; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=SSE2
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX2
; RUN: llc < %s -mtriple=i686-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=SSE2
; RUN: llc < %s -mtriple=i686-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX2

;------------------------------ 32-bit shuffles -------------------------------;

define <4 x i32> @shuffle_i32_of_shl_i16(<8 x i16> %x, <8 x i16> %y) nounwind {
; SSE2-LABEL: shuffle_i32_of_shl_i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psllw $15, %xmm0
; SSE2-NEXT:    psllw $15, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,2],xmm0[1,0]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i32_of_shl_i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsllw $15, %xmm0, %xmm0
; AVX2-NEXT:    vpsllw $15, %xmm1, %xmm1
; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[3,2],xmm0[1,0]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <8 x i16> @llvm.x86.sse2.pslli.w(<8 x i16> %x, i32 15)
  %i2 = tail call <8 x i16> @llvm.x86.sse2.pslli.w(<8 x i16> %y, i32 15)
  %i3 = bitcast <8 x i16> %i1 to <4 x i32>
  %i4 = bitcast <8 x i16> %i2 to <4 x i32>
  %i5 = shufflevector <4 x i32> %i3, <4 x i32> %i4, <4 x i32> <i32 7, i32 6, i32 1, i32 0>
  ret <4 x i32> %i5
}
define <4 x i32> @shuffle_i32_of_lshr_i16(<8 x i16> %x, <8 x i16> %y) nounwind {
; SSE2-LABEL: shuffle_i32_of_lshr_i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrlw $15, %xmm0
; SSE2-NEXT:    psrlw $15, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,2],xmm0[1,0]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i32_of_lshr_i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsrlw $15, %xmm0, %xmm0
; AVX2-NEXT:    vpsrlw $15, %xmm1, %xmm1
; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[3,2],xmm0[1,0]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16> %x, i32 15)
  %i2 = tail call <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16> %y, i32 15)
  %i3 = bitcast <8 x i16> %i1 to <4 x i32>
  %i4 = bitcast <8 x i16> %i2 to <4 x i32>
  %i5 = shufflevector <4 x i32> %i3, <4 x i32> %i4, <4 x i32> <i32 7, i32 6, i32 1, i32 0>
  ret <4 x i32> %i5
}
define <4 x i32> @shuffle_i32_of_ashr_i16(<8 x i16> %x, <8 x i16> %y) nounwind {
; SSE2-LABEL: shuffle_i32_of_ashr_i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psraw $15, %xmm0
; SSE2-NEXT:    psraw $15, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,2],xmm0[1,0]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i32_of_ashr_i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsraw $15, %xmm0, %xmm0
; AVX2-NEXT:    vpsraw $15, %xmm1, %xmm1
; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[3,2],xmm0[1,0]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16> %x, i32 15)
  %i2 = tail call <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16> %y, i32 15)
  %i3 = bitcast <8 x i16> %i1 to <4 x i32>
  %i4 = bitcast <8 x i16> %i2 to <4 x i32>
  %i5 = shufflevector <4 x i32> %i3, <4 x i32> %i4, <4 x i32> <i32 7, i32 6, i32 1, i32 0>
  ret <4 x i32> %i5
}

