; 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,FALLBACK0
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx  | FileCheck %s --check-prefixes=AVX,AVX1,AVX1-ONLY,FALLBACK1
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX1,AVX2,AVX2-ONLY,AVX2-SLOW,FALLBACK2
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX1,AVX2,AVX2-ONLY,AVX2-FAST,FALLBACK3
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX1,AVX2,AVX2-ONLY,AVX2-FAST-PERLANE,FALLBACK4
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512F,AVX512-SLOW,AVX512F-SLOW,AVX512F-ONLY-SLOW,FALLBACK5
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512F,AVX512-FAST,AVX512F-FAST,AVX512F-ONLY-FAST,FALLBACK6
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512F,AVX512-SLOW,AVX512F-SLOW,AVX512DQ-SLOW,FALLBACK7
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512F,AVX512-FAST,AVX512F-FAST,AVX512DQ-FAST,FALLBACK8
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512BW,AVX512-SLOW,AVX512BW-SLOW,AVX512BW-ONLY-SLOW,FALLBACK9
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512bw,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512BW,AVX512-FAST,AVX512BW-FAST,AVX512BW-ONLY-FAST,FALLBACK10
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq,+avx512bw | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512BW,AVX512-SLOW,AVX512BW-SLOW,AVX512DQBW-SLOW,FALLBACK11
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq,+avx512bw,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512BW,AVX512-FAST,AVX512BW-FAST,AVX512DQBW-FAST,FALLBACK12

; These patterns are produced by LoopVectorizer for interleaved loads.

define void @load_i16_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
; SSE-LABEL: load_i16_stride8_vf2:
; SSE:       # %bb.0:
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    movdqa 16(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm0, %xmm2
; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[2,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[3,3,3,3]
; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[2,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm0[3,3,3,3]
; SSE-NEXT:    movd %xmm2, (%rsi)
; SSE-NEXT:    movd %xmm3, (%rdx)
; SSE-NEXT:    movd %xmm4, (%rcx)
; SSE-NEXT:    movd %xmm5, (%r8)
; SSE-NEXT:    movd %xmm0, (%r9)
; SSE-NEXT:    movd %xmm1, (%r11)
; SSE-NEXT:    movd %xmm6, (%r10)
; SSE-NEXT:    movd %xmm7, (%rax)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i16_stride8_vf2:
; AVX:       # %bb.0:
; AVX-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; AVX-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; AVX-NEXT:    vmovdqa (%rdi), %xmm0
; AVX-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; AVX-NEXT:    vmovd %xmm2, (%rsi)
; AVX-NEXT:    vpextrd $1, %xmm2, (%rdx)
; AVX-NEXT:    vpextrd $2, %xmm2, (%rcx)
; AVX-NEXT:    vpextrd $3, %xmm2, (%r8)
; AVX-NEXT:    vmovd %xmm0, (%r9)
; AVX-NEXT:    vpextrd $1, %xmm0, (%r11)
; AVX-NEXT:    vpextrd $2, %xmm0, (%r10)
; AVX-NEXT:    vpextrd $3, %xmm0, (%rax)
; AVX-NEXT:    retq
  %wide.vec = load <16 x i16>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 0, i32 8>
  %strided.vec1 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 1, i32 9>
  %strided.vec2 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 2, i32 10>
  %strided.vec3 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 3, i32 11>
  %strided.vec4 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 4, i32 12>
  %strided.vec5 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 5, i32 13>
  %strided.vec6 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 6, i32 14>
  %strided.vec7 = shufflevector <16 x i16> %wide.vec, <16 x i16> poison, <2 x i32> <i32 7, i32 15>
  store <2 x i16> %strided.vec0, ptr %out.vec0, align 64
  store <2 x i16> %strided.vec1, ptr %out.vec1, align 64
  store <2 x i16> %strided.vec2, ptr %out.vec2, align 64
  store <2 x i16> %strided.vec3, ptr %out.vec3, align 64
  store <2 x i16> %strided.vec4, ptr %out.vec4, align 64
  store <2 x i16> %strided.vec5, ptr %out.vec5, align 64
  store <2 x i16> %strided.vec6, ptr %out.vec6, align 64
  store <2 x i16> %strided.vec7, ptr %out.vec7, align 64
  ret void
}

define void @load_i16_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
; SSE-LABEL: load_i16_stride8_vf4:
; SSE:       # %bb.0:
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    movdqa 16(%rdi), %xmm1
; SSE-NEXT:    movdqa 32(%rdi), %xmm2
; SSE-NEXT:    movdqa 48(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm2, %xmm4
; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; SSE-NEXT:    movdqa %xmm0, %xmm5
; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
; SSE-NEXT:    movdqa %xmm5, %xmm6
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm4[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm5[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm5[3,3,3,3]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; SSE-NEXT:    movdqa %xmm0, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; SSE-NEXT:    movq %xmm6, (%rsi)
; SSE-NEXT:    movq %xmm8, (%rdx)
; SSE-NEXT:    movq %xmm5, (%rcx)
; SSE-NEXT:    movq %xmm7, (%r8)
; SSE-NEXT:    movq %xmm1, (%r9)
; SSE-NEXT:    movq %xmm4, (%r11)
; SSE-NEXT:    movq %xmm0, (%r10)
; SSE-NEXT:    movq %xmm3, (%rax)
; SSE-NEXT:    retq
;
; AVX1-ONLY-LABEL: load_i16_stride8_vf4:
; AVX1-ONLY:       # %bb.0:
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; AVX1-ONLY-NEXT:    vmovdqa (%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa 48(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm7 = xmm5[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm7 = xmm7[0,1],xmm4[2,3],xmm7[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm8 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm5 = xmm5[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3],xmm5[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
; AVX1-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovq %xmm6, (%rsi)
; AVX1-ONLY-NEXT:    vmovq %xmm7, (%rdx)
; AVX1-ONLY-NEXT:    vmovq %xmm8, (%rcx)
; AVX1-ONLY-NEXT:    vmovq %xmm4, (%r8)
; AVX1-ONLY-NEXT:    vmovq %xmm1, (%r9)
; AVX1-ONLY-NEXT:    vmovq %xmm3, (%r11)
; AVX1-ONLY-NEXT:    vmovq %xmm5, (%r10)
; AVX1-ONLY-NEXT:    vmovq %xmm0, (%rax)
; AVX1-ONLY-NEXT:    retq
;
; AVX2-ONLY-LABEL: load_i16_stride8_vf4:
; AVX2-ONLY:       # %bb.0:
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; AVX2-ONLY-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-ONLY-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX2-ONLY-NEXT:    vmovdqa 48(%rdi), %xmm3
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm7 = xmm5[1,1,1,1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm7 = xmm7[0],xmm4[1],xmm7[2,3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm8 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm5 = xmm5[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3]
; AVX2-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
; AVX2-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,1,1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0],xmm2[1],xmm3[2,3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vmovq %xmm6, (%rsi)
; AVX2-ONLY-NEXT:    vmovq %xmm7, (%rdx)
; AVX2-ONLY-NEXT:    vmovq %xmm8, (%rcx)
; AVX2-ONLY-NEXT:    vmovq %xmm4, (%r8)
; AVX2-ONLY-NEXT:    vmovq %xmm1, (%r9)
; AVX2-ONLY-NEXT:    vmovq %xmm3, (%r11)
; AVX2-ONLY-NEXT:    vmovq %xmm5, (%r10)
; AVX2-ONLY-NEXT:    vmovq %xmm0, (%rax)
; AVX2-ONLY-NEXT:    retq
;
; AVX512F-SLOW-LABEL: load_i16_stride8_vf4:
; AVX512F-SLOW:       # %bb.0:
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; AVX512F-SLOW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa 48(%rdi), %xmm3
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm7 = xmm5[1,1,1,1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm7 = xmm7[0],xmm4[1],xmm7[2,3]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm8 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; AVX512F-SLOW-NEXT:    vmovdqa {{.*#+}} xmm9 = [3,7,3,3]
; AVX512F-SLOW-NEXT:    vpermt2d %xmm4, %xmm9, %xmm5
; AVX512F-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
; AVX512F-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,1,1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0],xmm2[1],xmm3[2,3]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm4 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; AVX512F-SLOW-NEXT:    vpermt2d %xmm2, %xmm9, %xmm0
; AVX512F-SLOW-NEXT:    vmovq %xmm6, (%rsi)
; AVX512F-SLOW-NEXT:    vmovq %xmm7, (%rdx)
; AVX512F-SLOW-NEXT:    vmovq %xmm8, (%rcx)
; AVX512F-SLOW-NEXT:    vmovq %xmm5, (%r8)
; AVX512F-SLOW-NEXT:    vmovq %xmm1, (%r9)
; AVX512F-SLOW-NEXT:    vmovq %xmm3, (%r11)
; AVX512F-SLOW-NEXT:    vmovq %xmm4, (%r10)
; AVX512F-SLOW-NEXT:    vmovq %xmm0, (%rax)
; AVX512F-SLOW-NEXT:    retq
;
; AVX512F-FAST-LABEL: load_i16_stride8_vf4:
; AVX512F-FAST:       # %bb.0:
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; AVX512F-FAST-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512F-FAST-NEXT:    vmovdqa 48(%rdi), %xmm3
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX512F-FAST-NEXT:    vmovdqa {{.*#+}} xmm7 = [1,5,1,1]
; AVX512F-FAST-NEXT:    vmovdqa %xmm5, %xmm8
; AVX512F-FAST-NEXT:    vpermt2d %xmm4, %xmm7, %xmm8
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm9 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; AVX512F-FAST-NEXT:    vmovdqa {{.*#+}} xmm10 = [3,7,3,3]
; AVX512F-FAST-NEXT:    vpermt2d %xmm4, %xmm10, %xmm5
; AVX512F-FAST-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
; AVX512F-FAST-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; AVX512F-FAST-NEXT:    vpermi2d %xmm2, %xmm0, %xmm7
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; AVX512F-FAST-NEXT:    vpermt2d %xmm2, %xmm10, %xmm0
; AVX512F-FAST-NEXT:    vmovq %xmm6, (%rsi)
; AVX512F-FAST-NEXT:    vmovq %xmm8, (%rdx)
; AVX512F-FAST-NEXT:    vmovq %xmm9, (%rcx)
; AVX512F-FAST-NEXT:    vmovq %xmm5, (%r8)
; AVX512F-FAST-NEXT:    vmovq %xmm1, (%r9)
; AVX512F-FAST-NEXT:    vmovq %xmm7, (%r11)
; AVX512F-FAST-NEXT:    vmovq %xmm3, (%r10)
; AVX512F-FAST-NEXT:    vmovq %xmm0, (%rax)
; AVX512F-FAST-NEXT:    retq
;
; AVX512BW-LABEL: load_i16_stride8_vf4:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} xmm0 = [0,8,16,24,0,8,16,24]
; AVX512BW-NEXT:    vmovdqa (%rdi), %ymm1
; AVX512BW-NEXT:    vmovdqa 32(%rdi), %ymm2
; AVX512BW-NEXT:    vpermi2w %ymm2, %ymm1, %ymm0
; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} xmm3 = [1,9,17,25,1,9,17,25]
; AVX512BW-NEXT:    vpermi2w %ymm2, %ymm1, %ymm3
; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} xmm4 = [2,10,18,26,2,10,18,26]
; AVX512BW-NEXT:    vpermi2w %ymm2, %ymm1, %ymm4
; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} xmm5 = [3,11,19,27,3,11,19,27]
; AVX512BW-NEXT:    vpermi2w %ymm2, %ymm1, %ymm5
; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} xmm6 = [4,12,20,28,4,12,20,28]
; AVX512BW-NEXT:    vpermi2w %ymm2, %ymm1, %ymm6
; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} xmm7 = [5,13,21,29,5,13,21,29]
; AVX512BW-NEXT:    vpermi2w %ymm2, %ymm1, %ymm7
; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} xmm8 = [6,14,22,30,6,14,22,30]
; AVX512BW-NEXT:    vpermi2w %ymm2, %ymm1, %ymm8
; AVX512BW-NEXT:    vpbroadcastq {{.*#+}} xmm9 = [7,15,23,31,7,15,23,31]
; AVX512BW-NEXT:    vpermi2w %ymm2, %ymm1, %ymm9
; AVX512BW-NEXT:    vmovq %xmm0, (%rsi)
; AVX512BW-NEXT:    vmovq %xmm3, (%rdx)
; AVX512BW-NEXT:    vmovq %xmm4, (%rcx)
; AVX512BW-NEXT:    vmovq %xmm5, (%r8)
; AVX512BW-NEXT:    vmovq %xmm6, (%r9)
; AVX512BW-NEXT:    vmovq %xmm7, (%r11)
; AVX512BW-NEXT:    vmovq %xmm8, (%r10)
; AVX512BW-NEXT:    vmovq %xmm9, (%rax)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
  %wide.vec = load <32 x i16>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
  %strided.vec1 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 1, i32 9, i32 17, i32 25>
  %strided.vec2 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 2, i32 10, i32 18, i32 26>
  %strided.vec3 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 3, i32 11, i32 19, i32 27>
  %strided.vec4 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 4, i32 12, i32 20, i32 28>
  %strided.vec5 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 5, i32 13, i32 21, i32 29>
  %strided.vec6 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 6, i32 14, i32 22, i32 30>
  %strided.vec7 = shufflevector <32 x i16> %wide.vec, <32 x i16> poison, <4 x i32> <i32 7, i32 15, i32 23, i32 31>
  store <4 x i16> %strided.vec0, ptr %out.vec0, align 64
  store <4 x i16> %strided.vec1, ptr %out.vec1, align 64
  store <4 x i16> %strided.vec2, ptr %out.vec2, align 64
  store <4 x i16> %strided.vec3, ptr %out.vec3, align 64
  store <4 x i16> %strided.vec4, ptr %out.vec4, align 64
  store <4 x i16> %strided.vec5, ptr %out.vec5, align 64
  store <4 x i16> %strided.vec6, ptr %out.vec6, align 64
  store <4 x i16> %strided.vec7, ptr %out.vec7, align 64
  ret void
}

define void @load_i16_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
; SSE-LABEL: load_i16_stride8_vf8:
; SSE:       # %bb.0:
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    movdqa 16(%rdi), %xmm7
; SSE-NEXT:    movdqa 32(%rdi), %xmm2
; SSE-NEXT:    movdqa 48(%rdi), %xmm9
; SSE-NEXT:    movdqa 80(%rdi), %xmm10
; SSE-NEXT:    movdqa 64(%rdi), %xmm1
; SSE-NEXT:    movdqa 112(%rdi), %xmm11
; SSE-NEXT:    movdqa 96(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, %xmm13
; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm11[0],xmm13[1],xmm11[1],xmm13[2],xmm11[2],xmm13[3],xmm11[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[0,0,0,0]
; SSE-NEXT:    movdqa %xmm1, %xmm12
; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm10[0],xmm12[1],xmm10[1],xmm12[2],xmm10[2],xmm12[3],xmm10[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm12[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm4[2],xmm6[3],xmm4[3]
; SSE-NEXT:    movdqa %xmm2, %xmm14
; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1],xmm14[2],xmm9[2],xmm14[3],xmm9[3]
; SSE-NEXT:    movdqa %xmm0, %xmm4
; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1],xmm4[2],xmm7[2],xmm4[3],xmm7[3]
; SSE-NEXT:    movdqa %xmm4, %xmm5
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm14[0],xmm5[1],xmm14[1]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
; SSE-NEXT:    movdqa %xmm12, %xmm8
; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm13[0],xmm8[1],xmm13[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm14[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm15[0],xmm6[1],xmm15[1]
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm13[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm12[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm8[2],xmm15[3],xmm8[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm4[3,3,3,3]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm14[2],xmm4[3],xmm14[3]
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm15[2,3]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm13[2],xmm12[3],xmm13[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm14[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm13[0],xmm8[1],xmm13[1]
; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm12[2,3]
; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm11[4],xmm3[5],xmm11[5],xmm3[6],xmm11[6],xmm3[7],xmm11[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm3[0,0,0,0]
; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm1[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm11[2],xmm10[3],xmm11[3]
; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm9[4],xmm2[5],xmm9[5],xmm2[6],xmm9[6],xmm2[7],xmm9[7]
; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
; SSE-NEXT:    movdqa %xmm0, %xmm7
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1]
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3]
; SSE-NEXT:    movdqa %xmm1, %xmm9
; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm3[0],xmm9[1],xmm3[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm2[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm0[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm9[2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm3[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm1[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[3,3,3,3]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm10[2,3]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm3[2],xmm1[3],xmm3[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm2[0],xmm9[1],xmm2[1]
; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm1[2,3]
; SSE-NEXT:    movaps %xmm5, (%rsi)
; SSE-NEXT:    movaps %xmm6, (%rdx)
; SSE-NEXT:    movaps %xmm4, (%rcx)
; SSE-NEXT:    movaps %xmm8, (%r8)
; SSE-NEXT:    movaps %xmm7, (%r9)
; SSE-NEXT:    movaps %xmm11, (%r11)
; SSE-NEXT:    movaps %xmm0, (%r10)
; SSE-NEXT:    movaps %xmm9, (%rax)
; SSE-NEXT:    retq
;
; AVX1-ONLY-LABEL: load_i16_stride8_vf8:
; AVX1-ONLY:       # %bb.0:
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; AVX1-ONLY-NEXT:    vmovdqa 112(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa 96(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm4[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa 80(%rdi), %xmm5
; AVX1-ONLY-NEXT:    vmovdqa 64(%rdi), %xmm6
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm7 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm7[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vmovdqa (%rdi), %xmm8
; AVX1-ONLY-NEXT:    vmovdqa 16(%rdi), %xmm9
; AVX1-ONLY-NEXT:    vmovdqa 32(%rdi), %xmm10
; AVX1-ONLY-NEXT:    vmovdqa 48(%rdi), %xmm11
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm12 = xmm10[0],xmm11[0],xmm10[1],xmm11[1],xmm10[2],xmm11[2],xmm10[3],xmm11[3]
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm13 = xmm8[0],xmm9[0],xmm8[1],xmm9[1],xmm8[2],xmm9[2],xmm8[3],xmm9[3]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm14 = xmm13[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm14 = xmm14[0,1],xmm12[2,3],xmm14[4,5,6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm14[0,1,2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm14 = xmm4[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm14 = xmm7[0,1,2,3,4,5],xmm14[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm15 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm14 = xmm15[0,1,2,3],xmm14[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm4 = xmm7[2],xmm4[2],xmm7[3],xmm4[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm7 = xmm12[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm12 = xmm13[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm7 = xmm12[0,1],xmm7[2,3],xmm12[4,5,6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm4 = xmm7[0,1,2,3],xmm4[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
; AVX1-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm5 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm6 = xmm5[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm6[0,1,2,3,4,5],xmm3[6,7]
; AVX1-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm6 = xmm10[4],xmm11[4],xmm10[5],xmm11[5],xmm10[6],xmm11[6],xmm10[7],xmm11[7]
; AVX1-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm7 = xmm8[4],xmm9[4],xmm8[5],xmm9[5],xmm8[6],xmm9[6],xmm8[7],xmm9[7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm8 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm8[0,1,2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm8 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm9 = xmm7[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm9 = xmm9[0,1],xmm6[2,3],xmm9[4,5,6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm8 = xmm9[0,1,2,3],xmm8[4,5,6,7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm9 = xmm2[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm9 = xmm5[0,1,2,3,4,5],xmm9[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm10 = xmm7[2],xmm6[2],xmm7[3],xmm6[3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm9 = xmm10[0,1,2,3],xmm9[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm5 = xmm6[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm6 = xmm7[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3],xmm6[4,5,6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm2 = xmm5[0,1,2,3],xmm2[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, (%rdx)
; AVX1-ONLY-NEXT:    vmovdqa %xmm14, (%rcx)
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, (%r8)
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, (%r9)
; AVX1-ONLY-NEXT:    vmovdqa %xmm8, (%r11)
; AVX1-ONLY-NEXT:    vmovdqa %xmm9, (%r10)
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, (%rax)
; AVX1-ONLY-NEXT:    retq
;
; AVX2-ONLY-LABEL: load_i16_stride8_vf8:
; AVX2-ONLY:       # %bb.0:
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; AVX2-ONLY-NEXT:    vmovdqa 112(%rdi), %xmm2
; AVX2-ONLY-NEXT:    vmovdqa 96(%rdi), %xmm3
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm4, %xmm0
; AVX2-ONLY-NEXT:    vmovdqa 80(%rdi), %xmm5
; AVX2-ONLY-NEXT:    vmovdqa 64(%rdi), %xmm6
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm7 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm7, %xmm1
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT:    vmovdqa (%rdi), %xmm8
; AVX2-ONLY-NEXT:    vmovdqa 16(%rdi), %xmm9
; AVX2-ONLY-NEXT:    vmovdqa 32(%rdi), %xmm10
; AVX2-ONLY-NEXT:    vmovdqa 48(%rdi), %xmm11
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm12 = xmm10[0],xmm11[0],xmm10[1],xmm11[1],xmm10[2],xmm11[2],xmm10[3],xmm11[3]
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm13 = xmm8[0],xmm9[0],xmm8[1],xmm9[1],xmm8[2],xmm9[2],xmm8[3],xmm9[3]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm14 = xmm13[1,1,1,1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm14 = xmm14[0],xmm12[1],xmm14[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm14[0,1],xmm1[2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm14 = xmm4[2,2,2,2]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm14 = xmm7[0,1,2],xmm14[3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm15 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm4 = xmm7[2],xmm4[2],xmm7[3],xmm4[3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm7 = xmm12[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm12 = xmm13[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm7 = xmm12[0],xmm7[1],xmm12[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
; AVX2-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm2, %xmm3
; AVX2-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm5 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm5, %xmm6
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
; AVX2-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm6 = xmm10[4],xmm11[4],xmm10[5],xmm11[5],xmm10[6],xmm11[6],xmm10[7],xmm11[7]
; AVX2-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm7 = xmm8[4],xmm9[4],xmm8[5],xmm9[5],xmm8[6],xmm9[6],xmm8[7],xmm9[7]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm8 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm8[0,1],xmm3[2,3]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm8 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm9 = xmm7[1,1,1,1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm9 = xmm9[0],xmm6[1],xmm9[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm9 = xmm2[2,2,2,2]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm9 = xmm5[0,1,2],xmm9[3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm10 = xmm7[2],xmm6[2],xmm7[3],xmm6[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm5 = xmm6[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm6 = xmm7[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, (%rdx)
; AVX2-ONLY-NEXT:    vmovdqa %xmm14, (%rcx)
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, (%r8)
; AVX2-ONLY-NEXT:    vmovdqa %xmm3, (%r9)
; AVX2-ONLY-NEXT:    vmovdqa %xmm8, (%r11)
; AVX2-ONLY-NEXT:    vmovdqa %xmm9, (%r10)
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, (%rax)
; AVX2-ONLY-NEXT:    retq
;
; AVX512F-SLOW-LABEL: load_i16_stride8_vf8:
; AVX512F-SLOW:       # %bb.0:
; AVX512F-SLOW-NEXT:    vmovdqa 112(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa 96(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT:    vmovdqa 80(%rdi), %xmm3
; AVX512F-SLOW-NEXT:    vmovdqa 64(%rdi), %xmm4
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm6 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; AVX512F-SLOW-NEXT:    vpbroadcastq {{.*#+}} xmm0 = [0,4,0,4]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm7 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm8 = xmm5[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm8 = xmm6[0,1,2],xmm8[3]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm9 = xmm6[2],xmm5[2],xmm6[3],xmm5[3]
; AVX512F-SLOW-NEXT:    vpermt2d %xmm5, %xmm0, %xmm6
; AVX512F-SLOW-NEXT:    vmovdqa (%rdi), %xmm5
; AVX512F-SLOW-NEXT:    vmovdqa 16(%rdi), %xmm10
; AVX512F-SLOW-NEXT:    vmovdqa 32(%rdi), %xmm11
; AVX512F-SLOW-NEXT:    vmovdqa 48(%rdi), %xmm12
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm13 = xmm11[0],xmm12[0],xmm11[1],xmm12[1],xmm11[2],xmm12[2],xmm11[3],xmm12[3]
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm14 = xmm5[0],xmm10[0],xmm5[1],xmm10[1],xmm5[2],xmm10[2],xmm5[3],xmm10[3]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm15 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm6 = xmm15[0,1],xmm6[2,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm15 = xmm14[1,1,1,1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm15 = xmm15[0],xmm13[1],xmm15[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm7 = xmm15[0,1],xmm7[2,3]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm15 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm8 = xmm15[0,1],xmm8[2,3]
; AVX512F-SLOW-NEXT:    vpbroadcastq {{.*#+}} xmm15 = [3,7,3,7]
; AVX512F-SLOW-NEXT:    vpermt2d %xmm13, %xmm15, %xmm14
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm9 = xmm14[0,1],xmm9[2,3]
; AVX512F-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
; AVX512F-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
; AVX512F-SLOW-NEXT:    vpermi2d %xmm1, %xmm2, %xmm0
; AVX512F-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm3 = xmm11[4],xmm12[4],xmm11[5],xmm12[5],xmm11[6],xmm12[6],xmm11[7],xmm12[7]
; AVX512F-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm5[4],xmm10[4],xmm5[5],xmm10[5],xmm5[6],xmm10[6],xmm5[7],xmm10[7]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm5[0,1],xmm0[2,3]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm10 = xmm4[1,1,1,1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm10 = xmm10[0],xmm3[1],xmm10[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm10 = xmm1[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm10 = xmm2[0,1,2],xmm10[3]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm11 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm10 = xmm11[0,1],xmm10[2,3]
; AVX512F-SLOW-NEXT:    vpermt2d %xmm3, %xmm15, %xmm4
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm4[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm6, (%rsi)
; AVX512F-SLOW-NEXT:    vmovdqa %xmm7, (%rdx)
; AVX512F-SLOW-NEXT:    vmovdqa %xmm8, (%rcx)
; AVX512F-SLOW-NEXT:    vmovdqa %xmm9, (%r8)
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, (%r9)
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT:    vmovdqa %xmm5, (%rax)
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT:    vmovdqa %xmm10, (%rax)
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, (%rax)
; AVX512F-SLOW-NEXT:    retq
;
; AVX512F-FAST-LABEL: load_i16_stride8_vf8:
; AVX512F-FAST:       # %bb.0:
; AVX512F-FAST-NEXT:    vmovdqa 112(%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa 96(%rdi), %xmm4
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm10 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm0, %xmm17
; AVX512F-FAST-NEXT:    vmovdqa 80(%rdi), %xmm5
; AVX512F-FAST-NEXT:    vmovdqa 64(%rdi), %xmm6
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm11 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm3 = [0,4,0,4]
; AVX512F-FAST-NEXT:    vmovdqa %xmm11, %xmm0
; AVX512F-FAST-NEXT:    vpermt2d %xmm10, %xmm3, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa (%rdi), %xmm7
; AVX512F-FAST-NEXT:    vmovdqa 16(%rdi), %xmm8
; AVX512F-FAST-NEXT:    vmovdqa 32(%rdi), %xmm12
; AVX512F-FAST-NEXT:    vmovdqa 48(%rdi), %xmm13
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm14 = xmm12[0],xmm13[0],xmm12[1],xmm13[1],xmm12[2],xmm13[2],xmm12[3],xmm13[3]
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm15 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm0, %xmm16
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm9 = [1,5,1,5]
; AVX512F-FAST-NEXT:    vmovdqa %xmm15, %xmm2
; AVX512F-FAST-NEXT:    vpermt2d %xmm14, %xmm9, %xmm2
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vmovdqa %xmm11, %xmm0
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm11 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [2,6,2,6]
; AVX512F-FAST-NEXT:    vpermt2d %xmm10, %xmm1, %xmm0
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm10 = xmm15[2],xmm14[2],xmm15[3],xmm14[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm10[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm10 = [3,7,3,7]
; AVX512F-FAST-NEXT:    vpermt2d %xmm14, %xmm10, %xmm15
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm11 = xmm15[0,1],xmm11[2,3]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm17, %xmm14
; AVX512F-FAST-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm14[4],xmm4[5],xmm14[5],xmm4[6],xmm14[6],xmm4[7],xmm14[7]
; AVX512F-FAST-NEXT:    vpunpckhwd {{.*#+}} xmm5 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
; AVX512F-FAST-NEXT:    vpunpckhwd {{.*#+}} xmm6 = xmm12[4],xmm13[4],xmm12[5],xmm13[5],xmm12[6],xmm13[6],xmm12[7],xmm13[7]
; AVX512F-FAST-NEXT:    vpermi2d %xmm4, %xmm5, %xmm3
; AVX512F-FAST-NEXT:    vpunpckhwd {{.*#+}} xmm7 = xmm7[4],xmm8[4],xmm7[5],xmm8[5],xmm7[6],xmm8[6],xmm7[7],xmm8[7]
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm8 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm3 = xmm8[0,1],xmm3[2,3]
; AVX512F-FAST-NEXT:    vpermi2d %xmm6, %xmm7, %xmm9
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm8 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
; AVX512F-FAST-NEXT:    vpermi2d %xmm4, %xmm5, %xmm1
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm9 = xmm7[2],xmm6[2],xmm7[3],xmm6[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm9[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vpermt2d %xmm6, %xmm10, %xmm7
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm4 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm16, (%rsi)
; AVX512F-FAST-NEXT:    vmovdqa %xmm2, (%rdx)
; AVX512F-FAST-NEXT:    vmovdqa %xmm0, (%rcx)
; AVX512F-FAST-NEXT:    vmovdqa %xmm11, (%r8)
; AVX512F-FAST-NEXT:    vmovdqa %xmm3, (%r9)
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT:    vmovdqa %xmm8, (%rax)
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, (%rax)
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT:    vmovdqa %xmm4, (%rax)
; AVX512F-FAST-NEXT:    retq
;
; AVX512BW-LABEL: load_i16_stride8_vf8:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm0 = [0,8,16,24,32,40,48,56]
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm2
; AVX512BW-NEXT:    vpermi2w %zmm2, %zmm1, %zmm0
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm3 = [1,9,17,25,33,41,49,57]
; AVX512BW-NEXT:    vpermi2w %zmm2, %zmm1, %zmm3
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm4 = [2,10,18,26,34,42,50,58]
; AVX512BW-NEXT:    vpermi2w %zmm2, %zmm1, %zmm4
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm5 = [3,11,19,27,35,43,51,59]
; AVX512BW-NEXT:    vpermi2w %zmm2, %zmm1, %zmm5
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm6 = [4,12,20,28,36,44,52,60]
; AVX512BW-NEXT:    vpermi2w %zmm2, %zmm1, %zmm6
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm7 = [5,13,21,29,37,45,53,61]
; AVX512BW-NEXT:    vpermi2w %zmm2, %zmm1, %zmm7
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm8 = [6,14,22,30,38,46,54,62]
; AVX512BW-NEXT:    vpermi2w %zmm2, %zmm1, %zmm8
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm9 = [7,15,23,31,39,47,55,63]
; AVX512BW-NEXT:    vpermi2w %zmm2, %zmm1, %zmm9
; AVX512BW-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX512BW-NEXT:    vmovdqa %xmm3, (%rdx)
; AVX512BW-NEXT:    vmovdqa %xmm4, (%rcx)
; AVX512BW-NEXT:    vmovdqa %xmm5, (%r8)
; AVX512BW-NEXT:    vmovdqa %xmm6, (%r9)
; AVX512BW-NEXT:    vmovdqa %xmm7, (%r11)
; AVX512BW-NEXT:    vmovdqa %xmm8, (%r10)
; AVX512BW-NEXT:    vmovdqa %xmm9, (%rax)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
  %wide.vec = load <64 x i16>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56>
  %strided.vec1 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57>
  %strided.vec2 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58>
  %strided.vec3 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 3, i32 11, i32 19, i32 27, i32 35, i32 43, i32 51, i32 59>
  %strided.vec4 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 4, i32 12, i32 20, i32 28, i32 36, i32 44, i32 52, i32 60>
  %strided.vec5 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61>
  %strided.vec6 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62>
  %strided.vec7 = shufflevector <64 x i16> %wide.vec, <64 x i16> poison, <8 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63>
  store <8 x i16> %strided.vec0, ptr %out.vec0, align 64
  store <8 x i16> %strided.vec1, ptr %out.vec1, align 64
  store <8 x i16> %strided.vec2, ptr %out.vec2, align 64
  store <8 x i16> %strided.vec3, ptr %out.vec3, align 64
  store <8 x i16> %strided.vec4, ptr %out.vec4, align 64
  store <8 x i16> %strided.vec5, ptr %out.vec5, align 64
  store <8 x i16> %strided.vec6, ptr %out.vec6, align 64
  store <8 x i16> %strided.vec7, ptr %out.vec7, align 64
  ret void
}

define void @load_i16_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
; SSE-LABEL: load_i16_stride8_vf16:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $168, %rsp
; SSE-NEXT:    movdqa 112(%rdi), %xmm6
; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 96(%rdi), %xmm14
; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 208(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; SSE-NEXT:    movdqa 192(%rdi), %xmm5
; SSE-NEXT:    movdqa 240(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 224(%rdi), %xmm12
; SSE-NEXT:    movdqa 144(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 128(%rdi), %xmm10
; SSE-NEXT:    movdqa 176(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 160(%rdi), %xmm11
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm2[0],xmm11[1],xmm2[1],xmm11[2],xmm2[2],xmm11[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm10, %xmm2
; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; SSE-NEXT:    movdqa %xmm12, %xmm4
; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,0,0,0]
; SSE-NEXT:    movdqa %xmm5, %xmm1
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
; SSE-NEXT:    movdqa %xmm2, %xmm3
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm11[0],xmm3[1],xmm11[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm6[0],xmm14[1],xmm6[1],xmm14[2],xmm6[2],xmm14[3],xmm6[3]
; SSE-NEXT:    movdqa 80(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 64(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, %xmm7
; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm14[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm7[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm6[2],xmm15[3],xmm6[3]
; SSE-NEXT:    movdqa 32(%rdi), %xmm9
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 48(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
; SSE-NEXT:    movdqa (%rdi), %xmm6
; SSE-NEXT:    movdqa 16(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm6, %xmm8
; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm8, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm15[2,3]
; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm11[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
; SSE-NEXT:    movdqa %xmm1, %xmm15
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm4[0],xmm15[1],xmm4[1]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm15[2,3]
; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm9[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
; SSE-NEXT:    movdqa %xmm7, %xmm15
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm15[2,3]
; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm4[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
; SSE-NEXT:    movdqa %xmm2, %xmm15
; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm11[2],xmm15[3],xmm11[3]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm15[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm14[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm8[3,3,3,3]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm0[2,3]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm14[2],xmm7[3],xmm14[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm7[2,3]
; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
; SSE-NEXT:    # xmm13 = xmm13[4],mem[4],xmm13[5],mem[5],xmm13[6],mem[6],xmm13[7],mem[7]
; SSE-NEXT:    movdqa %xmm10, %xmm15
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; SSE-NEXT:    # xmm15 = xmm15[4],mem[4],xmm15[5],mem[5],xmm15[6],mem[6],xmm15[7],mem[7]
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; SSE-NEXT:    # xmm12 = xmm12[4],mem[4],xmm12[5],mem[5],xmm12[6],mem[6],xmm12[7],mem[7]
; SSE-NEXT:    movdqa %xmm5, %xmm2
; SSE-NEXT:    punpckhwd (%rsp), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[0,0,0,0]
; SSE-NEXT:    movdqa %xmm12, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm2[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm1[2],xmm12[3],xmm1[3]
; SSE-NEXT:    movdqa %xmm15, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1]
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm1[0],xmm12[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; SSE-NEXT:    # xmm10 = xmm10[4],mem[4],xmm10[5],mem[5],xmm10[6],mem[6],xmm10[7],mem[7]
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,0,0,0]
; SSE-NEXT:    movdqa %xmm10, %xmm14
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; SSE-NEXT:    # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT:    # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
; SSE-NEXT:    movdqa %xmm6, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm15[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm13[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
; SSE-NEXT:    movdqa %xmm2, %xmm7
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm7[2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm5[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1]
; SSE-NEXT:    movdqa %xmm3, %xmm9
; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm0[2,2,2,2]
; SSE-NEXT:    movdqa %xmm0, %xmm10
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm2[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm11[2],xmm9[3],xmm11[3]
; SSE-NEXT:    movdqa %xmm15, %xmm11
; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm13[2],xmm11[3],xmm13[3]
; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm11[0],xmm9[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm14[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm11[2],xmm0[3],xmm11[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm6[3,3,3,3]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm5[2],xmm6[3],xmm5[3]
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[2,3]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm10[2],xmm2[3],xmm10[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm13[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm5[0],xmm14[1],xmm5[1]
; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm3[2,3]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    movaps %xmm2, (%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 16(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    movaps %xmm2, (%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    movaps %xmm2, 16(%rdx)
; SSE-NEXT:    movaps %xmm8, (%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    movaps %xmm2, 16(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    movaps %xmm2, (%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    movaps %xmm2, 16(%r8)
; SSE-NEXT:    movaps %xmm1, (%r9)
; SSE-NEXT:    movapd %xmm12, 16(%r9)
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movaps %xmm7, (%rax)
; SSE-NEXT:    movaps %xmm4, 16(%rax)
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movaps %xmm6, (%rax)
; SSE-NEXT:    movapd %xmm9, 16(%rax)
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movaps %xmm0, 16(%rax)
; SSE-NEXT:    movaps %xmm14, (%rax)
; SSE-NEXT:    addq $168, %rsp
; SSE-NEXT:    retq
;
; AVX1-ONLY-LABEL: load_i16_stride8_vf16:
; AVX1-ONLY:       # %bb.0:
; AVX1-ONLY-NEXT:    subq $152, %rsp
; AVX1-ONLY-NEXT:    vmovdqa 176(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 160(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm8 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vmovdqa 144(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 128(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vmovdqa 240(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 224(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm5[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa 208(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 192(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm3[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5],xmm1[6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vmovdqa 112(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 96(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm4[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa 80(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 64(%rdi), %xmm15
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm15[0],xmm1[0],xmm15[1],xmm1[1],xmm15[2],xmm1[2],xmm15[3],xmm1[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm6 = xmm1[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vmovdqa (%rdi), %xmm14
; AVX1-ONLY-NEXT:    vmovdqa 16(%rdi), %xmm12
; AVX1-ONLY-NEXT:    vmovdqa 32(%rdi), %xmm11
; AVX1-ONLY-NEXT:    vmovdqa 48(%rdi), %xmm10
