llvm/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse2 | FileCheck %s --check-prefixes=SSE
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx  | FileCheck %s --check-prefixes=AVX
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX2
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX2-FP
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX2-FCP
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl | FileCheck %s --check-prefixes=AVX512
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512-FCP
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq | FileCheck %s --check-prefixes=AVX512DQ
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512DQ-FCP
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefixes=AVX512BW
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512bw,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512BW-FCP
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq,+avx512bw | FileCheck %s --check-prefixes=AVX512DQ-BW
; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq,+avx512bw,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512DQ-BW-FCP

; These patterns are produced by LoopVectorizer for interleaved loads.

define void @load_i32_stride6_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
; SSE-LABEL: load_i32_stride6_vf2:
; SSE:       # %bb.0:
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movdqa (%rdi), %xmm1
; SSE-NEXT:    movdqa 16(%rdi), %xmm0
; SSE-NEXT:    movdqa 32(%rdi), %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[2,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[2,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm0[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1]
; SSE-NEXT:    movq %xmm1, (%rsi)
; SSE-NEXT:    movq %xmm4, (%rdx)
; SSE-NEXT:    movq %xmm5, (%rcx)
; SSE-NEXT:    movq %xmm6, (%r8)
; SSE-NEXT:    movq %xmm0, (%r9)
; SSE-NEXT:    movq %xmm7, (%rax)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i32_stride6_vf2:
; AVX:       # %bb.0:
; AVX-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX-NEXT:    vmovaps (%rdi), %xmm0
; AVX-NEXT:    vmovaps 16(%rdi), %xmm1
; AVX-NEXT:    vmovaps 32(%rdi), %xmm2
; AVX-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[2,2,3,3]
; AVX-NEXT:    vblendps {{.*#+}} xmm3 = xmm0[0],xmm3[1],xmm0[2,3]
; AVX-NEXT:    vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} xmm5 = xmm2[0,1],xmm0[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[2,0,2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm6 = xmm2[2,2,3,3]
; AVX-NEXT:    vblendps {{.*#+}} xmm6 = xmm1[0],xmm6[1],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3,2,3]
; AVX-NEXT:    vmovlps %xmm3, (%rsi)
; AVX-NEXT:    vmovlps %xmm4, (%rdx)
; AVX-NEXT:    vmovlps %xmm5, (%rcx)
; AVX-NEXT:    vmovlps %xmm0, (%r8)
; AVX-NEXT:    vmovlps %xmm6, (%r9)
; AVX-NEXT:    vmovlps %xmm1, (%rax)
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i32_stride6_vf2:
; AVX2:       # %bb.0:
; AVX2-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm0
; AVX2-NEXT:    vmovaps (%rdi), %xmm1
; AVX2-NEXT:    vmovaps 16(%rdi), %xmm2
; AVX2-NEXT:    vmovaps 32(%rdi), %xmm3
; AVX2-NEXT:    vshufps {{.*#+}} xmm4 = xmm2[2,2,3,3]
; AVX2-NEXT:    vblendps {{.*#+}} xmm4 = xmm1[0],xmm4[1],xmm1[2,3]
; AVX2-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0,1],xmm2[2,3]
; AVX2-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} xmm5 = xmm3[0,1],xmm1[2,3]
; AVX2-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[2,0,2,3]
; AVX2-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX2-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
; AVX2-NEXT:    vmovsd {{.*#+}} xmm3 = [4,2,0,0]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm0, %ymm3, %ymm3
; AVX2-NEXT:    vmovsd {{.*#+}} xmm6 = [5,3,0,0]
; AVX2-NEXT:    vpermps %ymm0, %ymm6, %ymm0
; AVX2-NEXT:    vmovlps %xmm4, (%rsi)
; AVX2-NEXT:    vmovlps %xmm2, (%rdx)
; AVX2-NEXT:    vmovlps %xmm5, (%rcx)
; AVX2-NEXT:    vmovlps %xmm1, (%r8)
; AVX2-NEXT:    vmovlps %xmm3, (%r9)
; AVX2-NEXT:    vmovlps %xmm0, (%rax)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i32_stride6_vf2:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovaps (%rdi), %xmm1
; AVX2-FP-NEXT:    vmovaps 16(%rdi), %xmm2
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %xmm3
; AVX2-FP-NEXT:    vshufps {{.*#+}} xmm4 = xmm2[2,2,3,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm4 = xmm1[0],xmm4[1],xmm1[2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0,1],xmm2[2,3]
; AVX2-FP-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm5 = xmm3[0,1],xmm1[2,3]
; AVX2-FP-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[2,0,2,3]
; AVX2-FP-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
; AVX2-FP-NEXT:    vmovsd {{.*#+}} xmm3 = [4,2,0,0]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm0, %ymm3, %ymm3
; AVX2-FP-NEXT:    vmovsd {{.*#+}} xmm6 = [5,3,0,0]
; AVX2-FP-NEXT:    vpermps %ymm0, %ymm6, %ymm0
; AVX2-FP-NEXT:    vmovlps %xmm4, (%rsi)
; AVX2-FP-NEXT:    vmovlps %xmm2, (%rdx)
; AVX2-FP-NEXT:    vmovlps %xmm5, (%rcx)
; AVX2-FP-NEXT:    vmovlps %xmm1, (%r8)
; AVX2-FP-NEXT:    vmovlps %xmm3, (%r9)
; AVX2-FP-NEXT:    vmovlps %xmm0, (%rax)
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i32_stride6_vf2:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovaps (%rdi), %xmm1
; AVX2-FCP-NEXT:    vmovaps 16(%rdi), %xmm2
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %xmm3
; AVX2-FCP-NEXT:    vshufps {{.*#+}} xmm4 = xmm2[2,2,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm4 = xmm1[0],xmm4[1],xmm1[2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm2 = xmm1[0,1],xmm2[2,3]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm5 = xmm3[0,1],xmm1[2,3]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[2,0,2,3]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
; AVX2-FCP-NEXT:    vmovsd {{.*#+}} xmm3 = [4,2,0,0]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm3, %ymm3
; AVX2-FCP-NEXT:    vmovsd {{.*#+}} xmm6 = [5,3,0,0]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm6, %ymm0
; AVX2-FCP-NEXT:    vmovlps %xmm4, (%rsi)
; AVX2-FCP-NEXT:    vmovlps %xmm2, (%rdx)
; AVX2-FCP-NEXT:    vmovlps %xmm5, (%rcx)
; AVX2-FCP-NEXT:    vmovlps %xmm1, (%r8)
; AVX2-FCP-NEXT:    vmovlps %xmm3, (%r9)
; AVX2-FCP-NEXT:    vmovlps %xmm0, (%rax)
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i32_stride6_vf2:
; AVX512:       # %bb.0:
; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-NEXT:    vmovaps 16(%rdi), %xmm1
; AVX512-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512-NEXT:    vextractps $2, %xmm1, %r10d
; AVX512-NEXT:    vpinsrd $1, %r10d, %xmm0, %xmm3
; AVX512-NEXT:    vextractps $3, %xmm1, %r10d
; AVX512-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512-NEXT:    vpinsrd $1, %r10d, %xmm1, %xmm1
; AVX512-NEXT:    vpbroadcastd 8(%rdi), %xmm4
; AVX512-NEXT:    vmovd %xmm2, %r10d
; AVX512-NEXT:    vpinsrd $1, %r10d, %xmm4, %xmm4
; AVX512-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3]
; AVX512-NEXT:    vmovsd {{.*#+}} xmm2 = [4,2,0,0]
; AVX512-NEXT:    vmovaps 32(%rdi), %ymm5
; AVX512-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vpermps %ymm5, %ymm2, %ymm2
; AVX512-NEXT:    vmovsd {{.*#+}} xmm6 = [5,3,0,0]
; AVX512-NEXT:    vpermps %ymm5, %ymm6, %ymm5
; AVX512-NEXT:    vmovq %xmm3, (%rsi)
; AVX512-NEXT:    vmovq %xmm1, (%rdx)
; AVX512-NEXT:    vmovq %xmm4, (%rcx)
; AVX512-NEXT:    vmovq %xmm0, (%r8)
; AVX512-NEXT:    vmovlps %xmm2, (%r9)
; AVX512-NEXT:    vmovlps %xmm5, (%rax)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i32_stride6_vf2:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-FCP-NEXT:    vpbroadcastq {{.*#+}} xmm0 = [0,6,0,6]
; AVX512-FCP-NEXT:    vmovdqa (%rdi), %xmm1
; AVX512-FCP-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX512-FCP-NEXT:    vpermi2d %xmm2, %xmm1, %xmm0
; AVX512-FCP-NEXT:    vpbroadcastq {{.*#+}} xmm4 = [1,7,1,7]
; AVX512-FCP-NEXT:    vpermi2d %xmm2, %xmm1, %xmm4
; AVX512-FCP-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [2,4,2,4]
; AVX512-FCP-NEXT:    vpermi2d %xmm3, %xmm1, %xmm2
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [3,5,0,0]
; AVX512-FCP-NEXT:    vpermi2d %xmm3, %xmm1, %xmm5
; AVX512-FCP-NEXT:    vmovsd {{.*#+}} xmm1 = [4,2,0,0]
; AVX512-FCP-NEXT:    vmovaps 32(%rdi), %ymm3
; AVX512-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vpermps %ymm3, %ymm1, %ymm1
; AVX512-FCP-NEXT:    vmovsd {{.*#+}} xmm6 = [5,3,0,0]
; AVX512-FCP-NEXT:    vpermps %ymm3, %ymm6, %ymm3
; AVX512-FCP-NEXT:    vmovq %xmm0, (%rsi)
; AVX512-FCP-NEXT:    vmovq %xmm4, (%rdx)
; AVX512-FCP-NEXT:    vmovq %xmm2, (%rcx)
; AVX512-FCP-NEXT:    vmovq %xmm5, (%r8)
; AVX512-FCP-NEXT:    vmovlps %xmm1, (%r9)
; AVX512-FCP-NEXT:    vmovlps %xmm3, (%rax)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i32_stride6_vf2:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-NEXT:    vmovaps 16(%rdi), %xmm1
; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-NEXT:    vextractps $2, %xmm1, %r10d
; AVX512DQ-NEXT:    vpinsrd $1, %r10d, %xmm0, %xmm3
; AVX512DQ-NEXT:    vextractps $3, %xmm1, %r10d
; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512DQ-NEXT:    vpinsrd $1, %r10d, %xmm1, %xmm1
; AVX512DQ-NEXT:    vpbroadcastd 8(%rdi), %xmm4
; AVX512DQ-NEXT:    vmovd %xmm2, %r10d
; AVX512DQ-NEXT:    vpinsrd $1, %r10d, %xmm4, %xmm4
; AVX512DQ-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512DQ-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3]
; AVX512DQ-NEXT:    vmovsd {{.*#+}} xmm2 = [4,2,0,0]
; AVX512DQ-NEXT:    vmovaps 32(%rdi), %ymm5
; AVX512DQ-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vpermps %ymm5, %ymm2, %ymm2
; AVX512DQ-NEXT:    vmovsd {{.*#+}} xmm6 = [5,3,0,0]
; AVX512DQ-NEXT:    vpermps %ymm5, %ymm6, %ymm5
; AVX512DQ-NEXT:    vmovq %xmm3, (%rsi)
; AVX512DQ-NEXT:    vmovq %xmm1, (%rdx)
; AVX512DQ-NEXT:    vmovq %xmm4, (%rcx)
; AVX512DQ-NEXT:    vmovq %xmm0, (%r8)
; AVX512DQ-NEXT:    vmovlps %xmm2, (%r9)
; AVX512DQ-NEXT:    vmovlps %xmm5, (%rax)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i32_stride6_vf2:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-FCP-NEXT:    vpbroadcastq {{.*#+}} xmm0 = [0,6,0,6]
; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %xmm1
; AVX512DQ-FCP-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX512DQ-FCP-NEXT:    vpermi2d %xmm2, %xmm1, %xmm0
; AVX512DQ-FCP-NEXT:    vpbroadcastq {{.*#+}} xmm4 = [1,7,1,7]
; AVX512DQ-FCP-NEXT:    vpermi2d %xmm2, %xmm1, %xmm4
; AVX512DQ-FCP-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [2,4,2,4]
; AVX512DQ-FCP-NEXT:    vpermi2d %xmm3, %xmm1, %xmm2
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [3,5,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2d %xmm3, %xmm1, %xmm5
; AVX512DQ-FCP-NEXT:    vmovsd {{.*#+}} xmm1 = [4,2,0,0]
; AVX512DQ-FCP-NEXT:    vmovaps 32(%rdi), %ymm3
; AVX512DQ-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vpermps %ymm3, %ymm1, %ymm1
; AVX512DQ-FCP-NEXT:    vmovsd {{.*#+}} xmm6 = [5,3,0,0]
; AVX512DQ-FCP-NEXT:    vpermps %ymm3, %ymm6, %ymm3
; AVX512DQ-FCP-NEXT:    vmovq %xmm0, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovq %xmm4, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovq %xmm2, (%rcx)
; AVX512DQ-FCP-NEXT:    vmovq %xmm5, (%r8)
; AVX512DQ-FCP-NEXT:    vmovlps %xmm1, (%r9)
; AVX512DQ-FCP-NEXT:    vmovlps %xmm3, (%rax)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i32_stride6_vf2:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT:    vmovaps 16(%rdi), %xmm1
; AVX512BW-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512BW-NEXT:    vextractps $2, %xmm1, %r10d
; AVX512BW-NEXT:    vpinsrd $1, %r10d, %xmm0, %xmm3
; AVX512BW-NEXT:    vextractps $3, %xmm1, %r10d
; AVX512BW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512BW-NEXT:    vpinsrd $1, %r10d, %xmm1, %xmm1
; AVX512BW-NEXT:    vpbroadcastd 8(%rdi), %xmm4
; AVX512BW-NEXT:    vmovd %xmm2, %r10d
; AVX512BW-NEXT:    vpinsrd $1, %r10d, %xmm4, %xmm4
; AVX512BW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512BW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3]
; AVX512BW-NEXT:    vmovsd {{.*#+}} xmm2 = [4,2,0,0]
; AVX512BW-NEXT:    vmovaps 32(%rdi), %ymm5
; AVX512BW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vpermps %ymm5, %ymm2, %ymm2
; AVX512BW-NEXT:    vmovsd {{.*#+}} xmm6 = [5,3,0,0]
; AVX512BW-NEXT:    vpermps %ymm5, %ymm6, %ymm5
; AVX512BW-NEXT:    vmovq %xmm3, (%rsi)
; AVX512BW-NEXT:    vmovq %xmm1, (%rdx)
; AVX512BW-NEXT:    vmovq %xmm4, (%rcx)
; AVX512BW-NEXT:    vmovq %xmm0, (%r8)
; AVX512BW-NEXT:    vmovlps %xmm2, (%r9)
; AVX512BW-NEXT:    vmovlps %xmm5, (%rax)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i32_stride6_vf2:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-FCP-NEXT:    vpbroadcastq {{.*#+}} xmm0 = [0,6,0,6]
; AVX512BW-FCP-NEXT:    vmovdqa (%rdi), %xmm1
; AVX512BW-FCP-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX512BW-FCP-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX512BW-FCP-NEXT:    vpermi2d %xmm2, %xmm1, %xmm0
; AVX512BW-FCP-NEXT:    vpbroadcastq {{.*#+}} xmm4 = [1,7,1,7]
; AVX512BW-FCP-NEXT:    vpermi2d %xmm2, %xmm1, %xmm4
; AVX512BW-FCP-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [2,4,2,4]
; AVX512BW-FCP-NEXT:    vpermi2d %xmm3, %xmm1, %xmm2
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [3,5,0,0]
; AVX512BW-FCP-NEXT:    vpermi2d %xmm3, %xmm1, %xmm5
; AVX512BW-FCP-NEXT:    vmovsd {{.*#+}} xmm1 = [4,2,0,0]
; AVX512BW-FCP-NEXT:    vmovaps 32(%rdi), %ymm3
; AVX512BW-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vpermps %ymm3, %ymm1, %ymm1
; AVX512BW-FCP-NEXT:    vmovsd {{.*#+}} xmm6 = [5,3,0,0]
; AVX512BW-FCP-NEXT:    vpermps %ymm3, %ymm6, %ymm3
; AVX512BW-FCP-NEXT:    vmovq %xmm0, (%rsi)
; AVX512BW-FCP-NEXT:    vmovq %xmm4, (%rdx)
; AVX512BW-FCP-NEXT:    vmovq %xmm2, (%rcx)
; AVX512BW-FCP-NEXT:    vmovq %xmm5, (%r8)
; AVX512BW-FCP-NEXT:    vmovlps %xmm1, (%r9)
; AVX512BW-FCP-NEXT:    vmovlps %xmm3, (%rax)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i32_stride6_vf2:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-BW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-BW-NEXT:    vmovaps 16(%rdi), %xmm1
; AVX512DQ-BW-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-BW-NEXT:    vextractps $2, %xmm1, %r10d
; AVX512DQ-BW-NEXT:    vpinsrd $1, %r10d, %xmm0, %xmm3
; AVX512DQ-BW-NEXT:    vextractps $3, %xmm1, %r10d
; AVX512DQ-BW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512DQ-BW-NEXT:    vpinsrd $1, %r10d, %xmm1, %xmm1
; AVX512DQ-BW-NEXT:    vpbroadcastd 8(%rdi), %xmm4
; AVX512DQ-BW-NEXT:    vmovd %xmm2, %r10d
; AVX512DQ-BW-NEXT:    vpinsrd $1, %r10d, %xmm4, %xmm4
; AVX512DQ-BW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512DQ-BW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3]
; AVX512DQ-BW-NEXT:    vmovsd {{.*#+}} xmm2 = [4,2,0,0]
; AVX512DQ-BW-NEXT:    vmovaps 32(%rdi), %ymm5
; AVX512DQ-BW-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vpermps %ymm5, %ymm2, %ymm2
; AVX512DQ-BW-NEXT:    vmovsd {{.*#+}} xmm6 = [5,3,0,0]
; AVX512DQ-BW-NEXT:    vpermps %ymm5, %ymm6, %ymm5
; AVX512DQ-BW-NEXT:    vmovq %xmm3, (%rsi)
; AVX512DQ-BW-NEXT:    vmovq %xmm1, (%rdx)
; AVX512DQ-BW-NEXT:    vmovq %xmm4, (%rcx)
; AVX512DQ-BW-NEXT:    vmovq %xmm0, (%r8)
; AVX512DQ-BW-NEXT:    vmovlps %xmm2, (%r9)
; AVX512DQ-BW-NEXT:    vmovlps %xmm5, (%rax)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i32_stride6_vf2:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-BW-FCP-NEXT:    vpbroadcastq {{.*#+}} xmm0 = [0,6,0,6]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa (%rdi), %xmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 16(%rdi), %xmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 32(%rdi), %xmm3
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %xmm2, %xmm1, %xmm0
; AVX512DQ-BW-FCP-NEXT:    vpbroadcastq {{.*#+}} xmm4 = [1,7,1,7]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %xmm2, %xmm1, %xmm4
; AVX512DQ-BW-FCP-NEXT:    vpbroadcastq {{.*#+}} xmm2 = [2,4,2,4]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %xmm3, %xmm1, %xmm2
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [3,5,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %xmm3, %xmm1, %xmm5
; AVX512DQ-BW-FCP-NEXT:    vmovsd {{.*#+}} xmm1 = [4,2,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovaps 32(%rdi), %ymm3
; AVX512DQ-BW-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vpermps %ymm3, %ymm1, %ymm1
; AVX512DQ-BW-FCP-NEXT:    vmovsd {{.*#+}} xmm6 = [5,3,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermps %ymm3, %ymm6, %ymm3
; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm0, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm4, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm2, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm5, (%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovlps %xmm1, (%r9)
; AVX512DQ-BW-FCP-NEXT:    vmovlps %xmm3, (%rax)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <12 x i32>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <2 x i32> <i32 0, i32 6>
  %strided.vec1 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <2 x i32> <i32 1, i32 7>
  %strided.vec2 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <2 x i32> <i32 2, i32 8>
  %strided.vec3 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <2 x i32> <i32 3, i32 9>
  %strided.vec4 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <2 x i32> <i32 4, i32 10>
  %strided.vec5 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <2 x i32> <i32 5, i32 11>
  store <2 x i32> %strided.vec0, ptr %out.vec0, align 64
  store <2 x i32> %strided.vec1, ptr %out.vec1, align 64
  store <2 x i32> %strided.vec2, ptr %out.vec2, align 64
  store <2 x i32> %strided.vec3, ptr %out.vec3, align 64
  store <2 x i32> %strided.vec4, ptr %out.vec4, align 64
  store <2 x i32> %strided.vec5, ptr %out.vec5, align 64
  ret void
}

define void @load_i32_stride6_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
; SSE-LABEL: load_i32_stride6_vf4:
; SSE:       # %bb.0:
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movdqa 80(%rdi), %xmm1
; SSE-NEXT:    movdqa 64(%rdi), %xmm0
; SSE-NEXT:    movdqa (%rdi), %xmm4
; SSE-NEXT:    movdqa 16(%rdi), %xmm2
; SSE-NEXT:    movdqa 48(%rdi), %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[2,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm4[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm4[2,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm4[3,3,3,3]
; SSE-NEXT:    movdqa %xmm4, %xmm6
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm6[0],xmm4[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm0[2,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1]
; SSE-NEXT:    movdqa 32(%rdi), %xmm10
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm7[0],xmm3[1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm10[0],xmm8[1],xmm10[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1]
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm8[0],xmm5[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm10[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm7[0],xmm9[1],xmm7[1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm9[0],xmm6[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm10[2,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm2[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm1[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm7[0],xmm9[1],xmm7[1]
; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm2[0],xmm9[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm8[0],xmm0[1]
; SSE-NEXT:    movapd %xmm4, (%rsi)
; SSE-NEXT:    movapd %xmm3, (%rdx)
; SSE-NEXT:    movapd %xmm5, (%rcx)
; SSE-NEXT:    movapd %xmm6, (%r8)
; SSE-NEXT:    movapd %xmm9, (%r9)
; SSE-NEXT:    movapd %xmm0, (%rax)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i32_stride6_vf4:
; AVX:       # %bb.0:
; AVX-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX-NEXT:    vmovaps 32(%rdi), %ymm0
; AVX-NEXT:    vmovaps (%rdi), %ymm1
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm3
; AVX-NEXT:    vblendps {{.*#+}} xmm4 = xmm2[0,1],xmm3[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,2],xmm3[0,3]
; AVX-NEXT:    vmovaps 64(%rdi), %xmm5
; AVX-NEXT:    vinsertps {{.*#+}} xmm4 = xmm4[0,1,2],xmm5[2]
; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,0],xmm3[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[1,3]
; AVX-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm5[3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[2,0],xmm1[2,3]
; AVX-NEXT:    vmovaps 80(%rdi), %xmm6
; AVX-NEXT:    vinsertps {{.*#+}} xmm3 = xmm3[0,1,2],xmm6[0]
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[3,3]
; AVX-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm6[1]
; AVX-NEXT:    vmovaps 32(%rdi), %xmm1
; AVX-NEXT:    vshufps {{.*#+}} xmm7 = xmm1[2,2,3,3]
; AVX-NEXT:    vmovaps 16(%rdi), %xmm8
; AVX-NEXT:    vblendps {{.*#+}} xmm7 = xmm8[0],xmm7[1],xmm8[2,3]
; AVX-NEXT:    vblendps {{.*#+}} xmm9 = xmm5[0,1],xmm6[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm9 = xmm9[0,1,0,2]
; AVX-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3]
; AVX-NEXT:    vpmovzxdq {{.*#+}} xmm5 = xmm5[0],zero,xmm5[1],zero
; AVX-NEXT:    vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5],xmm6[6,7]
; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3]
; AVX-NEXT:    vmovaps %xmm4, (%rsi)
; AVX-NEXT:    vmovaps %xmm2, (%rdx)
; AVX-NEXT:    vmovaps %xmm3, (%rcx)
; AVX-NEXT:    vmovaps %xmm0, (%r8)
; AVX-NEXT:    vmovaps %xmm7, (%r9)
; AVX-NEXT:    vmovaps %xmm1, (%rax)
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i32_stride6_vf4:
; AVX2:       # %bb.0:
; AVX2-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-NEXT:    vmovdqa (%rdi), %ymm1
; AVX2-NEXT:    vmovdqa 32(%rdi), %ymm2
; AVX2-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,6,4,0]
; AVX2-NEXT:    vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
; AVX2-NEXT:    vpermd %ymm3, %ymm0, %ymm0
; AVX2-NEXT:    vmovdqa 64(%rdi), %xmm4
; AVX2-NEXT:    vpshufd {{.*#+}} xmm5 = xmm4[2,2,2,2]
; AVX2-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm5[3]
; AVX2-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [1,7,5,0]
; AVX2-NEXT:    vpermd %ymm3, %ymm5, %ymm3
; AVX2-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3]
; AVX2-NEXT:    vmovdqa 80(%rdi), %xmm5
; AVX2-NEXT:    vpbroadcastd %xmm5, %xmm6
; AVX2-NEXT:    vpblendd {{.*#+}} ymm7 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vpshufd {{.*#+}} ymm8 = ymm7[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,3,2,3]
; AVX2-NEXT:    vpblendd {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
; AVX2-NEXT:    vpbroadcastd 84(%rdi), %xmm8
; AVX2-NEXT:    vpshufd {{.*#+}} ymm7 = ymm7[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[0,3,2,3]
; AVX2-NEXT:    vpblendd {{.*#+}} xmm7 = xmm7[0,1,2],xmm8[3]
; AVX2-NEXT:    vpblendd {{.*#+}} xmm8 = xmm4[0,1],xmm5[2,3]
; AVX2-NEXT:    vpshufd {{.*#+}} xmm8 = xmm8[0,1,0,2]
; AVX2-NEXT:    vpmovsxbd {{.*#+}} xmm9 = [4,2,0,0]
; AVX2-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vpermd %ymm1, %ymm9, %ymm2
; AVX2-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm8[2,3]
; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX2-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0,1,2],xmm5[3]
; AVX2-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [5,3,0,0]
; AVX2-NEXT:    vpermd %ymm1, %ymm5, %ymm1
; AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
; AVX2-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX2-NEXT:    vmovdqa %xmm3, (%rdx)
; AVX2-NEXT:    vmovdqa %xmm6, (%rcx)
; AVX2-NEXT:    vmovdqa %xmm7, (%r8)
; AVX2-NEXT:    vmovdqa %xmm2, (%r9)
; AVX2-NEXT:    vmovdqa %xmm1, (%rax)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i32_stride6_vf4:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FP-NEXT:    vmovdqa (%rdi), %ymm1
; AVX2-FP-NEXT:    vmovdqa 32(%rdi), %ymm2
; AVX2-FP-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,6,4,0]
; AVX2-FP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vpermd %ymm3, %ymm0, %ymm0
; AVX2-FP-NEXT:    vmovdqa 64(%rdi), %xmm4
; AVX2-FP-NEXT:    vpshufd {{.*#+}} xmm5 = xmm4[2,2,2,2]
; AVX2-FP-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm5[3]
; AVX2-FP-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [1,7,5,0]
; AVX2-FP-NEXT:    vpermd %ymm3, %ymm5, %ymm3
; AVX2-FP-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3]
; AVX2-FP-NEXT:    vmovdqa 80(%rdi), %xmm5
; AVX2-FP-NEXT:    vpbroadcastd %xmm5, %xmm6
; AVX2-FP-NEXT:    vpblendd {{.*#+}} ymm7 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vpshufd {{.*#+}} ymm8 = ymm7[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermq {{.*#+}} ymm8 = ymm8[0,3,2,3]
; AVX2-FP-NEXT:    vpblendd {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
; AVX2-FP-NEXT:    vpbroadcastd 84(%rdi), %xmm8
; AVX2-FP-NEXT:    vpshufd {{.*#+}} ymm7 = ymm7[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[0,3,2,3]
; AVX2-FP-NEXT:    vpblendd {{.*#+}} xmm7 = xmm7[0,1,2],xmm8[3]
; AVX2-FP-NEXT:    vpblendd {{.*#+}} xmm8 = xmm4[0,1],xmm5[2,3]
; AVX2-FP-NEXT:    vpshufd {{.*#+}} xmm8 = xmm8[0,1,0,2]
; AVX2-FP-NEXT:    vpmovsxbd {{.*#+}} xmm9 = [4,2,0,0]
; AVX2-FP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vpermd %ymm1, %ymm9, %ymm2
; AVX2-FP-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm8[2,3]
