test/CodeGen/X86/vector-interleaved-store-i16-stride-6.ll

; 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=AVX1
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX2-SLOW
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX2-FAST,AVX2-FAST-ALL
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX2-FAST,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 %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %out.vec) nounwind {
; SSE-LABEL: vf2:
; SSE:       # %bb.0:
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    movdqa (%rdx), %xmm1
; SSE-NEXT:    movdqa (%r8), %xmm2
; SSE-NEXT:    movdqa (%r9), %xmm3
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; 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[0,2,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,5,4,6]
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm4[3,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[0,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,0,3,3,4,5,6,7]
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,0,3,4,5,6,7]
; SSE-NEXT:    movaps %xmm0, (%rax)
; SSE-NEXT:    movq %xmm1, 16(%rax)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf2:
; AVX1:       # %bb.0:
; AVX1-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-NEXT:    vmovdqa (%rdi), %xmm0
; AVX1-NEXT:    vmovdqa (%rdx), %xmm1
; AVX1-NEXT:    vmovdqa (%r8), %xmm2
; AVX1-NEXT:    vmovdqa (%r9), %xmm3
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,5,7,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[2,2,2,2]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,u,u,u,u,2,3,6,7]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm2[4,5],xmm0[6,7]
; AVX1-NEXT:    vmovdqa %xmm0, (%rax)
; AVX1-NEXT:    vmovq %xmm1, 16(%rax)
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf2:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-SLOW-NEXT:    vmovdqa (%rdx), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa (%r8), %xmm2
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],mem[0],xmm2[1],mem[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,4,5,8,9,12,13,u,u,u,u,2,3,6,7,u,u,u,u,18,19,22,23,u,u,u,u,u,u,u,u]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3],ymm0[4],ymm1[5,6,7]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm1
; AVX2-SLOW-NEXT:    vmovq %xmm1, 16(%rax)
; AVX2-SLOW-NEXT:    vmovdqa %xmm0, (%rax)
; AVX2-SLOW-NEXT:    vzeroupper
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-LABEL: vf2:
; AVX2-FAST:       # %bb.0:
; AVX2-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-FAST-NEXT:    vmovdqa (%rdx), %xmm1
; AVX2-FAST-NEXT:    vmovdqa (%r8), %xmm2
; AVX2-FAST-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; AVX2-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; AVX2-FAST-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; AVX2-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],mem[0],xmm2[1],mem[1]
; AVX2-FAST-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,4,5,8,9,12,13,u,u,u,u,2,3,6,7,u,u,u,u,18,19,22,23,u,u,u,u,u,u,u,u]
; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,8,9,12,13,u,u,u,u,18,19,22,23,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-NEXT:    vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-NEXT:    vmovq %xmm1, 16(%rax)
; AVX2-FAST-NEXT:    vmovdqa %xmm0, (%rax)
; AVX2-FAST-NEXT:    vzeroupper
; AVX2-FAST-NEXT:    retq
;
; AVX512-LABEL: vf2:
; AVX512:       # %bb.0:
; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-NEXT:    vmovdqa (%rdx), %xmm1
; AVX512-NEXT:    vmovdqa (%r8), %xmm2
; AVX512-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; AVX512-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; AVX512-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; AVX512-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],mem[0],xmm2[1],mem[1]
; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm1 = <0,2,4,6,8,10,1,3,5,7,9,11,u,u,u,u>
; AVX512-NEXT:    vpermw %ymm0, %ymm1, %ymm0
; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT:    vmovq %xmm1, 16(%rax)
; AVX512-NEXT:    vmovdqa %xmm0, (%rax)
; AVX512-NEXT:    vzeroupper
; 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
  %in.vec3 = load <2 x i16>, ptr %in.vecptr3, align 32
  %in.vec4 = load <2 x i16>, ptr %in.vecptr4, align 32
  %in.vec5 = load <2 x i16>, ptr %in.vecptr5, 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>
  %concat23 = shufflevector <2 x i16> %in.vec2, <2 x i16> %in.vec3, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %concat45 = shufflevector <2 x i16> %in.vec4, <2 x i16> %in.vec5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %concat0123 = shufflevector <4 x i16> %concat01, <4 x i16> %concat23, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %concat45uu = shufflevector <4 x i16> %concat45, <4 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
  %concat012345 = shufflevector <8 x i16> %concat0123, <8 x i16> %concat45uu, <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> %concat012345, <12 x i16> poison, <12 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11>

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

  ret void
}

define void @vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %out.vec) nounwind {
; SSE-LABEL: vf4:
; SSE:       # %bb.0:
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
; SSE-NEXT:    movq {{.*#+}} xmm1 = mem[0],zero
; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE-NEXT:    movq {{.*#+}} xmm1 = mem[0],zero
; SSE-NEXT:    movq {{.*#+}} xmm2 = mem[0],zero
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; SSE-NEXT:    movq {{.*#+}} xmm2 = mem[0],zero
; SSE-NEXT:    movq {{.*#+}} xmm3 = mem[0],zero
; SSE-NEXT:    movdqa %xmm3, %xmm4
; SSE-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm2[0]
; SSE-NEXT:    movdqa %xmm1, %xmm5
; SSE-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm0[0]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[0,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,4,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,1],xmm0[1,3]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm6[0,2]
; SSE-NEXT:    movdqa %xmm0, %xmm6
; SSE-NEXT:    punpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm1[1]
; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm1[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm6[0,2]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm1[3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[3,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,7]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[0,3]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0,1,3]
; SSE-NEXT:    movaps %xmm3, 16(%rax)
; SSE-NEXT:    movaps %xmm5, (%rax)
; SSE-NEXT:    movaps %xmm0, 32(%rax)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf4:
; AVX1:       # %bb.0:
; AVX1-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX1-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX1-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX1-NEXT:    vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm3 = xmm2[0],xmm1[0]
; AVX1-NEXT:    vmovq {{.*#+}} xmm4 = mem[0],zero
; AVX1-NEXT:    vmovq {{.*#+}} xmm5 = mem[0],zero
; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm8 = xmm5[0],xmm4[0]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm7 = xmm3[2,3,10,11,u,u,u,u,u,u,u,u,4,5,12,13]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm0[0,1,1,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,6,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm6 = xmm7[0,1,2,3],xmm6[4,5],xmm7[6,7]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3],xmm6[4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,1,0,2,4,5,6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[0,1,8,9,u,u,u,u,u,u,u,u,2,3,10,11]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm8[0,1,2,0]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,4,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5],xmm1[6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm1, %ymm1
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm3[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm2 = xmm8[4,5,12,13,u,u,u,u,u,u,u,u,6,7,14,15]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3,4,5],xmm2[6,7]
; AVX1-NEXT:    vmovdqa %xmm0, 32(%rax)
; AVX1-NEXT:    vmovaps %ymm1, (%rax)
; AVX1-NEXT:    vzeroupper
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf4:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-SLOW-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX2-SLOW-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX2-SLOW-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX2-SLOW-NEXT:    vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm2
; AVX2-SLOW-NEXT:    vmovq {{.*#+}} xmm3 = mem[0],zero
; AVX2-SLOW-NEXT:    vmovq {{.*#+}} xmm4 = mem[0],zero
; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm5 = xmm4[0],xmm3[0]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm6 = ymm2[0,1,8,9,u,u,u,u,u,u,u,u,2,3,10,11,18,19,26,27,u,u,u,u,u,u,u,u,20,21,28,29]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,0,1,8,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,28,29,u,u,u,u]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0],ymm2[1],ymm6[2,3,4,5],ymm2[6],ymm6[7]
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; AVX2-SLOW-NEXT:    vpbroadcastq %xmm3, %ymm3
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm1 = xmm5[4,5,12,13,u,u,u,u,u,u,u,u,6,7,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
; AVX2-SLOW-NEXT:    vmovdqa %xmm0, 32(%rax)
; AVX2-SLOW-NEXT:    vmovdqa %ymm2, (%rax)
; AVX2-SLOW-NEXT:    vzeroupper
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-ALL-LABEL: vf4:
; AVX2-FAST-ALL:       # %bb.0:
; AVX2-FAST-ALL-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-ALL-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX2-FAST-ALL-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX2-FAST-ALL-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX2-FAST-ALL-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX2-FAST-ALL-NEXT:    vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX2-FAST-ALL-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX2-FAST-ALL-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm2
; AVX2-FAST-ALL-NEXT:    vmovq {{.*#+}} xmm3 = mem[0],zero
; AVX2-FAST-ALL-NEXT:    vmovq {{.*#+}} xmm4 = mem[0],zero
; AVX2-FAST-ALL-NEXT:    vpunpcklqdq {{.*#+}} xmm5 = xmm4[0],xmm3[0]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm6 = ymm2[0,1,8,9,u,u,u,u,u,u,u,u,2,3,10,11,18,19,26,27,u,u,u,u,u,u,u,u,20,21,28,29]
; AVX2-FAST-ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm7 = [4,6,1,3,4,6,1,3]
; AVX2-FAST-ALL-NEXT:    # ymm7 = mem[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm2, %ymm7, %ymm2
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,0,1,4,5,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,28,29,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0],ymm2[1],ymm6[2,3,4,5],ymm2[6],ymm6[7]
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; AVX2-FAST-ALL-NEXT:    vpbroadcastq %xmm3, %ymm3
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[6,7,14,15,4,5,6,7,u,u,u,u,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[12,13,14,15,6,7,14,15,u,u,u,u,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm1 = xmm5[4,5,12,13,u,u,u,u,u,u,u,u,6,7,14,15]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm0, 32(%rax)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm2, (%rax)
; AVX2-FAST-ALL-NEXT:    vzeroupper
; AVX2-FAST-ALL-NEXT:    retq
;
; AVX2-FAST-PERLANE-LABEL: vf4:
; AVX2-FAST-PERLANE:       # %bb.0:
; AVX2-FAST-PERLANE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-PERLANE-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX2-FAST-PERLANE-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX2-FAST-PERLANE-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX2-FAST-PERLANE-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX2-FAST-PERLANE-NEXT:    vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX2-FAST-PERLANE-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vmovq {{.*#+}} xmm3 = mem[0],zero
; AVX2-FAST-PERLANE-NEXT:    vmovq {{.*#+}} xmm4 = mem[0],zero
; AVX2-FAST-PERLANE-NEXT:    vpunpcklqdq {{.*#+}} xmm5 = xmm4[0],xmm3[0]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm6 = ymm2[0,1,8,9,u,u,u,u,u,u,u,u,2,3,10,11,18,19,26,27,u,u,u,u,u,u,u,u,20,21,28,29]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,0,1,8,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,28,29,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0],ymm2[1],ymm6[2,3,4,5],ymm2[6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq %xmm3, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[6,7,14,15,4,5,6,7,u,u,u,u,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[12,13,14,15,6,7,14,15,u,u,u,u,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm1 = xmm5[4,5,12,13,u,u,u,u,u,u,u,u,6,7,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm0, 32(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm2, (%rax)
; AVX2-FAST-PERLANE-NEXT:    vzeroupper
; AVX2-FAST-PERLANE-NEXT:    retq
;
; AVX512-LABEL: vf4:
; AVX512:       # %bb.0:
; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX512-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX512-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX512-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX512-NEXT:    vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX512-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512-NEXT:    vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX512-NEXT:    vmovq {{.*#+}} xmm3 = mem[0],zero
; AVX512-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX512-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm1 = <0,4,8,12,16,20,1,5,9,13,17,21,2,6,10,14,18,22,3,7,11,15,19,23,u,u,u,u,u,u,u,u>
; AVX512-NEXT:    vpermw %zmm0, %zmm1, %zmm0
; AVX512-NEXT:    vextracti32x4 $2, %zmm0, 32(%rax)
; AVX512-NEXT:    vmovdqa %ymm0, (%rax)
; 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
  %in.vec3 = load <4 x i16>, ptr %in.vecptr3, align 32
  %in.vec4 = load <4 x i16>, ptr %in.vecptr4, align 32
  %in.vec5 = load <4 x i16>, ptr %in.vecptr5, 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>
  %concat23 = shufflevector <4 x i16> %in.vec2, <4 x i16> %in.vec3, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %concat45 = shufflevector <4 x i16> %in.vec4, <4 x i16> %in.vec5, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %concat0123 = shufflevector <8 x i16> %concat01, <8 x i16> %concat23, <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>
  %concat45uu = shufflevector <8 x i16> %concat45, <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>
  %concat012345 = shufflevector <16 x i16> %concat0123, <16 x i16> %concat45uu, <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> %concat012345, <24 x i16> poison, <24 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 3, i32 7, i32 11, i32 15, i32 19, i32 23>

