; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse2 | FileCheck %s --check-prefixes=SSE
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx  | FileCheck %s --check-prefixes=AVX,AVX1
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2-SLOW
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX2-FAST,AVX2-FAST-ALL
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,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(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: vf2:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,2,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
; SSE-NEXT:    movd %xmm2, (%rsi)
; SSE-NEXT:    movd %xmm1, (%rdx)
; SSE-NEXT:    movd %xmm3, (%rcx)
; SSE-NEXT:    movd %xmm0, (%r8)
; SSE-NEXT:    retq
;
; AVX-LABEL: vf2:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovdqa (%rdi), %xmm0
; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
; AVX-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm1[0,2,2,3,4,5,6,7]
; AVX-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; AVX-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm0[2,0,2,3,4,5,6,7]
; AVX-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
; AVX-NEXT:    vmovd %xmm2, (%rsi)
; AVX-NEXT:    vmovd %xmm1, (%rdx)
; AVX-NEXT:    vmovd %xmm3, (%rcx)
; AVX-NEXT:    vmovd %xmm0, (%r8)
; AVX-NEXT:    retq
;
; AVX512-LABEL: vf2:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-NEXT:    vpmovqw %xmm0, %xmm1
; AVX512-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
; AVX512-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; AVX512-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; AVX512-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm0[2,0,2,3,4,5,6,7]
; AVX512-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
; AVX512-NEXT:    vmovd %xmm1, (%rsi)
; AVX512-NEXT:    vmovd %xmm2, (%rdx)
; AVX512-NEXT:    vmovd %xmm3, (%rcx)
; AVX512-NEXT:    vmovd %xmm0, (%r8)
; AVX512-NEXT:    retq
  %wide.vec = load <8 x i16>, ptr %in.vec, align 32

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

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

  ret void
}

define void @vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: vf4:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    movdqa 16(%rdi), %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm2[0,2,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm4[0,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm4[1,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    movq %xmm5, (%rsi)
; SSE-NEXT:    movq %xmm3, (%rdx)
; SSE-NEXT:    movq %xmm4, (%rcx)
; SSE-NEXT:    movq %xmm0, (%r8)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf4:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vpxor %xmm0, %xmm0, %xmm0
; AVX1-NEXT:    vmovdqa (%rdi), %xmm1
; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm2[0],xmm0[1,2,3],xmm2[4],xmm0[5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3],xmm1[4],xmm0[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
; AVX1-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm2[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm1[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX1-NEXT:    vmovq %xmm0, (%rsi)
; AVX1-NEXT:    vmovq %xmm3, (%rdx)
; AVX1-NEXT:    vmovq %xmm4, (%rcx)
; AVX1-NEXT:    vmovq %xmm1, (%r8)
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf4:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    vpxor %xmm0, %xmm0, %xmm0
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm3 = xmm2[0],xmm0[1,2,3],xmm2[4],xmm0[5,6,7]
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3],xmm1[4],xmm0[5,6,7]
; AVX2-SLOW-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-SLOW-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX2-SLOW-NEXT:    vmovq %xmm0, (%rsi)
; AVX2-SLOW-NEXT:    vmovq %xmm3, (%rdx)
; AVX2-SLOW-NEXT:    vmovq %xmm4, (%rcx)
; AVX2-SLOW-NEXT:    vmovq %xmm1, (%r8)
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-LABEL: vf4:
; AVX2-FAST:       # %bb.0:
; AVX2-FAST-NEXT:    vpxor %xmm0, %xmm0, %xmm0
; AVX2-FAST-NEXT:    vmovdqa (%rdi), %xmm1
; AVX2-FAST-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm3 = xmm2[0],xmm0[1,2,3],xmm2[4],xmm0[5,6,7]
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3],xmm1[4],xmm0[5,6,7]
; AVX2-FAST-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-FAST-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} xmm3 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-NEXT:    vpshufb %xmm3, %xmm2, %xmm4
; AVX2-FAST-NEXT:    vpshufb %xmm3, %xmm1, %xmm3
; AVX2-FAST-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-FAST-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX2-FAST-NEXT:    vmovq %xmm0, (%rsi)
; AVX2-FAST-NEXT:    vmovq %xmm3, (%rdx)
; AVX2-FAST-NEXT:    vmovq %xmm4, (%rcx)
; AVX2-FAST-NEXT:    vmovq %xmm1, (%r8)
; AVX2-FAST-NEXT:    retq
;
; AVX512-LABEL: vf4:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa (%rdi), %ymm0
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm1 = <1,5,9,13,u,u,u,u>
; AVX512-NEXT:    vmovdqa (%rdi), %xmm2
; AVX512-NEXT:    vmovdqa 16(%rdi), %xmm3
; AVX512-NEXT:    vpermi2w %xmm3, %xmm2, %xmm1
; AVX512-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX512-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm3[2,0,2,3,4,5,6,7]
; AVX512-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX512-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm2[2,0,2,3,4,5,6,7]
; AVX512-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX512-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX512-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX512-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX512-NEXT:    vpmovqw %ymm0, (%rsi)
; AVX512-NEXT:    vmovq %xmm1, (%rdx)
; AVX512-NEXT:    vmovq %xmm4, (%rcx)
; AVX512-NEXT:    vmovq %xmm2, (%r8)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %wide.vec = load <16 x i16>, ptr %in.vec, align 32