; AVX2-FP-NEXT:    vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX2-FP-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0,1,2],xmm5[3]
; AVX2-FP-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [5,3,0,0]
; AVX2-FP-NEXT:    vpermd %ymm1, %ymm5, %ymm1
; AVX2-FP-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
; AVX2-FP-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX2-FP-NEXT:    vmovdqa %xmm3, (%rdx)
; AVX2-FP-NEXT:    vmovdqa %xmm6, (%rcx)
; AVX2-FP-NEXT:    vmovdqa %xmm7, (%r8)
; AVX2-FP-NEXT:    vmovdqa %xmm2, (%r9)
; AVX2-FP-NEXT:    vmovdqa %xmm1, (%rax)
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i32_stride6_vf4:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FCP-NEXT:    vmovdqa (%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovdqa 32(%rdi), %ymm2
; AVX2-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm0 = [0,6,4,0]
; AVX2-FCP-NEXT:    vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vpermd %ymm3, %ymm0, %ymm0
; AVX2-FCP-NEXT:    vmovdqa 64(%rdi), %xmm4
; AVX2-FCP-NEXT:    vpshufd {{.*#+}} xmm5 = xmm4[2,2,2,2]
; AVX2-FCP-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm5[3]
; AVX2-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [1,7,5,0]
; AVX2-FCP-NEXT:    vpermd %ymm3, %ymm5, %ymm3
; AVX2-FCP-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3]
; AVX2-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [2,0,6,7]
; AVX2-FCP-NEXT:    vpblendd {{.*#+}} ymm6 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vpermd %ymm6, %ymm5, %ymm5
; AVX2-FCP-NEXT:    vmovdqa 80(%rdi), %xmm7
; AVX2-FCP-NEXT:    vpbroadcastd %xmm7, %xmm8
; AVX2-FCP-NEXT:    vpblendd {{.*#+}} xmm5 = xmm5[0,1,2],xmm8[3]
; AVX2-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm8 = [3,1,7,7]
; AVX2-FCP-NEXT:    vpermd %ymm6, %ymm8, %ymm6
; AVX2-FCP-NEXT:    vpbroadcastd 84(%rdi), %xmm8
; AVX2-FCP-NEXT:    vpblendd {{.*#+}} xmm6 = xmm6[0,1,2],xmm8[3]
; AVX2-FCP-NEXT:    vpblendd {{.*#+}} xmm8 = xmm4[0,1],xmm7[2,3]
; AVX2-FCP-NEXT:    vpshufd {{.*#+}} xmm8 = xmm8[0,1,0,2]
; AVX2-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm9 = [4,2,0,0]
; AVX2-FCP-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vpermd %ymm1, %ymm9, %ymm2
; AVX2-FCP-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm8[2,3]
; AVX2-FCP-NEXT:    vpmovzxdq {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero
; AVX2-FCP-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0,1,2],xmm7[3]
; AVX2-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm7 = [5,3,0,0]
; AVX2-FCP-NEXT:    vpermd %ymm1, %ymm7, %ymm1
; AVX2-FCP-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
; AVX2-FCP-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX2-FCP-NEXT:    vmovdqa %xmm3, (%rdx)
; AVX2-FCP-NEXT:    vmovdqa %xmm5, (%rcx)
; AVX2-FCP-NEXT:    vmovdqa %xmm6, (%r8)
; AVX2-FCP-NEXT:    vmovdqa %xmm2, (%r9)
; AVX2-FCP-NEXT:    vmovdqa %xmm1, (%rax)
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i32_stride6_vf4:
; AVX512:       # %bb.0:
; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512-NEXT:    vpmovsxbd {{.*#+}} xmm2 = [0,6,12,18]
; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm2
; AVX512-NEXT:    vpmovsxbd {{.*#+}} xmm3 = [1,7,13,19]
; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512-NEXT:    vpmovsxbd {{.*#+}} xmm4 = [2,8,14,20]
; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [3,9,15,21]
; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm5
; AVX512-NEXT:    vpmovsxbd {{.*#+}} xmm6 = [4,10,16,22]
; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm6
; AVX512-NEXT:    vpmovsxbd {{.*#+}} xmm7 = [5,11,17,23]
; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm7
; AVX512-NEXT:    vmovdqa %xmm2, (%rsi)
; AVX512-NEXT:    vmovdqa %xmm3, (%rdx)
; AVX512-NEXT:    vmovdqa %xmm4, (%rcx)
; AVX512-NEXT:    vmovdqa %xmm5, (%r8)
; AVX512-NEXT:    vmovdqa %xmm6, (%r9)
; AVX512-NEXT:    vmovdqa %xmm7, (%rax)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i32_stride6_vf4:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm2 = [0,6,12,18]
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm2
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm3 = [1,7,13,19]
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm4 = [2,8,14,20]
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [3,9,15,21]
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm5
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm6 = [4,10,16,22]
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm6
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm7 = [5,11,17,23]
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm7
; AVX512-FCP-NEXT:    vmovdqa %xmm2, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa %xmm3, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa %xmm4, (%rcx)
; AVX512-FCP-NEXT:    vmovdqa %xmm5, (%r8)
; AVX512-FCP-NEXT:    vmovdqa %xmm6, (%r9)
; AVX512-FCP-NEXT:    vmovdqa %xmm7, (%rax)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i32_stride6_vf4:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} xmm2 = [0,6,12,18]
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm0, %zmm2
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} xmm3 = [1,7,13,19]
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} xmm4 = [2,8,14,20]
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [3,9,15,21]
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm0, %zmm5
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} xmm6 = [4,10,16,22]
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm0, %zmm6
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} xmm7 = [5,11,17,23]
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm0, %zmm7
; AVX512DQ-NEXT:    vmovdqa %xmm2, (%rsi)
; AVX512DQ-NEXT:    vmovdqa %xmm3, (%rdx)
; AVX512DQ-NEXT:    vmovdqa %xmm4, (%rcx)
; AVX512DQ-NEXT:    vmovdqa %xmm5, (%r8)
; AVX512DQ-NEXT:    vmovdqa %xmm6, (%r9)
; AVX512DQ-NEXT:    vmovdqa %xmm7, (%rax)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i32_stride6_vf4:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm2 = [0,6,12,18]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm2
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm3 = [1,7,13,19]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm4 = [2,8,14,20]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [3,9,15,21]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm5
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm6 = [4,10,16,22]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm6
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm7 = [5,11,17,23]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqa %xmm2, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa %xmm3, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa %xmm4, (%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa %xmm5, (%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa %xmm6, (%r9)
; AVX512DQ-FCP-NEXT:    vmovdqa %xmm7, (%rax)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i32_stride6_vf4:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} xmm2 = [0,6,12,18]
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm2
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} xmm3 = [1,7,13,19]
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} xmm4 = [2,8,14,20]
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [3,9,15,21]
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm5
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} xmm6 = [4,10,16,22]
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm6
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} xmm7 = [5,11,17,23]
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm7
; AVX512BW-NEXT:    vmovdqa %xmm2, (%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, (%rax)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i32_stride6_vf4:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm2 = [0,6,12,18]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm2
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm3 = [1,7,13,19]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm4 = [2,8,14,20]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [3,9,15,21]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm5
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm6 = [4,10,16,22]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm6
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm7 = [5,11,17,23]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqa %xmm2, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa %xmm3, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa %xmm4, (%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa %xmm5, (%r8)
; AVX512BW-FCP-NEXT:    vmovdqa %xmm6, (%r9)
; AVX512BW-FCP-NEXT:    vmovdqa %xmm7, (%rax)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i32_stride6_vf4:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} xmm2 = [0,6,12,18]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm2
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} xmm3 = [1,7,13,19]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} xmm4 = [2,8,14,20]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [3,9,15,21]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm5
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} xmm6 = [4,10,16,22]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm6
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} xmm7 = [5,11,17,23]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqa %xmm2, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa %xmm3, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa %xmm4, (%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa %xmm5, (%r8)
; AVX512DQ-BW-NEXT:    vmovdqa %xmm6, (%r9)
; AVX512DQ-BW-NEXT:    vmovdqa %xmm7, (%rax)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i32_stride6_vf4:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm2 = [0,6,12,18]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm3 = [1,7,13,19]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm4 = [2,8,14,20]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm5 = [3,9,15,21]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm6 = [4,10,16,22]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} xmm7 = [5,11,17,23]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm2, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm3, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm4, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm5, (%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm6, (%r9)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm7, (%rax)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <24 x i32>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <4 x i32> <i32 0, i32 6, i32 12, i32 18>
  %strided.vec1 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <4 x i32> <i32 1, i32 7, i32 13, i32 19>
  %strided.vec2 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <4 x i32> <i32 2, i32 8, i32 14, i32 20>
  %strided.vec3 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <4 x i32> <i32 3, i32 9, i32 15, i32 21>
  %strided.vec4 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <4 x i32> <i32 4, i32 10, i32 16, i32 22>
  %strided.vec5 = shufflevector <24 x i32> %wide.vec, <24 x i32> poison, <4 x i32> <i32 5, i32 11, i32 17, i32 23>
  store <4 x i32> %strided.vec0, ptr %out.vec0, align 64
  store <4 x i32> %strided.vec1, ptr %out.vec1, align 64
  store <4 x i32> %strided.vec2, ptr %out.vec2, align 64
  store <4 x i32> %strided.vec3, ptr %out.vec3, align 64
  store <4 x i32> %strided.vec4, ptr %out.vec4, align 64
  store <4 x i32> %strided.vec5, ptr %out.vec5, align 64
  ret void
}

define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
; SSE-LABEL: load_i32_stride6_vf8:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa 144(%rdi), %xmm4
; SSE-NEXT:    movdqa 160(%rdi), %xmm2
; SSE-NEXT:    movdqa 96(%rdi), %xmm6
; SSE-NEXT:    movdqa 112(%rdi), %xmm3
; SSE-NEXT:    movdqa 64(%rdi), %xmm5
; SSE-NEXT:    movdqa (%rdi), %xmm10
; SSE-NEXT:    movdqa 16(%rdi), %xmm0
; SSE-NEXT:    movdqa 48(%rdi), %xmm8
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE-NEXT:    movdqa %xmm0, %xmm11
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm10[2,3,2,3]
; SSE-NEXT:    movdqa %xmm10, %xmm7
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm8[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm7[0],xmm9[1]
; SSE-NEXT:    movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,3,2,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm6[1,1,1,1]
; SSE-NEXT:    movdqa %xmm6, %xmm9
; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm4[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm9[0],xmm7[1]
; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
; SSE-NEXT:    movdqa %xmm5, %xmm9
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm8[2,2,3,3]
; SSE-NEXT:    movdqa %xmm8, %xmm11
; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm0[0],xmm11[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[3,3,3,3]
; SSE-NEXT:    movdqa %xmm3, %xmm1
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
; SSE-NEXT:    movdqa %xmm4, %xmm12
; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
; SSE-NEXT:    movdqa 80(%rdi), %xmm14
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm15[0],xmm12[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; SSE-NEXT:    movdqa 32(%rdi), %xmm7
; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm7[0],xmm13[1],xmm7[1]
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm13[0],xmm5[1]
; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm4[2,2,3,3]
; SSE-NEXT:    movdqa 176(%rdi), %xmm15
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
; SSE-NEXT:    movdqa 128(%rdi), %xmm5
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm7[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm14[0],xmm8[1],xmm14[1]
; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm0[0],xmm8[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm15[0],xmm4[1],xmm15[1]
; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm3, %xmm10
; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm9[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm10[0],xmm6[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm5[2,3,2,3]
; SSE-NEXT:    movdqa %xmm1, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm2[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm0[0],xmm9[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:    pshufd {{.*#+}} xmm1 = xmm5[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rsi)
; SSE-NEXT:    movapd %xmm12, 16(%rdx)
; SSE-NEXT:    movapd %xmm11, (%rdx)
; SSE-NEXT:    movapd %xmm13, 16(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rcx)
; SSE-NEXT:    movapd %xmm4, 16(%r8)
; SSE-NEXT:    movapd %xmm8, (%r8)
; SSE-NEXT:    movapd %xmm10, 16(%r9)
; SSE-NEXT:    movapd %xmm6, (%r9)
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movapd %xmm2, 16(%rax)
; SSE-NEXT:    movapd %xmm9, (%rax)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i32_stride6_vf8:
; AVX:       # %bb.0:
; AVX-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX-NEXT:    vmovapd 160(%rdi), %ymm3
; AVX-NEXT:    vmovapd 128(%rdi), %ymm4
; AVX-NEXT:    vmovaps 32(%rdi), %ymm6
; AVX-NEXT:    vmovaps (%rdi), %ymm7
; AVX-NEXT:    vmovaps 96(%rdi), %ymm0
; AVX-NEXT:    vmovaps 64(%rdi), %ymm1
; AVX-NEXT:    vinsertf128 $1, 96(%rdi), %ymm1, %ymm5
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm0[2,0],ymm5[0,0],ymm0[6,4],ymm5[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm5[2,2],ymm2[6,4],ymm5[6,6]
; AVX-NEXT:    vblendps {{.*#+}} ymm8 = ymm7[0,1,2,3],ymm6[4,5],ymm7[6,7]
; AVX-NEXT:    vextractf128 $1, %ymm8, %xmm9
; AVX-NEXT:    vblendps {{.*#+}} xmm10 = xmm8[0,1],xmm9[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[0,2],xmm9[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm10[0,1,2],ymm2[3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm4[2,3],ymm3[0,1]
; AVX-NEXT:    vshufpd {{.*#+}} ymm11 = ymm10[0],ymm4[0],ymm10[3],ymm4[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm11[6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm11 = ymm0[3,0],ymm5[1,0],ymm0[7,4],ymm5[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm11[2,0],ymm5[2,3],ymm11[6,4],ymm5[6,7]
; AVX-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[1,0],xmm9[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[0,2],xmm9[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1,2],ymm5[3,4,5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm8 = ymm10[3,1],ymm4[1,3],ymm10[7,5],ymm4[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm8[6,7]
; AVX-NEXT:    vblendpd {{.*#+}} ymm8 = ymm4[0,1],ymm3[2],ymm4[3]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm8[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm10 = ymm9[0,0],ymm8[2,0],ymm9[4,4],ymm8[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm7 = ymm6[0,1],ymm7[2,3],ymm6[4,5,6,7]
; AVX-NEXT:    vextractf128 $1, %ymm7, %xmm11
; AVX-NEXT:    vshufps {{.*#+}} xmm6 = xmm7[2,0],xmm11[2,3]
; AVX-NEXT:    vshufps {{.*#+}} ymm12 = ymm0[2,1],ymm1[2,0],ymm0[6,5],ymm1[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm12[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm10[5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm8 = ymm9[0,1],ymm8[3,1],ymm9[4,5],ymm8[7,5]
; AVX-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[3,1],xmm11[3,3]
; AVX-NEXT:    vshufps {{.*#+}} ymm9 = ymm0[3,1],ymm1[2,1],ymm0[7,5],ymm1[6,5]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm9[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm8[5,6,7]
; AVX-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm3[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm8 = ymm3[2,0],ymm4[0,0],ymm3[6,4],ymm4[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm8 = ymm4[0,2],ymm8[2,0],ymm4[4,6],ymm8[6,4]
; AVX-NEXT:    vmovaps 32(%rdi), %xmm9
; AVX-NEXT:    vshufps {{.*#+}} xmm10 = xmm9[2,2,3,3]
; AVX-NEXT:    vmovaps 16(%rdi), %xmm11
; AVX-NEXT:    vblendps {{.*#+}} xmm10 = xmm11[0],xmm10[1],xmm11[2,3]
; AVX-NEXT:    vmovapd 80(%rdi), %xmm12
; AVX-NEXT:    vshufpd {{.*#+}} ymm13 = ymm12[1],ymm1[0],ymm12[2],ymm1[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm13 = ymm0[0,1],ymm13[2,0],ymm0[4,5],ymm13[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm13[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1,2,3,4],ymm8[5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,0],ymm4[1,0],ymm3[7,4],ymm4[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[0,3],ymm3[2,0],ymm4[4,7],ymm3[6,4]
; AVX-NEXT:    vblendps {{.*#+}} xmm4 = xmm11[0,1],xmm9[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[1,3,2,3]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[3,1],ymm1[1,3],ymm12[7,5],ymm1[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm1[2,0],ymm0[5,5],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX-NEXT:    vmovaps %ymm2, (%rsi)
; AVX-NEXT:    vmovaps %ymm5, (%rdx)
; AVX-NEXT:    vmovaps %ymm6, (%rcx)
; AVX-NEXT:    vmovaps %ymm7, (%r8)
; AVX-NEXT:    vmovaps %ymm8, (%r9)
; AVX-NEXT:    vmovaps %ymm0, (%rax)
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i32_stride6_vf8:
; AVX2:       # %bb.0:
; AVX2-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-NEXT:    vmovaps 128(%rdi), %ymm0
; AVX2-NEXT:    vmovaps 160(%rdi), %ymm1
; AVX2-NEXT:    vmovaps 96(%rdi), %ymm5
; AVX2-NEXT:    vmovaps (%rdi), %ymm2
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm4
; AVX2-NEXT:    vmovaps 64(%rdi), %ymm6
; AVX2-NEXT:    vmovaps {{.*#+}} xmm3 = [0,6,4,u]
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm4[4,5],ymm2[6,7]
; AVX2-NEXT:    vpermps %ymm7, %ymm3, %ymm3
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm6[0,1],ymm5[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm9 = ymm8[0,1,2,3,4,5],ymm5[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm8 = ymm9[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm8 = [4,2,4,2,4,2,4,2]
; AVX2-NEXT:    vpermps %ymm10, %ymm8, %ymm11
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
; AVX2-NEXT:    vmovaps {{.*#+}} xmm11 = [1,7,5,u]
; AVX2-NEXT:    vpermps %ymm7, %ymm11, %ymm7
; AVX2-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm9 = [5,3,5,3,5,3,5,3]
; AVX2-NEXT:    vpermps %ymm10, %ymm9, %ymm10
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm10[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm10 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,2,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm11 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm12 = ymm11[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1,2],ymm10[3,4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm13 = ymm12[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm13[5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm13 = ymm5[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm6[1],ymm13[2,3,4],ymm6[5],ymm13[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,2,0,3]
; AVX2-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm13[3,4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm12[5,6,7]
; AVX2-NEXT:    vmovaps 80(%rdi), %xmm12
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm6 = ymm5[0,1],ymm12[2,3],ymm5[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm2, %ymm8, %ymm4
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm6[2,3,4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [0,2,0,6,0,2,0,6]
; AVX2-NEXT:    # ymm1 = mem[0,1,0,1]
; AVX2-NEXT:    vpermps %ymm0, %ymm1, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm4 = ymm5[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm12[3],ymm4[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm2, %ymm9, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3,4,5,6,7]
; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm4 = [0,3,1,7,0,3,1,7]
; AVX2-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX2-NEXT:    vpermps %ymm0, %ymm4, %ymm0
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
; AVX2-NEXT:    vmovaps %ymm3, (%rsi)
; AVX2-NEXT:    vmovaps %ymm7, (%rdx)
; AVX2-NEXT:    vmovaps %ymm10, (%rcx)
; AVX2-NEXT:    vmovaps %ymm11, (%r8)
; AVX2-NEXT:    vmovaps %ymm1, (%r9)
; AVX2-NEXT:    vmovaps %ymm0, (%rax)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i32_stride6_vf8:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovaps 160(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovaps 96(%rdi), %ymm5
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm2
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm4
; AVX2-FP-NEXT:    vmovaps 64(%rdi), %ymm6
; AVX2-FP-NEXT:    vmovaps {{.*#+}} xmm3 = [0,6,4,u]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm4[4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vpermps %ymm7, %ymm3, %ymm3
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm6[0,1],ymm5[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm9 = ymm8[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm8 = ymm9[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vbroadcastsd {{.*#+}} ymm8 = [4,2,4,2,4,2,4,2]
; AVX2-FP-NEXT:    vpermps %ymm10, %ymm8, %ymm11
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
; AVX2-FP-NEXT:    vmovaps {{.*#+}} xmm11 = [1,7,5,u]
; AVX2-FP-NEXT:    vpermps %ymm7, %ymm11, %ymm7
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
; AVX2-FP-NEXT:    vbroadcastsd {{.*#+}} ymm9 = [5,3,5,3,5,3,5,3]
; AVX2-FP-NEXT:    vpermps %ymm10, %ymm9, %ymm10
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm10 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,2,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm11 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm12 = ymm11[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1,2],ymm10[3,4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm13 = ymm12[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm13[5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm13 = ymm5[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm6[1],ymm13[2,3,4],ymm6[5],ymm13[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,2,0,3]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm13[3,4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm12[5,6,7]
; AVX2-FP-NEXT:    vmovaps 80(%rdi), %xmm12
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm6 = ymm5[0,1],ymm12[2,3],ymm5[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm8, %ymm4
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm6[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [0,2,0,6,0,2,0,6]
; AVX2-FP-NEXT:    # ymm1 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpermps %ymm0, %ymm1, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm4 = ymm5[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm12[3],ymm4[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm9, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vbroadcastf128 {{.*#+}} ymm4 = [0,3,1,7,0,3,1,7]
; AVX2-FP-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpermps %ymm0, %ymm4, %ymm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FP-NEXT:    vmovaps %ymm3, (%rsi)
; AVX2-FP-NEXT:    vmovaps %ymm7, (%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm10, (%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm11, (%r8)
; AVX2-FP-NEXT:    vmovaps %ymm1, (%r9)
; AVX2-FP-NEXT:    vmovaps %ymm0, (%rax)