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

  ret void
}

define void @vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %out.vec) nounwind {
; SSE-LABEL: vf8:
; SSE:       # %bb.0:
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movdqa (%rdi), %xmm10
; SSE-NEXT:    movdqa (%rsi), %xmm0
; SSE-NEXT:    movdqa (%rdx), %xmm2
; SSE-NEXT:    movdqa (%rcx), %xmm1
; SSE-NEXT:    movdqa (%r8), %xmm3
; SSE-NEXT:    movdqa (%r9), %xmm5
; SSE-NEXT:    movdqa %xmm2, %xmm9
; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3]
; SSE-NEXT:    movdqa %xmm10, %xmm12
; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1],xmm12[2],xmm0[2],xmm12[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm12, %xmm4
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,3],xmm9[3,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm3[2,1,3,3,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,2],xmm6[0,1]
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0,1,3]
; SSE-NEXT:    movaps {{.*#+}} xmm11 = [65535,0,65535,65535,65535,65535,65535,0]
; SSE-NEXT:    andps %xmm11, %xmm4
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm5[0,2,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,2,1]
; SSE-NEXT:    movaps %xmm11, %xmm8
; SSE-NEXT:    andnps %xmm6, %xmm8
; SSE-NEXT:    orps %xmm4, %xmm8
; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
; SSE-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm0[4],xmm10[5],xmm0[5],xmm10[6],xmm0[6],xmm10[7],xmm0[7]
; SSE-NEXT:    movdqa %xmm10, %xmm0
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm2[3,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,6,5,7,7]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0,1,3]
; SSE-NEXT:    andps %xmm11, %xmm0
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm5[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
; SSE-NEXT:    andnps %xmm1, %xmm11
; SSE-NEXT:    orps %xmm0, %xmm11
; SSE-NEXT:    movdqa %xmm9, %xmm0
; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm12[0]
; SSE-NEXT:    movdqa %xmm3, %xmm1
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm12[1,3]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[0,2]
; SSE-NEXT:    movaps {{.*#+}} xmm4 = [65535,65535,65535,65535,65535,0,65535,65535]
; SSE-NEXT:    andps %xmm4, %xmm0
; SSE-NEXT:    movdqa %xmm5, %xmm6
; SSE-NEXT:    pslldq {{.*#+}} xmm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm6[0,1,2,3,4,5]
; SSE-NEXT:    movaps %xmm4, %xmm1
; SSE-NEXT:    andnps %xmm6, %xmm1
; SSE-NEXT:    orps %xmm0, %xmm1
; SSE-NEXT:    movdqa %xmm10, %xmm0
; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; SSE-NEXT:    movdqa %xmm3, %xmm7
; SSE-NEXT:    psrldq {{.*#+}} xmm7 = xmm7[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,1],xmm2[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm0[0,2]
; SSE-NEXT:    movaps {{.*#+}} xmm6 = [65535,65535,65535,0,65535,65535,65535,65535]
; SSE-NEXT:    andps %xmm6, %xmm7
; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm5[2,2,3,3]
; SSE-NEXT:    movaps %xmm6, %xmm0
; SSE-NEXT:    andnps %xmm13, %xmm0
; SSE-NEXT:    orps %xmm7, %xmm0
; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm10[0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm3[1,1,1,1,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,1],xmm10[1,3]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm3[0,2]
; SSE-NEXT:    andps %xmm4, %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[0,0,1,1]
; SSE-NEXT:    pslld $16, %xmm5
; SSE-NEXT:    andnps %xmm5, %xmm4
; SSE-NEXT:    orps %xmm2, %xmm4
; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm9[1]
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,1],xmm9[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm12[0,2]
; SSE-NEXT:    andps %xmm6, %xmm7
; SSE-NEXT:    andnps %xmm3, %xmm6
; SSE-NEXT:    orps %xmm7, %xmm6
; SSE-NEXT:    movaps %xmm6, 16(%rax)
; SSE-NEXT:    movaps %xmm4, 48(%rax)
; SSE-NEXT:    movaps %xmm0, 64(%rax)
; SSE-NEXT:    movaps %xmm1, (%rax)
; SSE-NEXT:    movaps %xmm11, 80(%rax)
; SSE-NEXT:    movaps %xmm8, 32(%rax)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf8:
; AVX1:       # %bb.0:
; AVX1-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-NEXT:    vmovdqa (%rdi), %xmm8
; AVX1-NEXT:    vmovdqa (%rsi), %xmm9
; AVX1-NEXT:    vmovdqa (%rdx), %xmm2
; AVX1-NEXT:    vmovdqa (%rcx), %xmm3
; AVX1-NEXT:    vmovdqa (%r8), %xmm11
; AVX1-NEXT:    vmovdqa (%r9), %xmm5
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm9[0],xmm8[1],xmm9[1],xmm8[2],xmm9[2],xmm8[3],xmm9[3]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm7 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm7[1,1,2,2]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm6[4,5],xmm0[6,7]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm11[0],xmm5[0],xmm11[1],xmm5[1],xmm11[2],xmm5[2],xmm11[3],xmm5[3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm10 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm7[0,0,1,1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm6[0,1,0,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm4[0,1],xmm0[2,3],xmm4[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm1[0,1,0,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm4[4,5],xmm0[6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm8[4],xmm9[4],xmm8[5],xmm9[5],xmm8[6],xmm9[6],xmm8[7],xmm9[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm4[0,1,0,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3],xmm0[4,5,6,7]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm3 = xmm11[4],xmm5[4],xmm11[5],xmm5[5],xmm11[6],xmm5[6],xmm11[7],xmm5[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm3[0,1,0,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm5[4,5],xmm0[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm7[2,2,3,3]
; AVX1-NEXT:    vpunpckhqdq {{.*#+}} xmm5 = xmm6[1],xmm5[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3,4,5],xmm1[6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[2,2,3,3]
; AVX1-NEXT:    vpunpckhqdq {{.*#+}} xmm1 = xmm4[1],xmm1[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm3[2,3,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3,4,5],xmm5[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,2,2]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm4[4,5],xmm2[6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
; AVX1-NEXT:    vmovaps %ymm1, 64(%rax)
; AVX1-NEXT:    vmovaps %ymm0, 32(%rax)
; AVX1-NEXT:    vmovaps %ymm10, (%rax)
; AVX1-NEXT:    vzeroupper
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf8:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-SLOW-NEXT:    vmovdqa (%rdx), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa (%r8), %xmm2
; AVX2-SLOW-NEXT:    vinserti128 $1, (%rsi), %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vinserti128 $1, (%rcx), %ymm1, %ymm1
; AVX2-SLOW-NEXT:    vinserti128 $1, (%r9), %ymm2, %ymm2
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[0,2,0,2]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm3 = ymm3[u,u,u,u,0,1,8,9,u,u,u,u,u,u,u,u,18,19,26,27,u,u,u,u,u,u,u,u,20,21,28,29]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[0,2,0,2]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[0,1,8,9,u,u,u,u,u,u,u,u,2,3,10,11,u,u,u,u,u,u,u,u,20,21,28,29,u,u,u,u]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[0,2,0,2]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u,u,u,u,u,0,1,8,9,u,u,u,u,u,u,u,u,18,19,26,27,u,u,u,u,u,u,u,u]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[0,2,1,3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u,u,u,u,u,6,7,14,15,u,u,u,u,u,u,u,u,16,17,24,25,u,u,u,u,u,u,u,u]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[0,2,1,3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,u,6,7,14,15,u,u,u,u,u,u,u,u,16,17,24,25,u,u,u,u,u,u,u,u,18,19,26,27]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3,4],ymm4[5],ymm5[6,7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm2[0,2,1,3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[4,5,12,13,u,u,u,u,u,u,u,u,6,7,14,15,u,u,u,u,u,u,u,u,16,17,24,25,u,u,u,u]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0],ymm4[1,2],ymm5[3],ymm4[4,5],ymm5[6],ymm4[7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,3,1,3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,4,5,12,13,u,u,u,u,u,u,u,u,22,23,30,31,u,u,u,u,u,u,u,u]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[1,3,1,3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[2,3,10,11,u,u,u,u,u,u,u,u,4,5,12,13,u,u,u,u,u,u,u,u,22,23,30,31,u,u,u,u]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm2[1,3,1,3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,2,3,10,11,u,u,u,u,u,u,u,u,20,21,28,29,u,u,u,u,u,u,u,u,22,23,30,31]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX2-SLOW-NEXT:    vmovdqa %ymm0, 64(%rax)
; AVX2-SLOW-NEXT:    vmovdqa %ymm4, 32(%rax)
; AVX2-SLOW-NEXT:    vmovdqa %ymm3, (%rax)
; AVX2-SLOW-NEXT:    vzeroupper
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-ALL-LABEL: vf8:
; AVX2-FAST-ALL:       # %bb.0:
; AVX2-FAST-ALL-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdx), %xmm1
; AVX2-FAST-ALL-NEXT:    vmovdqa (%r8), %xmm2
; AVX2-FAST-ALL-NEXT:    vinserti128 $1, (%rsi), %ymm0, %ymm0
; AVX2-FAST-ALL-NEXT:    vinserti128 $1, (%rcx), %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vinserti128 $1, (%r9), %ymm2, %ymm2
; AVX2-FAST-ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,4,1,5,0,4,1,5]
; AVX2-FAST-ALL-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm1, %ymm3, %ymm3
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm3 = ymm3[u,u,u,u,0,1,4,5,u,u,u,u,u,u,u,u,18,19,22,23,u,u,u,u,u,u,u,u,24,25,28,29]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[0,2,0,2]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[0,1,8,9,u,u,u,u,u,u,u,u,2,3,10,11,u,u,u,u,u,u,u,u,20,21,28,29,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7]
; AVX2-FAST-ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm4 = [0,4,4,0,0,4,4,0]
; AVX2-FAST-ALL-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm2, %ymm4, %ymm4
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u,u,u,u,u,12,13,8,9,u,u,u,u,u,u,u,u,18,19,22,23,u,u,u,u,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7]
; AVX2-FAST-ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm4 = [2,6,1,5,2,6,1,5]
; AVX2-FAST-ALL-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm1, %ymm4, %ymm4
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u,u,u,u,u,10,11,14,15,u,u,u,u,u,u,u,u,16,17,20,21,u,u,u,u,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[0,2,1,3]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,u,6,7,14,15,u,u,u,u,u,u,u,u,16,17,24,25,u,u,u,u,u,u,u,u,18,19,26,27]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3,4],ymm4[5],ymm5[6,7]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm5 = ymm2[0,2,1,3]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[4,5,12,13,u,u,u,u,u,u,u,u,6,7,14,15,u,u,u,u,u,u,u,u,16,17,24,25,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0],ymm4[1,2],ymm5[3],ymm4[4,5],ymm5[6],ymm4[7]
; AVX2-FAST-ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [7,3,3,7,7,3,3,7]
; AVX2-FAST-ALL-NEXT:    # ymm5 = mem[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm5, %ymm0
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,8,9,12,13,u,u,u,u,u,u,u,u,22,23,18,19,u,u,u,u,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[1,3,1,3]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[2,3,10,11,u,u,u,u,u,u,u,u,4,5,12,13,u,u,u,u,u,u,u,u,22,23,30,31,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm2[1,3,1,3]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,2,3,10,11,u,u,u,u,u,u,u,u,20,21,28,29,u,u,u,u,u,u,u,u,22,23,30,31]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm0, 64(%rax)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm4, 32(%rax)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm3, (%rax)
; AVX2-FAST-ALL-NEXT:    vzeroupper
; AVX2-FAST-ALL-NEXT:    retq
;
; AVX2-FAST-PERLANE-LABEL: vf8:
; AVX2-FAST-PERLANE:       # %bb.0:
; AVX2-FAST-PERLANE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdx), %xmm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%r8), %xmm2
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, (%rsi), %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, (%rcx), %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, (%r9), %ymm2, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[0,2,0,2]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm3 = ymm3[u,u,u,u,0,1,8,9,u,u,u,u,u,u,u,u,18,19,26,27,u,u,u,u,u,u,u,u,20,21,28,29]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[0,2,0,2]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[0,1,8,9,u,u,u,u,u,u,u,u,2,3,10,11,u,u,u,u,u,u,u,u,20,21,28,29,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[0,2,0,2]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u,u,u,u,u,0,1,8,9,u,u,u,u,u,u,u,u,18,19,26,27,u,u,u,u,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm4 = ymm1[0,2,1,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u,u,u,u,u,6,7,14,15,u,u,u,u,u,u,u,u,16,17,24,25,u,u,u,u,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm5 = ymm0[0,2,1,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,u,6,7,14,15,u,u,u,u,u,u,u,u,16,17,24,25,u,u,u,u,u,u,u,u,18,19,26,27]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3,4],ymm4[5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm5 = ymm2[0,2,1,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[4,5,12,13,u,u,u,u,u,u,u,u,6,7,14,15,u,u,u,u,u,u,u,u,16,17,24,25,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0],ymm4[1,2],ymm5[3],ymm4[4,5],ymm5[6],ymm4[7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,3,1,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,4,5,12,13,u,u,u,u,u,u,u,u,22,23,30,31,u,u,u,u,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[1,3,1,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[2,3,10,11,u,u,u,u,u,u,u,u,4,5,12,13,u,u,u,u,u,u,u,u,22,23,30,31,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm2[1,3,1,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,2,3,10,11,u,u,u,u,u,u,u,u,20,21,28,29,u,u,u,u,u,u,u,u,22,23,30,31]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm0, 64(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm4, 32(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm3, (%rax)
; AVX2-FAST-PERLANE-NEXT:    vzeroupper
; AVX2-FAST-PERLANE-NEXT:    retq
;
; AVX512-LABEL: vf8:
; AVX512:       # %bb.0:
; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-NEXT:    vmovdqa (%rdx), %xmm1
; AVX512-NEXT:    vmovdqa (%r8), %xmm2
; AVX512-NEXT:    vinserti128 $1, (%rcx), %ymm1, %ymm1
; AVX512-NEXT:    vinserti128 $1, (%rsi), %ymm0, %ymm0
; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
; AVX512-NEXT:    vinserti32x4 $1, (%r9), %zmm2, %zmm1
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm2 = [21,29,37,45,6,14,22,30,38,46,7,15,23,31,39,47]
; AVX512-NEXT:    vpermi2w %zmm1, %zmm0, %zmm2
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,8,16,24,32,40,1,9,17,25,33,41,2,10,18,26,34,42,3,11,19,27,35,43,4,12,20,28,36,44,5,13]
; AVX512-NEXT:    vpermi2w %zmm1, %zmm0, %zmm3
; AVX512-NEXT:    vmovdqu64 %zmm3, (%rax)
; AVX512-NEXT:    vmovdqa %ymm2, 64(%rax)
; 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
  %in.vec3 = load <8 x i16>, ptr %in.vecptr3, align 32
  %in.vec4 = load <8 x i16>, ptr %in.vecptr4, align 32
  %in.vec5 = load <8 x i16>, ptr %in.vecptr5, 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>
  %concat23 = shufflevector <8 x i16> %in.vec2, <8 x i16> %in.vec3, <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>
  %concat45 = shufflevector <8 x i16> %in.vec4, <8 x i16> %in.vec5, <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>
  %concat0123 = shufflevector <16 x i16> %concat01, <16 x i16> %concat23, <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>
  %concat45uu = shufflevector <16 x i16> %concat45, <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>
  %concat012345 = shufflevector <32 x i16> %concat0123, <32 x i16> %concat45uu, <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> %concat012345, <48 x i16> poison, <48 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 3, i32 11, i32 19, i32 27, i32 35, i32 43, i32 4, i32 12, i32 20, i32 28, i32 36, i32 44, i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 7, i32 15, i32 23, i32 31, i32 39, i32 47>