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

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

  ret void
}

define void @vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: vf8:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa (%rdi), %xmm10
; SSE-NEXT:    movdqa 16(%rdi), %xmm3
; SSE-NEXT:    movdqa 32(%rdi), %xmm8
; SSE-NEXT:    movdqa 48(%rdi), %xmm9
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[0,2,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm10[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm6[0,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm1[0,1,0,2,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,1,0,2,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm5[1,3,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm6[1,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm5[0],xmm2[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm10[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm4[2,0,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm3[0,1,2,0,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm8[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm7[0,1,2,0,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1]
; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm5[0],xmm6[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[0,1,3,1,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm7[0,1,3,1,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm4[0],xmm3[1]
; SSE-NEXT:    movapd %xmm0, (%rsi)
; SSE-NEXT:    movapd %xmm2, (%rdx)
; SSE-NEXT:    movapd %xmm6, (%rcx)
; SSE-NEXT:    movapd %xmm3, (%r8)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf8:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vpxor %xmm0, %xmm0, %xmm0
; AVX1-NEXT:    vmovdqa (%rdi), %xmm1
; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX1-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX1-NEXT:    vmovdqa 48(%rdi), %xmm4
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm4[0],xmm0[1,2,3],xmm4[4],xmm0[5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm6 = xmm3[0],xmm0[1,2,3],xmm3[4],xmm0[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm5, %xmm6, %xmm5
; AVX1-NEXT:    vpblendw {{.*#+}} xmm6 = xmm2[0],xmm0[1,2,3],xmm2[4],xmm0[5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3],xmm1[4],xmm0[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm6, %xmm0, %xmm0
; AVX1-NEXT:    vpackusdw %xmm5, %xmm0, %xmm8
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm4[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm5[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm3[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm2[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm1[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm5 = xmm6[0,1,2,3],xmm5[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm4[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm3[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm2[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm1[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm6[4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm3[4,5,6,7]
; AVX1-NEXT:    vmovdqa %xmm8, (%rsi)
; AVX1-NEXT:    vmovdqa %xmm5, (%rdx)
; AVX1-NEXT:    vmovdqa %xmm0, (%rcx)
; AVX1-NEXT:    vmovdqa %xmm1, (%r8)
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf8:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    vpxor %xmm0, %xmm0, %xmm0
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm2
; AVX2-SLOW-NEXT:    vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm2
; AVX2-SLOW-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
; AVX2-SLOW-NEXT:    vpackusdw %xmm1, %xmm0, %xmm8
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX2-SLOW-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX2-SLOW-NEXT:    vmovdqa 48(%rdi), %xmm4
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm5 = xmm4[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm5[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm6 = xmm3[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm6 = xmm2[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm7 = xmm1[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm4[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm3[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-SLOW-NEXT:    vmovdqa %xmm8, (%rsi)
; AVX2-SLOW-NEXT:    vmovdqa %xmm5, (%rdx)
; AVX2-SLOW-NEXT:    vmovdqa %xmm0, (%rcx)
; AVX2-SLOW-NEXT:    vmovdqa %xmm1, (%r8)
; AVX2-SLOW-NEXT:    vzeroupper
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-LABEL: vf8:
; AVX2-FAST:       # %bb.0:
; AVX2-FAST-NEXT:    vpxor %xmm0, %xmm0, %xmm0
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-NEXT:    vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-NEXT:    vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-NEXT:    vextracti128 $1, %ymm0, %xmm2
; AVX2-FAST-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
; AVX2-FAST-NEXT:    vpackusdw %xmm1, %xmm0, %xmm8
; AVX2-FAST-NEXT:    vmovdqa (%rdi), %xmm1
; AVX2-FAST-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX2-FAST-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX2-FAST-NEXT:    vmovdqa 48(%rdi), %xmm4
; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} xmm5 = [0,1,2,3,2,3,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-NEXT:    vpshufb %xmm5, %xmm4, %xmm6
; AVX2-FAST-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
; AVX2-FAST-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-NEXT:    vpshufb %xmm6, %xmm2, %xmm7
; AVX2-FAST-NEXT:    vpshufb %xmm6, %xmm1, %xmm6
; AVX2-FAST-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm4[0,1,2,0,4,5,6,7]
; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm3[0,1,2,0,4,5,6,7]
; AVX2-FAST-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-FAST-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3]
; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
; AVX2-FAST-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-FAST-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-FAST-NEXT:    vmovdqa %xmm8, (%rsi)
; AVX2-FAST-NEXT:    vmovdqa %xmm5, (%rdx)
; AVX2-FAST-NEXT:    vmovdqa %xmm0, (%rcx)
; AVX2-FAST-NEXT:    vmovdqa %xmm1, (%r8)
; AVX2-FAST-NEXT:    vzeroupper
; AVX2-FAST-NEXT:    retq
;
; AVX512-LABEL: vf8:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqu64 (%rdi), %zmm0
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm1 = [1,5,9,13,17,21,25,29]
; AVX512-NEXT:    vmovdqa (%rdi), %ymm2
; AVX512-NEXT:    vmovdqa 32(%rdi), %ymm3
; AVX512-NEXT:    vpermi2w %ymm3, %ymm2, %ymm1
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm4 = [2,6,10,14,18,22,26,30]
; AVX512-NEXT:    vpermi2w %ymm3, %ymm2, %ymm4
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm5 = [3,7,11,15,19,23,27,31]
; AVX512-NEXT:    vpermi2w %ymm3, %ymm2, %ymm5
; AVX512-NEXT:    vpmovqw %zmm0, (%rsi)
; AVX512-NEXT:    vmovdqa %xmm1, (%rdx)
; AVX512-NEXT:    vmovdqa %xmm4, (%rcx)
; AVX512-NEXT:    vmovdqa %xmm5, (%r8)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %wide.vec = load <32 x i16>, ptr %in.vec, align 32