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i32_stride6_vf8:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovaps 160(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovaps 96(%rdi), %ymm5
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm4
; AVX2-FCP-NEXT:    vmovaps 64(%rdi), %ymm6
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm3 = [0,6,4,u]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm4[4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm7, %ymm3, %ymm3
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm6[0,1],ymm5[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm9 = ymm8[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm8 = ymm9[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastsd {{.*#+}} ymm8 = [4,2,4,2,4,2,4,2]
; AVX2-FCP-NEXT:    vpermps %ymm10, %ymm8, %ymm11
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm11 = [1,7,5,u]
; AVX2-FCP-NEXT:    vpermps %ymm7, %ymm11, %ymm7
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastsd {{.*#+}} ymm9 = [5,3,5,3,5,3,5,3]
; AVX2-FCP-NEXT:    vpermps %ymm10, %ymm9, %ymm10
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm10 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
; AVX2-FCP-NEXT:    vbroadcastsd {{.*#+}} ymm11 = [2,4,2,4,2,4,2,4]
; AVX2-FCP-NEXT:    vpermps %ymm10, %ymm11, %ymm10
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm11 = [2,0,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm12 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm12, %ymm11, %ymm11
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2],ymm10[3,4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm13 = [0,0,6,4,0,0,6,4]
; AVX2-FCP-NEXT:    # ymm13 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm11, %ymm13, %ymm13
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm13[5,6,7]
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm13 = [0,1,7,5,0,1,7,5]
; AVX2-FCP-NEXT:    # ymm13 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm11, %ymm13, %ymm11
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm13 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0],ymm6[1],ymm13[2,3,4],ymm6[5],ymm13[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0],ymm12[1,2,3,4],ymm13[5],ymm12[6,7]
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} ymm13 = [3,1,7,5,0,u,u,u]
; AVX2-FCP-NEXT:    vpermps %ymm12, %ymm13, %ymm12
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4],ymm11[5,6,7]
; AVX2-FCP-NEXT:    vmovaps 80(%rdi), %xmm12
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm5[0,1],ymm12[2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm8, %ymm4
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm6[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm1 = [0,2,0,6,0,2,0,6]
; AVX2-FCP-NEXT:    # ymm1 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm1, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm4 = ymm5[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm12[3],ymm4[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm9, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm4 = [0,3,1,7,0,3,1,7]
; AVX2-FCP-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm4, %ymm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FCP-NEXT:    vmovaps %ymm3, (%rsi)
; AVX2-FCP-NEXT:    vmovaps %ymm7, (%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm10, (%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm11, (%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm1, (%r9)
; AVX2-FCP-NEXT:    vmovaps %ymm0, (%rax)
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i32_stride6_vf8:
; AVX512:       # %bb.0:
; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512-NEXT:    vmovdqa 128(%rdi), %ymm0
; AVX512-NEXT:    vmovdqa 160(%rdi), %ymm1
; AVX512-NEXT:    vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,6,12,18,24,30,0,0]
; AVX512-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm6 = [0,1,2,3,4,5,12,10]
; AVX512-NEXT:    vpermi2d %ymm4, %ymm5, %ymm6
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,7,13,19,25,31,0,0]
; AVX512-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,2,3,4,5,13,11]
; AVX512-NEXT:    vpermi2d %ymm4, %ymm5, %ymm7
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [0,0,0,0,0,0,6,12]
; AVX512-NEXT:    vpermi2d %ymm1, %ymm0, %ymm4
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [2,8,14,20,26,0,0,0]
; AVX512-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7]
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,0,0,0,0,1,7,13]
; AVX512-NEXT:    vpermi2d %ymm1, %ymm0, %ymm5
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512-NEXT:    vpermi2d %zmm3, %zmm2, %zmm8
; AVX512-NEXT:    vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3,4],ymm5[5,6,7]
; AVX512-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [20,26,0,6,12,0,0,0]
; AVX512-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [0,1,2,3,4,10,8,14]
; AVX512-NEXT:    vpermi2d %ymm0, %ymm1, %ymm8
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [21,27,1,7,13,0,0,0]
; AVX512-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,1,2,3,4,11,9,15]
; AVX512-NEXT:    vpermi2d %ymm0, %ymm1, %ymm2
; AVX512-NEXT:    vmovdqa %ymm6, (%rsi)
; AVX512-NEXT:    vmovdqa %ymm7, (%rdx)
; AVX512-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512-NEXT:    vmovdqa %ymm5, (%r8)
; AVX512-NEXT:    vmovdqa %ymm8, (%r9)
; AVX512-NEXT:    vmovdqa %ymm2, (%rax)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i32_stride6_vf8:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512-FCP-NEXT:    vmovdqa 128(%rdi), %ymm0
; AVX512-FCP-NEXT:    vmovdqa 160(%rdi), %ymm1
; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,6,12,18,24,30,0,0]
; AVX512-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm6 = [0,1,2,3,4,5,12,10]
; AVX512-FCP-NEXT:    vpermi2d %ymm4, %ymm5, %ymm6
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,7,13,19,25,31,0,0]
; AVX512-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,2,3,4,5,13,11]
; AVX512-FCP-NEXT:    vpermi2d %ymm4, %ymm5, %ymm7
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [0,0,0,0,0,0,6,12]
; AVX512-FCP-NEXT:    vpermi2d %ymm1, %ymm0, %ymm4
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [2,8,14,20,26,0,0,0]
; AVX512-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7]
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,0,0,0,0,1,7,13]
; AVX512-FCP-NEXT:    vpermi2d %ymm1, %ymm0, %ymm5
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm8
; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3,4],ymm5[5,6,7]
; AVX512-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [20,26,0,6,12,0,0,0]
; AVX512-FCP-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [0,1,2,3,4,10,8,14]
; AVX512-FCP-NEXT:    vpermi2d %ymm0, %ymm1, %ymm8
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [21,27,1,7,13,0,0,0]
; AVX512-FCP-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,1,2,3,4,11,9,15]
; AVX512-FCP-NEXT:    vpermi2d %ymm0, %ymm1, %ymm2
; AVX512-FCP-NEXT:    vmovdqa %ymm6, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa %ymm7, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512-FCP-NEXT:    vmovdqa %ymm5, (%r8)
; AVX512-FCP-NEXT:    vmovdqa %ymm8, (%r9)
; AVX512-FCP-NEXT:    vmovdqa %ymm2, (%rax)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i32_stride6_vf8:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512DQ-NEXT:    vmovdqa 128(%rdi), %ymm0
; AVX512DQ-NEXT:    vmovdqa 160(%rdi), %ymm1
; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,6,12,18,24,30,0,0]
; AVX512DQ-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm6 = [0,1,2,3,4,5,12,10]
; AVX512DQ-NEXT:    vpermi2d %ymm4, %ymm5, %ymm6
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,7,13,19,25,31,0,0]
; AVX512DQ-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,2,3,4,5,13,11]
; AVX512DQ-NEXT:    vpermi2d %ymm4, %ymm5, %ymm7
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [0,0,0,0,0,0,6,12]
; AVX512DQ-NEXT:    vpermi2d %ymm1, %ymm0, %ymm4
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [2,8,14,20,26,0,0,0]
; AVX512DQ-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7]
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,0,0,0,0,1,7,13]
; AVX512DQ-NEXT:    vpermi2d %ymm1, %ymm0, %ymm5
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512DQ-NEXT:    vpermi2d %zmm3, %zmm2, %zmm8
; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3,4],ymm5[5,6,7]
; AVX512DQ-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [20,26,0,6,12,0,0,0]
; AVX512DQ-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [0,1,2,3,4,10,8,14]
; AVX512DQ-NEXT:    vpermi2d %ymm0, %ymm1, %ymm8
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [21,27,1,7,13,0,0,0]
; AVX512DQ-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,1,2,3,4,11,9,15]
; AVX512DQ-NEXT:    vpermi2d %ymm0, %ymm1, %ymm2
; AVX512DQ-NEXT:    vmovdqa %ymm6, (%rsi)
; AVX512DQ-NEXT:    vmovdqa %ymm7, (%rdx)
; AVX512DQ-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512DQ-NEXT:    vmovdqa %ymm5, (%r8)
; AVX512DQ-NEXT:    vmovdqa %ymm8, (%r9)
; AVX512DQ-NEXT:    vmovdqa %ymm2, (%rax)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i32_stride6_vf8:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa 128(%rdi), %ymm0
; AVX512DQ-FCP-NEXT:    vmovdqa 160(%rdi), %ymm1
; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,6,12,18,24,30,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm6 = [0,1,2,3,4,5,12,10]
; AVX512DQ-FCP-NEXT:    vpermi2d %ymm4, %ymm5, %ymm6
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,7,13,19,25,31,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,2,3,4,5,13,11]
; AVX512DQ-FCP-NEXT:    vpermi2d %ymm4, %ymm5, %ymm7
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [0,0,0,0,0,0,6,12]
; AVX512DQ-FCP-NEXT:    vpermi2d %ymm1, %ymm0, %ymm4
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [2,8,14,20,26,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7]
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,0,0,0,0,1,7,13]
; AVX512DQ-FCP-NEXT:    vpermi2d %ymm1, %ymm0, %ymm5
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm8
; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3,4],ymm5[5,6,7]
; AVX512DQ-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [20,26,0,6,12,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [0,1,2,3,4,10,8,14]
; AVX512DQ-FCP-NEXT:    vpermi2d %ymm0, %ymm1, %ymm8
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [21,27,1,7,13,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,1,2,3,4,11,9,15]
; AVX512DQ-FCP-NEXT:    vpermi2d %ymm0, %ymm1, %ymm2
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm6, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm7, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm5, (%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm8, (%r9)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm2, (%rax)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i32_stride6_vf8:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512BW-NEXT:    vmovdqa 128(%rdi), %ymm0
; AVX512BW-NEXT:    vmovdqa 160(%rdi), %ymm1
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,6,12,18,24,30,0,0]
; AVX512BW-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm6 = [0,1,2,3,4,5,12,10]
; AVX512BW-NEXT:    vpermi2d %ymm4, %ymm5, %ymm6
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,7,13,19,25,31,0,0]
; AVX512BW-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,2,3,4,5,13,11]
; AVX512BW-NEXT:    vpermi2d %ymm4, %ymm5, %ymm7
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [0,0,0,0,0,0,6,12]
; AVX512BW-NEXT:    vpermi2d %ymm1, %ymm0, %ymm4
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [2,8,14,20,26,0,0,0]
; AVX512BW-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7]
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,0,0,0,0,1,7,13]
; AVX512BW-NEXT:    vpermi2d %ymm1, %ymm0, %ymm5
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512BW-NEXT:    vpermi2d %zmm3, %zmm2, %zmm8
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3,4],ymm5[5,6,7]
; AVX512BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [20,26,0,6,12,0,0,0]
; AVX512BW-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [0,1,2,3,4,10,8,14]
; AVX512BW-NEXT:    vpermi2d %ymm0, %ymm1, %ymm8
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [21,27,1,7,13,0,0,0]
; AVX512BW-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,1,2,3,4,11,9,15]
; AVX512BW-NEXT:    vpermi2d %ymm0, %ymm1, %ymm2
; AVX512BW-NEXT:    vmovdqa %ymm6, (%rsi)
; AVX512BW-NEXT:    vmovdqa %ymm7, (%rdx)
; AVX512BW-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512BW-NEXT:    vmovdqa %ymm5, (%r8)
; AVX512BW-NEXT:    vmovdqa %ymm8, (%r9)
; AVX512BW-NEXT:    vmovdqa %ymm2, (%rax)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i32_stride6_vf8:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa 128(%rdi), %ymm0
; AVX512BW-FCP-NEXT:    vmovdqa 160(%rdi), %ymm1
; AVX512BW-FCP-NEXT:    vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,6,12,18,24,30,0,0]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm6 = [0,1,2,3,4,5,12,10]
; AVX512BW-FCP-NEXT:    vpermi2d %ymm4, %ymm5, %ymm6
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,7,13,19,25,31,0,0]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,2,3,4,5,13,11]
; AVX512BW-FCP-NEXT:    vpermi2d %ymm4, %ymm5, %ymm7
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [0,0,0,0,0,0,6,12]
; AVX512BW-FCP-NEXT:    vpermi2d %ymm1, %ymm0, %ymm4
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [2,8,14,20,26,0,0,0]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512BW-FCP-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7]
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,0,0,0,0,1,7,13]
; AVX512BW-FCP-NEXT:    vpermi2d %ymm1, %ymm0, %ymm5
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm8
; AVX512BW-FCP-NEXT:    vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3,4],ymm5[5,6,7]
; AVX512BW-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [20,26,0,6,12,0,0,0]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [0,1,2,3,4,10,8,14]
; AVX512BW-FCP-NEXT:    vpermi2d %ymm0, %ymm1, %ymm8
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [21,27,1,7,13,0,0,0]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,1,2,3,4,11,9,15]
; AVX512BW-FCP-NEXT:    vpermi2d %ymm0, %ymm1, %ymm2
; AVX512BW-FCP-NEXT:    vmovdqa %ymm6, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa %ymm7, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa %ymm5, (%r8)
; AVX512BW-FCP-NEXT:    vmovdqa %ymm8, (%r9)
; AVX512BW-FCP-NEXT:    vmovdqa %ymm2, (%rax)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i32_stride6_vf8:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-BW-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa 128(%rdi), %ymm0
; AVX512DQ-BW-NEXT:    vmovdqa 160(%rdi), %ymm1
; AVX512DQ-BW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,6,12,18,24,30,0,0]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm6 = [0,1,2,3,4,5,12,10]
; AVX512DQ-BW-NEXT:    vpermi2d %ymm4, %ymm5, %ymm6
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,7,13,19,25,31,0,0]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,2,3,4,5,13,11]
; AVX512DQ-BW-NEXT:    vpermi2d %ymm4, %ymm5, %ymm7
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [0,0,0,0,0,0,6,12]
; AVX512DQ-BW-NEXT:    vpermi2d %ymm1, %ymm0, %ymm4
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [2,8,14,20,26,0,0,0]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512DQ-BW-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7]
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,0,0,0,0,1,7,13]
; AVX512DQ-BW-NEXT:    vpermi2d %ymm1, %ymm0, %ymm5
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm3, %zmm2, %zmm8
; AVX512DQ-BW-NEXT:    vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3,4],ymm5[5,6,7]
; AVX512DQ-BW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [20,26,0,6,12,0,0,0]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [0,1,2,3,4,10,8,14]
; AVX512DQ-BW-NEXT:    vpermi2d %ymm0, %ymm1, %ymm8
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [21,27,1,7,13,0,0,0]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,1,2,3,4,11,9,15]
; AVX512DQ-BW-NEXT:    vpermi2d %ymm0, %ymm1, %ymm2
; AVX512DQ-BW-NEXT:    vmovdqa %ymm6, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa %ymm7, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa %ymm5, (%r8)
; AVX512DQ-BW-NEXT:    vmovdqa %ymm8, (%r9)
; AVX512DQ-BW-NEXT:    vmovdqa %ymm2, (%rax)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i32_stride6_vf8:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 128(%rdi), %ymm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 160(%rdi), %ymm1
; AVX512DQ-BW-FCP-NEXT:    vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,6,12,18,24,30,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm6 = [0,1,2,3,4,5,12,10]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %ymm4, %ymm5, %ymm6
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [1,7,13,19,25,31,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,1,2,3,4,5,13,11]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %ymm4, %ymm5, %ymm7
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [0,0,0,0,0,0,6,12]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %ymm1, %ymm0, %ymm4
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [2,8,14,20,26,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [0,0,0,0,0,1,7,13]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %ymm1, %ymm0, %ymm5
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm3, %zmm2, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3,4],ymm5[5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [20,26,0,6,12,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [0,1,2,3,4,10,8,14]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %ymm0, %ymm1, %ymm8
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm1 = [21,27,1,7,13,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm2, %zmm3, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,1,2,3,4,11,9,15]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %ymm0, %ymm1, %ymm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm6, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm7, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm5, (%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm8, (%r9)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm2, (%rax)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <48 x i32>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <8 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42>
  %strided.vec1 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <8 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43>
  %strided.vec2 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <8 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44>
  %strided.vec3 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <8 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45>
  %strided.vec4 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <8 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46>
  %strided.vec5 = shufflevector <48 x i32> %wide.vec, <48 x i32> poison, <8 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47>
  store <8 x i32> %strided.vec0, ptr %out.vec0, align 64
  store <8 x i32> %strided.vec1, ptr %out.vec1, align 64
  store <8 x i32> %strided.vec2, ptr %out.vec2, align 64
  store <8 x i32> %strided.vec3, ptr %out.vec3, align 64
  store <8 x i32> %strided.vec4, ptr %out.vec4, align 64
  store <8 x i32> %strided.vec5, ptr %out.vec5, align 64
  ret void
}

define void @load_i32_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
; SSE-LABEL: load_i32_stride6_vf16:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $408, %rsp # imm = 0x198
; SSE-NEXT:    movdqa 240(%rdi), %xmm9
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 256(%rdi), %xmm3
; SSE-NEXT:    movdqa 192(%rdi), %xmm10
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 208(%rdi), %xmm4
; SSE-NEXT:    movdqa 336(%rdi), %xmm14
; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 352(%rdi), %xmm5
; SSE-NEXT:    movdqa 288(%rdi), %xmm15
; SSE-NEXT:    movdqa 304(%rdi), %xmm7
; SSE-NEXT:    movdqa 64(%rdi), %xmm12
; SSE-NEXT:    movdqa (%rdi), %xmm8
; SSE-NEXT:    movdqa 16(%rdi), %xmm1
; SSE-NEXT:    movdqa 48(%rdi), %xmm13
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
; SSE-NEXT:    movdqa %xmm1, %xmm11
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm8, %xmm1
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm15, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
; SSE-NEXT:    movdqa %xmm4, %xmm14
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm10, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa %xmm3, %xmm2
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 96(%rdi), %xmm4
; SSE-NEXT:    movdqa 112(%rdi), %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
; SSE-NEXT:    movdqa %xmm1, %xmm10
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm4, %xmm1
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 144(%rdi), %xmm9
; SSE-NEXT:    movdqa 160(%rdi), %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
; SSE-NEXT:    movdqa %xmm3, %xmm6
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,3,2,3]
; SSE-NEXT:    movdqa %xmm13, %xmm3
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm5, %xmm3
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm3, %xmm2
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,3,2,3]
; SSE-NEXT:    movdqa %xmm9, %xmm14
; SSE-NEXT:    movdqa %xmm9, %xmm2
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,2,3,3]
; SSE-NEXT:    movdqa 80(%rdi), %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,0,1,1]
; SSE-NEXT:    movdqa %xmm2, %xmm10
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,3,2,3]
; SSE-NEXT:    movdqa 32(%rdi), %xmm6
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
; SSE-NEXT:    movdqa 368(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa %xmm15, %xmm9
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
; SSE-NEXT:    movdqa 320(%rdi), %xmm8
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,2,3,3]
; SSE-NEXT:    movdqa 272(%rdi), %xmm15
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
; SSE-NEXT:    movdqa 224(%rdi), %xmm3
; 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:    pshufd {{.*#+}} xmm1 = xmm14[2,2,3,3]
; SSE-NEXT:    movdqa 176(%rdi), %xmm11
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3]
; SSE-NEXT:    movdqa 128(%rdi), %xmm4
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[3,3,3,3]
; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,3,2,3]
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[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:    pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[1,1,1,1]
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[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:    pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,1,1]
; SSE-NEXT:    movdqa %xmm3, %xmm9
; SSE-NEXT:    movdqa %xmm3, (%rsp) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; SSE-NEXT:    # xmm12 = mem[2,3,2,3]
; SSE-NEXT:    movdqa %xmm15, %xmm3
; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm15[0],xmm12[1],xmm15[1]
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm0[0],xmm12[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
; SSE-NEXT:    movdqa %xmm4, %xmm7
; 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:    pshufd {{.*#+}} xmm15 = xmm14[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm11[0],xmm15[1],xmm11[1]
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm6, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm10[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm14, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,2,3,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm7, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm0[0],xmm8[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:    pshufd $255, (%rsp), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rcx)
; SSE-NEXT:    movapd %xmm15, 16(%r8)
; SSE-NEXT:    movapd %xmm12, 32(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%r8)
; SSE-NEXT:    movapd %xmm2, 16(%r9)
; SSE-NEXT:    movapd %xmm3, 32(%r9)
; SSE-NEXT:    movapd %xmm4, 48(%r9)
; SSE-NEXT:    movapd %xmm5, (%r9)
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movapd %xmm13, 16(%rax)
; SSE-NEXT:    movapd %xmm9, 32(%rax)
; SSE-NEXT:    movapd %xmm8, 48(%rax)
; SSE-NEXT:    movapd %xmm10, (%rax)
; SSE-NEXT:    addq $408, %rsp # imm = 0x198
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i32_stride6_vf16:
; AVX:       # %bb.0:
; AVX-NEXT:    subq $328, %rsp # imm = 0x148
; AVX-NEXT:    vmovaps 224(%rdi), %ymm12
; AVX-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 288(%rdi), %ymm10
; AVX-NEXT:    vmovaps 256(%rdi), %ymm4
; AVX-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 160(%rdi), %ymm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 128(%rdi), %ymm13
; AVX-NEXT:    vmovaps 32(%rdi), %ymm6
; AVX-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps (%rdi), %ymm14
; AVX-NEXT:    vmovaps 96(%rdi), %ymm3
; AVX-NEXT:    vmovaps 64(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 96(%rdi), %ymm0, %ymm2