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

  ret void
}

define void @vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %out.vec) nounwind {
; SSE-LABEL: vf16:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa (%rdi), %xmm11
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 16(%rdi), %xmm14
; SSE-NEXT:    movdqa (%rsi), %xmm15
; SSE-NEXT:    movdqa 16(%rsi), %xmm4
; SSE-NEXT:    movdqa (%rdx), %xmm8
; SSE-NEXT:    movdqa 16(%rdx), %xmm1
; SSE-NEXT:    movdqa (%rcx), %xmm10
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 16(%rcx), %xmm6
; SSE-NEXT:    movdqa 16(%r8), %xmm3
; SSE-NEXT:    movdqa 16(%r9), %xmm13
; SSE-NEXT:    movdqa %xmm1, %xmm7
; SSE-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]
; SSE-NEXT:    movdqa %xmm14, %xmm9
; SSE-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm4[4],xmm9[5],xmm4[5],xmm9[6],xmm4[6],xmm9[7],xmm4[7]
; SSE-NEXT:    movdqa %xmm9, %xmm0
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm7[3,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,6,5,7,7]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,2],xmm2[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0,1,3]
; SSE-NEXT:    movaps {{.*#+}} xmm12 = [65535,0,65535,65535,65535,65535,65535,0]
; SSE-NEXT:    andps %xmm12, %xmm0
; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm13[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
; SSE-NEXT:    movaps %xmm12, %xmm2
; SSE-NEXT:    andnps %xmm5, %xmm2
; SSE-NEXT:    orps %xmm0, %xmm2
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm4[0],xmm14[1],xmm4[1],xmm14[2],xmm4[2],xmm14[3],xmm4[3]
; SSE-NEXT:    movdqa %xmm14, %xmm0
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm1[3,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm3[2,1,3,3,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,2],xmm5[0,1]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0,1,3]
; SSE-NEXT:    andps %xmm12, %xmm0
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm13[0,2,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
; SSE-NEXT:    movaps %xmm12, %xmm2
; SSE-NEXT:    andnps %xmm5, %xmm2
; SSE-NEXT:    orps %xmm0, %xmm2
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm8, %xmm4
; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm10[4],xmm4[5],xmm10[5],xmm4[6],xmm10[6],xmm4[7],xmm10[7]
; SSE-NEXT:    movdqa %xmm11, %xmm2
; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
; SSE-NEXT:    movdqa %xmm15, %xmm10
; SSE-NEXT:    movdqa %xmm2, %xmm6
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,3],xmm4[3,3]
; SSE-NEXT:    movdqa (%r8), %xmm11
; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm11[0,1,2,3,6,5,7,7]
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,2],xmm5[2,3]
; SSE-NEXT:    movdqa (%r9), %xmm15
; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm15[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
; SSE-NEXT:    movaps %xmm12, %xmm0
; SSE-NEXT:    andnps %xmm5, %xmm0
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0,1,3]
; SSE-NEXT:    andps %xmm12, %xmm6
; SSE-NEXT:    orps %xmm6, %xmm0
; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = xmm8[0],mem[0],xmm8[1],mem[1],xmm8[2],mem[2],xmm8[3],mem[3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm10[0],xmm5[1],xmm10[1],xmm5[2],xmm10[2],xmm5[3],xmm10[3]
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,3],xmm8[3,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm11[2,1,3,3,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,2],xmm6[0,1]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0,1,3]
; SSE-NEXT:    andps %xmm12, %xmm5
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm15[0,2,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,2,1]
; SSE-NEXT:    andnps %xmm6, %xmm12
; SSE-NEXT:    orps %xmm5, %xmm12
; SSE-NEXT:    movdqa %xmm9, %xmm6
; SSE-NEXT:    punpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm7[1]
; SSE-NEXT:    movdqa %xmm3, %xmm5
; SSE-NEXT:    psrldq {{.*#+}} xmm5 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm6[0,2]
; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,0,65535,65535,65535,65535]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm13[2,2,3,3]
; SSE-NEXT:    movdqa %xmm6, %xmm0
; SSE-NEXT:    pandn %xmm10, %xmm0
; SSE-NEXT:    andps %xmm6, %xmm5
; SSE-NEXT:    por %xmm5, %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklqdq {{.*#+}} xmm7 = xmm7[0],xmm9[0]
; SSE-NEXT:    movdqa %xmm3, %xmm5
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm9[1,3]
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm5[0,2]
; SSE-NEXT:    movdqa %xmm13, %xmm5
; SSE-NEXT:    pslld $16, %xmm5
; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,65535,65535,0,65535,65535]
; SSE-NEXT:    movdqa %xmm9, %xmm10
; SSE-NEXT:    pandn %xmm5, %xmm10
; SSE-NEXT:    andps %xmm9, %xmm7
; SSE-NEXT:    por %xmm7, %xmm10
; SSE-NEXT:    movdqa %xmm14, %xmm5
; SSE-NEXT:    punpckhqdq {{.*#+}} xmm5 = xmm5[1],xmm1[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[1,1,1,1,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm5[0,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm13[0,0,1,1]
; SSE-NEXT:    movdqa %xmm6, %xmm7
; SSE-NEXT:    pandn %xmm5, %xmm7
; SSE-NEXT:    andps %xmm6, %xmm0
; SSE-NEXT:    por %xmm0, %xmm7
; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm14[0]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm14[1,3]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[0,2]
; SSE-NEXT:    pslldq {{.*#+}} xmm13 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm13[0,1,2,3,4,5]
; SSE-NEXT:    movdqa %xmm9, %xmm3
; SSE-NEXT:    pandn %xmm13, %xmm3
; SSE-NEXT:    andps %xmm9, %xmm1
; SSE-NEXT:    por %xmm1, %xmm3
; SSE-NEXT:    movdqa %xmm2, %xmm0
; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm4[1]
; SSE-NEXT:    movdqa %xmm11, %xmm5
; SSE-NEXT:    psrldq {{.*#+}} xmm5 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm4[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm0[0,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
; SSE-NEXT:    movdqa %xmm6, %xmm1
; SSE-NEXT:    pandn %xmm0, %xmm1
; SSE-NEXT:    andps %xmm6, %xmm5
; SSE-NEXT:    por %xmm5, %xmm1
; SSE-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm2[0]
; SSE-NEXT:    movdqa %xmm11, %xmm0
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,1],xmm2[1,3]
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm0[0,2]
; SSE-NEXT:    movdqa %xmm15, %xmm0
; SSE-NEXT:    pslld $16, %xmm0
; SSE-NEXT:    movdqa %xmm9, %xmm5
; SSE-NEXT:    pandn %xmm0, %xmm5
; SSE-NEXT:    andps %xmm9, %xmm4
; SSE-NEXT:    por %xmm4, %xmm5
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    movaps %xmm2, %xmm0
; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm8[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm11[1,1,1,1,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm8[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm0[0,2]
; SSE-NEXT:    andps %xmm6, %xmm4
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,0,1,1]
; SSE-NEXT:    pandn %xmm0, %xmm6
; SSE-NEXT:    por %xmm4, %xmm6
; SSE-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm2[0]
; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm2[1,3]
; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm11[0,2]
; SSE-NEXT:    andps %xmm9, %xmm8
; SSE-NEXT:    pslldq {{.*#+}} xmm15 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm15[0,1,2,3,4,5]
; SSE-NEXT:    pandn %xmm15, %xmm9
; SSE-NEXT:    por %xmm8, %xmm9
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movdqa %xmm9, (%rax)
; SSE-NEXT:    movdqa %xmm6, 16(%rax)
; SSE-NEXT:    movdqa %xmm5, 48(%rax)
; SSE-NEXT:    movdqa %xmm1, 64(%rax)
; SSE-NEXT:    movdqa %xmm3, 96(%rax)
; SSE-NEXT:    movdqa %xmm7, 112(%rax)
; SSE-NEXT:    movdqa %xmm10, 144(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 160(%rax)
; SSE-NEXT:    movaps %xmm12, 32(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 80(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 128(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 176(%rax)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf16:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vmovdqa (%rcx), %xmm8
; AVX1-NEXT:    vmovdqa 16(%rcx), %xmm0
; AVX1-NEXT:    vmovdqa (%rdx), %xmm9
; AVX1-NEXT:    vmovdqa 16(%rdx), %xmm1
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm12 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm12[2,2,3,3]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm10 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm10[0,0,1,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
; AVX1-NEXT:    vmovdqa (%rsi), %xmm15
; AVX1-NEXT:    vmovdqa 16(%rsi), %xmm2
; AVX1-NEXT:    vmovdqa (%rdi), %xmm11
; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm4
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm3[2,3,2,3]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[0,1,0,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm7, %ymm4
; AVX1-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2],ymm4[3,4],ymm1[5],ymm4[6,7]
; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm7
; AVX1-NEXT:    vmovdqa 16(%r8), %xmm4
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm7[0,1,2,3],xmm4[4,5],xmm7[6,7]
; AVX1-NEXT:    vmovdqa 16(%r9), %xmm7
; AVX1-NEXT:    vpslld $16, %xmm7, %xmm6
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm6[5],xmm0[6,7]
; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm4[2,1,3,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2],xmm0[3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm7[0,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5,6],xmm1[7]
; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm10[1,1,2,2]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm10[2,2,3,3]
; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm4[0,1,2,3,6,5,7,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3,4,5],xmm2[6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm7[0,1,2,3,4,6,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3,4,5,6],xmm2[7]
; AVX1-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpsrldq {{.*#+}} xmm1 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm7[2,2,3,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3],xmm0[4,5,6,7]
; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm13 = xmm11[4],xmm15[4],xmm11[5],xmm15[5],xmm11[6],xmm15[6],xmm11[7],xmm15[7]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm9[4],xmm8[4],xmm9[5],xmm8[5],xmm9[6],xmm8[6],xmm9[7],xmm8[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm1[1,1,2,2]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[2,2,3,3]
; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm13[2,3,2,3]
; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm13, %ymm2
; AVX1-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1],ymm2[2],ymm0[3,4],ymm2[5],ymm0[6,7]
; AVX1-NEXT:    vmovdqa (%r8), %xmm2
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,6,5,7,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
; AVX1-NEXT:    vextractf128 $1, %ymm14, %xmm6
; AVX1-NEXT:    vpblendw {{.*#+}} xmm6 = xmm0[0,1],xmm6[2,3,4,5],xmm0[6,7]
; AVX1-NEXT:    vmovdqa (%r9), %xmm0
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm0[0,1,2,3,4,6,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2,3,4,5,6],xmm5[7]
; AVX1-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpsrldq {{.*#+}} xmm6 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-NEXT:    vblendps {{.*#+}} xmm6 = xmm14[0],xmm6[1],xmm14[2,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm0[2,2,3,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm14 = xmm6[0,1,2],xmm5[3],xmm6[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm12[0,0,1,1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm12[1,1,2,2]
; AVX1-NEXT:    vinsertf128 $1, %xmm6, %ymm5, %ymm5
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm3[0,1,0,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm6, %ymm3
; AVX1-NEXT:    vblendps {{.*#+}} ymm5 = ymm3[0],ymm5[1],ymm3[2,3],ymm5[4],ymm3[5,6],ymm5[7]
; AVX1-NEXT:    vextractf128 $1, %ymm5, %xmm3
; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm6 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm6[2,3],xmm3[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm7[0,0,1,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm12 = xmm3[0,1,2],xmm6[3],xmm3[4,5,6,7]
; AVX1-NEXT:    vinsertps {{.*#+}} xmm4 = xmm5[0,1],xmm4[0],xmm5[3]
; AVX1-NEXT:    vpslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm7[0,1,2,3,4,5]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm10 = xmm4[0,1,2,3,4],xmm5[5],xmm4[6,7]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm5[0,0,1,1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm5[1,1,2,2]
; AVX1-NEXT:    vinsertf128 $1, %xmm7, %ymm6, %ymm6
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm7 = xmm11[0],xmm15[0],xmm11[1],xmm15[1],xmm11[2],xmm15[2],xmm11[3],xmm15[3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm8 = xmm7[0,1,0,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm7, %ymm8, %ymm8
; AVX1-NEXT:    vblendps {{.*#+}} ymm6 = ymm8[0],ymm6[1],ymm8[2,3],ymm6[4],ymm8[5,6],ymm6[7]
; AVX1-NEXT:    vextractf128 $1, %ymm6, %xmm3
; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[0,0,1,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3],xmm3[4,5,6,7]
; AVX1-NEXT:    vinsertps {{.*#+}} xmm4 = xmm6[0,1],xmm2[0],xmm6[3]
; AVX1-NEXT:    vpslldq {{.*#+}} xmm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm6[5],xmm4[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm5[2,2,3,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm5, %ymm1
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm7[2,3,2,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm13[0,1,0,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm6, %ymm5, %ymm5
; AVX1-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm1[2],ymm5[3,4],ymm1[5],ymm5[6,7]
; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm5
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3],xmm2[4,5],xmm5[6,7]
; AVX1-NEXT:    vpslld $16, %xmm0, %xmm6
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm6[5],xmm5[6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[2,1,3,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; AVX1-NEXT:    vblendps {{.*#+}} xmm1 = xmm2[0],xmm1[1,2],xmm2[3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3,4,5,6],xmm0[7]
; AVX1-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-NEXT:    vmovdqa %xmm0, 32(%rax)
; AVX1-NEXT:    vmovdqa %xmm5, 48(%rax)
; AVX1-NEXT:    vmovdqa %xmm4, (%rax)
; AVX1-NEXT:    vmovdqa %xmm3, 16(%rax)
; AVX1-NEXT:    vmovdqa %xmm10, 96(%rax)
; AVX1-NEXT:    vmovdqa %xmm12, 112(%rax)
; AVX1-NEXT:    vmovdqa %xmm14, 64(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 80(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 160(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 176(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 128(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 144(%rax)
; AVX1-NEXT:    vzeroupper
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf16:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %ymm8
; AVX2-SLOW-NEXT:    vmovdqa (%rsi), %ymm11
; AVX2-SLOW-NEXT:    vmovdqa (%rdx), %ymm15
; AVX2-SLOW-NEXT:    vmovdqa (%rcx), %ymm9
; AVX2-SLOW-NEXT:    vmovdqa (%r8), %ymm10
; AVX2-SLOW-NEXT:    vmovdqa (%rcx), %xmm5
; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-SLOW-NEXT:    vmovdqa (%rdx), %xmm6
; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} xmm1 = xmm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX2-SLOW-NEXT:    vpbroadcastq %xmm0, %ymm1
; AVX2-SLOW-NEXT:    vmovdqa (%rsi), %xmm7
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm7[0,1,2,1]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,7,6,5]
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[0,1,2,1]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,5]
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7]
; AVX2-SLOW-NEXT:    vmovdqa (%r8), %xmm1
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm1[2,1,3,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,2,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0],ymm2[1,2],ymm3[3],ymm2[4,5],ymm3[6],ymm2[7]
; AVX2-SLOW-NEXT:    vmovdqa (%r9), %xmm2
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm2[0,2,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[0,0,2,1]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm13 = [255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255]
; AVX2-SLOW-NEXT:    vpblendvb %ymm13, %ymm3, %ymm4, %ymm3
; AVX2-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} ymm4 = ymm9[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm9[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} ymm12 = ymm15[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm15[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} ymm4 = ymm12[0],ymm4[0],ymm12[1],ymm4[1],ymm12[2],ymm4[2],ymm12[3],ymm4[3],ymm12[8],ymm4[8],ymm12[9],ymm4[9],ymm12[10],ymm4[10],ymm12[11],ymm4[11]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm12 = ymm11[2,1,2,3,6,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm12[0,3,2,1,4,5,6,7,8,11,10,9,12,13,14,15]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm14 = ymm8[2,1,2,3,6,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm14 = ymm14[0,3,2,1,4,5,6,7,8,11,10,9,12,13,14,15]
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} ymm14 = ymm14[0],ymm12[0],ymm14[1],ymm12[1],ymm14[2],ymm12[2],ymm14[3],ymm12[3],ymm14[8],ymm12[8],ymm14[9],ymm12[9],ymm14[10],ymm12[10],ymm14[11],ymm12[11]
; AVX2-SLOW-NEXT:    vmovdqa (%r9), %ymm12
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[2,2,2,2]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[2,1,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm14[0,1],ymm4[2],ymm14[3,4],ymm4[5],ymm14[6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm14 = ymm10[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[2,2,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm14[0],ymm4[1,2],ymm14[3],ymm4[4,5],ymm14[6],ymm4[7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm14 = ymm12[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm14 = ymm14[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[2,2,2,3]
; AVX2-SLOW-NEXT:    vpblendvb %ymm13, %ymm4, %ymm14, %ymm3
; AVX2-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[1,1,1,1]
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm3 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[1,2,3,3]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,2,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm4 = xmm1[12,13,10,11,10,11,14,15,14,15,14,15,14,15,14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[0,1,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[2,3,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,2,2,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[0,1,0,1]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm13 = [255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-SLOW-NEXT:    vpblendvb %ymm13, %ymm3, %ymm4, %ymm14
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} ymm3 = ymm8[4],ymm11[4],ymm8[5],ymm11[5],ymm8[6],ymm11[6],ymm8[7],ymm11[7],ymm8[12],ymm11[12],ymm8[13],ymm11[13],ymm8[14],ymm11[14],ymm8[15],ymm11[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[3,3,3,3]
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} ymm4 = ymm15[4],ymm9[4],ymm15[5],ymm9[5],ymm15[6],ymm9[6],ymm15[7],ymm9[7],ymm15[12],ymm9[12],ymm15[13],ymm9[13],ymm15[14],ymm9[14],ymm15[15],ymm9[15]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm4 = ymm4[1,2,3,3,5,6,7,7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[2,2,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2],ymm4[3,4],ymm3[5],ymm4[6,7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm4 = ymm10[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,28,29,26,27,26,27,30,31,30,31,30,31,30,31,30,31]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[2,1,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5,6],ymm4[7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm4 = ymm12[2,3,2,3,6,7,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm4[0,2,2,1,4,5,6,7,8,10,10,9,12,13,14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[2,1,2,3]
; AVX2-SLOW-NEXT:    vpblendvb %ymm13, %ymm3, %ymm4, %ymm3
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,2,1]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[1,0,2,2]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[0,1,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3],ymm4[4],ymm0[5,6],ymm4[7]
