; 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 loads.

define void @vf2(<10 x i16>* %in.vec, <2 x i16>* %out.vec0, <2 x i16>* %out.vec1, <2 x i16>* %out.vec2, <2 x i16>* %out.vec3, <2 x i16>* %out.vec4) nounwind {
; SSE-LABEL: vf2:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    movdqa 16(%rdi), %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,3,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[2,3,2,3]
; SSE-NEXT:    psrlq $48, %xmm0
; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE-NEXT:    psrld $16, %xmm1
; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
; SSE-NEXT:    movd %xmm2, (%rsi)
; SSE-NEXT:    movd %xmm3, (%rdx)
; SSE-NEXT:    movd %xmm4, (%rcx)
; SSE-NEXT:    movd %xmm0, (%r8)
; SSE-NEXT:    movd %xmm5, (%r9)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf2:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vmovdqa (%rdi), %xmm0
; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[0,3,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[2,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpsrlq $48, %xmm0, %xmm5
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5,6,7]
; AVX1-NEXT:    vmovd %xmm2, (%rsi)
; AVX1-NEXT:    vmovd %xmm3, (%rdx)
; AVX1-NEXT:    vmovd %xmm4, (%rcx)
; AVX1-NEXT:    vmovd %xmm5, (%r8)
; AVX1-NEXT:    vmovd %xmm0, (%r9)
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf2:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-SLOW-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[0,3,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[2,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpsrlq $48, %xmm0, %xmm0
; AVX2-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX2-SLOW-NEXT:    vpbroadcastw 8(%rdi), %xmm5
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm1 = xmm5[0],xmm1[1],xmm5[2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vmovd %xmm2, (%rsi)
; AVX2-SLOW-NEXT:    vmovd %xmm3, (%rdx)
; AVX2-SLOW-NEXT:    vmovd %xmm4, (%rcx)
; AVX2-SLOW-NEXT:    vmovd %xmm0, (%r8)
; AVX2-SLOW-NEXT:    vmovd %xmm1, (%r9)
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-LABEL: vf2:
; AVX2-FAST:       # %bb.0:
; AVX2-FAST-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-FAST-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[0,1,10,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[2,3,12,13,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[4,5,14,15,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FAST-NEXT:    vpsrlq $48, %xmm0, %xmm0
; AVX2-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX2-FAST-NEXT:    vpbroadcastw 8(%rdi), %xmm5
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm1 = xmm5[0],xmm1[1],xmm5[2,3,4,5,6,7]
; AVX2-FAST-NEXT:    vmovd %xmm2, (%rsi)
; AVX2-FAST-NEXT:    vmovd %xmm3, (%rdx)
; AVX2-FAST-NEXT:    vmovd %xmm4, (%rcx)
; AVX2-FAST-NEXT:    vmovd %xmm0, (%r8)
; AVX2-FAST-NEXT:    vmovd %xmm1, (%r9)
; AVX2-FAST-NEXT:    retq
;
; AVX512-LABEL: vf2:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
; AVX512-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
; AVX512-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[0,3,2,3]
; AVX512-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
; AVX512-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[3,1,2,3]
; AVX512-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[2,1,2,3,4,5,6,7]
; AVX512-NEXT:    vpsrlq $48, %xmm0, %xmm0
; AVX512-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX512-NEXT:    vpbroadcastw 8(%rdi), %xmm5
; AVX512-NEXT:    vpblendw {{.*#+}} xmm1 = xmm5[0],xmm1[1],xmm5[2,3,4,5,6,7]
; AVX512-NEXT:    vmovd %xmm2, (%rsi)
; AVX512-NEXT:    vmovd %xmm3, (%rdx)
; AVX512-NEXT:    vmovd %xmm4, (%rcx)
; AVX512-NEXT:    vmovd %xmm0, (%r8)
; AVX512-NEXT:    vmovd %xmm1, (%r9)
; AVX512-NEXT:    retq
  %wide.vec = load <10 x i16>, <10 x i16>* %in.vec, align 32

  %strided.vec0 = shufflevector <10 x i16> %wide.vec, <10 x i16> poison, <2 x i32> <i32 0, i32 5>
  %strided.vec1 = shufflevector <10 x i16> %wide.vec, <10 x i16> poison, <2 x i32> <i32 1, i32 6>
  %strided.vec2 = shufflevector <10 x i16> %wide.vec, <10 x i16> poison, <2 x i32> <i32 2, i32 7>
  %strided.vec3 = shufflevector <10 x i16> %wide.vec, <10 x i16> poison, <2 x i32> <i32 3, i32 8>
  %strided.vec4 = shufflevector <10 x i16> %wide.vec, <10 x i16> poison, <2 x i32> <i32 4, i32 9>

  store <2 x i16> %strided.vec0, <2 x i16>* %out.vec0, align 32
  store <2 x i16> %strided.vec1, <2 x i16>* %out.vec1, align 32
  store <2 x i16> %strided.vec2, <2 x i16>* %out.vec2, align 32
  store <2 x i16> %strided.vec3, <2 x i16>* %out.vec3, align 32
  store <2 x i16> %strided.vec4, <2 x i16>* %out.vec4, align 32

  ret void
}

define void @vf4(<20 x i16>* %in.vec, <4 x i16>* %out.vec0, <4 x i16>* %out.vec1, <4 x i16>* %out.vec2, <4 x i16>* %out.vec3, <4 x i16>* %out.vec4) nounwind {
; SSE-LABEL: vf4:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa (%rdi), %xmm3
; SSE-NEXT:    movdqa 16(%rdi), %xmm2
; SSE-NEXT:    movdqa 32(%rdi), %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm4[0,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
; SSE-NEXT:    movdqa %xmm0, %xmm9
; SSE-NEXT:    psllq $48, %xmm9
; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,0,65535,65535,65535,65535]
; SSE-NEXT:    movdqa %xmm2, %xmm6
; SSE-NEXT:    psrlq $48, %xmm6
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[0,3,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
; SSE-NEXT:    pand %xmm10, %xmm5
; SSE-NEXT:    movdqa %xmm10, %xmm11
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm3[0,1,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,7,6,7]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm2[2],xmm7[3],xmm2[3]
; SSE-NEXT:    pand %xmm10, %xmm7
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
; SSE-NEXT:    movdqa %xmm10, %xmm4
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm2[3,0]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm3[0,2]
; SSE-NEXT:    movdqa %xmm10, %xmm3
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm10, %xmm1
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm10, %xmm2
; SSE-NEXT:    pandn %xmm9, %xmm10
; SSE-NEXT:    por %xmm10, %xmm5
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[0,0,1,1]
; SSE-NEXT:    pandn %xmm6, %xmm11
; SSE-NEXT:    por %xmm11, %xmm7
; SSE-NEXT:    pandn %xmm0, %xmm4
; SSE-NEXT:    pslld $16, %xmm0
; SSE-NEXT:    pandn %xmm0, %xmm3
; SSE-NEXT:    por %xmm3, %xmm1
; SSE-NEXT:    por %xmm4, %xmm2
; SSE-NEXT:    movq %xmm8, (%rsi)
; SSE-NEXT:    movq %xmm5, (%rdx)
; SSE-NEXT:    movq %xmm7, (%rcx)
; SSE-NEXT:    movq %xmm1, (%r8)
; SSE-NEXT:    movq %xmm2, (%r9)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf4:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = mem[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
; AVX1-NEXT:    vmovdqa (%rdi), %xmm1
; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX1-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; AVX1-NEXT:    vpsrlq $48, %xmm2, %xmm4
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm1[0,3,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm5[1,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,4,5,8,9,2,3,u,u,u,u,u,u,u,u]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm1[0,1,1,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,7]
; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[0,1,4,5,8,9,6,7,u,u,u,u,u,u,u,u]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1],xmm1[2,3],xmm2[4,5,6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[6,7,0,1,10,11,10,11,u,u,u,u,u,u,u,u]
; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm6[0,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5],xmm2[6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[8,9,2,3,12,13,u,u,u,u,u,u,u,u,u,u]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm3[3],xmm1[4,5,6,7]
; AVX1-NEXT:    vmovq %xmm0, (%rsi)
; AVX1-NEXT:    vmovq %xmm4, (%rdx)
; AVX1-NEXT:    vmovq %xmm5, (%rcx)
; AVX1-NEXT:    vmovq %xmm6, (%r8)
; AVX1-NEXT:    vmovq %xmm1, (%r9)
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf4:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-SLOW-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm1[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0],xmm1[1],xmm4[2,3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[2,3,12,13,6,7,0,1,u,u,u,u,u,u,u,u]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm2[0],xmm0[1,2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1],xmm1[2],xmm5[3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[4,5,14,15,8,9,2,3,u,u,u,u,u,u,u,u]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm6 = xmm0[0,1],xmm2[2],xmm0[3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm6 = xmm1[0],xmm6[1],xmm1[2,3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[6,7,0,1,10,11,4,5,u,u,u,u,u,u,u,u]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[8,9,2,3,12,13,u,u,u,u,u,u,u,u,u,u]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm2[3],xmm0[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovq %xmm3, (%rsi)
; AVX2-SLOW-NEXT:    vmovq %xmm4, (%rdx)
; AVX2-SLOW-NEXT:    vmovq %xmm5, (%rcx)
; AVX2-SLOW-NEXT:    vmovq %xmm6, (%r8)
; AVX2-SLOW-NEXT:    vmovq %xmm0, (%r9)
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-LABEL: vf4:
; AVX2-FAST:       # %bb.0:
; AVX2-FAST-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-FAST-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-FAST-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm3 = xmm1[4,5,14,15,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[0,1,10,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FAST-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1,2,3,4,5,6,7]
; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0],xmm1[1],xmm4[2,3]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[2,3,12,13,6,7,0,1,u,u,u,u,u,u,u,u]
; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm5 = xmm2[0],xmm0[1,2,3]
; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1],xmm1[2],xmm5[3]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[4,5,14,15,8,9,2,3,u,u,u,u,u,u,u,u]
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm6 = xmm0[0,1],xmm2[2],xmm0[3,4,5,6,7]
; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm6 = xmm1[0],xmm6[1],xmm1[2,3]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[6,7,0,1,10,11,4,5,u,u,u,u,u,u,u,u]
; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[8,9,2,3,12,13,u,u,u,u,u,u,u,u,u,u]
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm2[3],xmm0[4,5,6,7]
; AVX2-FAST-NEXT:    vmovq %xmm3, (%rsi)
; AVX2-FAST-NEXT:    vmovq %xmm4, (%rdx)
; AVX2-FAST-NEXT:    vmovq %xmm5, (%rcx)
; AVX2-FAST-NEXT:    vmovq %xmm6, (%r8)
; AVX2-FAST-NEXT:    vmovq %xmm0, (%r9)
; AVX2-FAST-NEXT:    retq
;
; AVX512-LABEL: vf4:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa (%rdi), %ymm0
; AVX512-NEXT:    vmovdqa 32(%rdi), %ymm1
; AVX512-NEXT:    vmovdqa (%rdi), %xmm2
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,5,10,15,u,u,u,u>
; AVX512-NEXT:    vpermi2w 16(%rdi), %xmm2, %xmm3
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm2 = <1,6,11,16,u,u,u,u>
; AVX512-NEXT:    vpermi2w %ymm1, %ymm0, %ymm2
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm4 = <2,7,12,17,u,u,u,u>
; AVX512-NEXT:    vpermi2w %ymm1, %ymm0, %ymm4
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm5 = <3,8,13,18,u,u,u,u>
; AVX512-NEXT:    vpermi2w %ymm1, %ymm0, %ymm5
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm6 = <4,9,14,19,u,u,u,u>
; AVX512-NEXT:    vpermi2w %ymm1, %ymm0, %ymm6
; AVX512-NEXT:    vmovq %xmm3, (%rsi)
; AVX512-NEXT:    vmovq %xmm2, (%rdx)
; AVX512-NEXT:    vmovq %xmm4, (%rcx)
; AVX512-NEXT:    vmovq %xmm5, (%r8)
; AVX512-NEXT:    vmovq %xmm6, (%r9)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %wide.vec = load <20 x i16>, <20 x i16>* %in.vec, align 32

  %strided.vec0 = shufflevector <20 x i16> %wide.vec, <20 x i16> poison, <4 x i32> <i32 0, i32 5, i32 10, i32 15>
  %strided.vec1 = shufflevector <20 x i16> %wide.vec, <20 x i16> poison, <4 x i32> <i32 1, i32 6, i32 11, i32 16>
  %strided.vec2 = shufflevector <20 x i16> %wide.vec, <20 x i16> poison, <4 x i32> <i32 2, i32 7, i32 12, i32 17>
  %strided.vec3 = shufflevector <20 x i16> %wide.vec, <20 x i16> poison, <4 x i32> <i32 3, i32 8, i32 13, i32 18>
  %strided.vec4 = shufflevector <20 x i16> %wide.vec, <20 x i16> poison, <4 x i32> <i32 4, i32 9, i32 14, i32 19>

  store <4 x i16> %strided.vec0, <4 x i16>* %out.vec0, align 32
  store <4 x i16> %strided.vec1, <4 x i16>* %out.vec1, align 32
  store <4 x i16> %strided.vec2, <4 x i16>* %out.vec2, align 32
  store <4 x i16> %strided.vec3, <4 x i16>* %out.vec3, align 32
  store <4 x i16> %strided.vec4, <4 x i16>* %out.vec4, align 32