define <4 x i32> @shuffle_i32_of_shl_i32(<4 x i32> %x, <4 x i32> %y) nounwind {
; SSE2-LABEL: shuffle_i32_of_shl_i32:
; SSE2:       # %bb.0:
; SSE2-NEXT:    pslld $31, %xmm0
; SSE2-NEXT:    pslld $31, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,2],xmm0[1,0]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i32_of_shl_i32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpslld $31, %xmm0, %xmm0
; AVX2-NEXT:    vpslld $31, %xmm1, %xmm1
; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[3,2],xmm0[1,0]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <4 x i32> @llvm.x86.sse2.pslli.d(<4 x i32> %x, i32 31)
  %i2 = tail call <4 x i32> @llvm.x86.sse2.pslli.d(<4 x i32> %y, i32 31)
  %i3 = shufflevector <4 x i32> %i1, <4 x i32> %i2, <4 x i32> <i32 7, i32 6, i32 1, i32 0>
  ret <4 x i32> %i3
}
define <4 x i32> @shuffle_i32_of_lshr_i32(<4 x i32> %x, <4 x i32> %y) nounwind {
; SSE2-LABEL: shuffle_i32_of_lshr_i32:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrld $31, %xmm0
; SSE2-NEXT:    psrld $31, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,2],xmm0[1,0]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i32_of_lshr_i32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsrld $31, %xmm0, %xmm0
; AVX2-NEXT:    vpsrld $31, %xmm1, %xmm1
; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[3,2],xmm0[1,0]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <4 x i32> @llvm.x86.sse2.psrli.d(<4 x i32> %x, i32 31)
  %i2 = tail call <4 x i32> @llvm.x86.sse2.psrli.d(<4 x i32> %y, i32 31)
  %i3 = shufflevector <4 x i32> %i1, <4 x i32> %i2, <4 x i32> <i32 7, i32 6, i32 1, i32 0>
  ret <4 x i32> %i3
}
define <4 x i32> @shuffle_i32_of_ashr_i32(<4 x i32> %x, <4 x i32> %y) nounwind {
; SSE2-LABEL: shuffle_i32_of_ashr_i32:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrad $31, %xmm0
; SSE2-NEXT:    psrad $31, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,2],xmm0[1,0]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i32_of_ashr_i32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsrad $31, %xmm0, %xmm0
; AVX2-NEXT:    vpsrad $31, %xmm1, %xmm1
; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[3,2],xmm0[1,0]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32> %x, i32 31)
  %i2 = tail call <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32> %y, i32 31)
  %i3 = shufflevector <4 x i32> %i1, <4 x i32> %i2, <4 x i32> <i32 7, i32 6, i32 1, i32 0>
  ret <4 x i32> %i3
}

define <4 x i32> @shuffle_i32_of_shl_i64(<2 x i64> %x, <2 x i64> %y) nounwind {
; SSE2-LABEL: shuffle_i32_of_shl_i64:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psllq $63, %xmm0
; SSE2-NEXT:    psllq $63, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,2],xmm0[1,0]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i32_of_shl_i64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsllq $63, %xmm0, %xmm0
; AVX2-NEXT:    vpsllq $63, %xmm1, %xmm1
; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[3,2],xmm0[1,0]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <2 x i64> @llvm.x86.sse2.pslli.q(<2 x i64> %x, i32 63)
  %i2 = tail call <2 x i64> @llvm.x86.sse2.pslli.q(<2 x i64> %y, i32 63)
  %i3 = bitcast <2 x i64> %i1 to <4 x i32>
  %i4 = bitcast <2 x i64> %i2 to <4 x i32>
  %i5 = shufflevector <4 x i32> %i3, <4 x i32> %i4, <4 x i32> <i32 7, i32 6, i32 1, i32 0>
  ret <4 x i32> %i5
}
define <4 x i32> @shuffle_i32_of_lshr_i64(<2 x i64> %x, <2 x i64> %y) nounwind {
; SSE2-LABEL: shuffle_i32_of_lshr_i64:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrlq $63, %xmm0
; SSE2-NEXT:    psrlq $63, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,2],xmm0[1,0]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i32_of_lshr_i64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsrlq $63, %xmm0, %xmm0
; AVX2-NEXT:    vpsrlq $63, %xmm1, %xmm1
; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm1[3,2],xmm0[1,0]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <2 x i64> @llvm.x86.sse2.psrli.q(<2 x i64> %x, i32 63)
  %i2 = tail call <2 x i64> @llvm.x86.sse2.psrli.q(<2 x i64> %y, i32 63)
  %i3 = bitcast <2 x i64> %i1 to <4 x i32>
  %i4 = bitcast <2 x i64> %i2 to <4 x i32>
  %i5 = shufflevector <4 x i32> %i3, <4 x i32> %i4, <4 x i32> <i32 7, i32 6, i32 1, i32 0>
  ret <4 x i32> %i5
}