; AVX2-SLOW-NEXT:    vpmovzxwd {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
; AVX2-SLOW-NEXT:    vpbroadcastq %xmm1, %ymm1
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm2[0,0,2,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpbroadcastq %xmm1, %ymm1
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255]
; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm8[0],ymm11[0],ymm8[1],ymm11[1],ymm8[2],ymm11[2],ymm8[3],ymm11[3],ymm8[8],ymm11[8],ymm8[9],ymm11[9],ymm8[10],ymm11[10],ymm8[11],ymm11[11]
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} ymm4 = ymm15[0],ymm9[0],ymm15[1],ymm9[1],ymm15[2],ymm9[2],ymm15[3],ymm9[3],ymm15[8],ymm9[8],ymm15[9],ymm9[9],ymm15[10],ymm9[10],ymm15[11],ymm9[11]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,3]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm4 = ymm4[1,0,2,2,5,4,6,6]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[2,1,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3],ymm4[4],ymm1[5,6],ymm4[7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm10[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[2,2,2,2]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm12[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[2,2,2,2]
; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm1, %ymm4, %ymm1
; AVX2-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-SLOW-NEXT:    vmovdqa %ymm1, 96(%rax)
; AVX2-SLOW-NEXT:    vmovdqa %ymm3, 160(%rax)
; AVX2-SLOW-NEXT:    vmovdqa %ymm0, (%rax)
; AVX2-SLOW-NEXT:    vmovdqa %ymm14, 64(%rax)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm0, 128(%rax)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm0, 32(%rax)
; AVX2-SLOW-NEXT:    vzeroupper
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-ALL-LABEL: vf16:
; AVX2-FAST-ALL:       # %bb.0:
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdi), %ymm9
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rsi), %ymm10
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdx), %ymm15
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rcx), %ymm4
; AVX2-FAST-ALL-NEXT:    vmovdqa (%r8), %ymm11
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rsi), %xmm5
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm0 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm0, %xmm5, %xmm1
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdi), %xmm6
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm0, %xmm6, %xmm0
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rcx), %xmm7
; AVX2-FAST-ALL-NEXT:    vpsrldq {{.*#+}} xmm2 = xmm7[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdx), %xmm0
; AVX2-FAST-ALL-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX2-FAST-ALL-NEXT:    vpbroadcastq %xmm2, %ymm2
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa (%r8), %xmm1
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm1[2,1,3,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,2,1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0],ymm2[1,2],ymm3[3],ymm2[4,5],ymm3[6],ymm2[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa (%r9), %xmm2
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm8 = xmm2[0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,0,2,1]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm13 = [255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm13, %ymm3, %ymm8, %ymm3
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm3 = <u,u,u,u,4,5,10,11,u,u,u,u,u,u,u,u,24,25,22,23,20,21,26,27,u,u,u,u,u,u,u,u>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm3, %ymm10, %ymm12
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm3, %ymm9, %ymm3
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} ymm3 = ymm3[0],ymm12[0],ymm3[1],ymm12[1],ymm3[2],ymm12[2],ymm3[3],ymm12[3],ymm3[8],ymm12[8],ymm3[9],ymm12[9],ymm3[10],ymm12[10],ymm3[11],ymm12[11]
; AVX2-FAST-ALL-NEXT:    vpsrldq {{.*#+}} ymm12 = ymm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm4[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-FAST-ALL-NEXT:    vpsrldq {{.*#+}} ymm14 = ymm15[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm15[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} ymm14 = ymm14[0],ymm12[0],ymm14[1],ymm12[1],ymm14[2],ymm12[2],ymm14[3],ymm12[3],ymm14[8],ymm12[8],ymm14[9],ymm12[9],ymm14[10],ymm12[10],ymm14[11],ymm12[11]
; AVX2-FAST-ALL-NEXT:    vmovdqa (%r9), %ymm12
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[2,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[2,2,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm14[2],ymm3[3,4],ymm14[5],ymm3[6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} ymm14 = ymm11[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[2,2,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm14[0],ymm3[1,2],ymm14[3],ymm3[4,5],ymm14[6],ymm3[7]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm14 = ymm12[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,20,21,20,21,22,23,24,25,24,25,24,25,24,25]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[2,2,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm13, %ymm3, %ymm14, %ymm3
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} xmm3 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm14 = <1,u,u,2,u,u,3,u>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm3, %ymm14, %ymm3
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} xmm14 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[1,1,1,1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm14[2],ymm3[3,4],ymm14[5],ymm3[6,7]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm14 = xmm1[12,13,10,11,10,11,14,15,14,15,14,15,14,15,14,15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0],ymm14[1],ymm3[2,3],ymm14[4],ymm3[5,6],ymm14[7]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm14 = xmm2[8,9,12,13,12,13,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm8 = [255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm8, %ymm3, %ymm14, %ymm14
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} ymm3 = ymm15[4],ymm4[4],ymm15[5],ymm4[5],ymm15[6],ymm4[6],ymm15[7],ymm4[7],ymm15[12],ymm4[12],ymm15[13],ymm4[13],ymm15[14],ymm4[14],ymm15[15],ymm4[15]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm13 = <5,u,u,6,u,u,7,u>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm3, %ymm13, %ymm3
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} ymm13 = ymm9[4],ymm10[4],ymm9[5],ymm10[5],ymm9[6],ymm10[6],ymm9[7],ymm10[7],ymm9[12],ymm10[12],ymm9[13],ymm10[13],ymm9[14],ymm10[14],ymm9[15],ymm10[15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[3,3,3,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm13[2],ymm3[3,4],ymm13[5],ymm3[6,7]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm13 = ymm11[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,28,29,26,27,26,27,30,31,30,31,30,31,30,31,30,31]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[2,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0],ymm13[1],ymm3[2,3],ymm13[4],ymm3[5,6],ymm13[7]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm13 = ymm12[u,u,u,u,u,u,u,u,8,9,10,11,12,13,14,15,24,25,28,29,28,29,26,27,24,25,26,27,28,29,30,31]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[2,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm8, %ymm3, %ymm13, %ymm3
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
; AVX2-FAST-ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm7 = [1,0,2,2,1,0,2,2]
; AVX2-FAST-ALL-NEXT:    # ymm7 = mem[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm7, %ymm0
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[0,0,2,1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0],ymm0[1],ymm5[2,3],ymm0[4],ymm5[5,6],ymm0[7]
; AVX2-FAST-ALL-NEXT:    vpmovzxwd {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
; AVX2-FAST-ALL-NEXT:    vpbroadcastq %xmm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm2[0,0,2,1,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpbroadcastq %xmm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm15[0],ymm4[0],ymm15[1],ymm4[1],ymm15[2],ymm4[2],ymm15[3],ymm4[3],ymm15[8],ymm4[8],ymm15[9],ymm4[9],ymm15[10],ymm4[10],ymm15[11],ymm4[11]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm4 = <u,4,u,u,5,u,u,6>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm1, %ymm4, %ymm1
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} ymm4 = ymm9[0],ymm10[0],ymm9[1],ymm10[1],ymm9[2],ymm10[2],ymm9[3],ymm10[3],ymm9[8],ymm10[8],ymm9[9],ymm10[9],ymm9[10],ymm10[10],ymm9[11],ymm10[11]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[2,2,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0],ymm1[1],ymm4[2,3],ymm1[4],ymm4[5,6],ymm1[7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm11[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[2,2,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm12[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[2,2,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm2, %ymm1, %ymm4, %ymm1
; AVX2-FAST-ALL-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm1, 96(%rax)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm3, 160(%rax)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm0, (%rax)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm14, 64(%rax)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm0, 128(%rax)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm0, 32(%rax)
; 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 (%rdi), %ymm15
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rsi), %ymm10
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdx), %ymm13
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rcx), %ymm11
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%r8), %ymm12
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rsi), %xmm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm0 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm0, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdi), %xmm5
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm0, %xmm5, %xmm0
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rcx), %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpsrldq {{.*#+}} xmm2 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdx), %xmm7
; AVX2-FAST-PERLANE-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm7[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq %xmm2, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%r8), %xmm9
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm9[2,1,3,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,2,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0],ymm2[1,2],ymm3[3],ymm2[4,5],ymm3[6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%r9), %xmm6
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm8 = xmm6[0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,0,2,1]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm2, %ymm3, %ymm8, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm3 = <u,u,u,u,4,5,10,11,u,u,u,u,u,u,u,u,24,25,22,23,20,21,26,27,u,u,u,u,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm3, %ymm10, %ymm14
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm15, %ymm8
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm3, %ymm15, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} ymm3 = ymm3[0],ymm14[0],ymm3[1],ymm14[1],ymm3[2],ymm14[2],ymm3[3],ymm14[3],ymm3[8],ymm14[8],ymm3[9],ymm14[9],ymm3[10],ymm14[10],ymm3[11],ymm14[11]
; AVX2-FAST-PERLANE-NEXT:    vpsrldq {{.*#+}} ymm14 = ymm11[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm11[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-FAST-PERLANE-NEXT:    vpsrldq {{.*#+}} ymm15 = ymm13[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm13[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} ymm14 = ymm15[0],ymm14[0],ymm15[1],ymm14[1],ymm15[2],ymm14[2],ymm15[3],ymm14[3],ymm15[8],ymm14[8],ymm15[9],ymm14[9],ymm15[10],ymm14[10],ymm15[11],ymm14[11]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%r9), %ymm15
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[2,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm14[2],ymm3[3,4],ymm14[5],ymm3[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm12, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} ymm14 = ymm12[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[2,2,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm3 = ymm14[0],ymm3[1,2],ymm14[3],ymm3[4,5],ymm14[6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm14 = ymm15[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,20,21,20,21,22,23,24,25,24,25,24,25,24,25]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[2,2,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm2, %ymm3, %ymm14, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} xmm12 = xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[1,1,1,1]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm4, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm7[4],xmm4[4],xmm7[5],xmm4[5],xmm7[6],xmm4[6],xmm7[7],xmm4[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[1,2,3,3]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[0,0,2,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm12[2],ymm4[3,4],ymm12[5],ymm4[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm12 = xmm9[12,13,10,11,10,11,14,15,14,15,14,15,14,15,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0],ymm12[1],ymm4[2,3],ymm12[4],ymm4[5,6],ymm12[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm12 = xmm6[8,9,12,13,12,13,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm14 = [255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm14, %ymm4, %ymm12, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} ymm12 = ymm8[4],ymm10[4],ymm8[5],ymm10[5],ymm8[6],ymm10[6],ymm8[7],ymm10[7],ymm8[12],ymm10[12],ymm8[13],ymm10[13],ymm8[14],ymm10[14],ymm8[15],ymm10[15]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm8, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} ymm8 = ymm13[4],ymm11[4],ymm13[5],ymm11[5],ymm13[6],ymm11[6],ymm13[7],ymm11[7],ymm13[12],ymm11[12],ymm13[13],ymm11[13],ymm13[14],ymm11[14],ymm13[15],ymm11[15]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm8 = ymm8[1,2,3,3,5,6,7,7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[2,2,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1],ymm12[2],ymm8[3,4],ymm12[5],ymm8[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm12 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,28,29,26,27,26,27,30,31,30,31,30,31,30,31,30,31]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm0, %ymm4
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[2,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0],ymm12[1],ymm8[2,3],ymm12[4],ymm8[5,6],ymm12[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm12 = ymm15[u,u,u,u,u,u,u,u,8,9,10,11,12,13,14,15,24,25,28,29,28,29,26,27,24,25,26,27,28,29,30,31]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[2,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm14, %ymm8, %ymm12, %ymm8
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm7[0],xmm1[0],xmm7[1],xmm1[1],xmm7[2],xmm1[2],xmm7[3],xmm1[3]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[0,0,2,1]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,0,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0],ymm0[1],ymm5[2,3],ymm0[4],ymm5[5,6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT:    vpmovzxwd {{.*#+}} xmm1 = xmm9[0],zero,xmm9[1],zero,xmm9[2],zero,xmm9[3],zero
; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq %xmm1, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm6[0,0,2,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq %xmm1, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm6 = [255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm6, %ymm0, %ymm1, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm3[0],ymm10[0],ymm3[1],ymm10[1],ymm3[2],ymm10[2],ymm3[3],ymm10[3],ymm3[8],ymm10[8],ymm3[9],ymm10[9],ymm3[10],ymm10[10],ymm3[11],ymm10[11]
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} ymm5 = ymm13[0],ymm11[0],ymm13[1],ymm11[1],ymm13[2],ymm11[2],ymm13[3],ymm11[3],ymm13[8],ymm11[8],ymm13[9],ymm11[9],ymm13[10],ymm11[10],ymm13[11],ymm11[11]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm5 = ymm5[1,0,2,2,5,4,6,6]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[2,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3],ymm5[4],ymm1[5,6],ymm5[7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm4[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm5[2],ymm1[3,4],ymm5[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm15[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm6, %ymm1, %ymm5, %ymm1
; AVX2-FAST-PERLANE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm1, 96(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm8, 160(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, 128(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm0, (%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 64(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 32(%rax)
; AVX2-FAST-PERLANE-NEXT:    vzeroupper
; AVX2-FAST-PERLANE-NEXT:    retq
;
; AVX512-LABEL: vf16:
; AVX512:       # %bb.0:
; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-NEXT:    vmovdqa (%rdi), %ymm0
; AVX512-NEXT:    vmovdqa (%rdx), %ymm1
; AVX512-NEXT:    vmovdqa (%r8), %ymm2
; AVX512-NEXT:    vinserti64x4 $1, (%rsi), %zmm0, %zmm0
; AVX512-NEXT:    vinserti64x4 $1, (%rcx), %zmm1, %zmm1
; AVX512-NEXT:    vinserti64x4 $1, (%r9), %zmm2, %zmm2
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm3 = <0,16,32,48,u,u,1,17,33,49,u,u,2,18,34,50,u,u,3,19,35,51,u,u,4,20,36,52,u,u,5,21>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm0, %zmm3
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm4 = [0,1,2,3,32,48,6,7,8,9,33,49,12,13,14,15,34,50,18,19,20,21,35,51,24,25,26,27,36,52,30,31]
; AVX512-NEXT:    vpermi2w %zmm2, %zmm3, %zmm4
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm3 = <5,21,u,u,38,54,6,22,u,u,39,55,7,23,u,u,40,56,8,24,u,u,41,57,9,25,u,u,42,58,10,26>
; AVX512-NEXT:    vpermi2w %zmm0, %zmm1, %zmm3
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm5 = [0,1,37,53,4,5,6,7,38,54,10,11,12,13,39,55,16,17,18,19,40,56,22,23,24,25,41,57,28,29,30,31]
; AVX512-NEXT:    vpermi2w %zmm2, %zmm3, %zmm5
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm3 = <u,u,11,27,43,59,u,u,12,28,44,60,u,u,13,29,45,61,u,u,14,30,46,62,u,u,15,31,47,63,u,u>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm0, %zmm3
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [42,58,2,3,4,5,43,59,8,9,10,11,44,60,14,15,16,17,45,61,20,21,22,23,46,62,26,27,28,29,47,63]
; AVX512-NEXT:    vpermi2w %zmm2, %zmm3, %zmm0
; AVX512-NEXT:    vmovdqu64 %zmm0, 128(%rax)
; AVX512-NEXT:    vmovdqu64 %zmm5, 64(%rax)
; AVX512-NEXT:    vmovdqu64 %zmm4, (%rax)
; 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
  %in.vec3 = load <16 x i16>, ptr %in.vecptr3, align 32
  %in.vec4 = load <16 x i16>, ptr %in.vecptr4, align 32
  %in.vec5 = load <16 x i16>, ptr %in.vecptr5, 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>
  %concat23 = shufflevector <16 x i16> %in.vec2, <16 x i16> %in.vec3, <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>
  %concat45 = shufflevector <16 x i16> %in.vec4, <16 x i16> %in.vec5, <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>
  %concat0123 = shufflevector <32 x i16> %concat01, <32 x i16> %concat23, <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>
  %concat45uu = shufflevector <32 x i16> %concat45, <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>
  %concat012345 = shufflevector <64 x i16> %concat0123, <64 x i16> %concat45uu, <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> %concat012345, <96 x i16> poison, <96 x i32> <i32 0, i32 16, i32 32, i32 48, i32 64, i32 80, i32 1, i32 17, i32 33, i32 49, i32 65, i32 81, i32 2, i32 18, i32 34, i32 50, i32 66, i32 82, i32 3, i32 19, i32 35, i32 51, i32 67, i32 83, i32 4, i32 20, i32 36, i32 52, i32 68, i32 84, i32 5, i32 21, i32 37, i32 53, i32 69, i32 85, i32 6, i32 22, i32 38, i32 54, i32 70, i32 86, i32 7, i32 23, i32 39, i32 55, i32 71, i32 87, i32 8, i32 24, i32 40, i32 56, i32 72, i32 88, i32 9, i32 25, i32 41, i32 57, i32 73, i32 89, i32 10, i32 26, i32 42, i32 58, i32 74, i32 90, i32 11, i32 27, i32 43, i32 59, i32 75, i32 91, i32 12, i32 28, i32 44, i32 60, i32 76, i32 92, i32 13, i32 29, i32 45, i32 61, i32 77, i32 93, i32 14, i32 30, i32 46, i32 62, i32 78, i32 94, i32 15, i32 31, i32 47, i32 63, i32 79, i32 95>