  ret void
}

define void @vf8(<40 x i16>* %in.vec, <8 x i16>* %out.vec0, <8 x i16>* %out.vec1, <8 x i16>* %out.vec2, <8 x i16>* %out.vec3, <8 x i16>* %out.vec4) nounwind {
; SSE-LABEL: vf8:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa 64(%rdi), %xmm12
; SSE-NEXT:    movdqa (%rdi), %xmm14
; SSE-NEXT:    movdqa 16(%rdi), %xmm1
; SSE-NEXT:    movdqa 32(%rdi), %xmm2
; SSE-NEXT:    movdqa 48(%rdi), %xmm11
; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535,0,65535,65535,65535]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[0,1,0,3]
; SSE-NEXT:    pand %xmm3, %xmm4
; SSE-NEXT:    pandn %xmm2, %xmm3
; SSE-NEXT:    por %xmm4, %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm14[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm3[2,3]
; SSE-NEXT:    movaps {{.*#+}} xmm3 = [65535,65535,65535,65535,65535,65535,65535,0]
; SSE-NEXT:    andps %xmm3, %xmm7
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm12[0,1,0,1]
; SSE-NEXT:    movaps %xmm3, %xmm10
; SSE-NEXT:    pandn %xmm4, %xmm10
; SSE-NEXT:    por %xmm7, %xmm10
; SSE-NEXT:    movdqa %xmm1, %xmm4
; SSE-NEXT:    psrlq $48, %xmm4
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm14[0,3,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm7[1,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [0,0,0,65535,65535,65535,65535,65535]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[1,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm6[0,1,2,3,7,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,4,7]
; SSE-NEXT:    pand %xmm7, %xmm4
; SSE-NEXT:    pandn %xmm5, %xmm7
; SSE-NEXT:    por %xmm4, %xmm7
; SSE-NEXT:    pand %xmm3, %xmm7
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm12[0,1,2,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm12[0,1,0,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm12[0,1,1,3]
; SSE-NEXT:    psllq $48, %xmm12
; SSE-NEXT:    pandn %xmm12, %xmm3
; SSE-NEXT:    por %xmm7, %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm14[0,1,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,7]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm1[2],xmm5[3],xmm1[3]
; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,0,0,0,65535,65535]
; SSE-NEXT:    pand %xmm6, %xmm5
; SSE-NEXT:    movdqa %xmm11, %xmm7
; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm11[0,2,2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,0],xmm2[0,0]
; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,0],xmm2[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[1,1,1,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
; SSE-NEXT:    movdqa %xmm6, %xmm4
; SSE-NEXT:    pandn %xmm0, %xmm4
; SSE-NEXT:    por %xmm5, %xmm4
; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm8[0,1,2,3,4,5,6,5]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,1],xmm0[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,0]
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm2[3,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm7[0,2]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,4,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm14[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm1[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1]
; SSE-NEXT:    movdqa %xmm6, %xmm5
; SSE-NEXT:    pandn %xmm2, %xmm5
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[1,0,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm6, %xmm7
; SSE-NEXT:    por %xmm5, %xmm7
; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm9[0,1,2,3,4,5,5,6]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,1],xmm2[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm5[2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm12[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm13[0,1,2,3,4,5,4,7]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,1],xmm2[1,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[2,0],xmm1[3,0]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm14[0,2]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm6, %xmm0
; SSE-NEXT:    pandn %xmm2, %xmm6
; SSE-NEXT:    por %xmm0, %xmm6
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm5[2,0]
; SSE-NEXT:    movdqa %xmm10, (%rsi)
; SSE-NEXT:    movdqa %xmm3, (%rdx)
; SSE-NEXT:    movaps %xmm4, (%rcx)
; SSE-NEXT:    movaps %xmm7, (%r8)
; SSE-NEXT:    movaps %xmm6, (%r9)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf8:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vmovdqa (%rdi), %xmm0
; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX1-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX1-NEXT:    vmovdqa 48(%rdi), %xmm3
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[0,1,0,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm2[4],xmm4[5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm1[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm5[2,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm0[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm6[0,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm5[0,1,2,3],xmm4[4,5,6,7]
; AVX1-NEXT:    vmovdqa 64(%rdi), %xmm5
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm5[0,1,0,1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm8 = xmm4[0,1,2,3,4,5,6],xmm6[7]
; AVX1-NEXT:    vpsrlq $48, %xmm1, %xmm6
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm0[0,3,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm7[1,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm7 = xmm2[0,1],xmm3[2,3],xmm2[4,5],xmm3[6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,0,1,10,11,4,5,14,15,u,u]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2],xmm7[3,4,5,6,7]
; AVX1-NEXT:    vpsllq $48, %xmm5, %xmm7
; AVX1-NEXT:    vpblendw {{.*#+}} xmm9 = xmm6[0,1,2,3,4,5,6],xmm7[7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm0[0,1,1,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,7,6,7]
; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm7 = xmm7[2],xmm1[2],xmm7[3],xmm1[3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm2[0,1],xmm3[2,3],xmm2[4,5,6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,2,3,12,13,6,7,u,u,u,u]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3,4,5],xmm7[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm5[0,1,2,0]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,6,5]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5],xmm7[6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm7 = xmm1[0,1],xmm0[2,3],xmm1[4,5,6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm7 = xmm7[6,7,0,1,10,11,u,u,u,u,u,u,12,13,14,15]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0,1,2,3],xmm3[4,5],xmm2[6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm6[2,2,2,2,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,4,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm6 = xmm7[0,1,2],xmm6[3,4,5],xmm7[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm5[0,1,0,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,5,6]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5],xmm7[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5],xmm1[6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[8,9,2,3,12,13,u,u,u,u,u,u,12,13,14,15]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm2[3,4,5],xmm0[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm5[0,1,1,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm1[6,7]
; AVX1-NEXT:    vmovdqa %xmm8, (%rsi)
; AVX1-NEXT:    vmovdqa %xmm9, (%rdx)
; AVX1-NEXT:    vmovdqa %xmm4, (%rcx)
; AVX1-NEXT:    vmovdqa %xmm6, (%r8)
; AVX1-NEXT:    vmovdqa %xmm0, (%r9)
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf8:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %ymm0
; AVX2-SLOW-NEXT:    vmovdqa 32(%rdi), %ymm2
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm1 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5],ymm2[6],ymm0[7,8],ymm2[9],ymm0[10,11],ymm2[12],ymm0[13],ymm2[14],ymm0[15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm3
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm3[1,2,3],xmm1[4,5],xmm3[6,7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,10,11,4,5,14,15,8,9,2,3,12,13,u,u]
; AVX2-SLOW-NEXT:    vpbroadcastw 70(%rdi), %xmm3
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6],xmm3[7]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm3 = ymm2[0],ymm0[1],ymm2[2],ymm0[3],ymm2[4,5],ymm0[6],ymm2[7,8],ymm0[9],ymm2[10],ymm0[11],ymm2[12,13],ymm0[14],ymm2[15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm3, %xmm4
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6],xmm4[7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[2,3,12,13,6,7,0,1,10,11,4,5,14,15,u,u]
; AVX2-SLOW-NEXT:    vmovdqa 64(%rdi), %xmm4
; AVX2-SLOW-NEXT:    vpsllq $48, %xmm4, %xmm5
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],xmm5[7]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm2[0,1],ymm0[2],ymm2[3],ymm0[4],ymm2[5,6],ymm0[7],ymm2[8,9],ymm0[10],ymm2[11],ymm0[12],ymm2[13,14],ymm0[15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm5, %xmm6
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm6[3,4],xmm5[5,6,7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm6 = xmm4[0,1,2,0]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,6,5]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1,2],xmm6[3]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm6 = ymm0[0],ymm2[1,2],ymm0[3],ymm2[4],ymm0[5],ymm2[6,7],ymm0[8],ymm2[9,10],ymm0[11],ymm2[12],ymm0[13],ymm2[14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm6, %xmm7
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm6 = xmm7[0],xmm6[1],xmm7[2],xmm6[3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm7 = xmm4[0,1,0,3]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,5,6]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm6 = xmm6[0,1,2],xmm7[3]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3],ymm0[4],ymm2[5],ymm0[6],ymm2[7,8],ymm0[9],ymm2[10,11],ymm0[12],ymm2[13],ymm0[14],ymm2[15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm2
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3,4],xmm2[5,6,7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[8,9,2,3,12,13,6,7,0,1,10,11,u,u,u,u]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm4[0,1,1,3]
; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm2[3]
; AVX2-SLOW-NEXT:    vmovdqa %xmm1, (%rsi)
; AVX2-SLOW-NEXT:    vmovdqa %xmm3, (%rdx)
; AVX2-SLOW-NEXT:    vmovdqa %xmm5, (%rcx)
; AVX2-SLOW-NEXT:    vmovdqa %xmm6, (%r8)
; AVX2-SLOW-NEXT:    vmovdqa %xmm0, (%r9)
; AVX2-SLOW-NEXT:    vzeroupper
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-LABEL: vf8:
; AVX2-FAST:       # %bb.0:
; AVX2-FAST-NEXT:    vmovdqa (%rdi), %ymm0
; AVX2-FAST-NEXT:    vmovdqa 32(%rdi), %ymm1
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm2 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5],ymm1[6],ymm0[7,8],ymm1[9],ymm0[10,11],ymm1[12],ymm0[13],ymm1[14],ymm0[15]
; AVX2-FAST-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1,2,3],xmm2[4,5],xmm3[6,7]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[0,1,10,11,4,5,14,15,8,9,2,3,12,13,u,u]
; AVX2-FAST-NEXT:    vpbroadcastw 70(%rdi), %xmm3
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6],xmm3[7]
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm3 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4,5],ymm0[6],ymm1[7,8],ymm0[9],ymm1[10],ymm0[11],ymm1[12,13],ymm0[14],ymm1[15]
; AVX2-FAST-NEXT:    vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6],xmm4[7]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[2,3,12,13,6,7,0,1,10,11,4,5,14,15,u,u]
; AVX2-FAST-NEXT:    vmovdqa 64(%rdi), %xmm4
; AVX2-FAST-NEXT:    vpsllq $48, %xmm4, %xmm5
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],xmm5[7]
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm5 = ymm1[0,1],ymm0[2],ymm1[3],ymm0[4],ymm1[5,6],ymm0[7],ymm1[8,9],ymm0[10],ymm1[11],ymm0[12],ymm1[13,14],ymm0[15]
; AVX2-FAST-NEXT:    vextracti128 $1, %ymm5, %xmm6
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm6[3,4],xmm5[5,6,7]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm6 = xmm4[u,u,u,u,u,u,u,u,u,u,u,u,0,1,10,11]
; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1,2],xmm6[3]
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm6 = ymm0[0],ymm1[1,2],ymm0[3],ymm1[4],ymm0[5],ymm1[6,7],ymm0[8],ymm1[9,10],ymm0[11],ymm1[12],ymm0[13],ymm1[14,15]
; AVX2-FAST-NEXT:    vextracti128 $1, %ymm6, %xmm7
; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm6 = xmm7[0],xmm6[1],xmm7[2],xmm6[3]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm7 = xmm4[u,u,u,u,u,u,u,u,u,u,u,u,2,3,12,13]
; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm6 = xmm6[0,1,2],xmm7[3]
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5],ymm0[6],ymm1[7,8],ymm0[9],ymm1[10,11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
; AVX2-FAST-NEXT:    vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3,4],xmm1[5,6,7]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[8,9,2,3,12,13,6,7,0,1,10,11,u,u,u,u]
; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm1 = xmm4[u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15]
; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
; AVX2-FAST-NEXT:    vmovdqa %xmm2, (%rsi)
; AVX2-FAST-NEXT:    vmovdqa %xmm3, (%rdx)
; AVX2-FAST-NEXT:    vmovdqa %xmm5, (%rcx)
; AVX2-FAST-NEXT:    vmovdqa %xmm6, (%r8)
; AVX2-FAST-NEXT:    vmovdqa %xmm0, (%r9)
; AVX2-FAST-NEXT:    vzeroupper
; AVX2-FAST-NEXT:    retq
;
; AVX512-LABEL: vf8:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm0 = <4,9,14,u,u,u,u,u>
; AVX512-NEXT:    vmovdqu64 (%rdi), %zmm1
; AVX512-NEXT:    vpermw %zmm1, %zmm0, %zmm0
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm2 = <3,8,13,u,u,u,u,u>
; AVX512-NEXT:    vpermw %zmm1, %zmm2, %zmm4
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm2 = <2,7,12,u,u,u,u,u>
; AVX512-NEXT:    vpermw %zmm1, %zmm2, %zmm5
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm2 = <1,6,11,u,u,u,u,u>
; AVX512-NEXT:    vpermw %zmm1, %zmm2, %zmm6
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,5,10,u,u,u,u,u>
; AVX512-NEXT:    vpermw %zmm1, %zmm2, %zmm1
; AVX512-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX512-NEXT:    vpextrw $7, %xmm2, %eax
; AVX512-NEXT:    vpinsrw $3, %eax, %xmm1, %xmm1
; AVX512-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512-NEXT:    vpblendw {{.*#+}} xmm3 = xmm1[0,1,2,3],xmm2[4],xmm1[5,6,7]
; AVX512-NEXT:    vmovdqa 48(%rdi), %xmm1
; AVX512-NEXT:    vpextrw $1, %xmm1, %eax
; AVX512-NEXT:    vpinsrw $5, %eax, %xmm3, %xmm3
; AVX512-NEXT:    vpblendw {{.*#+}} xmm7 = xmm3[0,1,2,3,4,5],xmm1[6],xmm3[7]
; AVX512-NEXT:    vmovdqa 64(%rdi), %xmm3
; AVX512-NEXT:    vpextrw $3, %xmm3, %eax
; AVX512-NEXT:    vpinsrw $7, %eax, %xmm7, %xmm7
; AVX512-NEXT:    vmovd %xmm2, %eax
; AVX512-NEXT:    vpinsrw $3, %eax, %xmm6, %xmm6
; AVX512-NEXT:    vpextrw $5, %xmm2, %eax
; AVX512-NEXT:    vpinsrw $4, %eax, %xmm6, %xmm6
; AVX512-NEXT:    vpextrw $2, %xmm1, %eax
; AVX512-NEXT:    vpinsrw $5, %eax, %xmm6, %xmm6
; AVX512-NEXT:    vpextrw $7, %xmm1, %eax
; AVX512-NEXT:    vpinsrw $6, %eax, %xmm6, %xmm6
; AVX512-NEXT:    vpextrw $4, %xmm3, %eax
; AVX512-NEXT:    vpinsrw $7, %eax, %xmm6, %xmm6
; AVX512-NEXT:    vpextrw $1, %xmm2, %eax
; AVX512-NEXT:    vpinsrw $3, %eax, %xmm5, %xmm5
; AVX512-NEXT:    vpextrw $6, %xmm2, %eax
; AVX512-NEXT:    vpinsrw $4, %eax, %xmm5, %xmm5
; AVX512-NEXT:    vpextrw $3, %xmm1, %eax
; AVX512-NEXT:    vpinsrw $5, %eax, %xmm5, %xmm5
; AVX512-NEXT:    vmovd %xmm3, %eax
; AVX512-NEXT:    vpinsrw $6, %eax, %xmm5, %xmm5
; AVX512-NEXT:    vpextrw $5, %xmm3, %eax
; AVX512-NEXT:    vpinsrw $7, %eax, %xmm5, %xmm5
; AVX512-NEXT:    vpextrw $2, %xmm2, %eax
; AVX512-NEXT:    vpinsrw $3, %eax, %xmm4, %xmm4
; AVX512-NEXT:    vpextrw $7, %xmm2, %eax
; AVX512-NEXT:    vpinsrw $4, %eax, %xmm4, %xmm4
; AVX512-NEXT:    vpextrw $4, %xmm1, %eax
; AVX512-NEXT:    vpinsrw $5, %eax, %xmm4, %xmm4
; AVX512-NEXT:    vpextrw $1, %xmm3, %eax
; AVX512-NEXT:    vpinsrw $6, %eax, %xmm4, %xmm4
; AVX512-NEXT:    vpextrw $6, %xmm3, %eax
; AVX512-NEXT:    vpinsrw $7, %eax, %xmm4, %xmm4
; AVX512-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm2[3],xmm0[4,5,6,7]
; AVX512-NEXT:    vmovd %xmm1, %eax
; AVX512-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
; AVX512-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm1[5],xmm0[6,7]
; AVX512-NEXT:    vpextrw $2, %xmm3, %eax
; AVX512-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
; AVX512-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm3[7]
; AVX512-NEXT:    vmovdqa %xmm7, (%rsi)
; AVX512-NEXT:    vmovdqa %xmm6, (%rdx)
; AVX512-NEXT:    vmovdqa %xmm5, (%rcx)
; AVX512-NEXT:    vmovdqa %xmm4, (%r8)
; AVX512-NEXT:    vmovdqa %xmm0, (%r9)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %wide.vec = load <40 x i16>, <40 x i16>* %in.vec, align 32