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm14[0],xmm12[0],xmm14[1],xmm12[1],xmm14[2],xmm12[2],xmm14[3],xmm12[3]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm13 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm6 = xmm13[0,1,2,3],xmm6[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm9[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm6 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm9 = xmm7[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm9 = xmm9[0,1],xmm8[2,3],xmm9[4,5,6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3,4,5],ymm6[6,7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm9 = xmm1[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm9 = xmm9[0,1],xmm0[2,3],xmm9[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm13 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm9 = xmm9[0,1,2,3],xmm13[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm6 = xmm7[2],xmm8[2],xmm7[3],xmm8[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm9 = xmm5[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm9 = xmm3[0,1,2,3,4,5],xmm9[6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm9[6,7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm9 = xmm4[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm9 = xmm2[0,1,2,3,4,5],xmm9[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm13 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm9 = xmm13[0,1,2,3],xmm9[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm9[0,1,2,3],ymm6[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm5 = xmm8[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm6 = xmm7[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3],xmm6[4,5,6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm2[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa (%rsp), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[0,0,0,0]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm5 = xmm3[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm4 = xmm5[0,1,2,3,4,5],xmm4[6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm4[6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm9 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm9 = xmm15[4],mem[4],xmm15[5],mem[5],xmm15[6],mem[6],xmm15[7],mem[7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm6 = xmm4[0,0,0,0]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm7 = xmm9[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm8 = xmm7[0,1,2,3,4,5],xmm6[6,7]
; AVX1-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm6 = xmm11[4],xmm10[4],xmm11[5],xmm10[5],xmm11[6],xmm10[6],xmm11[7],xmm10[7]
; AVX1-ONLY-NEXT:    vpunpckhwd {{.*#+}} xmm7 = xmm14[4],xmm12[4],xmm14[5],xmm12[5],xmm14[6],xmm12[6],xmm14[7],xmm12[7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm10 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm8 = xmm10[0,1,2,3],xmm8[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1,2,3],ymm5[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm8 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm10 = xmm1[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm10 = xmm10[0,1],xmm0[2,3],xmm10[4,5,6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1,2,3,4,5],ymm8[6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm10 = xmm9[0],xmm4[0],xmm9[1],xmm4[1]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm11 = xmm7[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm11 = xmm11[0,1],xmm6[2,3],xmm11[4,5,6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm10 = xmm11[0,1,2,3],xmm10[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1,2,3],ymm8[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm10 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm11 = xmm2[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm11 = xmm3[0,1,2,3,4,5],xmm11[6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm11
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm11 = xmm4[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm11 = xmm9[0,1,2,3,4,5],xmm11[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm12 = xmm7[2],xmm6[2],xmm7[3],xmm6[3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm11 = xmm12[0,1,2,3],xmm11[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm9[2],xmm4[2],xmm9[3],xmm4[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm6[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm7[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%rsi)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%rdx)
; AVX1-ONLY-NEXT:    vmovaps %ymm13, (%rcx)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%r8)
; AVX1-ONLY-NEXT:    vmovaps %ymm5, (%r9)
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT:    vmovaps %ymm8, (%rax)
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT:    vmovaps %ymm10, (%rax)
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT:    vmovaps %ymm0, (%rax)
; AVX1-ONLY-NEXT:    addq $152, %rsp
; AVX1-ONLY-NEXT:    vzeroupper
; AVX1-ONLY-NEXT:    retq
;
; AVX2-ONLY-LABEL: load_i16_stride8_vf16:
; AVX2-ONLY:       # %bb.0:
; AVX2-ONLY-NEXT:    subq $264, %rsp # imm = 0x108
; AVX2-ONLY-NEXT:    vmovdqa 192(%rdi), %ymm2
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 224(%rdi), %ymm5
; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 112(%rdi), %xmm0
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 96(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm12 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm12, %xmm0
; AVX2-ONLY-NEXT:    vmovdqa 80(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 64(%rdi), %xmm3
; AVX2-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm9 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm9, %xmm1
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 32(%rdi), %xmm4
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 48(%rdi), %xmm6
; AVX2-ONLY-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm8 = xmm4[0],xmm6[0],xmm4[1],xmm6[1],xmm4[2],xmm6[2],xmm4[3],xmm6[3]
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm7 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm14 = xmm4[0,1],xmm3[2,3]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm13 = ymm5[0,1,0,2]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm6 = ymm13[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm6[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm11 = ymm2[0,1,0,2]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm5 = ymm11[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm5[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT:    vmovdqa 128(%rdi), %ymm0
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 160(%rdi), %ymm1
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, (%rsp) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm10 = ymm1[0,1,0,2]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm1 = ymm10[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm1[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[0,1,0,2]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm0 = ymm4[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm0[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm15[0,1,2,3,4],ymm2[5],ymm15[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm14[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm7[1,1,1,1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0],xmm8[1],xmm2[2,3]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm9[0],xmm12[0],xmm9[1],xmm12[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm6[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm5[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm12[2,2,2,2]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm9[0,1,2],xmm0[3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm7[2],xmm8[2],xmm7[3],xmm8[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm1 = ymm13[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm2 = ymm11[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm2[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm5 = ymm10[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm4 = ymm4[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm5[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm10 = ymm4[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm10[0,1,2,3,4],ymm6[5],ymm10[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm9[2],xmm12[2],xmm9[3],xmm12[3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm8[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm6 = xmm7[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm6[0],xmm3[1],xmm6[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm5[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm4[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm0, %xmm1
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm5 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm5 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm5, %xmm2
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm6 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm4 = ymm6[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm8 = ymm4[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm7 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm10 = ymm7[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm9 = ymm10[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm11 = ymm9[0,1,2,3,4,5,6],ymm8[7]
; AVX2-ONLY-NEXT:    vpermq $212, (%rsp), %ymm8 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm8 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm9 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm13 = ymm8[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm14 = ymm13[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm15 = ymm9[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm15[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm14[5],ymm12[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm11[4,5,6,7]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm11 = xmm2[1,1,1,1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm11 = xmm11[0],xmm1[1],xmm11[2,3]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm12 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm11 = xmm11[0,1],xmm12[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm10 = ymm10[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm10[0,1,2,3,4,5,6],ymm4[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm10 = ymm13[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm15[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3,4],ymm10[5],ymm12[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm10[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm11[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm10 = xmm0[2,2,2,2]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm10 = xmm5[0,1,2],xmm10[3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm11 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm10 = xmm11[0,1],xmm10[2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm6 = ymm6[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm11 = ymm6[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm7 = ymm7[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm12 = ymm7[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5,6],ymm11[7]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm8 = ymm8[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm9 = ymm9[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm8[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm9[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3,4],ymm12[5],ymm13[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm7[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm8[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm9[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3,4],ymm2[5],ymm5[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rsi)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rdx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rcx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%r8)
; AVX2-ONLY-NEXT:    vmovdqa %ymm3, (%r9)
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT:    vmovdqa %ymm4, (%rax)
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT:    vmovdqa %ymm10, (%rax)
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT:    vmovdqa %ymm0, (%rax)
; AVX2-ONLY-NEXT:    addq $264, %rsp # imm = 0x108
; AVX2-ONLY-NEXT:    vzeroupper
; AVX2-ONLY-NEXT:    retq
;
; AVX512F-SLOW-LABEL: load_i16_stride8_vf16:
; AVX512F-SLOW:       # %bb.0:
; AVX512F-SLOW-NEXT:    vmovdqa 112(%rdi), %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa 96(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm1, %xmm24
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm0, %xmm25
; AVX512F-SLOW-NEXT:    vmovdqa 80(%rdi), %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa 64(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm6 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm1, %xmm26
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm0, %xmm27
; AVX512F-SLOW-NEXT:    vpbroadcastq {{.*#+}} xmm7 = [0,4,0,4]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm29 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm5[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm6[0,1,2],xmm0[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm0, %xmm30
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm11 = xmm6[2],xmm5[2],xmm6[3],xmm5[3]
; AVX512F-SLOW-NEXT:    vpermt2d %xmm5, %xmm7, %xmm6
; AVX512F-SLOW-NEXT:    vmovdqa (%rdi), %xmm9
; AVX512F-SLOW-NEXT:    vmovdqa 16(%rdi), %xmm10
; AVX512F-SLOW-NEXT:    vmovdqa 32(%rdi), %xmm12
; AVX512F-SLOW-NEXT:    vmovdqa 48(%rdi), %xmm13
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm12[0],xmm13[0],xmm12[1],xmm13[1],xmm12[2],xmm13[2],xmm12[3],xmm13[3]
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm14 = xmm9[0],xmm10[0],xmm9[1],xmm10[1],xmm9[2],xmm10[2],xmm9[3],xmm10[3]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm14[0],xmm4[0],xmm14[1],xmm4[1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa64 224(%rdi), %ymm18
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm17 = ymm18[0,1,0,2]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm6 = ymm17[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vmovdqa64 192(%rdi), %ymm20
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm19 = ymm20[0,1,0,2]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm15 = ymm6[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm19[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm1[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm15[7]
; AVX512F-SLOW-NEXT:    vmovdqa64 160(%rdi), %ymm23
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm21 = ymm23[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqa64 128(%rdi), %ymm16
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm22 = ymm16[0,1,0,2]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm15 = ymm21[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm15[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm22[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm8 = ymm0[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3,4],ymm2[5],ymm8[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm2, %ymm28
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm15[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm14[1,1,1,1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm29, %xmm2
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm0, %ymm29
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm17[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm19[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm2[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm21[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm3[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm8 = ymm22[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm8[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm15[0,1,2,3,4],ymm5[5],ymm15[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm14[2],xmm4[2],xmm14[3],xmm4[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm30, %xmm6
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm1, %ymm19
; AVX512F-SLOW-NEXT:    vpbroadcastq {{.*#+}} xmm17 = [3,7,3,7]
; AVX512F-SLOW-NEXT:    vpermt2d %xmm4, %xmm17, %xmm14
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm14[0,1],xmm11[2,3]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm3[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm8[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm8 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm24, %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm25, %xmm1
; AVX512F-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm26, %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm27, %xmm2
; AVX512F-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
; AVX512F-SLOW-NEXT:    vpermi2d %xmm1, %xmm2, %xmm7
; AVX512F-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm12[4],xmm13[4],xmm12[5],xmm13[5],xmm12[6],xmm13[6],xmm12[7],xmm13[7]
; AVX512F-SLOW-NEXT:    vpunpckhwd {{.*#+}} xmm3 = xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm0[0,1],xmm7[2,3]
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm9 = ymm18[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm7 = ymm9[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm10 = ymm7[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm11 = ymm20[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm12 = ymm11[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm13 = ymm12[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm10 = ymm13[0,1,2,3,4,5,6],ymm10[7]
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm13 = ymm23[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm14 = ymm13[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm14[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm16 = ymm16[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm16[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm0[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm15[5],ymm6[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm10[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm10 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm7[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm6 = ymm12[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5,6],ymm5[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm14[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm6[5],ymm0[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm6 = xmm3[1,1,1,1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm7 = ymm5[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm9[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm0[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm6 = ymm11[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm9 = ymm6[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3,4,5,6],ymm5[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm9 = ymm13[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm11 = ymm9[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm12 = ymm16[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm12[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm11 = ymm13[0,1,2,3,4],ymm11[5],ymm13[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm11[0,1,2,3,4,5],ymm5[6,7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm11 = xmm1[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm11 = xmm2[0,1,2],xmm11[3]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm13 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm11 = xmm13[0,1],xmm11[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm11[0,1,2,3],ymm5[4,5,6,7]
; AVX512F-SLOW-NEXT:    vpermt2d %xmm4, %xmm17, %xmm3
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm6[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm9[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm12[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm28, (%rsi)
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm29, (%rdx)
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm19, (%rcx)
; AVX512F-SLOW-NEXT:    vmovdqa %ymm8, (%r8)
; AVX512F-SLOW-NEXT:    vmovdqa %ymm10, (%r9)
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT:    vmovdqa %ymm7, (%rax)
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT:    vmovdqa %ymm5, (%rax)
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT:    vmovdqa %ymm0, (%rax)
; AVX512F-SLOW-NEXT:    vzeroupper
; AVX512F-SLOW-NEXT:    retq
;
; AVX512F-FAST-LABEL: load_i16_stride8_vf16:
; AVX512F-FAST:       # %bb.0:
; AVX512F-FAST-NEXT:    vmovdqa 112(%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa 96(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm1, %xmm26
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm0, %xmm27
; AVX512F-FAST-NEXT:    vmovdqa 80(%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa 64(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm12 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm1, %xmm28
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm0, %xmm29
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm7 = [0,4,0,4]
; AVX512F-FAST-NEXT:    vmovdqa %xmm12, %xmm13
; AVX512F-FAST-NEXT:    vpermt2d %xmm4, %xmm7, %xmm13
; AVX512F-FAST-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa 32(%rdi), %xmm9
; AVX512F-FAST-NEXT:    vmovdqa 48(%rdi), %xmm10
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm9[0],xmm10[0],xmm9[1],xmm10[1],xmm9[2],xmm10[2],xmm9[3],xmm10[3]
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm11 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm1, %xmm30
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm0, %xmm31
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm14 = xmm11[0],xmm5[0],xmm11[1],xmm5[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm15 = xmm14[0,1],xmm13[2,3]
; AVX512F-FAST-NEXT:    vmovdqa64 224(%rdi), %ymm16
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm18 = ymm16[0,1,0,2]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm13 = ymm18[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm14 = ymm13[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vmovdqa64 192(%rdi), %ymm17
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm20 = ymm17[0,1,0,2]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm2 = ymm20[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm2[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm14[7]
; AVX512F-FAST-NEXT:    vmovdqa64 160(%rdi), %ymm19
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm21 = ymm19[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqa64 128(%rdi), %ymm23
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm22 = ymm23[0,1,0,2]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm14 = ymm21[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm14[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm0 = ymm22[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm0[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm6[0,1,2,3,4],ymm1[5],ymm6[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm15[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm1, %ymm24
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm15 = [1,5,1,5]
; AVX512F-FAST-NEXT:    vmovdqa %xmm11, %xmm1
; AVX512F-FAST-NEXT:    vpermt2d %xmm5, %xmm15, %xmm1
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm12[0],xmm4[0],xmm12[1],xmm4[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm13[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm14[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5],ymm0[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm0, %ymm25
; AVX512F-FAST-NEXT:    vmovdqa %xmm12, %xmm0
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm12[2],xmm4[2],xmm12[3],xmm4[3]
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm12 = [2,6,2,6]
; AVX512F-FAST-NEXT:    vpermt2d %xmm4, %xmm12, %xmm0
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm11[2],xmm5[2],xmm11[3],xmm5[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm3 = ymm18[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm4 = ymm20[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm6 = ymm4[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm6 = ymm21[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm6[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm14 = ymm22[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm8 = ymm14[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm13[5],ymm8[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm0, %ymm21
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm18 = [3,7,3,7]
; AVX512F-FAST-NEXT:    vpermt2d %xmm5, %xmm18, %xmm11
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm11[0,1],xmm2[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm4[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm6[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm14[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5],ymm4[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm0, %ymm22
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm26, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm27, %xmm1
; AVX512F-FAST-NEXT:    vpunpckhwd {{.*#+}} xmm3 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm28, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm29, %xmm1
; AVX512F-FAST-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; AVX512F-FAST-NEXT:    vpermi2d %xmm3, %xmm4, %xmm7
; AVX512F-FAST-NEXT:    vpunpckhwd {{.*#+}} xmm6 = xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm30, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm31, %xmm1
; AVX512F-FAST-NEXT:    vpunpckhwd {{.*#+}} xmm5 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm7[2,3]
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm20 = ymm16[0,1,1,3]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm7 = ymm20[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm9 = ymm7[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm10 = ymm17[0,1,1,3]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm11 = ymm10[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm13 = ymm11[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm9 = ymm13[0,1,2,3,4,5,6],ymm9[7]
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm13 = ymm19[0,1,1,3]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm14 = ymm13[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm14[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm16 = ymm23[0,1,1,3]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm8 = ymm16[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm8[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm9[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-FAST-NEXT:    vpermi2d %xmm6, %xmm5, %xmm15
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm15[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm7 = ymm11[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm14[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm8 = ymm8[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm7 = ymm8[0,1,2,3,4],ymm7[5],ymm8[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm7 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-FAST-NEXT:    vpermi2d %xmm3, %xmm4, %xmm12
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm5[2],xmm6[2],xmm5[3],xmm6[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm12[2,3]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm1 = ymm20[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm8 = ymm1[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm9 = ymm10[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm10 = ymm9[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm8 = ymm10[0,1,2,3,4,5,6],ymm8[7]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm10 = ymm13[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm11 = ymm10[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm12 = ymm16[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm12[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm11 = ymm13[0,1,2,3,4],ymm11[5],ymm13[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm8 = ymm11[0,1,2,3,4,5],ymm8[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm8[4,5,6,7]
; AVX512F-FAST-NEXT:    vpermt2d %xmm6, %xmm18, %xmm5
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm9[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm10[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm12[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5],ymm5[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm24, (%rsi)
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm25, (%rdx)
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm21, (%rcx)
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm22, (%r8)
; AVX512F-FAST-NEXT:    vmovdqa %ymm2, (%r9)
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT:    vmovdqa %ymm7, (%rax)
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT:    vmovdqa %ymm0, (%rax)
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT:    vmovdqa %ymm1, (%rax)
; AVX512F-FAST-NEXT:    vzeroupper
; AVX512F-FAST-NEXT:    retq
;
; AVX512BW-LABEL: load_i16_stride8_vf16:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm4 = [0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56]
; AVX512BW-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm3, %zmm2, %zmm4
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm5 = [0,8,16,24,32,40,48,56]
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm5
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512BW-NEXT:    # ymm5 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm3, %zmm2, %zmm5
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm6 = [1,9,17,25,33,41,49,57]
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm6
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm6 = [2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58]
; AVX512BW-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm3, %zmm2, %zmm6
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm7 = [2,10,18,26,34,42,50,58]
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm7
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm7 = [3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59]
; AVX512BW-NEXT:    # ymm7 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm3, %zmm2, %zmm7
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm8 = [3,11,19,27,35,43,51,59]
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm8
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm8 = [4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60]
; AVX512BW-NEXT:    # ymm8 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm3, %zmm2, %zmm8
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm9 = [4,12,20,28,36,44,52,60]
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm9
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm9 = [5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61]
; AVX512BW-NEXT:    # ymm9 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm3, %zmm2, %zmm9
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm10 = [5,13,21,29,37,45,53,61]
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm10
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm9 = ymm10[0,1,2,3],ymm9[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm10 = [6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62]
; AVX512BW-NEXT:    # ymm10 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm3, %zmm2, %zmm10
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm11 = [6,14,22,30,38,46,54,62]
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm11
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm11 = [7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63]
; AVX512BW-NEXT:    # ymm11 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm3, %zmm2, %zmm11
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm2 = [7,15,23,31,39,47,55,63]
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm2
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm11[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa %ymm4, (%rsi)
; AVX512BW-NEXT:    vmovdqa %ymm5, (%rdx)
; AVX512BW-NEXT:    vmovdqa %ymm6, (%rcx)
; AVX512BW-NEXT:    vmovdqa %ymm7, (%r8)
; AVX512BW-NEXT:    vmovdqa %ymm8, (%r9)
; AVX512BW-NEXT:    vmovdqa %ymm9, (%r11)
; AVX512BW-NEXT:    vmovdqa %ymm10, (%r10)
; AVX512BW-NEXT:    vmovdqa %ymm0, (%rax)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
  %wide.vec = load <128 x i16>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120>
  %strided.vec1 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121>
  %strided.vec2 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122>
  %strided.vec3 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 3, i32 11, i32 19, i32 27, i32 35, i32 43, i32 51, i32 59, i32 67, i32 75, i32 83, i32 91, i32 99, i32 107, i32 115, i32 123>
  %strided.vec4 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 4, i32 12, i32 20, i32 28, i32 36, i32 44, i32 52, i32 60, i32 68, i32 76, i32 84, i32 92, i32 100, i32 108, i32 116, i32 124>
  %strided.vec5 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125>
  %strided.vec6 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126>
  %strided.vec7 = shufflevector <128 x i16> %wide.vec, <128 x i16> poison, <16 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127>
  store <16 x i16> %strided.vec0, ptr %out.vec0, align 64
  store <16 x i16> %strided.vec1, ptr %out.vec1, align 64
  store <16 x i16> %strided.vec2, ptr %out.vec2, align 64
  store <16 x i16> %strided.vec3, ptr %out.vec3, align 64
  store <16 x i16> %strided.vec4, ptr %out.vec4, align 64
  store <16 x i16> %strided.vec5, ptr %out.vec5, align 64
  store <16 x i16> %strided.vec6, ptr %out.vec6, align 64
  store <16 x i16> %strided.vec7, ptr %out.vec7, align 64
  ret void
}

define void @load_i16_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
; SSE-LABEL: load_i16_stride8_vf32:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $696, %rsp # imm = 0x2B8
; SSE-NEXT:    movdqa 496(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 480(%rdi), %xmm14
; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 208(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 192(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 240(%rdi), %xmm5
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 224(%rdi), %xmm15
; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 144(%rdi), %xmm6
; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 128(%rdi), %xmm4
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 176(%rdi), %xmm7
; SSE-NEXT:    movdqa %xmm7, (%rsp) # 16-byte Spill
; SSE-NEXT:    movdqa 160(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
; SSE-NEXT:    movdqa %xmm0, %xmm7
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1],xmm4[2],xmm6[2],xmm4[3],xmm6[3]
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm5[0],xmm15[1],xmm5[1],xmm15[2],xmm5[2],xmm15[3],xmm5[3]
; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,0,0,0]
; SSE-NEXT:    movdqa %xmm3, %xmm15
; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1],xmm15[2],xmm1[2],xmm15[3],xmm1[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm4, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm2[0],xmm14[1],xmm2[1],xmm14[2],xmm2[2],xmm14[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 464(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 448(%rdi), %xmm8
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa 432(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 416(%rdi), %xmm11
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1],xmm11[2],xmm0[2],xmm11[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 400(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 384(%rdi), %xmm10
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1],xmm10[2],xmm0[2],xmm10[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm10, %xmm0
; SSE-NEXT:    movdqa %xmm10, %xmm14
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 368(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 352(%rdi), %xmm9
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 336(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 320(%rdi), %xmm12
; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1],xmm12[2],xmm0[2],xmm12[3],xmm0[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa 304(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 288(%rdi), %xmm11
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1],xmm11[2],xmm0[2],xmm11[3],xmm0[3]
; SSE-NEXT:    movdqa 272(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 256(%rdi), %xmm10
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1],xmm10[2],xmm0[2],xmm10[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm10, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 112(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 96(%rdi), %xmm9
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
; SSE-NEXT:    movdqa 80(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 64(%rdi), %xmm13
; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1],xmm13[2],xmm0[2],xmm13[3],xmm0[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
; SSE-NEXT:    movdqa 32(%rdi), %xmm7
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 48(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3]
; SSE-NEXT:    movdqa (%rdi), %xmm6
; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 16(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1],xmm6[2],xmm0[2],xmm6[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm6, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa %xmm15, %xmm0
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[1,1,1,1]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa %xmm8, %xmm0
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa %xmm12, %xmm0
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa %xmm13, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm4, %xmm0
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movapd %xmm3, %xmm0
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm10, %xmm0
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm11[2],xmm0[3],xmm11[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE-NEXT:    movdqa %xmm6, %xmm1
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; SSE-NEXT:    # xmm15 = xmm15[2],mem[2],xmm15[3],mem[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm8[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm14[2],xmm12[3],xmm14[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm12[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm9[2],xmm13[3],xmm9[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    punpckhwd (%rsp), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; SSE-NEXT:    # xmm15 = xmm15[4],mem[4],xmm15[5],mem[5],xmm15[6],mem[6],xmm15[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
; SSE-NEXT:    movdqa %xmm1, %xmm12
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm15, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    movdqa %xmm3, %xmm7
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = xmm8[4],mem[4],xmm8[5],mem[5],xmm8[6],mem[6],xmm8[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,0,0,0]
; SSE-NEXT:    movdqa %xmm8, %xmm11
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm4, %xmm0
; SSE-NEXT:    movdqa %xmm4, %xmm14
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    movdqa %xmm1, %xmm13
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm0[0],xmm5[1]
; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; SSE-NEXT:    # xmm10 = xmm10[4],mem[4],xmm10[5],mem[5],xmm10[6],mem[6],xmm10[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; SSE-NEXT:    # xmm9 = xmm9[4],mem[4],xmm9[5],mem[5],xmm9[6],mem[6],xmm9[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm9, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm0[0],xmm5[1]
; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = xmm8[4],mem[4],xmm8[5],mem[5],xmm8[6],mem[6],xmm8[7],mem[7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,0,0,0]
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
; SSE-NEXT:    movdqa %xmm4, %xmm3
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2,3]
; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa %xmm2, %xmm0
; SSE-NEXT:    movdqa %xmm2, %xmm5
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm12, %xmm3
; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm1, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
; SSE-NEXT:    movdqa %xmm11, %xmm6
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm2, (%rsp) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm9, %xmm14
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[1,1,1,1]
; SSE-NEXT:    movdqa %xmm10, %xmm13
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    movaps %xmm7, %xmm0
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm0[2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm4[1,1,1,1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
; SSE-NEXT:    movdqa %xmm8, %xmm0
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm0[2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm5[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm0[2],xmm11[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm15, %xmm0
; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm0[0],xmm11[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm7[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm0[2],xmm12[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm14, %xmm0
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm13[2],xmm0[3],xmm13[3]
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm0[0],xmm12[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm1[2,2,2,2]
; SSE-NEXT:    movdqa %xmm1, %xmm3
; SSE-NEXT:    punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm0[2],xmm13[3],xmm0[3]
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    movapd %xmm6, %xmm0
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,2,2]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm4, %xmm14
; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm2[2],xmm14[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm2, %xmm8
; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm1[2,3]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[3,3,3,3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm1[2,3]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3]
; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm7[2,3]
; 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, 48(%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), %xmm8 # 16-byte Reload
; SSE-NEXT:    movaps %xmm8, 16(%rsi)
; 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, (%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 16(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 32(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 48(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, (%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 16(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 32(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 48(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, (%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 16(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 32(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 48(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, (%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 16(%r9)
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movaps %xmm9, 32(%rax)
; SSE-NEXT:    movaps (%rsp), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 48(%rax)
; SSE-NEXT:    movaps %xmm10, (%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 16(%rax)
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movapd %xmm13, 48(%rax)
; SSE-NEXT:    movapd %xmm12, 32(%rax)
; SSE-NEXT:    movapd %xmm11, 16(%rax)
; SSE-NEXT:    movaps %xmm14, (%rax)
