; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse2 | FileCheck %s --check-prefixes=SSE
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx  | FileCheck %s --check-prefixes=AVX,AVX1
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2,AVX2-SLOW
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX2,AVX2-FAST-ALL
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX2,AVX2-FAST-PERLANE
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512bw,+avx512vl | FileCheck %s --check-prefixes=AVX512

; These patterns are produced by LoopVectorizer for interleaved stores.

define void @vf2(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %out.vec) nounwind {
; SSE-LABEL: vf2:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,7,6,7]
; SSE-NEXT:    movq %xmm0, (%rcx)
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
; SSE-NEXT:    movd %xmm0, 8(%rcx)
; SSE-NEXT:    retq
;
; AVX-LABEL: vf2:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovdqa (%rdi), %xmm0
; AVX-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; AVX-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,2,3,4,5,12,13,6,7,4,5,6,7]
; AVX-NEXT:    vpextrd $2, %xmm0, 8(%rcx)
; AVX-NEXT:    vmovq %xmm0, (%rcx)
; AVX-NEXT:    retq
;
; AVX512-LABEL: vf2:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm1 = [0,2,8,1,3,9,1,9]
; AVX512-NEXT:    vpermi2w (%rdx), %xmm0, %xmm1
; AVX512-NEXT:    vpextrd $2, %xmm1, 8(%rcx)
; AVX512-NEXT:    vmovq %xmm1, (%rcx)
; AVX512-NEXT:    retq
  %in.vec0 = load <2 x i16>, ptr %in.vecptr0, align 32
  %in.vec1 = load <2 x i16>, ptr %in.vecptr1, align 32
  %in.vec2 = load <2 x i16>, ptr %in.vecptr2, align 32

  %concat01 = shufflevector <2 x i16> %in.vec0, <2 x i16> %in.vec1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %concat2u = shufflevector <2 x i16> %in.vec2, <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
  %concat012 = shufflevector <4 x i16> %concat01, <4 x i16> %concat2u, <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
  %interleaved.vec = shufflevector <6 x i16> %concat012, <6 x i16> poison, <6 x i32> <i32 0, i32 2, i32 4, i32 1, i32 3, i32 5>

  store <6 x i16> %interleaved.vec, ptr %out.vec, align 32

  ret void
}

define void @vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %out.vec) nounwind {
; SSE-LABEL: vf4:
; SSE:       # %bb.0:
; SSE-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
; SSE-NEXT:    movq {{.*#+}} xmm1 = mem[0],zero
; SSE-NEXT:    movq {{.*#+}} xmm2 = mem[0],zero
; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm1[0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,0,0]
; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,0,65535,65535,0,65535,65535]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,2,2,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,6,4]
; SSE-NEXT:    pand %xmm3, %xmm4
; SSE-NEXT:    pandn %xmm1, %xmm3
; SSE-NEXT:    por %xmm4, %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,65535,65535,0,65535,65535,65535,65535]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,1,3,4,5,6,7]
; SSE-NEXT:    pand %xmm1, %xmm2
; SSE-NEXT:    pandn %xmm0, %xmm1
; SSE-NEXT:    por %xmm2, %xmm1
; SSE-NEXT:    movq %xmm1, 16(%rcx)
; SSE-NEXT:    movdqa %xmm3, (%rcx)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf4:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX1-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX1-NEXT:    vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,3,1,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,1,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1,2],xmm3[3],xmm2[4,5,6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,8,9,u,u,2,3,10,11,u,u,4,5,12,13]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3,4],xmm0[5],xmm1[6,7]
; AVX1-NEXT:    vmovdqa %xmm0, (%rcx)
; AVX1-NEXT:    vmovq %xmm2, 16(%rcx)
; AVX1-NEXT:    retq
;
; AVX2-LABEL: vf4:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX2-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX2-NEXT:    vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,8,9],zero,zero,ymm0[2,3,10,11],zero,zero,ymm0[4,5,12,13,20,21],zero,zero,zero,zero,ymm0[22,23],zero,zero,zero,zero,zero,zero,zero,zero
; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,ymm0[0,1],zero,zero,zero,zero,ymm0[2,3],zero,zero,zero,zero,zero,zero,ymm0[22,23,30,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX2-NEXT:    vpor %ymm0, %ymm1, %ymm0
; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT:    vmovq %xmm1, 16(%rcx)
; AVX2-NEXT:    vmovdqa %xmm0, (%rcx)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX512-LABEL: vf4:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX512-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX512-NEXT:    vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX512-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm1 = <0,4,8,1,5,9,2,6,10,3,7,11,u,u,u,u>
; AVX512-NEXT:    vpermw %ymm0, %ymm1, %ymm0
; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT:    vmovq %xmm1, 16(%rcx)
; AVX512-NEXT:    vmovdqa %xmm0, (%rcx)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %in.vec0 = load <4 x i16>, ptr %in.vecptr0, align 32
  %in.vec1 = load <4 x i16>, ptr %in.vecptr1, align 32
  %in.vec2 = load <4 x i16>, ptr %in.vecptr2, align 32

  %concat01 = shufflevector <4 x i16> %in.vec0, <4 x i16> %in.vec1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %concat2u = shufflevector <4 x i16> %in.vec2, <4 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
  %concat012 = shufflevector <8 x i16> %concat01, <8 x i16> %concat2u, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
  %interleaved.vec = shufflevector <12 x i16> %concat012, <12 x i16> poison, <12 x i32> <i32 0, i32 4, i32 8, i32 1, i32 5, i32 9, i32 2, i32 6, i32 10, i32 3, i32 7, i32 11>