  %strided.vec0 = shufflevector <40 x i16> %wide.vec, <40 x i16> poison, <8 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35>
  %strided.vec1 = shufflevector <40 x i16> %wide.vec, <40 x i16> poison, <8 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36>
  %strided.vec2 = shufflevector <40 x i16> %wide.vec, <40 x i16> poison, <8 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37>
  %strided.vec3 = shufflevector <40 x i16> %wide.vec, <40 x i16> poison, <8 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38>
  %strided.vec4 = shufflevector <40 x i16> %wide.vec, <40 x i16> poison, <8 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39>

  store <8 x i16> %strided.vec0, <8 x i16>* %out.vec0, align 32
  store <8 x i16> %strided.vec1, <8 x i16>* %out.vec1, align 32
  store <8 x i16> %strided.vec2, <8 x i16>* %out.vec2, align 32
  store <8 x i16> %strided.vec3, <8 x i16>* %out.vec3, align 32
  store <8 x i16> %strided.vec4, <8 x i16>* %out.vec4, align 32

  ret void
}

define void @vf16(<80 x i16>* %in.vec, <16 x i16>* %out.vec0, <16 x i16>* %out.vec1, <16 x i16>* %out.vec2, <16 x i16>* %out.vec3, <16 x i16>* %out.vec4) nounwind {
; SSE-LABEL: vf16:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $40, %rsp
; SSE-NEXT:    movdqa 144(%rdi), %xmm9
; SSE-NEXT:    movdqa 80(%rdi), %xmm11
; SSE-NEXT:    movaps 96(%rdi), %xmm0
; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 128(%rdi), %xmm5
; SSE-NEXT:    movdqa 112(%rdi), %xmm14
; SSE-NEXT:    movdqa 64(%rdi), %xmm6
; SSE-NEXT:    movdqa (%rdi), %xmm8
; SSE-NEXT:    movdqa 16(%rdi), %xmm15
; SSE-NEXT:    movdqa 32(%rdi), %xmm13
; SSE-NEXT:    movdqa 48(%rdi), %xmm12
; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,0,65535,65535,65535]
; SSE-NEXT:    movdqa %xmm0, %xmm1
; SSE-NEXT:    pandn %xmm13, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[0,1,0,3]
; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pand %xmm0, %xmm2
; SSE-NEXT:    por %xmm1, %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
; SSE-NEXT:    movaps {{.*#+}} xmm10 = [65535,65535,65535,65535,65535,65535,65535,0]
; SSE-NEXT:    andps %xmm10, %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,1,0,1]
; SSE-NEXT:    movaps %xmm10, %xmm2
; SSE-NEXT:    pandn %xmm1, %xmm2
; SSE-NEXT:    por %xmm3, %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,1,0,3]
; SSE-NEXT:    pand %xmm0, %xmm1
; SSE-NEXT:    pandn %xmm14, %xmm0
; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    por %xmm1, %xmm0
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[0,2,2,3]
; SSE-NEXT:    movdqa %xmm11, %xmm7
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
; SSE-NEXT:    andps %xmm10, %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[0,1,0,1]
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm10, %xmm1
; SSE-NEXT:    andnps %xmm0, %xmm1
; SSE-NEXT:    orps %xmm2, %xmm1
; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm15, %xmm0
; SSE-NEXT:    psrlq $48, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,3,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [0,0,0,65535,65535,65535,65535,65535]
; SSE-NEXT:    movdqa %xmm0, %xmm2
; SSE-NEXT:    pandn %xmm1, %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,7,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
; SSE-NEXT:    pand %xmm0, %xmm1
; SSE-NEXT:    por %xmm2, %xmm1
; SSE-NEXT:    movdqa %xmm6, %xmm2
; SSE-NEXT:    psllq $48, %xmm2
; SSE-NEXT:    movaps %xmm10, %xmm3
; SSE-NEXT:    andnps %xmm2, %xmm3
; SSE-NEXT:    pand %xmm10, %xmm1
; SSE-NEXT:    orps %xmm1, %xmm3
; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm4, %xmm1
; SSE-NEXT:    movdqa %xmm4, %xmm11
; SSE-NEXT:    psrlq $48, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,3,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    movdqa %xmm5, %xmm4
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm14[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,7,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
; SSE-NEXT:    pand %xmm0, %xmm1
; SSE-NEXT:    pandn %xmm2, %xmm0
; SSE-NEXT:    por %xmm1, %xmm0
; SSE-NEXT:    pand %xmm10, %xmm0
; SSE-NEXT:    movdqa %xmm9, %xmm1
; SSE-NEXT:    psllq $48, %xmm1
; SSE-NEXT:    pandn %xmm1, %xmm10
; SSE-NEXT:    por %xmm0, %xmm10
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,1,1,3]
; SSE-NEXT:    movdqa %xmm8, %xmm12
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,7,6,7]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm15[2],xmm1[3],xmm15[3]
; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,0,0,0,65535,65535]
; SSE-NEXT:    pand %xmm5, %xmm1
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, %xmm2
; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm13[0,0]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm13[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,1,1,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
; SSE-NEXT:    movdqa %xmm5, %xmm14
; SSE-NEXT:    pandn %xmm2, %xmm14
; SSE-NEXT:    por %xmm1, %xmm14
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,1,2,0]
; SSE-NEXT:    movdqa %xmm6, %xmm9
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm2[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm1[2,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[0,1,1,3]
; SSE-NEXT:    movdqa %xmm7, %xmm13
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm11[2],xmm1[3],xmm11[3]
; SSE-NEXT:    pand %xmm5, %xmm1
; SSE-NEXT:    movdqa %xmm4, %xmm2
; SSE-NEXT:    movdqa %xmm4, %xmm8
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm4[0,0]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm4[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,1,1,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
; SSE-NEXT:    movdqa %xmm5, %xmm3
; SSE-NEXT:    pandn %xmm2, %xmm3
; SSE-NEXT:    por %xmm1, %xmm3
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[0,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm2[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2,0]
; SSE-NEXT:    movaps %xmm0, %xmm1
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm11[3,0]
; SSE-NEXT:    movaps %xmm11, %xmm2
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,4,6,7]
; SSE-NEXT:    movdqa %xmm5, %xmm6
; SSE-NEXT:    pandn %xmm1, %xmm6
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm0[1,0,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm5, %xmm12
; SSE-NEXT:    por %xmm6, %xmm12
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[0,1,0,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,5,6]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm0[2,0]
; SSE-NEXT:    movdqa %xmm8, %xmm0
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[3,0]
; SSE-NEXT:    movaps %xmm4, %xmm1
; SSE-NEXT:    movaps %xmm4, %xmm8
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,4,6,7]
; SSE-NEXT:    movdqa %xmm5, %xmm4
; SSE-NEXT:    pandn %xmm0, %xmm4
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm5, %xmm1
; SSE-NEXT:    por %xmm4, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[0,1,0,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,5,6]
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,1],xmm0[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[0,2,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm9[0,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,4,7]
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[3,1],xmm4[1,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; SSE-NEXT:    movdqa %xmm5, %xmm0
; SSE-NEXT:    pandn %xmm4, %xmm0
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm15[3,0]
; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm4[0,2]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm15[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm5, %xmm4
; SSE-NEXT:    por %xmm0, %xmm4
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[1,1,1,1]
; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT:    # xmm6 = mem[0,2,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm6[1,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm8[0],xmm6[1],xmm8[1]
; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[2,0],xmm2[3,0]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm13[0,2]
; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm2[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[2,1,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm5, %xmm7
; SSE-NEXT:    pandn %xmm6, %xmm5
; SSE-NEXT:    por %xmm7, %xmm5
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[2,0]
; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rsi)
; SSE-NEXT:    movdqa %xmm10, 16(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rdx)
; SSE-NEXT:    movaps %xmm3, 16(%rcx)
; SSE-NEXT:    movaps %xmm14, (%rcx)
; SSE-NEXT:    movaps %xmm1, 16(%r8)
; SSE-NEXT:    movaps %xmm12, (%r8)
; SSE-NEXT:    movaps %xmm5, 16(%r9)
; SSE-NEXT:    movaps %xmm4, (%r9)
; SSE-NEXT:    addq $40, %rsp
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf16:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vmovdqa 64(%rdi), %xmm9
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm9[0,1,0,1]
; AVX1-NEXT:    vmovdqa 96(%rdi), %xmm10
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm10[0,1,1,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; AVX1-NEXT:    vmovdqa 112(%rdi), %xmm13
; AVX1-NEXT:    vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm13[1]
; AVX1-NEXT:    vmovdqa 80(%rdi), %xmm14
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm14[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm4[0,1],xmm1[2,3,4,5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX1-NEXT:    vmovaps {{.*#+}} ymm11 = [65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,65535,65535,65535]
; AVX1-NEXT:    vandnps %ymm0, %ymm11, %ymm0
; AVX1-NEXT:    vmovdqa (%rdi), %xmm15
; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm5
; AVX1-NEXT:    vmovdqa 32(%rdi), %xmm6
; AVX1-NEXT:    vmovdqa 48(%rdi), %xmm7
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm7[0,1,0,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm6[4],xmm1[5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm5[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm15[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
; AVX1-NEXT:    vandps %ymm1, %ymm11, %ymm1
; AVX1-NEXT:    vorps %ymm0, %ymm1, %ymm2
; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT:    vmovdqa 144(%rdi), %xmm1
; AVX1-NEXT:    vmovdqa 128(%rdi), %xmm0
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,u,u,u,u,2,3,12,13,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm4[5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm8
; AVX1-NEXT:    vpsllq $48, %xmm9, %xmm2
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm13[0,1],xmm10[2,3],xmm13[4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,0,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,5,5,5]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm14[0,3,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[1,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3,4,5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
; AVX1-NEXT:    vandnps %ymm2, %ymm11, %ymm2
; AVX1-NEXT:    vpsrlq $48, %xmm5, %xmm3
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm15[0,3,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[1,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm6[0,1],xmm7[2,3],xmm6[4,5],xmm7[6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,0,1,10,11,4,5,14,15,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3,4,5,6,7]
; AVX1-NEXT:    vandps %ymm3, %ymm11, %ymm3
; AVX1-NEXT:    vorps %ymm2, %ymm3, %ymm3
; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm2
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm0[0,1,2,3],xmm1[4,5],xmm0[6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,u,u,u,u,4,5,14,15,8,9]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm4[5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm11
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm13[0,1,2,3],xmm10[4,5],xmm13[6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,8,9,2,3,12,13,12,13,u,u,u,u]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm14[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm1[0,1],xmm0[2,3],xmm1[4,5,6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,u,u,u,6,7,0,1,10,11]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm3[5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm15[0,1,1,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm6[0,1],xmm7[2,3],xmm6[4,5,6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,2,3,12,13,6,7,u,u,u,u]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3,4,5],xmm3[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm9[0,1,2,0]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6,5]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm4[6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm12
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm10[0,1],xmm13[2,3],xmm10[4,5],xmm13[6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[u,u,0,1,10,11,4,5,14,15,u,u,u,u,u,u]
; AVX1-NEXT:    vpsrlq $48, %xmm14, %xmm3
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm1[0,1,2,3],xmm0[4,5],xmm1[6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,u,u,u,8,9,2,3,12,13]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm3[5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm5[0,1],xmm15[2,3],xmm5[4,5,6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[6,7,0,1,10,11,u,u,u,u,u,u,u,u,u,u]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm6[0,1,2,3],xmm7[4,5],xmm6[6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,4,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3,4,5],xmm3[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm9[0,1,0,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,5,6]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm4[6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm10[0,3,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm13[3],xmm1[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm14[2,3,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm3[0],xmm1[1,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm7[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm5[0,1,2,3],xmm15[4,5],xmm5[6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[8,9,2,3,12,13,u,u,u,u,u,u,u,u,u,u]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3,4,5],xmm3[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm9[0,1,1,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm3[6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-NEXT:    vmovaps %ymm8, (%rsi)
; AVX1-NEXT:    vmovaps %ymm11, (%rdx)
; AVX1-NEXT:    vmovaps %ymm12, (%rcx)
; AVX1-NEXT:    vmovaps %ymm2, (%r8)
; AVX1-NEXT:    vmovaps %ymm0, (%r9)
; AVX1-NEXT:    vzeroupper
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf16:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %ymm2
; AVX2-SLOW-NEXT:    vmovdqa 32(%rdi), %ymm3
; AVX2-SLOW-NEXT:    vmovdqa 64(%rdi), %ymm0
; AVX2-SLOW-NEXT:    vmovdqa 96(%rdi), %ymm1
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm4 = ymm2[0],ymm3[1],ymm2[2,3],ymm3[4],ymm2[5],ymm3[6],ymm2[7,8],ymm3[9],ymm2[10,11],ymm3[12],ymm2[13],ymm3[14],ymm2[15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm4, %xmm5
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm5[1,2,3],xmm4[4,5],xmm5[6,7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,10,11,4,5,14,15,8,9,2,3,12,13,6,7]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm0[0],ymm1[1,2],ymm0[3],ymm1[4],ymm0[5],ymm1[6,7],ymm0[8],ymm1[9,10],ymm0[11],ymm1[12],ymm0[13],ymm1[14,15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm6 = ymm5[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5],ymm5[6],ymm6[7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[0,1,10,11,4,5,14,15,8,9,10,11,4,5,6,7,16,17,26,27,20,21,30,31,24,25,26,27,20,21,22,23]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-SLOW-NEXT:    vpblendvb %ymm8, %ymm4, %ymm5, %ymm5
; AVX2-SLOW-NEXT:    vmovdqa 144(%rdi), %xmm6
; AVX2-SLOW-NEXT:    vmovdqa 128(%rdi), %xmm4
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm7 = xmm4[0],xmm6[1],xmm4[2,3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u,u,u,u,2,3,12,13,6,7]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm7 = ymm5[0,1,2,3,4],ymm7[5,6,7],ymm5[8,9,10,11,12],ymm7[13,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm9 = ymm5[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm7 = ymm3[0],ymm2[1],ymm3[2],ymm2[3],ymm3[4,5],ymm2[6],ymm3[7,8],ymm2[9],ymm3[10],ymm2[11],ymm3[12,13],ymm2[14],ymm3[15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm7, %xmm5
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3],xmm7[4,5,6],xmm5[7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[2,3,12,13,6,7,0,1,10,11,4,5,14,15,10,11]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm7 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5],ymm0[6],ymm1[7,8],ymm0[9],ymm1[10,11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm10 = ymm7[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm10[5],ymm7[6,7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[2,3,12,13,6,7,0,1,10,11,6,7,8,9,8,9,18,19,28,29,22,23,16,17,26,27,22,23,24,25,24,25]
; AVX2-SLOW-NEXT:    vpblendvb %ymm8, %ymm5, %ymm7, %ymm5
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm7 = xmm4[0,1],xmm6[2],xmm4[3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u,u,u,u,4,5,14,15,8,9]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm7 = ymm5[0,1,2,3,4],ymm7[5,6,7],ymm5[8,9,10,11,12],ymm7[13,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm10 = ymm5[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm3[0,1],ymm2[2],ymm3[3],ymm2[4],ymm3[5,6],ymm2[7],ymm3[8,9],ymm2[10],ymm3[11],ymm2[12],ymm3[13,14],ymm2[15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm5, %xmm7
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm7[3,4],xmm5[5,6,7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm7 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5],ymm1[6],ymm0[7,8],ymm1[9],ymm0[10,11],ymm1[12],ymm0[13],ymm1[14],ymm0[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm8 = ymm7[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6],ymm7[7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,0,1,10,11,20,21,30,31,24,25,18,19,28,29,26,27,16,17,26,27]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm7 = xmm6[0],xmm4[1],xmm6[2,3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u,u,u,u,6,7,0,1,10,11]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm7 = ymm5[0,1,2,3,4],ymm7[5,6,7],ymm5[8,9,10,11,12],ymm7[13,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm8 = ymm5[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm2[0],ymm3[1,2],ymm2[3],ymm3[4],ymm2[5],ymm3[6,7],ymm2[8],ymm3[9,10],ymm2[11],ymm3[12],ymm2[13],ymm3[14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm5, %xmm7
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm7[0],xmm5[1],xmm7[2],xmm5[3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm7 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4,5],ymm0[6],ymm1[7,8],ymm0[9],ymm1[10],ymm0[11],ymm1[12,13],ymm0[14],ymm1[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm11 = ymm7[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm11[4],ymm7[5],ymm11[6],ymm7[7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,2,3,12,13,22,23,16,17,26,27,20,21,30,31,30,31,18,19,28,29]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm7 = xmm6[0,1],xmm4[2],xmm6[3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u,u,u,u,8,9,2,3,12,13]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm7 = ymm5[0,1,2,3,4],ymm7[5,6,7],ymm5[8,9,10,11,12],ymm7[13,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5],ymm2[6],ymm3[7,8],ymm2[9],ymm3[10,11],ymm2[12],ymm3[13],ymm2[14],ymm3[15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3,4],xmm3[5,6,7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[8,9,2,3,12,13,6,7,0,1,10,11,u,u,u,u]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3],ymm0[4],ymm1[5,6],ymm0[7],ymm1[8,9],ymm0[10],ymm1[11],ymm0[12],ymm1[13,14],ymm0[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15,24,25,18,19,28,29,22,23,u,u,u,u,u,u,u,u]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm6[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm4[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT:    vmovdqa %ymm9, (%rsi)
; AVX2-SLOW-NEXT:    vmovdqa %ymm10, (%rdx)
; AVX2-SLOW-NEXT:    vmovdqa %ymm8, (%rcx)
; AVX2-SLOW-NEXT:    vmovdqa %ymm5, (%r8)
; AVX2-SLOW-NEXT:    vmovdqa %ymm0, (%r9)
; AVX2-SLOW-NEXT:    vzeroupper
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-ALL-LABEL: vf16:
; AVX2-FAST-ALL:       # %bb.0:
; AVX2-FAST-ALL-NEXT:    vmovdqa 128(%rdi), %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdi), %ymm3
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rdi), %ymm4
; AVX2-FAST-ALL-NEXT:    vmovdqa 64(%rdi), %ymm1
; AVX2-FAST-ALL-NEXT:    vmovdqa 96(%rdi), %ymm2
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm5 = ymm3[0],ymm4[1],ymm3[2,3],ymm4[4],ymm3[5],ymm4[6],ymm3[7,8],ymm4[9],ymm3[10,11],ymm4[12],ymm3[13],ymm4[14],ymm3[15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm5, %xmm6
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0],xmm6[1,2,3],xmm5[4,5],xmm6[6,7]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[0,1,10,11,4,5,14,15,8,9,2,3,12,13,6,7]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm6 = ymm1[0],ymm2[1,2],ymm1[3],ymm2[4],ymm1[5],ymm2[6,7],ymm1[8],ymm2[9,10],ymm1[11],ymm2[12],ymm1[13],ymm2[14,15]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm7 = [1,3,0,2,4,6,1,3]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm6, %ymm7, %ymm6
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[0,1,6,7,8,9,14,15,4,5,14,15,4,5,2,3,16,17,22,23,24,25,30,31,20,21,30,31,20,21,18,19]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm8, %ymm5, %ymm6, %ymm5
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm6 = <u,u,u,u,0,3,5,u>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm6, %ymm6
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm11 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm11, %ymm6, %ymm6
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2,3,4],ymm6[5,6,7],ymm5[8,9,10,11,12],ymm6[13,14,15]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm9 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm6 = ymm4[0],ymm3[1],ymm4[2],ymm3[3],ymm4[4,5],ymm3[6],ymm4[7,8],ymm3[9],ymm4[10],ymm3[11],ymm4[12,13],ymm3[14],ymm4[15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm6, %xmm5
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3],xmm6[4,5,6],xmm5[7]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[2,3,12,13,6,7,0,1,10,11,4,5,14,15,10,11]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm6 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5],ymm1[6],ymm2[7,8],ymm1[9],ymm2[10,11],ymm1[12],ymm2[13],ymm1[14],ymm2[15]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm10 = <2,u,u,u,4,7,1,6>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm6, %ymm10, %ymm6
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[2,3,4,5,10,11,0,1,14,15,2,3,12,13,0,1,18,19,20,21,26,27,16,17,30,31,18,19,28,29,16,17]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm8, %ymm5, %ymm6, %ymm5
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm6 = <u,u,u,u,1,3,6,u>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm6, %ymm6
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm8 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm8, %ymm6, %ymm6
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2,3,4],ymm6[5,6,7],ymm5[8,9,10,11,12],ymm6[13,14,15]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm12 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm5 = ymm4[0,1],ymm3[2],ymm4[3],ymm3[4],ymm4[5,6],ymm3[7],ymm4[8,9],ymm3[10],ymm4[11],ymm3[12],ymm4[13,14],ymm3[15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm5, %xmm7
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm7[3,4],xmm5[5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm7 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5],ymm2[6],ymm1[7,8],ymm2[9],ymm1[10,11],ymm2[12],ymm1[13],ymm2[14],ymm1[15]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm10 = <0,2,u,u,5,7,2,4>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm7, %ymm10, %ymm7
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,20,21,22,23,16,17,22,23]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm7 = [1,4,6,0,1,4,6,0]
; AVX2-FAST-ALL-NEXT:    # ymm7 = mem[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm7, %ymm7
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm11, %ymm7, %ymm7
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm7 = ymm5[0,1,2,3,4],ymm7[5,6,7],ymm5[8,9,10,11,12],ymm7[13,14,15]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm7 = ymm5[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm5 = ymm3[0],ymm4[1,2],ymm3[3],ymm4[4],ymm3[5],ymm4[6,7],ymm3[8],ymm4[9,10],ymm3[11],ymm4[12],ymm3[13],ymm4[14,15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm5, %xmm6
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2],xmm5[3]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm6 = ymm2[0],ymm1[1],ymm2[2],ymm1[3],ymm2[4,5],ymm1[6],ymm2[7,8],ymm1[9],ymm2[10],ymm1[11],ymm2[12,13],ymm1[14],ymm2[15]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm10 = <0,3,u,u,5,0,2,7>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm6, %ymm10, %ymm6
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,u,u,u,u,u,u,u,u,2,3,4,5,18,19,20,21,26,27,16,17,30,31,30,31,18,19,20,21]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm6 = [2,4,7,0,2,4,7,0]
; AVX2-FAST-ALL-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm6, %ymm6
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm8, %ymm6, %ymm6
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm6 = ymm5[0,1,2,3,4],ymm6[5,6,7],ymm5[8,9,10,11,12],ymm6[13,14,15]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5],ymm3[6],ymm4[7,8],ymm3[9],ymm4[10,11],ymm3[12],ymm4[13],ymm3[14],ymm4[15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3,4],xmm4[5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} xmm3 = xmm3[8,9,2,3,12,13,6,7,0,1,10,11,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3],ymm1[4],ymm2[5,6],ymm1[7],ymm2[8,9],ymm1[10],ymm2[11],ymm1[12],ymm2[13,14],ymm1[15]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm2 = <1,3,u,u,6,0,3,5>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm1, %ymm2, %ymm1
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,u,u,u,u,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5],ymm3[6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm2 = <u,u,u,u,0,2,5,7>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm2, %ymm0
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25,30,31]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm9, (%rsi)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm12, (%rdx)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm7, (%rcx)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm5, (%r8)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm0, (%r9)
; 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), %ymm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 64(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 96(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm4 = ymm2[0],ymm3[1],ymm2[2,3],ymm3[4],ymm2[5],ymm3[6],ymm2[7,8],ymm3[9],ymm2[10,11],ymm3[12],ymm2[13],ymm3[14],ymm2[15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm4, %xmm5
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0],xmm5[1,2,3],xmm4[4,5],xmm5[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm4 = xmm4[0,1,10,11,4,5,14,15,8,9,2,3,12,13,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm5 = ymm0[0],ymm1[1,2],ymm0[3],ymm1[4],ymm0[5],ymm1[6,7],ymm0[8],ymm1[9,10],ymm0[11],ymm1[12],ymm0[13],ymm1[14,15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm6 = ymm5[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5],ymm5[6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm5 = ymm5[0,1,10,11,4,5,14,15,8,9,10,11,4,5,6,7,16,17,26,27,20,21,30,31,24,25,26,27,20,21,22,23]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm8, %ymm4, %ymm5, %ymm6
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 144(%rdi), %xmm4
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 128(%rdi), %xmm5
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm7 = xmm5[0],xmm4[1],xmm5[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u,u,u,u,2,3,12,13,6,7]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm7 = ymm6[0,1,2,3,4],ymm7[5,6,7],ymm6[8,9,10,11,12],ymm7[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm9 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm7 = ymm3[0],ymm2[1],ymm3[2],ymm2[3],ymm3[4,5],ymm2[6],ymm3[7,8],ymm2[9],ymm3[10],ymm2[11],ymm3[12,13],ymm2[14],ymm3[15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm7, %xmm6
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3],xmm7[4,5,6],xmm6[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[2,3,12,13,6,7,0,1,10,11,4,5,14,15,10,11]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm7 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5],ymm0[6],ymm1[7,8],ymm0[9],ymm1[10,11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm10 = ymm7[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm10[5],ymm7[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[2,3,12,13,6,7,0,1,10,11,6,7,8,9,8,9,18,19,28,29,22,23,16,17,26,27,22,23,24,25,24,25]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm8, %ymm6, %ymm7, %ymm6
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm7 = xmm5[0,1],xmm4[2],xmm5[3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u,u,u,u,4,5,14,15,8,9]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm7 = ymm6[0,1,2,3,4],ymm7[5,6,7],ymm6[8,9,10,11,12],ymm7[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm10 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm6 = ymm3[0,1],ymm2[2],ymm3[3],ymm2[4],ymm3[5,6],ymm2[7],ymm3[8,9],ymm2[10],ymm3[11],ymm2[12],ymm3[13,14],ymm2[15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm6, %xmm7
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm6 = xmm6[0,1,2],xmm7[3,4],xmm6[5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm7 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5],ymm1[6],ymm0[7,8],ymm1[9],ymm0[10,11],ymm1[12],ymm0[13],ymm1[14],ymm0[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm8 = ymm7[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6],ymm7[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,0,1,10,11,20,21,30,31,24,25,18,19,28,29,26,27,16,17,26,27]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm7 = xmm4[0],xmm5[1],xmm4[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u,u,u,u,6,7,0,1,10,11]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm7 = ymm6[0,1,2,3,4],ymm7[5,6,7],ymm6[8,9,10,11,12],ymm7[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm8 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm6 = ymm2[0],ymm3[1,2],ymm2[3],ymm3[4],ymm2[5],ymm3[6,7],ymm2[8],ymm3[9,10],ymm2[11],ymm3[12],ymm2[13],ymm3[14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm6, %xmm7
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm6 = xmm7[0],xmm6[1],xmm7[2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm6 = xmm6[6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm7 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4,5],ymm0[6],ymm1[7,8],ymm0[9],ymm1[10],ymm0[11],ymm1[12,13],ymm0[14],ymm1[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm11 = ymm7[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm11[4],ymm7[5],ymm11[6],ymm7[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u,u,u,u,u,u,u,u,2,3,12,13,22,23,16,17,26,27,20,21,30,31,30,31,18,19,28,29]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm7 = xmm4[0,1],xmm5[2],xmm4[3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u,u,u,u,8,9,2,3,12,13]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm7 = ymm6[0,1,2,3,4],ymm7[5,6,7],ymm6[8,9,10,11,12],ymm7[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5],ymm2[6],ymm3[7,8],ymm2[9],ymm3[10,11],ymm2[12],ymm3[13],ymm2[14],ymm3[15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3,4],xmm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[8,9,2,3,12,13,6,7,0,1,10,11,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3],ymm0[4],ymm1[5,6],ymm0[7],ymm1[8,9],ymm0[10],ymm1[11],ymm0[12],ymm1[13,14],ymm0[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15,24,25,18,19,28,29,22,23,u,u,u,u,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5],ymm2[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm1 = xmm4[12,13,14,15,4,5,14,15,u,u,u,u,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm2 = xmm5[0,1,2,3,0,1,10,11,u,u,u,u,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm9, (%rsi)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm10, (%rdx)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm8, (%rcx)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm6, (%r8)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm0, (%r9)
; AVX2-FAST-PERLANE-NEXT:    vzeroupper
; AVX2-FAST-PERLANE-NEXT:    retq
;
; AVX512-LABEL: vf16:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa 128(%rdi), %ymm0
; AVX512-NEXT:    vmovdqu64 (%rdi), %zmm1
; AVX512-NEXT:    vmovdqu64 64(%rdi), %zmm2
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm3 = <0,5,10,15,20,25,30,35,40,45,50,55,60,u,u,u>
; AVX512-NEXT:    vpermi2w %zmm2, %zmm1, %zmm3
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,2,3,4,5,6,7,8,9,10,11,12,17,22,27]
; AVX512-NEXT:    vpermi2w %ymm0, %ymm3, %ymm4
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm3 = <1,6,11,16,21,26,31,36,41,46,51,56,61,u,u,u>
; AVX512-NEXT:    vpermi2w %zmm2, %zmm1, %zmm3
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm5 = [0,1,2,3,4,5,6,7,8,9,10,11,12,18,23,28]
; AVX512-NEXT:    vpermi2w %ymm0, %ymm3, %ymm5
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm3 = <34,39,44,49,54,59,0,5,10,15,20,25,30,u,u,u>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm2, %zmm3
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm6 = [0,1,2,3,4,5,6,7,8,9,10,11,12,19,24,29]
; AVX512-NEXT:    vpermi2w %ymm0, %ymm3, %ymm6
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm3 = <35,40,45,50,55,60,1,6,11,16,21,26,31,u,u,u>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm2, %zmm3
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm7 = [0,1,2,3,4,5,6,7,8,9,10,11,12,20,25,30]
; AVX512-NEXT:    vpermi2w %ymm0, %ymm3, %ymm7
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm3 = <4,9,14,19,24,29,34,39,44,49,54,59,u,u,u,u>
; AVX512-NEXT:    vpermi2w %zmm2, %zmm1, %zmm3
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,1,2,3,4,5,6,7,8,9,10,11,16,21,26,31]
; AVX512-NEXT:    vpermi2w %ymm0, %ymm3, %ymm1
; AVX512-NEXT:    vmovdqa %ymm4, (%rsi)
; AVX512-NEXT:    vmovdqa %ymm5, (%rdx)
; AVX512-NEXT:    vmovdqa %ymm6, (%rcx)
; AVX512-NEXT:    vmovdqa %ymm7, (%r8)
; AVX512-NEXT:    vmovdqa %ymm1, (%r9)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %wide.vec = load <80 x i16>, <80 x i16>* %in.vec, align 32