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

  ret void
}

define void @vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %out.vec) nounwind {
; SSE-LABEL: vf32:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $296, %rsp # imm = 0x128
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    movdqa 16(%rdi), %xmm13
; SSE-NEXT:    movdqa (%rsi), %xmm2
; SSE-NEXT:    movdqa 16(%rsi), %xmm14
; SSE-NEXT:    movdqa (%rdx), %xmm1
; SSE-NEXT:    movdqa 16(%rdx), %xmm9
; SSE-NEXT:    movdqa (%rcx), %xmm7
; SSE-NEXT:    movdqa 16(%rcx), %xmm15
; SSE-NEXT:    movdqa (%r8), %xmm5
; SSE-NEXT:    movdqa (%r9), %xmm12
; SSE-NEXT:    movdqa %xmm1, %xmm4
; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1],xmm4[2],xmm7[2],xmm4[3],xmm7[3]
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm0, %xmm8
; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1],xmm8[2],xmm2[2],xmm8[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm8, %xmm3
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,3],xmm4[3,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm5[2,1,3,3,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,2],xmm4[0,1]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0,1,3]
; SSE-NEXT:    movaps {{.*#+}} xmm10 = [65535,0,65535,65535,65535,65535,65535,0]
; SSE-NEXT:    andps %xmm10, %xmm3
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm12[0,2,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,2,1]
; SSE-NEXT:    movaps %xmm10, %xmm4
; SSE-NEXT:    andnps %xmm6, %xmm4
; SSE-NEXT:    orps %xmm3, %xmm4
; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm0, %xmm2
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,3],xmm1[3,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,6,5,7,7]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,2],xmm3[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0,1,3]
; SSE-NEXT:    andps %xmm10, %xmm2
; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm12[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,2,3]
; SSE-NEXT:    movaps %xmm10, %xmm4
; SSE-NEXT:    andnps %xmm3, %xmm4
; SSE-NEXT:    orps %xmm2, %xmm4
; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm9, %xmm0
; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1],xmm0[2],xmm15[2],xmm0[3],xmm15[3]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm13, %xmm11
; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm14[0],xmm11[1],xmm14[1],xmm11[2],xmm14[2],xmm11[3],xmm14[3]
; SSE-NEXT:    movdqa %xmm11, %xmm3
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,3],xmm0[3,3]
; SSE-NEXT:    movdqa 16(%r8), %xmm4
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm4[2,1,3,3,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,2],xmm6[0,1]
; SSE-NEXT:    movdqa 16(%r9), %xmm0
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,2,1]
; SSE-NEXT:    movaps %xmm10, %xmm7
; SSE-NEXT:    andnps %xmm6, %xmm7
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0,1,3]
; SSE-NEXT:    andps %xmm10, %xmm3
; SSE-NEXT:    orps %xmm3, %xmm7
; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm15[4],xmm9[5],xmm15[5],xmm9[6],xmm15[6],xmm9[7],xmm15[7]
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm13, %xmm3
; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm14[4],xmm3[5],xmm14[5],xmm3[6],xmm14[6],xmm3[7],xmm14[7]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,3],xmm9[3,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm4[0,1,2,3,6,5,7,7]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,2],xmm6[2,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm0[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,2,3]
; SSE-NEXT:    movaps %xmm10, %xmm0
; SSE-NEXT:    andnps %xmm6, %xmm0
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0,1,3]
; SSE-NEXT:    andps %xmm10, %xmm3
; SSE-NEXT:    orps %xmm3, %xmm0
; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 32(%rdx), %xmm2
; SSE-NEXT:    movdqa 32(%rcx), %xmm15
; SSE-NEXT:    movdqa %xmm2, %xmm0
; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1],xmm0[2],xmm15[2],xmm0[3],xmm15[3]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 32(%rdi), %xmm9
; SSE-NEXT:    movdqa 32(%rsi), %xmm14
; SSE-NEXT:    movdqa %xmm9, %xmm3
; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm14[0],xmm3[1],xmm14[1],xmm3[2],xmm14[2],xmm3[3],xmm14[3]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,3],xmm0[3,3]
; SSE-NEXT:    movdqa 32(%r8), %xmm6
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[2,1,3,3,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,2],xmm0[0,1]
; SSE-NEXT:    movdqa 32(%r9), %xmm1
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
; SSE-NEXT:    movdqa %xmm1, %xmm7
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
; SSE-NEXT:    movaps %xmm10, %xmm1
; SSE-NEXT:    andnps %xmm0, %xmm1
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0,1,3]
; SSE-NEXT:    andps %xmm10, %xmm3
; SSE-NEXT:    orps %xmm3, %xmm1
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm9, %xmm0
; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm14[4],xmm0[5],xmm14[5],xmm0[6],xmm14[6],xmm0[7],xmm14[7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm2[3,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm6[0,1,2,3,6,5,7,7]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,2],xmm3[2,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm7[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,2,3]
; SSE-NEXT:    movaps %xmm10, %xmm1
; SSE-NEXT:    andnps %xmm3, %xmm1
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0,1,3]
; SSE-NEXT:    andps %xmm10, %xmm0
; SSE-NEXT:    orps %xmm0, %xmm1
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 48(%rdx), %xmm0
; SSE-NEXT:    movdqa 48(%rcx), %xmm3
; SSE-NEXT:    movdqa %xmm0, %xmm1
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 48(%rdi), %xmm2
; SSE-NEXT:    movdqa 48(%rsi), %xmm14
; SSE-NEXT:    movdqa %xmm2, %xmm15
; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm14[0],xmm15[1],xmm14[1],xmm15[2],xmm14[2],xmm15[3],xmm14[3]
; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[2,3],xmm1[3,3]
; SSE-NEXT:    movdqa 48(%r8), %xmm9
; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm9[2,1,3,3,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[1,2],xmm13[0,1]
; SSE-NEXT:    movdqa 48(%r9), %xmm1
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm1[0,2,2,3,4,5,6,7]
; SSE-NEXT:    movdqa %xmm1, %xmm13
; SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,2,1]
; SSE-NEXT:    movaps %xmm10, %xmm1
; SSE-NEXT:    andnps %xmm7, %xmm1
; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[2,0,1,3]
; SSE-NEXT:    andps %xmm10, %xmm15
; SSE-NEXT:    orps %xmm15, %xmm1
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm0, %xmm1
; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm14[4],xmm2[5],xmm14[5],xmm2[6],xmm14[6],xmm2[7],xmm14[7]
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,3],xmm1[3,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm9[0,1,2,3,6,5,7,7]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,2],xmm7[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0,1,3]
; SSE-NEXT:    andps %xmm10, %xmm2
; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm13[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[2,1,2,3]
; SSE-NEXT:    andnps %xmm7, %xmm10
; SSE-NEXT:    orps %xmm2, %xmm10
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm2, %xmm0
; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm8[0]
; SSE-NEXT:    movdqa %xmm5, %xmm7
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm8[1,3]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[0,2]
; SSE-NEXT:    movdqa %xmm12, %xmm7
; SSE-NEXT:    pslldq {{.*#+}} xmm7 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm7[0,1,2,3,4,5]
; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [65535,65535,65535,65535,65535,0,65535,65535]
; SSE-NEXT:    movdqa %xmm14, %xmm1
; SSE-NEXT:    pandn %xmm7, %xmm1
; SSE-NEXT:    andps %xmm14, %xmm0
; SSE-NEXT:    por %xmm0, %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhqdq {{.*#+}} xmm8 = xmm8[1],xmm2[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm5[1,1,1,1,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,1],xmm2[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm8[0,2]
; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,65535,65535,0,65535,65535,65535,65535]
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm12[0,0,1,1]
; SSE-NEXT:    movdqa %xmm15, %xmm0
; SSE-NEXT:    pandn %xmm8, %xmm0
; SSE-NEXT:    andps %xmm15, %xmm7
; SSE-NEXT:    por %xmm7, %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, %xmm7
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm0[0]
; SSE-NEXT:    movdqa %xmm5, %xmm1
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,1],xmm0[1,3]
; SSE-NEXT:    movaps %xmm0, %xmm2
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm1[0,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,2,3,3]
; SSE-NEXT:    pslld $16, %xmm12
; SSE-NEXT:    movdqa %xmm14, %xmm0
; SSE-NEXT:    pandn %xmm12, %xmm0
; SSE-NEXT:    andps %xmm14, %xmm7
; SSE-NEXT:    por %xmm7, %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
; SSE-NEXT:    psrldq {{.*#+}} xmm5 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm3[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm2[0,2]
; SSE-NEXT:    movdqa %xmm15, %xmm0
; SSE-NEXT:    pandn %xmm1, %xmm0
; SSE-NEXT:    andps %xmm15, %xmm5
; SSE-NEXT:    por %xmm5, %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm2, %xmm1
; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm11[0]
; SSE-NEXT:    movdqa %xmm4, %xmm5
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm11[1,3]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm5[0,2]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm0, %xmm5
; SSE-NEXT:    pslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5]
; SSE-NEXT:    movdqa %xmm14, %xmm3
; SSE-NEXT:    pandn %xmm5, %xmm3
; SSE-NEXT:    andps %xmm14, %xmm1
; SSE-NEXT:    por %xmm1, %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhqdq {{.*#+}} xmm11 = xmm11[1],xmm2[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm4[1,1,1,1,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm2[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm11[0,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm0[0,0,1,1]
; SSE-NEXT:    movdqa %xmm0, %xmm3
; SSE-NEXT:    movdqa %xmm15, %xmm0
; SSE-NEXT:    pandn %xmm7, %xmm0
; SSE-NEXT:    andps %xmm15, %xmm1
; SSE-NEXT:    por %xmm1, %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    movaps %xmm5, %xmm1
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT:    movdqa %xmm4, %xmm7
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,1],xmm0[1,3]
; SSE-NEXT:    movaps %xmm0, %xmm2
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm7[0,2]
; SSE-NEXT:    movdqa %xmm3, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,2,3,3]
; SSE-NEXT:    pslld $16, %xmm0
; SSE-NEXT:    movdqa %xmm14, %xmm7
; SSE-NEXT:    pandn %xmm0, %xmm7
; SSE-NEXT:    andps %xmm14, %xmm1
; SSE-NEXT:    por %xmm1, %xmm7
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm5[1]
; SSE-NEXT:    psrldq {{.*#+}} xmm4 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm5[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm2[0,2]
; SSE-NEXT:    movdqa %xmm15, %xmm8
; SSE-NEXT:    pandn %xmm3, %xmm8
; SSE-NEXT:    andps %xmm15, %xmm4
; SSE-NEXT:    por %xmm4, %xmm8
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    movaps %xmm4, %xmm1
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT:    movdqa %xmm6, %xmm2
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[1,3]
; SSE-NEXT:    movaps %xmm0, %xmm3
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[0,2]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm0, %xmm5
; SSE-NEXT:    pslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5]
; SSE-NEXT:    movdqa %xmm14, %xmm7
; SSE-NEXT:    pandn %xmm5, %xmm7
; SSE-NEXT:    andps %xmm14, %xmm1
; SSE-NEXT:    por %xmm1, %xmm7
; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm4[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[1,1,1,1,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm4[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[0,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[0,0,1,1]
; SSE-NEXT:    movdqa %xmm0, %xmm3
; SSE-NEXT:    movdqa %xmm15, %xmm11
; SSE-NEXT:    pandn %xmm5, %xmm11
; SSE-NEXT:    andps %xmm15, %xmm1
; SSE-NEXT:    por %xmm1, %xmm11
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    movaps %xmm4, %xmm5
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm0[0]
; SSE-NEXT:    movdqa %xmm6, %xmm1
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,1],xmm0[1,3]
; SSE-NEXT:    movaps %xmm0, %xmm2
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm1[0,2]
; SSE-NEXT:    movdqa %xmm3, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm3[2,2,3,3]
; SSE-NEXT:    pslld $16, %xmm0
; SSE-NEXT:    movdqa %xmm14, %xmm3
; SSE-NEXT:    pandn %xmm0, %xmm3
; SSE-NEXT:    andps %xmm14, %xmm5
; SSE-NEXT:    por %xmm5, %xmm3
; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm4[1]
; SSE-NEXT:    psrldq {{.*#+}} xmm6 = xmm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,1],xmm4[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm2[0,2]
; SSE-NEXT:    movdqa %xmm15, %xmm12
; SSE-NEXT:    pandn %xmm13, %xmm12
; SSE-NEXT:    andps %xmm15, %xmm6
; SSE-NEXT:    por %xmm6, %xmm12
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, %xmm5
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm0[0]
; SSE-NEXT:    movdqa %xmm9, %xmm6
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[1,3]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm6[0,2]
; SSE-NEXT:    movdqa (%rsp), %xmm4 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm4, %xmm6
; SSE-NEXT:    pslldq {{.*#+}} xmm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm6[0,1,2,3,4,5]
; SSE-NEXT:    movdqa %xmm14, %xmm13
; SSE-NEXT:    pandn %xmm6, %xmm13
; SSE-NEXT:    andps %xmm14, %xmm5
; SSE-NEXT:    por %xmm5, %xmm13
; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm9[1,1,1,1,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm1[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm0[0,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,0,1,1]
; SSE-NEXT:    movdqa %xmm15, %xmm6
; SSE-NEXT:    pandn %xmm1, %xmm6
; SSE-NEXT:    andps %xmm15, %xmm5
; SSE-NEXT:    por %xmm5, %xmm6
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    movaps %xmm2, %xmm1
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT:    movdqa %xmm9, %xmm5
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm0[1,3]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm5[0,2]
; SSE-NEXT:    andps %xmm14, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[2,2,3,3]
; SSE-NEXT:    pslld $16, %xmm4
; SSE-NEXT:    pandn %xmm4, %xmm14
; SSE-NEXT:    por %xmm1, %xmm14
; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; SSE-NEXT:    psrldq {{.*#+}} xmm9 = xmm9[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,1],xmm2[1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0],xmm0[0,2]
; SSE-NEXT:    andps %xmm15, %xmm9
; SSE-NEXT:    pandn %xmm5, %xmm15
; SSE-NEXT:    por %xmm9, %xmm15
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movdqa %xmm15, 352(%rax)
; SSE-NEXT:    movdqa %xmm14, 336(%rax)
; SSE-NEXT:    movdqa %xmm6, 304(%rax)
; SSE-NEXT:    movdqa %xmm13, 288(%rax)
; SSE-NEXT:    movdqa %xmm12, 256(%rax)
; SSE-NEXT:    movdqa %xmm3, 240(%rax)
; SSE-NEXT:    movdqa %xmm11, 208(%rax)
; SSE-NEXT:    movdqa %xmm7, 192(%rax)
; SSE-NEXT:    movdqa %xmm8, 160(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 144(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 112(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 96(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 64(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rax)
; SSE-NEXT:    movaps %xmm10, 368(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 320(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 272(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 224(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 176(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 128(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 80(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%rax)
; SSE-NEXT:    addq $296, %rsp # imm = 0x128
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf32:
; AVX1:       # %bb.0:
; AVX1-NEXT:    subq $184, %rsp
; AVX1-NEXT:    vmovdqa 32(%rcx), %xmm8
; AVX1-NEXT:    vmovdqa 48(%rcx), %xmm0
; AVX1-NEXT:    vmovdqa 32(%rdx), %xmm10
; AVX1-NEXT:    vmovdqa 48(%rdx), %xmm1
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm13 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm13[2,2,3,3]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm1[0,0,1,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
; AVX1-NEXT:    vmovdqa 32(%rsi), %xmm11
; AVX1-NEXT:    vmovdqa 48(%rsi), %xmm2
; AVX1-NEXT:    vmovdqa 32(%rdi), %xmm12
; AVX1-NEXT:    vmovdqa 48(%rdi), %xmm6
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm14 = xmm6[0],xmm2[0],xmm6[1],xmm2[1],xmm6[2],xmm2[2],xmm6[3],xmm2[3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm14[2,3,2,3]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm6[4],xmm2[4],xmm6[5],xmm2[5],xmm6[6],xmm2[6],xmm6[7],xmm2[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm2[0,1,0,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm6, %ymm7, %ymm6
; AVX1-NEXT:    vblendps {{.*#+}} ymm7 = ymm6[0,1],ymm0[2],ymm6[3,4],ymm0[5],ymm6[6,7]
; AVX1-NEXT:    vmovdqa 48(%r8), %xmm6
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm6[2,1,3,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
; AVX1-NEXT:    vblendps {{.*#+}} xmm3 = xmm0[0],xmm7[1,2],xmm0[3]
; AVX1-NEXT:    vmovdqa 48(%r9), %xmm0
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm0[0,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3,4,5,6],xmm4[7]
; AVX1-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vextractf128 $1, %ymm7, %xmm3
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm6[4,5],xmm3[6,7]
; AVX1-NEXT:    vpslld $16, %xmm0, %xmm4
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm4[5],xmm3[6,7]
; AVX1-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[2,3,2,3]
; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm1[1,1,2,2]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
; AVX1-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
; AVX1-NEXT:    vpsrldq {{.*#+}} xmm2 = xmm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0],xmm2[1],xmm1[2,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[2,2,3,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3],xmm2[4,5,6,7]
; AVX1-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm6[0,1,2,3,6,5,7,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3,4,5],xmm2[6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,6,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3,4,5,6],xmm2[7]
; AVX1-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm9 = xmm12[4],xmm11[4],xmm12[5],xmm11[5],xmm12[6],xmm11[6],xmm12[7],xmm11[7]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm10[4],xmm8[4],xmm10[5],xmm8[5],xmm10[6],xmm8[6],xmm10[7],xmm8[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[1,1,2,2]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[2,2,3,3]
; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm9[2,3,2,3]
; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm9, %ymm3
; AVX1-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7]
; AVX1-NEXT:    vmovdqa 32(%r8), %xmm7
; AVX1-NEXT:    vpsrldq {{.*#+}} xmm1 = xmm7[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-NEXT:    vblendps {{.*#+}} xmm4 = xmm3[0],xmm1[1],xmm3[2,3]
; AVX1-NEXT:    vmovdqa 32(%r9), %xmm1
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm1[2,2,3,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2],xmm5[3],xmm4[4,5,6,7]
; AVX1-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm3
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm7[0,1,2,3,6,5,7,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3,4,5],xmm4[6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm1[0,1,2,3,4,6,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3,4,5,6],xmm4[7]
; AVX1-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm13[0,0,1,1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm13[1,1,2,2]
; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm14[0,1,0,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm14, %ymm4, %ymm4
; AVX1-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7]
; AVX1-NEXT:    vinsertps {{.*#+}} xmm4 = xmm3[0,1],xmm6[0],xmm3[3]
; AVX1-NEXT:    vpslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm5[5],xmm4[6,7]
; AVX1-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm3
; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm4 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm3[0,1,2],xmm0[3],xmm3[4,5,6,7]
; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm10[0],xmm8[0],xmm10[1],xmm8[1],xmm10[2],xmm8[2],xmm10[3],xmm8[3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[0,0,1,1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[1,1,2,2]
; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm12[0],xmm11[0],xmm12[1],xmm11[1],xmm12[2],xmm11[2],xmm12[3],xmm11[3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm4[0,1,0,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm5, %ymm5
; AVX1-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0],ymm3[1],ymm5[2,3],ymm3[4],ymm5[5,6],ymm3[7]
; AVX1-NEXT:    vinsertps {{.*#+}} xmm5 = xmm3[0,1],xmm7[0],xmm3[3]
; AVX1-NEXT:    vpslldq {{.*#+}} xmm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4,5]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm6[5],xmm5[6,7]
; AVX1-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm3
; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm5 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm5[2,3],xmm3[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm1[0,0,1,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm5[3],xmm3[4,5,6,7]
; AVX1-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-NEXT:    vmovdqa 16(%rcx), %xmm2
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm4[2,3,2,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm9[0,1,0,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2],ymm3[3,4],ymm0[5],ymm3[6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm7[2,1,3,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
; AVX1-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0],xmm0[1,2],xmm3[3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm1[0,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3,4,5,6],xmm4[7]
; AVX1-NEXT:    vmovdqa %xmm3, (%rsp) # 16-byte Spill
; AVX1-NEXT:    vmovdqa 16(%rdx), %xmm3
; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm7[4,5],xmm0[6,7]
; AVX1-NEXT:    vpslld $16, %xmm1, %xmm1
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm1[5],xmm0[6,7]
; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[0,0,1,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-NEXT:    vmovdqa 16(%rsi), %xmm2
; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm3
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm5[2,3,2,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm4[0,1,0,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2],ymm2[3,4],ymm0[5],ymm2[6,7]
; AVX1-NEXT:    vmovdqa 16(%r8), %xmm3
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm3[2,1,3,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; AVX1-NEXT:    vblendps {{.*#+}} xmm5 = xmm2[0],xmm0[1,2],xmm2[3]
; AVX1-NEXT:    vmovdqa 16(%r9), %xmm2
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm2[0,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm6[0,1,2,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0],xmm6[1],xmm5[2,3,4,5,6],xmm6[7]
; AVX1-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm3[4,5],xmm0[6,7]
; AVX1-NEXT:    vpslld $16, %xmm2, %xmm5
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5],xmm0[6,7]
; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm4, %ymm0
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm1[1,1,2,2]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm4, %ymm1
; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
; AVX1-NEXT:    vpsrldq {{.*#+}} xmm1 = xmm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-NEXT:    vblendps {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[2,2,3,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3],xmm1[4,5,6,7]
; AVX1-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,6,5,7,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3,4,5],xmm1[6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,6,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5,6],xmm1[7]
; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vmovdqa (%rcx), %xmm14
; AVX1-NEXT:    vmovdqa (%rdx), %xmm11
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm10 = xmm11[4],xmm14[4],xmm11[5],xmm14[5],xmm11[6],xmm14[6],xmm11[7],xmm14[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm10[1,1,2,2]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm10[2,2,3,3]
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX1-NEXT:    vmovdqa (%rsi), %xmm9
; AVX1-NEXT:    vmovdqa (%rdi), %xmm8
; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm8[4],xmm9[4],xmm8[5],xmm9[5],xmm8[6],xmm9[6],xmm8[7],xmm9[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm4, %ymm1
; AVX1-NEXT:    vblendps {{.*#+}} ymm12 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
; AVX1-NEXT:    vmovdqa (%r8), %xmm1
; AVX1-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-NEXT:    vblendps {{.*#+}} xmm13 = xmm12[0],xmm0[1],xmm12[2,3]
; AVX1-NEXT:    vmovdqa (%r9), %xmm0
; AVX1-NEXT:    vpshufd {{.*#+}} xmm15 = xmm0[2,2,3,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm13[0,1,2],xmm15[3],xmm13[4,5,6,7]
; AVX1-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vextractf128 $1, %ymm12, %xmm12
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm1[0,1,2,3,6,5,7,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm7[2,1,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1],xmm12[2,3,4,5],xmm7[6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm0[0,1,2,3,4,6,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm6[2,1,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm13 = xmm7[0],xmm6[1],xmm7[2,3,4,5,6],xmm6[7]
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-NEXT:    vpermilps {{.*#+}} xmm6 = xmm5[0,0,1,1]
; AVX1-NEXT:    vpermilps {{.*#+}} xmm7 = xmm5[1,1,2,2]
; AVX1-NEXT:    vinsertf128 $1, %xmm7, %ymm6, %ymm6
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-NEXT:    vpermilps {{.*#+}} xmm7 = xmm5[0,1,0,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm5, %ymm7, %ymm7
; AVX1-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0],ymm6[1],ymm7[2,3],ymm6[4],ymm7[5,6],ymm6[7]
; AVX1-NEXT:    vinsertps {{.*#+}} xmm7 = xmm6[0,1],xmm3[0],xmm6[3]
; AVX1-NEXT:    vpslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm15 = xmm7[0,1,2,3,4],xmm5[5],xmm7[6,7]
; AVX1-NEXT:    vextractf128 $1, %ymm6, %xmm5
; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3],xmm5[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm12 = xmm3[0,1,2],xmm2[3],xmm3[4,5,6,7]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm11[0],xmm14[0],xmm11[1],xmm14[1],xmm11[2],xmm14[2],xmm11[3],xmm14[3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm3[0,0,1,1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm3[1,1,2,2]
; AVX1-NEXT:    vinsertf128 $1, %xmm6, %ymm5, %ymm5
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm9[0],xmm8[1],xmm9[1],xmm8[2],xmm9[2],xmm8[3],xmm9[3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm8 = xmm6[0,1,0,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm6, %ymm8, %ymm8
; AVX1-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0],ymm5[1],ymm8[2,3],ymm5[4],ymm8[5,6],ymm5[7]
; AVX1-NEXT:    vinsertps {{.*#+}} xmm7 = xmm5[0,1],xmm1[0],xmm5[3]
; AVX1-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm7[0,1,2,3,4],xmm2[5],xmm7[6,7]
; AVX1-NEXT:    vextractf128 $1, %ymm5, %xmm5
; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm7 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1],xmm7[2,3],xmm5[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm0[0,0,1,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm7[3],xmm5[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[2,2,3,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm10[0,0,1,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm7, %ymm3, %ymm3
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm6[2,3,2,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[0,1,0,1]
; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm6, %ymm4
; AVX1-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2],ymm4[3,4],ymm3[5],ymm4[6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm1[2,1,3,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
; AVX1-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0],xmm3[1,2],xmm4[3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm0[0,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm6[0,1,2,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm6[1],xmm4[2,3,4,5,6],xmm6[7]
; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm3
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5],xmm3[6,7]
; AVX1-NEXT:    vpslld $16, %xmm0, %xmm0
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4],xmm0[5],xmm1[6,7]
; AVX1-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-NEXT:    vmovdqa %xmm0, 48(%rax)
; AVX1-NEXT:    vmovdqa %xmm4, 32(%rax)
; AVX1-NEXT:    vmovdqa %xmm5, 16(%rax)
; AVX1-NEXT:    vmovdqa %xmm2, (%rax)
; AVX1-NEXT:    vmovdqa %xmm12, 112(%rax)
; AVX1-NEXT:    vmovdqa %xmm15, 96(%rax)
; AVX1-NEXT:    vmovdqa %xmm13, 80(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 64(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 176(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 160(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 144(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 128(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 240(%rax)
; AVX1-NEXT:    vmovaps (%rsp), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 224(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 208(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 192(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 304(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 288(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 272(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 256(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 368(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 352(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 336(%rax)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm0, 320(%rax)
; AVX1-NEXT:    addq $184, %rsp
; AVX1-NEXT:    vzeroupper
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf32:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    subq $648, %rsp # imm = 0x288
; AVX2-SLOW-NEXT:    vmovdqa (%rcx), %xmm0
; AVX2-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa 32(%rcx), %xmm9
; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-SLOW-NEXT:    vmovdqa (%rdx), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa 32(%rdx), %xmm8
; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} xmm1 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX2-SLOW-NEXT:    vpbroadcastq %xmm0, %ymm0
; AVX2-SLOW-NEXT:    vmovdqa (%rsi), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa 32(%rsi), %xmm7
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %xmm2
; AVX2-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa 32(%rdi), %xmm4
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,5]
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
; AVX2-SLOW-NEXT:    vmovdqa (%r8), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa 32(%r8), %xmm3
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[2,1,3,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3],ymm0[4,5],ymm1[6],ymm0[7]
; AVX2-SLOW-NEXT:    vmovdqa (%r9), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm12 = [255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255]
; AVX2-SLOW-NEXT:    vpblendvb %ymm12, %ymm0, %ymm1, %ymm0
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm9[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-SLOW-NEXT:    vmovdqa %xmm9, %xmm11
; AVX2-SLOW-NEXT:    vmovdqa %xmm9, (%rsp) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} xmm1 = xmm8[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-SLOW-NEXT:    vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm7[0,1,2,1]
; AVX2-SLOW-NEXT:    vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm4[0,1,2,1]
; AVX2-SLOW-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,5]
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
; AVX2-SLOW-NEXT:    vmovdqa 32(%r9), %xmm2
; AVX2-SLOW-NEXT:    vpbroadcastq %xmm0, %ymm0
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm3[2,1,3,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3],ymm0[4,5],ymm1[6],ymm0[7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm2[0,2,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa %xmm2, %xmm6
; AVX2-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
; AVX2-SLOW-NEXT:    vpblendvb %ymm12, %ymm0, %ymm1, %ymm0
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa 32(%rdx), %ymm2
; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa 32(%rcx), %ymm1
; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} ymm0 = ymm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm1[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-SLOW-NEXT:    vmovdqa %ymm1, %ymm9
; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} ymm1 = ymm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm2[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
; AVX2-SLOW-NEXT:    vmovdqa 32(%rsi), %ymm1
; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[2,1,2,3,6,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,3,2,1,4,5,6,7,8,11,10,9,12,13,14,15]
; AVX2-SLOW-NEXT:    vmovdqa 32(%rdi), %ymm2
; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[2,1,2,3,6,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[0,3,2,1,4,5,6,7,8,11,10,9,12,13,14,15]
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[8],ymm1[8],ymm2[9],ymm1[9],ymm2[10],ymm1[10],ymm2[11],ymm1[11]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,2,2,2]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,1,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
; AVX2-SLOW-NEXT:    vmovdqa 32(%r8), %ymm1
; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3],ymm0[4,5],ymm1[6],ymm0[7]
; AVX2-SLOW-NEXT:    vmovdqa 32(%r9), %ymm1
; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,3]
; AVX2-SLOW-NEXT:    vpblendvb %ymm12, %ymm0, %ymm1, %ymm0
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa (%rdx), %ymm1
; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa (%rcx), %ymm0
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} ymm0 = ymm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm0[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} ymm1 = ymm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm1[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
; AVX2-SLOW-NEXT:    vmovdqa (%rsi), %ymm15
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm15[2,1,2,3,6,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,3,2,1,4,5,6,7,8,11,10,9,12,13,14,15]
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %ymm2
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm5 = ymm2[2,1,2,3,6,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm5[0,3,2,1,4,5,6,7,8,11,10,9,12,13,14,15]
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm5[0],ymm1[0],ymm5[1],ymm1[1],ymm5[2],ymm1[2],ymm5[3],ymm1[3],ymm5[8],ymm1[8],ymm5[9],ymm1[9],ymm5[10],ymm1[10],ymm5[11],ymm1[11]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,2,2,2]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,1,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
; AVX2-SLOW-NEXT:    vmovdqa (%r8), %ymm1
; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm1[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[2,2,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7]
; AVX2-SLOW-NEXT:    vmovdqa (%r9), %ymm0
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm0[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm13 = ymm13[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[2,2,2,3]
; AVX2-SLOW-NEXT:    vpblendvb %ymm12, %ymm5, %ymm13, %ymm0
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm7[4],xmm4[5],xmm7[5],xmm4[6],xmm7[6],xmm4[7],xmm7[7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[1,1,1,1]
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm5 = xmm8[4],xmm11[4],xmm8[5],xmm11[5],xmm8[6],xmm11[6],xmm8[7],xmm11[7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm5 = xmm5[1,2,3,3]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[0,0,2,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1],ymm4[2],ymm5[3,4],ymm4[5],ymm5[6,7]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm5 = [12,13,10,11,10,11,14,15,14,15,14,15,14,15,14,15]
; AVX2-SLOW-NEXT:    vpshufb %xmm5, %xmm3, %xmm13
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,1,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm13 = ymm4[0],ymm13[1],ymm4[2,3],ymm13[4],ymm4[5,6],ymm13[7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm6[2,3,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,2,2,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm4[0,1,0,1]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-SLOW-NEXT:    vpblendvb %ymm4, %ymm13, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,1,1,1]
; AVX2-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm7 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm7 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm7 = xmm7[1,2,3,3]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[0,0,2,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1],ymm0[2],ymm7[3,4],ymm0[5],ymm7[6,7]
; AVX2-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-SLOW-NEXT:    vpshufb %xmm5, %xmm1, %xmm5
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[0,1,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3],ymm5[4],ymm0[5,6],ymm5[7]
; AVX2-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm5 = xmm14[2,3,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm5[0,2,2,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[0,1,0,1]
; AVX2-SLOW-NEXT:    vpblendvb %ymm4, %ymm0, %ymm5, %ymm0
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} ymm0 = ymm11[4],ymm12[4],ymm11[5],ymm12[5],ymm11[6],ymm12[6],ymm11[7],ymm12[7],ymm11[12],ymm12[12],ymm11[13],ymm12[13],ymm11[14],ymm12[14],ymm11[15],ymm12[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT:    vmovdqa %ymm9, %ymm10
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} ymm5 = ymm9[4],ymm10[4],ymm9[5],ymm10[5],ymm9[6],ymm10[6],ymm9[7],ymm10[7],ymm9[12],ymm10[12],ymm9[13],ymm10[13],ymm9[14],ymm10[14],ymm9[15],ymm10[15]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm5 = ymm5[1,2,3,3,5,6,7,7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[2,2,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1],ymm0[2],ymm5[3,4],ymm0[5],ymm5[6,7]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,14,15,14,15,14,15,14,15,28,29,26,27,26,27,30,31,30,31,30,31,30,31,30,31>
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-SLOW-NEXT:    vpshufb %ymm7, %ymm8, %ymm5
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[2,1,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3],ymm5[4],ymm0[5,6],ymm5[7]
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm5 = ymm6[2,3,2,3,6,7,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm5[0,2,2,1,4,5,6,7,8,10,10,9,12,13,14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[2,1,2,3]
; AVX2-SLOW-NEXT:    vpblendvb %ymm4, %ymm0, %ymm5, %ymm0
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} ymm0 = ymm2[4],ymm15[4],ymm2[5],ymm15[5],ymm2[6],ymm15[6],ymm2[7],ymm15[7],ymm2[12],ymm15[12],ymm2[13],ymm15[13],ymm2[14],ymm15[14],ymm2[15],ymm15[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} ymm13 = ymm3[4],ymm5[4],ymm3[5],ymm5[5],ymm3[6],ymm5[6],ymm3[7],ymm5[7],ymm3[12],ymm5[12],ymm3[13],ymm5[13],ymm3[14],ymm5[14],ymm3[15],ymm5[15]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm13 = ymm13[1,2,3,3,5,6,7,7]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[2,2,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm13[0,1],ymm0[2],ymm13[3,4],ymm0[5],ymm13[6,7]
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT:    vpshufb %ymm7, %ymm2, %ymm7
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[2,1,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3],ymm7[4],ymm0[5,6],ymm7[7]
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm7 = ymm1[2,3,2,3,6,7,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm7[0,2,2,1,4,5,6,7,8,10,10,9,12,13,14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[2,1,2,3]
; AVX2-SLOW-NEXT:    vpblendvb %ymm4, %ymm0, %ymm7, %ymm15
; AVX2-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-SLOW-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm4 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; AVX2-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-SLOW-NEXT:    vpunpcklwd (%rsp), %xmm0, %xmm7 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm7 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[0,0,2,1]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm7 = xmm7[1,0,2,2]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[0,1,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0],ymm7[1],ymm4[2,3],ymm7[4],ymm4[5,6],ymm7[7]
; AVX2-SLOW-NEXT:    vpmovzxwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm7 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
; AVX2-SLOW-NEXT:    vpbroadcastq %xmm7, %ymm7
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm7[2],ymm4[3,4],ymm7[5],ymm4[6,7]
; AVX2-SLOW-NEXT:    vpshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm7 = mem[0,0,2,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpbroadcastq %xmm7, %ymm7
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm13 = [255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255]
; AVX2-SLOW-NEXT:    vpblendvb %ymm13, %ymm4, %ymm7, %ymm4