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

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

  ret void
}

define void @vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: vf16:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa 96(%rdi), %xmm7
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 64(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 80(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa (%rdi), %xmm13
; SSE-NEXT:    movdqa 16(%rdi), %xmm14
; SSE-NEXT:    movdqa 32(%rdi), %xmm4
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 48(%rdi), %xmm12
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm14[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[0,2,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm13[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm9[0,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm12[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm8[0,1,0,2,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm5[0,1,0,2,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm6[0],xmm11[1],xmm6[1]
; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm2[0],xmm11[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm4[0,2,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[0,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE-NEXT:    movdqa 112(%rdi), %xmm15
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm15[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[0,1,0,2,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm2[0,1,0,2,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm0[0],xmm7[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[1,3,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm9[1,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[0,1,1,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,1,1,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[1,3,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm6[1,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[0,1,1,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[0,1,1,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm4[0],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm14[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm8[2,0,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm13[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm9[2,0,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm12[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[0,1,2,0,4,5,6,7]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; SSE-NEXT:    # xmm12 = mem[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm12[0,1,2,0,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm6[0],xmm3[1]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
; SSE-NEXT:    # xmm13 = mem[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm13[2,0,2,3,4,5,6,7]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; SSE-NEXT:    # xmm14 = mem[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm14[2,0,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm15[0,1,2,0,4,5,6,7]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm2[0,1,2,0,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[3,1,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm9[3,1,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm10[0,1,3,1,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm12[0,1,3,1,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm13[3,1,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm14[3,1,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm15[0,1,3,1,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm4[0],xmm2[1]
; SSE-NEXT:    movapd %xmm7, 16(%rsi)
; SSE-NEXT:    movapd %xmm11, (%rsi)
; SSE-NEXT:    movapd %xmm1, 16(%rdx)
; SSE-NEXT:    movapd %xmm5, (%rdx)
; SSE-NEXT:    movapd %xmm6, 16(%rcx)
; SSE-NEXT:    movapd %xmm3, (%rcx)
; SSE-NEXT:    movapd %xmm2, 16(%r8)
; SSE-NEXT:    movapd %xmm0, (%r8)
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf16:
; AVX1:       # %bb.0:
; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
; AVX1-NEXT:    vmovdqa 112(%rdi), %xmm5
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm5[0],xmm2[1,2,3],xmm5[4],xmm2[5,6,7]
; AVX1-NEXT:    vmovdqa 96(%rdi), %xmm6
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm6[0],xmm2[1,2,3],xmm6[4],xmm2[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm0, %xmm1, %xmm1
; AVX1-NEXT:    vmovdqa 80(%rdi), %xmm7
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm7[0],xmm2[1,2,3],xmm7[4],xmm2[5,6,7]
; AVX1-NEXT:    vmovdqa 64(%rdi), %xmm0
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm0[0],xmm2[1,2,3],xmm0[4],xmm2[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm3, %xmm4, %xmm3
; AVX1-NEXT:    vpackusdw %xmm1, %xmm3, %xmm1
; AVX1-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vmovdqa (%rdi), %xmm11
; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm12
; AVX1-NEXT:    vmovdqa 32(%rdi), %xmm13
; AVX1-NEXT:    vmovdqa 48(%rdi), %xmm1
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm1[0],xmm2[1,2,3],xmm1[4],xmm2[5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm13[0],xmm2[1,2,3],xmm13[4],xmm2[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm3, %xmm4, %xmm3
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm12[0],xmm2[1,2,3],xmm12[4],xmm2[5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm11[0],xmm2[1,2,3],xmm11[4],xmm2[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm4, %xmm2, %xmm2
; AVX1-NEXT:    vpackusdw %xmm3, %xmm2, %xmm9
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm5[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm6[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm7[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm10
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm13[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm12[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm11[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4,5,6,7]
; AVX1-NEXT:    vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm10[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm15 = xmm5[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm15[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm8 = xmm6[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm8[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm7[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm5[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm0[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm6[0,1,2,3],xmm3[4,5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm14
; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm1[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm13[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm7[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm13 = xmm3[0],xmm6[0],xmm3[1],xmm6[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm12[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm6[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm11[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm3[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm13[4,5,6,7]
; AVX1-NEXT:    vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm14[4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm15[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm8[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm5[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm2[4,5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm7[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm6[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; 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:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-NEXT:    vmovdqa %xmm9, (%rsi)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm1, 16(%rsi)
; AVX1-NEXT:    vmovaps %ymm10, (%rdx)
; AVX1-NEXT:    vmovaps %ymm11, (%rcx)
; AVX1-NEXT:    vmovaps %ymm0, (%r8)
; AVX1-NEXT:    vzeroupper
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf16:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    vpxor %xmm0, %xmm0, %xmm0
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm2
; AVX2-SLOW-NEXT:    vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-SLOW-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-SLOW-NEXT:    vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm3
; AVX2-SLOW-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-SLOW-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %xmm9
; AVX2-SLOW-NEXT:    vmovdqa 16(%rdi), %xmm10
; AVX2-SLOW-NEXT:    vmovdqa 32(%rdi), %xmm11
; AVX2-SLOW-NEXT:    vmovdqa 48(%rdi), %xmm12
; AVX2-SLOW-NEXT:    vmovdqa 112(%rdi), %xmm6
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm6[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa 96(%rdi), %xmm7
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm7[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vmovdqa 80(%rdi), %xmm4
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm4[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa 64(%rdi), %xmm2
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