  %strided.vec0 = shufflevector <80 x i16> %wide.vec, <80 x i16> poison, <16 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75>
  %strided.vec1 = shufflevector <80 x i16> %wide.vec, <80 x i16> poison, <16 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76>
  %strided.vec2 = shufflevector <80 x i16> %wide.vec, <80 x i16> poison, <16 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77>
  %strided.vec3 = shufflevector <80 x i16> %wide.vec, <80 x i16> poison, <16 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78>
  %strided.vec4 = shufflevector <80 x i16> %wide.vec, <80 x i16> poison, <16 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79>

  store <16 x i16> %strided.vec0, <16 x i16>* %out.vec0, align 32
  store <16 x i16> %strided.vec1, <16 x i16>* %out.vec1, align 32
  store <16 x i16> %strided.vec2, <16 x i16>* %out.vec2, align 32
  store <16 x i16> %strided.vec3, <16 x i16>* %out.vec3, align 32
  store <16 x i16> %strided.vec4, <16 x i16>* %out.vec4, align 32

  ret void
}

define void @vf32(<160 x i16>* %in.vec, <32 x i16>* %out.vec0, <32 x i16>* %out.vec1, <32 x i16>* %out.vec2, <32 x i16>* %out.vec3, <32 x i16>* %out.vec4) nounwind {
; SSE-LABEL: vf32:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $344, %rsp # imm = 0x158
; SSE-NEXT:    movdqa 304(%rdi), %xmm9
; SSE-NEXT:    movdqa 240(%rdi), %xmm10
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 256(%rdi), %xmm11
; SSE-NEXT:    movdqa 288(%rdi), %xmm12
; SSE-NEXT:    movdqa 272(%rdi), %xmm7
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 144(%rdi), %xmm4
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 80(%rdi), %xmm6
; SSE-NEXT:    movdqa 96(%rdi), %xmm8
; SSE-NEXT:    movdqa 128(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 112(%rdi), %xmm15
; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,0,65535,65535,65535]
; SSE-NEXT:    movdqa %xmm0, %xmm1
; SSE-NEXT:    pandn %xmm15, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,1,0,3]
; SSE-NEXT:    pand %xmm0, %xmm2
; SSE-NEXT:    por %xmm1, %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[0,2,2,3]
; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
; SSE-NEXT:    movaps {{.*#+}} xmm13 = [65535,65535,65535,65535,65535,65535,65535,0]
; SSE-NEXT:    andps %xmm13, %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,1,0,1]
; SSE-NEXT:    movaps %xmm13, %xmm2
; SSE-NEXT:    pandn %xmm1, %xmm2
; SSE-NEXT:    por %xmm3, %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm0, %xmm1
; SSE-NEXT:    pandn %xmm7, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[0,1,0,3]
; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pand %xmm0, %xmm2
; SSE-NEXT:    por %xmm1, %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,1,2,3]
; SSE-NEXT:    movdqa %xmm11, %xmm4
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,1,0,1]
; SSE-NEXT:    movaps %xmm13, %xmm2
; SSE-NEXT:    andnps %xmm1, %xmm2
; SSE-NEXT:    movdqa 32(%rdi), %xmm5
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    andps %xmm13, %xmm3
; SSE-NEXT:    orps %xmm3, %xmm2
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm0, %xmm1
; SSE-NEXT:    pandn %xmm5, %xmm1
; SSE-NEXT:    movdqa 48(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
; SSE-NEXT:    pand %xmm0, %xmm2
; SSE-NEXT:    por %xmm1, %xmm2
; SSE-NEXT:    movdqa 16(%rdi), %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[3,1,2,3]
; SSE-NEXT:    movdqa %xmm3, %xmm14
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT:    movdqa (%rdi), %xmm5
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
; SSE-NEXT:    movdqa 64(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
; SSE-NEXT:    movaps %xmm13, %xmm2
; SSE-NEXT:    andnps %xmm1, %xmm2
; SSE-NEXT:    andps %xmm13, %xmm3
; SSE-NEXT:    orps %xmm3, %xmm2
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 208(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
; SSE-NEXT:    pand %xmm0, %xmm1
; SSE-NEXT:    movdqa 192(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pandn %xmm2, %xmm0
; SSE-NEXT:    por %xmm1, %xmm0
; SSE-NEXT:    movdqa 176(%rdi), %xmm11
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,1,2,3]
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT:    movdqa 160(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
; SSE-NEXT:    movdqa 224(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
; SSE-NEXT:    movaps %xmm13, %xmm1
; SSE-NEXT:    andnps %xmm0, %xmm1
; SSE-NEXT:    andps %xmm13, %xmm2
; SSE-NEXT:    orps %xmm2, %xmm1
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm8, %xmm10
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm8, %xmm0
; SSE-NEXT:    psrlq $48, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,3,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [0,0,0,65535,65535,65535,65535,65535]
; SSE-NEXT:    movdqa %xmm0, %xmm2
; SSE-NEXT:    pandn %xmm1, %xmm2
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[1,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm15[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,7,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
; SSE-NEXT:    pand %xmm0, %xmm1
; SSE-NEXT:    por %xmm2, %xmm1
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm7, %xmm2
; SSE-NEXT:    psllq $48, %xmm2
; SSE-NEXT:    movaps %xmm13, %xmm3
; SSE-NEXT:    andnps %xmm2, %xmm3
; SSE-NEXT:    pand %xmm13, %xmm1
; SSE-NEXT:    orps %xmm1, %xmm3
; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm4, %xmm1
; SSE-NEXT:    psrlq $48, %xmm1
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[0,3,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    movdqa %xmm0, %xmm1
; SSE-NEXT:    pandn %xmm2, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[1,3,2,3]
; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,4,7]
; SSE-NEXT:    pand %xmm0, %xmm2
; SSE-NEXT:    por %xmm1, %xmm2
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm9, %xmm1
; SSE-NEXT:    psllq $48, %xmm1
; SSE-NEXT:    movdqa %xmm13, %xmm3
; SSE-NEXT:    pandn %xmm1, %xmm3
; SSE-NEXT:    pand %xmm13, %xmm2
; SSE-NEXT:    por %xmm2, %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm14, %xmm1
; SSE-NEXT:    psrlq $48, %xmm1
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,3,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    movdqa %xmm0, %xmm1
; SSE-NEXT:    pandn %xmm2, %xmm1
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[1,3,2,3]
; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,4,7]
; SSE-NEXT:    pand %xmm0, %xmm2
; SSE-NEXT:    por %xmm1, %xmm2
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    psllq $48, %xmm1
; SSE-NEXT:    movdqa %xmm13, %xmm3
; SSE-NEXT:    pandn %xmm1, %xmm3
; SSE-NEXT:    pand %xmm13, %xmm2
; SSE-NEXT:    por %xmm2, %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    psrlq $48, %xmm11
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,3,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,3,2,3]
; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,7,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
; SSE-NEXT:    pand %xmm0, %xmm1
; SSE-NEXT:    pandn %xmm2, %xmm0
; SSE-NEXT:    por %xmm1, %xmm0
; SSE-NEXT:    pand %xmm13, %xmm0
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    psllq $48, %xmm1
; SSE-NEXT:    pandn %xmm1, %xmm13
; SSE-NEXT:    por %xmm0, %xmm13
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,1,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,7,6,7]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm10[2],xmm1[3],xmm10[3]
; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,65535,0,0,0,65535,65535]
; SSE-NEXT:    pand %xmm11, %xmm1
; SSE-NEXT:    movdqa %xmm15, (%rsp) # 16-byte Spill
; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,0],xmm15[0,0]
; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm15[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm8[1,1,1,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
; SSE-NEXT:    movdqa %xmm11, %xmm0
; SSE-NEXT:    pandn %xmm2, %xmm0
; SSE-NEXT:    por %xmm1, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[0,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm2[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,1,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3]
; SSE-NEXT:    pand %xmm11, %xmm1
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, %xmm2
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm6[0,0]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm6[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,1,1,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
; SSE-NEXT:    movdqa %xmm11, %xmm6
; SSE-NEXT:    pandn %xmm2, %xmm6
; SSE-NEXT:    por %xmm1, %xmm6
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm2[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm1[2,0]
; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,1,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm8[2],xmm1[3],xmm8[3]
; SSE-NEXT:    pand %xmm11, %xmm1
; SSE-NEXT:    movdqa %xmm14, %xmm2
; SSE-NEXT:    movdqa %xmm14, %xmm10
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm5[0,0]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm5[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,1,1,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
; SSE-NEXT:    movdqa %xmm11, %xmm14
; SSE-NEXT:    pandn %xmm2, %xmm14
; SSE-NEXT:    por %xmm1, %xmm14
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm2[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm1[2,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,1,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm9[2],xmm1[3],xmm9[3]
; SSE-NEXT:    pand %xmm11, %xmm1
; SSE-NEXT:    movdqa %xmm12, %xmm2
; SSE-NEXT:    movdqa %xmm12, %xmm15
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm4[0,0]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm4[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,1,1,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
; SSE-NEXT:    movdqa %xmm11, %xmm12
; SSE-NEXT:    pandn %xmm2, %xmm12
; SSE-NEXT:    por %xmm1, %xmm12
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm2[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm1[2,0]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps (%rsp), %xmm2 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,1,1]
; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,4,6,7]
; SSE-NEXT:    movdqa %xmm11, %xmm2
; SSE-NEXT:    pandn %xmm1, %xmm2
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm3[1,0,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm11, %xmm4
; SSE-NEXT:    por %xmm2, %xmm4
; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[0,1,0,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,6]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm1[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,0]
; SSE-NEXT:    movaps %xmm0, %xmm1
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,4,6,7]
; SSE-NEXT:    movdqa %xmm11, %xmm2
; SSE-NEXT:    pandn %xmm1, %xmm2
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,0,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm11, %xmm3
; SSE-NEXT:    por %xmm2, %xmm3
; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[0,1,0,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,6]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm1[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[2,0],xmm2[3,0]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm10[0,2]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,4,6,7]
; SSE-NEXT:    movdqa %xmm11, %xmm8
; SSE-NEXT:    pandn %xmm1, %xmm8
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[1,0,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm11, %xmm2
; SSE-NEXT:    por %xmm8, %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[0,1,0,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,5,6]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[2,0],xmm1[3,0]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm15[0,2]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = mem[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[0,2,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm1[0,1,2,3,7,4,6,7]
; SSE-NEXT:    movdqa %xmm11, %xmm1
; SSE-NEXT:    pandn %xmm8, %xmm1
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm11, %xmm0
; SSE-NEXT:    por %xmm1, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,1,0,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm8[2,3]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
; SSE-NEXT:    pshufd $85, (%rsp), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; SSE-NEXT:    # xmm15 = mem[0,2,2,3]
; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[0,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm15[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm15[1,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm8[0],xmm15[1],xmm8[1]
; SSE-NEXT:    movdqa %xmm11, %xmm8
; SSE-NEXT:    pandn %xmm15, %xmm8
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm5[3,0]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm7[0,2]
; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm5[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm7[2,1,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm11, %xmm10
; SSE-NEXT:    por %xmm8, %xmm10
; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm1[2,0]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT:    # xmm7 = mem[0,2,2,3]
; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[0,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm7[1,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
; SSE-NEXT:    movdqa %xmm11, %xmm8
; SSE-NEXT:    pandn %xmm7, %xmm8
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm7[3,0]
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm5[0,2]
; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm7[2,1,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm11, %xmm15
; SSE-NEXT:    por %xmm8, %xmm15
; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm1[2,0]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT:    # xmm7 = mem[0,2,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm7[1,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
; SSE-NEXT:    movdqa %xmm11, %xmm8
; SSE-NEXT:    pandn %xmm7, %xmm8
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm9[3,0]
; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm5[0,2]
; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm9[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm7[2,1,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm11, %xmm9
; SSE-NEXT:    por %xmm8, %xmm9
; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm1[2,0]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT:    # xmm7 = mem[0,2,2,3]
; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[0,1,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm7[1,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm6[3,0]
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm5[0,2]
; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,6,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[2,1,3,3,4,5,6,7]
; SSE-NEXT:    pand %xmm11, %xmm6
; SSE-NEXT:    pandn %xmm7, %xmm11
; SSE-NEXT:    por %xmm6, %xmm11
; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm1[2,0]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 32(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, (%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 48(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 16(%rsi)
; SSE-NEXT:    movdqa %xmm13, 32(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, (%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 48(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 16(%rdx)
; SSE-NEXT:    movaps %xmm12, 32(%rcx)
; SSE-NEXT:    movaps %xmm14, (%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 48(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 16(%rcx)
; SSE-NEXT:    movaps %xmm0, 32(%r8)
; SSE-NEXT:    movaps %xmm2, (%r8)
; SSE-NEXT:    movaps %xmm3, 48(%r8)
; SSE-NEXT:    movaps %xmm4, 16(%r8)
; SSE-NEXT:    movaps %xmm11, (%r9)
; SSE-NEXT:    movaps %xmm9, 32(%r9)
; SSE-NEXT:    movaps %xmm15, 48(%r9)
; SSE-NEXT:    movaps %xmm10, 16(%r9)
; SSE-NEXT:    addq $344, %rsp # imm = 0x158
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf32:
; AVX1:       # %bb.0:
; AVX1-NEXT:    subq $424, %rsp # imm = 0x1A8
; AVX1-NEXT:    vmovaps 224(%rdi), %xmm0
; AVX1-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
; AVX1-NEXT:    vmovdqa 256(%rdi), %xmm1
; AVX1-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; AVX1-NEXT:    vmovdqa 272(%rdi), %xmm2
; AVX1-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX1-NEXT:    vmovdqa 240(%rdi), %xmm2
; AVX1-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3,4,5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX1-NEXT:    vmovaps {{.*#+}} ymm8 = [65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,65535,65535,65535]
; AVX1-NEXT:    vandnps %ymm0, %ymm8, %ymm0
; AVX1-NEXT:    vmovdqa 208(%rdi), %xmm1
; AVX1-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
; AVX1-NEXT:    vmovdqa 192(%rdi), %xmm2
; AVX1-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4],xmm1[5,6,7]
; AVX1-NEXT:    vmovdqa 176(%rdi), %xmm2
; AVX1-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
; AVX1-NEXT:    vmovdqa 160(%rdi), %xmm2
; AVX1-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
; AVX1-NEXT:    vandps %ymm1, %ymm8, %ymm1
; AVX1-NEXT:    vorps %ymm0, %ymm1, %ymm1
; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT:    vmovdqa 304(%rdi), %xmm3
; AVX1-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vmovdqa 288(%rdi), %xmm2
; AVX1-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm2[0,1],xmm3[2,3],xmm2[4,5,6,7]
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm13 = [6,7,2,3,4,5,6,7,6,7,2,3,12,13,6,7]
; AVX1-NEXT:    vpshufb %xmm13, %xmm3, %xmm3