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movaps %xmm1, 48(%rax)
; SSE-NEXT:    movaps %xmm2, 32(%rax)
; SSE-NEXT:    movaps %xmm15, 16(%rax)
; SSE-NEXT:    movaps %xmm0, (%rax)
; SSE-NEXT:    addq $696, %rsp # imm = 0x2B8
; SSE-NEXT:    retq
;
; AVX1-ONLY-LABEL: load_i16_stride8_vf32:
; AVX1-ONLY:       # %bb.0:
; AVX1-ONLY-NEXT:    subq $872, %rsp # imm = 0x368
; AVX1-ONLY-NEXT:    vmovdqa 304(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 288(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vmovdqa 272(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 256(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm8 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm8[0],xmm5[0],xmm8[1],xmm5[1]
; AVX1-ONLY-NEXT:    vmovdqa 368(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 352(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, %xmm6
; AVX1-ONLY-NEXT:    vmovdqa 336(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 320(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm14 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm14[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5],xmm1[6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa 496(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 480(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, %xmm7
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 464(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 448(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm10 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm10[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5],xmm1[6,7]
; AVX1-ONLY-NEXT:    vmovdqa 432(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 416(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm13 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vmovdqa 400(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 384(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm11 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm11[0],xmm13[0],xmm11[1],xmm13[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 240(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 224(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa 208(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 192(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vmovdqa 176(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 160(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 144(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 128(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT:    vmovdqa 112(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 96(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa 80(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 64(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vmovdqa (%rdi), %xmm15
; AVX1-ONLY-NEXT:    vmovdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 48(%rdi), %xmm4
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm12 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm15[0],xmm1[0],xmm15[1],xmm1[1],xmm15[2],xmm1[2],xmm15[3],xmm1[3]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm15 = xmm1[0],xmm12[0],xmm1[1],xmm12[1]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, %xmm9
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm15[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm6, %xmm4
; AVX1-ONLY-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm14[0],xmm6[0],xmm14[1],xmm6[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm10[0],xmm7[0],xmm10[1],xmm7[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm15 = xmm11[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm15 = xmm15[0,1],xmm13[2,3],xmm15[4,5,6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm15[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[1,1,1,1]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0],xmm1[1],xmm3[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm9[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm12[2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm15 = xmm15[0],xmm9[0],xmm15[1],xmm9[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm15[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm14[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm8[2],xmm5[2],xmm8[3],xmm5[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm5, %xmm4
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm3[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm11[2],xmm13[2],xmm11[3],xmm13[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm15 = xmm11[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm15 = xmm10[0,1,2,3,4,5],xmm15[6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm15[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm6[2,2,2,2]
; AVX1-ONLY-NEXT:    vmovaps %xmm6, %xmm5
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm7[0,1,2],xmm3[3]
; AVX1-ONLY-NEXT:    vmovaps %xmm7, %xmm6
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm9, %xmm7
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm9[2,2,2,2]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm8[0,1,2,3,4,5],xmm3[6,7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm12, %xmm9
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm15 = xmm12[2],xmm9[2],xmm12[3],xmm9[3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1,2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm14[2],mem[2],xmm14[3],mem[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm14 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm14[0,1],xmm3[2,3],xmm14[4,5,6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm3[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm10[2],xmm11[2],xmm10[3],xmm11[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm11 = xmm13[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm10 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm10 = xmm10[0,1],xmm11[2,3],xmm10[4,5,6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm6[2],xmm5[2],xmm6[3],xmm5[3]
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm2[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm6[0],xmm3[1],xmm6[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm9[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm12[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm6 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm10 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm4 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, %xmm8
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm4[0,1,0,1]
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, %xmm7
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm10[0],xmm6[0],xmm10[1],xmm6[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm4 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd (%rsp), %xmm1, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm2 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm5 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[0,0,0,0]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm5[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5],xmm1[6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm11 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm11 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm12 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm9 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm9 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm12[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm9[0,1,0,1]
; AVX1-ONLY-NEXT:    vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm11[0],xmm2[0],xmm11[1],xmm2[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm2[0,1,2,3,4,5],xmm1[6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm13 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm13 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm14 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm15 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm15[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
; AVX1-ONLY-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm8, %xmm5
; AVX1-ONLY-NEXT:    vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm3 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm4[1,1,1,1]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0],xmm1[1],xmm15[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm15[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm9[0],xmm12[0],xmm9[1],xmm12[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm11[1,1,1,1]
; AVX1-ONLY-NEXT:    vmovdqa %xmm11, %xmm9
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm11[2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm14[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm13[2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa (%rsp), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm15 = xmm12[0],xmm2[0],xmm12[1],xmm2[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm15[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm5[2,2,2,2]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm5[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm10[2],xmm6[2],xmm10[3],xmm6[3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm3[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vmovaps %xmm8, %xmm6
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm8[2,2,2,2]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm7[0,1,2],xmm15[3]
; AVX1-ONLY-NEXT:    vmovaps %xmm7, %xmm8
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm15[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa %xmm9, %xmm4
; AVX1-ONLY-NEXT:    vmovdqa %xmm11, %xmm1
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm9[2],xmm11[2],xmm9[3],xmm11[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm9[2,2,2,2]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm11[0,1,2],xmm3[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm12[0,1,2,3,4,5],xmm3[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm15 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1,2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm5[2],mem[2],xmm5[3],mem[3]
; AVX1-ONLY-NEXT:    vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = mem[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm7 = xmm10[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm7[0,1],xmm3[2,3],xmm7[4,5,6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm3[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm8[2],xmm6[2],xmm8[3],xmm6[3]
; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm5 = mem[2,3,2,3]
; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm6 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm11[2],xmm9[2],xmm11[3],xmm9[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm5 = xmm1[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm6 = xmm4[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3],xmm6[4,5,6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm4 = xmm12[2],xmm2[2],xmm12[3],xmm2[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm13[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm14[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm4[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%rsi)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm2, 32(%rsi)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%rdx)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm2, 32(%rdx)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%rcx)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm2, 32(%rcx)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%r8)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm2, 32(%r8)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%r9)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm2, 32(%r9)
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%rax)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm2, 32(%rax)
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT:    vmovaps %ymm15, (%rax)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm2, 32(%rax)
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT:    vmovaps %ymm1, (%rax)
; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%rax)
; AVX1-ONLY-NEXT:    addq $872, %rsp # imm = 0x368
; AVX1-ONLY-NEXT:    vzeroupper
; AVX1-ONLY-NEXT:    retq
;
; AVX2-ONLY-LABEL: load_i16_stride8_vf32:
; AVX2-ONLY:       # %bb.0:
; AVX2-ONLY-NEXT:    subq $1000, %rsp # imm = 0x3E8
; AVX2-ONLY-NEXT:    vmovdqa 448(%rdi), %ymm2
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 480(%rdi), %ymm3
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 272(%rdi), %xmm0
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 256(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 304(%rdi), %xmm4
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 288(%rdi), %xmm5
; AVX2-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm8 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX2-ONLY-NEXT:    vmovdqa 368(%rdi), %xmm0
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 352(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm1, %xmm0
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, %xmm9
; AVX2-ONLY-NEXT:    vmovdqa 336(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 320(%rdi), %xmm4
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm1, %xmm1
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm8[0],xmm5[0],xmm8[1],xmm5[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm3[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vmovdqa 384(%rdi), %ymm2
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 416(%rdi), %ymm3
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm13 = ymm3[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm13[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm7 = ymm2[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm7[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5],ymm5[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 112(%rdi), %xmm0
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 96(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm0, %xmm0
; AVX2-ONLY-NEXT:    vmovdqa 80(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 64(%rdi), %xmm2
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm1, %xmm1
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT:    vmovdqa (%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX2-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 48(%rdi), %xmm4
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm15 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm15[0],xmm2[1],xmm15[1]
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, %xmm10
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm6 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vmovdqa 192(%rdi), %ymm0
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 224(%rdi), %ymm1
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm12 = ymm1[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm4 = ymm0[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm14 = ymm4[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5,6],ymm11[7]
; AVX2-ONLY-NEXT:    vmovdqa 128(%rdi), %ymm2
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 160(%rdi), %ymm0
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm0[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm3[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm14 = ymm2[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3,4],ymm0[5],ymm14[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm5[1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm9, %xmm14
; AVX2-ONLY-NEXT:    vmovdqa %xmm9, (%rsp) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm9[0],xmm14[0],xmm9[1],xmm14[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX2-ONLY-NEXT:    vpshufhw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm6 = mem[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm13[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm7[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5],ymm7[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
; AVX2-ONLY-NEXT:    vmovdqa %xmm10, %xmm11
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm15[1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm10[0],xmm13[0],xmm10[1],xmm13[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm4[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5],ymm2[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm14[2,2,2,2]
; AVX2-ONLY-NEXT:    vmovdqa %xmm9, %xmm3
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm9[0,1,2],xmm0[3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm8[2],xmm5[2],xmm8[3],xmm5[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm8 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm8[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm1[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm2[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5],ymm7[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm4 = xmm13[2,2,2,2]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm10[0,1,2],xmm4[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm11, %xmm12
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm11[2],xmm15[2],xmm11[3],xmm15[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm7 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm9 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm7[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm6 = ymm9[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5,6],ymm5[7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm6 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm11 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm6[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm14 = ymm11[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5],ymm14[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm13[0,1,2,3,4,5],ymm5[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpunpckhdq (%rsp), %xmm3, %xmm4 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm4 = xmm3[2],mem[2],xmm3[3],mem[3]
; AVX2-ONLY-NEXT:    vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm5 = mem[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm13 = mem[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = xmm13[0],xmm5[1],xmm13[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm8[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm0[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5],ymm2[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm0 = xmm10[2],mem[2],xmm10[3],mem[3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm15[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm12[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm9[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm6[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm11[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm4 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm1, %xmm0
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm4, %xmm1
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm12 = ymm1[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm10 = ymm2[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm10[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm9 = ymm2[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm9[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm8 = ymm3[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm8[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5],ymm7[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm11 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm11 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm1, %xmm0
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm11, %xmm1
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm2 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm14 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm14 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm14[0],xmm2[0],xmm14[1],xmm2[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm5 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm4 = ymm0[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm15 = ymm4[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm1[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm13 = ymm3[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm15[7]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm7 = ymm0[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm7[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm1[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3,4],ymm0[5],ymm15[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm13[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm6[1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm2[0],xmm13[0],xmm2[1],xmm13[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm12[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm10 = ymm10[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm10[0,1,2,3,4,5,6],ymm5[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm9 = ymm9[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm8 = ymm8[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5],ymm8[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3,4,5],ymm5[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm12[1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm7[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm4[5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm13[2,2,2,2]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm15[2],xmm6[2],xmm15[3],xmm6[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm3[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm4 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm5 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm8 = ymm4[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm9 = ymm5[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3,4],ymm8[5],ymm9[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm10[2,2,2,2]
; AVX2-ONLY-NEXT:    vmovdqa %xmm11, %xmm7
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm11[0,1,2],xmm0[3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm14[2],xmm12[2],xmm14[3],xmm12[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm8 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm9 = ymm0[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm10 = ymm1[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm9 = ymm10[0,1,2,3,4,5,6],ymm9[7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm10 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm11 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm10[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm11[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4],ymm13[5],ymm15[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm9 = ymm13[0,1,2,3,4,5],ymm9[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm9 = xmm9[2],mem[2],xmm9[3],mem[3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm13 = xmm6[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm15 = mem[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm13 = xmm15[0],xmm13[1],xmm15[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm9 = xmm13[0,1],xmm9[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm4[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm5[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5],ymm4[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm9[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm3 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm3 = xmm7[2],mem[2],xmm7[3],mem[3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm4 = xmm12[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm5 = xmm14[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm10[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm11[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5],ymm4[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rsi)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rsi)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rdx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rdx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rcx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rcx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%r8)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%r8)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%r9)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%r9)
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, (%rax)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rax)
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT:    vmovdqa %ymm8, (%rax)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm1, 32(%rax)
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT:    vmovdqa %ymm0, (%rax)
; AVX2-ONLY-NEXT:    vmovdqa %ymm2, 32(%rax)
; AVX2-ONLY-NEXT:    addq $1000, %rsp # imm = 0x3E8
; AVX2-ONLY-NEXT:    vzeroupper
; AVX2-ONLY-NEXT:    retq
;
; AVX512F-SLOW-LABEL: load_i16_stride8_vf32:
; AVX512F-SLOW:       # %bb.0:
; AVX512F-SLOW-NEXT:    subq $616, %rsp # imm = 0x268
; AVX512F-SLOW-NEXT:    vmovdqa 368(%rdi), %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 352(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-SLOW-NEXT:    vmovdqa 336(%rdi), %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 320(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-SLOW-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [0,4,0,4]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm27 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm5[0,1,2],xmm0[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm0, %xmm26
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm25 = xmm5[2],xmm3[2],xmm5[3],xmm3[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm5, %xmm0
; AVX512F-SLOW-NEXT:    vpermt2d %xmm3, %xmm1, %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa 304(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 288(%rdi), %xmm3
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm11 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX512F-SLOW-NEXT:    vmovdqa 272(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, (%rsp) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 256(%rdi), %xmm3
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm7 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm7[0],xmm11[0],xmm7[1],xmm11[1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm16
; AVX512F-SLOW-NEXT:    vmovdqa 480(%rdi), %ymm0
; AVX512F-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm18 = ymm0[0,1,0,2]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm18[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm2, %ymm21
; AVX512F-SLOW-NEXT:    vmovdqa 448(%rdi), %ymm2
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm20 = ymm2[0,1,0,2]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm12 = ymm20[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm12[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-SLOW-NEXT:    vmovdqa 416(%rdi), %ymm2
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm22 = ymm2[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqa 384(%rdi), %ymm2
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm29 = ymm2[0,1,0,2]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm10 = ymm22[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm10[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm9 = ymm29[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm8 = ymm9[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3,4],ymm2[5],ymm8[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-SLOW-NEXT:    movb $-64, %al
; AVX512F-SLOW-NEXT:    kmovw %eax, %k1
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm16 {%k1}
; AVX512F-SLOW-NEXT:    vmovdqa 112(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 96(%rdi), %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; AVX512F-SLOW-NEXT:    vmovdqa 80(%rdi), %xmm3
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 64(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm8 = xmm0[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm3 = xmm2[0,1,2],xmm8[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm3, %xmm23
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm19 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm24 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX512F-SLOW-NEXT:    vpermt2d %xmm0, %xmm1, %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 48(%rdi), %xmm4
; AVX512F-SLOW-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm8 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm15 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm6 = xmm15[0,1],xmm2[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa 224(%rdi), %ymm0
; AVX512F-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm31 = ymm0[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqa 192(%rdi), %ymm0
; AVX512F-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm30 = ymm0[0,1,0,2]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm15 = ymm31[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm5 = ymm30[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm14 = ymm5[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm14[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-SLOW-NEXT:    vmovdqa 160(%rdi), %ymm0
; AVX512F-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm14 = ymm0[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqa64 128(%rdi), %ymm28
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm17 = ymm28[0,1,0,2]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm14[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm2[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm17[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm1[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3,4],ymm0[5],ymm13[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm16, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm11[1],xmm0[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm27, %xmm4
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm21, %ymm4
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm6 = ymm12[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5,6],ymm4[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm10[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm9 = ymm9[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm9[0,1,2,3,4],ymm6[5],ymm9[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm4[6,7]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm4, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm15[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm5[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5,6],ymm4[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm3[1,1,1,1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0],xmm8[1],xmm2[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm19, %xmm4
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm4[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm7[2],xmm11[2],xmm7[3],xmm11[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm26, %xmm1
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm18[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm4 = ymm20[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm4[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm5 = ymm22[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm5[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm9 = ymm29[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm10 = ymm9[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm10[0,1,2,3,4],ymm6[5],ymm10[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm31[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm6 = ymm2[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm10 = ymm30[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm12 = ymm10[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2,3,4,5,6],ymm6[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm12 = ymm14[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm12[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm14 = ymm17[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm14[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4],ymm13[5],ymm15[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm13[0,1,2,3,4,5],ymm6[6,7]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm13 = xmm3[2],xmm8[2],xmm3[3],xmm8[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm23, %xmm15
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm13 = xmm13[0,1],xmm15[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm13[0,1,2,3],ymm6[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm6, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vpbroadcastq {{.*#+}} xmm0 = [3,7,3,7]
; AVX512F-SLOW-NEXT:    vpermt2d %xmm11, %xmm0, %xmm7
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, %xmm6
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm25, %xmm0
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm7[0,1],xmm0[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm4[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm5[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm9[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5],ymm5[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vpermt2d %xmm8, %xmm6, %xmm3
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm24, %xmm1
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm10[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm12[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm14[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5],ymm4[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0,1,2],xmm2[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm2, %xmm30
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm22 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm31 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, %xmm2
; AVX512F-SLOW-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [0,4,0,4]
; AVX512F-SLOW-NEXT:    vpermt2d %xmm0, %xmm1, %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm5 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd (%rsp), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm4 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm5, %xmm20
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm2, %zmm0, %zmm12
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm19 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm19 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm7 = ymm19[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm7[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm21 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm21 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm11 = ymm21[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm11[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm29 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm29 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm23 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm23 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm10 = ymm29[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm10[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm13 = ymm23[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm13[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm3[5],ymm5[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm12 {%k1}
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, %xmm0
; AVX512F-SLOW-NEXT:    vpermi2d %xmm2, %xmm3, %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm3, %xmm16
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm2, %xmm18
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm8 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm8 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm5 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm5 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm15 = xmm5[0],xmm8[0],xmm5[1],xmm8[1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm15 = xmm15[0,1],xmm0[2,3]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm17 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm17 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm14 = ymm17[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm14[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm24 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm24 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm24[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm9 = ymm3[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm25 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm25 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm26 = ymm28[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm9 = ymm25[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm9[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm26[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm1[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm0[5],ymm6[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm12, %zmm27
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm20, %xmm12
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm12[1],xmm0[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm22, %xmm2
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm7[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm6 = ymm11[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm10[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm13[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5],ymm7[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm14[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm9[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm16, %xmm9
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm16[0],xmm18[0],xmm16[1],xmm18[1]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm5[1,1,1,1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0],xmm8[1],xmm3[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm20
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm4[2],xmm12[2],xmm4[3],xmm12[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm12, %xmm16
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm30, %xmm1
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm1
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm19[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm21[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm6 = ymm3[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm6 = ymm29[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm6[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm10 = ymm23[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm11 = ymm10[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm7 = ymm11[0,1,2,3,4],ymm7[5],ymm11[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm1 {%k1}
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm17[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm11 = ymm0[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm7 = ymm24[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm13 = ymm7[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm11 = ymm13[0,1,2,3,4,5,6],ymm11[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm13 = ymm25[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm14 = ymm13[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm15 = ymm26[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm15[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm14[5],ymm12[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm12 = xmm18[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm12 = xmm9[0,1,2],xmm12[3]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm14 = xmm5[2],xmm8[2],xmm5[3],xmm8[3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm12 = xmm14[0,1],xmm12[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2,3],ymm11[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm11, %zmm1, %zmm1
; AVX512F-SLOW-NEXT:    vpbroadcastq {{.*#+}} xmm12 = [3,7,3,7]
; AVX512F-SLOW-NEXT:    vpermt2d %xmm16, %xmm12, %xmm4
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm31, %xmm11
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm11[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm4, %zmm0, %zmm4
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm6[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm10[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3,4],ymm3[5],ymm6[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm4 {%k1}
; AVX512F-SLOW-NEXT:    vpermt2d %xmm8, %xmm12, %xmm5
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm9[2],xmm18[2],xmm9[3],xmm18[3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm7[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm13[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm15[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm3[5],ymm5[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm4, %zmm0
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm2, (%rsi)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm2, (%rdx)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm2, (%rcx)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm2, (%r8)
; AVX512F-SLOW-NEXT:    vmovdqa64 %zmm27, (%r9)
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT:    vmovdqa64 %zmm20, (%rax)
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT:    vmovdqa64 %zmm1, (%rax)
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT:    vmovdqa64 %zmm0, (%rax)
; AVX512F-SLOW-NEXT:    addq $616, %rsp # imm = 0x268
; AVX512F-SLOW-NEXT:    vzeroupper
; AVX512F-SLOW-NEXT:    retq
;
; AVX512F-FAST-LABEL: load_i16_stride8_vf32:
; AVX512F-FAST:       # %bb.0:
; AVX512F-FAST-NEXT:    subq $552, %rsp # imm = 0x228
; AVX512F-FAST-NEXT:    vmovdqa 368(%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 352(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-FAST-NEXT:    vmovdqa 336(%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 320(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [0,4,0,4]
; AVX512F-FAST-NEXT:    vmovdqa %xmm3, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm16
; AVX512F-FAST-NEXT:    vpermt2d %xmm2, %xmm1, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm2, %xmm22
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, %xmm2
; AVX512F-FAST-NEXT:    vmovdqa 304(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 288(%rdi), %xmm3
; AVX512F-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
; AVX512F-FAST-NEXT:    vmovdqa 272(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, (%rsp) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 256(%rdi), %xmm3
; AVX512F-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm9 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm9[0],xmm4[0],xmm9[1],xmm4[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm4, %xmm20
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm11
; AVX512F-FAST-NEXT:    vmovdqa 480(%rdi), %ymm0
; AVX512F-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm21 = ymm0[0,1,0,2]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm12 = ymm21[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm12[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vmovdqa 448(%rdi), %ymm1