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm3[2,0],ymm2[0,0],ymm3[6,4],ymm2[4,4]
; AVX-NEXT:    vmovaps %ymm3, %ymm9
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,0],ymm2[2,2],ymm5[6,4],ymm2[6,6]
; AVX-NEXT:    vblendps {{.*#+}} ymm6 = ymm14[0,1,2,3],ymm6[4,5],ymm14[6,7]
; AVX-NEXT:    vextractf128 $1, %ymm6, %xmm7
; AVX-NEXT:    vblendps {{.*#+}} xmm11 = xmm6[0,1],xmm7[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[0,2],xmm7[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm5 = ymm11[0,1,2],ymm5[3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm13[2,3],ymm1[0,1]
; AVX-NEXT:    vshufpd {{.*#+}} ymm11 = ymm8[0],ymm13[0],ymm8[3],ymm13[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm11[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 288(%rdi), %ymm4, %ymm1
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm10[2,0],ymm1[0,0],ymm10[6,4],ymm1[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,0],ymm1[2,2],ymm5[6,4],ymm1[6,6]
; AVX-NEXT:    vmovaps 192(%rdi), %ymm15
; AVX-NEXT:    vblendps {{.*#+}} ymm3 = ymm15[0,1,2,3],ymm12[4,5],ymm15[6,7]
; AVX-NEXT:    vextractf128 $1, %ymm3, %xmm0
; AVX-NEXT:    vblendps {{.*#+}} xmm11 = xmm3[0,1],xmm0[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[0,2],xmm0[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm5 = ymm11[0,1,2],ymm5[3,4,5,6,7]
; AVX-NEXT:    vmovapd 352(%rdi), %ymm4
; AVX-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 320(%rdi), %ymm12
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm12[2,3],ymm4[0,1]
; AVX-NEXT:    vshufpd {{.*#+}} ymm11 = ymm4[0],ymm12[0],ymm4[3],ymm12[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm11[6,7]
; AVX-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups %ymm9, (%rsp) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm11 = ymm9[3,0],ymm2[1,0],ymm9[7,4],ymm2[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm11[2,0],ymm2[2,3],ymm11[6,4],ymm2[6,7]
; AVX-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[1,0],xmm7[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,2],xmm7[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1,2],ymm2[3,4,5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm6 = ymm8[3,1],ymm13[1,3],ymm8[7,5],ymm13[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm6[6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm10[3,0],ymm1[1,0],ymm10[7,4],ymm1[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm1[2,3],ymm2[6,4],ymm1[6,7]
; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm3[1,0],xmm0[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[0,2],xmm0[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[3,1],ymm12[1,3],ymm4[7,5],ymm12[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = mem[0,1],ymm14[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm9[2,1],ymm7[2,0],ymm9[6,5],ymm7[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm5
; AVX-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[2,0],xmm5[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX-NEXT:    vblendps {{.*#+}} ymm4 = ymm13[0,1,2,3],ymm9[4,5],ymm13[6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm4[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm8[0,0],ymm4[2,0],ymm8[4,4],ymm4[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm3[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm11 # 32-byte Folded Reload
; AVX-NEXT:    # ymm11 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[2,1],ymm3[2,0],ymm10[6,5],ymm3[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm11, %xmm15
; AVX-NEXT:    vshufps {{.*#+}} xmm14 = xmm11[2,0],xmm15[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm14[0,1,2],ymm1[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vblendps {{.*#+}} ymm14 = ymm12[0,1,2,3],ymm10[4,5],ymm12[6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm14[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm6 = ymm0[0,0],ymm14[2,0],ymm0[4,4],ymm14[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm6[5,6,7]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm4 = ymm8[0,1],ymm4[3,1],ymm8[4,5],ymm4[7,5]
; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[3,1],xmm5[3,3]
; AVX-NEXT:    vmovaps %ymm7, %ymm1
; AVX-NEXT:    vmovups (%rsp), %ymm7 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm7[3,1],ymm1[2,1],ymm7[7,5],ymm1[6,5]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm5[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm14[3,1],ymm0[4,5],ymm14[7,5]
; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm11[3,1],xmm15[3,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm4[3,1],ymm3[2,1],ymm4[7,5],ymm3[6,5]
; AVX-NEXT:    vmovaps %ymm3, %ymm15
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm5[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm13[2,3],ymm9[4,5,6,7]
; AVX-NEXT:    vmovaps 32(%rdi), %xmm3
; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm3[2,2,3,3]
; AVX-NEXT:    vmovaps 16(%rdi), %xmm5
; AVX-NEXT:    vblendps {{.*#+}} xmm2 = xmm5[0],xmm2[1],xmm5[2,3]
; AVX-NEXT:    vmovapd 80(%rdi), %xmm6
; AVX-NEXT:    vshufpd {{.*#+}} ymm8 = ymm6[1],ymm1[0],ymm6[2],ymm1[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm8 = ymm7[0,1],ymm8[2,0],ymm7[4,5],ymm8[6,4]
; AVX-NEXT:    vmovaps %ymm7, %ymm13
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm8[2,3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm9[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm11 = ymm9[2,0],ymm8[0,0],ymm9[6,4],ymm8[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm11 = ymm8[0,2],ymm11[2,0],ymm8[4,6],ymm11[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm11[5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm11 = ymm10[0,1],ymm12[2,3],ymm10[4,5,6,7]
; AVX-NEXT:    vmovaps 224(%rdi), %xmm12
; AVX-NEXT:    vshufps {{.*#+}} xmm14 = xmm12[2,2,3,3]
; AVX-NEXT:    vmovaps 208(%rdi), %xmm0
; AVX-NEXT:    vblendps {{.*#+}} xmm14 = xmm0[0],xmm14[1],xmm0[2,3]
; AVX-NEXT:    vmovapd 272(%rdi), %xmm1
; AVX-NEXT:    vshufpd {{.*#+}} ymm7 = ymm1[1],ymm15[0],ymm1[2],ymm15[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm7 = ymm4[0,1],ymm7[2,0],ymm4[4,5],ymm7[6,4]
; AVX-NEXT:    vmovaps %ymm4, %ymm10
; AVX-NEXT:    vblendps {{.*#+}} ymm7 = ymm14[0,1],ymm7[2,3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm11[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm4 = ymm11[2,0],ymm14[0,0],ymm11[6,4],ymm14[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm4 = ymm14[0,2],ymm4[2,0],ymm14[4,6],ymm4[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1,2,3,4],ymm4[5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm7 = ymm9[3,0],ymm8[1,0],ymm9[7,4],ymm8[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm7 = ymm8[0,3],ymm7[2,0],ymm8[4,7],ymm7[6,4]
; AVX-NEXT:    vblendps {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3]
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm5 # 32-byte Folded Reload
; AVX-NEXT:    # ymm5 = ymm6[3,1],mem[1,3],ymm6[7,5],mem[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm13[1,1],ymm5[2,0],ymm13[5,5],ymm5[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm7[5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm11[3,0],ymm14[1,0],ymm11[7,4],ymm14[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm14[0,3],ymm5[2,0],ymm14[4,7],ymm5[6,4]
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm12[2,3]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,1],ymm15[1,3],ymm1[7,5],ymm15[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[1,1],ymm1[2,0],ymm10[5,5],ymm1[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, 32(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, (%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, 32(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, (%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, 32(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, (%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, 32(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, (%r8)
; AVX-NEXT:    vmovaps %ymm4, 32(%r9)
; AVX-NEXT:    vmovaps %ymm2, (%r9)
; AVX-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX-NEXT:    vmovaps %ymm0, 32(%rax)
; AVX-NEXT:    vmovaps %ymm3, (%rax)
; AVX-NEXT:    addq $328, %rsp # imm = 0x148
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i32_stride6_vf16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    subq $392, %rsp # imm = 0x188
; AVX2-NEXT:    vmovaps 288(%rdi), %ymm10
; AVX2-NEXT:    vmovaps 224(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 192(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 128(%rdi), %ymm0
; AVX2-NEXT:    vmovaps 160(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 96(%rdi), %ymm15
; AVX2-NEXT:    vmovaps (%rdi), %ymm4
; AVX2-NEXT:    vmovups %ymm4, (%rsp) # 32-byte Spill
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm5
; AVX2-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 64(%rdi), %ymm13
; AVX2-NEXT:    vmovaps {{.*#+}} xmm6 = [0,6,4,u]
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
; AVX2-NEXT:    vpermps %ymm8, %ymm6, %ymm7
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm13[0,1],ymm15[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm15[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm11 = ymm9[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm11 = ymm7[0,1,2],ymm11[3,4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vmovaps %ymm0, %ymm7
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm12 = [4,2,4,2,4,2,4,2]
; AVX2-NEXT:    vpermps %ymm4, %ymm12, %ymm14
; AVX2-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4,5],ymm14[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX2-NEXT:    vpermps %ymm3, %ymm6, %ymm0
; AVX2-NEXT:    vmovaps 256(%rdi), %ymm11
; AVX2-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm11[0,1],ymm10[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm10[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm14 = ymm1[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2],ymm14[3,4,5,6,7]
; AVX2-NEXT:    vmovaps 320(%rdi), %ymm10
; AVX2-NEXT:    vmovaps 352(%rdi), %ymm6
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm10[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps %ymm10, %ymm5
; AVX2-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm0, %ymm12, %ymm10
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps {{.*#+}} xmm2 = [1,7,5,u]
; AVX2-NEXT:    vpermps %ymm8, %ymm2, %ymm8
; AVX2-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm14 = [5,3,5,3,5,3,5,3]
; AVX2-NEXT:    vpermps %ymm4, %ymm14, %ymm4
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm8[0,1,2,3,4,5],ymm4[6,7]
; AVX2-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm3, %ymm2, %ymm2
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm0, %ymm14, %ymm0
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm15[2,3],ymm13[4,5],ymm15[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vmovups (%rsp), %ymm9 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm10[0,1],ymm9[2,3],ymm10[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm12[4,5],ymm7[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1],ymm7[2,3],ymm11[4,5],ymm7[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = ymm3[0,1],mem[2,3],ymm3[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm4 = ymm3[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm8 = ymm4[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm15[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1],ymm0[2,3,4],ymm13[5],ymm0[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vmovaps %ymm7, %ymm6
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm4[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm13[0,1,2,3],ymm15[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm9[4,5,6,7]
; AVX2-NEXT:    vmovaps 80(%rdi), %xmm4
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm4[2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-NEXT:    vpermps %ymm3, %ymm7, %ymm5
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3,4,5,6,7]
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm5 = ymm12[0,1],mem[2,3],ymm12[4,5,6,7]
; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm10 = [0,2,0,6,0,2,0,6]
; AVX2-NEXT:    # ymm10 = mem[0,1,0,1]
; AVX2-NEXT:    vpermps %ymm5, %ymm10, %ymm13
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm13[5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm6 = ymm11[0,1,2,3],ymm6[4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm11 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm11 = ymm9[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovaps 272(%rdi), %xmm13
; AVX2-NEXT:    vblendps {{.*#+}} ymm15 = ymm6[0,1],ymm13[2,3],ymm6[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vpermps %ymm11, %ymm7, %ymm7
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm12 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm12 = mem[0,1],ymm9[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm12, %ymm10, %ymm10
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm10[5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm3, %ymm14, %ymm3
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = [0,3,1,7,0,3,1,7]
; AVX2-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX2-NEXT:    vpermps %ymm5, %ymm3, %ymm4
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm4 = ymm6[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm13[3],ymm4[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm11, %ymm14, %ymm5
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm12, %ymm3, %ymm3
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 32(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, (%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 32(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, (%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 32(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, (%rcx)
; AVX2-NEXT:    vmovaps %ymm8, 32(%r8)
; AVX2-NEXT:    vmovaps %ymm0, (%r8)
; AVX2-NEXT:    vmovaps %ymm7, 32(%r9)
; AVX2-NEXT:    vmovaps %ymm1, (%r9)
; AVX2-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-NEXT:    vmovaps %ymm3, 32(%rax)
; AVX2-NEXT:    vmovaps %ymm2, (%rax)
; AVX2-NEXT:    addq $392, %rsp # imm = 0x188
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i32_stride6_vf16:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    subq $392, %rsp # imm = 0x188
; AVX2-FP-NEXT:    vmovaps 288(%rdi), %ymm10
; AVX2-FP-NEXT:    vmovaps 224(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 192(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovaps 160(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 96(%rdi), %ymm15
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm4
; AVX2-FP-NEXT:    vmovups %ymm4, (%rsp) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm5
; AVX2-FP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 64(%rdi), %ymm13
; AVX2-FP-NEXT:    vmovaps {{.*#+}} xmm6 = [0,6,4,u]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
; AVX2-FP-NEXT:    vpermps %ymm8, %ymm6, %ymm7
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm13[0,1],ymm15[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm11 = ymm9[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm11 = ymm7[0,1,2],ymm11[3,4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps %ymm0, %ymm7
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vbroadcastsd {{.*#+}} ymm12 = [4,2,4,2,4,2,4,2]
; AVX2-FP-NEXT:    vpermps %ymm4, %ymm12, %ymm14
; AVX2-FP-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4,5],ymm14[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vpermps %ymm3, %ymm6, %ymm0
; AVX2-FP-NEXT:    vmovaps 256(%rdi), %ymm11
; AVX2-FP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm11[0,1],ymm10[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm14 = ymm1[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2],ymm14[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 320(%rdi), %ymm10
; AVX2-FP-NEXT:    vmovaps 352(%rdi), %ymm6
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps %ymm10, %ymm5
; AVX2-FP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm0, %ymm12, %ymm10
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps {{.*#+}} xmm2 = [1,7,5,u]
; AVX2-FP-NEXT:    vpermps %ymm8, %ymm2, %ymm8
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
; AVX2-FP-NEXT:    vbroadcastsd {{.*#+}} ymm14 = [5,3,5,3,5,3,5,3]
; AVX2-FP-NEXT:    vpermps %ymm4, %ymm14, %ymm4
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm8[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm3, %ymm2, %ymm2
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm0, %ymm14, %ymm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm15[2,3],ymm13[4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vmovups (%rsp), %ymm9 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm10[0,1],ymm9[2,3],ymm10[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm12[4,5],ymm7[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1],ymm7[2,3],ymm11[4,5],ymm7[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = ymm3[0,1],mem[2,3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm4 = ymm3[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm8 = ymm4[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm15[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1],ymm0[2,3,4],ymm13[5],ymm0[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vmovaps %ymm7, %ymm6
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm4[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm13[0,1,2,3],ymm15[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 80(%rdi), %xmm4
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm4[2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FP-NEXT:    vpermps %ymm3, %ymm7, %ymm5
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm5 = ymm12[0,1],mem[2,3],ymm12[4,5,6,7]
; AVX2-FP-NEXT:    vbroadcastf128 {{.*#+}} ymm10 = [0,2,0,6,0,2,0,6]
; AVX2-FP-NEXT:    # ymm10 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpermps %ymm5, %ymm10, %ymm13
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm13[5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm6 = ymm11[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm11 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm11 = ymm9[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 272(%rdi), %xmm13
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm15 = ymm6[0,1],ymm13[2,3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vpermps %ymm11, %ymm7, %ymm7
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm12 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm12 = mem[0,1],ymm9[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm12, %ymm10, %ymm10
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm10[5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm3, %ymm14, %ymm3
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = [0,3,1,7,0,3,1,7]
; AVX2-FP-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpermps %ymm5, %ymm3, %ymm4
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm4 = ymm6[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm13[3],ymm4[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm11, %ymm14, %ymm5
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm12, %ymm3, %ymm3
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 32(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, (%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 32(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, (%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 32(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, (%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm8, 32(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm0, (%r8)
; AVX2-FP-NEXT:    vmovaps %ymm7, 32(%r9)
; AVX2-FP-NEXT:    vmovaps %ymm1, (%r9)
; AVX2-FP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FP-NEXT:    vmovaps %ymm3, 32(%rax)
; AVX2-FP-NEXT:    vmovaps %ymm2, (%rax)
; AVX2-FP-NEXT:    addq $392, %rsp # imm = 0x188
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i32_stride6_vf16:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    subq $360, %rsp # imm = 0x168
; AVX2-FCP-NEXT:    vmovaps 288(%rdi), %ymm10
; AVX2-FCP-NEXT:    vmovaps 224(%rdi), %ymm6
; AVX2-FCP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 192(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 160(%rdi), %ymm13
; AVX2-FCP-NEXT:    vmovaps 96(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm4
; AVX2-FCP-NEXT:    vmovups %ymm4, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm5
; AVX2-FCP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 64(%rdi), %ymm15
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm12 = [0,6,4,u]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm8, %ymm12, %ymm7
; AVX2-FCP-NEXT:    vmovaps %ymm1, %ymm5
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm15[0,1],ymm1[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm11 = ymm9[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm11 = ymm7[0,1,2],ymm11[3,4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm13[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastsd {{.*#+}} ymm1 = [4,2,4,2,4,2,4,2]
; AVX2-FCP-NEXT:    vpermps %ymm4, %ymm1, %ymm14
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4,5],ymm14[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm6[4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm3, %ymm12, %ymm12
; AVX2-FCP-NEXT:    vmovaps 256(%rdi), %ymm11
; AVX2-FCP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm11[0,1],ymm10[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm14[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm14 = ymm6[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm12[0,1,2],ymm14[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 320(%rdi), %ymm7
; AVX2-FCP-NEXT:    vmovaps 352(%rdi), %ymm12
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm1, %ymm10
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm2 = [1,7,5,u]
; AVX2-FCP-NEXT:    vpermps %ymm8, %ymm2, %ymm8
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastsd {{.*#+}} ymm14 = [5,3,5,3,5,3,5,3]
; AVX2-FCP-NEXT:    vpermps %ymm4, %ymm14, %ymm4
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm8[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm3, %ymm2, %ymm2
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm14, %ymm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm5[2,3],ymm15[4,5],ymm5[6,7]
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} ymm1 = [2,0,6,4,2,0,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm1, %ymm0
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm2 = [2,0,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovups (%rsp), %ymm9 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm9[2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm3, %ymm2, %ymm4
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FCP-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm4 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm4 = mem[0,1,2,3],ymm13[4,5],mem[6,7]
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm8 = [0,0,6,4,0,0,6,4]
; AVX2-FCP-NEXT:    # ymm8 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm4, %ymm8, %ymm10
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm10[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1],ymm6[2,3],ymm11[4,5],ymm6[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm1, %ymm0
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm1 = ymm1[0,1],mem[2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm1, %ymm2, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm12[4,5],ymm7[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm8, %ymm8
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm15[1],ymm0[2,3,4],ymm15[5],ymm0[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1,2,3,4],ymm0[5],ymm3[6,7]
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = [0,1,7,5,0,1,7,5]
; AVX2-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm4, %ymm3, %ymm4
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} ymm8 = [3,1,7,5,0,u,u,u]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm8, %ymm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm4 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3,4],ymm11[5],ymm4[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0],ymm1[1,2,3,4],ymm4[5],ymm1[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm1, %ymm8, %ymm1
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm3, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm15[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 80(%rdi), %xmm4
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm4[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FCP-NEXT:    vpermps %ymm3, %ymm7, %ymm5
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm5 = ymm13[0,1],mem[2,3],ymm13[4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm10 = [0,2,0,6,0,2,0,6]
; AVX2-FCP-NEXT:    # ymm10 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm5, %ymm10, %ymm13
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm13[5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm11[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm11 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm11 = ymm9[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 272(%rdi), %xmm13
