; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i686-unknown -mattr=+sse4.2 | FileCheck %s --check-prefix=X86
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse4.2 | FileCheck %s --check-prefix=X64

;
; AND/XOR/OR i24 as v3i8
;

define i24 @and_i24_as_v3i8(i24 %a, i24 %b) nounwind {
; X86-LABEL: and_i24_as_v3i8:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    retl
;
; X64-LABEL: and_i24_as_v3i8:
; X64:       # %bb.0:
; X64-NEXT:    movl %edi, %eax
; X64-NEXT:    andl %esi, %eax
; X64-NEXT:    retq
  %1 = bitcast i24 %a to <3 x i8>
  %2 = bitcast i24 %b to <3 x i8>
  %3 = and <3 x i8> %1, %2
  %4 = bitcast <3 x i8> %3 to i24
  ret i24 %4
}

define i24 @xor_i24_as_v3i8(i24 %a, i24 %b) nounwind {
; X86-LABEL: xor_i24_as_v3i8:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    xorl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    retl
;
; X64-LABEL: xor_i24_as_v3i8:
; X64:       # %bb.0:
; X64-NEXT:    movl %edi, %eax
; X64-NEXT:    xorl %esi, %eax
; X64-NEXT:    retq
  %1 = bitcast i24 %a to <3 x i8>
  %2 = bitcast i24 %b to <3 x i8>
  %3 = xor <3 x i8> %1, %2
  %4 = bitcast <3 x i8> %3 to i24
  ret i24 %4
}

define i24 @or_i24_as_v3i8(i24 %a, i24 %b) nounwind {
; X86-LABEL: or_i24_as_v3i8:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    orl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    retl
;
; X64-LABEL: or_i24_as_v3i8:
; X64:       # %bb.0:
; X64-NEXT:    movl %edi, %eax
; X64-NEXT:    orl %esi, %eax
; X64-NEXT:    retq
  %1 = bitcast i24 %a to <3 x i8>
  %2 = bitcast i24 %b to <3 x i8>
  %3 = or <3 x i8> %1, %2
  %4 = bitcast <3 x i8> %3 to i24
  ret i24 %4
}

;
; AND/XOR/OR i24 as v8i3
;

define i24 @and_i24_as_v8i3(i24 %a, i24 %b) nounwind {
; X86-LABEL: and_i24_as_v8i3:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    retl
;
; X64-LABEL: and_i24_as_v8i3:
; X64:       # %bb.0:
; X64-NEXT:    movl %edi, %eax
; X64-NEXT:    andl %esi, %eax
; X64-NEXT:    retq
  %1 = bitcast i24 %a to <8 x i3>
  %2 = bitcast i24 %b to <8 x i3>
  %3 = and <8 x i3> %1, %2
  %4 = bitcast <8 x i3> %3 to i24
  ret i24 %4
}

define i24 @xor_i24_as_v8i3(i24 %a, i24 %b) nounwind {
; X86-LABEL: xor_i24_as_v8i3:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    xorl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    retl
;
; X64-LABEL: xor_i24_as_v8i3:
; X64:       # %bb.0:
; X64-NEXT:    movl %edi, %eax
; X64-NEXT:    xorl %esi, %eax
; X64-NEXT:    retq
  %1 = bitcast i24 %a to <8 x i3>
  %2 = bitcast i24 %b to <8 x i3>
  %3 = xor <8 x i3> %1, %2
  %4 = bitcast <8 x i3> %3 to i24
  ret i24 %4
}

define i24 @or_i24_as_v8i3(i24 %a, i24 %b) nounwind {
; X86-LABEL: or_i24_as_v8i3:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    orl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    retl
;
; X64-LABEL: or_i24_as_v8i3:
; X64:       # %bb.0:
; X64-NEXT:    movl %edi, %eax
; X64-NEXT:    orl %esi, %eax
; X64-NEXT:    retq
  %1 = bitcast i24 %a to <8 x i3>
  %2 = bitcast i24 %b to <8 x i3>
  %3 = or <8 x i3> %1, %2
  %4 = bitcast <8 x i3> %3 to i24
  ret i24 %4
}