  store <12 x i16> %interleaved.vec, ptr %out.vec, align 32

  ret void
}

define void @vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %out.vec) nounwind {
; SSE-LABEL: vf8:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    movdqa (%rsi), %xmm2
; SSE-NEXT:    movdqa (%rdx), %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,0,0,0]
; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,0,65535,65535]
; SSE-NEXT:    movdqa %xmm4, %xmm5
; SSE-NEXT:    pandn %xmm1, %xmm5
; SSE-NEXT:    movdqa %xmm0, %xmm1
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,4,5]
; SSE-NEXT:    pand %xmm4, %xmm1
; SSE-NEXT:    por %xmm5, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[1,1,2,2]
; SSE-NEXT:    pand %xmm4, %xmm5
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm2[3,3,3,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,4,4]
; SSE-NEXT:    pandn %xmm6, %xmm4
; SSE-NEXT:    por %xmm5, %xmm4
; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [0,65535,65535,0,65535,65535,0,65535]
; SSE-NEXT:    pand %xmm5, %xmm4
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[1,1,2,2]
; SSE-NEXT:    pandn %xmm6, %xmm5
; SSE-NEXT:    por %xmm4, %xmm5
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,2,3,3]
; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,0,65535,65535,0,65535,65535,0]
; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,1,3,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,1,0,4,5,6,7]
; SSE-NEXT:    pand %xmm4, %xmm0
; SSE-NEXT:    pandn %xmm3, %xmm4
; SSE-NEXT:    por %xmm0, %xmm4
; SSE-NEXT:    movdqa %xmm4, 32(%rcx)
; SSE-NEXT:    movdqa %xmm5, 16(%rcx)
; SSE-NEXT:    movdqa %xmm1, (%rcx)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf8:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vmovdqa (%rdi), %xmm0
; AVX1-NEXT:    vmovdqa (%rsi), %xmm1
; AVX1-NEXT:    vmovdqa (%rdx), %xmm2
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,2,2]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm1[3,3,3,3,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3,4],xmm4[5],xmm3[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[1,1,2,2]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1,2],xmm4[3],xmm3[4,5],xmm4[6],xmm3[7]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[4,5,u,u,10,11,8,9,u,u,14,15,12,13,u,u]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm2[2,2,3,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm5[1],xmm4[2,3],xmm5[4],xmm4[5,6],xmm5[7]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,u,u,4,5,6,7,u,u,8,9,10,11]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[0,0,0,0]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4],xmm1[5],xmm0[6,7]
; AVX1-NEXT:    vmovdqa %xmm0, (%rcx)
; AVX1-NEXT:    vmovdqa %xmm4, 32(%rcx)
; AVX1-NEXT:    vmovdqa %xmm3, 16(%rcx)
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf8:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-SLOW-NEXT:    vmovdqa (%rsi), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa (%rdx), %xmm2
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm3
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm4 = ymm3[0,1,u,u,6,7,2,3,u,u,8,9,4,5,u,u,16,17,u,u,22,23,18,19,u,u,24,25,20,21,u,u]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[2,3,0,1]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm3 = ymm3[u,u,0,1,u,u,u,u,2,3,u,u,u,u,4,5,u,u,22,23,u,u,u,u,24,25,u,u,u,u,26,27]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7],ymm4[8],ymm3[9],ymm4[10,11],ymm3[12],ymm4[13,14],ymm3[15]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm4 = <u,0,0,u,1,1,u,2>
; AVX2-SLOW-NEXT:    vpermd %ymm2, %ymm4, %ymm4
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm5 = [255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255]
; AVX2-SLOW-NEXT:    vpblendvb %ymm5, %ymm3, %ymm4, %ymm3
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,u,u,10,11,8,9,u,u,14,15,12,13,u,u]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[2,2,3,3]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6],xmm1[7]
; AVX2-SLOW-NEXT:    vmovdqa %xmm0, 32(%rcx)
; AVX2-SLOW-NEXT:    vmovdqa %ymm3, (%rcx)
; AVX2-SLOW-NEXT:    vzeroupper
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-ALL-LABEL: vf8:
; AVX2-FAST-ALL:       # %bb.0:
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rsi), %xmm1
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdx), %xmm2
; AVX2-FAST-ALL-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm3
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm4 = <u,0,0,u,1,1,u,2>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm2, %ymm4, %ymm4
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm5 = [0,4,1,5,1,5,2,6]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm3, %ymm5, %ymm3
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm3 = ymm3[0,1,4,5,u,u,2,3,6,7,u,u,8,9,12,13,u,u,18,19,22,23,u,u,24,25,28,29,u,u,26,27]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm5 = [255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm5, %ymm3, %ymm4, %ymm3
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,u,u,10,11,8,9,u,u,14,15,12,13,u,u]
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[2,2,3,3]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6],xmm1[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm0, 32(%rcx)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm3, (%rcx)
; AVX2-FAST-ALL-NEXT:    vzeroupper
; AVX2-FAST-ALL-NEXT:    retq
;
; AVX2-FAST-PERLANE-LABEL: vf8:
; AVX2-FAST-PERLANE:       # %bb.0:
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rsi), %xmm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdx), %xmm2
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm4 = ymm3[0,1,u,u,6,7,2,3,u,u,8,9,4,5,u,u,16,17,u,u,22,23,18,19,u,u,24,25,20,21,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm3 = ymm3[u,u,0,1,u,u,u,u,2,3,u,u,u,u,4,5,u,u,22,23,u,u,u,u,24,25,u,u,u,u,26,27]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7],ymm4[8],ymm3[9],ymm4[10,11],ymm3[12],ymm4[13,14],ymm3[15]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm4 = <u,0,0,u,1,1,u,2>
; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm2, %ymm4, %ymm4
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm5 = [255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm5, %ymm3, %ymm4, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,5,u,u,10,11,8,9,u,u,14,15,12,13,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[2,2,3,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6],xmm1[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm0, 32(%rcx)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm3, (%rcx)
; AVX2-FAST-PERLANE-NEXT:    vzeroupper
; AVX2-FAST-PERLANE-NEXT:    retq
;
; AVX512-LABEL: vf8:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-NEXT:    vmovdqa (%rdx), %xmm1
; AVX512-NEXT:    vinserti128 $1, (%rsi), %ymm0, %ymm0
; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm1 = <0,8,16,1,9,17,2,10,18,3,11,19,4,12,20,5,13,21,6,14,22,7,15,23,u,u,u,u,u,u,u,u>
; AVX512-NEXT:    vpermw %zmm0, %zmm1, %zmm0
; AVX512-NEXT:    vextracti32x4 $2, %zmm0, 32(%rcx)
; AVX512-NEXT:    vmovdqa %ymm0, (%rcx)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %in.vec0 = load <8 x i16>, ptr %in.vecptr0, align 32
  %in.vec1 = load <8 x i16>, ptr %in.vecptr1, align 32
  %in.vec2 = load <8 x i16>, ptr %in.vecptr2, align 32

  %concat01 = shufflevector <8 x i16> %in.vec0, <8 x i16> %in.vec1, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
  %concat2u = shufflevector <8 x i16> %in.vec2, <8 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
  %concat012 = shufflevector <16 x i16> %concat01, <16 x i16> %concat2u, <24 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
  %interleaved.vec = shufflevector <24 x i16> %concat012, <24 x i16> poison, <24 x i32> <i32 0, i32 8, i32 16, i32 1, i32 9, i32 17, i32 2, i32 10, i32 18, i32 3, i32 11, i32 19, i32 4, i32 12, i32 20, i32 5, i32 13, i32 21, i32 6, i32 14, i32 22, i32 7, i32 15, i32 23>