; AVX2-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-SLOW-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm7 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm7 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; AVX2-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-SLOW-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[0,0,2,1]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,0,2,2]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0],ymm0[1],ymm7[2,3],ymm0[4],ymm7[5,6],ymm0[7]
; AVX2-SLOW-NEXT:    vpmovzxwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm7 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
; AVX2-SLOW-NEXT:    vpbroadcastq %xmm7, %ymm7
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2],ymm0[3,4],ymm7[5],ymm0[6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm14[0,0,2,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpbroadcastq %xmm7, %ymm7
; AVX2-SLOW-NEXT:    vpblendvb %ymm13, %ymm0, %ymm7, %ymm0
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} ymm7 = ymm11[0],ymm12[0],ymm11[1],ymm12[1],ymm11[2],ymm12[2],ymm11[3],ymm12[3],ymm11[8],ymm12[8],ymm11[9],ymm12[9],ymm11[10],ymm12[10],ymm11[11],ymm12[11]
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} ymm14 = ymm9[0],ymm10[0],ymm9[1],ymm10[1],ymm9[2],ymm10[2],ymm9[3],ymm10[3],ymm9[8],ymm10[8],ymm9[9],ymm10[9],ymm9[10],ymm10[10],ymm9[11],ymm10[11]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[2,2,2,3]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm14 = ymm14[1,0,2,2,5,4,6,6]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[2,1,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0],ymm14[1],ymm7[2,3],ymm14[4],ymm7[5,6],ymm14[7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm8[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[2,2,2,2]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1],ymm12[2],ymm7[3,4],ymm12[5],ymm7[6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm11 = ymm6[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm11 = ymm11[2,2,2,2]
; AVX2-SLOW-NEXT:    vpblendvb %ymm13, %ymm7, %ymm11, %ymm7
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-SLOW-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
; AVX2-SLOW-NEXT:    # ymm6 = ymm6[0],mem[0],ymm6[1],mem[1],ymm6[2],mem[2],ymm6[3],mem[3],ymm6[8],mem[8],ymm6[9],mem[9],ymm6[10],mem[10],ymm6[11],mem[11]
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} ymm8 = ymm3[0],ymm5[0],ymm3[1],ymm5[1],ymm3[2],ymm5[2],ymm3[3],ymm5[3],ymm3[8],ymm5[8],ymm3[9],ymm5[9],ymm3[10],ymm5[10],ymm3[11],ymm5[11]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm6 = ymm6[2,2,2,3]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm8 = ymm8[1,0,2,2,5,4,6,6]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[2,1,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0],ymm8[1],ymm6[2,3],ymm8[4],ymm6[5,6],ymm8[7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm2[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[2,2,2,2]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1],ymm3[2],ymm6[3,4],ymm3[5],ymm6[6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm1[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2]
; AVX2-SLOW-NEXT:    vpblendvb %ymm13, %ymm3, %ymm2, %ymm2
; AVX2-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-SLOW-NEXT:    vmovdqa %ymm2, 96(%rax)
; AVX2-SLOW-NEXT:    vmovdqa %ymm15, 160(%rax)
; AVX2-SLOW-NEXT:    vmovdqa %ymm7, 288(%rax)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm1, 352(%rax)
; AVX2-SLOW-NEXT:    vmovdqa %ymm0, (%rax)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm0, 64(%rax)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm0, 128(%rax)
; AVX2-SLOW-NEXT:    vmovdqa %ymm4, 192(%rax)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm0, 256(%rax)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm0, 320(%rax)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm0, 224(%rax)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm0, 32(%rax)
; AVX2-SLOW-NEXT:    addq $648, %rsp # imm = 0x288
; AVX2-SLOW-NEXT:    vzeroupper
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-ALL-LABEL: vf32:
; AVX2-FAST-ALL:       # %bb.0:
; AVX2-FAST-ALL-NEXT:    subq $648, %rsp # imm = 0x288
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rsi), %xmm1
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rsi), %xmm0
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm8 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm8, %xmm1, %xmm1
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdi), %xmm2
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rdi), %xmm11
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm8, %xmm2, %xmm2
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rcx), %xmm2
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rcx), %xmm4
; AVX2-FAST-ALL-NEXT:    vpsrldq {{.*#+}} xmm2 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdx), %xmm3
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rdx), %xmm6
; AVX2-FAST-ALL-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX2-FAST-ALL-NEXT:    vpbroadcastq %xmm2, %ymm2
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa (%r8), %xmm2
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[2,1,3,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,0,2,1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3],ymm1[4,5],ymm2[6],ymm1[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa (%r9), %xmm3
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm2, %xmm3, %xmm3
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,2,1]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm10 = [255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm10, %ymm1, %ymm3, %ymm1
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm8, %xmm0, %xmm1
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm0, %xmm13
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm0, (%rsp) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm8, %xmm11, %xmm0
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; AVX2-FAST-ALL-NEXT:    vpsrldq {{.*#+}} xmm1 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm4, %xmm14
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%r8), %xmm3
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpbroadcastq %xmm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm3[2,1,3,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm3, %xmm8
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3],ymm0[4,5],ymm1[6],ymm0[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%r9), %xmm3
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm2, %xmm3, %xmm1
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm3, %xmm7
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm10, %ymm0, %ymm1, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rdi), %ymm2
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rsi), %ymm1
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm0 = <u,u,u,u,4,5,10,11,u,u,u,u,u,u,u,u,24,25,22,23,20,21,26,27,u,u,u,u,u,u,u,u>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm0, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm0, %ymm2, %ymm2
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[8],ymm1[8],ymm2[9],ymm1[9],ymm2[10],ymm1[10],ymm2[11],ymm1[11]
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rdx), %ymm12
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rcx), %ymm9
; AVX2-FAST-ALL-NEXT:    vpsrldq {{.*#+}} ymm2 = ymm9[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm9[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpsrldq {{.*#+}} ymm3 = ymm12[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm12[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%r8), %ymm2
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,2,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3],ymm1[4,5],ymm2[6],ymm1[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%r9), %ymm2
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm4 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,20,21,20,21,22,23,24,25,24,25,24,25,24,25>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,2,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm10, %ymm1, %ymm2, %ymm1
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdi), %ymm2
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rsi), %ymm1
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm0, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm0, %ymm2, %ymm0
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdx), %ymm2
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rcx), %ymm3
; AVX2-FAST-ALL-NEXT:    vpsrldq {{.*#+}} ymm1 = ymm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm3[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpsrldq {{.*#+}} ymm5 = ymm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm2[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm5[0],ymm1[0],ymm5[1],ymm1[1],ymm5[2],ymm1[2],ymm5[3],ymm1[3],ymm5[8],ymm1[8],ymm5[9],ymm1[9],ymm5[10],ymm1[10],ymm5[11],ymm1[11]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa (%r8), %ymm1
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm1[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[2,2,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0],ymm0[1,2],ymm5[3],ymm0[4,5],ymm5[6],ymm0[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa (%r9), %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm4, %ymm0, %ymm4
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[2,2,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm10, %ymm5, %ymm4, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm6[4],xmm14[4],xmm6[5],xmm14[5],xmm6[6],xmm14[6],xmm6[7],xmm14[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm5 = <1,2,1,2,u,u,3,3>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm4, %ymm5, %ymm4
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} xmm15 = xmm11[4],xmm13[4],xmm11[5],xmm13[5],xmm11[6],xmm13[6],xmm11[7],xmm13[7]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm15 = ymm15[1,1,1,1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm15[2],ymm4[3,4],ymm15[5],ymm4[6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm15 = [12,13,10,11,10,11,14,15,14,15,14,15,14,15,14,15]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm15, %xmm8, %xmm13
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm13 = ymm4[0],ymm13[1],ymm4[2,3],ymm13[4],ymm4[5,6],ymm13[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,9,12,13,12,13,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm1, %xmm7, %xmm4
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm4[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm4, %ymm13, %ymm0, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm5, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[1,1,1,1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2],ymm0[3,4],ymm5[5],ymm0[6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm15, %xmm5, %xmm5
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3],ymm5[4],ymm0[5,6],ymm5[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm1, %xmm14, %xmm1
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm4, %ymm0, %ymm1, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} ymm0 = ymm12[4],ymm9[4],ymm12[5],ymm9[5],ymm12[6],ymm9[6],ymm12[7],ymm9[7],ymm12[12],ymm9[12],ymm12[13],ymm9[13],ymm12[14],ymm9[14],ymm12[15],ymm9[15]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm1 = [5,6,5,6,5,6,7,7]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm1, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} ymm5 = ymm10[4],ymm11[4],ymm10[5],ymm11[5],ymm10[6],ymm11[6],ymm10[7],ymm11[7],ymm10[12],ymm11[12],ymm10[13],ymm11[13],ymm10[14],ymm11[14],ymm10[15],ymm11[15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[3,3,3,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2],ymm0[3,4],ymm5[5],ymm0[6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm13 = <u,u,u,u,u,u,u,u,14,15,14,15,14,15,14,15,28,29,26,27,26,27,30,31,30,31,30,31,30,31,30,31>
; AVX2-FAST-ALL-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm13, %ymm9, %ymm5
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[2,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3],ymm5[4],ymm0[5,6],ymm5[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,8,9,10,11,12,13,14,15,24,25,28,29,28,29,26,27,24,25,26,27,28,29,30,31>
; AVX2-FAST-ALL-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm15, %ymm6, %ymm5
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[2,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm4, %ymm0, %ymm5, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} ymm0 = ymm2[4],ymm3[4],ymm2[5],ymm3[5],ymm2[6],ymm3[6],ymm2[7],ymm3[7],ymm2[12],ymm3[12],ymm2[13],ymm3[13],ymm2[14],ymm3[14],ymm2[15],ymm3[15]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm1, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vpunpckhwd {{.*#+}} ymm1 = ymm7[4],ymm8[4],ymm7[5],ymm8[5],ymm7[6],ymm8[6],ymm7[7],ymm8[7],ymm7[12],ymm8[12],ymm7[13],ymm8[13],ymm7[14],ymm8[14],ymm7[15],ymm8[15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[3,3,3,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm13, %ymm3, %ymm1
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX2-FAST-ALL-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm15, %ymm2, %ymm1
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm4, %ymm0, %ymm1, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX2-FAST-ALL-NEXT:    vpunpcklwd (%rsp), %xmm4, %xmm4 # 16-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # xmm4 = xmm4[0],mem[0],xmm4[1],mem[1],xmm4[2],mem[2],xmm4[3],mem[3]
; AVX2-FAST-ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm13 = [1,0,2,2,1,0,2,2]
; AVX2-FAST-ALL-NEXT:    # ymm13 = mem[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm1, %ymm13, %ymm1
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[0,0,2,1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0],ymm1[1],ymm4[2,3],ymm1[4],ymm4[5,6],ymm1[7]
; AVX2-FAST-ALL-NEXT:    vpmovzxwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # xmm4 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
; AVX2-FAST-ALL-NEXT:    vpbroadcastq %xmm4, %ymm4
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # xmm4 = mem[0,0,2,1,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpbroadcastq %xmm4, %ymm4
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm15 = [255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm15, %ymm1, %ymm4, %ymm1
; AVX2-FAST-ALL-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # xmm4 = xmm4[0],mem[0],xmm4[1],mem[1],xmm4[2],mem[2],xmm4[3],mem[3]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm4, %ymm13, %ymm4
; AVX2-FAST-ALL-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm13 # 16-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # xmm13 = xmm5[0],mem[0],xmm5[1],mem[1],xmm5[2],mem[2],xmm5[3],mem[3]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[0,0,2,1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm4 = ymm13[0],ymm4[1],ymm13[2,3],ymm4[4],ymm13[5,6],ymm4[7]
; AVX2-FAST-ALL-NEXT:    vpmovzxwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # xmm5 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
; AVX2-FAST-ALL-NEXT:    vpbroadcastq %xmm5, %ymm5
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm14[0,0,2,1,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpbroadcastq %xmm5, %ymm5
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm15, %ymm4, %ymm5, %ymm4
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # ymm5 = ymm12[0],mem[0],ymm12[1],mem[1],ymm12[2],mem[2],ymm12[3],mem[3],ymm12[8],mem[8],ymm12[9],mem[9],ymm12[10],mem[10],ymm12[11],mem[11]
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} ymm13 = ymm10[0],ymm11[0],ymm10[1],ymm11[1],ymm10[2],ymm11[2],ymm10[3],ymm11[3],ymm10[8],ymm11[8],ymm10[9],ymm11[9],ymm10[10],ymm11[10],ymm10[11],ymm11[11]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm14 = [5,4,2,2,5,4,6,6]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm5, %ymm14, %ymm5
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm13 = ymm13[2,2,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm5 = ymm13[0],ymm5[1],ymm13[2,3],ymm5[4],ymm13[5,6],ymm5[7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm9[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[2,2,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1],ymm12[2],ymm5[3,4],ymm12[5],ymm5[6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} ymm11 = ymm6[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm11 = ymm11[2,2,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm15, %ymm5, %ymm11, %ymm5
; AVX2-FAST-ALL-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # ymm6 = ymm6[0],mem[0],ymm6[1],mem[1],ymm6[2],mem[2],ymm6[3],mem[3],ymm6[8],mem[8],ymm6[9],mem[9],ymm6[10],mem[10],ymm6[11],mem[11]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm6, %ymm14, %ymm6
; AVX2-FAST-ALL-NEXT:    vpunpcklwd {{.*#+}} ymm7 = ymm7[0],ymm8[0],ymm7[1],ymm8[1],ymm7[2],ymm8[2],ymm7[3],ymm8[3],ymm7[8],ymm8[8],ymm7[9],ymm8[9],ymm7[10],ymm8[10],ymm7[11],ymm8[11]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[2,2,2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm6 = ymm7[0],ymm6[1],ymm7[2,3],ymm6[4],ymm7[5,6],ymm6[7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[2,2,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1],ymm3[2],ymm6[3,4],ymm3[5],ymm6[6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm15, %ymm3, %ymm2, %ymm2
; AVX2-FAST-ALL-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm2, 96(%rax)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm0, 160(%rax)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm5, 288(%rax)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm0, 352(%rax)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm4, (%rax)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm0, 64(%rax)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm0, 128(%rax)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm1, 192(%rax)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm0, 256(%rax)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm0, 320(%rax)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm0, 224(%rax)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm0, 32(%rax)
; AVX2-FAST-ALL-NEXT:    addq $648, %rsp # imm = 0x288
; AVX2-FAST-ALL-NEXT:    vzeroupper
; AVX2-FAST-ALL-NEXT:    retq
;
; AVX2-FAST-PERLANE-LABEL: vf32:
; AVX2-FAST-PERLANE:       # %bb.0:
; AVX2-FAST-PERLANE-NEXT:    subq $664, %rsp # imm = 0x298
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rsi), %xmm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rsi), %xmm12
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm8 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm8, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rdi), %xmm5
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm8, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rcx), %xmm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rcx), %xmm7
; AVX2-FAST-PERLANE-NEXT:    vpsrldq {{.*#+}} xmm2 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdx), %xmm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm3, (%rsp) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rdx), %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq %xmm2, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%r8), %xmm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[2,1,3,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,0,2,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0],ymm1[1,2],ymm2[3],ymm1[4,5],ymm2[6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%r9), %xmm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm1 = [0,1,4,5,4,5,6,7,8,9,8,9,8,9,8,9]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm1, %xmm3, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,0,2,1]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm9 = [255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm9, %ymm2, %ymm3, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm8, %xmm12, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm8, %xmm5, %xmm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
; AVX2-FAST-PERLANE-NEXT:    vpsrldq {{.*#+}} xmm2 = xmm7[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm7, %xmm6
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%r8), %xmm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq %xmm2, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2],ymm0[3,4],ymm2[5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm3[2,1,3,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,0,2,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0],ymm0[1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%r9), %xmm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm1, %xmm2, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm9, %ymm0, %ymm1, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rsi), %ymm15
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm0 = <u,u,u,u,4,5,10,11,u,u,u,u,u,u,u,u,24,25,22,23,20,21,26,27,u,u,u,u,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm0, %ymm15, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm0, %ymm3, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[8],ymm1[8],ymm2[9],ymm1[9],ymm2[10],ymm1[10],ymm2[11],ymm1[11]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rdx), %ymm7