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm27 = ymm1[0,1,0,2]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm10 = ymm27[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm10[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-FAST-NEXT:    vmovdqa 416(%rdi), %ymm1
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm28 = ymm1[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqa 384(%rdi), %ymm1
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm29 = ymm1[0,1,0,2]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm8 = ymm28[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm8[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm7 = ymm29[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm7[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm1[5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-FAST-NEXT:    movb $-64, %al
; AVX512F-FAST-NEXT:    kmovw %eax, %k1
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm11 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa 112(%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 96(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-FAST-NEXT:    vmovdqa 80(%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 64(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm13 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-FAST-NEXT:    vmovdqa %xmm13, %xmm0
; AVX512F-FAST-NEXT:    vpermt2d %xmm3, %xmm2, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm25
; AVX512F-FAST-NEXT:    vmovdqa (%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX512F-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX512F-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 48(%rdi), %xmm4
; AVX512F-FAST-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm6 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm17
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm5 = xmm1[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vmovdqa 224(%rdi), %ymm0
; AVX512F-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm30 = ymm0[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqa64 192(%rdi), %ymm23
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm31 = ymm23[0,1,0,2]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm4 = ymm30[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm3 = ymm31[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm14 = ymm3[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-FAST-NEXT:    vmovdqa 160(%rdi), %ymm0
; AVX512F-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm19 = ymm0[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqa 128(%rdi), %ymm0
; AVX512F-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm18 = ymm0[0,1,0,2]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm14 = ymm19[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm14[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm2 = ymm18[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm2[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3,4],ymm0[5],ymm15[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm11, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [1,5,1,5]
; AVX512F-FAST-NEXT:    vmovdqa %xmm9, %xmm0
; AVX512F-FAST-NEXT:    vpermt2d %xmm20, %xmm1, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, %xmm11
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm16[0],xmm22[0],xmm16[1],xmm22[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm10[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm8[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm7[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm5 = ymm7[0,1,2,3,4],ymm5[5],ymm7[6,7]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa %xmm6, %xmm1
; AVX512F-FAST-NEXT:    vpermt2d %xmm17, %xmm11, %xmm1
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm13[0],xmm25[0],xmm13[1],xmm25[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm14[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm16, %xmm2
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm16 = xmm16[2],xmm22[2],xmm16[3],xmm22[3]
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm0 = [2,6,2,6]
; AVX512F-FAST-NEXT:    vpermt2d %xmm22, %xmm0, %xmm2
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm9[2],xmm20[2],xmm9[3],xmm20[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm2, %zmm0, %zmm3
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm2 = ymm21[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm2[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm4 = ymm27[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm7 = ymm4[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm5[7]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm5 = ymm28[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm10 = ymm5[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm8 = ymm29[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm8[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3,4],ymm10[5],ymm12[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm7, %zmm0, %zmm3 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa %xmm13, %xmm7
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm10 = xmm13[2],xmm25[2],xmm13[3],xmm25[3]
; AVX512F-FAST-NEXT:    vpermt2d %xmm25, %xmm0, %xmm7
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm12 = xmm6[2],xmm17[2],xmm6[3],xmm17[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm7 = xmm12[0,1],xmm7[2,3]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm12 = ymm30[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm13 = ymm12[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm14 = ymm31[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm15 = ymm14[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5,6],ymm13[7]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm15 = ymm19[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm15[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm1 = ymm18[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm11 = ymm1[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm11[0,1,2,3,4],ymm0[5],ymm11[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm13[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm3, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm27 = [3,7,3,7]
; AVX512F-FAST-NEXT:    vpermt2d %xmm20, %xmm27, %xmm9
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm16, %xmm0
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm9[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm4[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm5[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm8[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5],ymm4[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT:    vpermt2d %xmm17, %xmm27, %xmm6
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm2 = xmm6[0,1],xmm10[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm12[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm14[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5,6],ymm3[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm15[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm4[5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm28
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm12 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd (%rsp), %xmm0, %xmm5 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm5 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa %xmm12, %xmm1
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm0 = [0,4,0,4]
; AVX512F-FAST-NEXT:    vpermt2d %xmm3, %xmm0, %xmm1
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm24
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm2, %xmm18
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm1, %zmm0, %zmm31
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm29 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm29 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm9 = ymm29[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm26 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm26 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm6 = ymm26[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm6[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5,6],ymm3[7]
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm30 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm30 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm25 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm25 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm8 = ymm30[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm8[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm13 = ymm25[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm13[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm4 = ymm7[0,1,2,3,4],ymm4[5],ymm7[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm31 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm2 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm3 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm4 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm4 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-FAST-NEXT:    vpermi2d %xmm2, %xmm3, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm17
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm2, %xmm22
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm7 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm7 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm10 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm4, %xmm16
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm11 = xmm10[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm20 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm20 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm10 = ymm20[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm23 = ymm23[0,1,1,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm10[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm14 = ymm23[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm15 = ymm14[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm19 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm19 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm21 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm21 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm15 = ymm19[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm15[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm2 = ymm21[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm2[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm1[5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm31, %zmm31
; AVX512F-FAST-NEXT:    vmovdqa %xmm5, %xmm0
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm4 = [1,5,1,5]
; AVX512F-FAST-NEXT:    vpermt2d %xmm18, %xmm4, %xmm0
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm12[0],xmm24[0],xmm12[1],xmm24[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm9[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm6[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm8[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm13[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3,4],ymm3[5],ymm6[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa %xmm4, %xmm3
; AVX512F-FAST-NEXT:    vpermi2d %xmm16, %xmm7, %xmm3
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm17[0],xmm22[0],xmm17[1],xmm22[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm6 = xmm3[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm14[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm15[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm9 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vmovdqa %xmm12, %xmm8
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm24, %xmm2
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm10 = xmm12[2],xmm24[2],xmm12[3],xmm24[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm9[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm24
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm0 = [2,6,2,6]
; AVX512F-FAST-NEXT:    vpermt2d %xmm2, %xmm0, %xmm8
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm5[2],xmm18[2],xmm5[3],xmm18[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm8[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm1, %zmm0, %zmm1
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm2 = ymm29[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm2[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm6 = ymm26[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm8 = ymm6[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm8[0,1,2,3,4,5,6],ymm3[7]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm8 = ymm30[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm9 = ymm8[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm11 = ymm25[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm11[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm9 = ymm12[0,1,2,3,4],ymm9[5],ymm12[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm1 {%k1}
; AVX512F-FAST-NEXT:    vpermi2d %xmm22, %xmm17, %xmm0
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm7[2],xmm16[2],xmm7[3],xmm16[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm9 = ymm20[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm12 = ymm9[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm13 = ymm23[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm14 = ymm13[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm12 = ymm14[0,1,2,3,4,5,6],ymm12[7]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm14 = ymm19[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm14[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm0 = ymm21[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm0[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm15[5],ymm4[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm12[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm3, %zmm1, %zmm1
; AVX512F-FAST-NEXT:    vpermt2d %xmm18, %xmm27, %xmm5
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm3 = xmm5[0,1],xmm10[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm3, %zmm0, %zmm3
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm6[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm8[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm11[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5],ymm5[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm3 {%k1}
; AVX512F-FAST-NEXT:    vpermt2d %xmm16, %xmm27, %xmm7
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm17[2],xmm22[2],xmm17[3],xmm22[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm2 = xmm7[0,1],xmm2[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm9[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm13[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5,6],ymm4[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm14[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5],ymm0[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm3, %zmm0
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, (%rsi)
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, (%rdx)
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, (%rcx)
; AVX512F-FAST-NEXT:    vmovdqa64 %zmm28, (%r8)
; AVX512F-FAST-NEXT:    vmovdqa64 %zmm31, (%r9)
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT:    vmovdqa64 %zmm24, (%rax)
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT:    vmovdqa64 %zmm1, (%rax)
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT:    vmovdqa64 %zmm0, (%rax)
; AVX512F-FAST-NEXT:    addq $552, %rsp # imm = 0x228
; AVX512F-FAST-NEXT:    vzeroupper
; AVX512F-FAST-NEXT:    retq
;
; AVX512BW-LABEL: load_i16_stride8_vf32:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %r10
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %r11
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm3
; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm4
; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm5
; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm7
; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm6
; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm8 = [0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56]
; AVX512BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm9
; AVX512BW-NEXT:    vpermt2w %zmm7, %zmm8, %zmm9
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512BW-NEXT:    vpermt2w %zmm5, %zmm8, %zmm10
; AVX512BW-NEXT:    movb $-64, %dil
; AVX512BW-NEXT:    kmovd %edi, %k1
; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm10 {%k1}
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm9 = [0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56]
; AVX512BW-NEXT:    # ymm9 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm4, %zmm3, %zmm9
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm8
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7]
; AVX512BW-NEXT:    vinserti64x4 $0, %ymm8, %zmm10, %zmm8
; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm9 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512BW-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm10
; AVX512BW-NEXT:    vpermt2w %zmm7, %zmm9, %zmm10
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm11
; AVX512BW-NEXT:    vpermt2w %zmm5, %zmm9, %zmm11
; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm11 {%k1}
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm10 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512BW-NEXT:    # ymm10 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm4, %zmm3, %zmm10
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm9
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
; AVX512BW-NEXT:    vinserti64x4 $0, %ymm9, %zmm11, %zmm9
; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm10 = [2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58]
; AVX512BW-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm11
; AVX512BW-NEXT:    vpermt2w %zmm7, %zmm10, %zmm11
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm12
; AVX512BW-NEXT:    vpermt2w %zmm5, %zmm10, %zmm12
; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm12 {%k1}
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm11 = [2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58]
; AVX512BW-NEXT:    # ymm11 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm4, %zmm3, %zmm11
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm10
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
; AVX512BW-NEXT:    vinserti64x4 $0, %ymm10, %zmm12, %zmm10
; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm11 = [3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59]
; AVX512BW-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm12
; AVX512BW-NEXT:    vpermt2w %zmm7, %zmm11, %zmm12
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm13
; AVX512BW-NEXT:    vpermt2w %zmm5, %zmm11, %zmm13
; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm13 {%k1}
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm12 = [3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59]
; AVX512BW-NEXT:    # ymm12 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm4, %zmm3, %zmm12
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm11
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
; AVX512BW-NEXT:    vinserti64x4 $0, %ymm11, %zmm13, %zmm11
; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm12 = [4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60]
; AVX512BW-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm13
; AVX512BW-NEXT:    vpermt2w %zmm7, %zmm12, %zmm13
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm14
; AVX512BW-NEXT:    vpermt2w %zmm5, %zmm12, %zmm14
; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm14 {%k1}
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm13 = [4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60]
; AVX512BW-NEXT:    # ymm13 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm4, %zmm3, %zmm13
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm12
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6,7]
; AVX512BW-NEXT:    vinserti64x4 $0, %ymm12, %zmm14, %zmm12
; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm13 = [5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61]
; AVX512BW-NEXT:    # zmm13 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm14
; AVX512BW-NEXT:    vpermt2w %zmm7, %zmm13, %zmm14
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm15
; AVX512BW-NEXT:    vpermt2w %zmm5, %zmm13, %zmm15
; AVX512BW-NEXT:    vmovdqa64 %zmm14, %zmm15 {%k1}
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm14 = [5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61]
; AVX512BW-NEXT:    # ymm14 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm4, %zmm3, %zmm14
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm13
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm14[4,5,6,7]
; AVX512BW-NEXT:    vinserti64x4 $0, %ymm13, %zmm15, %zmm13
; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm14 = [6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62]
; AVX512BW-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm15
; AVX512BW-NEXT:    vpermt2w %zmm7, %zmm14, %zmm15
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm16
; AVX512BW-NEXT:    vpermt2w %zmm5, %zmm14, %zmm16
; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm16 {%k1}
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm15 = [6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62]
; AVX512BW-NEXT:    # ymm15 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm4, %zmm3, %zmm15
; AVX512BW-NEXT:    vpermi2w %zmm1, %zmm0, %zmm14
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5,6,7]
; AVX512BW-NEXT:    vinserti64x4 $0, %ymm14, %zmm16, %zmm14
; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm15 = [7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63]
; AVX512BW-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermt2w %zmm7, %zmm15, %zmm6
; AVX512BW-NEXT:    vpermt2w %zmm5, %zmm15, %zmm2
; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm2 {%k1}
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63]
; AVX512BW-NEXT:    # ymm5 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpermi2w %zmm4, %zmm3, %zmm5
; AVX512BW-NEXT:    vpermt2w %zmm1, %zmm15, %zmm0
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX512BW-NEXT:    vinserti64x4 $0, %ymm0, %zmm2, %zmm0
; AVX512BW-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm9, (%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm10, (%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm11, (%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm12, (%r9)
; AVX512BW-NEXT:    vmovdqa64 %zmm13, (%r11)
; AVX512BW-NEXT:    vmovdqa64 %zmm14, (%r10)
; AVX512BW-NEXT:    vmovdqa64 %zmm0, (%rax)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
  %wide.vec = load <256 x i16>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120, i32 128, i32 136, i32 144, i32 152, i32 160, i32 168, i32 176, i32 184, i32 192, i32 200, i32 208, i32 216, i32 224, i32 232, i32 240, i32 248>
  %strided.vec1 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121, i32 129, i32 137, i32 145, i32 153, i32 161, i32 169, i32 177, i32 185, i32 193, i32 201, i32 209, i32 217, i32 225, i32 233, i32 241, i32 249>
  %strided.vec2 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122, i32 130, i32 138, i32 146, i32 154, i32 162, i32 170, i32 178, i32 186, i32 194, i32 202, i32 210, i32 218, i32 226, i32 234, i32 242, i32 250>
  %strided.vec3 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 3, i32 11, i32 19, i32 27, i32 35, i32 43, i32 51, i32 59, i32 67, i32 75, i32 83, i32 91, i32 99, i32 107, i32 115, i32 123, i32 131, i32 139, i32 147, i32 155, i32 163, i32 171, i32 179, i32 187, i32 195, i32 203, i32 211, i32 219, i32 227, i32 235, i32 243, i32 251>
  %strided.vec4 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 4, i32 12, i32 20, i32 28, i32 36, i32 44, i32 52, i32 60, i32 68, i32 76, i32 84, i32 92, i32 100, i32 108, i32 116, i32 124, i32 132, i32 140, i32 148, i32 156, i32 164, i32 172, i32 180, i32 188, i32 196, i32 204, i32 212, i32 220, i32 228, i32 236, i32 244, i32 252>
  %strided.vec5 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125, i32 133, i32 141, i32 149, i32 157, i32 165, i32 173, i32 181, i32 189, i32 197, i32 205, i32 213, i32 221, i32 229, i32 237, i32 245, i32 253>
  %strided.vec6 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126, i32 134, i32 142, i32 150, i32 158, i32 166, i32 174, i32 182, i32 190, i32 198, i32 206, i32 214, i32 222, i32 230, i32 238, i32 246, i32 254>
  %strided.vec7 = shufflevector <256 x i16> %wide.vec, <256 x i16> poison, <32 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127, i32 135, i32 143, i32 151, i32 159, i32 167, i32 175, i32 183, i32 191, i32 199, i32 207, i32 215, i32 223, i32 231, i32 239, i32 247, i32 255>
  store <32 x i16> %strided.vec0, ptr %out.vec0, align 64
  store <32 x i16> %strided.vec1, ptr %out.vec1, align 64
  store <32 x i16> %strided.vec2, ptr %out.vec2, align 64
  store <32 x i16> %strided.vec3, ptr %out.vec3, align 64
  store <32 x i16> %strided.vec4, ptr %out.vec4, align 64
  store <32 x i16> %strided.vec5, ptr %out.vec5, align 64
  store <32 x i16> %strided.vec6, ptr %out.vec6, align 64
  store <32 x i16> %strided.vec7, ptr %out.vec7, align 64
  ret void
}

define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
; SSE-LABEL: load_i16_stride8_vf64:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $1800, %rsp # imm = 0x708
; SSE-NEXT:    movdqa 752(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 736(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 208(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 192(%rdi), %xmm4
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 240(%rdi), %xmm5
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 224(%rdi), %xmm6
; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 144(%rdi), %xmm7
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 128(%rdi), %xmm8
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 176(%rdi), %xmm9
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 160(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3]
; SSE-NEXT:    movdqa %xmm0, %xmm9
; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3]
; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,0,0,0]
; SSE-NEXT:    movdqa %xmm6, %xmm12
; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm8, %xmm0
; SSE-NEXT:    movdqa %xmm8, %xmm4
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
; SSE-NEXT:    movdqa %xmm9, %xmm7
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 720(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 704(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa 688(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 672(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm0, %xmm3
; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 656(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 640(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 624(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 608(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm1, %xmm2
; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 592(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 576(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa 560(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 544(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm0, %xmm3
; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 528(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 512(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 496(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 480(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 464(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 448(%rdi), %xmm10
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1],xmm10[2],xmm0[2],xmm10[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa 432(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 416(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm0, %xmm3
; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 400(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 384(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1008(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 992(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 976(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 960(%rdi), %xmm13
; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1],xmm13[2],xmm0[2],xmm13[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa 944(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 928(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm0, %xmm3
; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 912(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 896(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 368(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 352(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 336(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 320(%rdi), %xmm10
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1],xmm10[2],xmm0[2],xmm10[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa 304(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 288(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm0, %xmm3
; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 272(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 256(%rdi), %xmm14
; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1],xmm14[2],xmm0[2],xmm14[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm14, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 880(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 864(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 848(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 832(%rdi), %xmm9
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa 816(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 800(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm0, %xmm3
; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 784(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 768(%rdi), %xmm15
; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1],xmm15[2],xmm0[2],xmm15[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm15, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 112(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 96(%rdi), %xmm13
; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1],xmm13[2],xmm0[2],xmm13[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm13, {{[-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:    movdqa 64(%rdi), %xmm5
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
; SSE-NEXT:    movdqa 32(%rdi), %xmm8
; SSE-NEXT:    movdqa %xmm8, {{[-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:    movdqa %xmm8, %xmm2
; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa (%rdi), %xmm6
; SSE-NEXT:    movdqa %xmm6, {{[-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:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1],xmm6[2],xmm1[2],xmm6[3],xmm1[3]
; SSE-NEXT:    movdqa %xmm6, %xmm1
; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
; SSE-NEXT:    movdqa %xmm7, %xmm14
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm7, %xmm0
; SSE-NEXT:    movdqa %xmm12, %xmm4
; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[1,1,1,1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[1,1,1,1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    movaps %xmm5, %xmm0
; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm14[2],xmm0[3],xmm14[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm13[2],xmm12[3],xmm13[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm12[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,2,2,2]
; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm11[2],xmm10[3],xmm11[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm10[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,2,2]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,2,2]
; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm9[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,2,2]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    movapd %xmm4, %xmm0
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    movapd %xmm7, %xmm0
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm8[2],xmm0[3],xmm8[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,2,2,2]
; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm5, %xmm1
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm6[2],xmm1[3],xmm6[3]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm15, %xmm2
; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm13[2],xmm2[3],xmm13[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm10[2],xmm14[3],xmm10[3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm14[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm11[2],xmm2[3],xmm11[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm9[2],xmm2[3],xmm9[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT:    # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; SSE-NEXT:    # xmm11 = xmm11[4],mem[4],xmm11[5],mem[5],xmm11[6],mem[6],xmm11[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm11, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
; SSE-NEXT:    movdqa %xmm6, %xmm13
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; SSE-NEXT:    # xmm14 = xmm14[4],mem[4],xmm14[5],mem[5],xmm14[6],mem[6],xmm14[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm2, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; SSE-NEXT:    # xmm10 = xmm10[4],mem[4],xmm10[5],mem[5],xmm10[6],mem[6],xmm10[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT:    # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm6, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = xmm8[4],mem[4],xmm8[5],mem[5],xmm8[6],mem[6],xmm8[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm3[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm0[2],xmm7[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm2, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm0[0],xmm7[1]
; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; SSE-NEXT:    # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm2[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm0[2],xmm9[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm5, %xmm0
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
; SSE-NEXT:    movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT:    # xmm7 = xmm7[4],mem[4],xmm7[5],mem[5],xmm7[6],mem[6],xmm7[7],mem[7]
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; SSE-NEXT:    # xmm12 = xmm12[4],mem[4],xmm12[5],mem[5],xmm12[6],mem[6],xmm12[7],mem[7]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[0,0,0,0]
; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm2, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa (%rsp), %xmm2 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; SSE-NEXT:    movdqa %xmm2, (%rsp) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm3, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT:    # xmm7 = xmm7[4],mem[4],xmm7[5],mem[5],xmm7[6],mem[6],xmm7[7],mem[7]
; SSE-NEXT:    movdqa %xmm7, %xmm2
; SSE-NEXT:    movdqa %xmm7, %xmm3
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[1,1,1,1]
; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT:    movaps %xmm13, %xmm0
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[1,1,1,1]
; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    movaps %xmm7, %xmm0
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    movaps %xmm10, %xmm0
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    movaps %xmm8, %xmm0
; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm6, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $85, (%rsp), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    movaps %xmm5, %xmm0
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,1,1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, %xmm0
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm13[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm0[2],xmm11[3],xmm0[3]
; SSE-NEXT:    movdqa %xmm11, %xmm1
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; SSE-NEXT:    # xmm11 = mem[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm0[2],xmm11[3],xmm0[3]
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    movapd %xmm14, %xmm0
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm0[0],xmm11[1]
; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm6[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm0[2],xmm10[3],xmm0[3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm5[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm0[2],xmm9[3],xmm0[3]
; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    unpckhps (%rsp), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,2,2]
; SSE-NEXT:    movaps %xmm2, %xmm15
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,2,2,2]
; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm13, %xmm7
; SSE-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm12[2],xmm7[3],xmm12[3]
; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm1[2,3]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm1[2,3]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT:    # xmm6 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm1[2,3]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; SSE-NEXT:    # xmm5 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm1[2,3]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm1[2,3]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm14[3,3,3,3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2,3]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $255, (%rsp), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm12[2,3]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    unpckhps {{.*#+}} xmm12 = xmm12[2],xmm15[2],xmm12[3],xmm15[3]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[3,3,3,3]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm12[2,3]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 96(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 32(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 112(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 48(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 64(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, (%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 80(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 16(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 96(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 32(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 112(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 48(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 64(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, (%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 80(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 16(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 96(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 32(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 112(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 48(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 64(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, (%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 80(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 16(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 112(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 96(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 80(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 64(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 48(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 32(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 16(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, (%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 112(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 96(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 80(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 64(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 48(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 32(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 16(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, (%r9)
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    movaps %xmm12, 112(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    movaps %xmm12, 96(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    movaps %xmm12, 80(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    movaps %xmm12, 64(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    movaps %xmm12, 48(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    movaps %xmm12, 32(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movaps %xmm15, 16(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    movaps %xmm12, (%rax)
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movapd %xmm9, 112(%rax)
; SSE-NEXT:    movapd %xmm10, 96(%rax)
; SSE-NEXT:    movapd %xmm11, 80(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT:    movaps %xmm9, 64(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT:    movaps %xmm9, 48(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT:    movaps %xmm9, 32(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT:    movaps %xmm9, 16(%rax)
; SSE-NEXT:    movaps %xmm7, (%rax)
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movaps %xmm1, 112(%rax)
; SSE-NEXT:    movaps %xmm2, 96(%rax)
; SSE-NEXT:    movaps %xmm3, 80(%rax)
; SSE-NEXT:    movaps %xmm4, 64(%rax)
; SSE-NEXT:    movaps %xmm5, 48(%rax)
; SSE-NEXT:    movaps %xmm6, 32(%rax)
; SSE-NEXT:    movaps %xmm8, 16(%rax)
; SSE-NEXT:    movaps %xmm0, (%rax)
; SSE-NEXT:    addq $1800, %rsp # imm = 0x708
; SSE-NEXT:    retq
;
; AVX1-ONLY-LABEL: load_i16_stride8_vf64:
; AVX1-ONLY:       # %bb.0:
; AVX1-ONLY-NEXT:    subq $2056, %rsp # imm = 0x808
; AVX1-ONLY-NEXT:    vmovdqa 304(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 288(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm14 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vmovdqa 272(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 256(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm11 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm11[0],xmm14[0],xmm11[1],xmm14[1]
; AVX1-ONLY-NEXT:    vmovdqa 368(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 352(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa 336(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 320(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm13 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm13[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5],xmm1[6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa 496(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 480(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm10 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm10[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 464(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 448(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm3[0,1,0,1]