; 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 {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm12[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm11[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm10[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm5 = xmm9[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm15 = xmm6[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm15[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm8 = xmm7[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm8[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm4[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm14 = ymm6[0,1,2,3,4,5],ymm3[6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm12 = xmm12[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm12[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm11[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm0[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm7 = xmm10[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm7[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm6 = xmm9[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm6[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm15[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm8[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm12[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm7[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm6[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm2, (%rsi)
; AVX2-SLOW-NEXT:    vmovdqa %ymm13, (%rdx)
; AVX2-SLOW-NEXT:    vmovdqa %ymm1, (%rcx)
; AVX2-SLOW-NEXT:    vmovdqa %ymm0, (%r8)
; AVX2-SLOW-NEXT:    vzeroupper
; AVX2-SLOW-NEXT:    retq
;
; AVX2-FAST-ALL-LABEL: vf16:
; AVX2-FAST-ALL:       # %bb.0:
; AVX2-FAST-ALL-NEXT:    vmovdqa 64(%rdi), %ymm11
; AVX2-FAST-ALL-NEXT:    vmovdqa 96(%rdi), %ymm12
; AVX2-FAST-ALL-NEXT:    vpxor %xmm0, %xmm0, %xmm0
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm3 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-ALL-NEXT:    vpackusdw %xmm4, %xmm3, %xmm3
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm0, %xmm4
; AVX2-FAST-ALL-NEXT:    vpackusdw %xmm4, %xmm0, %xmm0
; AVX2-FAST-ALL-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,2,2,3,0,2,4,6]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm12, %ymm3, %ymm4
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm10 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,20,21,24,25,28,29>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm10, %ymm4, %ymm6
; AVX2-FAST-ALL-NEXT:    vpermd %ymm11, %ymm3, %ymm7
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm8 = <0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,u,u,u,u,u,u,u,u>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm8, %ymm7, %ymm3
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm6[6,7]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdi), %xmm13
; AVX2-FAST-ALL-NEXT:    vmovdqa 16(%rdi), %xmm6
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rdi), %xmm0
; AVX2-FAST-ALL-NEXT:    vmovdqa 48(%rdi), %xmm5
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm1 = [0,1,2,3,2,3,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm1, %xmm5, %xmm2
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm1, %xmm0, %xmm1
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm2 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm2, %xmm6, %xmm3
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm2, %xmm13, %xmm2
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; 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,18,19,22,23,26,27,30,31>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm15, %ymm4, %ymm3
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm14 = <2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,u,u,u,u,u,u,u,u>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm14, %ymm7, %ymm4
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm1 = [1,3,2,3,1,3,5,7]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm12, %ymm1, %ymm3
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm10, %ymm3, %ymm7
; AVX2-FAST-ALL-NEXT:    vpermd %ymm11, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm8, %ymm1, %ymm8
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm5 = xmm5[3,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm5[0,1,2,0,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm0[0,1,2,0,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm10 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm4 = xmm6[3,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm4[2,0,2,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm7 = xmm13[3,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm7[2,0,2,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm10[2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm15, %ymm3, %ymm3
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm14, %ymm1, %ymm1
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm5[0,1,3,1,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm4[3,1,2,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm7[3,1,2,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vmovaps %ymm1, (%rsi)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm9, (%rdx)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm2, (%rcx)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm0, (%r8)
; AVX2-FAST-ALL-NEXT:    vzeroupper
; AVX2-FAST-ALL-NEXT:    retq
;
; AVX2-FAST-PERLANE-LABEL: vf16:
; AVX2-FAST-PERLANE:       # %bb.0:
; AVX2-FAST-PERLANE-NEXT:    vpxor %xmm0, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm0, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdi), %xmm9
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 16(%rdi), %xmm10
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rdi), %xmm11
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 48(%rdi), %xmm12
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,2,3,2,3,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 112(%rdi), %xmm6
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm2, %xmm6, %xmm7
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 96(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm2, %xmm0, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm13
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm7 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 80(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm7, %xmm3, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 64(%rdi), %xmm5
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm7, %xmm5, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm13[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm2, %xmm12, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm2, %xmm11, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm7, %xmm10, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm7, %xmm9, %xmm7
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm15 = xmm6[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm15[0,1,2,0,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm8 = xmm0[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm8[0,1,2,0,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm3[2,0,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm5 = xmm5[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm5[2,0,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm14 = ymm6[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm12 = xmm12[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm12[0,1,2,0,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm2 = xmm11[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm7 = xmm10[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm7[2,0,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm6 = xmm9[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm6[2,0,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm14[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm15[0,1,3,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm8[0,1,3,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm5[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm12[0,1,3,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm7[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm6[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; 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:    vmovdqa %ymm13, (%rdx)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm0, (%rcx)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm1, (%r8)
; AVX2-FAST-PERLANE-NEXT:    vzeroupper
; AVX2-FAST-PERLANE-NEXT:    retq
;
; AVX512-LABEL: vf16:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm0 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512-NEXT:    vmovdqu64 (%rdi), %zmm1
; AVX512-NEXT:    vmovdqu64 64(%rdi), %zmm2
; AVX512-NEXT:    vpermi2w %zmm2, %zmm1, %zmm0
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm3 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512-NEXT:    vpermi2w %zmm2, %zmm1, %zmm3
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm4 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512-NEXT:    vpermi2w %zmm2, %zmm1, %zmm4
; AVX512-NEXT:    vmovdqa {{.*#+}} ymm5 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512-NEXT:    vpermi2w %zmm2, %zmm1, %zmm5
; AVX512-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX512-NEXT:    vmovdqa %ymm3, (%rdx)
; AVX512-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512-NEXT:    vmovdqa %ymm5, (%r8)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %wide.vec = load <64 x i16>, ptr %in.vec, align 32