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm15 = ymm6[0,1],ymm13[2,3],ymm6[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vpermps %ymm11, %ymm7, %ymm7
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm12 = ymm12[0,1],mem[2,3],ymm12[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm12, %ymm10, %ymm10
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm10[5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm3, %ymm14, %ymm3
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = [0,3,1,7,0,3,1,7]
; AVX2-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm5, %ymm3, %ymm4
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm4 = ymm6[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm13[3],ymm4[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm11, %ymm14, %ymm5
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm12, %ymm3, %ymm3
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm4, 32(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm4, (%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm4, 32(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm4, (%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm4, 32(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm4, (%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm8, 32(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm0, (%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm7, 32(%r9)
; AVX2-FCP-NEXT:    vmovaps %ymm1, (%r9)
; AVX2-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FCP-NEXT:    vmovaps %ymm3, 32(%rax)
; AVX2-FCP-NEXT:    vmovaps %ymm2, (%rax)
; AVX2-FCP-NEXT:    addq $360, %rsp # imm = 0x168
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i32_stride6_vf16:
; AVX512:       # %bb.0:
; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm3
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm4
; AVX512-NEXT:    vmovdqa64 128(%rdi), %zmm5
; AVX512-NEXT:    vmovdqa64 192(%rdi), %zmm6
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2d %zmm5, %zmm6, %zmm7
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,6,12,18,24,30,0,0]
; AVX512-NEXT:    vpermi2d %zmm4, %zmm3, %zmm2
; AVX512-NEXT:    movb $56, %dil
; AVX512-NEXT:    kmovw %edi, %k2
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm2 {%k2}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2d %zmm0, %zmm1, %zmm7
; AVX512-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512-NEXT:    kmovw %edi, %k1
; AVX512-NEXT:    vmovdqa32 %zmm7, %zmm2 {%k1}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2d %zmm5, %zmm6, %zmm8
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [1,7,13,19,25,31,0,0]
; AVX512-NEXT:    vpermi2d %zmm4, %zmm3, %zmm7
; AVX512-NEXT:    vmovdqa64 %zmm8, %zmm7 {%k2}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512-NEXT:    vmovdqa32 %zmm8, %zmm7 {%k1}
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [2,8,14,20,26,0,0,0]
; AVX512-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2d %zmm6, %zmm5, %zmm9
; AVX512-NEXT:    movw $31, %di
; AVX512-NEXT:    kmovw %edi, %k2
; AVX512-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k2}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k1}
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2d %zmm6, %zmm5, %zmm10
; AVX512-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k2}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k1}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm11 = [20,26,0,6,12,0,0,0]
; AVX512-NEXT:    vpermi2d %zmm3, %zmm4, %zmm11
; AVX512-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512-NEXT:    kmovw %edi, %k1
; AVX512-NEXT:    vmovdqa32 %zmm8, %zmm11 {%k1}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512-NEXT:    movb $-32, %dil
; AVX512-NEXT:    kmovw %edi, %k2
; AVX512-NEXT:    vmovdqa64 %zmm8, %zmm11 {%k2}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [21,27,1,7,13,0,0,0]
; AVX512-NEXT:    vpermi2d %zmm3, %zmm4, %zmm5
; AVX512-NEXT:    vmovdqa32 %zmm8, %zmm5 {%k1}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermi2d %zmm0, %zmm1, %zmm3
; AVX512-NEXT:    vmovdqa64 %zmm3, %zmm5 {%k2}
; AVX512-NEXT:    vmovdqa64 %zmm2, (%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm7, (%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm9, (%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm10, (%r8)
; AVX512-NEXT:    vmovdqa64 %zmm11, (%r9)
; AVX512-NEXT:    vmovdqa64 %zmm5, (%rax)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i32_stride6_vf16:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm4
; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm5
; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm6
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermi2d %zmm5, %zmm6, %zmm7
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,6,12,18,24,30,0,0]
; AVX512-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm2
; AVX512-FCP-NEXT:    movb $56, %dil
; AVX512-FCP-NEXT:    kmovw %edi, %k2
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm2 {%k2}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm7
; AVX512-FCP-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512-FCP-NEXT:    kmovw %edi, %k1
; AVX512-FCP-NEXT:    vmovdqa32 %zmm7, %zmm2 {%k1}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermi2d %zmm5, %zmm6, %zmm8
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [1,7,13,19,25,31,0,0]
; AVX512-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm7
; AVX512-FCP-NEXT:    vmovdqa64 %zmm8, %zmm7 {%k2}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512-FCP-NEXT:    vmovdqa32 %zmm8, %zmm7 {%k1}
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [2,8,14,20,26,0,0,0]
; AVX512-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512-FCP-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm9
; AVX512-FCP-NEXT:    movw $31, %di
; AVX512-FCP-NEXT:    kmovw %edi, %k2
; AVX512-FCP-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k2}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512-FCP-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k1}
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512-FCP-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm10
; AVX512-FCP-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k2}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512-FCP-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k1}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm11 = [20,26,0,6,12,0,0,0]
; AVX512-FCP-NEXT:    vpermi2d %zmm3, %zmm4, %zmm11
; AVX512-FCP-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512-FCP-NEXT:    kmovw %edi, %k1
; AVX512-FCP-NEXT:    vmovdqa32 %zmm8, %zmm11 {%k1}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512-FCP-NEXT:    movb $-32, %dil
; AVX512-FCP-NEXT:    kmovw %edi, %k2
; AVX512-FCP-NEXT:    vmovdqa64 %zmm8, %zmm11 {%k2}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [21,27,1,7,13,0,0,0]
; AVX512-FCP-NEXT:    vpermi2d %zmm3, %zmm4, %zmm5
; AVX512-FCP-NEXT:    vmovdqa32 %zmm8, %zmm5 {%k1}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512-FCP-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, %zmm5 {%k2}
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm9, (%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm10, (%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, (%r9)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, (%rax)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i32_stride6_vf16:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm3
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm4
; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %zmm5
; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %zmm6
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512DQ-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermi2d %zmm5, %zmm6, %zmm7
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,6,12,18,24,30,0,0]
; AVX512DQ-NEXT:    vpermi2d %zmm4, %zmm3, %zmm2
; AVX512DQ-NEXT:    movb $56, %dil
; AVX512DQ-NEXT:    kmovw %edi, %k2
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm2 {%k2}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512DQ-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermi2d %zmm0, %zmm1, %zmm7
; AVX512DQ-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512DQ-NEXT:    kmovw %edi, %k1
; AVX512DQ-NEXT:    vmovdqa32 %zmm7, %zmm2 {%k1}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512DQ-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermi2d %zmm5, %zmm6, %zmm8
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [1,7,13,19,25,31,0,0]
; AVX512DQ-NEXT:    vpermi2d %zmm4, %zmm3, %zmm7
; AVX512DQ-NEXT:    vmovdqa64 %zmm8, %zmm7 {%k2}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512DQ-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512DQ-NEXT:    vmovdqa32 %zmm8, %zmm7 {%k1}
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [2,8,14,20,26,0,0,0]
; AVX512DQ-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512DQ-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermi2d %zmm6, %zmm5, %zmm9
; AVX512DQ-NEXT:    movw $31, %di
; AVX512DQ-NEXT:    kmovw %edi, %k2
; AVX512DQ-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k2}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512DQ-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512DQ-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k1}
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512DQ-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512DQ-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermi2d %zmm6, %zmm5, %zmm10
; AVX512DQ-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k2}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512DQ-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512DQ-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k1}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512DQ-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm11 = [20,26,0,6,12,0,0,0]
; AVX512DQ-NEXT:    vpermi2d %zmm3, %zmm4, %zmm11
; AVX512DQ-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512DQ-NEXT:    kmovw %edi, %k1
; AVX512DQ-NEXT:    vmovdqa32 %zmm8, %zmm11 {%k1}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512DQ-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512DQ-NEXT:    movb $-32, %dil
; AVX512DQ-NEXT:    kmovw %edi, %k2
; AVX512DQ-NEXT:    vmovdqa64 %zmm8, %zmm11 {%k2}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512DQ-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [21,27,1,7,13,0,0,0]
; AVX512DQ-NEXT:    vpermi2d %zmm3, %zmm4, %zmm5
; AVX512DQ-NEXT:    vmovdqa32 %zmm8, %zmm5 {%k1}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512DQ-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermi2d %zmm0, %zmm1, %zmm3
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, %zmm5 {%k2}
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, (%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, (%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm9, (%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm10, (%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, (%r9)
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, (%rax)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i32_stride6_vf16:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm5
; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm6
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512DQ-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm5, %zmm6, %zmm7
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,6,12,18,24,30,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm2
; AVX512DQ-FCP-NEXT:    movb $56, %dil
; AVX512DQ-FCP-NEXT:    kmovw %edi, %k2
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm2 {%k2}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512DQ-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm7
; AVX512DQ-FCP-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512DQ-FCP-NEXT:    kmovw %edi, %k1
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm7, %zmm2 {%k1}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512DQ-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm5, %zmm6, %zmm8
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [1,7,13,19,25,31,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm8, %zmm7 {%k2}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512DQ-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm8, %zmm7 {%k1}
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [2,8,14,20,26,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512DQ-FCP-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm9
; AVX512DQ-FCP-NEXT:    movw $31, %di
; AVX512DQ-FCP-NEXT:    kmovw %edi, %k2
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k2}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512DQ-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k1}
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512DQ-FCP-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm10
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k2}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512DQ-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k1}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512DQ-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm11 = [20,26,0,6,12,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm3, %zmm4, %zmm11
; AVX512DQ-FCP-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512DQ-FCP-NEXT:    kmovw %edi, %k1
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm8, %zmm11 {%k1}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512DQ-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512DQ-FCP-NEXT:    movb $-32, %dil
; AVX512DQ-FCP-NEXT:    kmovw %edi, %k2
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm8, %zmm11 {%k2}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512DQ-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [21,27,1,7,13,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm3, %zmm4, %zmm5
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm8, %zmm5 {%k1}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512DQ-FCP-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, %zmm5 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm9, (%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm10, (%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, (%r9)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, (%rax)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i32_stride6_vf16:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm3
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm4
; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm5
; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm6
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512BW-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermi2d %zmm5, %zmm6, %zmm7
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,6,12,18,24,30,0,0]
; AVX512BW-NEXT:    vpermi2d %zmm4, %zmm3, %zmm2
; AVX512BW-NEXT:    movb $56, %dil
; AVX512BW-NEXT:    kmovd %edi, %k2
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm2 {%k2}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512BW-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm7
; AVX512BW-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512BW-NEXT:    kmovd %edi, %k1
; AVX512BW-NEXT:    vmovdqa32 %zmm7, %zmm2 {%k1}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermi2d %zmm5, %zmm6, %zmm8
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [1,7,13,19,25,31,0,0]
; AVX512BW-NEXT:    vpermi2d %zmm4, %zmm3, %zmm7
; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm7 {%k2}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512BW-NEXT:    vmovdqa32 %zmm8, %zmm7 {%k1}
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [2,8,14,20,26,0,0,0]
; AVX512BW-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512BW-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermi2d %zmm6, %zmm5, %zmm9
; AVX512BW-NEXT:    movw $31, %di
; AVX512BW-NEXT:    kmovd %edi, %k2
; AVX512BW-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k2}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512BW-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k1}
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512BW-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512BW-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermi2d %zmm6, %zmm5, %zmm10
; AVX512BW-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k2}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512BW-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k1}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm11 = [20,26,0,6,12,0,0,0]
; AVX512BW-NEXT:    vpermi2d %zmm3, %zmm4, %zmm11
; AVX512BW-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512BW-NEXT:    kmovd %edi, %k1
; AVX512BW-NEXT:    vmovdqa32 %zmm8, %zmm11 {%k1}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512BW-NEXT:    movb $-32, %dil
; AVX512BW-NEXT:    kmovd %edi, %k2
; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm11 {%k2}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [21,27,1,7,13,0,0,0]
; AVX512BW-NEXT:    vpermi2d %zmm3, %zmm4, %zmm5
; AVX512BW-NEXT:    vmovdqa32 %zmm8, %zmm5 {%k1}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512BW-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm3
; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm5 {%k2}
; AVX512BW-NEXT:    vmovdqa64 %zmm2, (%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm7, (%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm9, (%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm10, (%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm11, (%r9)
; AVX512BW-NEXT:    vmovdqa64 %zmm5, (%rax)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i32_stride6_vf16:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm4
; AVX512BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm5
; AVX512BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm6
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512BW-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm5, %zmm6, %zmm7
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,6,12,18,24,30,0,0]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm2
; AVX512BW-FCP-NEXT:    movb $56, %dil
; AVX512BW-FCP-NEXT:    kmovd %edi, %k2
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm2 {%k2}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512BW-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm7
; AVX512BW-FCP-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512BW-FCP-NEXT:    kmovd %edi, %k1
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm7, %zmm2 {%k1}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm5, %zmm6, %zmm8
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [1,7,13,19,25,31,0,0]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm8, %zmm7 {%k2}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm7 {%k1}
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [2,8,14,20,26,0,0,0]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512BW-FCP-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm9
; AVX512BW-FCP-NEXT:    movw $31, %di
; AVX512BW-FCP-NEXT:    kmovd %edi, %k2
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k2}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k1}
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512BW-FCP-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm10
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k2}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k1}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm11 = [20,26,0,6,12,0,0,0]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm3, %zmm4, %zmm11
; AVX512BW-FCP-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512BW-FCP-NEXT:    kmovd %edi, %k1
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm11 {%k1}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512BW-FCP-NEXT:    movb $-32, %dil
; AVX512BW-FCP-NEXT:    kmovd %edi, %k2
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm8, %zmm11 {%k2}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [21,27,1,7,13,0,0,0]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm3, %zmm4, %zmm5
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm5 {%k1}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512BW-FCP-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm5 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm9, (%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm10, (%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, (%r9)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, (%rax)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i32_stride6_vf16:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-BW-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 (%rdi), %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 64(%rdi), %zmm4
; AVX512DQ-BW-NEXT:    vmovdqa64 128(%rdi), %zmm5
; AVX512DQ-BW-NEXT:    vmovdqa64 192(%rdi), %zmm6
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512DQ-BW-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm5, %zmm6, %zmm7
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,6,12,18,24,30,0,0]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm4, %zmm3, %zmm2
; AVX512DQ-BW-NEXT:    movb $56, %dil
; AVX512DQ-BW-NEXT:    kmovd %edi, %k2
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm2 {%k2}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512DQ-BW-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm7
; AVX512DQ-BW-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512DQ-BW-NEXT:    kmovd %edi, %k1
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm7, %zmm2 {%k1}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512DQ-BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm5, %zmm6, %zmm8
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [1,7,13,19,25,31,0,0]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm4, %zmm3, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm8, %zmm7 {%k2}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512DQ-BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm8, %zmm7 {%k1}
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [2,8,14,20,26,0,0,0]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512DQ-BW-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm6, %zmm5, %zmm9
; AVX512DQ-BW-NEXT:    movw $31, %di
; AVX512DQ-BW-NEXT:    kmovd %edi, %k2
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k2}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512DQ-BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k1}
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512DQ-BW-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm6, %zmm5, %zmm10
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k2}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512DQ-BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k1}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512DQ-BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm11 = [20,26,0,6,12,0,0,0]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm3, %zmm4, %zmm11
; AVX512DQ-BW-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512DQ-BW-NEXT:    kmovd %edi, %k1
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm8, %zmm11 {%k1}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512DQ-BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512DQ-BW-NEXT:    movb $-32, %dil
; AVX512DQ-BW-NEXT:    kmovd %edi, %k2
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm8, %zmm11 {%k2}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512DQ-BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [21,27,1,7,13,0,0,0]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm3, %zmm4, %zmm5
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm8, %zmm5 {%k1}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512DQ-BW-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, %zmm5 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm9, (%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm10, (%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, (%r9)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, (%rax)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i32_stride6_vf16:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm5
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm6
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512DQ-BW-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm5, %zmm6, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm2 = [0,6,12,18,24,30,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm2
; AVX512DQ-BW-FCP-NEXT:    movb $56, %dil
; AVX512DQ-BW-FCP-NEXT:    kmovd %edi, %k2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm2 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512DQ-BW-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm7
; AVX512DQ-BW-FCP-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512DQ-BW-FCP-NEXT:    kmovd %edi, %k1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm7, %zmm2 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512DQ-BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm5, %zmm6, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [1,7,13,19,25,31,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm8, %zmm7 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512DQ-BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm7 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [2,8,14,20,26,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512DQ-BW-FCP-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm9