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm3[5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-NEXT:    vmovdqa 96(%rdi), %xmm11
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm11[0,1,1,3]
; AVX1-NEXT:    vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
; AVX1-NEXT:    vmovdqa 112(%rdi), %xmm9
; AVX1-NEXT:    vpunpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm9[1]
; AVX1-NEXT:    vmovdqa 80(%rdi), %xmm10
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm10[0,2,2,3]
; AVX1-NEXT:    vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3,4,5,6,7]
; AVX1-NEXT:    vmovdqa 64(%rdi), %xmm2
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[0,1,0,1]
; AVX1-NEXT:    vmovdqa %xmm2, %xmm12
; AVX1-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm3
; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm0
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
; AVX1-NEXT:    vmovdqa %xmm0, %xmm7
; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; AVX1-NEXT:    vmovdqa (%rdi), %xmm0
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
; AVX1-NEXT:    vmovdqa %xmm0, %xmm4
; AVX1-NEXT:    vmovdqa %xmm0, (%rsp) # 16-byte Spill
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX1-NEXT:    vmovdqa 32(%rdi), %xmm6
; AVX1-NEXT:    vmovdqa 48(%rdi), %xmm0
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[0,1,0,3]
; AVX1-NEXT:    vmovdqa %xmm0, %xmm5
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm6[4],xmm2[5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm2[4,5,6,7]
; AVX1-NEXT:    vmovdqa 144(%rdi), %xmm2
; AVX1-NEXT:    vmovdqa 128(%rdi), %xmm1
; AVX1-NEXT:    vpblendw {{.*#+}} xmm15 = xmm1[0,1],xmm2[2,3],xmm1[4,5,6,7]
; AVX1-NEXT:    vpshufb %xmm13, %xmm15, %xmm15
; AVX1-NEXT:    vandnps %ymm3, %ymm8, %ymm3
; AVX1-NEXT:    vandps %ymm0, %ymm8, %ymm0
; AVX1-NEXT:    vorps %ymm3, %ymm0, %ymm0
; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm3
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm15[5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-NEXT:    vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm9[0,1],xmm11[2,3],xmm9[4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,0,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm10[0,3,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3,4,5,6,7]
; AVX1-NEXT:    vpsllq $48, %xmm12, %xmm3
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm3, %ymm0
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm4[0,3,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpsrlq $48, %xmm7, %xmm7
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm7 = xmm6[0,1],xmm5[2,3],xmm6[4,5],xmm5[6,7]
; AVX1-NEXT:    vmovdqa %xmm6, %xmm15
; AVX1-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vmovdqa %xmm5, %xmm10
; AVX1-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm4 = [0,1,0,1,0,1,0,1,10,11,4,5,14,15,6,7]
; AVX1-NEXT:    vpshufb %xmm4, %xmm7, %xmm7
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm7[3,4,5,6,7]
; AVX1-NEXT:    vandnps %ymm0, %ymm8, %ymm0
; AVX1-NEXT:    vandps %ymm3, %ymm8, %ymm3
; AVX1-NEXT:    vorps %ymm0, %ymm3, %ymm3
; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT:    vpblendw {{.*#+}} xmm7 = xmm1[0,1,2,3],xmm2[4,5],xmm1[6,7]
; AVX1-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufb {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,6,7,8,9,4,5,14,15,8,9]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm7[5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm3, %ymm0
; AVX1-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,0,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm14[0,3,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3,4,5,6,7]
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX1-NEXT:    vpsllq $48, %xmm3, %xmm3
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm3, %ymm0
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX1-NEXT:    vpblendw $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm3 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm3 = mem[0,1],xmm13[2,3],mem[4,5],xmm13[6,7]
; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm3
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm11[0,3,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm7[1,2,2,3,4,5,6,7]
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX1-NEXT:    vpsrlq $48, %xmm12, %xmm6
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm6[0,1,2],xmm3[3,4,5,6,7]
; AVX1-NEXT:    vandnps %ymm0, %ymm8, %ymm0
; AVX1-NEXT:    vandps %ymm3, %ymm8, %ymm3
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm4[0,1,2,3],xmm8[4,5],xmm4[6,7]
; AVX1-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6,7,8,9,4,5,14,15,8,9]
; AVX1-NEXT:    vorps %ymm0, %ymm3, %ymm3
; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm3, %ymm0
; AVX1-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-NEXT:    vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm0 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm0 = xmm9[0,1,2,3],mem[4,5],xmm9[6,7]
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm7 = <u,u,u,u,8,9,2,3,12,13,12,13,12,13,12,13>
; AVX1-NEXT:    vpshufb %xmm7, %xmm0, %xmm0
; AVX1-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm5 = mem[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm5[2,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm5[0,1],xmm0[2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm2[0,1],xmm1[2,3],xmm2[4,5,6,7]
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
; AVX1-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm15[0,1],xmm10[2,3],xmm15[4,5,6,7]
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,3,2,3,2,3,2,3,12,13,6,7,12,13,14,15]
; AVX1-NEXT:    vpshufb %xmm1, %xmm5, %xmm5
; AVX1-NEXT:    vpshufd $212, (%rsp), %xmm3 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm3 = mem[0,1,1,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
; AVX1-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm5[3,4,5],xmm3[6,7]
; AVX1-NEXT:    vpshufd $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm5 = mem[0,1,2,0]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6,5]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm5[6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm3, %ymm0
; AVX1-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm8[0,1],xmm4[2,3],xmm8[4,5,6,7]
; AVX1-NEXT:    vmovdqa %xmm8, %xmm10
; AVX1-NEXT:    vmovdqa %xmm4, %xmm9
; AVX1-NEXT:    vpshufb %xmm6, %xmm0, %xmm0
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm6[0,1,2,3],xmm4[4,5],xmm6[6,7]
; AVX1-NEXT:    vpshufb %xmm7, %xmm3, %xmm3
; AVX1-NEXT:    vmovdqa %xmm14, %xmm2
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm14[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm5[2,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm14 = xmm3[0,1,2,3,4],xmm0[5,6,7]
; AVX1-NEXT:    vmovdqa %xmm13, %xmm8
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm7[0,1],xmm13[2,3],xmm7[4,5,6,7]
; AVX1-NEXT:    vpshufb %xmm1, %xmm3, %xmm3
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm11[0,1,1,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,7]
; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm5[2],xmm12[2],xmm5[3],xmm12[3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm5[0,1,2],xmm3[3,4,5],xmm5[6,7]
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm15[0,1,2,0]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6,5]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm5[6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm14, %ymm3, %ymm3
; AVX1-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm4[0,1],xmm6[2,3],xmm4[4,5],xmm6[6,7]
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = <u,u,0,1,10,11,4,5,14,15,14,15,14,15,14,15>
; AVX1-NEXT:    vpshufb %xmm3, %xmm5, %xmm4
; AVX1-NEXT:    vpsrlq $48, %xmm2, %xmm5
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm5[0],xmm4[1,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm10[0,1,2,3],xmm9[4,5],xmm10[6,7]
; AVX1-NEXT:    vmovdqa %xmm10, %xmm13
; AVX1-NEXT:    vmovdqa %xmm9, %xmm14
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
; AVX1-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm5[5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm12[0,1],xmm11[2,3],xmm12[4,5,6,7]
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm9 = [6,7,0,1,10,11,10,11,8,9,10,11,12,13,14,15]
; AVX1-NEXT:    vpshufb %xmm9, %xmm5, %xmm5
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0,1,2,3],xmm8[4,5],xmm7[6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,4,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3,4,5],xmm5[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm15[0,1,0,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,5,6]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5],xmm5[6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm0
; AVX1-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm2[4,5],xmm1[6,7]
; AVX1-NEXT:    vpshufb %xmm6, %xmm0, %xmm0
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm6[0,1],xmm12[2,3],xmm6[4,5],xmm12[6,7]
; AVX1-NEXT:    vpshufb %xmm3, %xmm4, %xmm3
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-NEXT:    vpsrlq $48, %xmm5, %xmm4
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm8 = xmm3[0,1,2,3,4],xmm0[5,6,7]
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vmovdqa (%rsp), %xmm7 # 16-byte Reload
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm0[0,1],xmm7[2,3],xmm0[4,5,6,7]
; AVX1-NEXT:    vpshufb %xmm9, %xmm3, %xmm3
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm11[0,1,2,3],xmm10[4,5],xmm11[6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,4,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3,4,5],xmm3[6,7]
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm9[0,1,0,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,5,6]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm4[6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm8, %ymm3, %ymm8
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm6[0,3,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm12[3],xmm2[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm5[2,3,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm0[0,1,2,3],xmm7[4,5],xmm0[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm11[1,1,1,1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm10[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX1-NEXT:    vmovdqa {{.*#+}} xmm4 = [8,9,2,3,12,13,12,13,8,9,12,13,12,13,14,15]
; AVX1-NEXT:    vpshufb %xmm4, %xmm2, %xmm2
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3,4,5],xmm2[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm9[0,1,1,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm3[6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm13[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,1,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm14[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX1-NEXT:    vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm3 = mem[0,3,2,3]
; AVX1-NEXT:    vpblendw $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm3 = xmm3[0,1,2],mem[3],xmm3[4,5,6,7]
; AVX1-NEXT:    vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm5 = mem[2,3,2,3]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm5[0],xmm3[1,2,3,4,5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-NEXT:    vpblendw $207, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm3 = mem[0,1,2,3],xmm0[4,5],mem[6,7]
; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm3
; AVX1-NEXT:    vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm4 = mem[1,1,1,1]
; AVX1-NEXT:    vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm5 = mem[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm5[0,1,0,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3,4,5],xmm3[6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm15[0,1,1,3]
; AVX1-NEXT:    vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm4[6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-NEXT:    vmovaps %ymm3, (%rsi)
; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-NEXT:    vmovaps %ymm3, 32(%rsi)
; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-NEXT:    vmovaps %ymm0, 32(%rdx)
; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-NEXT:    vmovaps %ymm0, (%rdx)
; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-NEXT:    vmovaps %ymm0, 32(%rcx)
; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-NEXT:    vmovaps %ymm0, (%rcx)
; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-NEXT:    vmovaps %ymm0, 32(%r8)
; AVX1-NEXT:    vmovaps %ymm8, (%r8)
; AVX1-NEXT:    vmovaps %ymm2, 32(%r9)
; AVX1-NEXT:    vmovaps %ymm1, (%r9)
; AVX1-NEXT:    addq $424, %rsp # imm = 0x1A8
; AVX1-NEXT:    vzeroupper
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf32:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    subq $280, %rsp # imm = 0x118
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %ymm4
; AVX2-SLOW-NEXT:    vmovdqa 32(%rdi), %ymm5
; AVX2-SLOW-NEXT:    vmovdqa 64(%rdi), %ymm15
; AVX2-SLOW-NEXT:    vmovdqa 96(%rdi), %ymm9
; AVX2-SLOW-NEXT:    vmovdqa 192(%rdi), %ymm3
; AVX2-SLOW-NEXT:    vmovdqa 160(%rdi), %ymm10
; AVX2-SLOW-NEXT:    vmovdqa 224(%rdi), %ymm1
; AVX2-SLOW-NEXT:    vmovdqa 256(%rdi), %ymm2
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0],ymm2[1,2],ymm1[3],ymm2[4],ymm1[5],ymm2[6,7],ymm1[8],ymm2[9,10],ymm1[11],ymm2[12],ymm1[13],ymm2[14,15]
; AVX2-SLOW-NEXT:    vmovdqa %ymm2, %ymm8
; AVX2-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa %ymm1, %ymm7
; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6],ymm1[7]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,10,11,4,5,14,15,8,9,10,11,4,5,6,7,16,17,26,27,20,21,30,31,24,25,26,27,20,21,22,23]
; AVX2-SLOW-NEXT:    vpshufb %ymm2, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm1 = ymm10[0],ymm3[1],ymm10[2,3],ymm3[4],ymm10[5],ymm3[6],ymm10[7,8],ymm3[9],ymm10[10,11],ymm3[12],ymm10[13],ymm3[14],ymm10[15]
; AVX2-SLOW-NEXT:    vmovdqa %ymm10, %ymm13
; AVX2-SLOW-NEXT:    vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa %ymm3, %ymm14
; AVX2-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm3
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm3[1,2,3],xmm1[4,5],xmm3[6,7]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm12 = [0,1,10,11,4,5,14,15,8,9,2,3,12,13,6,7]
; AVX2-SLOW-NEXT:    vpshufb %xmm12, %xmm1, %xmm1
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-SLOW-NEXT:    vpblendvb %ymm10, %ymm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm15[0],ymm9[1,2],ymm15[3],ymm9[4],ymm15[5],ymm9[6,7],ymm15[8],ymm9[9,10],ymm15[11],ymm9[12],ymm15[13],ymm9[14,15]
; AVX2-SLOW-NEXT:    vmovdqa %ymm9, %ymm11
; AVX2-SLOW-NEXT:    vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm9 = ymm0[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm9[5],ymm0[6],ymm9[7]
; AVX2-SLOW-NEXT:    vpshufb %ymm2, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm2 = ymm4[0],ymm5[1],ymm4[2,3],ymm5[4],ymm4[5],ymm5[6],ymm4[7,8],ymm5[9],ymm4[10,11],ymm5[12],ymm4[13],ymm5[14],ymm4[15]
; AVX2-SLOW-NEXT:    vmovdqa %ymm4, %ymm6
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1,2,3],xmm2[4,5],xmm3[6,7]
; AVX2-SLOW-NEXT:    vpshufb %xmm12, %xmm2, %xmm2
; AVX2-SLOW-NEXT:    vpblendvb %ymm10, %ymm2, %ymm0, %ymm9
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm8[0],ymm7[1],ymm8[2,3],ymm7[4],ymm8[5],ymm7[6],ymm8[7,8],ymm7[9],ymm8[10,11],ymm7[12],ymm8[13],ymm7[14],ymm8[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,12,13,6,7,0,1,10,11,6,7,8,9,8,9,18,19,28,29,22,23,16,17,26,27,22,23,24,25,24,25]
; AVX2-SLOW-NEXT:    vpshufb %ymm2, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm3 = ymm14[0],ymm13[1],ymm14[2],ymm13[3],ymm14[4,5],ymm13[6],ymm14[7,8],ymm13[9],ymm14[10],ymm13[11],ymm14[12,13],ymm13[14],ymm14[15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm3, %xmm4
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6],xmm4[7]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [2,3,12,13,6,7,0,1,10,11,4,5,14,15,10,11]
; AVX2-SLOW-NEXT:    vpshufb %xmm4, %xmm3, %xmm3
; AVX2-SLOW-NEXT:    vpblendvb %ymm10, %ymm3, %ymm0, %ymm12
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm11[0],ymm15[1],ymm11[2,3],ymm15[4],ymm11[5],ymm15[6],ymm11[7,8],ymm15[9],ymm11[10,11],ymm15[12],ymm11[13],ymm15[14],ymm11[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm0[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5],ymm0[6,7]
; AVX2-SLOW-NEXT:    vpshufb %ymm2, %ymm0, %ymm3
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm5[0],ymm6[1],ymm5[2],ymm6[3],ymm5[4,5],ymm6[6],ymm5[7,8],ymm6[9],ymm5[10],ymm6[11],ymm5[12,13],ymm6[14],ymm5[15]
; AVX2-SLOW-NEXT:    vmovdqa %ymm6, %ymm13
; AVX2-SLOW-NEXT:    vmovdqa %ymm5, %ymm14
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm2
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm2 = xmm0[0,1],xmm2[2,3],xmm0[4,5,6],xmm2[7]
; AVX2-SLOW-NEXT:    vmovdqa 304(%rdi), %xmm7
; AVX2-SLOW-NEXT:    vpshufb %xmm4, %xmm2, %xmm4
; AVX2-SLOW-NEXT:    vmovdqa 288(%rdi), %xmm8
; AVX2-SLOW-NEXT:    vpblendvb %ymm10, %ymm4, %ymm3, %ymm4
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm3 = xmm8[0],xmm7[1],xmm8[2,3]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm10 = [6,7,2,3,4,5,6,7,6,7,2,3,12,13,6,7]
; AVX2-SLOW-NEXT:    vpshufb %xmm10, %xmm3, %xmm3
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm3 = ymm1[0,1,2,3,4],ymm3[5,6,7],ymm1[8,9,10,11,12],ymm3[13,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa 144(%rdi), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa 128(%rdi), %xmm0
; AVX2-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm11 = xmm0[0],xmm1[1],xmm0[2,3]
; AVX2-SLOW-NEXT:    vpshufb %xmm10, %xmm11, %xmm5
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm9[0,1,2,3,4],ymm5[5,6,7],ymm9[8,9,10,11,12],ymm5[13,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm8[0,1],xmm7[2],xmm8[3]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm9 = [0,1,2,3,4,5,6,7,8,9,4,5,14,15,8,9]