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

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

  ret void
}

define void @vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: vf32:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $280, %rsp # imm = 0x118
; SSE-NEXT:    movdqa 224(%rdi), %xmm6
; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 192(%rdi), %xmm5
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 208(%rdi), %xmm7
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 96(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 112(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 64(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 80(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm4[0,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm3[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[0,1,0,2,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm2[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm10[0,1,0,2,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm7[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[0,2,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm5[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm8[0,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; SSE-NEXT:    movdqa 240(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm12[0,1,0,2,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm6[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm14[0,1,0,2,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 16(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm1[0,2,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1]
; SSE-NEXT:    movdqa 32(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 48(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm1[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm15[0,1,0,2,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[0,1,0,2,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 128(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 144(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm5[0,2,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm7[0,2,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
; SSE-NEXT:    movdqa 160(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 176(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[0,1,0,2,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm2[0,1,0,2,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm4[1,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[0,1,1,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm10[0,1,1,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[1,3,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm8[1,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm12[0,1,1,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm14[0,1,1,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[1,3,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm7[1,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[0,1,1,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw $237, (%rsp), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,3,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm15[0,1,1,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm11[0,1,1,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[3,1,2,3]
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[2,0,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm3[0,1,2,0,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm1[0],xmm11[1]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; SSE-NEXT:    # xmm9 = mem[3,1,2,3]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = mem[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[2,0,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm8[2,0,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm4[0,1,2,0,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm3[0,1,2,0,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm5[0],xmm2[1]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,1,2,3]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm0[2,0,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm1[2,0,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; SSE-NEXT:    # xmm14 = mem[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm0[0,1,2,0,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm14[0,1,2,0,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm5[0],xmm10[1],xmm5[1]
; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm6[0],xmm10[1]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
; SSE-NEXT:    # xmm13 = mem[3,1,2,3]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; SSE-NEXT:    # xmm12 = mem[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm13[2,0,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm12[2,0,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT:    # xmm6 = mem[3,1,2,3]
; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; SSE-NEXT:    # xmm5 = mem[3,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[0,1,2,0,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm5[0,1,2,0,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[3,1,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm8[3,1,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[0,1,3,1,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm3[0,1,3,1,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT:    pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm14[0,1,3,1,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm13[3,1,2,3,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm12[3,1,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,3,1,4,5,6,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,1,3,1,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm3[0],xmm5[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 32(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, (%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 48(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 16(%rsi)
; SSE-NEXT:    movapd %xmm7, (%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 32(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 48(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 16(%rdx)
; SSE-NEXT:    movapd %xmm15, 32(%rcx)
; SSE-NEXT:    movapd %xmm10, (%rcx)
; SSE-NEXT:    movapd %xmm2, 16(%rcx)
; SSE-NEXT:    movapd %xmm11, 48(%rcx)
; SSE-NEXT:    movapd %xmm5, 32(%r8)
; SSE-NEXT:    movapd %xmm0, (%r8)
; SSE-NEXT:    movapd %xmm1, 48(%r8)
; SSE-NEXT:    movapd %xmm4, 16(%r8)
; SSE-NEXT:    addq $280, %rsp # imm = 0x118
; SSE-NEXT:    retq
;
; AVX1-LABEL: vf32:
; AVX1:       # %bb.0:
; AVX1-NEXT:    subq $200, %rsp
; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
; AVX1-NEXT:    vmovdqa 240(%rdi), %xmm1
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0],xmm4[1,2,3],xmm1[4],xmm4[5,6,7]
; AVX1-NEXT:    vmovdqa %xmm1, %xmm12
; AVX1-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vmovdqa 224(%rdi), %xmm2
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0],xmm4[1,2,3],xmm2[4],xmm4[5,6,7]
; AVX1-NEXT:    vmovdqa %xmm2, %xmm14
; AVX1-NEXT:    vpackusdw %xmm0, %xmm1, %xmm0
; AVX1-NEXT:    vmovdqa 208(%rdi), %xmm2
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0],xmm4[1,2,3],xmm2[4],xmm4[5,6,7]
; AVX1-NEXT:    vmovdqa %xmm2, %xmm15
; AVX1-NEXT:    vmovdqa 192(%rdi), %xmm2
; AVX1-NEXT:    vmovdqa %xmm2, (%rsp) # 16-byte Spill
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0],xmm4[1,2,3],xmm2[4],xmm4[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
; AVX1-NEXT:    vpackusdw %xmm0, %xmm1, %xmm0
; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vmovdqa 176(%rdi), %xmm0
; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm4[1,2,3],xmm0[4],xmm4[5,6,7]
; AVX1-NEXT:    vmovdqa 160(%rdi), %xmm1
; AVX1-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm4[1,2,3],xmm1[4],xmm4[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm0, %xmm1, %xmm0
; AVX1-NEXT:    vmovdqa 144(%rdi), %xmm1
; AVX1-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm4[1,2,3],xmm1[4],xmm4[5,6,7]
; AVX1-NEXT:    vmovdqa 128(%rdi), %xmm2
; AVX1-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0],xmm4[1,2,3],xmm2[4],xmm4[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
; AVX1-NEXT:    vpackusdw %xmm0, %xmm1, %xmm0
; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vmovdqa 112(%rdi), %xmm8
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm8[0],xmm4[1,2,3],xmm8[4],xmm4[5,6,7]
; AVX1-NEXT:    vmovdqa 96(%rdi), %xmm5
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm5[0],xmm4[1,2,3],xmm5[4],xmm4[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm0, %xmm1, %xmm0
; AVX1-NEXT:    vmovdqa 80(%rdi), %xmm9
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm9[0],xmm4[1,2,3],xmm9[4],xmm4[5,6,7]
; AVX1-NEXT:    vmovdqa 64(%rdi), %xmm3
; AVX1-NEXT:    vpblendw {{.*#+}} xmm6 = xmm3[0],xmm4[1,2,3],xmm3[4],xmm4[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm1, %xmm6, %xmm1
; AVX1-NEXT:    vpackusdw %xmm0, %xmm1, %xmm0
; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vmovdqa 32(%rdi), %xmm10
; AVX1-NEXT:    vmovdqa 48(%rdi), %xmm0
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm4[1,2,3],xmm0[4],xmm4[5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm7 = xmm10[0],xmm4[1,2,3],xmm10[4],xmm4[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm1, %xmm7, %xmm13