; AVX512DQ-BW-FCP-NEXT:    movw $31, %di
; AVX512DQ-BW-FCP-NEXT:    kmovd %edi, %k2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512DQ-BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm9 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm8 = [3,9,15,21,27,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm4, %zmm3, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512DQ-BW-FCP-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm10
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512DQ-BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm10 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512DQ-BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm11 = [20,26,0,6,12,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm3, %zmm4, %zmm11
; AVX512DQ-BW-FCP-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512DQ-BW-FCP-NEXT:    kmovd %edi, %k1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm11 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512DQ-BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm8
; AVX512DQ-BW-FCP-NEXT:    movb $-32, %dil
; AVX512DQ-BW-FCP-NEXT:    kmovd %edi, %k2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm8, %zmm11 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512DQ-BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm6, %zmm5, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [21,27,1,7,13,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm3, %zmm4, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm8, %zmm5 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512DQ-BW-FCP-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm5 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm9, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm10, (%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, (%r9)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, (%rax)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <96 x i32>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <16 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90>
  %strided.vec1 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <16 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91>
  %strided.vec2 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <16 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92>
  %strided.vec3 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <16 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93>
  %strided.vec4 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <16 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94>
  %strided.vec5 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <16 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95>
  store <16 x i32> %strided.vec0, ptr %out.vec0, align 64
  store <16 x i32> %strided.vec1, ptr %out.vec1, align 64
  store <16 x i32> %strided.vec2, ptr %out.vec2, align 64
  store <16 x i32> %strided.vec3, ptr %out.vec3, align 64
  store <16 x i32> %strided.vec4, ptr %out.vec4, align 64
  store <16 x i32> %strided.vec5, ptr %out.vec5, align 64
  ret void
}

define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
; SSE-LABEL: load_i32_stride6_vf32:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $1032, %rsp # imm = 0x408
; SSE-NEXT:    movdqa 64(%rdi), %xmm5
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa (%rdi), %xmm12
; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 16(%rdi), %xmm13
; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 48(%rdi), %xmm9
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 528(%rdi), %xmm7
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 544(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 480(%rdi), %xmm8
; SSE-NEXT:    movdqa 496(%rdi), %xmm4
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 144(%rdi), %xmm10
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 160(%rdi), %xmm2
; SSE-NEXT:    movdqa 96(%rdi), %xmm6
; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 112(%rdi), %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
; SSE-NEXT:    movdqa %xmm1, %xmm11
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm6, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
; SSE-NEXT:    movdqa %xmm2, %xmm6
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
; SSE-NEXT:    movdqa %xmm8, %xmm1
; SSE-NEXT:    movdqa %xmm8, %xmm4
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,3,2,3]
; SSE-NEXT:    movdqa %xmm12, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 384(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 400(%rdi), %xmm14
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,3,2,3]
; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm2, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 432(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, (%rsp) # 16-byte Spill
; SSE-NEXT:    movdqa 448(%rdi), %xmm9
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 288(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 304(%rdi), %xmm15
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm2, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 336(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 352(%rdi), %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
; SSE-NEXT:    movdqa %xmm2, %xmm12
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 672(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 688(%rdi), %xmm8
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,3,2,3]
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm2, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 720(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 736(%rdi), %xmm5
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 192(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 208(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 240(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 256(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 576(%rdi), %xmm7
; SSE-NEXT:    movdqa 592(%rdi), %xmm13
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,3,2,3]
; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm7, %xmm1
; 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 624(%rdi), %xmm10
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 640(%rdi), %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm10, %xmm2
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm11, %xmm2
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm4, %xmm2
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,3,2,3]
; SSE-NEXT:    movdqa (%rsp), %xmm2 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm15, %xmm2
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm8, %xmm2
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm14, %xmm5
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm0[0],xmm5[1]
; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm5, %xmm3
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[2,2,3,3]
; SSE-NEXT:    movdqa 176(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
; SSE-NEXT:    movdqa 128(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; 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:    pshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
; SSE-NEXT:    movdqa 80(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
; SSE-NEXT:    movdqa 32(%rdi), %xmm13
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,3,3]
; SSE-NEXT:    movdqa 368(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
; SSE-NEXT:    movdqa 320(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; 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:    pshufd {{.*#+}} xmm1 = xmm14[2,2,3,3]
; SSE-NEXT:    movdqa 272(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,3,2,3]
; SSE-NEXT:    movdqa %xmm12, %xmm14
; SSE-NEXT:    movdqa 224(%rdi), %xmm6
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
; SSE-NEXT:    movdqa %xmm6, {{[-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:    pshufd {{.*#+}} xmm1 = xmm11[2,2,3,3]
; SSE-NEXT:    movdqa 560(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa 512(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; 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 (%rsp), %xmm11 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,2,3,3]
; SSE-NEXT:    movdqa 464(%rdi), %xmm15
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,0,1,1]
; 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:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
; SSE-NEXT:    movdqa 416(%rdi), %xmm4
; 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 {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,2,3,3]
; SSE-NEXT:    movdqa 752(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa 704(%rdi), %xmm12
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
; SSE-NEXT:    movdqa %xmm12, {{[-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:    pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
; SSE-NEXT:    movdqa 656(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
; SSE-NEXT:    movdqa 608(%rdi), %xmm5
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    movdqa %xmm13, %xmm9
; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[3,3,3,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[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:    pshufd {{.*#+}} xmm0 = xmm14[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,3,2,3]
; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, (%rsp) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[3,3,3,3]
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[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:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[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:    pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm12, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; SSE-NEXT:    # xmm9 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
; SSE-NEXT:    movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm13, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm14[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm10, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT:    # xmm6 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm1[0],xmm6[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
; SSE-NEXT:    movapd %xmm15, %xmm7
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
; SSE-NEXT:    movapd %xmm15, %xmm11
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
; SSE-NEXT:    movapd %xmm14, %xmm12
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
; SSE-NEXT:    movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
; SSE-NEXT:    movapd %xmm15, %xmm10
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[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:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[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:    pshufd {{.*#+}} xmm1 = xmm5[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 96(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 112(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 64(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 80(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 96(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 112(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 64(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 80(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 96(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 112(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 64(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 80(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 112(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 96(%r8)
; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 80(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 64(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%r8)
; SSE-NEXT:    movapd %xmm2, 112(%r9)
; SSE-NEXT:    movapd %xmm3, 96(%r9)
; SSE-NEXT:    movapd %xmm4, 80(%r9)
; SSE-NEXT:    movapd %xmm6, 64(%r9)
; SSE-NEXT:    movapd %xmm8, 48(%r9)
; SSE-NEXT:    movapd %xmm9, 32(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%r9)
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movapd %xmm14, 112(%rax)
; SSE-NEXT:    movapd %xmm13, 96(%rax)
; SSE-NEXT:    movapd %xmm15, 80(%rax)
; SSE-NEXT:    movapd %xmm10, 64(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%rax)
; SSE-NEXT:    movapd %xmm12, 32(%rax)
; SSE-NEXT:    movapd %xmm11, 16(%rax)
; SSE-NEXT:    movapd %xmm7, (%rax)
; SSE-NEXT:    addq $1032, %rsp # imm = 0x408
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i32_stride6_vf32:
; AVX:       # %bb.0:
; AVX-NEXT:    subq $1032, %rsp # imm = 0x408
; AVX-NEXT:    vmovaps 416(%rdi), %ymm9
; AVX-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 480(%rdi), %ymm4
; AVX-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 448(%rdi), %ymm5
; AVX-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 160(%rdi), %ymm2
; AVX-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 128(%rdi), %ymm3
; AVX-NEXT:    vmovupd %ymm3, (%rsp) # 32-byte Spill
; AVX-NEXT:    vmovaps 32(%rdi), %ymm6
; AVX-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps (%rdi), %ymm7
; AVX-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 96(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 64(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 96(%rdi), %ymm0, %ymm8
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm8[0,0],ymm1[6,4],ymm8[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm8[2,2],ymm0[6,4],ymm8[6,6]
; AVX-NEXT:    vblendps {{.*#+}} ymm12 = ymm7[0,1,2,3],ymm6[4,5],ymm7[6,7]
; AVX-NEXT:    vextractf128 $1, %ymm12, %xmm7
; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm12[0,1],xmm7[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm7[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm3[2,3],ymm2[0,1]
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[3],ymm3[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 480(%rdi), %ymm5, %ymm6
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm4[2,0],ymm6[0,0],ymm4[6,4],ymm6[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm6[2,2],ymm0[6,4],ymm6[6,6]
; AVX-NEXT:    vmovaps 384(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm9[4,5],ymm1[6,7]
; AVX-NEXT:    vextractf128 $1, %ymm13, %xmm5
; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm5[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vmovapd 544(%rdi), %ymm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 512(%rdi), %ymm2
; AVX-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 288(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 256(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 288(%rdi), %ymm0, %ymm3
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm3[0,0],ymm1[6,4],ymm3[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm3[2,2],ymm0[6,4],ymm3[6,6]
; AVX-NEXT:    vmovaps 224(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 192(%rdi), %ymm2
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX-NEXT:    vextractf128 $1, %ymm11, %xmm2
; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm11[0,1],xmm2[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vmovapd 352(%rdi), %ymm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 320(%rdi), %ymm4
; AVX-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[0,1]
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[3],ymm4[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 672(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 640(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 672(%rdi), %ymm0, %ymm9
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm9[0,0],ymm1[6,4],ymm9[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm10 = ymm0[2,0],ymm9[2,2],ymm0[6,4],ymm9[6,6]
; AVX-NEXT:    vmovaps 608(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 576(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX-NEXT:    vextractf128 $1, %ymm4, %xmm1
; AVX-NEXT:    vblendps {{.*#+}} xmm15 = xmm4[0,1],xmm1[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[0,2],xmm1[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2],ymm10[3,4,5,6,7]
; AVX-NEXT:    vmovapd 736(%rdi), %ymm10
; AVX-NEXT:    vmovupd %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 704(%rdi), %ymm0
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm0[2,3],ymm10[0,1]
; AVX-NEXT:    vshufpd {{.*#+}} ymm14 = ymm10[0],ymm0[0],ymm10[3],ymm0[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6,7]
; AVX-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm14 = ymm15[3,0],ymm8[1,0],ymm15[7,4],ymm8[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm8 = ymm14[2,0],ymm8[2,3],ymm14[6,4],ymm8[6,7]
; AVX-NEXT:    vshufps {{.*#+}} xmm12 = xmm12[1,0],xmm7[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm7 = xmm12[0,2],xmm7[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3,4,5,6,7]
; AVX-NEXT:    vmovups (%rsp), %ymm12 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm8 = ymm0[3,1],ymm12[1,3],ymm0[7,5],ymm12[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
; AVX-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm7 = ymm8[3,0],ymm6[1,0],ymm8[7,4],ymm6[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm6 = ymm7[2,0],ymm6[2,3],ymm7[6,4],ymm6[6,7]
; AVX-NEXT:    vshufps {{.*#+}} xmm7 = xmm13[1,0],xmm5[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm5 = xmm7[0,2],xmm5[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm6 = ymm0[3,1],ymm14[1,3],ymm0[7,5],ymm14[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm6[6,7]
; AVX-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[3,0],ymm3[1,0],ymm0[7,4],ymm3[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm5[2,0],ymm3[2,3],ymm5[6,4],ymm3[6,7]
; AVX-NEXT:    vshufps {{.*#+}} xmm5 = xmm11[1,0],xmm2[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm5[0,2],xmm2[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
; AVX-NEXT:    # ymm3 = ymm0[3,1],mem[1,3],ymm0[7,5],mem[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm11[3,0],ymm9[1,0],ymm11[7,4],ymm9[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[2,0],ymm9[2,3],ymm2[6,4],ymm9[6,7]
; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm4[1,0],xmm1[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[0,2],xmm1[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[3,1],ymm13[1,3],ymm10[7,5],ymm13[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
; AVX-NEXT:    # ymm4 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm1 # 32-byte Folded Reload
; AVX-NEXT:    # ymm1 = ymm15[2,1],mem[2,0],ymm15[6,5],mem[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm4, %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm3 = xmm4[2,0],xmm0[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm3 # 32-byte Folded Reload
; AVX-NEXT:    # ymm3 = ymm12[0,1,2,3],mem[4,5],ymm12[6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm3[2,3,0,1]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[0,0],ymm3[2,0],ymm0[4,4],ymm3[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm5[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX-NEXT:    # ymm6 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm5 # 32-byte Folded Reload
; AVX-NEXT:    # ymm5 = ymm8[2,1],mem[2,0],ymm8[6,5],mem[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm6, %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm7 = xmm6[2,0],xmm0[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2],ymm5[3,4,5,6,7]
; AVX-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = ymm14[0,1,2,3],mem[4,5],ymm14[6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm2[2,3,0,1]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm7 = ymm0[0,0],ymm2[2,0],ymm0[4,4],ymm2[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4],ymm7[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
; AVX-NEXT:    # ymm9 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm7 # 32-byte Folded Reload
; AVX-NEXT:    # ymm7 = ymm11[2,1],mem[2,0],ymm11[6,5],mem[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm7[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm9, %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm11 = xmm9[2,0],xmm0[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm12 = ymm11[0,1,2],ymm10[3,4,5,6,7]
; AVX-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm10 # 32-byte Folded Reload
; AVX-NEXT:    # ymm10 = ymm13[0,1,2,3],mem[4,5],ymm13[6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm10[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm13 = ymm7[0,0],ymm10[2,0],ymm7[4,4],ymm10[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3,4],ymm13[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX-NEXT:    # ymm12 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm13 = ymm5[2,1],ymm8[2,0],ymm5[6,5],ymm8[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm13[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm12, %xmm14
; AVX-NEXT:    vshufps {{.*#+}} xmm15 = xmm12[2,0],xmm14[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[0,0],ymm15[2,0],ymm0[4,4],ymm15[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1],ymm3[3,1],ymm1[4,5],ymm3[7,5]
; AVX-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm3 # 16-byte Folded Reload
; AVX-NEXT:    # xmm3 = xmm4[3,1],mem[3,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm4 = ymm11[3,1],ymm13[2,1],ymm11[7,5],ymm13[6,5]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1],ymm2[3,1],ymm1[4,5],ymm2[7,5]
; AVX-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm2 # 16-byte Folded Reload
; AVX-NEXT:    # xmm2 = xmm6[3,1],mem[3,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[3,1],ymm6[2,1],ymm4[7,5],ymm6[6,5]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm15[3,1],ymm0[4,5],ymm15[7,5]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm12[3,1],xmm14[3,3]
; AVX-NEXT:    vmovaps %ymm5, %ymm3
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm5[3,1],ymm8[2,1],ymm5[7,5],ymm8[6,5]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm7[0,1],ymm10[3,1],ymm7[4,5],ymm10[7,5]
; AVX-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm1 # 16-byte Folded Reload
; AVX-NEXT:    # xmm1 = xmm9[3,1],mem[3,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[3,1],ymm12[2,1],ymm2[7,5],ymm12[6,5]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovaps 416(%rdi), %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX-NEXT:    vmovaps 400(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX-NEXT:    vmovapd 464(%rdi), %xmm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm6[0],ymm1[2],ymm6[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,0],ymm4[4,5],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm2[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm7[0,0],ymm2[6,4],ymm7[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[0,2],ymm1[2,0],ymm7[4,6],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX-NEXT:    # ymm6 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovaps 32(%rdi), %xmm0
; AVX-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX-NEXT:    vmovaps 16(%rdi), %xmm15
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm15[0],xmm0[1],xmm15[2,3]
; AVX-NEXT:    vmovapd 80(%rdi), %xmm10
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm10[1],ymm13[0],ymm10[2],ymm13[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,0],ymm11[4,5],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm6[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm6[2,0],ymm1[0,0],ymm6[6,4],ymm1[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[0,2],ymm2[2,0],ymm1[4,6],ymm2[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX-NEXT:    # ymm5 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovaps 224(%rdi), %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX-NEXT:    vmovaps 208(%rdi), %xmm13
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm13[0],xmm0[1],xmm13[2,3]
; AVX-NEXT:    vmovapd 272(%rdi), %xmm2
; AVX-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[1],ymm8[0],ymm2[2],ymm8[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,0],ymm3[4,5],ymm2[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm5[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm5[2,0],ymm2[0,0],ymm5[6,4],ymm2[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[0,2],ymm3[2,0],ymm2[4,6],ymm3[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
; AVX-NEXT:    # ymm3 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovaps 608(%rdi), %xmm11
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm11[2,2,3,3]
; AVX-NEXT:    vmovaps 592(%rdi), %xmm8
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm8[0],xmm0[1],xmm8[2,3]
; AVX-NEXT:    vmovapd 656(%rdi), %xmm9
; AVX-NEXT:    vshufpd {{.*#+}} ymm14 = ymm9[1],ymm12[0],ymm9[2],ymm12[2]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm14 = ymm4[0,1],ymm14[2,0],ymm4[4,5],ymm14[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1],ymm14[2,3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm3[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm12 = ymm3[2,0],ymm0[0,0],ymm3[6,4],ymm0[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm12 = ymm0[0,2],ymm12[2,0],ymm0[4,6],ymm12[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm12[5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[3,0],ymm1[1,0],ymm6[7,4],ymm1[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,3],ymm6[2,0],ymm1[4,7],ymm6[6,4]