;
; AND/XOR/OR v3i8 as i24
;

define <3 x i8> @and_v3i8_as_i24(<3 x i8> %a, <3 x i8> %b) nounwind {
; X86-LABEL: and_v3i8_as_i24:
; X86:       # %bb.0:
; X86-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X86-NEXT:    pinsrb $1, {{[0-9]+}}(%esp), %xmm0
; X86-NEXT:    pinsrb $2, {{[0-9]+}}(%esp), %xmm0
; X86-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
; X86-NEXT:    pinsrb $1, {{[0-9]+}}(%esp), %xmm1
; X86-NEXT:    pinsrb $2, {{[0-9]+}}(%esp), %xmm1
; X86-NEXT:    pand %xmm0, %xmm1
; X86-NEXT:    movd %xmm1, %eax
; X86-NEXT:    pextrb $1, %xmm1, %edx
; X86-NEXT:    pextrb $2, %xmm1, %ecx
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    # kill: def $dl killed $dl killed $edx
; X86-NEXT:    # kill: def $cl killed $cl killed $ecx
; X86-NEXT:    retl
;
; X64-LABEL: and_v3i8_as_i24:
; X64:       # %bb.0:
; X64-NEXT:    movd %ecx, %xmm0
; X64-NEXT:    pinsrb $1, %r8d, %xmm0
; X64-NEXT:    pinsrb $2, %r9d, %xmm0
; X64-NEXT:    movd %edi, %xmm1
; X64-NEXT:    pinsrb $1, %esi, %xmm1
; X64-NEXT:    pinsrb $2, %edx, %xmm1
; X64-NEXT:    pand %xmm0, %xmm1
; X64-NEXT:    movd %xmm1, %eax
; X64-NEXT:    pextrb $1, %xmm1, %edx
; X64-NEXT:    pextrb $2, %xmm1, %ecx
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    # kill: def $dl killed $dl killed $edx
; X64-NEXT:    # kill: def $cl killed $cl killed $ecx
; X64-NEXT:    retq
  %1 = bitcast <3 x i8> %a to i24
  %2 = bitcast <3 x i8> %b to i24
  %3 = and i24 %1, %2
  %4 = bitcast i24 %3 to <3 x i8>
  ret <3 x i8>  %4
}

define <3 x i8> @xor_v3i8_as_i24(<3 x i8> %a, <3 x i8> %b) nounwind {
; X86-LABEL: xor_v3i8_as_i24:
; X86:       # %bb.0:
; X86-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X86-NEXT:    pinsrb $1, {{[0-9]+}}(%esp), %xmm0
; X86-NEXT:    pinsrb $2, {{[0-9]+}}(%esp), %xmm0
; X86-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
; X86-NEXT:    pinsrb $1, {{[0-9]+}}(%esp), %xmm1
; X86-NEXT:    pinsrb $2, {{[0-9]+}}(%esp), %xmm1
; X86-NEXT:    pxor %xmm0, %xmm1
; X86-NEXT:    movd %xmm1, %eax
; X86-NEXT:    pextrb $1, %xmm1, %edx
; X86-NEXT:    pextrb $2, %xmm1, %ecx
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    # kill: def $dl killed $dl killed $edx
; X86-NEXT:    # kill: def $cl killed $cl killed $ecx
; X86-NEXT:    retl
;
; X64-LABEL: xor_v3i8_as_i24:
; X64:       # %bb.0:
; X64-NEXT:    movd %ecx, %xmm0
; X64-NEXT:    pinsrb $1, %r8d, %xmm0
; X64-NEXT:    pinsrb $2, %r9d, %xmm0
; X64-NEXT:    movd %edi, %xmm1
; X64-NEXT:    pinsrb $1, %esi, %xmm1
; X64-NEXT:    pinsrb $2, %edx, %xmm1
; X64-NEXT:    pxor %xmm0, %xmm1
; X64-NEXT:    movd %xmm1, %eax
; X64-NEXT:    pextrb $1, %xmm1, %edx
; X64-NEXT:    pextrb $2, %xmm1, %ecx
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    # kill: def $dl killed $dl killed $edx
; X64-NEXT:    # kill: def $cl killed $cl killed $ecx
; X64-NEXT:    retq
  %1 = bitcast <3 x i8> %a to i24
  %2 = bitcast <3 x i8> %b to i24
  %3 = xor i24 %1, %2
  %4 = bitcast i24 %3 to <3 x i8>
  ret <3 x i8>  %4
}