; AVX2-SLOW-NEXT:    vpshufb %xmm9, %xmm5, %xmm5
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm12[0,1,2,3,4],ymm5[5,6,7],ymm12[8,9,10,11,12],ymm5[13,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm12[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm0[0,1],xmm1[2],xmm0[3]
; AVX2-SLOW-NEXT:    vmovdqa %xmm1, %xmm10
; AVX2-SLOW-NEXT:    vpshufb %xmm9, %xmm5, %xmm5
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm4 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5],ymm1[6],ymm0[7,8],ymm1[9],ymm0[10,11],ymm1[12],ymm0[13],ymm1[14],ymm0[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6],ymm4[7]
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm12[0,1],ymm2[2],ymm12[3],ymm2[4],ymm12[5,6],ymm2[7],ymm12[8,9],ymm2[10],ymm12[11],ymm2[12],ymm12[13,14],ymm2[15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm5, %xmm6
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm6[3,4],xmm5[5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,0,1,10,11,20,21,30,31,24,25,18,19,28,29,26,27,16,17,26,27>
; AVX2-SLOW-NEXT:    vpshufb %ymm6, %ymm4, %ymm4
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm9 = <4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u>
; AVX2-SLOW-NEXT:    vpshufb %xmm9, %xmm5, %xmm5
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm7[0],xmm8[1],xmm7[2,3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm15[0],ymm3[1],ymm15[2,3],ymm3[4],ymm15[5],ymm3[6],ymm15[7,8],ymm3[9],ymm15[10,11],ymm3[12],ymm15[13],ymm3[14],ymm15[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm11 = ymm5[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm11[6],ymm5[7]
; AVX2-SLOW-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm6 = ymm14[0,1],ymm13[2],ymm14[3],ymm13[4],ymm14[5,6],ymm13[7],ymm14[8,9],ymm13[10],ymm14[11],ymm13[12],ymm14[13,14],ymm13[15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm6, %xmm4
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm4 = xmm6[0,1,2],xmm4[3,4],xmm6[5,6,7]
; AVX2-SLOW-NEXT:    vpshufb %xmm9, %xmm4, %xmm4
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm10[0],xmm9[1],xmm10[2,3]
; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm4 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4,5],ymm0[6],ymm1[7,8],ymm0[9],ymm1[10],ymm0[11],ymm1[12,13],ymm0[14],ymm1[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5],ymm5[6],ymm4[7]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm2[0],ymm12[1,2],ymm2[3],ymm12[4],ymm2[5],ymm12[6,7],ymm2[8],ymm12[9,10],ymm2[11],ymm12[12],ymm2[13],ymm12[14,15]
; AVX2-SLOW-NEXT:    vmovdqa %ymm12, %ymm0
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm5, %xmm6
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2],xmm5[3]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,2,3,12,13,22,23,16,17,26,27,20,21,30,31,30,31,18,19,28,29>
; AVX2-SLOW-NEXT:    vpshufb %ymm6, %ymm4, %ymm4
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm2 = <6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u>
; AVX2-SLOW-NEXT:    vpshufb %xmm2, %xmm5, %xmm5
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm7[0,1],xmm8[2],xmm7[3]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm11 = [0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
; AVX2-SLOW-NEXT:    vpshufb %xmm11, %xmm5, %xmm5
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm4 = ymm3[0],ymm15[1],ymm3[2],ymm15[3],ymm3[4,5],ymm15[6],ymm3[7,8],ymm15[9],ymm3[10],ymm15[11],ymm3[12,13],ymm15[14],ymm3[15]
; AVX2-SLOW-NEXT:    vmovdqa %ymm3, %ymm12
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5],ymm5[6],ymm4[7]
; AVX2-SLOW-NEXT:    vpshufb %ymm6, %ymm4, %ymm4
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm13[0],ymm14[1,2],ymm13[3],ymm14[4],ymm13[5],ymm14[6,7],ymm13[8],ymm14[9,10],ymm13[11],ymm14[12],ymm13[13],ymm14[14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm5, %xmm6
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2],xmm5[3]
; AVX2-SLOW-NEXT:    vpshufb %xmm2, %xmm5, %xmm5
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa %xmm9, %xmm3
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm10[0,1],xmm9[2],xmm10[3]
; AVX2-SLOW-NEXT:    vpshufb %xmm11, %xmm5, %xmm5
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm9 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT:    vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
; AVX2-SLOW-NEXT:    # ymm4 = ymm1[0,1],mem[2],ymm1[3],mem[4],ymm1[5,6],mem[7],ymm1[8,9],mem[10],ymm1[11],mem[12],ymm1[13,14],mem[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6],ymm5[7]
; AVX2-SLOW-NEXT:    vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX2-SLOW-NEXT:    # ymm5 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5],mem[6],ymm0[7,8],mem[9],ymm0[10,11],mem[12],ymm0[13],mem[14],ymm0[15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm5, %xmm6
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3,4],xmm6[5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15,24,25,18,19,28,29,22,23,u,u,u,u,u,u,u,u>
; AVX2-SLOW-NEXT:    vpshufb %ymm6, %ymm4, %ymm4
; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
; AVX2-SLOW-NEXT:    vpshufb %xmm1, %xmm5, %xmm5
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5],ymm5[6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm7[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm8[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm2 = ymm12[0,1],ymm15[2],ymm12[3],ymm15[4],ymm12[5,6],ymm15[7],ymm12[8,9],ymm15[10],ymm12[11],ymm15[12],ymm12[13,14],ymm15[15]
; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4],ymm2[5,6],ymm4[7]
; AVX2-SLOW-NEXT:    vpshufb %ymm6, %ymm2, %ymm2
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm4 = ymm14[0],ymm13[1],ymm14[2,3],ymm13[4],ymm14[5],ymm13[6],ymm14[7,8],ymm13[9],ymm14[10,11],ymm13[12],ymm14[13],ymm13[14],ymm14[15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm4, %xmm5
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3,4],xmm5[5,6,7]
; AVX2-SLOW-NEXT:    vpshufb %xmm1, %xmm4, %xmm4
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3,4,5],ymm4[6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm10[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm2, (%rsi)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm2, 32(%rsi)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm2, 32(%rdx)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm2, (%rdx)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm2, 32(%rcx)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm2, (%rcx)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm2, 32(%r8)
; AVX2-SLOW-NEXT:    vmovdqa %ymm9, (%r8)
; AVX2-SLOW-NEXT:    vmovdqa %ymm0, 32(%r9)
; AVX2-SLOW-NEXT:    vmovdqa %ymm1, (%r9)
; AVX2-SLOW-NEXT:    addq $280, %rsp # imm = 0x118
; AVX2-SLOW-NEXT:    vzeroupper
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-ALL-LABEL: vf32:
; AVX2-FAST-ALL:       # %bb.0:
; AVX2-FAST-ALL-NEXT:    subq $200, %rsp
; AVX2-FAST-ALL-NEXT:    vmovdqa 224(%rdi), %ymm12
; AVX2-FAST-ALL-NEXT:    vmovdqa 256(%rdi), %ymm2
; AVX2-FAST-ALL-NEXT:    vmovdqa 192(%rdi), %ymm14
; AVX2-FAST-ALL-NEXT:    vmovdqa 160(%rdi), %ymm3
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdi), %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rdi), %ymm4
; AVX2-FAST-ALL-NEXT:    vmovdqa 64(%rdi), %ymm6
; AVX2-FAST-ALL-NEXT:    vmovdqa 96(%rdi), %ymm7
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm8 = ymm6[0],ymm7[1,2],ymm6[3],ymm7[4],ymm6[5],ymm7[6,7],ymm6[8],ymm7[9,10],ymm6[11],ymm7[12],ymm6[13],ymm7[14,15]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm10 = [1,3,0,2,4,6,1,3]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm8, %ymm10, %ymm8
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm11 = [0,1,6,7,8,9,14,15,4,5,14,15,4,5,2,3,16,17,22,23,24,25,30,31,20,21,30,31,20,21,18,19]
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm11, %ymm8, %ymm9
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm8 = ymm0[0],ymm4[1],ymm0[2,3],ymm4[4],ymm0[5],ymm4[6],ymm0[7,8],ymm4[9],ymm0[10,11],ymm4[12],ymm0[13],ymm4[14],ymm0[15]
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm0, %ymm5
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm8, %xmm0
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm0 = xmm8[0],xmm0[1,2,3],xmm8[4,5],xmm0[6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm13 = [0,1,10,11,4,5,14,15,8,9,2,3,12,13,6,7]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm13, %xmm0, %xmm0
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm8, %ymm0, %ymm9, %ymm9
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm0 = ymm3[0],ymm14[1],ymm3[2,3],ymm14[4],ymm3[5],ymm14[6],ymm3[7,8],ymm14[9],ymm3[10,11],ymm14[12],ymm3[13],ymm14[14],ymm3[15]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3],xmm0[4,5],xmm1[6,7]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm13, %xmm0, %xmm0
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm1 = ymm12[0],ymm2[1,2],ymm12[3],ymm2[4],ymm12[5],ymm2[6,7],ymm12[8],ymm2[9,10],ymm12[11],ymm2[12],ymm12[13],ymm2[14,15]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm1, %ymm10, %ymm1
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm11, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm8, %ymm0, %ymm1, %ymm10
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm0 = ymm7[0],ymm6[1],ymm7[2,3],ymm6[4],ymm7[5],ymm6[6],ymm7[7,8],ymm6[9],ymm7[10,11],ymm6[12],ymm7[13],ymm6[14],ymm7[15]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm1 = <2,u,u,u,4,7,1,6>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm1, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm11 = [2,3,4,5,10,11,0,1,14,15,2,3,12,13,0,1,18,19,20,21,26,27,16,17,30,31,18,19,28,29,16,17]
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm11, %ymm0, %ymm0
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm13 = ymm4[0],ymm5[1],ymm4[2],ymm5[3],ymm4[4,5],ymm5[6],ymm4[7,8],ymm5[9],ymm4[10],ymm5[11],ymm4[12,13],ymm5[14],ymm4[15]
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm4, %ymm5
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm13, %xmm1
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3],xmm13[4,5,6],xmm1[7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm13 = [2,3,12,13,6,7,0,1,10,11,4,5,14,15,10,11]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm13, %xmm1, %xmm1
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm8, %ymm1, %ymm0, %ymm0
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm1 = ymm14[0],ymm3[1],ymm14[2],ymm3[3],ymm14[4,5],ymm3[6],ymm14[7,8],ymm3[9],ymm14[10],ymm3[11],ymm14[12,13],ymm3[14],ymm14[15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm1, %xmm15
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,3],xmm1[4,5,6],xmm15[7]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm13, %xmm1, %xmm1
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm13 = ymm2[0],ymm12[1],ymm2[2,3],ymm12[4],ymm2[5],ymm12[6],ymm2[7,8],ymm12[9],ymm2[10,11],ymm12[12],ymm2[13],ymm12[14],ymm2[15]
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm12, %ymm4
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm2, %ymm12
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm2 = <2,u,u,u,4,7,1,6>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm13, %ymm2, %ymm15
; AVX2-FAST-ALL-NEXT:    vmovdqa 128(%rdi), %ymm13
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm11, %ymm15, %ymm11
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,3,1,3,0,3,5,7]
; AVX2-FAST-ALL-NEXT:    vpblendvb %ymm8, %ymm1, %ymm11, %ymm1
; AVX2-FAST-ALL-NEXT:    vpermd %ymm13, %ymm2, %ymm8
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm15, %ymm8, %ymm8
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm8 = ymm9[0,1,2,3,4],ymm8[5,6,7],ymm9[8,9,10,11,12],ymm8[13,14,15]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm9[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 288(%rdi), %ymm11
; AVX2-FAST-ALL-NEXT:    vpermd %ymm11, %ymm2, %ymm2
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm15, %ymm2, %ymm2
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm2 = ymm10[0,1,2,3,4],ymm2[5,6,7],ymm10[8,9,10,11,12],ymm2[13,14,15]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm10[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm2 = [1,3,2,3,1,3,6,7]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm13, %ymm2, %ymm9
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm3, %ymm9, %ymm9
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm9 = ymm0[0,1,2,3,4],ymm9[5,6,7],ymm0[8,9,10,11,12],ymm9[13,14,15]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpermd %ymm11, %ymm2, %ymm0
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7],ymm1[8,9,10,11,12],ymm0[13,14,15]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm1 = ymm5[0,1],ymm8[2],ymm5[3],ymm8[4],ymm5[5,6],ymm8[7],ymm5[8,9],ymm8[10],ymm5[11],ymm8[12],ymm5[13,14],ymm8[15]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3,4],xmm1[5,6,7]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm2 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5],ymm7[6],ymm6[7,8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13],ymm7[14],ymm6[15]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm9 = <0,2,u,u,5,7,2,4>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm2, %ymm9, %ymm2
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,20,21,22,23,16,17,22,23]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm0 = <4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u>
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm0, %xmm1, %xmm1
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [1,4,6,0,1,4,6,0]
; AVX2-FAST-ALL-NEXT:    # ymm2 = mem[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm13, %ymm2, %ymm10
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm15, %ymm10, %ymm10
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm10 = ymm1[0,1,2,3,4],ymm10[5,6,7],ymm1[8,9,10,11,12],ymm10[13,14,15]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm10 = ymm14[0,1],ymm3[2],ymm14[3],ymm3[4],ymm14[5,6],ymm3[7],ymm14[8,9],ymm3[10],ymm14[11],ymm3[12],ymm14[13,14],ymm3[15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm10, %xmm1
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm1 = xmm10[0,1,2],xmm1[3,4],xmm10[5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm0, %xmm1, %xmm0
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm1 = ymm4[0],ymm12[1],ymm4[2,3],ymm12[4],ymm4[5],ymm12[6],ymm4[7,8],ymm12[9],ymm4[10,11],ymm12[12],ymm4[13],ymm12[14],ymm4[15]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm1, %ymm9, %ymm1
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,20,21,22,23,16,17,22,23]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm11, %ymm2, %ymm1
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm15, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm1 = ymm0[0,1,2,3,4],ymm1[5,6,7],ymm0[8,9,10,11,12],ymm1[13,14,15]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, (%rsp) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm0 = ymm8[0],ymm5[1,2],ymm8[3],ymm5[4],ymm8[5],ymm5[6,7],ymm8[8],ymm5[9,10],ymm8[11],ymm5[12],ymm8[13],ymm5[14,15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2],xmm0[3]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm1 = ymm7[0],ymm6[1],ymm7[2],ymm6[3],ymm7[4,5],ymm6[6],ymm7[7,8],ymm6[9],ymm7[10],ymm6[11],ymm7[12,13],ymm6[14],ymm7[15]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm10 = <0,3,u,u,5,0,2,7>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm1, %ymm10, %ymm1
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm8 = <u,u,u,u,u,u,u,u,u,u,u,u,2,3,4,5,18,19,20,21,26,27,16,17,30,31,30,31,18,19,20,21>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm8, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm2 = <6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u>
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [2,4,7,0,2,4,7,0]
; AVX2-FAST-ALL-NEXT:    # ymm1 = mem[0,1,0,1]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm13, %ymm1, %ymm9
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm5 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm5, %ymm9, %ymm9
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm9 = ymm0[0,1,2,3,4],ymm9[5,6,7],ymm0[8,9,10,11,12],ymm9[13,14,15]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm0 = ymm3[0],ymm14[1,2],ymm3[3],ymm14[4],ymm3[5],ymm14[6,7],ymm3[8],ymm14[9,10],ymm3[11],ymm14[12],ymm3[13],ymm14[14,15]
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm3, %ymm9
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm0, %xmm15
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm15[0],xmm0[1],xmm15[2],xmm0[3]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm4, %ymm15
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm2 = ymm12[0],ymm4[1],ymm12[2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7,8],ymm4[9],ymm12[10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm2, %ymm10, %ymm2
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm8, %ymm2, %ymm2
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm11, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm5, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm1 = ymm0[0,1,2,3,4],ymm1[5,6,7],ymm0[8,9,10,11,12],ymm1[13,14,15]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm10 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm0 = ymm7[0,1],ymm6[2],ymm7[3],ymm6[4],ymm7[5,6],ymm6[7],ymm7[8,9],ymm6[10],ymm7[11],ymm6[12],ymm7[13,14],ymm6[15]
; AVX2-FAST-ALL-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # ymm1 = mem[0],ymm1[1],mem[2,3],ymm1[4],mem[5],ymm1[6],mem[7,8],ymm1[9],mem[10,11],ymm1[12],mem[13],ymm1[14],mem[15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4],xmm2[5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm2 = <1,3,u,u,6,0,3,5>
; AVX2-FAST-ALL-NEXT:    vpermd %ymm0, %ymm2, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,u,u,u,u,u,u,u,u>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm4 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm4, %xmm1, %xmm1