; AVX-NEXT:    vblendps $12, (%rsp), %xmm15, %xmm6 # 16-byte Folded Reload
; AVX-NEXT:    # xmm6 = xmm15[0,1],mem[2,3]
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX-NEXT:    # ymm10 = ymm10[3,1],mem[1,3],ymm10[7,5],mem[5,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm10 = ymm12[1,1],ymm10[2,0],ymm12[5,5],ymm10[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm10[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm7[1,0],ymm1[7,4],ymm7[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[0,3],ymm1[2,0],ymm7[4,7],ymm1[6,4]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
; AVX-NEXT:    # xmm7 = xmm7[0,1],mem[2,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX-NEXT:    # ymm10 = ymm10[3,1],mem[1,3],ymm10[7,5],mem[5,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm10 = ymm12[1,1],ymm10[2,0],ymm12[5,5],ymm10[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm7 = xmm7[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm10[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[3,0],ymm2[1,0],ymm5[7,4],ymm2[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,3],ymm1[2,0],ymm2[4,7],ymm1[6,4]
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm2 # 16-byte Folded Reload
; AVX-NEXT:    # xmm2 = xmm13[0,1],mem[2,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX-NEXT:    # ymm5 = ymm5[3,1],mem[1,3],ymm5[7,5],mem[5,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm10[1,1],ymm5[2,0],ymm10[5,5],ymm5[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm5[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,0],ymm0[1,0],ymm3[7,4],ymm0[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm2[2,0],ymm0[4,7],ymm2[6,4]
; AVX-NEXT:    vblendps {{.*#+}} xmm2 = xmm8[0,1],xmm11[2,3]
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
; AVX-NEXT:    # ymm3 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[1,1],ymm3[2,0],ymm4[5,5],ymm3[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 96(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 32(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 64(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, (%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 96(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 32(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 64(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, (%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 32(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 96(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 64(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, (%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 96(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 32(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 64(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, (%r8)
; AVX-NEXT:    vmovaps %ymm14, 96(%r9)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 32(%r9)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, (%r9)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm2, 64(%r9)
; AVX-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX-NEXT:    vmovaps %ymm0, 96(%rax)
; AVX-NEXT:    vmovaps %ymm1, 32(%rax)
; AVX-NEXT:    vmovaps %ymm7, 64(%rax)
; AVX-NEXT:    vmovaps %ymm6, (%rax)
; AVX-NEXT:    addq $1032, %rsp # imm = 0x408
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i32_stride6_vf32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    subq $1224, %rsp # imm = 0x4C8
; AVX2-NEXT:    vmovaps 480(%rdi), %ymm9
; AVX2-NEXT:    vmovaps 448(%rdi), %ymm11
; AVX2-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 416(%rdi), %ymm8
; AVX2-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 128(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
; AVX2-NEXT:    vmovaps 160(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 96(%rdi), %ymm4
; AVX2-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps (%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 64(%rdi), %ymm5
; AVX2-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps {{.*#+}} xmm10 = [0,6,4,u]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm0, %ymm10, %ymm0
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[0,1],ymm4[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3,4,5],ymm4[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [4,2,4,2,4,2,4,2]
; AVX2-NEXT:    vpermps %ymm7, %ymm2, %ymm1
; AVX2-NEXT:    vmovaps %ymm2, %ymm6
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm11[0,1],ymm9[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3,4,5],ymm9[6,7]
; AVX2-NEXT:    vmovaps 384(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
; AVX2-NEXT:    vpermps %ymm3, %ymm10, %ymm0
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-NEXT:    vmovaps 512(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 544(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm2, %ymm6, %ymm1
; AVX2-NEXT:    vmovaps %ymm6, %ymm9
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 288(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 256(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovaps 224(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 192(%rdi), %ymm6
; AVX2-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm0[4,5],ymm6[6,7]
; AVX2-NEXT:    vpermps %ymm1, %ymm10, %ymm8
; AVX2-NEXT:    vshufps {{.*#+}} ymm11 = ymm13[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm11 = ymm8[0,1,2],ymm11[3,4,5,6,7]
; AVX2-NEXT:    vmovaps 320(%rdi), %ymm6
; AVX2-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 352(%rdi), %ymm8
; AVX2-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm6[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm8, %ymm9, %ymm14
; AVX2-NEXT:    vmovaps %ymm9, %ymm0
; AVX2-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5],ymm14[6,7]
; AVX2-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 608(%rdi), %ymm6
; AVX2-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 576(%rdi), %ymm9
; AVX2-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm9[0,1,2,3],ymm6[4,5],ymm9[6,7]
; AVX2-NEXT:    vpermps %ymm14, %ymm10, %ymm10
; AVX2-NEXT:    vmovaps 672(%rdi), %ymm6
; AVX2-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 640(%rdi), %ymm9
; AVX2-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm9[0,1],ymm6[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm6[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm15 = ymm12[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm15[3,4,5,6,7]
; AVX2-NEXT:    vmovaps 704(%rdi), %ymm6
; AVX2-NEXT:    vmovaps 736(%rdi), %ymm11
; AVX2-NEXT:    vblendps {{.*#+}} ymm15 = ymm11[0,1,2,3],ymm6[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm15, %ymm0, %ymm9
; AVX2-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4,5],ymm9[6,7]
; AVX2-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps {{.*#+}} xmm9 = [1,7,5,u]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm10 # 32-byte Folded Reload
; AVX2-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm10[0,1,2],ymm5[3,4,5,6,7]
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm10 = [5,3,5,3,5,3,5,3]
; AVX2-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm7, %ymm10, %ymm7
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm7[6,7]
; AVX2-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm3, %ymm9, %ymm3
; AVX2-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm2, %ymm10, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm1, %ymm9, %ymm0
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm8, %ymm10, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm14, %ymm9, %ymm0
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm15, %ymm10, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = mem[0,1],ymm14[2,3],mem[4,5],ymm14[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vmovups (%rsp), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm4 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm15 = ymm1[0,1],mem[2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm12[2,3],ymm13[4,5],ymm12[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm7 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm1[3,4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm11[4,5],ymm6[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm8 = ymm1[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3,4],ymm8[5,6,7]
; AVX2-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm6[0,1],ymm11[2,3],ymm6[4,5],ymm11[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm8 = mem[0,1],ymm8[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm9 = ymm8[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1,2],ymm7[3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm9 = mem[0,1,2,3],ymm9[4,5],mem[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm10 = ymm9[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3,4],ymm10[5,6,7]
; AVX2-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm7 = ymm14[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm10[1],ymm7[2,3,4],ymm10[5],ymm7[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm7[3,4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm5[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm7[1],ymm3[2,3,4],ymm7[5],ymm3[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
; AVX2-NEXT:    vshufps {{.*#+}} ymm4 = ymm15[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm11[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm6[1],ymm2[2,3,4],ymm6[5],ymm2[6,7]
; AVX2-NEXT:    vmovaps %ymm6, %ymm4
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm8[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm9[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm13[1],ymm2[2,3,4],ymm13[5],ymm2[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm5[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 464(%rdi), %xmm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vpermps %ymm3, %ymm2, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm7 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = [0,2,0,6,0,2,0,6]
; AVX2-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX2-NEXT:    vpermps %ymm7, %ymm3, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm14[4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vmovaps 80(%rdi), %xmm5
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vpermps %ymm14, %ymm2, %ymm1
; AVX2-NEXT:    vmovaps %ymm2, %ymm8
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-NEXT:    vmovups (%rsp), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm2, %ymm3, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm9 = ymm4[0,1,2,3],ymm11[4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm12 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vmovaps 272(%rdi), %xmm4
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1],ymm4[2,3],ymm9[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vpermps %ymm12, %ymm8, %ymm11
; AVX2-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm0[2,3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm1, %ymm3, %ymm15
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4],ymm15[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm11 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm11 = ymm13[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm15 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vmovaps 656(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm0[2,3],ymm11[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vpermps %ymm15, %ymm8, %ymm8
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm10[2,3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm10 = ymm10[0,1],mem[2,3],ymm10[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm10, %ymm3, %ymm3
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3,4],ymm3[5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vpermps %ymm14, %ymm13, %ymm6
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3,4,5,6,7]
; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm6 = [0,3,1,7,0,3,1,7]
; AVX2-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX2-NEXT:    vpermps %ymm2, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm5 = mem[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm5 = ymm5[0,1,2],mem[3],ymm5[4,5,6,7]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm8 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm7, %ymm6, %ymm7
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm7[5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm7 = ymm9[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3],ymm7[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm12, %ymm13, %ymm7
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1],ymm4[2,3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm1, %ymm6, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm4 = ymm11[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3],ymm4[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm15, %ymm13, %ymm4
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm10, %ymm6, %ymm4
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 96(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 32(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 64(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, (%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 96(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 32(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 64(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, (%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 32(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 96(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 64(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, (%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 96(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 32(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, 64(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm4, (%r8)
; AVX2-NEXT:    vmovaps %ymm3, 96(%r9)
; AVX2-NEXT:    vmovups (%rsp), %ymm3 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm3, 32(%r9)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm3, (%r9)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm3, 64(%r9)
; AVX2-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-NEXT:    vmovaps %ymm0, 96(%rax)
; AVX2-NEXT:    vmovaps %ymm1, 32(%rax)
; AVX2-NEXT:    vmovaps %ymm5, 64(%rax)
; AVX2-NEXT:    vmovaps %ymm2, (%rax)
; AVX2-NEXT:    addq $1224, %rsp # imm = 0x4C8
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i32_stride6_vf32:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    subq $1224, %rsp # imm = 0x4C8
; AVX2-FP-NEXT:    vmovaps 480(%rdi), %ymm9
; AVX2-FP-NEXT:    vmovaps 448(%rdi), %ymm11
; AVX2-FP-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 416(%rdi), %ymm8
; AVX2-FP-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 160(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 96(%rdi), %ymm4
; AVX2-FP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 64(%rdi), %ymm5
; AVX2-FP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps {{.*#+}} xmm10 = [0,6,4,u]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm0, %ymm10, %ymm0
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[0,1],ymm4[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [4,2,4,2,4,2,4,2]
; AVX2-FP-NEXT:    vpermps %ymm7, %ymm2, %ymm1
; AVX2-FP-NEXT:    vmovaps %ymm2, %ymm6
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm11[0,1],ymm9[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FP-NEXT:    vmovaps 384(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vpermps %ymm3, %ymm10, %ymm0
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 512(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 544(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm6, %ymm1
; AVX2-FP-NEXT:    vmovaps %ymm6, %ymm9
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 288(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 256(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovaps 224(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 192(%rdi), %ymm6
; AVX2-FP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm0[4,5],ymm6[6,7]
; AVX2-FP-NEXT:    vpermps %ymm1, %ymm10, %ymm8
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm11 = ymm13[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm11 = ymm8[0,1,2],ymm11[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 320(%rdi), %ymm6
; AVX2-FP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 352(%rdi), %ymm8
; AVX2-FP-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm8, %ymm9, %ymm14
; AVX2-FP-NEXT:    vmovaps %ymm9, %ymm0
; AVX2-FP-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5],ymm14[6,7]
; AVX2-FP-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 608(%rdi), %ymm6
; AVX2-FP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 576(%rdi), %ymm9
; AVX2-FP-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm9[0,1,2,3],ymm6[4,5],ymm9[6,7]
; AVX2-FP-NEXT:    vpermps %ymm14, %ymm10, %ymm10
; AVX2-FP-NEXT:    vmovaps 672(%rdi), %ymm6
; AVX2-FP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 640(%rdi), %ymm9
; AVX2-FP-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm9[0,1],ymm6[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm6[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm15 = ymm12[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 704(%rdi), %ymm6
; AVX2-FP-NEXT:    vmovaps 736(%rdi), %ymm11
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm15 = ymm11[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm15, %ymm0, %ymm9
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FP-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps {{.*#+}} xmm9 = [1,7,5,u]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm10 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm10[0,1,2],ymm5[3,4,5,6,7]
; AVX2-FP-NEXT:    vbroadcastsd {{.*#+}} ymm10 = [5,3,5,3,5,3,5,3]
; AVX2-FP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm7, %ymm10, %ymm7
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm3, %ymm9, %ymm3
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm10, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm1, %ymm9, %ymm0
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm13[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm8, %ymm10, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm14, %ymm9, %ymm0
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm15, %ymm10, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = mem[0,1],ymm14[2,3],mem[4,5],ymm14[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups (%rsp), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm4 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm15 = ymm1[0,1],mem[2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm12[2,3],ymm13[4,5],ymm12[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm7 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm11[4,5],ymm6[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm8 = ymm1[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3,4],ymm8[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm6[0,1],ymm11[2,3],ymm6[4,5],ymm11[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm8 = mem[0,1],ymm8[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm9 = ymm8[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm9 = mem[0,1,2,3],ymm9[4,5],mem[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm10 = ymm9[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3,4],ymm10[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm7 = ymm14[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm10[1],ymm7[2,3,4],ymm10[5],ymm7[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm3 = ymm5[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0],ymm7[1],ymm3[2,3,4],ymm7[5],ymm3[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm4 = ymm15[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm11[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm6[1],ymm2[2,3,4],ymm6[5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovaps %ymm6, %ymm4
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm3 = ymm8[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm3 = ymm9[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm13[1],ymm2[2,3,4],ymm13[5],ymm2[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 464(%rdi), %xmm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vpermps %ymm3, %ymm2, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm7 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = [0,2,0,6,0,2,0,6]
; AVX2-FP-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpermps %ymm7, %ymm3, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm14[4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 80(%rdi), %xmm5
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vpermps %ymm14, %ymm2, %ymm1
; AVX2-FP-NEXT:    vmovaps %ymm2, %ymm8
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups (%rsp), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm3, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm9 = ymm4[0,1,2,3],ymm11[4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm12 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 272(%rdi), %xmm4
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1],ymm4[2,3],ymm9[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vpermps %ymm12, %ymm8, %ymm11
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm1, %ymm3, %ymm15
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4],ymm15[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm11 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm11 = ymm13[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm15 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 656(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm0[2,3],ymm11[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vpermps %ymm15, %ymm8, %ymm8
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm10[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm10 = ymm10[0,1],mem[2,3],ymm10[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm10, %ymm3, %ymm3
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vpermps %ymm14, %ymm13, %ymm6
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vbroadcastf128 {{.*#+}} ymm6 = [0,3,1,7,0,3,1,7]
; AVX2-FP-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm5 = mem[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm5 = ymm5[0,1,2],mem[3],ymm5[4,5,6,7]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm8 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm7, %ymm6, %ymm7
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm7[5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm7 = ymm9[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3],ymm7[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm12, %ymm13, %ymm7
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1],ymm4[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm1, %ymm6, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm4 = ymm11[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3],ymm4[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm15, %ymm13, %ymm4
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm10, %ymm6, %ymm4
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 96(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 32(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 64(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, (%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 96(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 32(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 64(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, (%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 32(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 96(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 64(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, (%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 96(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 32(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, 64(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm4, (%r8)
; AVX2-FP-NEXT:    vmovaps %ymm3, 96(%r9)
; AVX2-FP-NEXT:    vmovups (%rsp), %ymm3 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm3, 32(%r9)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm3, (%r9)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm3, 64(%r9)
; AVX2-FP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FP-NEXT:    vmovaps %ymm0, 96(%rax)
; AVX2-FP-NEXT:    vmovaps %ymm1, 32(%rax)
; AVX2-FP-NEXT:    vmovaps %ymm5, 64(%rax)
; AVX2-FP-NEXT:    vmovaps %ymm2, (%rax)
; AVX2-FP-NEXT:    addq $1224, %rsp # imm = 0x4C8
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i32_stride6_vf32:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    subq $1192, %rsp # imm = 0x4A8
; AVX2-FCP-NEXT:    vmovaps 480(%rdi), %ymm6
; AVX2-FCP-NEXT:    vmovaps 448(%rdi), %ymm11
; AVX2-FCP-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 416(%rdi), %ymm10
; AVX2-FCP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 160(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 96(%rdi), %ymm4
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 64(%rdi), %ymm5