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, %xmm9
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5],xmm1[6,7]
; AVX1-ONLY-NEXT:    vmovdqa 432(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 416(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 400(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 384(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 880(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 864(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa 848(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 832(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vmovdqa 816(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 800(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 784(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 768(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm8 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa 1008(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 992(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm12 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm12[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa 976(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 960(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5],xmm1[6,7]
; AVX1-ONLY-NEXT:    vmovdqa 944(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 928(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm6 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 912(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 896(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 624(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 608(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa 592(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 576(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vmovdqa 560(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 544(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 528(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 512(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa 752(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 736(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa 720(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 704(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5],xmm1[6,7]
; AVX1-ONLY-NEXT:    vmovdqa 688(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 672(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 656(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 640(%rdi), %xmm3
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 240(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 224(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa 208(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 192(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm4[0,1,0,1]
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vmovdqa 176(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 160(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 144(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 128(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT:    vmovdqa 112(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 96(%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa 80(%rdi), %xmm0
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 64(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm2[0,1,2,3,4,5],xmm1[6,7]
; AVX1-ONLY-NEXT:    vmovdqa (%rdi), %xmm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 32(%rdi), %xmm6
; AVX1-ONLY-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa 48(%rdi), %xmm7
; AVX1-ONLY-NEXT:    vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
; AVX1-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm15 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, %xmm2
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm15[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
; AVX1-ONLY-NEXT:    vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm14[2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm13[0],xmm5[0],xmm13[1],xmm5[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm10[1,1,1,1]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm15[0],xmm9[1],xmm15[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm15[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm8[2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm1[0],xmm12[0],xmm1[1],xmm12[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm15 = mem[1,1,1,1]
; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm15 = xmm15[0],mem[1],xmm15[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm15[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = mem[1,1,1,1]
; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = xmm1[0],mem[0],xmm1[1],mem[1]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = xmm1[0],mem[0],xmm1[1],mem[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm15 = mem[1,1,1,1]
; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm15 = xmm15[0],mem[1],xmm15[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm15[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm4[0],mem[0],xmm4[1],mem[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm4[1,1,1,1]
; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0],xmm1[1],xmm3[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = xmm3[0,1],mem[2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm15 = xmm15[0],xmm2[0],xmm15[1],xmm2[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm15[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm5[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm13[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm11[2],xmm14[2],xmm11[3],xmm14[3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm3[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm9[2,2,2,2]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm14[0,1,2],xmm15[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm15[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm6[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm7[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm3[2],xmm8[2],xmm3[3],xmm8[3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm3[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm13[2],xmm11[2],xmm13[3],xmm11[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm15 = xmm12[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm15 = mem[0,1,2,3,4,5],xmm15[6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm15[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm7[2,2,2,2]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm6[0,1,2],xmm0[3]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm10[2],xmm5[2],xmm10[3],xmm5[3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm12[2,2,2,2]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm15 = xmm8[0,1,2],xmm15[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm15[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm4[2,2,2,2]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm1[0,1,2],xmm3[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[2,2,2,2]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm2[0,1,2,3,4,5],xmm3[6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm15 = xmm15[2],mem[2],xmm15[3],mem[3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm15[0,1,2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = mem[2,3,2,3]
; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm15 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm15[0],xmm3[1],xmm15[2,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm14[2],xmm9[2],xmm14[3],xmm9[3]
; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm15 = mem[2,3,2,3]
; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm14 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm14[0],xmm15[1],xmm14[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm14[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = mem[2,3,2,3]
; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm14 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm14[0],xmm3[1],xmm14[2,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm11[2,3,2,3]
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm15 = xmm13[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm15[0],xmm14[1],xmm15[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm14[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm6[2],xmm7[2],xmm6[3],xmm7[3]
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm3 = xmm5[2,3,2,3]
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm10[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm14[0],xmm3[1],xmm14[2,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm3 = xmm8[2],xmm12[2],xmm8[3],xmm12[3]
; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm11 = mem[2,3,2,3]
; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm10 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0],xmm11[1],xmm10[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
; AVX1-ONLY-NEXT:    vpermilps $238, (%rsp), %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = mem[2,3,2,3]
; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm6 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm6[0],xmm3[1],xmm6[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = xmm2[2],mem[2],xmm2[3],mem[3]
; AVX1-ONLY-NEXT:    vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm2 = mem[2,3,2,3]
; AVX1-ONLY-NEXT:    vpshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm4 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm10 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm9 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm9 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm7 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm9[0,0,0,0]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm7[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm10[0],xmm4[0],xmm10[1],xmm4[1]
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, %xmm8
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm4 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm2 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm5 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, %xmm11
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm5[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5],xmm1[6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, %xmm15
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm6 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm4 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm4[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
; AVX1-ONLY-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm4 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm2 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm5 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[0,0,0,0]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm5[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5],xmm1[6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm4 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm4[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm4 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm13 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm13[0,0,0,0]
; AVX1-ONLY-NEXT:    vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm5 = xmm0[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm3 = xmm5[0,1,2,3,4,5],xmm3[6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm12 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,5],xmm2[6,7]
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm12[0],xmm1[0],xmm12[1],xmm1[1]
; AVX1-ONLY-NEXT:    vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[0,1,0,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm3[0,1,2,3,4,5],xmm2[6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm14 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm14[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm8[2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm7[0],xmm9[0],xmm7[1],xmm9[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm11, %xmm8
; AVX1-ONLY-NEXT:    vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm11[0],xmm8[0],xmm11[1],xmm8[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vmovdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm14 = xmm15[1,1,1,1]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm14 = xmm14[0,1],xmm3[2,3],xmm14[4,5,6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
; AVX1-ONLY-NEXT:    vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm6 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm14 = mem[1,1,1,1]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm14[0],xmm2[1],xmm14[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = mem[1,1,1,1]
; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[1],xmm0[2,3]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm14 = mem[1,1,1,1]
; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm14 = xmm14[0],mem[1],xmm14[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm12[2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = mem[1,1,1,1]
; AVX1-ONLY-NEXT:    vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[1],xmm1[2,3]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm14 = xmm14[0],mem[0],xmm14[1],mem[1]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm14[2,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm7[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = xmm10[2],mem[2],xmm10[3],mem[3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm15[2],xmm3[2],xmm15[3],xmm3[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm14 = xmm8[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm14 = xmm11[0,1,2,3,4,5],xmm14[6,7]
; AVX1-ONLY-NEXT:    vmovdqa %xmm11, %xmm15
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm14[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm5[2,2,2,2]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm4[0,1,2,3,4,5],xmm0[6,7]
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
; AVX1-ONLY-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm6[2,2,2,2]
; AVX1-ONLY-NEXT:    vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm14 = mem[0,1,2],xmm14[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm14[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm0 = xmm7[2,2,2,2]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm5[0,1,2],xmm0[3]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm8[2],xmm4[2],xmm8[3],xmm4[3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm14 = xmm6[2,2,2,2]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm3[0,1,2],xmm14[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm14[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm13[2,2,2,2]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm11[0,1,2],xmm1[3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = mem[2,2,2,2]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm13[0,1,2],xmm1[3]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm14 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm14[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = mem[2,3,2,3]
; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm14 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm14[0],xmm1[1],xmm14[2,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = xmm15[2],mem[2],xmm15[3],mem[3]
; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm14 = mem[2,3,2,3]
; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm15 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm14 = xmm15[0],xmm14[1],xmm15[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm14
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm5[2],xmm7[2],xmm5[3],xmm7[3]
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[2,3,2,3]
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm7 = xmm8[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm7[0],xmm1[1],xmm7[2,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT:    vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm6[2],xmm3[3],xmm6[3]
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm4 = xmm9[2,3,2,3]
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm6 = xmm10[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm6[0],xmm4[1],xmm6[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm4 = mem[2,3,2,3]
; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm6 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm6[0],xmm4[1],xmm6[2,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm1 = xmm4[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT:    vunpckhps (%rsp), %xmm2, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm4 = xmm2[2],mem[2],xmm2[3],mem[3]
; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm6 = mem[2,3,2,3]
; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm7 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm7[0],xmm6[1],xmm7[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm4[6,7]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm4 = xmm11[2],mem[2],xmm11[3],mem[3]
; AVX1-ONLY-NEXT:    vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm6 = mem[2,3,2,3]
; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm7 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm7[0],xmm6[1],xmm7[2,3]
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm4[6,7]
; AVX1-ONLY-NEXT:    vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm5 = xmm13[2],mem[2],xmm13[3],mem[3]
; AVX1-ONLY-NEXT:    vshufps {{.*#+}} xmm2 = xmm12[2,3,2,3]
; AVX1-ONLY-NEXT:    vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT:    # xmm3 = mem[3,3,3,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm5[2,3]
; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%rsi)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%rsi)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%rsi)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%rsi)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%rdx)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%rdx)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%rdx)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%rdx)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%rcx)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%rcx)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%rcx)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%rcx)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%r8)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%r8)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%r8)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%r8)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%r9)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%r9)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%r9)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%r9)
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%rax)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%rax)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%rax)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%rax)
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%rax)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, (%rax)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%rax)
; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT:    vmovaps %ymm3, 32(%rax)
; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%rax)
; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
; AVX1-ONLY-NEXT:    vmovaps %ymm14, 32(%rax)
; AVX1-ONLY-NEXT:    vmovaps %ymm2, (%rax)
; AVX1-ONLY-NEXT:    addq $2056, %rsp # imm = 0x808
; AVX1-ONLY-NEXT:    vzeroupper
; AVX1-ONLY-NEXT:    retq
;
; AVX2-ONLY-LABEL: load_i16_stride8_vf64:
; AVX2-ONLY:       # %bb.0:
; AVX2-ONLY-NEXT:    subq $2408, %rsp # imm = 0x968
; AVX2-ONLY-NEXT:    vmovdqa 448(%rdi), %ymm2
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 480(%rdi), %ymm3
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 272(%rdi), %xmm0
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 256(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 304(%rdi), %xmm4
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 288(%rdi), %xmm5
; AVX2-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm6 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 368(%rdi), %xmm0
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 352(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm0, %xmm0
; AVX2-ONLY-NEXT:    vmovdqa 336(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 320(%rdi), %xmm4
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm1, %xmm1
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm3[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vmovdqa 384(%rdi), %ymm3
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 416(%rdi), %ymm2
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 880(%rdi), %xmm0
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 864(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm0, %xmm0
; AVX2-ONLY-NEXT:    vmovdqa 848(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 832(%rdi), %xmm2
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm1, %xmm1
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT:    vmovdqa 784(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 768(%rdi), %xmm2
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 816(%rdi), %xmm3
; AVX2-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 800(%rdi), %xmm4
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vmovdqa 960(%rdi), %ymm2
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 992(%rdi), %ymm1
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm11 = ymm1[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm11[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm10 = ymm2[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm10[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vmovdqa 896(%rdi), %ymm3
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 928(%rdi), %ymm2
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm7 = ymm2[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm7[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm6 = ymm3[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm6[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 112(%rdi), %xmm0
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 96(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm0, %xmm0
; AVX2-ONLY-NEXT:    vmovdqa 80(%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 64(%rdi), %xmm2
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm1, %xmm1
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT:    vmovdqa (%rdi), %xmm1
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX2-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 48(%rdi), %xmm4
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, (%rsp) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vmovdqa 192(%rdi), %ymm1
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 224(%rdi), %ymm0
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm0[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm2 = ymm1[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm2[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-ONLY-NEXT:    vmovdqa 128(%rdi), %ymm0
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 160(%rdi), %ymm1
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm8 = ymm1[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm9 = ymm0[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3,4],ymm8[5],ymm9[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3,4,5],ymm5[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 624(%rdi), %xmm4
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 608(%rdi), %xmm5
; AVX2-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm4, %xmm4
; AVX2-ONLY-NEXT:    vmovdqa 592(%rdi), %xmm5
; AVX2-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 576(%rdi), %xmm8
; AVX2-ONLY-NEXT:    vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm8[0],xmm5[0],xmm8[1],xmm5[1],xmm8[2],xmm5[2],xmm8[3],xmm5[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm5, %xmm5
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-ONLY-NEXT:    vmovdqa 528(%rdi), %xmm5
; AVX2-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 512(%rdi), %xmm8
; AVX2-ONLY-NEXT:    vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 560(%rdi), %xmm9
; AVX2-ONLY-NEXT:    vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 544(%rdi), %xmm12
; AVX2-ONLY-NEXT:    vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm9 = xmm12[0],xmm9[0],xmm12[1],xmm9[1],xmm12[2],xmm9[2],xmm12[3],xmm9[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm8[0],xmm5[0],xmm8[1],xmm5[1],xmm8[2],xmm5[2],xmm8[3],xmm5[3]
; AVX2-ONLY-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm12 = xmm5[0,1],xmm4[2,3]
; AVX2-ONLY-NEXT:    vmovdqa 704(%rdi), %ymm5
; AVX2-ONLY-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 736(%rdi), %ymm4
; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm8 = ymm4[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm4 = ymm5[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm9 = ymm4[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm9[0,1,2,3,4,5,6],ymm8[7]
; AVX2-ONLY-NEXT:    vmovdqa 640(%rdi), %ymm9
; AVX2-ONLY-NEXT:    vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa 672(%rdi), %ymm8
; AVX2-ONLY-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm4 = ymm8[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm4 = ymm4[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm14 = ymm4[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpermq {{.*#+}} ymm4 = ymm9[0,1,0,2]
; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm4 = ymm4[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm4[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5],ymm15[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm12 = xmm9[1,1,1,1]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm12 = xmm12[0],xmm8[1],xmm12[2,3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm13 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm12 = xmm12[0,1],xmm13[2,3]
; AVX2-ONLY-NEXT:    vpshufhw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm13 = mem[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5,6],ymm13[7]
; AVX2-ONLY-NEXT:    vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm15 = mem[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5],ymm15[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm12 = mem[1,1,1,1]
; AVX2-ONLY-NEXT:    vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm13 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm12 = xmm12[0,1],xmm13[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm10 = ymm10[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,6],ymm11[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm7[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm6[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5],ymm6[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm10[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm12[0,1,2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa (%rsp), %xmm13 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm6 = xmm13[1,1,1,1]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm6 = xmm6[0],xmm12[1],xmm6[2,3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm7 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm0 = mem[1,1,1,1]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm7[1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX2-ONLY-NEXT:    vpshufhw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm5[0,1,2],xmm0[3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm14[2,2,2,2]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm15[0,1,2],xmm0[3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm10[2,2,2,2]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm11[0,1,2],xmm0[3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm13 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm13[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm12 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm10 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm12[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm10[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm6[2,2,2,2]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm5[0,1,2],xmm0[3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm6[2],xmm7[2],xmm6[3],xmm7[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm9 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm8 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm9[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm8[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm4 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm4[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm2[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3,4],ymm0[5],ymm15[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; AVX2-ONLY-NEXT:    vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = mem[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm3 = mem[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm15 = mem[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm15[0,1,2,3,4],ymm3[5],ymm15[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; AVX2-ONLY-NEXT:    vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = mem[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm3 = mem[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm11 = mem[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm11[0,1,2,3,4],ymm3[5],ymm11[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; AVX2-ONLY-NEXT:    vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = mem[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd $255, (%rsp), %xmm3 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm3 = mem[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm13[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm12[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm10[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm7[0,1,2,3,4],ymm3[5],ymm7[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm0 = xmm5[2],mem[2],xmm5[3],mem[3]
; AVX2-ONLY-NEXT:    vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = mem[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm3 = xmm6[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm9[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm8[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm4[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5],ymm2[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm4 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm1, %xmm0
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm4, %xmm1
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm4 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm1, %xmm0
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm4, %xmm1
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm11 = ymm1[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm11[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm10 = ymm2[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm10[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm9 = ymm2[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm9[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm8 = ymm3[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm8[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm1, %xmm0
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm2, %xmm1
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm2 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, (%rsp) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm7 = ymm1[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm6 = ymm2[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm6[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm5 = ymm2[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm5[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm4 = ymm3[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm4[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm12 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX2-ONLY-NEXT:    vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm1, %xmm0
; AVX2-ONLY-NEXT:    vpbroadcastd %xmm12, %xmm1
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm12 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm3 = ymm0[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm2 = ymm1[0,1,2,0,4,5,6,4]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm2[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm15 = mem[0,1,1,3]
; AVX2-ONLY-NEXT:    vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm1 = ymm0[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm14 = ymm1[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} ymm0 = ymm15[0,2,2,3,4,6,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm0[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5],ymm15[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm12 = mem[1,1,1,1]
; AVX2-ONLY-NEXT:    vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm13 = xmm13[0],mem[0],xmm13[1],mem[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm12 = xmm12[0,1],xmm13[2,3]
; AVX2-ONLY-NEXT:    vpshufhw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm13 = mem[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5,6],ymm13[7]
; AVX2-ONLY-NEXT:    vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm15 = mem[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5],ymm15[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm12 = xmm14[1,1,1,1]
; AVX2-ONLY-NEXT:    vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm12 = xmm12[0],mem[1],xmm12[2,3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckldq {{.*#+}} xmm13 = xmm13[0],xmm15[0],xmm13[1],xmm15[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm12 = xmm12[0,1],xmm13[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm10 = ymm10[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,6],ymm11[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm9 = ymm9[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm8 = ymm8[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5],ymm8[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm10[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm12[0,1,2,3],ymm8[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm8 = xmm12[1,1,1,1]
; AVX2-ONLY-NEXT:    vmovdqa (%rsp), %xmm10 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm8 = xmm8[0],xmm10[1],xmm8[2,3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm9 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm9 = xmm11[0],mem[0],xmm11[1],mem[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm8 = xmm8[0,1],xmm9[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm6 = ymm6[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm5[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm4[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5],ymm4[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm4 = mem[1,1,1,1]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0],xmm13[1],xmm4[2,3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm5 = xmm5[0],mem[0],xmm5[1],mem[1]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm0 = mem[2,2,2,2]
; AVX2-ONLY-NEXT:    vpblendd $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm15[2,2,2,2]
; AVX2-ONLY-NEXT:    vpblendd $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm0 = mem[0,1,2],xmm0[3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = xmm14[2],mem[2],xmm14[3],mem[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm0 = xmm14[2,2,2,2]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm11[0,1,2],xmm0[3]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm12[2],xmm10[2],xmm12[3],xmm10[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm6 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm7 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm5 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm7[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm5[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm4 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm9 = ymm4[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm11 = ymm3[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm9 = ymm11[0,1,2,3,4],ymm9[5],ymm11[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3,4,5],ymm8[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm8[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm6 = xmm12[2,2,2,2]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm6 = xmm9[0,1,2],xmm6[3]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm8 = xmm10[2],xmm13[2],xmm10[3],xmm13[3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm8[0,1],xmm6[2,3]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm13 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm8 = mem[0,1,1,3,4,5,5,7]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm6 = ymm13[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm11 = ymm8[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm11[0,1,2,3,4,5,6],ymm6[7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm11 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm6 = mem[3,1,2,3,7,5,6,7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm11[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm6[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3,4],ymm0[5],ymm15[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm15 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm14[2],xmm0[3],xmm14[3]
; AVX2-ONLY-NEXT:    vpshufd $238, (%rsp), %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm1 = mem[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm2 = mem[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm5[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm4[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; AVX2-ONLY-NEXT:    vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm2 = mem[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm3 = mem[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT:    vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX2-ONLY-NEXT:    vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm4 = mem[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5],ymm4[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm9[2],xmm12[2],xmm9[3],xmm12[3]
; AVX2-ONLY-NEXT:    vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm3 = mem[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd {{.*#+}} xmm4 = xmm10[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm13[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm8[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm11[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm6[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5],ymm5[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX2-ONLY-NEXT:    vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
; AVX2-ONLY-NEXT:    vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm4 = mem[2,3,2,3]
; AVX2-ONLY-NEXT:    vpshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX2-ONLY-NEXT:    # xmm5 = mem[3,3,3,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
; AVX2-ONLY-NEXT:    vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm4 = mem[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX2-ONLY-NEXT:    vpshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm5 = mem[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5,6],ymm4[7]
; AVX2-ONLY-NEXT:    vpshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm5 = mem[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX2-ONLY-NEXT:    vpshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-ONLY-NEXT:    # ymm6 = mem[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5],ymm6[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 64(%rsi)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, (%rsi)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 96(%rsi)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 32(%rsi)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 64(%rdx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, (%rdx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 96(%rdx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 32(%rdx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 64(%rcx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, (%rcx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 96(%rcx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 32(%rcx)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 64(%r8)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, (%r8)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 96(%r8)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 32(%r8)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 64(%r9)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, (%r9)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 96(%r9)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 32(%r9)
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 64(%rax)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, (%rax)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 96(%rax)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 32(%rax)
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT:    vmovdqa %ymm15, 64(%rax)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, (%rax)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 96(%rax)
; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT:    vmovaps %ymm4, 32(%rax)
; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT:    vmovdqa %ymm3, 96(%rax)
; AVX2-ONLY-NEXT:    vmovdqa %ymm2, 64(%rax)
; AVX2-ONLY-NEXT:    vmovdqa %ymm0, 32(%rax)
; AVX2-ONLY-NEXT:    vmovdqa %ymm1, (%rax)
; AVX2-ONLY-NEXT:    addq $2408, %rsp # imm = 0x968
; AVX2-ONLY-NEXT:    vzeroupper
; AVX2-ONLY-NEXT:    retq
;
; AVX512F-SLOW-LABEL: load_i16_stride8_vf64:
; AVX512F-SLOW:       # %bb.0:
; AVX512F-SLOW-NEXT:    subq $2408, %rsp # imm = 0x968
; AVX512F-SLOW-NEXT:    vmovdqa 368(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 352(%rdi), %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX512F-SLOW-NEXT:    vmovdqa 336(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 320(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; AVX512F-SLOW-NEXT:    vpbroadcastq {{.*#+}} xmm4 = [0,4,0,4]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0,1,2],xmm2[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpermt2d %xmm0, %xmm4, %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa 304(%rdi), %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 288(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 272(%rdi), %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 256(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm2, %xmm31
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqa 480(%rdi), %ymm1
; AVX512F-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vmovdqa 448(%rdi), %ymm2