  store <24 x i16> %interleaved.vec, ptr %out.vec, align 32

  ret void
}

define void @vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %out.vec) nounwind {
; SSE-LABEL: vf16:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa (%rdi), %xmm8
; SSE-NEXT:    movdqa 16(%rdi), %xmm10
; SSE-NEXT:    movdqa (%rsi), %xmm12
; SSE-NEXT:    movdqa 16(%rsi), %xmm7
; SSE-NEXT:    movdqa (%rdx), %xmm9
; SSE-NEXT:    movdqa 16(%rdx), %xmm11
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[1,1,2,2]
; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,0,65535,65535,0,65535,65535]
; SSE-NEXT:    pand %xmm2, %xmm1
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm7[3,3,3,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
; SSE-NEXT:    movdqa %xmm2, %xmm0
; SSE-NEXT:    pandn %xmm3, %xmm0
; SSE-NEXT:    por %xmm1, %xmm0
; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [0,65535,65535,0,65535,65535,0,65535]
; SSE-NEXT:    pand %xmm3, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[1,1,2,2]
; SSE-NEXT:    movdqa %xmm3, %xmm5
; SSE-NEXT:    pandn %xmm1, %xmm5
; SSE-NEXT:    por %xmm0, %xmm5
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,0,0,0]
; SSE-NEXT:    movdqa %xmm2, %xmm4
; SSE-NEXT:    pandn %xmm0, %xmm4
; SSE-NEXT:    movdqa %xmm10, %xmm0
; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,7,5,4,5]
; SSE-NEXT:    pand %xmm2, %xmm1
; SSE-NEXT:    por %xmm4, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,2,2]
; SSE-NEXT:    pand %xmm2, %xmm0
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm12[3,3,3,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
; SSE-NEXT:    movdqa %xmm2, %xmm6
; SSE-NEXT:    pandn %xmm4, %xmm6
; SSE-NEXT:    por %xmm0, %xmm6
; SSE-NEXT:    pand %xmm3, %xmm6
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[1,1,2,2]
; SSE-NEXT:    pandn %xmm0, %xmm3
; SSE-NEXT:    por %xmm6, %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[0,0,0,0]
; SSE-NEXT:    movdqa %xmm8, %xmm4
; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm12[0],xmm4[1],xmm12[1],xmm4[2],xmm12[2],xmm4[3],xmm12[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,5,4,5]
; SSE-NEXT:    pand %xmm2, %xmm4
; SSE-NEXT:    pandn %xmm0, %xmm2
; SSE-NEXT:    por %xmm4, %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,0,65535,65535,0,65535,65535,0]
; SSE-NEXT:    movdqa %xmm4, %xmm6
; SSE-NEXT:    pandn %xmm0, %xmm6
; SSE-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm10[4],xmm7[5],xmm10[5],xmm7[6],xmm10[6],xmm7[7],xmm10[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,1,3,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,1,0,4,5,6,7]
; SSE-NEXT:    pand %xmm4, %xmm0
; SSE-NEXT:    por %xmm6, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm9[2,2,3,3]
; SSE-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm8[4],xmm12[5],xmm8[5],xmm12[6],xmm8[6],xmm12[7],xmm8[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm12[2,1,3,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[2,1,1,0,4,5,6,7]
; SSE-NEXT:    pand %xmm4, %xmm7
; SSE-NEXT:    pandn %xmm6, %xmm4
; SSE-NEXT:    por %xmm7, %xmm4
; SSE-NEXT:    movdqa %xmm4, 32(%rcx)
; SSE-NEXT:    movdqa %xmm0, 80(%rcx)
; SSE-NEXT:    movdqa %xmm2, (%rcx)
; SSE-NEXT:    movdqa %xmm3, 16(%rcx)
; SSE-NEXT:    movdqa %xmm1, 48(%rcx)
; SSE-NEXT:    movdqa %xmm5, 64(%rcx)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf16:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vmovdqa (%rdi), %xmm0
; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,2,2]
; AVX1-NEXT:    vmovdqa (%rsi), %xmm3
; AVX1-NEXT:    vmovdqa 16(%rsi), %xmm4
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm4[3,3,3,3,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,4,4]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm5[2],xmm2[3,4],xmm5[5],xmm2[6,7]
; AVX1-NEXT:    vmovdqa (%rdx), %xmm5
; AVX1-NEXT:    vmovdqa 16(%rdx), %xmm6
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm6[1,1,2,2]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm8 = xmm7[0],xmm2[1,2],xmm7[3],xmm2[4,5],xmm7[6],xmm2[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm0[1,1,2,2]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm3[3,3,3,3,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,4,4]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm7[0,1],xmm2[2],xmm7[3,4],xmm2[5],xmm7[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm5[1,1,2,2]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm10 = xmm7[0],xmm2[1,2],xmm7[3],xmm2[4,5],xmm7[6],xmm2[7]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm7 = xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7]
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,5,10,11,10,11,8,9,8,9,14,15,12,13,14,15]
; AVX1-NEXT:    vpshufb %xmm9, %xmm7, %xmm7
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm6[2,2,3,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm11 = xmm7[0],xmm2[1],xmm7[2,3],xmm2[4],xmm7[5,6],xmm2[7]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm7 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm12 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
; AVX1-NEXT:    vpshufb %xmm12, %xmm7, %xmm7
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm5[0,0,0,0]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm7[0,1],xmm2[2],xmm7[3,4],xmm2[5],xmm7[6,7]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
; AVX1-NEXT:    vpshufb %xmm12, %xmm1, %xmm1
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm6[0,0,0,0]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm4[2],xmm1[3,4],xmm4[5],xmm1[6,7]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
; AVX1-NEXT:    vpshufb %xmm9, %xmm0, %xmm0
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm5[2,2,3,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1],xmm0[2,3],xmm3[4],xmm0[5,6],xmm3[7]
; AVX1-NEXT:    vmovdqa %xmm0, 32(%rcx)
; AVX1-NEXT:    vmovdqa %xmm1, 48(%rcx)
; AVX1-NEXT:    vmovdqa %xmm2, (%rcx)
; AVX1-NEXT:    vmovdqa %xmm11, 80(%rcx)
; AVX1-NEXT:    vmovdqa %xmm10, 16(%rcx)
; AVX1-NEXT:    vmovdqa %xmm8, 64(%rcx)
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf16:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    vmovdqa (%rsi), %ymm1
; AVX2-SLOW-NEXT:    vmovdqa (%rdx), %ymm0
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %xmm2
; AVX2-SLOW-NEXT:    vmovdqa 16(%rdi), %xmm3
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,2,2]
; AVX2-SLOW-NEXT:    vmovdqa (%rsi), %xmm5
; AVX2-SLOW-NEXT:    vmovdqa 16(%rsi), %xmm6
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm6[3,3,3,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,4,4]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3,4],xmm7[5],xmm4[6,7]
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm3 = xmm6[4],xmm3[4],xmm6[5],xmm3[5],xmm6[6],xmm3[6],xmm6[7],xmm3[7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[4,5,10,11,10,11,8,9,8,9,14,15,12,13,14,15]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm4, %ymm3
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm4 = <5,5,u,6,6,u,7,7>
; AVX2-SLOW-NEXT:    vpermd %ymm0, %ymm4, %ymm4
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm6 = [0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0]
; AVX2-SLOW-NEXT:    vpblendvb %ymm6, %ymm3, %ymm4, %ymm3
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[1,1,2,2]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm5[3,3,3,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,4,4]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm4, %ymm2, %ymm2
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm4 = <u,0,0,u,1,1,u,2>
; AVX2-SLOW-NEXT:    vpermd %ymm0, %ymm4, %ymm4
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm5 = [255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255]
