    popl %esi
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    jmp bar # TAILCALL
; X86-NEXT:  .LBB58_1:
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    popl %esi
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl
;
; X64-LABEL: pr42118_i64:
; X64:       # %bb.0:
; X64-NEXT:    blsrq %rdi, %rax
; X64-NEXT:    je bar # TAILCALL
; X64-NEXT:  # %bb.1:
; X64-NEXT:    retq
;
; EGPR-LABEL: pr42118_i64:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsrq %rdi, %rax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf8,0xf3,0xcf]
; EGPR-NEXT:    testq %rax, %rax # encoding: [0x48,0x85,0xc0]
; EGPR-NEXT:    je bar # TAILCALL
; EGPR-NEXT:    # encoding: [0x74,A]
; EGPR-NEXT:    # fixup A - offset: 1, value: bar-1, kind: FK_PCRel_1
; EGPR-NEXT:  # %bb.1:
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tmp = sub i64 0, %x
  %tmp1 = and i64 %tmp, %x
  %cmp = icmp eq i64 %tmp1, %x
  br i1 %cmp, label %1, label %2

  tail call void @bar()
  br label %2

  ret void
}

define i32 @blsi_cflag_32(i32 %x, i32 %y) nounwind {
; X86-LABEL: blsi_cflag_32:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    testl %eax, %eax
; X86-NEXT:    jne .LBB59_1
; X86-NEXT:  # %bb.2:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    retl
; X86-NEXT:  .LBB59_1:
; X86-NEXT:    blsil %eax, %eax
; X86-NEXT:    retl
;
; X64-LABEL: blsi_cflag_32:
; X64:       # %bb.0:
; X64-NEXT:    blsil %edi, %eax
; X64-NEXT:    cmovael %esi, %eax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsi_cflag_32:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsil %edi, %eax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0xf3,0xdf]
; EGPR-NEXT:    testl %edi, %edi # encoding: [0x85,0xff]
; EGPR-NEXT:    cmovel %esi, %eax # encoding: [0x0f,0x44,0xc6]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tobool = icmp eq i32 %x, 0
  %sub = sub nsw i32 0, %x
  %and = and i32 %sub, %x
  %cond = select i1 %tobool, i32 %y, i32 %and
  ret i32 %cond
}

define i64 @blsi_cflag_64(i64 %x, i64 %y) nounwind {
; X86-LABEL: blsi_cflag_64:
; X86:       # %bb.0:
; X86-NEXT:    pushl %edi
; X86-NEXT:    pushl %esi
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
; X86-NEXT:    xorl %edx, %edx
; X86-NEXT:    movl %ecx, %eax
; X86-NEXT:    negl %eax
; X86-NEXT:    sbbl %esi, %edx
; X86-NEXT:    movl %ecx, %edi
; X86-NEXT:    orl %esi, %edi
; X86-NEXT:    jne .LBB60_1
; X86-NEXT:  # %bb.2:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    jmp .LBB60_3
; X86-NEXT:  .LBB60_1:
; X86-NEXT:    andl %esi, %edx
; X86-NEXT:    andl %ecx, %eax
; X86-NEXT:  .LBB60_3:
; X86-NEXT:    popl %esi
; X86-NEXT:    popl %edi
; X86-NEXT:    retl
;
; X64-LABEL: blsi_cflag_64:
; X64:       # %bb.0:
; X64-NEXT:    blsiq %rdi, %rax
; X64-NEXT:    cmovaeq %rsi, %rax
; X64-NEXT:    retq
;
; EGPR-LABEL: blsi_cflag_64:
; EGPR:       # %bb.0:
; EGPR-NEXT:    blsiq %rdi, %rax # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf8,0xf3,0xdf]
; EGPR-NEXT:    testq %rdi, %rdi # encoding: [0x48,0x85,0xff]
; EGPR-NEXT:    cmoveq %rsi, %rax # encoding: [0x48,0x0f,0x44,0xc6]
; EGPR-NEXT:    retq # encoding: [0xc3]
  %tobool = icmp eq i64 %x, 0
  %sub = sub nsw i64 0, %x
  %and = and i64 %sub, %x
  %cond = select i1 %tobool, i64 %y, i64 %and
  ret i64 %cond
}