; AVX1-NEXT:    vmovdqa (%rdi), %xmm11
; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX1-NEXT:    vpblendw {{.*#+}} xmm7 = xmm1[0],xmm4[1,2,3],xmm1[4],xmm4[5,6,7]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm11[0],xmm4[1,2,3],xmm11[4],xmm4[5,6,7]
; AVX1-NEXT:    vpackusdw %xmm7, %xmm4, %xmm4
; AVX1-NEXT:    vpackusdw %xmm13, %xmm4, %xmm2
; AVX1-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm8[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm5[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm9[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm3[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm6[0,1,2,3],xmm4[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm0[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm10[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm7 = xmm1[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm11[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm6[4,5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
; AVX1-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
; AVX1-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm12[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vmovdqa %xmm14, %xmm12
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm14[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
; AVX1-NEXT:    vmovdqa %xmm15, %xmm14
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm15[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vmovdqa (%rsp), %xmm15 # 16-byte Reload
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm15[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0,1,2,3],xmm2[4,5,6,7]
; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm13[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm6 = mem[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
; AVX1-NEXT:    vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm6 = mem[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm7 = mem[0,2,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm6[0,1,2,3],xmm4[4,5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX1-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX1-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm8[3,1,2,3]
; AVX1-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm5[3,1,2,3]
; AVX1-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm5[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm9[3,1,2,3]
; AVX1-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm3[3,1,2,3]
; AVX1-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm4[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm5[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm9 = xmm3[0,1,2,3],xmm2[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm8 = xmm0[3,1,2,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm10[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm8[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm6[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm1[3,1,2,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm11[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm5[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm4[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm1
; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm10 = mem[3,1,2,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm11 = xmm12[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm10[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm11[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm12 = xmm14[3,1,2,3]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm9 = xmm15[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm12[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm9[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm7 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX1-NEXT:    vpshufd {{.*#+}} xmm14 = xmm13[3,1,2,3]
; AVX1-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm15 = mem[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm14[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm15[0,1,2,0,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX1-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm3 = mem[3,1,2,3]
; AVX1-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm2 = mem[3,1,2,3]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm3[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm13 = xmm2[2,0,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm1
; AVX1-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-NEXT:    vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm0 = mem[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX1-NEXT:    vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm1 = mem[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX1-NEXT:    # xmm7 = mem[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm8[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm6[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm4[0,1,2,3],xmm1[4,5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm10[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm11[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm12[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm9[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm4[0,1,2,3],xmm1[4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm14[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm15[0,1,3,1,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm4[4,5,6,7]
; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm2, (%rsi)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm2, 16(%rsi)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm2, 32(%rsi)
; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-NEXT:    vmovaps %xmm2, 48(%rsi)
; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-NEXT:    vmovaps %ymm2, 32(%rdx)
; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-NEXT:    vmovaps %ymm2, (%rdx)
; AVX1-NEXT:    vmovaps %ymm13, 32(%rcx)
; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-NEXT:    vmovaps %ymm2, (%rcx)
; AVX1-NEXT:    vmovaps %ymm1, 32(%r8)
; AVX1-NEXT:    vmovaps %ymm0, (%r8)
; AVX1-NEXT:    addq $200, %rsp
; AVX1-NEXT:    vzeroupper
; AVX1-NEXT:    retq
;
; AVX2-SLOW-LABEL: vf32:
; AVX2-SLOW:       # %bb.0:
; AVX2-SLOW-NEXT:    subq $200, %rsp
; AVX2-SLOW-NEXT:    vpxor %xmm0, %xmm0, %xmm0
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm2
; AVX2-SLOW-NEXT:    vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-SLOW-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-SLOW-NEXT:    vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm3 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm3, %xmm4
; AVX2-SLOW-NEXT:    vpackusdw %xmm4, %xmm3, %xmm3
; AVX2-SLOW-NEXT:    vpackusdw %xmm3, %xmm3, %xmm3
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm2
; AVX2-SLOW-NEXT:    vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-SLOW-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-SLOW-NEXT:    vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm3
; AVX2-SLOW-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-SLOW-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa (%rdi), %xmm12
; AVX2-SLOW-NEXT:    vmovdqa 16(%rdi), %xmm7
; AVX2-SLOW-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX2-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa 48(%rdi), %xmm15
; AVX2-SLOW-NEXT:    vmovdqa 112(%rdi), %xmm11
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm11[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa 96(%rdi), %xmm10
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm10[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vmovdqa 80(%rdi), %xmm9
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm9[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa 64(%rdi), %xmm8
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm8[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; 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 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm15[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm7[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm12[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vmovdqa 240(%rdi), %xmm0
; AVX2-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa 224(%rdi), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa %xmm1, (%rsp) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-SLOW-NEXT:    vmovdqa 208(%rdi), %xmm6
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm6[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa 192(%rdi), %xmm3
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-SLOW-NEXT:    vmovdqa 176(%rdi), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa 160(%rdi), %xmm2
; AVX2-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; AVX2-SLOW-NEXT:    vmovdqa 144(%rdi), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm13 = xmm1[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm13 = xmm13[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqa 128(%rdi), %xmm1
; AVX2-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm14 = xmm1[0,2,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm14[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm5[0],xmm13[0],xmm5[1],xmm13[1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm11[3,1,2,3]
; AVX2-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm10[3,1,2,3]