; AVX2-SLOW-NEXT:    vpblendvb %ymm5, %ymm2, %ymm4, %ymm2
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm4 = <u,3,3,u,4,4,u,5>
; AVX2-SLOW-NEXT:    vpermd (%rdi), %ymm4, %ymm4
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[10,11,0,1,12,13,12,13,2,3,2,3,14,15,4,5,26,27,16,17,28,29,28,29,18,19,18,19,30,31,20,21]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm5 = <255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255>
; AVX2-SLOW-NEXT:    vpblendvb %ymm5, %ymm1, %ymm4, %ymm1
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm4 = <2,u,3,3,u,4,4,u>
; AVX2-SLOW-NEXT:    vpermd %ymm0, %ymm4, %ymm0
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm4 = [255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255]
; AVX2-SLOW-NEXT:    vpblendvb %ymm4, %ymm1, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vmovdqa %ymm0, 32(%rcx)
; AVX2-SLOW-NEXT:    vmovdqa %ymm2, (%rcx)
; AVX2-SLOW-NEXT:    vmovdqa %ymm3, 64(%rcx)
; AVX2-SLOW-NEXT:    vzeroupper
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-ALL-LABEL: vf16:
; AVX2-FAST-ALL:       # %bb.0:
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rsi), %ymm8
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdx), %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rsi), %xmm2
; AVX2-FAST-ALL-NEXT:    vmovdqa 16(%rsi), %xmm3
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm4 = <u,u,u,u,6,7,u,u,u,u,8,9,u,u,u,u>
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm4, %xmm3, %xmm5
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdi), %xmm6
; AVX2-FAST-ALL-NEXT:    vmovdqa 16(%rdi), %xmm7
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm7[1,1,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm5[2],xmm1[3,4],xmm5[5],xmm1[6,7]
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} xmm3 = xmm3[4],xmm7[4],xmm3[5],xmm7[5],xmm3[6],xmm7[6],xmm3[7],xmm7[7]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[4,5,10,11,10,11,8,9,8,9,14,15,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vinserti128 $1, %xmm3, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm3 = <5,5,u,6,6,u,7,7>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm3, %ymm3
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm5 = [0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm5, %ymm1, %ymm3, %ymm1
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm4, %xmm2, %xmm3
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm4 = xmm6[1,1,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm3[2],xmm4[3,4],xmm3[5],xmm4[6,7]
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1],xmm6[2],xmm2[2],xmm6[3],xmm2[3]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
; AVX2-FAST-ALL-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm3 = <u,0,0,u,1,1,u,2>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm3, %ymm3
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm4 = [255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm4, %ymm2, %ymm3, %ymm2
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm3 = <u,3,3,u,4,4,u,5>
; AVX2-FAST-ALL-NEXT:    vpermd (%rdi), %ymm3, %ymm3
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm4 = ymm8[10,11,0,1,12,13,12,13,2,3,2,3,14,15,4,5,26,27,16,17,28,29,28,29,18,19,18,19,30,31,20,21]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm5 = <255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255>
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm5, %ymm4, %ymm3, %ymm3
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm4 = <2,u,3,3,u,4,4,u>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm4, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm4 = [255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm4, %ymm3, %ymm0, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm0, 32(%rcx)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm2, (%rcx)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm1, 64(%rcx)
; AVX2-FAST-ALL-NEXT:    vzeroupper
; AVX2-FAST-ALL-NEXT:    retq
;
; AVX2-FAST-PERLANE-LABEL: vf16:
; AVX2-FAST-PERLANE:       # %bb.0:
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rsi), %ymm8
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdx), %ymm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rsi), %xmm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 16(%rsi), %xmm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm4 = <u,u,u,u,6,7,u,u,u,u,8,9,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm4, %xmm3, %xmm5
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdi), %xmm6
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 16(%rdi), %xmm7
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm1 = xmm7[1,1,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm5[2],xmm1[3,4],xmm5[5],xmm1[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} xmm3 = xmm3[4],xmm7[4],xmm3[5],xmm7[5],xmm3[6],xmm7[6],xmm3[7],xmm7[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[4,5,10,11,10,11,8,9,8,9,14,15,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm3, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm3 = <5,5,u,6,6,u,7,7>
; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm0, %ymm3, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm5 = [0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm5, %ymm1, %ymm3, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm4, %xmm2, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm4 = xmm6[1,1,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm3[2],xmm4[3,4],xmm3[5],xmm4[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1],xmm6[2],xmm2[2],xmm6[3],xmm2[3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm3 = <u,0,0,u,1,1,u,2>
; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm0, %ymm3, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm4 = [255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm4, %ymm2, %ymm3, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm3 = <u,3,3,u,4,4,u,5>
; AVX2-FAST-PERLANE-NEXT:    vpermd (%rdi), %ymm3, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm4 = ymm8[10,11,0,1,12,13,12,13,2,3,2,3,14,15,4,5,26,27,16,17,28,29,28,29,18,19,18,19,30,31,20,21]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm5 = <255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255>
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm5, %ymm4, %ymm3, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm4 = <2,u,3,3,u,4,4,u>
; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm0, %ymm4, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm4 = [255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm4, %ymm3, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm0, 32(%rcx)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm2, (%rcx)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm1, 64(%rcx)
; AVX2-FAST-PERLANE-NEXT:    vzeroupper
; AVX2-FAST-PERLANE-NEXT:    retq
;
; AVX512-LABEL: vf16:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa (%rdi), %ymm0
; AVX512-NEXT:    vmovdqa (%rdx), %ymm1
; AVX512-NEXT:    vinserti64x4 $1, (%rsi), %zmm0, %zmm0
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm2 = [42,11,27,43,12,28,44,13,29,45,14,30,46,15,31,47]
; AVX512-NEXT:    vpermi2w %zmm1, %zmm0, %zmm2
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,16,32,1,17,33,2,18,34,3,19,35,4,20,36,5,21,37,6,22,38,7,23,39,8,24,40,9,25,41,10,26]
; AVX512-NEXT:    vpermi2w %zmm1, %zmm0, %zmm3
; AVX512-NEXT:    vmovdqu64 %zmm3, (%rcx)
; AVX512-NEXT:    vmovdqa %ymm2, 64(%rcx)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %in.vec0 = load <16 x i16>, ptr %in.vecptr0, align 32
  %in.vec1 = load <16 x i16>, ptr %in.vecptr1, align 32
  %in.vec2 = load <16 x i16>, ptr %in.vecptr2, align 32