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT:    vmovdqa 416(%rdi), %ymm2
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm2[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 384(%rdi), %ymm2
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm2[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm3[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm5[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    movb $-64, %al
; AVX512F-SLOW-NEXT:    kmovw %eax, %k1
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vmovdqa 112(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 96(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; AVX512F-SLOW-NEXT:    vmovdqa 80(%rdi), %xmm3
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 64(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm1[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm3 = xmm2[0,1,2],xmm3[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpermt2d %xmm1, %xmm4, %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm4, %xmm6
; AVX512F-SLOW-NEXT:    vmovdqa (%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 16(%rdi), %xmm3
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 32(%rdi), %xmm4
; AVX512F-SLOW-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 48(%rdi), %xmm5
; AVX512F-SLOW-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm3, %xmm17
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa 224(%rdi), %ymm2
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm2[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 192(%rdi), %ymm2
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm3[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm4[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT:    vmovdqa 160(%rdi), %ymm3
; AVX512F-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm24 = ymm3[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqa 128(%rdi), %ymm3
; AVX512F-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm22 = ymm3[0,1,0,2]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm24[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm12 = ymm22[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm12[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5],ymm4[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 880(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 864(%rdi), %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; AVX512F-SLOW-NEXT:    vmovdqa 848(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 832(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0,1,2],xmm2[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpermt2d %xmm0, %xmm6, %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa %xmm6, %xmm10
; AVX512F-SLOW-NEXT:    vmovdqa 816(%rdi), %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 800(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX512F-SLOW-NEXT:    vmovdqa 784(%rdi), %xmm0
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 768(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm2, %xmm19
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm3, %xmm20
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqa 992(%rdi), %ymm1
; AVX512F-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 960(%rdi), %ymm1
; AVX512F-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm7 = ymm3[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm6 = ymm2[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm6[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT:    vmovdqa 928(%rdi), %ymm2
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm2[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 896(%rdi), %ymm2
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm18 = ymm2[0,1,0,2]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm4 = ymm3[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm4[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm18[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm9 = ymm3[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm9[0,1,2,3,4],ymm2[5],ymm9[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vmovdqa 624(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 608(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; AVX512F-SLOW-NEXT:    vmovdqa 592(%rdi), %xmm5
; AVX512F-SLOW-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 576(%rdi), %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm9 = xmm1[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm5 = xmm2[0,1,2],xmm9[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm5, (%rsp) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm16 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpermt2d %xmm1, %xmm10, %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa 560(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 544(%rdi), %xmm5
; AVX512F-SLOW-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm8 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
; AVX512F-SLOW-NEXT:    vmovdqa 528(%rdi), %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa 512(%rdi), %xmm5
; AVX512F-SLOW-NEXT:    vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpcklwd {{.*#+}} xmm15 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm15[0],xmm8[0],xmm15[1],xmm8[1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm9 = xmm1[0,1],xmm2[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa 736(%rdi), %ymm1
; AVX512F-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm29 = ymm1[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqa 704(%rdi), %ymm1
; AVX512F-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm29[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm10 = ymm2[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm11 = ymm1[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm13 = ymm11[0,1,2,3,4,5,6],ymm10[7]
; AVX512F-SLOW-NEXT:    vmovdqa 672(%rdi), %ymm5
; AVX512F-SLOW-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm28 = ymm5[0,1,0,2]
; AVX512F-SLOW-NEXT:    vmovdqa 640(%rdi), %ymm5
; AVX512F-SLOW-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq {{.*#+}} ymm30 = ymm5[0,1,0,2]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm10 = ymm28[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm14 = ymm10[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm11 = ymm30[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm11[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm14[5],ymm5[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm13[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3],ymm5[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm5, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm31[1,1,1,1]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm14[1],xmm0[2,3]
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vpshufhw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm5 = mem[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vpshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm9 = mem[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3,4,5,6],ymm5[7]
; AVX512F-SLOW-NEXT:    vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm9 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm13 = mem[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm9 = ymm13[0,1,2,3,4],ymm9[5],ymm13[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3,4,5],ymm5[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm5, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vpshufhw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm5 = mem[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vpshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm9 = mem[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3,4,5,6],ymm5[7]
; AVX512F-SLOW-NEXT:    vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm9 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm12[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm9 = ymm13[0,1,2,3,4],ymm9[5],ymm13[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3,4,5],ymm5[6,7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm9 = xmm17[1,1,1,1]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm9 = xmm9[0],xmm13[1],xmm9[2,3]
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm9 = xmm9[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3],ymm5[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm5, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm19[1,1,1,1]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm20, %xmm9
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm9[1],xmm0[2,3]
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm7[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm6 = ymm6[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5,6],ymm5[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm4[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm15[1,1,1,1]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm8, %xmm5
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0],xmm8[1],xmm3[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm16, %xmm4
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm10[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm11[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4],ymm2[5],ymm4[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm31[2],xmm14[2],xmm31[3],xmm14[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm14, %xmm16
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm2 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm2, %ymm27
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm3 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm3[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm3, %ymm26
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm3 = mem[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm3[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm3, %ymm23
; AVX512F-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm4 = mem[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm4[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm4, %ymm21
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm2 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm2, %ymm25
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm14 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm14[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm12 = ymm24[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm12[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm11 = ymm22[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm11[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm17[2],xmm13[2],xmm17[3],xmm13[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm13, %xmm22
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm2 = xmm2[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm19[2],xmm20[2],xmm19[3],xmm20[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm20, %xmm24
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm20
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm10 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm10[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm9 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm9[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm8 = mem[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm8[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm6 = ymm18[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm6[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5],ymm2[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm20 {%k1}
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm15[2],xmm5[2],xmm15[3],xmm5[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm5, %xmm18
; AVX512F-SLOW-NEXT:    vpblendd $12, (%rsp), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm3 = xmm0[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm7 = ymm29[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm5 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm5[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm4 = ymm28[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm4[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm30[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm2[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3,4],ymm0[5],ymm13[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm20, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vpbroadcastq {{.*#+}} xmm0 = [3,7,3,7]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm31, %xmm1
; AVX512F-SLOW-NEXT:    vpermt2d %xmm16, %xmm0, %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm0, %xmm16
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm0 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm0 = xmm1[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm27, %ymm1
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm26, %ymm3
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm23, %ymm3
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm21, %ymm13
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm13[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm13[0,1,2,3,4],ymm3[5],ymm13[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm17, %xmm1
; AVX512F-SLOW-NEXT:    vpermt2d %xmm22, %xmm16, %xmm1
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm1 = xmm1[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm25, %ymm3
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm13 = ymm14[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm13[0,1,2,3,4,5,6],ymm3[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm12[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm11 = ymm11[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm12[5],ymm11[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm11[0,1,2,3,4,5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm19, %xmm0
; AVX512F-SLOW-NEXT:    vpermt2d %xmm24, %xmm16, %xmm0
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm9[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm8[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm6[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3,4],ymm3[5],ymm6[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vpermt2d %xmm18, %xmm16, %xmm15
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm1 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm1 = xmm15[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm7[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm5[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5,6],ymm3[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm4[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5],ymm2[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm5 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm6 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm5[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm6[0,1,2],xmm0[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm6[2],xmm5[2],xmm6[3],xmm5[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpbroadcastq {{.*#+}} xmm10 = [0,4,0,4]
; AVX512F-SLOW-NEXT:    vpermt2d %xmm5, %xmm10, %xmm6
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm2, %xmm28
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm3 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm1[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm3 = xmm2[0,1,2],xmm3[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm3 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm3 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpermt2d %xmm1, %xmm10, %xmm2
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm3 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm3 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm4 = ymm3[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm4[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm4, %ymm17
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm30 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm30 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm4 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm9 = ymm30[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm9[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm7 = ymm4[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm7[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5],ymm4[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm1[0,1,2],xmm2[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm23 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-SLOW-NEXT:    vpermt2d %xmm0, %xmm10, %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm3, %xmm16
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm2, %xmm20
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm12
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm0 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm6 = ymm0[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm5 = ymm1[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm5[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm22 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm22 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm29 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm29 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm4 = ymm22[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm4[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm29[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm8 = ymm3[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4],ymm1[5],ymm8[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm12 {%k1}
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm8 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm8 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm11 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm11 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm13 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-SLOW-NEXT:    vmovdqa %xmm10, %xmm1
; AVX512F-SLOW-NEXT:    vpermi2d %xmm2, %xmm8, %xmm1
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm8, %xmm19
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm2, %xmm24
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm13[0],xmm11[0],xmm13[1],xmm11[1]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm13, %xmm31
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm11, %xmm21
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm14 = xmm0[0,1],xmm1[2,3]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm0 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm1[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm8 = ymm2[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm0[0,1,2,0,4,5,6,4]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm11 = ymm1[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm15 = ymm11[0,1,2,3,4,5,6],ymm8[7]
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm0 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm10 = mem[0,1,1,3]
; AVX512F-SLOW-NEXT:    vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm8 = ymm0[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm8[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm11 = ymm10[0,2,2,3,4,6,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm11[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3,4],ymm0[5],ymm13[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm12, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm28[1,1,1,1]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm10[1],xmm0[2,3]
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vpshufhw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm12 = mem[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vpshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm13 = mem[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5,6],ymm12[7]
; AVX512F-SLOW-NEXT:    vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm13 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm14 = mem[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5],ymm14[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm12, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vpshufhw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm12 = mem[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm17, %ymm13
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm13 = ymm13[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5,6],ymm12[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm9 = ymm9[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm7[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm9[5],ymm7[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm12[6,7]
; AVX512F-SLOW-NEXT:    vmovdqa64 {{[-0-9]+}}(%r{{[sb]}}p), %xmm17 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm9 = xmm17[1,1,1,1]
; AVX512F-SLOW-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm9 = xmm9[0],xmm15[1],xmm9[2,3]
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm9 = xmm9[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm7, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm16[1,1,1,1]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm20, %xmm7
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm7[1],xmm0[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm23, %xmm9
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm9[2,3]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm5[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm4[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm31[1,1,1,1]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm21, %xmm6
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0],xmm6[1],xmm3[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm19, %xmm5
; AVX512F-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm19[0],xmm24[0],xmm19[1],xmm24[1]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm8[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm11[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4],ymm2[5],ymm4[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm28[2],xmm10[2],xmm28[3],xmm10[3]
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm2 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm2, %ymm27
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm3 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm3[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm3, %ymm26
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm3 = mem[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm3[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm3, %ymm25
; AVX512F-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm13 = mem[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm13[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm2 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm2, %ymm21
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm14 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm14[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm12 = ymm30[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm12[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm11 = mem[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm11[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm17[2],xmm15[2],xmm17[3],xmm15[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm15, %xmm23
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm2 = xmm2[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm16[2],xmm20[2],xmm16[3],xmm20[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm16, %xmm19
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm20, %xmm30
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm0 = xmm0[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm20
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm10 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm10[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm9 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm9[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm8 = ymm22[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm8[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} ymm7 = ymm29[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm7[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm1[5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm20 {%k1}
; AVX512F-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm24[2,2,2,2]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm5[0,1,2],xmm0[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm5, %xmm22
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm31, %xmm4
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm31[2],xmm6[2],xmm31[3],xmm6[3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm6, %xmm29
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm3 = xmm1[0,1],xmm0[2,3]
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm0 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm0, %ymm18
; AVX512F-SLOW-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm5 = mem[0,1,1,3,4,5,5,7]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm5[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm2 = mem[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm2[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm2, %ymm31
; AVX512F-SLOW-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-SLOW-NEXT:    # ymm2 = mem[3,1,2,3,7,5,6,7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm2[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3,4],ymm0[5],ymm15[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm20, %zmm0
; AVX512F-SLOW-NEXT:    vpbroadcastq {{.*#+}} xmm16 = [3,7,3,7]
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm28, %xmm1
; AVX512F-SLOW-NEXT:    vpermt2d {{[-0-9]+}}(%r{{[sb]}}p), %xmm16, %xmm1 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm1 = xmm1[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm1, %zmm0, %zmm1
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm27, %ymm3
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm26, %ymm6
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm15 = ymm6[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm15[0,1,2,3,4,5,6],ymm3[7]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm25, %ymm6
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm6[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm13[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm15[5],ymm13[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm13[0,1,2,3,4,5],ymm3[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm1 {%k1}
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm17, %xmm3
; AVX512F-SLOW-NEXT:    vpermt2d %xmm23, %xmm16, %xmm3
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm3 = xmm3[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm21, %ymm6
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm13 = ymm6[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm14 = ymm14[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5,6],ymm13[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm12[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm11 = ymm11[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm12[5],ymm11[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm13[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm11[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm3, %zmm1, %zmm1
; AVX512F-SLOW-NEXT:    vmovdqa64 %xmm19, %xmm3
; AVX512F-SLOW-NEXT:    vpermt2d %xmm30, %xmm16, %xmm3
; AVX512F-SLOW-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX512F-SLOW-NEXT:    # xmm3 = xmm3[0,1],mem[2,3]
; AVX512F-SLOW-NEXT:    vinserti32x4 $2, %xmm3, %zmm0, %zmm3
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm9 = ymm9[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5,6],ymm10[7]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm8 = ymm8[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm7[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm8[5],ymm6[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm9[6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $1, %ymm6, %zmm0, %zmm3 {%k1}
; AVX512F-SLOW-NEXT:    vpermt2d %xmm29, %xmm16, %xmm4
; AVX512F-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm6 = xmm22[2],xmm24[2],xmm22[3],xmm24[3]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} xmm6 = xmm4[0,1],xmm6[2,3]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm18, %ymm7
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-SLOW-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm5[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm7[7]
; AVX512F-SLOW-NEXT:    vmovdqa64 %ymm31, %ymm4
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm4[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5],ymm2[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm5[6,7]
; AVX512F-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-SLOW-NEXT:    vinserti64x4 $0, %ymm2, %zmm3, %zmm2
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm3, 64(%rsi)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm3, (%rsi)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm3, 64(%rdx)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm3, (%rdx)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm3, 64(%rcx)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm3, (%rcx)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm3, 64(%r8)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm3, (%r8)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm3, 64(%r9)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm3, (%r9)
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm3, 64(%rax)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm3, (%rax)
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT:    vmovdqa64 %zmm0, 64(%rax)
; AVX512F-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512F-SLOW-NEXT:    vmovaps %zmm0, (%rax)
; AVX512F-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT:    vmovdqa64 %zmm2, 64(%rax)
; AVX512F-SLOW-NEXT:    vmovdqa64 %zmm1, (%rax)
; AVX512F-SLOW-NEXT:    addq $2408, %rsp # imm = 0x968
; AVX512F-SLOW-NEXT:    vzeroupper
; AVX512F-SLOW-NEXT:    retq
;
; AVX512F-FAST-LABEL: load_i16_stride8_vf64:
; AVX512F-FAST:       # %bb.0:
; AVX512F-FAST-NEXT:    subq $2312, %rsp # imm = 0x908
; AVX512F-FAST-NEXT:    vmovdqa 368(%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 352(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-FAST-NEXT:    vmovdqa 336(%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 320(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm14 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [0,4,0,4]
; AVX512F-FAST-NEXT:    vmovdqa %xmm14, %xmm0
; AVX512F-FAST-NEXT:    vpermt2d %xmm2, %xmm1, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm2, %xmm24
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, %xmm5
; AVX512F-FAST-NEXT:    vmovdqa 304(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 288(%rdi), %xmm2
; AVX512F-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-FAST-NEXT:    vmovdqa 272(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 256(%rdi), %xmm2
; AVX512F-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm2, %xmm30
; AVX512F-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm17
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vmovdqa 480(%rdi), %ymm1
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vmovdqa 448(%rdi), %ymm2
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vmovdqa 416(%rdi), %ymm2
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm3 = ymm2[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 384(%rdi), %ymm2
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm2 = ymm3[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm3 = ymm4[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    movb $-64, %al
; AVX512F-FAST-NEXT:    kmovw %eax, %k1
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa 112(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 96(%rdi), %xmm2
; AVX512F-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-FAST-NEXT:    vmovdqa 80(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 64(%rdi), %xmm2
; AVX512F-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-FAST-NEXT:    vmovdqa %xmm2, %xmm1
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm2, %xmm20
; AVX512F-FAST-NEXT:    vpermt2d %xmm3, %xmm5, %xmm1
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm19
; AVX512F-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa (%rdi), %xmm2
; AVX512F-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 16(%rdi), %xmm3
; AVX512F-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 32(%rdi), %xmm4
; AVX512F-FAST-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 48(%rdi), %xmm6
; AVX512F-FAST-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1],xmm4[2],xmm6[2],xmm4[3],xmm6[3]
; AVX512F-FAST-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm31
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vmovdqa 224(%rdi), %ymm2
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm3 = ymm2[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 192(%rdi), %ymm2
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm3, %ymm28
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm10 = ymm4[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm10[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-FAST-NEXT:    vmovdqa 160(%rdi), %ymm3
; AVX512F-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm4 = ymm3[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 128(%rdi), %ymm3
; AVX512F-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm6 = ymm3[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm8 = ymm4[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm8[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm7 = ymm6[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm7[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5],ymm4[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 880(%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 864(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-FAST-NEXT:    vmovdqa 848(%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 832(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm1, %xmm26
; AVX512F-FAST-NEXT:    vmovdqa %xmm5, %xmm2
; AVX512F-FAST-NEXT:    vpermt2d %xmm3, %xmm5, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm27
; AVX512F-FAST-NEXT:    vmovdqa 816(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 800(%rdi), %xmm3
; AVX512F-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
; AVX512F-FAST-NEXT:    vmovdqa 784(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 768(%rdi), %xmm3
; AVX512F-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm23
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm4, %xmm25
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm9
; AVX512F-FAST-NEXT:    vmovdqa 992(%rdi), %ymm0
; AVX512F-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, (%rsp) # 32-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 960(%rdi), %ymm0
; AVX512F-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm3 = ymm0[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm6 = ymm1[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm5 = ymm3[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm5[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-FAST-NEXT:    vmovdqa 928(%rdi), %ymm1
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 896(%rdi), %ymm1
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm11 = ymm1[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm4 = ymm3[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm4[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm3 = ymm11[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm3[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm12[0,1,2,3,4],ymm1[5],ymm12[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm9 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa 624(%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 608(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm11 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-FAST-NEXT:    vmovdqa 592(%rdi), %xmm0
; AVX512F-FAST-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 576(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm1, %xmm29
; AVX512F-FAST-NEXT:    vpermt2d %xmm11, %xmm2, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm11, %xmm16
; AVX512F-FAST-NEXT:    vmovdqa 560(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 544(%rdi), %xmm2
; AVX512F-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm11 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-FAST-NEXT:    vmovdqa 528(%rdi), %xmm1
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 512(%rdi), %xmm2
; AVX512F-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm11[0],xmm2[1],xmm11[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm2, %xmm22
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm11, %xmm21
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm12 = xmm1[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vmovdqa 736(%rdi), %ymm0
; AVX512F-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 704(%rdi), %ymm0
; AVX512F-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm1, %ymm18
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm13 = ymm2[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-FAST-NEXT:    vmovdqa 672(%rdi), %ymm0
; AVX512F-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa 640(%rdi), %ymm0
; AVX512F-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq {{.*#+}} ymm11 = ymm0[0,1,0,2]
; AVX512F-FAST-NEXT:    vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm15 = ymm1[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm15[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm1 = ymm11[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm11 = ymm1[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm11[0,1,2,3,4],ymm0[5],ymm11[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm13[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm9, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm9 = [1,5,1,5]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm30, %xmm0
; AVX512F-FAST-NEXT:    vpermt2d %xmm17, %xmm9, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm17, %xmm30
; AVX512F-FAST-NEXT:    vmovdqa %xmm9, %xmm13
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm9 = xmm14[0],xmm24[0],xmm14[1],xmm24[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm9[2,3]
; AVX512F-FAST-NEXT:    vpshufhw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm9 = mem[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vpshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm11 = mem[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5,6],ymm9[7]
; AVX512F-FAST-NEXT:    vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm11 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm12 = mem[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm11 = ymm12[0,1,2,3,4],ymm11[5],ymm12[6,7]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5],ymm9[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm9, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm31, %xmm12
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm31, %xmm9
; AVX512F-FAST-NEXT:    vmovdqa64 {{[-0-9]+}}(%r{{[sb]}}p), %xmm17 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpermt2d %xmm17, %xmm13, %xmm9
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm11 = xmm20[0],xmm19[0],xmm20[1],xmm19[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm9 = xmm9[0,1],xmm11[2,3]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm28, %ymm11
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm10 = ymm10[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,6],ymm11[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm8 = ymm8[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm7[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm8[5],ymm7[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm10[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm7, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm23, %xmm9
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm23, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm25, %xmm11
; AVX512F-FAST-NEXT:    vpermt2d %xmm25, %xmm13, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm27, %xmm8
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm7 = xmm26[0],xmm27[0],xmm26[1],xmm27[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm26, %xmm19
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm7[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm5[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm4[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm22, %xmm3
; AVX512F-FAST-NEXT:    vpermt2d %xmm21, %xmm13, %xmm3
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm29[0],xmm16[0],xmm29[1],xmm16[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm29, %xmm6
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm18, %ymm4
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm4[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm15[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm4[5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa %xmm14, %xmm0
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm14[2],xmm24[2],xmm14[3],xmm24[3]
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [2,6,2,6]
; AVX512F-FAST-NEXT:    vpermt2d %xmm24, %xmm1, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, %xmm5
; AVX512F-FAST-NEXT:    vmovdqa64 {{[-0-9]+}}(%r{{[sb]}}p), %xmm31 # 16-byte Reload
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm30, %xmm28
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm31[2],xmm30[2],xmm31[3],xmm30[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm1 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm3 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm3[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm3, %ymm29
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm3 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm3[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm3, %ymm30
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm4 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm4[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm4, %ymm27
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm20, %xmm1
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm26 = xmm20[2],xmm2[2],xmm20[3],xmm2[3]
; AVX512F-FAST-NEXT:    vpermt2d %xmm2, %xmm5, %xmm1
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm12[2],xmm17[2],xmm12[3],xmm17[3]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm12, %xmm25
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm17, %xmm23
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm3 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm3, %ymm18
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm13 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm13[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm12 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm12[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm10 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm10[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5],ymm4[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm19, %xmm0