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,1,3,0,2,5,7]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm13, %ymm1, %ymm5
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25,30,31>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm5 = ymm14[0],ymm9[1],ymm14[2,3],ymm9[4],ymm14[5],ymm9[6],ymm14[7,8],ymm9[9],ymm14[10,11],ymm9[12],ymm14[13],ymm9[14],ymm14[15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm5, %xmm7
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} xmm5 = xmm7[0,1,2],xmm5[3,4],xmm7[5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm4, %xmm5, %xmm4
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm5 = ymm12[0,1],ymm15[2],ymm12[3],ymm15[4],ymm12[5,6],ymm15[7],ymm12[8,9],ymm15[10],ymm12[11],ymm15[12],ymm12[13,14],ymm15[15]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm5, %ymm2, %ymm2
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3,4,5],ymm4[6,7]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm11, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm2, 32(%rsi)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm2, (%rsi)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm2, 32(%rdx)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm2, (%rdx)
; AVX2-FAST-ALL-NEXT:    vmovups (%rsp), %ymm2 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm2, 32(%rcx)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm2, (%rcx)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm10, 32(%r8)
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm2, (%r8)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm1, 32(%r9)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm0, (%r9)
; AVX2-FAST-ALL-NEXT:    addq $200, %rsp
; 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 $296, %rsp # imm = 0x128
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdi), %ymm15
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rdi), %ymm8
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 64(%rdi), %ymm7
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 96(%rdi), %ymm9
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 192(%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 160(%rdi), %ymm4
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 224(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 256(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0],ymm2[1,2],ymm1[3],ymm2[4],ymm1[5],ymm2[6,7],ymm1[8],ymm2[9,10],ymm1[11],ymm2[12],ymm1[13],ymm2[14,15]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm2, %ymm5
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm1, %ymm10
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,10,11,4,5,14,15,8,9,10,11,4,5,6,7,16,17,26,27,20,21,30,31,24,25,26,27,20,21,22,23]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm2, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm1 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5],ymm3[6],ymm4[7,8],ymm3[9],ymm4[10,11],ymm3[12],ymm4[13],ymm3[14],ymm4[15]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm4, %ymm11
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm3, %ymm4
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm1, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm3[1,2,3],xmm1[4,5],xmm3[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,10,11,4,5,14,15,8,9,2,3,12,13,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm3, %xmm1, %xmm12
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm14 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm14, %ymm12, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm12 = ymm7[0],ymm9[1,2],ymm7[3],ymm9[4],ymm7[5],ymm9[6,7],ymm7[8],ymm9[9,10],ymm7[11],ymm9[12],ymm7[13],ymm9[14,15]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm13 = ymm12[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm13[5],ymm12[6],ymm13[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm2, %ymm12, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm12 = ymm15[0],ymm8[1],ymm15[2,3],ymm8[4],ymm15[5],ymm8[6],ymm15[7,8],ymm8[9],ymm15[10,11],ymm8[12],ymm15[13],ymm8[14],ymm15[15]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm8, (%rsp) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm12, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm1 = xmm12[0],xmm1[1,2,3],xmm12[4,5],xmm1[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm14, %ymm1, %ymm2, %ymm6
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm1 = ymm5[0],ymm10[1],ymm5[2,3],ymm10[4],ymm5[5],ymm10[6],ymm5[7,8],ymm10[9],ymm5[10,11],ymm10[12],ymm5[13],ymm10[14],ymm5[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,3,12,13,6,7,0,1,10,11,6,7,8,9,8,9,18,19,28,29,22,23,16,17,26,27,22,23,24,25,24,25]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm3 = ymm4[0],ymm11[1],ymm4[2],ymm11[3],ymm4[4,5],ymm11[6],ymm4[7,8],ymm11[9],ymm4[10],ymm11[11],ymm4[12,13],ymm11[14],ymm4[15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6],xmm4[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm4 = [2,3,12,13,6,7,0,1,10,11,4,5,14,15,10,11]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm4, %xmm3, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm14, %ymm3, %ymm1, %ymm13
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm1 = ymm9[0],ymm7[1],ymm9[2,3],ymm7[4],ymm9[5],ymm7[6],ymm9[7,8],ymm7[9],ymm9[10,11],ymm7[12],ymm9[13],ymm7[14],ymm9[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm2 = ymm8[0],ymm15[1],ymm8[2],ymm15[3],ymm8[4,5],ymm15[6],ymm8[7,8],ymm15[9],ymm8[10],ymm15[11],ymm8[12,13],ymm15[14],ymm8[15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm3 = xmm2[0,1],xmm3[2,3],xmm2[4,5,6],xmm3[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 304(%rdi), %xmm10
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 288(%rdi), %xmm12
; AVX2-FAST-PERLANE-NEXT:    vpblendvb %ymm14, %ymm4, %ymm1, %ymm4
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm1 = xmm12[0],xmm10[1],xmm12[2,3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm14 = [6,7,2,3,4,5,6,7,6,7,2,3,12,13,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm14, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm1 = ymm0[0,1,2,3,4],ymm1[5,6,7],ymm0[8,9,10,11,12],ymm1[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 144(%rdi), %xmm8
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 128(%rdi), %xmm9
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm11 = xmm9[0],xmm8[1],xmm9[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm14, %xmm11, %xmm5
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6,7],ymm6[8,9,10,11,12],ymm5[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm5 = xmm12[0,1],xmm10[2],xmm12[3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm11 = [0,1,2,3,4,5,6,7,8,9,4,5,14,15,8,9]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm11, %xmm5, %xmm5
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm5 = ymm13[0,1,2,3,4],ymm5[5,6,7],ymm13[8,9,10,11,12],ymm5[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm5 = xmm9[0,1],xmm8[2],xmm9[3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm11, %xmm5, %xmm5
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm4 = ymm13[0],ymm0[1],ymm13[2,3],ymm0[4],ymm13[5],ymm0[6],ymm13[7,8],ymm0[9],ymm13[10,11],ymm0[12],ymm13[13],ymm0[14],ymm13[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6],ymm4[7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm5 = ymm1[0,1],ymm14[2],ymm1[3],ymm14[4],ymm1[5,6],ymm14[7],ymm1[8,9],ymm14[10],ymm1[11],ymm14[12],ymm1[13,14],ymm14[15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm5, %xmm6
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm6[3,4],xmm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,0,1,10,11,20,21,30,31,24,25,18,19,28,29,26,27,16,17,26,27>
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm6, %ymm4, %ymm4
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm11 = <4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm11, %xmm5, %xmm5
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm5 = xmm10[0],xmm12[1],xmm10[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm7, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm4 = ymm7[0],ymm3[1],ymm7[2,3],ymm3[4],ymm7[5],ymm3[6],ymm7[7,8],ymm3[9],ymm7[10,11],ymm3[12],ymm7[13],ymm3[14],ymm7[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6],ymm4[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm6, %ymm4, %ymm4
; AVX2-FAST-PERLANE-NEXT:    vmovdqu (%rsp), %ymm7 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm5 = ymm7[0,1],ymm15[2],ymm7[3],ymm15[4],ymm7[5,6],ymm15[7],ymm7[8,9],ymm15[10],ymm7[11],ymm15[12],ymm7[13,14],ymm15[15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm5, %xmm6
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm6[3,4],xmm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm11, %xmm5, %xmm5
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm5 = xmm8[0],xmm9[1],xmm8[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb {{.*#+}} xmm5 = xmm5[0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm5 = ymm0[0],ymm13[1],ymm0[2],ymm13[3],ymm0[4,5],ymm13[6],ymm0[7,8],ymm13[9],ymm0[10],ymm13[11],ymm0[12,13],ymm13[14],ymm0[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm6 = ymm5[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4],ymm5[5],ymm6[6],ymm5[7]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm6 = ymm14[0],ymm1[1,2],ymm14[3],ymm1[4],ymm14[5],ymm1[6,7],ymm14[8],ymm1[9,10],ymm14[11],ymm1[12],ymm14[13],ymm1[14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm6, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0],xmm6[1],xmm4[2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,2,3,12,13,22,23,16,17,26,27,20,21,30,31,30,31,18,19,28,29>
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm0 = <6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm0, %xmm4, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm5 = xmm10[0,1],xmm12[2],xmm10[3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm13 = [0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm13, %xmm5, %xmm5
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm5 = ymm3[0],ymm2[1],ymm3[2],ymm2[3],ymm3[4,5],ymm2[6],ymm3[7,8],ymm2[9],ymm3[10],ymm2[11],ymm3[12,13],ymm2[14],ymm3[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm14 = ymm5[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm14[4],ymm5[5],ymm14[6],ymm5[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm15, %ymm14
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm6 = ymm15[0],ymm7[1,2],ymm15[3],ymm7[4],ymm15[5],ymm7[6,7],ymm15[8],ymm7[9,10],ymm15[11],ymm7[12],ymm15[13],ymm7[14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm6, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0],xmm6[1],xmm4[2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm0, %xmm4, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm5 = xmm8[0,1],xmm9[2],xmm8[3]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm13, %xmm5, %xmm5
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm15 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm13 = [12,13,14,15,4,5,14,15,8,9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm13, %xmm8, %xmm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm5 = [0,1,2,3,0,1,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm5, %xmm9, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm1 = ymm3[0,1],ymm2[2],ymm3[3],ymm2[4],ymm3[5,6],ymm2[7],ymm3[8,9],ymm2[10],ymm3[11],ymm2[12],ymm3[13,14],ymm2[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm6 = ymm1[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4],ymm1[5,6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm6 = ymm7[0],ymm14[1],ymm7[2,3],ymm14[4],ymm7[5],ymm14[6],ymm7[7,8],ymm14[9],ymm7[10,11],ymm14[12],ymm7[13],ymm14[14],ymm7[15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm6, %xmm7
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm6 = xmm7[0,1,2],xmm6[3,4],xmm7[5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15,24,25,18,19,28,29,22,23,u,u,u,u,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm7, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm4 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm4, %xmm6, %xmm6
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm6[0,1,2],ymm1[3,4,5],ymm6[6,7]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpblendw $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # ymm1 = mem[0,1],ymm1[2],mem[3],ymm1[4],mem[5,6],ymm1[7],mem[8,9],ymm1[10],mem[11],ymm1[12],mem[13,14],ymm1[15]
; AVX2-FAST-PERLANE-NEXT:    vpermq {{.*#+}} ymm6 = ymm1[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4],ymm1[5,6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %ymm7, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # ymm6 = mem[0],ymm2[1],mem[2,3],ymm2[4],mem[5],ymm2[6],mem[7,8],ymm2[9],mem[10,11],ymm2[12],mem[13],ymm2[14],mem[15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm6, %xmm7
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} xmm6 = xmm7[0,1,2],xmm6[3,4],xmm7[5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm4, %xmm6, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3,4,5],ymm4[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm13, %xmm10, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm5, %xmm12, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, (%rsi)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, 32(%rsi)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, 32(%rdx)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, (%rdx)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, 32(%rcx)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, (%rcx)
; AVX2-FAST-PERLANE-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT:    vmovaps %ymm2, 32(%r8)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm15, (%r8)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm1, 32(%r9)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm0, (%r9)
; AVX2-FAST-PERLANE-NEXT:    addq $296, %rsp # imm = 0x128
; AVX2-FAST-PERLANE-NEXT:    vzeroupper
; AVX2-FAST-PERLANE-NEXT:    retq
;
; AVX512-LABEL: vf32:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqu64 256(%rdi), %zmm0
; AVX512-NEXT:    vmovdqu64 (%rdi), %zmm1
; AVX512-NEXT:    vmovdqu64 64(%rdi), %zmm2
; AVX512-NEXT:    vmovdqu64 128(%rdi), %zmm3
; AVX512-NEXT:    vmovdqu64 192(%rdi), %zmm4
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [16,21,26,31,36,41,46,51,56,61,0,0,0,1,6,11,16,21,26,31,36,41,46,51,56,61,0,0,0,1,6,11]
; AVX512-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm4, %zmm3, %zmm5
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm6 = <0,5,10,15,20,25,30,35,40,45,50,55,60,u,u,u>
; AVX512-NEXT:    vpermi2w %zmm2, %zmm1, %zmm6
; AVX512-NEXT:    movl $67100672, %eax # imm = 0x3FFE000
; AVX512-NEXT:    kmovd %eax, %k1
; AVX512-NEXT:    vmovdqu16 %zmm5, %zmm6 {%k1}
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm5 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,34,39,44,49,54,59]
; AVX512-NEXT:    vpermi2w %zmm0, %zmm6, %zmm5
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [49,54,59,0,5,10,15,20,25,30,0,0,0,34,39,44,49,54,59,0,5,10,15,20,25,30,0,0,0,34,39,44]
; AVX512-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm3, %zmm4, %zmm6
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm7 = <1,6,11,16,21,26,31,36,41,46,51,56,61,u,u,u>
; AVX512-NEXT:    vpermi2w %zmm2, %zmm1, %zmm7
; AVX512-NEXT:    vmovdqu16 %zmm6, %zmm7 {%k1}
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm6 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,35,40,45,50,55,60]
; AVX512-NEXT:    vpermi2w %zmm0, %zmm7, %zmm6
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [50,55,60,1,6,11,16,21,26,31,0,0,0,35,40,45,50,55,60,1,6,11,16,21,26,31,0,0,0,35,40,45]
; AVX512-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm3, %zmm4, %zmm7
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm8 = <34,39,44,49,54,59,0,5,10,15,20,25,30,u,u,u>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm2, %zmm8
; AVX512-NEXT:    vmovdqu16 %zmm7, %zmm8 {%k1}
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm7 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,36,41,46,51,56,61]
; AVX512-NEXT:    vpermi2w %zmm0, %zmm8, %zmm7
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [19,24,29,34,39,44,49,54,59,0,0,0,0,4,9,14,19,24,29,34,39,44,49,54,59,0,0,0,0,4,9,14]
; AVX512-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm4, %zmm3, %zmm8
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm9 = <35,40,45,50,55,60,1,6,11,16,21,26,31,u,u,u>
; AVX512-NEXT:    vpermi2w %zmm1, %zmm2, %zmm9
; AVX512-NEXT:    movl $33546240, %eax # imm = 0x1FFE000
; AVX512-NEXT:    kmovd %eax, %k1
; AVX512-NEXT:    vmovdqu16 %zmm8, %zmm9 {%k1}
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm8 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,32,37,42,47,52,57,62]
; AVX512-NEXT:    vpermi2w %zmm0, %zmm9, %zmm8
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm9 = <4,9,14,19,24,29,34,39,44,49,54,59,u,u,u,u>
; AVX512-NEXT:    vpermi2w %zmm2, %zmm1, %zmm9
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm1 = [20,25,30,35,40,45,50,55,60,0,0,0,0,5,10,15,20,25,30,35,40,45,50,55,60,0,0,0,0,5,10,15]
; AVX512-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2w %zmm4, %zmm3, %zmm1
; AVX512-NEXT:    movb $7, %al
; AVX512-NEXT:    kmovd %eax, %k1
; AVX512-NEXT:    vmovdqa64 %zmm9, %zmm1 {%k1}
; AVX512-NEXT:    vmovdqa64 {{.*#+}} zmm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,33,38,43,48,53,58,63]
; AVX512-NEXT:    vpermi2w %zmm0, %zmm1, %zmm2
; AVX512-NEXT:    vmovdqu64 %zmm5, (%rsi)
; AVX512-NEXT:    vmovdqu64 %zmm6, (%rdx)
; AVX512-NEXT:    vmovdqu64 %zmm7, (%rcx)
; AVX512-NEXT:    vmovdqu64 %zmm8, (%r8)
; AVX512-NEXT:    vmovdqu64 %zmm2, (%r9)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %wide.vec = load <160 x i16>, <160 x i16>* %in.vec, align 32