  %concat01 = shufflevector <16 x i16> %in.vec0, <16 x i16> %in.vec1, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
  %concat2u = shufflevector <16 x i16> %in.vec2, <16 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
  %concat012 = shufflevector <32 x i16> %concat01, <32 x i16> %concat2u, <48 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
  %interleaved.vec = shufflevector <48 x i16> %concat012, <48 x i16> poison, <48 x i32> <i32 0, i32 16, i32 32, i32 1, i32 17, i32 33, i32 2, i32 18, i32 34, i32 3, i32 19, i32 35, i32 4, i32 20, i32 36, i32 5, i32 21, i32 37, i32 6, i32 22, i32 38, i32 7, i32 23, i32 39, i32 8, i32 24, i32 40, i32 9, i32 25, i32 41, i32 10, i32 26, i32 42, i32 11, i32 27, i32 43, i32 12, i32 28, i32 44, i32 13, i32 29, i32 45, i32 14, i32 30, i32 46, i32 15, i32 31, i32 47>

  store <48 x i16> %interleaved.vec, ptr %out.vec, align 32

  ret void
}

define void @vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %out.vec) nounwind {
; SSE-LABEL: vf32:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa 16(%rdi), %xmm8
; SSE-NEXT:    movdqa 32(%rdi), %xmm11
; SSE-NEXT:    movdqa 48(%rdi), %xmm0
; SSE-NEXT:    movdqa 16(%rsi), %xmm7
; SSE-NEXT:    movdqa 32(%rsi), %xmm6
; SSE-NEXT:    movdqa 48(%rsi), %xmm10
; SSE-NEXT:    movdqa 32(%rdx), %xmm9
; SSE-NEXT:    movdqa 48(%rdx), %xmm4
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,2]
; SSE-NEXT:    movdqa %xmm0, %xmm12
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,0,65535,65535,0,65535,65535]
; SSE-NEXT:    pand %xmm5, %xmm1
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm10[3,3,3,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,4,4]
; SSE-NEXT:    movdqa %xmm5, %xmm3
; SSE-NEXT:    pandn %xmm2, %xmm3
; SSE-NEXT:    por %xmm1, %xmm3
; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [0,65535,65535,0,65535,65535,0,65535]
; SSE-NEXT:    pand %xmm2, %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,2,2]
; SSE-NEXT:    movdqa %xmm2, %xmm0
; SSE-NEXT:    pandn %xmm1, %xmm0
; SSE-NEXT:    por %xmm3, %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,0,0,0]
; SSE-NEXT:    movdqa %xmm5, %xmm3
; SSE-NEXT:    pandn %xmm1, %xmm3
; SSE-NEXT:    movdqa %xmm12, %xmm1
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1],xmm1[2],xmm10[2],xmm1[3],xmm10[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm1[0,1,2,3,7,5,4,5]
; SSE-NEXT:    pand %xmm5, %xmm15
; SSE-NEXT:    por %xmm3, %xmm15
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[1,1,2,2]
; SSE-NEXT:    pand %xmm5, %xmm1
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm6[3,3,3,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
; SSE-NEXT:    movdqa %xmm5, %xmm4
; SSE-NEXT:    pandn %xmm3, %xmm4
; SSE-NEXT:    por %xmm1, %xmm4
; SSE-NEXT:    pand %xmm2, %xmm4
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[1,1,2,2]
; SSE-NEXT:    movdqa %xmm2, %xmm13
; SSE-NEXT:    pandn %xmm1, %xmm13
; SSE-NEXT:    por %xmm4, %xmm13
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,0,0,0]
; SSE-NEXT:    movdqa %xmm5, %xmm3
; SSE-NEXT:    pandn %xmm1, %xmm3
; SSE-NEXT:    movdqa %xmm11, %xmm1
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm1[0,1,2,3,7,5,4,5]
; SSE-NEXT:    pand %xmm5, %xmm9
; SSE-NEXT:    por %xmm3, %xmm9
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm7[3,3,3,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
; SSE-NEXT:    movdqa %xmm5, %xmm3
; SSE-NEXT:    pandn %xmm1, %xmm3
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[1,1,2,2]
; SSE-NEXT:    pand %xmm5, %xmm1
; SSE-NEXT:    por %xmm1, %xmm3
; SSE-NEXT:    movdqa 16(%rdx), %xmm11
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[1,1,2,2]
; SSE-NEXT:    movdqa %xmm2, %xmm14
; SSE-NEXT:    pandn %xmm1, %xmm14
; SSE-NEXT:    pand %xmm2, %xmm3
; SSE-NEXT:    por %xmm3, %xmm14
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,0,0,0]
; SSE-NEXT:    movdqa %xmm5, %xmm3
; SSE-NEXT:    pandn %xmm1, %xmm3
; SSE-NEXT:    movdqa %xmm8, %xmm1
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm1[0,1,2,3,7,5,4,5]
; SSE-NEXT:    pand %xmm5, %xmm4
; SSE-NEXT:    por %xmm3, %xmm4
; SSE-NEXT:    movdqa (%rsi), %xmm1
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm1[3,3,3,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
; SSE-NEXT:    movdqa %xmm5, %xmm0
; SSE-NEXT:    pandn %xmm3, %xmm0
; SSE-NEXT:    movdqa (%rdi), %xmm8
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[1,1,2,2]
; SSE-NEXT:    pand %xmm5, %xmm3
; SSE-NEXT:    por %xmm3, %xmm0
; SSE-NEXT:    pand %xmm2, %xmm0
; SSE-NEXT:    movdqa (%rdx), %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm3[1,1,2,2]
; SSE-NEXT:    pandn %xmm12, %xmm2
; SSE-NEXT:    por %xmm0, %xmm2
; SSE-NEXT:    movdqa %xmm8, %xmm0
; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,4,5]
; SSE-NEXT:    pand %xmm5, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm3[0,0,0,0]
; SSE-NEXT:    pandn %xmm7, %xmm5
; SSE-NEXT:    por %xmm0, %xmm5
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,0,65535,65535,0,65535,65535,0]
; SSE-NEXT:    movdqa %xmm12, %xmm7
; SSE-NEXT:    pandn %xmm0, %xmm7
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; SSE-NEXT:    # xmm10 = xmm10[4],mem[4],xmm10[5],mem[5],xmm10[6],mem[6],xmm10[7],mem[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,1,3,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm0[2,1,1,0,4,5,6,7]
; SSE-NEXT:    pand %xmm12, %xmm10
; SSE-NEXT:    por %xmm7, %xmm10
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    movdqa %xmm12, %xmm7
; SSE-NEXT:    pandn %xmm0, %xmm7
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT:    # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,1,3,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,1,0,4,5,6,7]
; SSE-NEXT:    pand %xmm12, %xmm0
; SSE-NEXT:    por %xmm7, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[2,2,3,3]
; SSE-NEXT:    movdqa %xmm12, %xmm7
; SSE-NEXT:    pandn %xmm11, %xmm7
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT:    # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,3,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[2,1,1,0,4,5,6,7]
; SSE-NEXT:    pand %xmm12, %xmm6
; SSE-NEXT:    por %xmm7, %xmm6
; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,3,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,1,0,4,5,6,7]
; SSE-NEXT:    pand %xmm12, %xmm1
; SSE-NEXT:    pandn %xmm3, %xmm12
; SSE-NEXT:    por %xmm1, %xmm12
; SSE-NEXT:    movdqa %xmm12, 32(%rcx)
; SSE-NEXT:    movdqa %xmm6, 80(%rcx)
; SSE-NEXT:    movdqa %xmm0, 128(%rcx)
; SSE-NEXT:    movdqa %xmm10, 176(%rcx)
; SSE-NEXT:    movdqa %xmm5, (%rcx)
; SSE-NEXT:    movdqa %xmm2, 16(%rcx)
; SSE-NEXT:    movdqa %xmm4, 48(%rcx)
; SSE-NEXT:    movdqa %xmm14, 64(%rcx)
; SSE-NEXT:    movdqa %xmm9, 96(%rcx)
; SSE-NEXT:    movdqa %xmm13, 112(%rcx)
; SSE-NEXT:    movdqa %xmm15, 144(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 160(%rcx)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf32:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vmovdqa (%rdi), %xmm12
; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm15
; AVX1-NEXT:    vmovdqa 32(%rdi), %xmm9
; AVX1-NEXT:    vmovdqa 48(%rdi), %xmm4
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm4[1,1,2,2]
; AVX1-NEXT:    vmovdqa (%rsi), %xmm13
; AVX1-NEXT:    vmovdqa 16(%rsi), %xmm8
; AVX1-NEXT:    vmovdqa 32(%rsi), %xmm6
; AVX1-NEXT:    vmovdqa 48(%rsi), %xmm1
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm1[3,3,3,3,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,4,4]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7]
; AVX1-NEXT:    vmovdqa 16(%rdx), %xmm11
; AVX1-NEXT:    vmovdqa 32(%rdx), %xmm2
; AVX1-NEXT:    vmovdqa 48(%rdx), %xmm0
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm0[1,1,2,2]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm7[0],xmm3[1,2],xmm7[3],xmm3[4,5],xmm7[6],xmm3[7]
; AVX1-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm9[1,1,2,2]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm6[3,3,3,3,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,4,4]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm7[2],xmm3[3,4],xmm7[5],xmm3[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm2[1,1,2,2]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm7[0],xmm3[1,2],xmm7[3],xmm3[4,5],xmm7[6],xmm3[7]
; AVX1-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm15[1,1,2,2]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm8[3,3,3,3,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,4,4]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm7[2],xmm3[3,4],xmm7[5],xmm3[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm11[1,1,2,2]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm7[0],xmm3[1,2],xmm7[3],xmm3[4,5],xmm7[6],xmm3[7]
; AVX1-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm12[1,1,2,2]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm13[3,3,3,3,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,4,4]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm7 = xmm3[0,1],xmm7[2],xmm3[3,4],xmm7[5],xmm3[6,7]
; AVX1-NEXT:    vmovdqa (%rdx), %xmm3
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm3[1,1,2,2]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0],xmm7[1,2],xmm5[3],xmm7[4,5],xmm5[6],xmm7[7]
; AVX1-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm7 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm10 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
; AVX1-NEXT:    vpshufb %xmm10, %xmm7, %xmm7