;------------------------------ 64-bit shuffles -------------------------------;

define <2 x i64> @shuffle_i64_of_shl_i16(<8 x i16> %x, <8 x i16> %y) nounwind {
; SSE2-LABEL: shuffle_i64_of_shl_i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psllw $15, %xmm0
; SSE2-NEXT:    psllw $15, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,1]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i64_of_shl_i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsllw $15, %xmm0, %xmm0
; AVX2-NEXT:    vpsllw $15, %xmm1, %xmm1
; AVX2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <8 x i16> @llvm.x86.sse2.pslli.w(<8 x i16> %x, i32 15)
  %i2 = tail call <8 x i16> @llvm.x86.sse2.pslli.w(<8 x i16> %y, i32 15)
  %i3 = bitcast <8 x i16> %i1 to <2 x i64>
  %i4 = bitcast <8 x i16> %i2 to <2 x i64>
  %i5 = shufflevector <2 x i64> %i3, <2 x i64> %i4, <2 x i32> <i32 3, i32 0>
  ret <2 x i64> %i5
}
define <2 x i64> @shuffle_i64_of_lshr_i16(<8 x i16> %x, <8 x i16> %y) nounwind {
; SSE2-LABEL: shuffle_i64_of_lshr_i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrlw $15, %xmm0
; SSE2-NEXT:    psrlw $15, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,1]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i64_of_lshr_i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsrlw $15, %xmm0, %xmm0
; AVX2-NEXT:    vpsrlw $15, %xmm1, %xmm1
; AVX2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16> %x, i32 15)
  %i2 = tail call <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16> %y, i32 15)
  %i3 = bitcast <8 x i16> %i1 to <2 x i64>
  %i4 = bitcast <8 x i16> %i2 to <2 x i64>
  %i5 = shufflevector <2 x i64> %i3, <2 x i64> %i4, <2 x i32> <i32 3, i32 0>
  ret <2 x i64> %i5
}
define <2 x i64> @shuffle_i64_of_ashr_i16(<8 x i16> %x, <8 x i16> %y) nounwind {
; SSE2-LABEL: shuffle_i64_of_ashr_i16:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psraw $15, %xmm0
; SSE2-NEXT:    psraw $15, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,1]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i64_of_ashr_i16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsraw $15, %xmm0, %xmm0
; AVX2-NEXT:    vpsraw $15, %xmm1, %xmm1
; AVX2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16> %x, i32 15)
  %i2 = tail call <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16> %y, i32 15)
  %i3 = bitcast <8 x i16> %i1 to <2 x i64>
  %i4 = bitcast <8 x i16> %i2 to <2 x i64>
  %i5 = shufflevector <2 x i64> %i3, <2 x i64> %i4, <2 x i32> <i32 3, i32 0>
  ret <2 x i64> %i5
}