  %strided.vec0 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75, i32 80, i32 85, i32 90, i32 95, i32 100, i32 105, i32 110, i32 115, i32 120, i32 125, i32 130, i32 135, i32 140, i32 145, i32 150, i32 155>
  %strided.vec1 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76, i32 81, i32 86, i32 91, i32 96, i32 101, i32 106, i32 111, i32 116, i32 121, i32 126, i32 131, i32 136, i32 141, i32 146, i32 151, i32 156>
  %strided.vec2 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77, i32 82, i32 87, i32 92, i32 97, i32 102, i32 107, i32 112, i32 117, i32 122, i32 127, i32 132, i32 137, i32 142, i32 147, i32 152, i32 157>
  %strided.vec3 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78, i32 83, i32 88, i32 93, i32 98, i32 103, i32 108, i32 113, i32 118, i32 123, i32 128, i32 133, i32 138, i32 143, i32 148, i32 153, i32 158>
  %strided.vec4 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79, i32 84, i32 89, i32 94, i32 99, i32 104, i32 109, i32 114, i32 119, i32 124, i32 129, i32 134, i32 139, i32 144, i32 149, i32 154, i32 159>

  store <32 x i16> %strided.vec0, <32 x i16>* %out.vec0, align 32
  store <32 x i16> %strided.vec1, <32 x i16>* %out.vec1, align 32
  store <32 x i16> %strided.vec2, <32 x i16>* %out.vec2, align 32
  store <32 x i16> %strided.vec3, <32 x i16>* %out.vec3, align 32
  store <32 x i16> %strided.vec4, <32 x i16>* %out.vec4, align 32

  ret void
}