; AVX2-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm9[3,1,2,3]
; AVX2-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm8[3,1,2,3]
; AVX2-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm0[6,7]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm15 = xmm15[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm14 = mem[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm15[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm14[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm13 = xmm7[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm7 = xmm12[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm13[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm12 = xmm7[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm12 = mem[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshufd $231, (%rsp), %xmm10 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm10 = mem[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm12[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm10[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm9 = xmm6[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm8 = xmm3[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm9[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm8[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm6 = mem[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm5 = mem[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm6[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm5[0,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm4 = mem[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm2 = mem[3,1,2,3]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm4[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm11 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm11 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm0 = mem[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm1 = mem[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-SLOW-NEXT:    vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm1 = mem[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-SLOW-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm15[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm14[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm13[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm7[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm12[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm10[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm9[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm8[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm6[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm5[0,1,3,1,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; 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, (%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:    vmovdqa %ymm11, 32(%rcx)
; AVX2-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT:    vmovaps %ymm2, (%rcx)
; AVX2-SLOW-NEXT:    vmovdqa %ymm1, 32(%r8)
; AVX2-SLOW-NEXT:    vmovdqa %ymm0, (%r8)
; AVX2-SLOW-NEXT:    addq $200, %rsp
; 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 192(%rdi), %ymm6
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 224(%rdi), %ymm7
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 64(%rdi), %ymm5
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm5, (%rsp) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 96(%rdi), %ymm4
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpxor %xmm0, %xmm0, %xmm0
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-ALL-NEXT:    vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-ALL-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FAST-ALL-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,2,2,3,0,2,4,6]
; AVX2-FAST-ALL-NEXT:    vpermd %ymm4, %ymm3, %ymm10
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,20,21,24,25,28,29>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm2, %ymm10, %ymm4
; AVX2-FAST-ALL-NEXT:    vpermd %ymm5, %ymm3, %ymm8
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm9 = <0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,u,u,u,u,u,u,u,u>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm9, %ymm8, %ymm5
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm9, %ymm11
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm1, %xmm4
; AVX2-FAST-ALL-NEXT:    vpackusdw %xmm4, %xmm1, %xmm1
; AVX2-FAST-ALL-NEXT:    vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-ALL-NEXT:    vextracti128 $1, %ymm0, %xmm4
; AVX2-FAST-ALL-NEXT:    vpackusdw %xmm4, %xmm0, %xmm0
; AVX2-FAST-ALL-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
; AVX2-FAST-ALL-NEXT:    vpermd %ymm7, %ymm3, %ymm15
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm2, %ymm15, %ymm1
; AVX2-FAST-ALL-NEXT:    vpermd %ymm6, %ymm3, %ymm9
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm11, %ymm9, %ymm4
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 32(%rdi), %xmm12
; AVX2-FAST-ALL-NEXT:    vmovdqa 48(%rdi), %xmm2
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,2,3,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm6, %xmm2, %xmm1
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm6, %xmm12, %xmm4
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; AVX2-FAST-ALL-NEXT:    vmovdqa (%rdi), %xmm11
; AVX2-FAST-ALL-NEXT:    vmovdqa 16(%rdi), %xmm3
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm1, %xmm3, %xmm4
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm1, %xmm11, %xmm7
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3]
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm14 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,22,23,26,27,30,31>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm14, %ymm10, %ymm5
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm13 = <2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,u,u,u,u,u,u,u,u>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm13, %ymm8, %ymm7
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm5 = ymm7[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 160(%rdi), %xmm0
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 176(%rdi), %xmm4
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm6, %xmm4, %xmm7
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm6, %xmm0, %xmm6
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm8 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
; AVX2-FAST-ALL-NEXT:    vmovdqa 128(%rdi), %xmm5
; AVX2-FAST-ALL-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa 144(%rdi), %xmm10
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm1, %xmm10, %xmm0
; AVX2-FAST-ALL-NEXT:    vpshufb %xmm1, %xmm5, %xmm1
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm8[2,3]
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm14, %ymm15, %ymm1
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm13, %ymm9, %ymm8
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm1 = [1,3,2,3,1,3,5,7]
; AVX2-FAST-ALL-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpermd (%rsp), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,20,21,24,25,28,29>
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm7, %ymm0, %ymm0
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm9 = ymm15[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,u,u,u,u,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm14 = ymm9[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm13 = xmm2[3,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm12 = xmm12[3,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm13[0,1,2,0,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm12[0,1,2,0,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm8 = xmm11[3,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm3[2,0,2,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm8[2,0,2,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm14[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm7, %ymm9, %ymm5
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm11 = ymm1[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,u,u,u,u,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # xmm2 = mem[3,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm4[0,1,2,0,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm2[0,1,2,0,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm7[0],xmm5[0],xmm7[1],xmm5[1]
; AVX2-FAST-ALL-NEXT:    vpshufd {{.*#+}} xmm10 = xmm10[3,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX2-FAST-ALL-NEXT:    # xmm7 = mem[3,1,2,3]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm10[2,0,2,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm14 = xmm7[2,0,2,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm11[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-ALL-NEXT:    vmovdqa {{.*#+}} ymm14 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,22,23,26,27,30,31>
; AVX2-FAST-ALL-NEXT:    vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm14, %ymm0, %ymm5
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm11 = ymm15[2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,u,u,u,u,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm5 = ymm11[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm13[0,1,3,1,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm12[0,1,3,1,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm3[3,1,2,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm8[3,1,2,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshufb %ymm14, %ymm9, %ymm3