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm0[0,0,0,0]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm7[0,1],xmm5[2],xmm7[3,4],xmm5[5],xmm7[6,7]
; AVX1-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm5 = xmm6[4],xmm9[4],xmm6[5],xmm9[5],xmm6[6],xmm9[6],xmm6[7],xmm9[7]
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm7 = [4,5,10,11,10,11,8,9,8,9,14,15,12,13,14,15]
; AVX1-NEXT:    vpshufb %xmm7, %xmm5, %xmm5
; AVX1-NEXT:    vpshufd {{.*#+}} xmm14 = xmm2[2,2,3,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0],xmm14[1],xmm5[2,3],xmm14[4],xmm5[5,6],xmm14[7]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
; AVX1-NEXT:    vpshufb %xmm7, %xmm1, %xmm1
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3],xmm0[4],xmm1[5,6],xmm0[7]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm9[0],xmm6[0],xmm9[1],xmm6[1],xmm9[2],xmm6[2],xmm9[3],xmm6[3]
; AVX1-NEXT:    vpshufb %xmm10, %xmm1, %xmm1
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm8[4],xmm15[4],xmm8[5],xmm15[5],xmm8[6],xmm15[6],xmm8[7],xmm15[7]
; AVX1-NEXT:    vpshufb %xmm7, %xmm2, %xmm2
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm11[2,2,3,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0],xmm4[1],xmm2[2,3],xmm4[4],xmm2[5,6],xmm4[7]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
; AVX1-NEXT:    vpshufb %xmm10, %xmm4, %xmm4
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm11[0,0,0,0]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm6[2],xmm4[3,4],xmm6[5],xmm4[6,7]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm6 = xmm13[4],xmm12[4],xmm13[5],xmm12[5],xmm13[6],xmm12[6],xmm13[7],xmm12[7]
; AVX1-NEXT:    vpshufb %xmm7, %xmm6, %xmm6
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm3[2,2,3,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0],xmm7[1],xmm6[2,3],xmm7[4],xmm6[5,6],xmm7[7]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm7 = xmm12[0],xmm13[0],xmm12[1],xmm13[1],xmm12[2],xmm13[2],xmm12[3],xmm13[3]
; AVX1-NEXT:    vpshufb %xmm10, %xmm7, %xmm7
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm7[0,1],xmm3[2],xmm7[3,4],xmm3[5],xmm7[6,7]
; AVX1-NEXT:    vmovdqa %xmm3, (%rcx)
; AVX1-NEXT:    vmovdqa %xmm6, 32(%rcx)
; AVX1-NEXT:    vmovdqa %xmm4, 48(%rcx)
; AVX1-NEXT:    vmovdqa %xmm2, 80(%rcx)
; AVX1-NEXT:    vmovdqa %xmm1, 96(%rcx)
; AVX1-NEXT:    vmovdqa %xmm0, 176(%rcx)
; AVX1-NEXT:    vmovdqa %xmm5, 128(%rcx)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 144(%rcx)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 16(%rcx)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 64(%rcx)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 112(%rcx)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 160(%rcx)
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf32:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    vmovdqa (%rsi), %ymm9
; AVX2-SLOW-NEXT:    vmovdqa 32(%rsi), %ymm11
; AVX2-SLOW-NEXT:    vmovdqa (%rdx), %ymm14
; AVX2-SLOW-NEXT:    vmovdqa 32(%rdx), %ymm2
; AVX2-SLOW-NEXT:    vmovdqa 48(%rdi), %xmm3
; AVX2-SLOW-NEXT:    vmovdqa 48(%rsi), %xmm5
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm6 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm12 = [4,5,10,11,10,11,8,9,8,9,14,15,12,13,14,15]
; AVX2-SLOW-NEXT:    vpshufb %xmm12, %xmm6, %xmm6
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[1,1,2,2]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm5[3,3,3,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,4,4]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm5[2],xmm3[3,4],xmm5[5],xmm3[6,7]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm6, %ymm3, %ymm3
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm13 = <5,5,u,6,6,u,7,7>
; AVX2-SLOW-NEXT:    vpermd %ymm2, %ymm13, %ymm6
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm8 = [0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0]
; AVX2-SLOW-NEXT:    vpblendvb %ymm8, %ymm3, %ymm6, %ymm10
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %xmm6
; AVX2-SLOW-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX2-SLOW-NEXT:    vmovdqa (%rsi), %xmm4
; AVX2-SLOW-NEXT:    vmovdqa 16(%rsi), %xmm7
; AVX2-SLOW-NEXT:    vmovdqa 32(%rsi), %xmm5
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm7[4],xmm1[4],xmm7[5],xmm1[5],xmm7[6],xmm1[6],xmm7[7],xmm1[7]
; AVX2-SLOW-NEXT:    vpshufb %xmm12, %xmm0, %xmm0
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,2,2]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm7[3,3,3,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,4,4]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm7[2],xmm1[3,4],xmm7[5],xmm1[6,7]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
; AVX2-SLOW-NEXT:    vpermd %ymm14, %ymm13, %ymm1
; AVX2-SLOW-NEXT:    vpblendvb %ymm8, %ymm0, %ymm1, %ymm12
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm6[1,1,2,2]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm4[3,3,3,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,4,4]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm7[2],xmm0[3,4],xmm7[5],xmm0[6,7]
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
; AVX2-SLOW-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm4, %ymm0
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm4 = <u,0,0,u,1,1,u,2>
; AVX2-SLOW-NEXT:    vpermd %ymm14, %ymm4, %ymm7
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm8 = [255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255]
; AVX2-SLOW-NEXT:    vpblendvb %ymm8, %ymm0, %ymm7, %ymm0
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm7 = xmm3[1,1,2,2]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm5[3,3,3,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm1 = xmm7[0,1],xmm1[2],xmm7[3,4],xmm1[5],xmm7[6,7]
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
; AVX2-SLOW-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm3, %ymm1
; AVX2-SLOW-NEXT:    vpermd %ymm2, %ymm4, %ymm3
; AVX2-SLOW-NEXT:    vpblendvb %ymm8, %ymm1, %ymm3, %ymm1
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [10,11,0,1,12,13,12,13,2,3,2,3,14,15,4,5,26,27,16,17,28,29,28,29,18,19,18,19,30,31,20,21]
; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm11, %ymm4
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm5 = <u,3,3,u,4,4,u,5>
; AVX2-SLOW-NEXT:    vpermd 32(%rdi), %ymm5, %ymm6
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm7 = <255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255>
; AVX2-SLOW-NEXT:    vpblendvb %ymm7, %ymm4, %ymm6, %ymm4
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm6 = <2,u,3,3,u,4,4,u>
; AVX2-SLOW-NEXT:    vpermd %ymm2, %ymm6, %ymm2
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm8 = [255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255]
; AVX2-SLOW-NEXT:    vpblendvb %ymm8, %ymm4, %ymm2, %ymm2
; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm9, %ymm3
; AVX2-SLOW-NEXT:    vpermd (%rdi), %ymm5, %ymm4
; AVX2-SLOW-NEXT:    vpblendvb %ymm7, %ymm3, %ymm4, %ymm3
; AVX2-SLOW-NEXT:    vpermd %ymm14, %ymm6, %ymm4
; AVX2-SLOW-NEXT:    vpblendvb %ymm8, %ymm3, %ymm4, %ymm3
; AVX2-SLOW-NEXT:    vmovdqa %ymm3, 32(%rcx)
; AVX2-SLOW-NEXT:    vmovdqa %ymm2, 128(%rcx)
; AVX2-SLOW-NEXT:    vmovdqa %ymm1, 96(%rcx)
; AVX2-SLOW-NEXT:    vmovdqa %ymm0, (%rcx)
; AVX2-SLOW-NEXT:    vmovdqa %ymm12, 64(%rcx)
; AVX2-SLOW-NEXT:    vmovdqa %ymm10, 160(%rcx)
; AVX2-SLOW-NEXT:    vzeroupper
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-ALL-LABEL: vf32:
; AVX2-FAST-ALL:       # %bb.0:
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdx), %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rsi), %xmm1
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rsi), %xmm3
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm2 = <u,u,u,u,6,7,u,u,u,u,8,9,u,u,u,u>
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm2, %xmm1, %xmm4
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdi), %xmm5
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rdi), %xmm6
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm7 = xmm5[1,1,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0,1],xmm4[2],xmm7[3,4],xmm4[5],xmm7[6,7]
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm10 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm10, %xmm1, %xmm1
; AVX2-FAST-ALL-NEXT:    vinserti128 $1, %xmm4, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm11 = <u,0,0,u,1,1,u,2>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm11, %ymm7
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm8 = [255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm8, %ymm1, %ymm7, %ymm9
; AVX2-FAST-ALL-NEXT:    vmovdqa 48(%rdi), %xmm7
; AVX2-FAST-ALL-NEXT:    vmovdqa 48(%rsi), %xmm1
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} xmm5 = xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm12 = [4,5,10,11,10,11,8,9,8,9,14,15,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm12, %xmm5, %xmm5
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm7 = xmm7[1,1,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm1 = xmm7[0,1],xmm1[2],xmm7[3,4],xmm1[5],xmm7[6,7]
; AVX2-FAST-ALL-NEXT:    vinserti128 $1, %xmm5, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm2, %xmm3, %xmm5
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm7 = xmm6[1,1,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm5 = xmm7[0,1],xmm5[2],xmm7[3,4],xmm5[5],xmm7[6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rdx), %ymm7
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
; AVX2-FAST-ALL-NEXT:    vmovdqa 16(%rsi), %xmm6