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm17 = xmm19[2],xmm8[2],xmm19[3],xmm8[3]
; AVX512F-FAST-NEXT:    vpermt2d %xmm8, %xmm5, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm9, %xmm19
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm11, %xmm20
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm9[2],xmm11[2],xmm9[3],xmm11[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm24
; AVX512F-FAST-NEXT:    vpshufd $212, (%rsp), %ymm9 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm9 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm9[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm11 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm11[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm8 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm8[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm7 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm7[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm24 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa %xmm6, %xmm0
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm14 = xmm6[2],xmm16[2],xmm6[3],xmm16[3]
; AVX512F-FAST-NEXT:    vpermt2d %xmm16, %xmm5, %xmm0
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm22[2],xmm21[2],xmm22[3],xmm21[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm3 = xmm1[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm6 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm5 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm5[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm4 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm4[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm2 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm2[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3,4],ymm0[5],ymm15[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm24, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm0 = [3,7,3,7]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm31, %xmm1
; AVX512F-FAST-NEXT:    vpermt2d %xmm28, %xmm0, %xmm1
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm0, %xmm16
; AVX512F-FAST-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm0 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm0 = xmm1[0,1],mem[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm1 = mem[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm29, %ymm3
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm30, %ymm3
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm27, %ymm15
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm15[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm15[0,1,2,3,4],ymm3[5],ymm15[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm25, %xmm1
; AVX512F-FAST-NEXT:    vpermt2d %xmm23, %xmm16, %xmm1
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm26, %xmm3
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm18, %ymm3
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm13 = ymm13[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm13[0,1,2,3,4,5,6],ymm3[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm12 = ymm12[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm10 = ymm10[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm12[5],ymm10[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm10[0,1,2,3,4,5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm19, %xmm0
; AVX512F-FAST-NEXT:    vpermt2d %xmm20, %xmm16, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm17, %xmm1
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm9[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm11[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm8[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm7[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm7[0,1,2,3,4],ymm3[5],ymm7[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm22, %xmm1
; AVX512F-FAST-NEXT:    vpermt2d %xmm21, %xmm16, %xmm1
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm14[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm6[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm5[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5,6],ymm3[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm4[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm4 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm9 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm9 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa %xmm9, %xmm0
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm5 = [0,4,0,4]
; AVX512F-FAST-NEXT:    vpermt2d %xmm4, %xmm5, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm4, %xmm29
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm23
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm2, %xmm19
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm3 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm6 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm6 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm2 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm3 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm4 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm4 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa %xmm2, %xmm1
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm2, %xmm27
; AVX512F-FAST-NEXT:    vpermt2d %xmm6, %xmm5, %xmm1
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm6, %xmm18
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm4, %xmm22
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm24
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm3 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm4 = ymm3[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm4[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm4, %ymm26
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm3 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm4 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm11 = ymm3[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm11[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm10 = ymm4[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm10[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5],ymm4[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm4 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm1 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa %xmm1, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm1, %xmm31
; AVX512F-FAST-NEXT:    vpermt2d %xmm4, %xmm5, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm4, %xmm20
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm21
; AVX512F-FAST-NEXT:    vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm2, %xmm25
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm12
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm0 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm14 = ymm0[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm14[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm8 = ymm1[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm8[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm7 = ymm1[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm7[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm6 = ymm2[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm6[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm12 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm13 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm4 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX512F-FAST-NEXT:    vmovdqa %xmm5, %xmm1
; AVX512F-FAST-NEXT:    vpermi2d %xmm2, %xmm3, %xmm1
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm3, %xmm17
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm2, %xmm16
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm4[0],xmm13[0],xmm4[1],xmm13[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm13, %xmm28
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm4, %xmm30
; AVX512F-FAST-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm13 = xmm0[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm0 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm1 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm5 = ymm0[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5,6,4,8,9,10,11,12,13,14,12]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm4 = ymm1[0,1,2,0,4,5,6,4]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm4[0,1,2,3,6,4,6,7,8,9,10,11,14,12,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm0 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm2 = mem[0,1,1,3]
; AVX512F-FAST-NEXT:    vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm3 = ymm0[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm3[0,1,0,2,4,5,6,7,8,9,8,10,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm15 = ymm2[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3,4],ymm0[5],ymm15[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm12, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm23, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm23, (%rsp) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm15 = [1,5,1,5]
; AVX512F-FAST-NEXT:    vpermt2d %xmm19, %xmm15, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm9[0],xmm29[0],xmm9[1],xmm29[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vpshufhw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm1 = mem[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vpshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm12 = mem[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm12 = mem[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm13 = mem[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3,4],ymm12[5],ymm13[6,7]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm22, %xmm1
; AVX512F-FAST-NEXT:    vpermt2d %xmm24, %xmm15, %xmm1
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm12 = xmm27[0],xmm18[0],xmm27[1],xmm18[1]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm12[2,3]
; AVX512F-FAST-NEXT:    vpshufhw $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm12 = mem[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm26, %ymm13
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm13 = ymm13[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5,6],ymm12[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm11 = ymm11[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm10 = ymm10[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm11[5],ymm10[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm12[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm21, %xmm0
; AVX512F-FAST-NEXT:    vpermt2d %xmm25, %xmm15, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm20, %xmm13
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm31[0],xmm20[0],xmm31[1],xmm20[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm31, %xmm21
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm8 = ymm8[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm7 = ymm7[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm6[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5],ymm6[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa %xmm15, %xmm6
; AVX512F-FAST-NEXT:    vpermi2d %xmm28, %xmm30, %xmm6
; AVX512F-FAST-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm17[0],xmm16[0],xmm17[1],xmm16[1]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm16, %xmm7
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,7,5,8,9,10,11,12,13,15,13]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm4[0,1,2,3,7,5,6,7,8,9,10,11,15,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm2[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa %xmm9, %xmm0
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm9[2],xmm29[2],xmm9[3],xmm29[3]
; AVX512F-FAST-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm5 = [2,6,2,6]
; AVX512F-FAST-NEXT:    vpermt2d %xmm29, %xmm5, %xmm0
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm23[2],xmm19[2],xmm23[3],xmm19[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm2 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm2, %ymm31
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm3 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm3[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm3, %ymm30
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm3 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm2 = ymm3[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm3, %ymm26
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm4 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm4[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm4, %ymm23
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm27, %xmm1
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm20 = xmm27[2],xmm18[2],xmm27[3],xmm18[3]
; AVX512F-FAST-NEXT:    vpermt2d %xmm18, %xmm5, %xmm1
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm22[2],xmm24[2],xmm22[3],xmm24[3]
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm22, %xmm19
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm24, %xmm18
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm12 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm12[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm9 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm9[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm10 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm10[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm11 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm11[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5],ymm4[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm0, %zmm0
; AVX512F-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm21, %xmm0
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm16 = xmm21[2],xmm13[2],xmm21[3],xmm13[3]
; AVX512F-FAST-NEXT:    vpermt2d %xmm13, %xmm5, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm25, %xmm22
; AVX512F-FAST-NEXT:    vmovdqa64 {{[-0-9]+}}(%r{{[sb]}}p), %xmm24 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm24[2],xmm25[2],xmm24[3],xmm25[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm0, %zmm0, %zmm21
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm14 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm14[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm15 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm15[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm6 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm6[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm4 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm4[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm1[5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm21 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa %xmm5, %xmm1
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm7, %xmm25
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm17, %xmm27
; AVX512F-FAST-NEXT:    vpermi2d %xmm7, %xmm17, %xmm1
; AVX512F-FAST-NEXT:    vmovdqa64 {{[-0-9]+}}(%r{{[sb]}}p), %xmm29 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm29[2],xmm28[2],xmm29[3],xmm28[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm2 = xmm0[0,1],xmm1[2,3]
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm8 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
; AVX512F-FAST-NEXT:    vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm7 = mem[0,1,1,3,4,5,5,7]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm3 = ymm7[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm5 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm5[0,1,2,0,4,5,6,7,8,9,10,8,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX512F-FAST-NEXT:    # ymm3 = mem[3,1,2,3,7,5,6,7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm3[2,0,2,3,4,5,6,7,10,8,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3,4],ymm0[5],ymm13[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm21, %zmm21
; AVX512F-FAST-NEXT:    vpbroadcastq {{.*#+}} xmm17 = [3,7,3,7]
; AVX512F-FAST-NEXT:    vmovdqa (%rsp), %xmm0 # 16-byte Reload
; AVX512F-FAST-NEXT:    vpermt2d {{[-0-9]+}}(%r{{[sb]}}p), %xmm17, %xmm0 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
; AVX512F-FAST-NEXT:    # xmm1 = xmm0[0,1],mem[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm1, %zmm0, %zmm1
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm31, %ymm0
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm30, %ymm0
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm13 = ymm0[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm26, %ymm0
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm13 = ymm0[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vmovdqa64 %ymm23, %ymm0
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm13[5],ymm0[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm1 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm19, %xmm0
; AVX512F-FAST-NEXT:    vpermt2d %xmm18, %xmm17, %xmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm20, %xmm2
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm12[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm9 = ymm9[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm9 = ymm10[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm10 = ymm11[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5],ymm10[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm24, %xmm1
; AVX512F-FAST-NEXT:    vpermt2d %xmm22, %xmm17, %xmm1
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm16, %xmm2
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; AVX512F-FAST-NEXT:    vinserti32x4 $2, %xmm1, %zmm0, %zmm1
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm2 = ymm14[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm9 = ymm15[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5,6],ymm2[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm6 = ymm6[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm4 = ymm4[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm6[5],ymm4[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm1 {%k1}
; AVX512F-FAST-NEXT:    vmovdqa64 %xmm29, %xmm4
; AVX512F-FAST-NEXT:    vpermt2d %xmm28, %xmm17, %xmm4
; AVX512F-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm27[2],xmm25[2],xmm27[3],xmm25[3]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm4 = ymm8[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
; AVX512F-FAST-NEXT:    vpshufhw {{.*#+}} ymm6 = ymm7[0,1,2,3,5,7,6,7,8,9,10,11,13,15,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5,6],ymm4[7]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm5 = ymm5[0,1,3,1,4,5,6,7,8,9,11,9,12,13,14,15]
; AVX512F-FAST-NEXT:    vpshuflw {{.*#+}} ymm3 = ymm3[3,1,2,3,4,5,6,7,11,9,10,11,12,13,14,15]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5],ymm3[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
; AVX512F-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX512F-FAST-NEXT:    vinserti64x4 $0, %ymm2, %zmm1, %zmm1
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, 64(%rsi)
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, (%rsi)
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, 64(%rdx)
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, (%rdx)
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, 64(%rcx)
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, (%rcx)
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, 64(%r8)
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, (%r8)
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, 64(%r9)
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, (%r9)
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, 64(%rax)
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, (%rax)
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT:    vmovdqa64 %zmm21, 64(%rax)
; AVX512F-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512F-FAST-NEXT:    vmovaps %zmm2, (%rax)
; AVX512F-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512F-FAST-NEXT:    vmovdqa64 %zmm1, 64(%rax)
; AVX512F-FAST-NEXT:    vmovdqa64 %zmm0, (%rax)
; AVX512F-FAST-NEXT:    addq $2312, %rsp # imm = 0x908
; AVX512F-FAST-NEXT:    vzeroupper
; AVX512F-FAST-NEXT:    retq
;
; AVX512BW-ONLY-SLOW-LABEL: load_i16_stride8_vf64:
; AVX512BW-ONLY-SLOW:       # %bb.0:
; AVX512BW-ONLY-SLOW-NEXT:    subq $1096, %rsp # imm = 0x448
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 704(%rdi), %zmm6
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 640(%rdi), %zmm8
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 576(%rdi), %zmm28
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 512(%rdi), %zmm11
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 832(%rdi), %zmm5
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 768(%rdi), %zmm29
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 960(%rdi), %zmm21
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 896(%rdi), %zmm25
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 (%rdi), %zmm7
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 64(%rdi), %zmm17
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 128(%rdi), %zmm4
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 192(%rdi), %zmm9
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 320(%rdi), %zmm20
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 256(%rdi), %zmm26
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 448(%rdi), %zmm22
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 384(%rdi), %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56]
; AVX512BW-ONLY-SLOW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm12
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm22, %zmm0, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm20, %zmm0, %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    movb $-64, %al
; AVX512BW-ONLY-SLOW-NEXT:    kmovd %eax, %k1
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56]
; AVX512BW-ONLY-SLOW-NEXT:    # ymm1 = mem[0,1,0,1]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm4, %zmm3
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm4, %zmm10
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm9, %zmm1, %zmm3
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm4
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm17, %zmm0, %zmm4
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm3, %zmm2, %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm21, %zmm0, %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm3
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm0, %zmm3
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm28, %zmm11, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm6, %zmm8, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm3, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm4 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512BW-ONLY-SLOW-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm12, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm22, %zmm4, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm20, %zmm4, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512BW-ONLY-SLOW-NEXT:    # ymm2 = mem[0,1,0,1]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm10, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm9, %zmm2, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm17, %zmm4, %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58]
; AVX512BW-ONLY-SLOW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm22, %zmm0, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59]
; AVX512BW-ONLY-SLOW-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm22, %zmm2, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm3
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm10 = [4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60]
; AVX512BW-ONLY-SLOW-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm12, %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm22, %zmm10, %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm15 = [5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61]
; AVX512BW-ONLY-SLOW-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm22, %zmm15, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm13 = [6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62]
; AVX512BW-ONLY-SLOW-NEXT:    # zmm13 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm22, %zmm13, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63]
; AVX512BW-ONLY-SLOW-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm22, %zmm1, %zmm12
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm24
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm20, %zmm0, %zmm24
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm27
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm20, %zmm3, %zmm27
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm31
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm20, %zmm10, %zmm31
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm30
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm20, %zmm15, %zmm30
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm22
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm20, %zmm13, %zmm22
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm20, %zmm1, %zmm26
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm8
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm17, %zmm0, %zmm8
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm12
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm17, %zmm3, %zmm12
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm6
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm17, %zmm10, %zmm6
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm6, (%rsp) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm6
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm17, %zmm15, %zmm6
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm6
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm17, %zmm13, %zmm6
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm17, %zmm1, %zmm7
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm18
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm21, %zmm4, %zmm18
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm19
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm21, %zmm0, %zmm19
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm23
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm3, %zmm6
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm21, %zmm3, %zmm23
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm16
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm21, %zmm10, %zmm16
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm20
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm21, %zmm15, %zmm20
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm17
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm21, %zmm13, %zmm17
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm21, %zmm1, %zmm25
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm9
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm4, %zmm9
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm11
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm0, %zmm11
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm3
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm14
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm6, %zmm14
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm21
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm10, %zmm21
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm7
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm15, %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm15, %zmm7
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm15
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm13, %zmm15
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm1, %zmm29
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm28, %zmm0, %zmm4
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm28, %zmm0, %zmm3
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm28, %zmm0, %zmm6
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm28, %zmm0, %zmm10
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm28, %zmm0, %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm28, %zmm0, %zmm13
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm28, %zmm1, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm18, %zmm9 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512BW-ONLY-SLOW-NEXT:    # ymm0 = mem[0,1,0,1]
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm28, %zmm2, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm9, %zmm18
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm24 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm9 = [2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58]
; AVX512BW-ONLY-SLOW-NEXT:    # ymm9 = mem[0,1,0,1]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm9, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm24, %zmm24
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm19, %zmm11 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm28, %zmm2, %zmm9
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm9[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm1, %zmm11, %zmm1
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm27 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59]
; AVX512BW-ONLY-SLOW-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm4, %zmm9
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm3, %zmm9
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm12[0,1,2,3],ymm9[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm4, %zmm27, %zmm4
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm23, %zmm14 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm28, %zmm2, %zmm3
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm3, %zmm14, %zmm3
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm6 = [4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60]
; AVX512BW-ONLY-SLOW-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm8
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm6, %zmm8
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm9 = [5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61]
; AVX512BW-ONLY-SLOW-NEXT:    # ymm9 = mem[0,1,0,1]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm11
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm9, %zmm11
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm12 = [6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62]
; AVX512BW-ONLY-SLOW-NEXT:    # ymm12 = mem[0,1,0,1]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm14
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm12, %zmm14
; AVX512BW-ONLY-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} ymm19 = [7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63]
; AVX512BW-ONLY-SLOW-NEXT:    # ymm19 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm5, %zmm19, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm5
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm28, %zmm2, %zmm6
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm28, %zmm2, %zmm9
; AVX512BW-ONLY-SLOW-NEXT:    vpermi2w %zmm28, %zmm2, %zmm12
; AVX512BW-ONLY-SLOW-NEXT:    vpermt2w %zmm28, %zmm19, %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd $15, (%rsp), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX512BW-ONLY-SLOW-NEXT:    # ymm8 = mem[0,1,2,3],ymm8[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm31 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm8, %zmm31, %zmm8
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm6[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm16, %zmm21 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm6, %zmm21, %zmm6
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm10 # 32-byte Folded Reload
; AVX512BW-ONLY-SLOW-NEXT:    # ymm10 = mem[0,1,2,3],ymm11[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm30 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm10, %zmm30, %zmm10
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
; AVX512BW-ONLY-SLOW-NEXT:    # ymm9 = mem[0,1,2,3],ymm9[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm20, %zmm7 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm9, %zmm7, %zmm7
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm22 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm9 # 32-byte Folded Reload
; AVX512BW-ONLY-SLOW-NEXT:    # ymm9 = mem[0,1,2,3],ymm14[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm9, %zmm22, %zmm9
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm17, %zmm15 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3],ymm12[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm2, %zmm15, %zmm2
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm11, %zmm26 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm11 # 32-byte Folded Reload
; AVX512BW-ONLY-SLOW-NEXT:    # ymm11 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm11, %zmm26, %zmm5
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm29 {%k1}
; AVX512BW-ONLY-SLOW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
; AVX512BW-ONLY-SLOW-NEXT:    # ymm11 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-SLOW-NEXT:    vinserti64x4 $0, %ymm11, %zmm29, %zmm11
; AVX512BW-ONLY-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vmovaps %zmm0, 64(%rsi)
; AVX512BW-ONLY-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vmovaps %zmm0, (%rsi)
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm18, 64(%rdx)
; AVX512BW-ONLY-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-ONLY-SLOW-NEXT:    vmovaps %zmm0, (%rdx)
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm3, 64(%r8)
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm4, (%r8)
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm6, 64(%r9)
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm8, (%r9)
; AVX512BW-ONLY-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm7, 64(%rax)
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm10, (%rax)
; AVX512BW-ONLY-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm2, 64(%rax)
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm9, (%rax)
; AVX512BW-ONLY-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm11, 64(%rax)
; AVX512BW-ONLY-SLOW-NEXT:    vmovdqa64 %zmm5, (%rax)
; AVX512BW-ONLY-SLOW-NEXT:    addq $1096, %rsp # imm = 0x448
; AVX512BW-ONLY-SLOW-NEXT:    vzeroupper
; AVX512BW-ONLY-SLOW-NEXT:    retq
;
; AVX512BW-ONLY-FAST-LABEL: load_i16_stride8_vf64:
; AVX512BW-ONLY-FAST:       # %bb.0:
; AVX512BW-ONLY-FAST-NEXT:    subq $1096, %rsp # imm = 0x448
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 704(%rdi), %zmm6
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 640(%rdi), %zmm8
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 576(%rdi), %zmm28
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 512(%rdi), %zmm11
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 832(%rdi), %zmm5
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 768(%rdi), %zmm29
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 960(%rdi), %zmm21
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 896(%rdi), %zmm25
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 (%rdi), %zmm7
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 64(%rdi), %zmm17
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 128(%rdi), %zmm4
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 192(%rdi), %zmm9
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 320(%rdi), %zmm20
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 256(%rdi), %zmm26
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 448(%rdi), %zmm22
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 384(%rdi), %zmm2
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56]
; AVX512BW-ONLY-FAST-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm2, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm2, %zmm12
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm22, %zmm0, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm20, %zmm0, %zmm2
; AVX512BW-ONLY-FAST-NEXT:    movb $-64, %al
; AVX512BW-ONLY-FAST-NEXT:    kmovd %eax, %k1
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56]
; AVX512BW-ONLY-FAST-NEXT:    # ymm1 = mem[0,1,0,1]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm4, %zmm3
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm4, %zmm10
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm9, %zmm1, %zmm3
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm7, %zmm4
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm17, %zmm0, %zmm4
; AVX512BW-ONLY-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm3, %zmm2, %zmm2
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm25, %zmm2
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm21, %zmm0, %zmm2
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm29, %zmm3
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm0, %zmm3
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm28, %zmm11, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm6, %zmm8, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm3, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm4 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512BW-ONLY-FAST-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm12, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm22, %zmm4, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm26, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm20, %zmm4, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512BW-ONLY-FAST-NEXT:    # ymm2 = mem[0,1,0,1]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm10, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm9, %zmm2, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm7, %zmm2
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm17, %zmm4, %zmm2
; AVX512BW-ONLY-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58]
; AVX512BW-ONLY-FAST-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm22, %zmm0, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59]
; AVX512BW-ONLY-FAST-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm22, %zmm2, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm2, %zmm3
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm10 = [4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60]
; AVX512BW-ONLY-FAST-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm12, %zmm2
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm22, %zmm10, %zmm2
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm15 = [5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61]
; AVX512BW-ONLY-FAST-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm22, %zmm15, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm13 = [6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62]
; AVX512BW-ONLY-FAST-NEXT:    # zmm13 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm22, %zmm13, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63]
; AVX512BW-ONLY-FAST-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm22, %zmm1, %zmm12
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm26, %zmm24
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm20, %zmm0, %zmm24
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm26, %zmm27
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm20, %zmm3, %zmm27
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm26, %zmm31
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm20, %zmm10, %zmm31
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm26, %zmm30
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm20, %zmm15, %zmm30
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm26, %zmm22
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm20, %zmm13, %zmm22
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm20, %zmm1, %zmm26
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm7, %zmm8
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm17, %zmm0, %zmm8
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm7, %zmm12
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm17, %zmm3, %zmm12
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm7, %zmm6
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm17, %zmm10, %zmm6
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm6, (%rsp) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm7, %zmm6
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm17, %zmm15, %zmm6
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm7, %zmm6
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm17, %zmm13, %zmm6
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm17, %zmm1, %zmm7
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm25, %zmm18
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm21, %zmm4, %zmm18
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm25, %zmm19
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm21, %zmm0, %zmm19
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm25, %zmm23
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm3, %zmm6
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm21, %zmm3, %zmm23
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm25, %zmm16
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm21, %zmm10, %zmm16
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm25, %zmm20
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm21, %zmm15, %zmm20
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm25, %zmm17
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm21, %zmm13, %zmm17
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm21, %zmm1, %zmm25
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm29, %zmm9
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm4, %zmm9
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm29, %zmm11
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm0, %zmm11
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm0, %zmm3
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm29, %zmm14
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm6, %zmm14
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm29, %zmm21
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm10, %zmm21
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm29, %zmm7
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm15, %zmm2
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm15, %zmm7
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm29, %zmm15
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm13, %zmm15
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm1, %zmm29
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm28, %zmm0, %zmm4
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm28, %zmm0, %zmm3
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm28, %zmm0, %zmm6
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm28, %zmm0, %zmm10
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm28, %zmm0, %zmm2
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm28, %zmm0, %zmm13
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm28, %zmm1, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm18, %zmm9 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512BW-ONLY-FAST-NEXT:    # ymm0 = mem[0,1,0,1]