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rcx), %ymm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpsrldq {{.*#+}} ymm2 = ymm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm2[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-FAST-PERLANE-NEXT:    vpsrldq {{.*#+}} ymm3 = ymm7[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm7[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%r8), %ymm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,2,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3],ymm1[4,5],ymm2[6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%r9), %ymm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm8 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,20,21,20,21,22,23,24,25,24,25,24,25,24,25>
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm8, %ymm2, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,2,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm9, %ymm1, %ymm2, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rsi), %ymm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm0, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm0, %ymm2, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdx), %ymm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rcx), %ymm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpsrldq {{.*#+}} ymm1 = ymm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm1[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-FAST-PERLANE-NEXT:    vpsrldq {{.*#+}} ymm11 = ymm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm2[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} ymm1 = ymm11[0],ymm1[0],ymm11[1],ymm1[1],ymm11[2],ymm1[2],ymm11[3],ymm1[3],ymm11[8],ymm1[8],ymm11[9],ymm1[9],ymm11[10],ymm1[10],ymm11[11],ymm1[11]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%r8), %ymm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} ymm11 = ymm1[2,1,3,3,4,5,6,7,10,9,11,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm11 = ymm11[2,2,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm11 = ymm11[0],ymm0[1,2],ymm11[3],ymm0[4,5],ymm11[6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%r9), %ymm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm8, %ymm0, %ymm8
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[2,2,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm9, %ymm11, %ymm8, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm12, %xmm0
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} xmm8 = xmm5[4],xmm12[4],xmm5[5],xmm12[5],xmm5[6],xmm12[6],xmm5[7],xmm12[7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[1,1,1,1]
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} xmm7 = xmm4[4],xmm6[4],xmm4[5],xmm6[5],xmm4[6],xmm6[6],xmm4[7],xmm6[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm7 = xmm7[1,2,3,3]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[0,0,2,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1],ymm8[2],ymm7[3,4],ymm8[5],ymm7[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm8 = [12,13,10,11,10,11,14,15,14,15,14,15,14,15,14,15]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm8, %xmm1, %xmm11
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm11 = ymm11[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0],ymm11[1],ymm7[2,3],ymm11[4],ymm7[5,6],ymm11[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm11 = [8,9,12,13,12,13,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm11, %xmm1, %xmm14
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm2, %ymm7, %ymm14, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} xmm14 = xmm13[4],xmm10[4],xmm13[5],xmm10[5],xmm13[6],xmm10[6],xmm13[7],xmm10[7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[1,1,1,1]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rsp), %xmm5 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} xmm7 = xmm5[4],xmm1[4],xmm5[5],xmm1[5],xmm5[6],xmm1[6],xmm5[7],xmm1[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm7 = xmm7[1,2,3,3]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[0,0,2,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1],ymm14[2],ymm7[3,4],ymm14[5],ymm7[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm8, %xmm6, %xmm8
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3],ymm8[4],ymm7[5,6],ymm8[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm11, %xmm9, %xmm8
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm2, %ymm7, %ymm8, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} ymm7 = ymm12[4],ymm15[4],ymm12[5],ymm15[5],ymm12[6],ymm15[6],ymm12[7],ymm15[7],ymm12[12],ymm15[12],ymm12[13],ymm15[13],ymm12[14],ymm15[14],ymm12[15],ymm15[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{.*#+}} ymm8 = ymm3[4],ymm2[4],ymm3[5],ymm2[5],ymm3[6],ymm2[6],ymm3[7],ymm2[7],ymm3[12],ymm2[12],ymm3[13],ymm2[13],ymm3[14],ymm2[14],ymm3[15],ymm2[15]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm8 = ymm8[1,2,3,3,5,6,7,7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[2,2,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm7 = ymm8[0,1],ymm7[2],ymm8[3,4],ymm7[5],ymm8[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm14 = <u,u,u,u,u,u,u,u,14,15,14,15,14,15,14,15,28,29,26,27,26,27,30,31,30,31,30,31,30,31,30,31>
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm14, %ymm4, %ymm8
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[2,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3],ymm8[4],ymm7[5,6],ymm8[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm8 = <u,u,u,u,u,u,u,u,8,9,10,11,12,13,14,15,24,25,28,29,28,29,26,27,24,25,26,27,28,29,30,31>
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm8, %ymm11, %ymm15
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm15 = ymm15[2,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm11 = [255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm11, %ymm7, %ymm15, %ymm7
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # ymm15 = ymm7[4],mem[4],ymm7[5],mem[5],ymm7[6],mem[6],ymm7[7],mem[7],ymm7[12],mem[12],ymm7[13],mem[13],ymm7[14],mem[14],ymm7[15],mem[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm15 = ymm15[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm11 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # ymm11 = ymm7[4],mem[4],ymm7[5],mem[5],ymm7[6],mem[6],ymm7[7],mem[7],ymm7[12],mem[12],ymm7[13],mem[13],ymm7[14],mem[14],ymm7[15],mem[15]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm11 = ymm11[1,2,3,3,5,6,7,7]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm11 = ymm11[2,2,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm11 = ymm11[0,1],ymm15[2],ymm11[3,4],ymm15[5],ymm11[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm14, %ymm7, %ymm14
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm14 = ymm14[2,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm11 = ymm11[0],ymm14[1],ymm11[2,3],ymm14[4],ymm11[5,6],ymm14[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm8, %ymm15, %ymm8
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[2,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm7, %ymm11, %ymm8, %ymm11
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} xmm8 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,0,2,1]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,0,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm8[0],ymm0[1],ymm8[2,3],ymm0[4],ymm8[5,6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT:    vpmovzxwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # xmm7 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq %xmm7, %ymm7
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2],ymm0[3,4],ymm7[5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # xmm7 = mem[0,0,2,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq %xmm7, %ymm7
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm8 = [255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm8, %ymm0, %ymm7, %ymm14
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} xmm7 = xmm13[0],xmm10[0],xmm13[1],xmm10[1],xmm13[2],xmm10[2],xmm13[3],xmm10[3]
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[0,0,2,1]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,0,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0],ymm0[1],ymm7[2,3],ymm0[4],ymm7[5,6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT:    vpmovzxwd {{.*#+}} xmm7 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq %xmm7, %ymm7
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2],ymm0[3,4],ymm7[5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm9[0,0,2,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpbroadcastq %xmm7, %ymm7
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm8, %ymm0, %ymm7, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm7 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # ymm7 = ymm12[0],mem[0],ymm12[1],mem[1],ymm12[2],mem[2],ymm12[3],mem[3],ymm12[8],mem[8],ymm12[9],mem[9],ymm12[10],mem[10],ymm12[11],mem[11]
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{.*#+}} ymm12 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[2,2,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm12 = ymm12[1,0,2,2,5,4,6,6]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[2,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0],ymm12[1],ymm7[2,3],ymm12[4],ymm7[5,6],ymm12[7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm4[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm12 = ymm12[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1],ymm12[2],ymm7[3,4],ymm12[5],ymm7[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # ymm9 = mem[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm9 = ymm9[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm8, %ymm7, %ymm9, %ymm7
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # ymm5 = ymm1[0],mem[0],ymm1[1],mem[1],ymm1[2],mem[2],ymm1[3],mem[3],ymm1[8],mem[8],ymm1[9],mem[9],ymm1[10],mem[10],ymm1[11],mem[11]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # ymm3 = ymm1[0],mem[0],ymm1[1],mem[1],ymm1[2],mem[2],ymm1[3],mem[3],ymm1[8],mem[8],ymm1[9],mem[9],ymm1[10],mem[10],ymm1[11],mem[11]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm4 = ymm5[2,2,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[1,0,2,2,5,4,6,6]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[2,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # ymm2 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm15[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm8, %ymm2, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm1, 96(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm11, 160(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, 128(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm7, 288(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, 352(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm1, 320(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm0, (%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 64(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 224(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm14, 192(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 256(%rax)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm0, 32(%rax)
; AVX2-FAST-PERLANE-NEXT:    addq $664, %rsp # imm = 0x298
; AVX2-FAST-PERLANE-NEXT:    vzeroupper
; AVX2-FAST-PERLANE-NEXT:    retq
;
; AVX512-LABEL: vf32:
; AVX512:       # %bb.0:
; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-NEXT:    vmovdqu64 (%rdi), %zmm2
; AVX512-NEXT:    vmovdqu64 (%rsi), %zmm3
; AVX512-NEXT:    vmovdqu64 (%rdx), %zmm4
; AVX512-NEXT:    vmovdqu64 (%rcx), %zmm5
; AVX512-NEXT:    vmovdqu64 (%r8), %zmm1
; AVX512-NEXT:    vmovdqu64 (%r9), %zmm0
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [0,0,0,32,3,35,0,0,1,33,4,36,0,0,2,34,0,0,0,32,3,35,0,0,1,33,4,36,0,0,2,34]
; AVX512-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm5, %zmm4, %zmm6
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,32,3,35,0,0,1,33,4,36,0,0,2,34,5,37,0,32,3,35,0,0,1,33,4,36,0,0,2,34,5,37]
; AVX512-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm3, %zmm2, %zmm7
; AVX512-NEXT:    movw $9362, %cx # imm = 0x2492
; AVX512-NEXT:    kmovd %ecx, %k2
; AVX512-NEXT:    vmovdqa32 %zmm6, %zmm7 {%k2}
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm8 = <0,1,2,3,32,u,6,7,8,9,33,u,12,13,14,15,34,u,18,19,20,21,35,u,24,25,26,27,36,u,30,31>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm7, %zmm8
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm6 = [0,1,2,3,4,32,6,7,8,9,10,33,12,13,14,15,16,34,18,19,20,21,22,35,24,25,26,27,28,36,30,31]
; AVX512-NEXT:    vpermi2w %zmm0, %zmm8, %zmm6
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [8,40,0,0,6,38,9,41,0,0,7,39,10,42,0,0,8,40,0,0,6,38,9,41,0,0,7,39,10,42,0,0]
; AVX512-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm3, %zmm2, %zmm7
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [5,37,8,40,0,0,6,38,9,41,0,0,7,39,10,42,5,37,8,40,0,0,6,38,9,41,0,0,7,39,10,42]
; AVX512-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm5, %zmm4, %zmm8
; AVX512-NEXT:    movw $18724, %cx # imm = 0x4924
; AVX512-NEXT:    kmovd %ecx, %k1
; AVX512-NEXT:    vmovdqa32 %zmm7, %zmm8 {%k1}
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm7 = <0,1,37,u,4,5,6,7,38,u,10,11,12,13,39,u,16,17,18,19,40,u,22,23,24,25,41,u,28,29,30,31>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm8, %zmm7
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm8 = [0,1,2,37,4,5,6,7,8,38,10,11,12,13,14,39,16,17,18,19,20,40,22,23,24,25,26,41,28,29,30,31]
; AVX512-NEXT:    vpermi2w %zmm0, %zmm7, %zmm8
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [13,45,0,0,11,43,14,46,0,0,12,44,15,47,0,0,13,45,0,0,11,43,14,46,0,0,12,44,15,47,0,0]
; AVX512-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm5, %zmm4, %zmm7
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [0,0,11,43,14,46,0,0,12,44,15,47,0,0,13,45,0,0,11,43,14,46,0,0,12,44,15,47,0,0,13,45]
; AVX512-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm3, %zmm2, %zmm9
; AVX512-NEXT:    vmovdqa32 %zmm7, %zmm9 {%k1}
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm7 = <42,u,2,3,4,5,43,u,8,9,10,11,44,u,14,15,16,17,45,u,20,21,22,23,46,u,26,27,28,29,47,u>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm9, %zmm7
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm9 = [0,42,2,3,4,5,6,43,8,9,10,11,12,44,14,15,16,17,18,45,20,21,22,23,24,46,26,27,28,29,30,47]
; AVX512-NEXT:    vpermi2w %zmm0, %zmm7, %zmm9
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,16,48,19,51,0,0,17,49,20,52,0,0,18,50,0,0,16,48,19,51,0,0,17,49,20,52,0,0,18,50]
; AVX512-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm5, %zmm4, %zmm7
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [16,48,19,51,0,0,17,49,20,52,0,0,18,50,21,53,16,48,19,51,0,0,17,49,20,52,0,0,18,50,21,53]
; AVX512-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm3, %zmm2, %zmm10
; AVX512-NEXT:    vmovdqa32 %zmm7, %zmm10 {%k2}
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm7 = <0,1,2,3,48,u,6,7,8,9,49,u,12,13,14,15,50,u,18,19,20,21,51,u,24,25,26,27,52,u,30,31>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm10, %zmm7
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm10 = [0,1,2,3,4,48,6,7,8,9,10,49,12,13,14,15,16,50,18,19,20,21,22,51,24,25,26,27,28,52,30,31]
; AVX512-NEXT:    vpermi2w %zmm0, %zmm7, %zmm10
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [24,56,0,0,22,54,25,57,0,0,23,55,26,58,0,0,24,56,0,0,22,54,25,57,0,0,23,55,26,58,0,0]
; AVX512-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm3, %zmm2, %zmm7
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm11 = [21,53,24,56,0,0,22,54,25,57,0,0,23,55,26,58,21,53,24,56,0,0,22,54,25,57,0,0,23,55,26,58]
; AVX512-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm5, %zmm4, %zmm11
; AVX512-NEXT:    vmovdqa32 %zmm7, %zmm11 {%k1}
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm7 = <0,1,53,u,4,5,6,7,54,u,10,11,12,13,55,u,16,17,18,19,56,u,22,23,24,25,57,u,28,29,30,31>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm11, %zmm7
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm11 = [0,1,2,53,4,5,6,7,8,54,10,11,12,13,14,55,16,17,18,19,20,56,22,23,24,25,26,57,28,29,30,31]
; AVX512-NEXT:    vpermi2w %zmm0, %zmm7, %zmm11
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [29,61,0,0,27,59,30,62,0,0,28,60,31,63,0,0,29,61,0,0,27,59,30,62,0,0,28,60,31,63,0,0]
; AVX512-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm5, %zmm4, %zmm7
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [0,0,27,59,30,62,0,0,28,60,31,63,0,0,29,61,0,0,27,59,30,62,0,0,28,60,31,63,0,0,29,61]
; AVX512-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm3, %zmm2, %zmm4
; AVX512-NEXT:    vmovdqa32 %zmm7, %zmm4 {%k1}
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm2 = <58,u,2,3,4,5,59,u,8,9,10,11,60,u,14,15,16,17,61,u,20,21,22,23,62,u,26,27,28,29,63,u>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm4, %zmm2
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [0,58,2,3,4,5,6,59,8,9,10,11,12,60,14,15,16,17,18,61,20,21,22,23,24,62,26,27,28,29,30,63]
; AVX512-NEXT:    vpermi2w %zmm0, %zmm2, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, 320(%rax)
; AVX512-NEXT:    vmovdqu64 %zmm11, 256(%rax)
; AVX512-NEXT:    vmovdqu64 %zmm10, 192(%rax)
; AVX512-NEXT:    vmovdqu64 %zmm9, 128(%rax)
; AVX512-NEXT:    vmovdqu64 %zmm8, 64(%rax)
; AVX512-NEXT:    vmovdqu64 %zmm6, (%rax)
; 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
  %in.vec3 = load <32 x i16>, ptr %in.vecptr3, align 32
  %in.vec4 = load <32 x i16>, ptr %in.vecptr4, align 32
  %in.vec5 = load <32 x i16>, ptr %in.vecptr5, 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>
  %concat23 = shufflevector <32 x i16> %in.vec2, <32 x i16> %in.vec3, <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>
  %concat45 = shufflevector <32 x i16> %in.vec4, <32 x i16> %in.vec5, <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>
  %concat0123 = shufflevector <64 x i16> %concat01, <64 x i16> %concat23, <128 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, i32 96, i32 97, i32 98, i32 99, i32 100, i32 101, i32 102, i32 103, i32 104, i32 105, i32 106, i32 107, i32 108, i32 109, i32 110, i32 111, i32 112, i32 113, i32 114, i32 115, i32 116, i32 117, i32 118, i32 119, i32 120, i32 121, i32 122, i32 123, i32 124, i32 125, i32 126, i32 127>
  %concat45uu = shufflevector <64 x i16> %concat45, <64 x i16> poison, <128 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 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, 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>
  %concat012345 = shufflevector <128 x i16> %concat0123, <128 x i16> %concat45uu, <192 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, i32 96, i32 97, i32 98, i32 99, i32 100, i32 101, i32 102, i32 103, i32 104, i32 105, i32 106, i32 107, i32 108, i32 109, i32 110, i32 111, i32 112, i32 113, i32 114, i32 115, i32 116, i32 117, i32 118, i32 119, i32 120, i32 121, i32 122, i32 123, i32 124, i32 125, i32 126, i32 127, i32 128, i32 129, i32 130, i32 131, i32 132, i32 133, i32 134, i32 135, i32 136, i32 137, i32 138, i32 139, i32 140, i32 141, i32 142, i32 143, i32 144, i32 145, i32 146, i32 147, i32 148, i32 149, i32 150, i32 151, i32 152, i32 153, i32 154, i32 155, i32 156, i32 157, i32 158, i32 159, i32 160, i32 161, i32 162, i32 163, i32 164, i32 165, i32 166, i32 167, i32 168, i32 169, i32 170, i32 171, i32 172, i32 173, i32 174, i32 175, i32 176, i32 177, i32 178, i32 179, i32 180, i32 181, i32 182, i32 183, i32 184, i32 185, i32 186, i32 187, i32 188, i32 189, i32 190, i32 191>
  %interleaved.vec = shufflevector <192 x i16> %concat012345, <192 x i16> poison, <192 x i32> <i32 0, i32 32, i32 64, i32 96, i32 128, i32 160, i32 1, i32 33, i32 65, i32 97, i32 129, i32 161, i32 2, i32 34, i32 66, i32 98, i32 130, i32 162, i32 3, i32 35, i32 67, i32 99, i32 131, i32 163, i32 4, i32 36, i32 68, i32 100, i32 132, i32 164, i32 5, i32 37, i32 69, i32 101, i32 133, i32 165, i32 6, i32 38, i32 70, i32 102, i32 134, i32 166, i32 7, i32 39, i32 71, i32 103, i32 135, i32 167, i32 8, i32 40, i32 72, i32 104, i32 136, i32 168, i32 9, i32 41, i32 73, i32 105, i32 137, i32 169, i32 10, i32 42, i32 74, i32 106, i32 138, i32 170, i32 11, i32 43, i32 75, i32 107, i32 139, i32 171, i32 12, i32 44, i32 76, i32 108, i32 140, i32 172, i32 13, i32 45, i32 77, i32 109, i32 141, i32 173, i32 14, i32 46, i32 78, i32 110, i32 142, i32 174, i32 15, i32 47, i32 79, i32 111, i32 143, i32 175, i32 16, i32 48, i32 80, i32 112, i32 144, i32 176, i32 17, i32 49, i32 81, i32 113, i32 145, i32 177, i32 18, i32 50, i32 82, i32 114, i32 146, i32 178, i32 19, i32 51, i32 83, i32 115, i32 147, i32 179, i32 20, i32 52, i32 84, i32 116, i32 148, i32 180, i32 21, i32 53, i32 85, i32 117, i32 149, i32 181, i32 22, i32 54, i32 86, i32 118, i32 150, i32 182, i32 23, i32 55, i32 87, i32 119, i32 151, i32 183, i32 24, i32 56, i32 88, i32 120, i32 152, i32 184, i32 25, i32 57, i32 89, i32 121, i32 153, i32 185, i32 26, i32 58, i32 90, i32 122, i32 154, i32 186, i32 27, i32 59, i32 91, i32 123, i32 155, i32 187, i32 28, i32 60, i32 92, i32 124, i32 156, i32 188, i32 29, i32 61, i32 93, i32 125, i32 157, i32 189, i32 30, i32 62, i32 94, i32 126, i32 158, i32 190, i32 31, i32 63, i32 95, i32 127, i32 159, i32 191>

  store <192 x i16> %interleaved.vec, ptr %out.vec, align 32

  ret void
}