; AVX2-FAST-ALL-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[2,3,6,7,10,11,14,15,14,15,10,11,12,13,14,15,18,19,22,23,26,27,30,31,u,u,u,u,u,u,u,u]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm4[0,1,3,1,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm10[3,1,2,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm7[3,1,2,3,4,5,6,7]
; AVX2-FAST-ALL-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
; AVX2-FAST-ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; 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 {{[-0-9]+}}(%r{{[sb]}}p), %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 %ymm1, 32(%r8)
; AVX2-FAST-ALL-NEXT:    vmovdqa %ymm0, (%r8)
; 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 $184, %rsp
; AVX2-FAST-PERLANE-NEXT:    vpxor %xmm0, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm3 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm4, %xmm3, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm3, %xmm3, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm1 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm2, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm2 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm2, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpblendw {{.*#+}} ymm0 = mem[0],ymm0[1,2,3],mem[4],ymm0[5,6,7],mem[8],ymm0[9,10,11],mem[12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vextracti128 $1, %ymm0, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa (%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 16(%rdi), %xmm13
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 32(%rdi), %xmm14
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 48(%rdi), %xmm15
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm5 = [0,1,2,3,2,3,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 112(%rdi), %xmm11
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm5, %xmm11, %xmm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 96(%rdi), %xmm12
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm5, %xmm12, %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa {{.*#+}} xmm6 = [2,3,10,11,8,9,10,11,8,9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 80(%rdi), %xmm10
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm6, %xmm10, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 64(%rdi), %xmm1
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm6, %xmm1, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm5, %xmm15, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm5, %xmm14, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm6, %xmm13, %xmm4
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm6, %xmm2, %xmm7
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 240(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm5, %xmm0, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 224(%rdi), %xmm7
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm5, %xmm7, %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm8 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 208(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm6, %xmm0, %xmm9
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 192(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm6, %xmm3, %xmm2
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 160(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 176(%rdi), %xmm4
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm5, %xmm4, %xmm9
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm5, %xmm0, %xmm5
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 128(%rdi), %xmm4
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm4, (%rsp) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vmovdqa 144(%rdi), %xmm9
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm6, %xmm9, %xmm0
; AVX2-FAST-PERLANE-NEXT:    vpshufb %xmm6, %xmm4, %xmm6
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm0 = xmm11[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm4 = xmm12[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm4[0,1,2,0,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm2 = xmm10[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm4 = xmm1[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm4[2,0,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm15 = xmm15[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm14 = xmm14[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm15[0,1,2,0,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm14[0,1,2,0,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm13 = xmm13[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # xmm4 = mem[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm13[2,0,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm4[2,0,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # xmm11 = mem[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm8 = xmm7[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm11[0,1,2,0,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm8[0,1,2,0,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; AVX2-FAST-PERLANE-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # xmm2 = mem[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm2[2,0,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm7 = xmm3[2,0,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm6, %ymm0, %ymm6
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # xmm7 = mem[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # xmm10 = mem[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm7[0,1,2,0,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm10[0,1,2,0,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; AVX2-FAST-PERLANE-NEXT:    vpshufd {{.*#+}} xmm9 = xmm9[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshufd $231, (%rsp), %xmm1 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # xmm1 = mem[3,1,2,3]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm9[2,0,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm12 = xmm1[2,0,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # xmm0 = mem[0,1,3,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # xmm6 = mem[0,1,3,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # xmm6 = mem[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT:    # xmm5 = mem[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm15[0,1,3,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm14[0,1,3,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm6 = xmm13[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm11[0,1,3,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm5 = xmm8[0,1,3,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm4, %ymm0, %ymm3
; AVX2-FAST-PERLANE-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm7[0,1,3,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm10[0,1,3,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm4 = xmm9[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX2-FAST-PERLANE-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; 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, (%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:    vmovdqa %ymm12, 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:    vmovdqa %ymm1, 32(%r8)
; AVX2-FAST-PERLANE-NEXT:    vmovdqa %ymm0, (%r8)
; AVX2-FAST-PERLANE-NEXT:    addq $184, %rsp
; AVX2-FAST-PERLANE-NEXT:    vzeroupper
; AVX2-FAST-PERLANE-NEXT:    retq
;
; AVX512-LABEL: vf32:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqu64 (%rdi), %zmm0
; AVX512-NEXT:    vmovdqu64 64(%rdi), %zmm1
; AVX512-NEXT:    vmovdqu64 128(%rdi), %zmm2
; AVX512-NEXT:    vmovdqu64 192(%rdi), %zmm3
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,0,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60]
; AVX512-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512-NEXT:    vpermt2w %zmm3, %zmm4, %zmm5
; AVX512-NEXT:    vpermi2w %zmm1, %zmm0, %zmm4
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7]
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61,1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
; AVX512-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm6
; AVX512-NEXT:    vpermt2w %zmm3, %zmm5, %zmm6
; AVX512-NEXT:    vpermi2w %zmm1, %zmm0, %zmm5
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7]
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62,2,6,10,14,18,22,26,30,34,38,42,46,50,54,58,62]
; AVX512-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512-NEXT:    vpermt2w %zmm3, %zmm6, %zmm7
; AVX512-NEXT:    vpermi2w %zmm1, %zmm0, %zmm6
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7]
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63,3,7,11,15,19,23,27,31,35,39,43,47,51,55,59,63]
; AVX512-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermt2w %zmm3, %zmm7, %zmm2
; AVX512-NEXT:    vpermt2w %zmm1, %zmm7, %zmm0
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 %zmm4, (%rsi)
; AVX512-NEXT:    vmovdqu64 %zmm5, (%rdx)
; AVX512-NEXT:    vmovdqu64 %zmm6, (%rcx)
; AVX512-NEXT:    vmovdqu64 %zmm0, (%r8)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
  %wide.vec = load <128 x i16>, ptr %in.vec, align 32

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

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

  ret void
}