define <3 x i8> @or_v3i8_as_i24(<3 x i8> %a, <3 x i8> %b) nounwind {
; X86-LABEL: or_v3i8_as_i24:
; X86:       # %bb.0:
; X86-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X86-NEXT:    pinsrb $1, {{[0-9]+}}(%esp), %xmm0
; X86-NEXT:    pinsrb $2, {{[0-9]+}}(%esp), %xmm0
; X86-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
; X86-NEXT:    pinsrb $1, {{[0-9]+}}(%esp), %xmm1
; X86-NEXT:    pinsrb $2, {{[0-9]+}}(%esp), %xmm1
; X86-NEXT:    por %xmm0, %xmm1
; X86-NEXT:    movd %xmm1, %eax
; X86-NEXT:    pextrb $1, %xmm1, %edx
; X86-NEXT:    pextrb $2, %xmm1, %ecx
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    # kill: def $dl killed $dl killed $edx
; X86-NEXT:    # kill: def $cl killed $cl killed $ecx
; X86-NEXT:    retl
;
; X64-LABEL: or_v3i8_as_i24:
; X64:       # %bb.0:
; X64-NEXT:    movd %ecx, %xmm0
; X64-NEXT:    pinsrb $1, %r8d, %xmm0
; X64-NEXT:    pinsrb $2, %r9d, %xmm0
; X64-NEXT:    movd %edi, %xmm1
; X64-NEXT:    pinsrb $1, %esi, %xmm1
; X64-NEXT:    pinsrb $2, %edx, %xmm1
; X64-NEXT:    por %xmm0, %xmm1
; X64-NEXT:    movd %xmm1, %eax
; X64-NEXT:    pextrb $1, %xmm1, %edx
; X64-NEXT:    pextrb $2, %xmm1, %ecx
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    # kill: def $dl killed $dl killed $edx
; X64-NEXT:    # kill: def $cl killed $cl killed $ecx
; X64-NEXT:    retq
  %1 = bitcast <3 x i8> %a to i24
  %2 = bitcast <3 x i8> %b to i24
  %3 = or i24 %1, %2
  %4 = bitcast i24 %3 to <3 x i8>
  ret <3 x i8>  %4
}

;
; AND/XOR/OR v8i3 as i24
;

define <8 x i3> @and_v8i3_as_i24(<8 x i3> %a, <8 x i3> %b) nounwind {
; X86-LABEL: and_v8i3_as_i24:
; X86:       # %bb.0:
; X86-NEXT:    andps %xmm1, %xmm0
; X86-NEXT:    retl
;
; X64-LABEL: and_v8i3_as_i24:
; X64:       # %bb.0:
; X64-NEXT:    andps %xmm1, %xmm0
; X64-NEXT:    retq
  %1 = bitcast <8 x i3> %a to i24
  %2 = bitcast <8 x i3> %b to i24
  %3 = and i24 %1, %2
  %4 = bitcast i24 %3 to <8 x i3>
  ret <8 x i3>  %4
}

define <8 x i3> @xor_v8i3_as_i24(<8 x i3> %a, <8 x i3> %b) nounwind {
; X86-LABEL: xor_v8i3_as_i24:
; X86:       # %bb.0:
; X86-NEXT:    xorps %xmm1, %xmm0
; X86-NEXT:    retl
;
; X64-LABEL: xor_v8i3_as_i24:
; X64:       # %bb.0:
; X64-NEXT:    xorps %xmm1, %xmm0
; X64-NEXT:    retq
  %1 = bitcast <8 x i3> %a to i24
  %2 = bitcast <8 x i3> %b to i24
  %3 = xor i24 %1, %2
  %4 = bitcast i24 %3 to <8 x i3>
  ret <8 x i3>  %4
}

define <8 x i3> @or_v8i3_as_i24(<8 x i3> %a, <8 x i3> %b) nounwind {
; X86-LABEL: or_v8i3_as_i24:
; X86:       # %bb.0:
; X86-NEXT:    orps %xmm1, %xmm0
; X86-NEXT:    retl
;
; X64-LABEL: or_v8i3_as_i24:
; X64:       # %bb.0:
; X64-NEXT:    orps %xmm1, %xmm0
; X64-NEXT:    retq
  %1 = bitcast <8 x i3> %a to i24
  %2 = bitcast <8 x i3> %b to i24
  %3 = or i24 %1, %2
  %4 = bitcast i24 %3 to <8 x i3>
  ret <8 x i3>  %4
}