define <2 x i64> @shuffle_i64_of_shl_i32(<4 x i32> %x, <4 x i32> %y) nounwind {
; SSE2-LABEL: shuffle_i64_of_shl_i32:
; SSE2:       # %bb.0:
; SSE2-NEXT:    pslld $31, %xmm0
; SSE2-NEXT:    pslld $31, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,1]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i64_of_shl_i32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpslld $31, %xmm0, %xmm0
; AVX2-NEXT:    vpslld $31, %xmm1, %xmm1
; AVX2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <4 x i32> @llvm.x86.sse2.pslli.d(<4 x i32> %x, i32 31)
  %i2 = tail call <4 x i32> @llvm.x86.sse2.pslli.d(<4 x i32> %y, i32 31)
  %i3 = bitcast <4 x i32> %i1 to <2 x i64>
  %i4 = bitcast <4 x i32> %i2 to <2 x i64>
  %i5 = shufflevector <2 x i64> %i3, <2 x i64> %i4, <2 x i32> <i32 3, i32 0>
  ret <2 x i64> %i5
}
define <2 x i64> @shuffle_i64_of_lshr_i32(<4 x i32> %x, <4 x i32> %y) nounwind {
; SSE2-LABEL: shuffle_i64_of_lshr_i32:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrld $31, %xmm0
; SSE2-NEXT:    psrld $31, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,1]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i64_of_lshr_i32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsrld $31, %xmm0, %xmm0
; AVX2-NEXT:    vpsrld $31, %xmm1, %xmm1
; AVX2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <4 x i32> @llvm.x86.sse2.psrli.d(<4 x i32> %x, i32 31)
  %i2 = tail call <4 x i32> @llvm.x86.sse2.psrli.d(<4 x i32> %y, i32 31)
  %i3 = bitcast <4 x i32> %i1 to <2 x i64>
  %i4 = bitcast <4 x i32> %i2 to <2 x i64>
  %i5 = shufflevector <2 x i64> %i3, <2 x i64> %i4, <2 x i32> <i32 3, i32 0>
  ret <2 x i64> %i5
}
define <2 x i64> @shuffle_i64_of_ashr_i32(<4 x i32> %x, <4 x i32> %y) nounwind {
; SSE2-LABEL: shuffle_i64_of_ashr_i32:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrad $31, %xmm0
; SSE2-NEXT:    psrad $31, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,1]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i64_of_ashr_i32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsrad $31, %xmm0, %xmm0
; AVX2-NEXT:    vpsrad $31, %xmm1, %xmm1
; AVX2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32> %x, i32 31)
  %i2 = tail call <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32> %y, i32 31)
  %i3 = bitcast <4 x i32> %i1 to <2 x i64>
  %i4 = bitcast <4 x i32> %i2 to <2 x i64>
  %i5 = shufflevector <2 x i64> %i3, <2 x i64> %i4, <2 x i32> <i32 3, i32 0>
  ret <2 x i64> %i5
}

define <2 x i64> @shuffle_i64_of_shl_i64(<2 x i64> %x, <2 x i64> %y) nounwind {
; SSE2-LABEL: shuffle_i64_of_shl_i64:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psllq $63, %xmm0
; SSE2-NEXT:    psllq $63, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,1]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i64_of_shl_i64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsllq $63, %xmm0, %xmm0
; AVX2-NEXT:    vpsllq $63, %xmm1, %xmm1
; AVX2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <2 x i64> @llvm.x86.sse2.pslli.q(<2 x i64> %x, i32 63)
  %i2 = tail call <2 x i64> @llvm.x86.sse2.pslli.q(<2 x i64> %y, i32 63)
  %i3 = bitcast <2 x i64> %i1 to <2 x i64>
  %i4 = bitcast <2 x i64> %i2 to <2 x i64>
  %i5 = shufflevector <2 x i64> %i3, <2 x i64> %i4, <2 x i32> <i32 3, i32 0>
  ret <2 x i64> %i5
}
define <2 x i64> @shuffle_i64_of_lshr_i64(<2 x i64> %x, <2 x i64> %y) nounwind {
; SSE2-LABEL: shuffle_i64_of_lshr_i64:
; SSE2:       # %bb.0:
; SSE2-NEXT:    psrlq $63, %xmm0
; SSE2-NEXT:    psrlq $63, %xmm1
; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,1]
; SSE2-NEXT:    movaps %xmm1, %xmm0
; SSE2-NEXT:    ret{{[l|q]}}
;
; AVX2-LABEL: shuffle_i64_of_lshr_i64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vpsrlq $63, %xmm0, %xmm0
; AVX2-NEXT:    vpsrlq $63, %xmm1, %xmm1
; AVX2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; AVX2-NEXT:    ret{{[l|q]}}
  %i1 = tail call <2 x i64> @llvm.x86.sse2.psrli.q(<2 x i64> %x, i32 63)
  %i2 = tail call <2 x i64> @llvm.x86.sse2.psrli.q(<2 x i64> %y, i32 63)
  %i3 = bitcast <2 x i64> %i1 to <2 x i64>
  %i4 = bitcast <2 x i64> %i2 to <2 x i64>
  %i5 = shufflevector <2 x i64> %i3, <2 x i64> %i4, <2 x i32> <i32 3, i32 0>
  ret <2 x i64> %i5
}