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm10, %xmm3, %xmm3
; AVX2-FAST-ALL-NEXT:    vmovdqa 16(%rdi), %xmm4
; AVX2-FAST-ALL-NEXT:    vinserti128 $1, %xmm5, %ymm3, %ymm3
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm5 = <5,5,u,6,6,u,7,7>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm7, %ymm11, %ymm10
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm8, %ymm3, %ymm10, %ymm11
; AVX2-FAST-ALL-NEXT:    vpermd %ymm7, %ymm5, %ymm8
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm10, %ymm1, %ymm8, %ymm1
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} xmm3 = xmm6[4],xmm4[4],xmm6[5],xmm4[5],xmm6[6],xmm4[6],xmm6[7],xmm4[7]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm12, %xmm3, %xmm3
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm2, %xmm6, %xmm2
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rsi), %ymm6
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[1,1,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0,1],xmm2[2],xmm4[3,4],xmm2[5],xmm4[6,7]
; AVX2-FAST-ALL-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm5, %ymm3
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm10, %ymm2, %ymm3, %ymm2
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm3 = [10,11,0,1,12,13,12,13,2,3,2,3,14,15,4,5,26,27,16,17,28,29,28,29,18,19,18,19,30,31,20,21]
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm3, %ymm6, %ymm4
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm5 = <u,3,3,u,4,4,u,5>
; AVX2-FAST-ALL-NEXT:    vpermd 32(%rdi), %ymm5, %ymm6
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm8 = <255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255>
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm8, %ymm4, %ymm6, %ymm4
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rsi), %ymm6
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm3, %ymm6, %ymm3
; AVX2-FAST-ALL-NEXT:    vpermd (%rdi), %ymm5, %ymm5
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm8, %ymm3, %ymm5, %ymm3
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm5 = <2,u,3,3,u,4,4,u>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm7, %ymm5, %ymm6
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm7, %ymm4, %ymm6, %ymm4
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm5, %ymm0
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm7, %ymm3, %ymm0, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm0, 32(%rcx)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm4, 128(%rcx)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm2, 64(%rcx)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm11, 96(%rcx)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm1, 160(%rcx)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm9, (%rcx)
; AVX2-FAST-ALL-NEXT:    vzeroupper
; AVX2-FAST-ALL-NEXT:    retq
;
; AVX2-FAST-PERLANE-LABEL: vf32:
; AVX2-FAST-PERLANE:       # %bb.0:
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdx), %ymm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rsi), %xmm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rsi), %xmm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm2 = <u,u,u,u,6,7,u,u,u,u,8,9,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm2, %xmm1, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdi), %xmm5
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rdi), %xmm6
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm7 = xmm5[1,1,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0,1],xmm4[2],xmm7[3,4],xmm4[5],xmm7[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm10 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm10, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm4, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm11 = <u,0,0,u,1,1,u,2>
; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm0, %ymm11, %ymm7
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm8 = [255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm8, %ymm1, %ymm7, %ymm9
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 48(%rdi), %xmm7
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 48(%rsi), %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} xmm5 = xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm12 = [4,5,10,11,10,11,8,9,8,9,14,15,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm12, %xmm5, %xmm5
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm7 = xmm7[1,1,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm1 = xmm7[0,1],xmm1[2],xmm7[3,4],xmm1[5],xmm7[6,7]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm5, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm2, %xmm3, %xmm5
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm7 = xmm6[1,1,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm5 = xmm7[0,1],xmm5[2],xmm7[3,4],xmm5[5],xmm7[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rdx), %ymm7
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 16(%rsi), %xmm6
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm10, %xmm3, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 16(%rdi), %xmm4
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm5, %ymm3, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm5 = <5,5,u,6,6,u,7,7>
; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm7, %ymm11, %ymm10
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm8, %ymm3, %ymm10, %ymm11
; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm7, %ymm5, %ymm8
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm10 = [0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm10, %ymm1, %ymm8, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} xmm3 = xmm6[4],xmm4[4],xmm6[5],xmm4[5],xmm6[6],xmm4[6],xmm6[7],xmm4[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm12, %xmm3, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm2, %xmm6, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rsi), %ymm6
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[1,1,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0,1],xmm2[2],xmm4[3,4],xmm2[5],xmm4[6,7]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm0, %ymm5, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm10, %ymm2, %ymm3, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm3 = [10,11,0,1,12,13,12,13,2,3,2,3,14,15,4,5,26,27,16,17,28,29,28,29,18,19,18,19,30,31,20,21]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm3, %ymm6, %ymm4
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm5 = <u,3,3,u,4,4,u,5>
; AVX2-FAST-PERLANE-NEXT:    vpermd 32(%rdi), %ymm5, %ymm6
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm8 = <255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255,u,u,0,0,255,255>
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm8, %ymm4, %ymm6, %ymm4
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rsi), %ymm6
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm3, %ymm6, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vpermd (%rdi), %ymm5, %ymm5
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm8, %ymm3, %ymm5, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm5 = <2,u,3,3,u,4,4,u>
; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm7, %ymm5, %ymm6
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm7, %ymm4, %ymm6, %ymm4
; AVX2-FAST-PERLANE-NEXT:    vpermd %ymm0, %ymm5, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm7, %ymm3, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm0, 32(%rcx)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm4, 128(%rcx)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm2, 64(%rcx)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm11, 96(%rcx)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm1, 160(%rcx)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm9, (%rcx)
; AVX2-FAST-PERLANE-NEXT:    vzeroupper
; AVX2-FAST-PERLANE-NEXT:    retq
;
; AVX512-LABEL: vf32:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqu64 (%rdi), %zmm0
; AVX512-NEXT:    vmovdqu64 (%rsi), %zmm1
; AVX512-NEXT:    vmovdqu64 (%rdx), %zmm2
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm3 = <0,32,u,1,33,u,2,34,u,3,35,u,4,36,u,5,37,u,6,38,u,7,39,u,8,40,u,9,41,u,10,42>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm0, %zmm3
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm4 = [0,1,32,3,4,33,6,7,34,9,10,35,12,13,36,15,16,37,18,19,38,21,22,39,24,25,40,27,28,41,30,31]
; AVX512-NEXT:    vpermi2w %zmm2, %zmm3, %zmm4
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm3 = <u,11,43,u,12,44,u,13,45,u,14,46,u,15,47,u,16,48,u,17,49,u,18,50,u,19,51,u,20,52,u,21>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm0, %zmm3
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm5 = [42,1,2,43,4,5,44,7,8,45,10,11,46,13,14,47,16,17,48,19,20,49,22,23,50,25,26,51,28,29,52,31]
; AVX512-NEXT:    vpermi2w %zmm2, %zmm3, %zmm5
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm3 = <21,u,54,22,u,55,23,u,56,24,u,57,25,u,58,26,u,59,27,u,60,28,u,61,29,u,62,30,u,63,31,u>
; AVX512-NEXT:    vpermi2w %zmm0, %zmm1, %zmm3
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [0,53,2,3,54,5,6,55,8,9,56,11,12,57,14,15,58,17,18,59,20,21,60,23,24,61,26,27,62,29,30,63]
; AVX512-NEXT:    vpermi2w %zmm2, %zmm3, %zmm0
; AVX512-NEXT:    vmovdqu64 %zmm0, 128(%rcx)
; AVX512-NEXT:    vmovdqu64 %zmm5, 64(%rcx)
; AVX512-NEXT:    vmovdqu64 %zmm4, (%rcx)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %in.vec0 = load <32 x i16>, ptr %in.vecptr0, align 32
  %in.vec1 = load <32 x i16>, ptr %in.vecptr1, align 32
  %in.vec2 = load <32 x i16>, ptr %in.vecptr2, align 32