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm28, %zmm2, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm9, %zmm18
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm0, %zmm24 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm9 = [2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58]
; AVX512BW-ONLY-FAST-NEXT:    # ymm9 = mem[0,1,0,1]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm9, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm24, %zmm24
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm19, %zmm11 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm28, %zmm2, %zmm9
; AVX512BW-ONLY-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm9[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm11, %zmm1
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm0, %zmm27 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59]
; AVX512BW-ONLY-FAST-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm4, %zmm9
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm3, %zmm9
; AVX512BW-ONLY-FAST-NEXT:    vpblendd {{.*#+}} ymm4 = ymm12[0,1,2,3],ymm9[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm4, %zmm27, %zmm4
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm23, %zmm14 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm28, %zmm2, %zmm3
; AVX512BW-ONLY-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm3, %zmm14, %zmm3
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm6 = [4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60]
; AVX512BW-ONLY-FAST-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm0, %zmm8
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm6, %zmm8
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm9 = [5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61]
; AVX512BW-ONLY-FAST-NEXT:    # ymm9 = mem[0,1,0,1]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm0, %zmm11
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm9, %zmm11
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm12 = [6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62]
; AVX512BW-ONLY-FAST-NEXT:    # ymm12 = mem[0,1,0,1]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm0, %zmm14
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm12, %zmm14
; AVX512BW-ONLY-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} ymm19 = [7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63]
; AVX512BW-ONLY-FAST-NEXT:    # ymm19 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm5, %zmm19, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm0, %zmm5
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm28, %zmm2, %zmm6
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm28, %zmm2, %zmm9
; AVX512BW-ONLY-FAST-NEXT:    vpermi2w %zmm28, %zmm2, %zmm12
; AVX512BW-ONLY-FAST-NEXT:    vpermt2w %zmm28, %zmm19, %zmm2
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512BW-ONLY-FAST-NEXT:    vpblendd $15, (%rsp), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX512BW-ONLY-FAST-NEXT:    # ymm8 = mem[0,1,2,3],ymm8[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm2, %zmm31 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm8, %zmm31, %zmm8
; AVX512BW-ONLY-FAST-NEXT:    vpblendd {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm6[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm16, %zmm21 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm6, %zmm21, %zmm6
; AVX512BW-ONLY-FAST-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm10 # 32-byte Folded Reload
; AVX512BW-ONLY-FAST-NEXT:    # ymm10 = mem[0,1,2,3],ymm11[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm2, %zmm30 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm10, %zmm30, %zmm10
; AVX512BW-ONLY-FAST-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
; AVX512BW-ONLY-FAST-NEXT:    # ymm9 = mem[0,1,2,3],ymm9[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm20, %zmm7 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm9, %zmm7, %zmm7
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm2, %zmm22 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm9 # 32-byte Folded Reload
; AVX512BW-ONLY-FAST-NEXT:    # ymm9 = mem[0,1,2,3],ymm14[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm9, %zmm22, %zmm9
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm17, %zmm15 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3],ymm12[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm2, %zmm15, %zmm2
; AVX512BW-ONLY-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm11, %zmm26 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm11 # 32-byte Folded Reload
; AVX512BW-ONLY-FAST-NEXT:    # ymm11 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm11, %zmm26, %zmm5
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm25, %zmm29 {%k1}
; AVX512BW-ONLY-FAST-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
; AVX512BW-ONLY-FAST-NEXT:    # ymm11 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-ONLY-FAST-NEXT:    vinserti64x4 $0, %ymm11, %zmm29, %zmm11
; AVX512BW-ONLY-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vmovaps %zmm0, 64(%rsi)
; AVX512BW-ONLY-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vmovaps %zmm0, (%rsi)
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm18, 64(%rdx)
; AVX512BW-ONLY-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-ONLY-FAST-NEXT:    vmovaps %zmm0, (%rdx)
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm3, 64(%r8)
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm4, (%r8)
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm6, 64(%r9)
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm8, (%r9)
; AVX512BW-ONLY-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm7, 64(%rax)
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm10, (%rax)
; AVX512BW-ONLY-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm2, 64(%rax)
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm9, (%rax)
; AVX512BW-ONLY-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm11, 64(%rax)
; AVX512BW-ONLY-FAST-NEXT:    vmovdqa64 %zmm5, (%rax)
; AVX512BW-ONLY-FAST-NEXT:    addq $1096, %rsp # imm = 0x448
; AVX512BW-ONLY-FAST-NEXT:    vzeroupper
; AVX512BW-ONLY-FAST-NEXT:    retq
;
; AVX512DQBW-SLOW-LABEL: load_i16_stride8_vf64:
; AVX512DQBW-SLOW:       # %bb.0:
; AVX512DQBW-SLOW-NEXT:    subq $1096, %rsp # imm = 0x448
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 704(%rdi), %zmm6
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 640(%rdi), %zmm8
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 576(%rdi), %zmm28
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 512(%rdi), %zmm11
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 832(%rdi), %zmm5
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 768(%rdi), %zmm29
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 960(%rdi), %zmm21
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 896(%rdi), %zmm25
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 (%rdi), %zmm7
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 64(%rdi), %zmm17
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 128(%rdi), %zmm4
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 192(%rdi), %zmm9
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 320(%rdi), %zmm20
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 256(%rdi), %zmm26
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 448(%rdi), %zmm22
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 384(%rdi), %zmm2
; AVX512DQBW-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56]
; AVX512DQBW-SLOW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm1
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm12
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm22, %zmm0, %zmm1
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm20, %zmm0, %zmm2
; AVX512DQBW-SLOW-NEXT:    movb $-64, %al
; AVX512DQBW-SLOW-NEXT:    kmovd %eax, %k1
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
; AVX512DQBW-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56]
; AVX512DQBW-SLOW-NEXT:    # ymm1 = mem[0,1,0,1]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm4, %zmm3
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm4, %zmm10
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm9, %zmm1, %zmm3
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm4
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm17, %zmm0, %zmm4
; AVX512DQBW-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm3, %zmm2, %zmm2
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm2
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm21, %zmm0, %zmm2
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm3
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm0, %zmm3
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm28, %zmm11, %zmm0
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm6, %zmm8, %zmm1
; AVX512DQBW-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm3, %zmm0
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm4 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512DQBW-SLOW-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm12, %zmm0
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm22, %zmm4, %zmm0
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm1
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm20, %zmm4, %zmm1
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
; AVX512DQBW-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512DQBW-SLOW-NEXT:    # ymm2 = mem[0,1,0,1]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm10, %zmm0
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm9, %zmm2, %zmm0
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm2
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm17, %zmm4, %zmm2
; AVX512DQBW-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58]
; AVX512DQBW-SLOW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm22, %zmm0, %zmm1
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59]
; AVX512DQBW-SLOW-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm22, %zmm2, %zmm1
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm3
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm10 = [4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60]
; AVX512DQBW-SLOW-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm12, %zmm2
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm22, %zmm10, %zmm2
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm15 = [5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61]
; AVX512DQBW-SLOW-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm22, %zmm15, %zmm1
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm13 = [6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62]
; AVX512DQBW-SLOW-NEXT:    # zmm13 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm22, %zmm13, %zmm1
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63]
; AVX512DQBW-SLOW-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm22, %zmm1, %zmm12
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm24
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm20, %zmm0, %zmm24
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm27
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm20, %zmm3, %zmm27
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm31
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm20, %zmm10, %zmm31
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm30
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm20, %zmm15, %zmm30
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm26, %zmm22
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm20, %zmm13, %zmm22
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm20, %zmm1, %zmm26
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm8
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm17, %zmm0, %zmm8
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm12
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm17, %zmm3, %zmm12
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm6
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm17, %zmm10, %zmm6
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm6, (%rsp) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm6
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm17, %zmm15, %zmm6
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm7, %zmm6
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm17, %zmm13, %zmm6
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm17, %zmm1, %zmm7
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm18
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm21, %zmm4, %zmm18
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm19
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm21, %zmm0, %zmm19
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm23
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm3, %zmm6
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm21, %zmm3, %zmm23
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm16
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm21, %zmm10, %zmm16
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm20
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm21, %zmm15, %zmm20
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm17
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm21, %zmm13, %zmm17
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm21, %zmm1, %zmm25
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm9
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm4, %zmm9
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm11
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm0, %zmm11
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm3
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm14
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm6, %zmm14
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm21
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm10, %zmm21
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm7
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm15, %zmm2
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm15, %zmm7
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm29, %zmm15
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm13, %zmm15
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm1, %zmm29
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm28, %zmm0, %zmm4
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm28, %zmm0, %zmm3
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm28, %zmm0, %zmm6
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm28, %zmm0, %zmm10
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm28, %zmm0, %zmm2
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm28, %zmm0, %zmm13
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm28, %zmm1, %zmm0
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm18, %zmm9 {%k1}
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512DQBW-SLOW-NEXT:    # ymm0 = mem[0,1,0,1]
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm28, %zmm2, %zmm0
; AVX512DQBW-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm9, %zmm18
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm24 {%k1}
; AVX512DQBW-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm9 = [2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58]
; AVX512DQBW-SLOW-NEXT:    # ymm9 = mem[0,1,0,1]
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm9, %zmm0
; AVX512DQBW-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm0, %zmm24, %zmm24
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm19, %zmm11 {%k1}
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm28, %zmm2, %zmm9
; AVX512DQBW-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm9[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm1, %zmm11, %zmm1
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm27 {%k1}
; AVX512DQBW-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59]
; AVX512DQBW-SLOW-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm4, %zmm9
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm3, %zmm9
; AVX512DQBW-SLOW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm12[0,1,2,3],ymm9[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm4, %zmm27, %zmm4
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm23, %zmm14 {%k1}
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm28, %zmm2, %zmm3
; AVX512DQBW-SLOW-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm3, %zmm14, %zmm3
; AVX512DQBW-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm6 = [4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60]
; AVX512DQBW-SLOW-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm8
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm6, %zmm8
; AVX512DQBW-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm9 = [5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61]
; AVX512DQBW-SLOW-NEXT:    # ymm9 = mem[0,1,0,1]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm11
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm9, %zmm11
; AVX512DQBW-SLOW-NEXT:    vbroadcasti128 {{.*#+}} ymm12 = [6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62]
; AVX512DQBW-SLOW-NEXT:    # ymm12 = mem[0,1,0,1]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm14
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm12, %zmm14
; AVX512DQBW-SLOW-NEXT:    vbroadcasti64x2 {{.*#+}} ymm19 = [7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63]
; AVX512DQBW-SLOW-NEXT:    # ymm19 = mem[0,1,0,1]
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm5, %zmm19, %zmm0
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm0, %zmm5
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm28, %zmm2, %zmm6
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm28, %zmm2, %zmm9
; AVX512DQBW-SLOW-NEXT:    vpermi2w %zmm28, %zmm2, %zmm12
; AVX512DQBW-SLOW-NEXT:    vpermt2w %zmm28, %zmm19, %zmm2
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQBW-SLOW-NEXT:    vpblendd $15, (%rsp), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX512DQBW-SLOW-NEXT:    # ymm8 = mem[0,1,2,3],ymm8[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm31 {%k1}
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm8, %zmm31, %zmm8
; AVX512DQBW-SLOW-NEXT:    vpblendd {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm6[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm16, %zmm21 {%k1}
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm6, %zmm21, %zmm6
; AVX512DQBW-SLOW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm10 # 32-byte Folded Reload
; AVX512DQBW-SLOW-NEXT:    # ymm10 = mem[0,1,2,3],ymm11[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm30 {%k1}
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm10, %zmm30, %zmm10
; AVX512DQBW-SLOW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
; AVX512DQBW-SLOW-NEXT:    # ymm9 = mem[0,1,2,3],ymm9[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm20, %zmm7 {%k1}
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm9, %zmm7, %zmm7
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm2, %zmm22 {%k1}
; AVX512DQBW-SLOW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm9 # 32-byte Folded Reload
; AVX512DQBW-SLOW-NEXT:    # ymm9 = mem[0,1,2,3],ymm14[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm9, %zmm22, %zmm9
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm17, %zmm15 {%k1}
; AVX512DQBW-SLOW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3],ymm12[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm2, %zmm15, %zmm2
; AVX512DQBW-SLOW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm11, %zmm26 {%k1}
; AVX512DQBW-SLOW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm11 # 32-byte Folded Reload
; AVX512DQBW-SLOW-NEXT:    # ymm11 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm11, %zmm26, %zmm5
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm25, %zmm29 {%k1}
; AVX512DQBW-SLOW-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
; AVX512DQBW-SLOW-NEXT:    # ymm11 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-SLOW-NEXT:    vinserti64x4 $0, %ymm11, %zmm29, %zmm11
; AVX512DQBW-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vmovaps %zmm0, 64(%rsi)
; AVX512DQBW-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vmovaps %zmm0, (%rsi)
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm18, 64(%rdx)
; AVX512DQBW-SLOW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQBW-SLOW-NEXT:    vmovaps %zmm0, (%rdx)
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm3, 64(%r8)
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm4, (%r8)
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm6, 64(%r9)
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm8, (%r9)
; AVX512DQBW-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm7, 64(%rax)
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm10, (%rax)
; AVX512DQBW-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm2, 64(%rax)
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm9, (%rax)
; AVX512DQBW-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm11, 64(%rax)
; AVX512DQBW-SLOW-NEXT:    vmovdqa64 %zmm5, (%rax)
; AVX512DQBW-SLOW-NEXT:    addq $1096, %rsp # imm = 0x448
; AVX512DQBW-SLOW-NEXT:    vzeroupper
; AVX512DQBW-SLOW-NEXT:    retq
;
; AVX512DQBW-FAST-LABEL: load_i16_stride8_vf64:
; AVX512DQBW-FAST:       # %bb.0:
; AVX512DQBW-FAST-NEXT:    subq $1096, %rsp # imm = 0x448
; AVX512DQBW-FAST-NEXT:    vmovdqa64 704(%rdi), %zmm6
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vmovdqa64 640(%rdi), %zmm8
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vmovdqa64 576(%rdi), %zmm28
; AVX512DQBW-FAST-NEXT:    vmovdqa64 512(%rdi), %zmm11
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vmovdqa64 832(%rdi), %zmm5
; AVX512DQBW-FAST-NEXT:    vmovdqa64 768(%rdi), %zmm29
; AVX512DQBW-FAST-NEXT:    vmovdqa64 960(%rdi), %zmm21
; AVX512DQBW-FAST-NEXT:    vmovdqa64 896(%rdi), %zmm25
; AVX512DQBW-FAST-NEXT:    vmovdqa64 (%rdi), %zmm7
; AVX512DQBW-FAST-NEXT:    vmovdqa64 64(%rdi), %zmm17
; AVX512DQBW-FAST-NEXT:    vmovdqa64 128(%rdi), %zmm4
; AVX512DQBW-FAST-NEXT:    vmovdqa64 192(%rdi), %zmm9
; AVX512DQBW-FAST-NEXT:    vmovdqa64 320(%rdi), %zmm20
; AVX512DQBW-FAST-NEXT:    vmovdqa64 256(%rdi), %zmm26
; AVX512DQBW-FAST-NEXT:    vmovdqa64 448(%rdi), %zmm22
; AVX512DQBW-FAST-NEXT:    vmovdqa64 384(%rdi), %zmm2
; AVX512DQBW-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56]
; AVX512DQBW-FAST-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm2, %zmm1
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm2, %zmm12
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm22, %zmm0, %zmm1
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm20, %zmm0, %zmm2
; AVX512DQBW-FAST-NEXT:    movb $-64, %al
; AVX512DQBW-FAST-NEXT:    kmovd %eax, %k1
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm1, %zmm2 {%k1}
; AVX512DQBW-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm1 = [0,8,16,24,32,40,48,56,0,8,16,24,32,40,48,56]
; AVX512DQBW-FAST-NEXT:    # ymm1 = mem[0,1,0,1]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm4, %zmm3
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm4, %zmm10
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm9, %zmm1, %zmm3
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm7, %zmm4
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm17, %zmm0, %zmm4
; AVX512DQBW-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm3, %zmm2, %zmm2
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm25, %zmm2
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm21, %zmm0, %zmm2
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm29, %zmm3
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm0, %zmm3
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm2, %zmm3 {%k1}
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm28, %zmm11, %zmm0
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm6, %zmm8, %zmm1
; AVX512DQBW-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm3, %zmm0
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm4 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512DQBW-FAST-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm12, %zmm0
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm22, %zmm4, %zmm0
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm26, %zmm1
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm20, %zmm4, %zmm1
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
; AVX512DQBW-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm2 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512DQBW-FAST-NEXT:    # ymm2 = mem[0,1,0,1]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm10, %zmm0
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm9, %zmm2, %zmm0
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm7, %zmm2
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm17, %zmm4, %zmm2
; AVX512DQBW-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm1, %zmm0
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm0 = [2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58]
; AVX512DQBW-FAST-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm22, %zmm0, %zmm1
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm2 = [3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59]
; AVX512DQBW-FAST-NEXT:    # zmm2 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm22, %zmm2, %zmm1
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm2, %zmm3
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm10 = [4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60]
; AVX512DQBW-FAST-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm12, %zmm2
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm22, %zmm10, %zmm2
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm15 = [5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61]
; AVX512DQBW-FAST-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm22, %zmm15, %zmm1
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm13 = [6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62]
; AVX512DQBW-FAST-NEXT:    # zmm13 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm12, %zmm1
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm22, %zmm13, %zmm1
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vbroadcasti32x4 {{.*#+}} zmm1 = [7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63]
; AVX512DQBW-FAST-NEXT:    # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm22, %zmm1, %zmm12
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm26, %zmm24
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm20, %zmm0, %zmm24
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm26, %zmm27
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm20, %zmm3, %zmm27
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm26, %zmm31
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm20, %zmm10, %zmm31
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm26, %zmm30
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm20, %zmm15, %zmm30
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm26, %zmm22
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm20, %zmm13, %zmm22
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm20, %zmm1, %zmm26
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm7, %zmm8
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm17, %zmm0, %zmm8
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm7, %zmm12
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm17, %zmm3, %zmm12
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm7, %zmm6
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm17, %zmm10, %zmm6
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm6, (%rsp) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm7, %zmm6
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm17, %zmm15, %zmm6
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm7, %zmm6
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm17, %zmm13, %zmm6
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm17, %zmm1, %zmm7
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm25, %zmm18
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm21, %zmm4, %zmm18
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm25, %zmm19
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm21, %zmm0, %zmm19
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm25, %zmm23
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm3, %zmm6
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm21, %zmm3, %zmm23
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm25, %zmm16
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm21, %zmm10, %zmm16
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm25, %zmm20
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm21, %zmm15, %zmm20
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm25, %zmm17
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm21, %zmm13, %zmm17
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm21, %zmm1, %zmm25
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm29, %zmm9
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm4, %zmm9
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm29, %zmm11
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm0, %zmm11
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm0, %zmm3
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm29, %zmm14
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm6, %zmm14
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm29, %zmm21
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm10, %zmm21
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm29, %zmm7
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm15, %zmm2
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm15, %zmm7
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm29, %zmm15
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm13, %zmm15
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm1, %zmm29
; AVX512DQBW-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm28, %zmm0, %zmm4
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm28, %zmm0, %zmm3
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm28, %zmm0, %zmm6
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm28, %zmm0, %zmm10
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm28, %zmm0, %zmm2
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm28, %zmm0, %zmm13
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm28, %zmm1, %zmm0
; AVX512DQBW-FAST-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm18, %zmm9 {%k1}
; AVX512DQBW-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm0 = [1,9,17,25,33,41,49,57,1,9,17,25,33,41,49,57]
; AVX512DQBW-FAST-NEXT:    # ymm0 = mem[0,1,0,1]
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm28, %zmm2, %zmm0
; AVX512DQBW-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm9, %zmm18
; AVX512DQBW-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm0, %zmm24 {%k1}
; AVX512DQBW-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm9 = [2,10,18,26,34,42,50,58,2,10,18,26,34,42,50,58]
; AVX512DQBW-FAST-NEXT:    # ymm9 = mem[0,1,0,1]
; AVX512DQBW-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512DQBW-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm9, %zmm0
; AVX512DQBW-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm0, %zmm24, %zmm24
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm19, %zmm11 {%k1}
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm28, %zmm2, %zmm9
; AVX512DQBW-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm9[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm1, %zmm11, %zmm1
; AVX512DQBW-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm0, %zmm27 {%k1}
; AVX512DQBW-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [3,11,19,27,35,43,51,59,3,11,19,27,35,43,51,59]
; AVX512DQBW-FAST-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm4, %zmm9
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm3, %zmm9
; AVX512DQBW-FAST-NEXT:    vpblendd {{.*#+}} ymm4 = ymm12[0,1,2,3],ymm9[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm4, %zmm27, %zmm4
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm23, %zmm14 {%k1}
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm28, %zmm2, %zmm3
; AVX512DQBW-FAST-NEXT:    vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm3, %zmm14, %zmm3
; AVX512DQBW-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm6 = [4,12,20,28,36,44,52,60,4,12,20,28,36,44,52,60]
; AVX512DQBW-FAST-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm0, %zmm8
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm6, %zmm8
; AVX512DQBW-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm9 = [5,13,21,29,37,45,53,61,5,13,21,29,37,45,53,61]
; AVX512DQBW-FAST-NEXT:    # ymm9 = mem[0,1,0,1]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm0, %zmm11
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm9, %zmm11
; AVX512DQBW-FAST-NEXT:    vbroadcasti128 {{.*#+}} ymm12 = [6,14,22,30,38,46,54,62,6,14,22,30,38,46,54,62]
; AVX512DQBW-FAST-NEXT:    # ymm12 = mem[0,1,0,1]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm0, %zmm14
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm12, %zmm14
; AVX512DQBW-FAST-NEXT:    vbroadcasti64x2 {{.*#+}} ymm19 = [7,15,23,31,39,47,55,63,7,15,23,31,39,47,55,63]
; AVX512DQBW-FAST-NEXT:    # ymm19 = mem[0,1,0,1]
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm5, %zmm19, %zmm0
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm0, %zmm5
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm28, %zmm2, %zmm6
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm28, %zmm2, %zmm9
; AVX512DQBW-FAST-NEXT:    vpermi2w %zmm28, %zmm2, %zmm12
; AVX512DQBW-FAST-NEXT:    vpermt2w %zmm28, %zmm19, %zmm2
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQBW-FAST-NEXT:    vpblendd $15, (%rsp), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX512DQBW-FAST-NEXT:    # ymm8 = mem[0,1,2,3],ymm8[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm2, %zmm31 {%k1}
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm8, %zmm31, %zmm8
; AVX512DQBW-FAST-NEXT:    vpblendd {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm6[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm16, %zmm21 {%k1}
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm6, %zmm21, %zmm6
; AVX512DQBW-FAST-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm10 # 32-byte Folded Reload
; AVX512DQBW-FAST-NEXT:    # ymm10 = mem[0,1,2,3],ymm11[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm2, %zmm30 {%k1}
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm10, %zmm30, %zmm10
; AVX512DQBW-FAST-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
; AVX512DQBW-FAST-NEXT:    # ymm9 = mem[0,1,2,3],ymm9[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm20, %zmm7 {%k1}
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm9, %zmm7, %zmm7
; AVX512DQBW-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm2, %zmm22 {%k1}
; AVX512DQBW-FAST-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm9 # 32-byte Folded Reload
; AVX512DQBW-FAST-NEXT:    # ymm9 = mem[0,1,2,3],ymm14[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm9, %zmm22, %zmm9
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm17, %zmm15 {%k1}
; AVX512DQBW-FAST-NEXT:    vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3],ymm12[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm2, %zmm15, %zmm2
; AVX512DQBW-FAST-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm11, %zmm26 {%k1}
; AVX512DQBW-FAST-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm11 # 32-byte Folded Reload
; AVX512DQBW-FAST-NEXT:    # ymm11 = mem[0,1,2,3],ymm5[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm11, %zmm26, %zmm5
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm25, %zmm29 {%k1}
; AVX512DQBW-FAST-NEXT:    vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
; AVX512DQBW-FAST-NEXT:    # ymm11 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQBW-FAST-NEXT:    vinserti64x4 $0, %ymm11, %zmm29, %zmm11
; AVX512DQBW-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vmovaps %zmm0, 64(%rsi)
; AVX512DQBW-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vmovaps %zmm0, (%rsi)
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm18, 64(%rdx)
; AVX512DQBW-FAST-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQBW-FAST-NEXT:    vmovaps %zmm0, (%rdx)
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm3, 64(%r8)
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm4, (%r8)
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm6, 64(%r9)
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm8, (%r9)
; AVX512DQBW-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm7, 64(%rax)
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm10, (%rax)
; AVX512DQBW-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm2, 64(%rax)
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm9, (%rax)
; AVX512DQBW-FAST-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm11, 64(%rax)
; AVX512DQBW-FAST-NEXT:    vmovdqa64 %zmm5, (%rax)
; AVX512DQBW-FAST-NEXT:    addq $1096, %rsp # imm = 0x448
; AVX512DQBW-FAST-NEXT:    vzeroupper
; AVX512DQBW-FAST-NEXT:    retq
  %wide.vec = load <512 x i16>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 64, i32 72, i32 80, i32 88, i32 96, i32 104, i32 112, i32 120, i32 128, i32 136, i32 144, i32 152, i32 160, i32 168, i32 176, i32 184, i32 192, i32 200, i32 208, i32 216, i32 224, i32 232, i32 240, i32 248, i32 256, i32 264, i32 272, i32 280, i32 288, i32 296, i32 304, i32 312, i32 320, i32 328, i32 336, i32 344, i32 352, i32 360, i32 368, i32 376, i32 384, i32 392, i32 400, i32 408, i32 416, i32 424, i32 432, i32 440, i32 448, i32 456, i32 464, i32 472, i32 480, i32 488, i32 496, i32 504>
  %strided.vec1 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 65, i32 73, i32 81, i32 89, i32 97, i32 105, i32 113, i32 121, i32 129, i32 137, i32 145, i32 153, i32 161, i32 169, i32 177, i32 185, i32 193, i32 201, i32 209, i32 217, i32 225, i32 233, i32 241, i32 249, i32 257, i32 265, i32 273, i32 281, i32 289, i32 297, i32 305, i32 313, i32 321, i32 329, i32 337, i32 345, i32 353, i32 361, i32 369, i32 377, i32 385, i32 393, i32 401, i32 409, i32 417, i32 425, i32 433, i32 441, i32 449, i32 457, i32 465, i32 473, i32 481, i32 489, i32 497, i32 505>
  %strided.vec2 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 66, i32 74, i32 82, i32 90, i32 98, i32 106, i32 114, i32 122, i32 130, i32 138, i32 146, i32 154, i32 162, i32 170, i32 178, i32 186, i32 194, i32 202, i32 210, i32 218, i32 226, i32 234, i32 242, i32 250, i32 258, i32 266, i32 274, i32 282, i32 290, i32 298, i32 306, i32 314, i32 322, i32 330, i32 338, i32 346, i32 354, i32 362, i32 370, i32 378, i32 386, i32 394, i32 402, i32 410, i32 418, i32 426, i32 434, i32 442, i32 450, i32 458, i32 466, i32 474, i32 482, i32 490, i32 498, i32 506>
  %strided.vec3 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 3, i32 11, i32 19, i32 27, i32 35, i32 43, i32 51, i32 59, i32 67, i32 75, i32 83, i32 91, i32 99, i32 107, i32 115, i32 123, i32 131, i32 139, i32 147, i32 155, i32 163, i32 171, i32 179, i32 187, i32 195, i32 203, i32 211, i32 219, i32 227, i32 235, i32 243, i32 251, i32 259, i32 267, i32 275, i32 283, i32 291, i32 299, i32 307, i32 315, i32 323, i32 331, i32 339, i32 347, i32 355, i32 363, i32 371, i32 379, i32 387, i32 395, i32 403, i32 411, i32 419, i32 427, i32 435, i32 443, i32 451, i32 459, i32 467, i32 475, i32 483, i32 491, i32 499, i32 507>
  %strided.vec4 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 4, i32 12, i32 20, i32 28, i32 36, i32 44, i32 52, i32 60, i32 68, i32 76, i32 84, i32 92, i32 100, i32 108, i32 116, i32 124, i32 132, i32 140, i32 148, i32 156, i32 164, i32 172, i32 180, i32 188, i32 196, i32 204, i32 212, i32 220, i32 228, i32 236, i32 244, i32 252, i32 260, i32 268, i32 276, i32 284, i32 292, i32 300, i32 308, i32 316, i32 324, i32 332, i32 340, i32 348, i32 356, i32 364, i32 372, i32 380, i32 388, i32 396, i32 404, i32 412, i32 420, i32 428, i32 436, i32 444, i32 452, i32 460, i32 468, i32 476, i32 484, i32 492, i32 500, i32 508>
  %strided.vec5 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 69, i32 77, i32 85, i32 93, i32 101, i32 109, i32 117, i32 125, i32 133, i32 141, i32 149, i32 157, i32 165, i32 173, i32 181, i32 189, i32 197, i32 205, i32 213, i32 221, i32 229, i32 237, i32 245, i32 253, i32 261, i32 269, i32 277, i32 285, i32 293, i32 301, i32 309, i32 317, i32 325, i32 333, i32 341, i32 349, i32 357, i32 365, i32 373, i32 381, i32 389, i32 397, i32 405, i32 413, i32 421, i32 429, i32 437, i32 445, i32 453, i32 461, i32 469, i32 477, i32 485, i32 493, i32 501, i32 509>
  %strided.vec6 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 70, i32 78, i32 86, i32 94, i32 102, i32 110, i32 118, i32 126, i32 134, i32 142, i32 150, i32 158, i32 166, i32 174, i32 182, i32 190, i32 198, i32 206, i32 214, i32 222, i32 230, i32 238, i32 246, i32 254, i32 262, i32 270, i32 278, i32 286, i32 294, i32 302, i32 310, i32 318, i32 326, i32 334, i32 342, i32 350, i32 358, i32 366, i32 374, i32 382, i32 390, i32 398, i32 406, i32 414, i32 422, i32 430, i32 438, i32 446, i32 454, i32 462, i32 470, i32 478, i32 486, i32 494, i32 502, i32 510>
  %strided.vec7 = shufflevector <512 x i16> %wide.vec, <512 x i16> poison, <64 x i32> <i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63, i32 71, i32 79, i32 87, i32 95, i32 103, i32 111, i32 119, i32 127, i32 135, i32 143, i32 151, i32 159, i32 167, i32 175, i32 183, i32 191, i32 199, i32 207, i32 215, i32 223, i32 231, i32 239, i32 247, i32 255, i32 263, i32 271, i32 279, i32 287, i32 295, i32 303, i32 311, i32 319, i32 327, i32 335, i32 343, i32 351, i32 359, i32 367, i32 375, i32 383, i32 391, i32 399, i32 407, i32 415, i32 423, i32 431, i32 439, i32 447, i32 455, i32 463, i32 471, i32 479, i32 487, i32 495, i32 503, i32 511>
  store <64 x i16> %strided.vec0, ptr %out.vec0, align 64
  store <64 x i16> %strided.vec1, ptr %out.vec1, align 64
  store <64 x i16> %strided.vec2, ptr %out.vec2, align 64
  store <64 x i16> %strided.vec3, ptr %out.vec3, align 64
  store <64 x i16> %strided.vec4, ptr %out.vec4, align 64
  store <64 x i16> %strided.vec5, ptr %out.vec5, align 64
  store <64 x i16> %strided.vec6, ptr %out.vec6, align 64
  store <64 x i16> %strided.vec7, ptr %out.vec7, align 64
  ret void
}
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; AVX1: {{.*}}
; AVX2: {{.*}}
; AVX2-FAST: {{.*}}
; AVX2-FAST-PERLANE: {{.*}}
; AVX2-SLOW: {{.*}}
; AVX512: {{.*}}
; AVX512-FAST: {{.*}}
; AVX512-SLOW: {{.*}}
; AVX512BW-FAST: {{.*}}
; AVX512BW-SLOW: {{.*}}
; AVX512DQ-FAST: {{.*}}
; AVX512DQ-SLOW: {{.*}}
; AVX512F: {{.*}}
; AVX512F-ONLY-FAST: {{.*}}
; AVX512F-ONLY-SLOW: {{.*}}
; FALLBACK0: {{.*}}
; FALLBACK1: {{.*}}
; FALLBACK10: {{.*}}
; FALLBACK11: {{.*}}
; FALLBACK12: {{.*}}
; FALLBACK2: {{.*}}
; FALLBACK3: {{.*}}
; FALLBACK4: {{.*}}
; FALLBACK5: {{.*}}
; FALLBACK6: {{.*}}
; FALLBACK7: {{.*}}
; FALLBACK8: {{.*}}
; FALLBACK9: {{.*}}