  %concat01 = shufflevector <32 x i16> %in.vec0, <32 x i16> %in.vec1, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
  %concat2u = shufflevector <32 x i16> %in.vec2, <32 x i16> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
  %concat012 = shufflevector <64 x i16> %concat01, <64 x i16> %concat2u, <96 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64, i32 65, i32 66, i32 67, i32 68, i32 69, i32 70, i32 71, i32 72, i32 73, i32 74, i32 75, i32 76, i32 77, i32 78, i32 79, i32 80, i32 81, i32 82, i32 83, i32 84, i32 85, i32 86, i32 87, i32 88, i32 89, i32 90, i32 91, i32 92, i32 93, i32 94, i32 95>
  %interleaved.vec = shufflevector <96 x i16> %concat012, <96 x i16> poison, <96 x i32> <i32 0, i32 32, i32 64, i32 1, i32 33, i32 65, i32 2, i32 34, i32 66, i32 3, i32 35, i32 67, i32 4, i32 36, i32 68, i32 5, i32 37, i32 69, i32 6, i32 38, i32 70, i32 7, i32 39, i32 71, i32 8, i32 40, i32 72, i32 9, i32 41, i32 73, i32 10, i32 42, i32 74, i32 11, i32 43, i32 75, i32 12, i32 44, i32 76, i32 13, i32 45, i32 77, i32 14, i32 46, i32 78, i32 15, i32 47, i32 79, i32 16, i32 48, i32 80, i32 17, i32 49, i32 81, i32 18, i32 50, i32 82, i32 19, i32 51, i32 83, i32 20, i32 52, i32 84, i32 21, i32 53, i32 85, i32 22, i32 54, i32 86, i32 23, i32 55, i32 87, i32 24, i32 56, i32 88, i32 25, i32 57, i32 89, i32 26, i32 58, i32 90, i32 27, i32 59, i32 91, i32 28, i32 60, i32 92, i32 29, i32 61, i32 93, i32 30, i32 62, i32 94, i32 31, i32 63, i32 95>

  store <96 x i16> %interleaved.vec, ptr %out.vec, align 32

  ret void
}