; AVX2-FCP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm9 = [0,6,4,u]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm9, %ymm0
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[0,1],ymm4[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm1 = ymm8[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [4,2,4,2,4,2,4,2]
; AVX2-FCP-NEXT:    vpermps %ymm7, %ymm2, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm11[0,1],ymm6[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm6[6,7]
; AVX2-FCP-NEXT:    vmovaps 384(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3],ymm10[4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm4, %ymm9, %ymm0
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm1 = ymm6[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 512(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 544(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm3, %ymm2, %ymm1
; AVX2-FCP-NEXT:    vmovaps %ymm2, %ymm5
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 288(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 256(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovaps 224(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 192(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm1, %ymm9, %ymm0
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm14 = ymm2[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 320(%rdi), %ymm10
; AVX2-FCP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 352(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm13, %ymm5, %ymm15
; AVX2-FCP-NEXT:    vmovaps %ymm5, %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm10 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 608(%rdi), %ymm10
; AVX2-FCP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 576(%rdi), %ymm11
; AVX2-FCP-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm11[0,1,2,3],ymm10[4,5],ymm11[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm14, %ymm9, %ymm15
; AVX2-FCP-NEXT:    vmovaps 672(%rdi), %ymm5
; AVX2-FCP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 640(%rdi), %ymm9
; AVX2-FCP-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm9[0,1],ymm5[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm11 = ymm12[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm11 = ymm15[0,1,2],ymm11[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 704(%rdi), %ymm9
; AVX2-FCP-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 736(%rdi), %ymm10
; AVX2-FCP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm15 = ymm10[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm15, %ymm0, %ymm10
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm10 = [1,7,5,u]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm11 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm11[0,1,2],ymm8[3,4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastsd {{.*#+}} ymm5 = [5,3,5,3,5,3,5,3]
; AVX2-FCP-NEXT:    vpermps %ymm7, %ymm5, %ymm7
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm4, %ymm10, %ymm4
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm3, %ymm5, %ymm3
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm1, %ymm10, %ymm1
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm13, %ymm5, %ymm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm14, %ymm10, %ymm0
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm15, %ymm5, %ymm1
; AVX2-FCP-NEXT:    vmovaps %ymm5, %ymm13
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} ymm7 = [2,0,6,4,2,0,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm7, %ymm0
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm8 = [2,0,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, (%rsp), %ymm1, %ymm3 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm3 = ymm1[0,1],mem[2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm3, %ymm8, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm10 = [0,0,6,4,0,0,6,4]
; AVX2-FCP-NEXT:    # ymm10 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm10, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm4[0,1],mem[2,3],ymm4[4,5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm7, %ymm0
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm14 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm14, %ymm8, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm6 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm6, %ymm10, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm5[0,1],mem[2,3],ymm5[4,5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm7, %ymm1
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm8, %ymm11
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm1, %ymm10, %ymm12
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3,4],ymm12[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm11 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm11 = ymm9[0,1],mem[2,3],ymm9[4,5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm11, %ymm7, %ymm7
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm11 = ymm11[0,1],mem[2,3],ymm11[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm11, %ymm8, %ymm8
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm8 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm8, %ymm10, %ymm10
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm10[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm7 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm15[1],ymm7[2,3,4],ymm15[5],ymm7[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm7[0],ymm3[1,2,3,4],ymm7[5],ymm3[6,7]
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm7 = [0,1,7,5,0,1,7,5]
; AVX2-FCP-NEXT:    # ymm7 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm7, %ymm2
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} ymm10 = [3,1,7,5,0,u,u,u]
; AVX2-FCP-NEXT:    vpermps %ymm3, %ymm10, %ymm3
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm2 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2,3,4],ymm4[5],ymm2[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm14[1,2,3,4],ymm2[5],ymm14[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm10, %ymm2
; AVX2-FCP-NEXT:    vpermps %ymm6, %ymm7, %ymm3
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm2 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm9[1],ymm2[2,3,4],ymm9[5],ymm2[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm11[1,2,3,4],ymm2[5],ymm11[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm10, %ymm2
; AVX2-FCP-NEXT:    vpermps %ymm8, %ymm7, %ymm3
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm2 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4],ymm2[5],ymm0[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm10, %ymm0
; AVX2-FCP-NEXT:    vpermps %ymm1, %ymm7, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm1 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm1 = ymm4[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm3 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 464(%rdi), %xmm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vpermps %ymm3, %ymm2, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm6 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm4 = [0,2,0,6,0,2,0,6]
; AVX2-FCP-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm6, %ymm4, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm3 = ymm15[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, (%rsp), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm12 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 80(%rdi), %xmm14
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm14[2,3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vpermps %ymm12, %ymm2, %ymm1
; AVX2-FCP-NEXT:    vmovaps %ymm2, %ymm5
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm4, %ymm8
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm8[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm10 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm10 = ymm9[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm11 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 272(%rdi), %xmm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1],ymm1[2,3],ymm10[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vpermps %ymm11, %ymm5, %ymm9
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm8[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm8 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm8, %ymm4, %ymm15
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3,4],ymm15[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm9 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm15 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 656(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm0[2,3],ymm9[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vpermps %ymm15, %ymm5, %ymm5
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm7 = mem[0,1],ymm7[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm7, %ymm4, %ymm4
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm14[3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm12, %ymm13, %ymm5
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm5 = [0,3,1,7,0,3,1,7]
; AVX2-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm5, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm3 = mem[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm12 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm12[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm6, %ymm5, %ymm6
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm6 = ymm10[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2],ymm1[3],ymm6[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm11, %ymm13, %ymm6
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm8, %ymm5, %ymm6
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm6 = ymm9[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2],ymm0[3],ymm6[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm15, %ymm13, %ymm6
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm7, %ymm5, %ymm5
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, 96(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, 32(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, 64(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, (%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, 96(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, 32(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, 64(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, (%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, 32(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, 96(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, 64(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, (%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, 96(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, 32(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, 64(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, (%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm4, 96(%r9)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm4, 32(%r9)
; AVX2-FCP-NEXT:    vmovups (%rsp), %ymm4 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm4, (%r9)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm4, 64(%r9)
; AVX2-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FCP-NEXT:    vmovaps %ymm0, 96(%rax)
; AVX2-FCP-NEXT:    vmovaps %ymm1, 32(%rax)
; AVX2-FCP-NEXT:    vmovaps %ymm3, 64(%rax)
; AVX2-FCP-NEXT:    vmovaps %ymm2, (%rax)
; AVX2-FCP-NEXT:    addq $1192, %rsp # imm = 0x4A8
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i32_stride6_vf32:
; AVX512:       # %bb.0:
; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-NEXT:    vmovdqa64 448(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 384(%rdi), %zmm3
; AVX512-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512-NEXT:    vmovdqa64 704(%rdi), %zmm6
; AVX512-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512-NEXT:    vmovdqa64 320(%rdi), %zmm10
; AVX512-NEXT:    vmovdqa64 256(%rdi), %zmm9
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm11
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 128(%rdi), %zmm12
; AVX512-NEXT:    vmovdqa64 192(%rdi), %zmm13
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm13, %zmm15
; AVX512-NEXT:    vpermt2d %zmm12, %zmm14, %zmm15
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,6,12,18,24,30,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm8
; AVX512-NEXT:    vpermt2d %zmm1, %zmm7, %zmm8
; AVX512-NEXT:    movb $56, %dil
; AVX512-NEXT:    kmovw %edi, %k2
; AVX512-NEXT:    vmovdqa64 %zmm15, %zmm8 {%k2}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm9, %zmm16
; AVX512-NEXT:    vpermt2d %zmm10, %zmm15, %zmm16
; AVX512-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512-NEXT:    kmovw %edi, %k1
; AVX512-NEXT:    vmovdqa32 %zmm16, %zmm8 {%k1}
; AVX512-NEXT:    vpermi2d %zmm6, %zmm4, %zmm15
; AVX512-NEXT:    vpermi2d %zmm2, %zmm5, %zmm14
; AVX512-NEXT:    vpermi2d %zmm0, %zmm3, %zmm7
; AVX512-NEXT:    vmovdqa64 %zmm14, %zmm7 {%k2}
; AVX512-NEXT:    vmovdqa32 %zmm15, %zmm7 {%k1}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm16 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512-NEXT:    # zmm16 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm13, %zmm17
; AVX512-NEXT:    vpermt2d %zmm12, %zmm16, %zmm17
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [1,7,13,19,25,31,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm15
; AVX512-NEXT:    vpermt2d %zmm1, %zmm14, %zmm15
; AVX512-NEXT:    vmovdqa64 %zmm17, %zmm15 {%k2}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm9, %zmm18
; AVX512-NEXT:    vpermt2d %zmm10, %zmm17, %zmm18
; AVX512-NEXT:    vmovdqa32 %zmm18, %zmm15 {%k1}
; AVX512-NEXT:    vpermi2d %zmm6, %zmm4, %zmm17
; AVX512-NEXT:    vpermi2d %zmm2, %zmm5, %zmm16
; AVX512-NEXT:    vpermi2d %zmm0, %zmm3, %zmm14
; AVX512-NEXT:    vmovdqa64 %zmm16, %zmm14 {%k2}
; AVX512-NEXT:    vmovdqa32 %zmm17, %zmm14 {%k1}
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm18 = [2,8,14,20,26,0,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512-NEXT:    vpermt2d %zmm1, %zmm18, %zmm19
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm12, %zmm16
; AVX512-NEXT:    vpermt2d %zmm13, %zmm17, %zmm16
; AVX512-NEXT:    movw $31, %di
; AVX512-NEXT:    kmovw %edi, %k2
; AVX512-NEXT:    vmovdqa32 %zmm19, %zmm16 {%k2}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm10, %zmm20
; AVX512-NEXT:    vpermt2d %zmm9, %zmm19, %zmm20
; AVX512-NEXT:    vmovdqa32 %zmm20, %zmm16 {%k1}
; AVX512-NEXT:    vpermi2d %zmm4, %zmm6, %zmm19
; AVX512-NEXT:    vpermi2d %zmm5, %zmm2, %zmm17
; AVX512-NEXT:    vpermi2d %zmm0, %zmm3, %zmm18
; AVX512-NEXT:    vmovdqa32 %zmm18, %zmm17 {%k2}
; AVX512-NEXT:    vmovdqa32 %zmm19, %zmm17 {%k1}
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm20 = [3,9,15,21,27,0,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512-NEXT:    vpermt2d %zmm1, %zmm20, %zmm21
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm12, %zmm18
; AVX512-NEXT:    vpermt2d %zmm13, %zmm19, %zmm18
; AVX512-NEXT:    vmovdqa32 %zmm21, %zmm18 {%k2}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm10, %zmm22
; AVX512-NEXT:    vpermt2d %zmm9, %zmm21, %zmm22
; AVX512-NEXT:    vmovdqa32 %zmm22, %zmm18 {%k1}
; AVX512-NEXT:    vpermi2d %zmm4, %zmm6, %zmm21
; AVX512-NEXT:    vpermi2d %zmm5, %zmm2, %zmm19
; AVX512-NEXT:    vpermi2d %zmm0, %zmm3, %zmm20
; AVX512-NEXT:    vmovdqa32 %zmm20, %zmm19 {%k2}
; AVX512-NEXT:    vmovdqa32 %zmm21, %zmm19 {%k1}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm12, %zmm21
; AVX512-NEXT:    vpermt2d %zmm13, %zmm20, %zmm21
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm22 = [20,26,0,6,12,0,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm1, %zmm23
; AVX512-NEXT:    vpermt2d %zmm11, %zmm22, %zmm23
; AVX512-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512-NEXT:    kmovw %edi, %k1
; AVX512-NEXT:    vmovdqa32 %zmm21, %zmm23 {%k1}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm9, %zmm24
; AVX512-NEXT:    vpermt2d %zmm10, %zmm21, %zmm24
; AVX512-NEXT:    movb $-32, %dil
; AVX512-NEXT:    kmovw %edi, %k2
; AVX512-NEXT:    vmovdqa64 %zmm24, %zmm23 {%k2}
; AVX512-NEXT:    vpermi2d %zmm6, %zmm4, %zmm21
; AVX512-NEXT:    vpermi2d %zmm5, %zmm2, %zmm20
; AVX512-NEXT:    vpermi2d %zmm3, %zmm0, %zmm22
; AVX512-NEXT:    vmovdqa32 %zmm20, %zmm22 {%k1}
; AVX512-NEXT:    vmovdqa64 %zmm21, %zmm22 {%k2}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermt2d %zmm13, %zmm20, %zmm12
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [21,27,1,7,13,0,0,0]
; AVX512-NEXT:    vpermt2d %zmm11, %zmm13, %zmm1
; AVX512-NEXT:    vmovdqa32 %zmm12, %zmm1 {%k1}
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm11 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermt2d %zmm10, %zmm11, %zmm9
; AVX512-NEXT:    vmovdqa64 %zmm9, %zmm1 {%k2}
; AVX512-NEXT:    vpermt2d %zmm6, %zmm11, %zmm4
; AVX512-NEXT:    vpermt2d %zmm5, %zmm20, %zmm2
; AVX512-NEXT:    vpermt2d %zmm3, %zmm13, %zmm0
; AVX512-NEXT:    vmovdqa32 %zmm2, %zmm0 {%k1}
; AVX512-NEXT:    vmovdqa64 %zmm4, %zmm0 {%k2}
; AVX512-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm14, 64(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm15, (%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm17, 64(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm16, (%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm19, 64(%r8)
; AVX512-NEXT:    vmovdqa64 %zmm18, (%r8)
; AVX512-NEXT:    vmovdqa64 %zmm22, 64(%r9)
; AVX512-NEXT:    vmovdqa64 %zmm23, (%r9)
; AVX512-NEXT:    vmovdqa64 %zmm0, 64(%rax)
; AVX512-NEXT:    vmovdqa64 %zmm1, (%rax)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i32_stride6_vf32:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm6
; AVX512-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm10
; AVX512-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm9
; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %zmm11
; AVX512-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm12
; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm13
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512-FCP-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm13, %zmm15
; AVX512-FCP-NEXT:    vpermt2d %zmm12, %zmm14, %zmm15
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,6,12,18,24,30,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, %zmm8
; AVX512-FCP-NEXT:    vpermt2d %zmm1, %zmm7, %zmm8
; AVX512-FCP-NEXT:    movb $56, %dil
; AVX512-FCP-NEXT:    kmovw %edi, %k2
; AVX512-FCP-NEXT:    vmovdqa64 %zmm15, %zmm8 {%k2}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512-FCP-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm9, %zmm16
; AVX512-FCP-NEXT:    vpermt2d %zmm10, %zmm15, %zmm16
; AVX512-FCP-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512-FCP-NEXT:    kmovw %edi, %k1
; AVX512-FCP-NEXT:    vmovdqa32 %zmm16, %zmm8 {%k1}
; AVX512-FCP-NEXT:    vpermi2d %zmm6, %zmm4, %zmm15
; AVX512-FCP-NEXT:    vpermi2d %zmm2, %zmm5, %zmm14
; AVX512-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm7
; AVX512-FCP-NEXT:    vmovdqa64 %zmm14, %zmm7 {%k2}
; AVX512-FCP-NEXT:    vmovdqa32 %zmm15, %zmm7 {%k1}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm16 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512-FCP-NEXT:    # zmm16 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm13, %zmm17
; AVX512-FCP-NEXT:    vpermt2d %zmm12, %zmm16, %zmm17
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [1,7,13,19,25,31,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, %zmm15
; AVX512-FCP-NEXT:    vpermt2d %zmm1, %zmm14, %zmm15
; AVX512-FCP-NEXT:    vmovdqa64 %zmm17, %zmm15 {%k2}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512-FCP-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm9, %zmm18
; AVX512-FCP-NEXT:    vpermt2d %zmm10, %zmm17, %zmm18
; AVX512-FCP-NEXT:    vmovdqa32 %zmm18, %zmm15 {%k1}
; AVX512-FCP-NEXT:    vpermi2d %zmm6, %zmm4, %zmm17
; AVX512-FCP-NEXT:    vpermi2d %zmm2, %zmm5, %zmm16
; AVX512-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm14
; AVX512-FCP-NEXT:    vmovdqa64 %zmm16, %zmm14 {%k2}
; AVX512-FCP-NEXT:    vmovdqa32 %zmm17, %zmm14 {%k1}
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm18 = [2,8,14,20,26,0,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512-FCP-NEXT:    vpermt2d %zmm1, %zmm18, %zmm19
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512-FCP-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm12, %zmm16
; AVX512-FCP-NEXT:    vpermt2d %zmm13, %zmm17, %zmm16
; AVX512-FCP-NEXT:    movw $31, %di
; AVX512-FCP-NEXT:    kmovw %edi, %k2
; AVX512-FCP-NEXT:    vmovdqa32 %zmm19, %zmm16 {%k2}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512-FCP-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm10, %zmm20
; AVX512-FCP-NEXT:    vpermt2d %zmm9, %zmm19, %zmm20
; AVX512-FCP-NEXT:    vmovdqa32 %zmm20, %zmm16 {%k1}
; AVX512-FCP-NEXT:    vpermi2d %zmm4, %zmm6, %zmm19
; AVX512-FCP-NEXT:    vpermi2d %zmm5, %zmm2, %zmm17
; AVX512-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm18
; AVX512-FCP-NEXT:    vmovdqa32 %zmm18, %zmm17 {%k2}
; AVX512-FCP-NEXT:    vmovdqa32 %zmm19, %zmm17 {%k1}
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm20 = [3,9,15,21,27,0,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512-FCP-NEXT:    vpermt2d %zmm1, %zmm20, %zmm21
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512-FCP-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm12, %zmm18
; AVX512-FCP-NEXT:    vpermt2d %zmm13, %zmm19, %zmm18
; AVX512-FCP-NEXT:    vmovdqa32 %zmm21, %zmm18 {%k2}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512-FCP-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm10, %zmm22
; AVX512-FCP-NEXT:    vpermt2d %zmm9, %zmm21, %zmm22
; AVX512-FCP-NEXT:    vmovdqa32 %zmm22, %zmm18 {%k1}
; AVX512-FCP-NEXT:    vpermi2d %zmm4, %zmm6, %zmm21
; AVX512-FCP-NEXT:    vpermi2d %zmm5, %zmm2, %zmm19
; AVX512-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm20
; AVX512-FCP-NEXT:    vmovdqa32 %zmm20, %zmm19 {%k2}
; AVX512-FCP-NEXT:    vmovdqa32 %zmm21, %zmm19 {%k1}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512-FCP-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm12, %zmm21
; AVX512-FCP-NEXT:    vpermt2d %zmm13, %zmm20, %zmm21
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm22 = [20,26,0,6,12,0,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, %zmm23
; AVX512-FCP-NEXT:    vpermt2d %zmm11, %zmm22, %zmm23
; AVX512-FCP-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512-FCP-NEXT:    kmovw %edi, %k1
; AVX512-FCP-NEXT:    vmovdqa32 %zmm21, %zmm23 {%k1}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512-FCP-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm9, %zmm24
; AVX512-FCP-NEXT:    vpermt2d %zmm10, %zmm21, %zmm24
; AVX512-FCP-NEXT:    movb $-32, %dil
; AVX512-FCP-NEXT:    kmovw %edi, %k2
; AVX512-FCP-NEXT:    vmovdqa64 %zmm24, %zmm23 {%k2}
; AVX512-FCP-NEXT:    vpermi2d %zmm6, %zmm4, %zmm21
; AVX512-FCP-NEXT:    vpermi2d %zmm5, %zmm2, %zmm20
; AVX512-FCP-NEXT:    vpermi2d %zmm3, %zmm0, %zmm22
; AVX512-FCP-NEXT:    vmovdqa32 %zmm20, %zmm22 {%k1}
; AVX512-FCP-NEXT:    vmovdqa64 %zmm21, %zmm22 {%k2}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512-FCP-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermt2d %zmm13, %zmm20, %zmm12
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [21,27,1,7,13,0,0,0]
; AVX512-FCP-NEXT:    vpermt2d %zmm11, %zmm13, %zmm1
; AVX512-FCP-NEXT:    vmovdqa32 %zmm12, %zmm1 {%k1}
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm11 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512-FCP-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermt2d %zmm10, %zmm11, %zmm9
; AVX512-FCP-NEXT:    vmovdqa64 %zmm9, %zmm1 {%k2}
; AVX512-FCP-NEXT:    vpermt2d %zmm6, %zmm11, %zmm4
; AVX512-FCP-NEXT:    vpermt2d %zmm5, %zmm20, %zmm2
; AVX512-FCP-NEXT:    vpermt2d %zmm3, %zmm13, %zmm0
; AVX512-FCP-NEXT:    vmovdqa32 %zmm2, %zmm0 {%k1}
; AVX512-FCP-NEXT:    vmovdqa64 %zmm4, %zmm0 {%k2}
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm14, 64(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm15, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm17, 64(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm16, (%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm19, 64(%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm18, (%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm22, 64(%r9)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm23, (%r9)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, 64(%rax)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, (%rax)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i32_stride6_vf32:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-NEXT:    vmovdqa64 448(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 384(%rdi), %zmm3
; AVX512DQ-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512DQ-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512DQ-NEXT:    vmovdqa64 704(%rdi), %zmm6
; AVX512DQ-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %zmm10
; AVX512DQ-NEXT:    vmovdqa64 256(%rdi), %zmm9
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm11
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %zmm12
; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %zmm13
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512DQ-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm13, %zmm15
; AVX512DQ-NEXT:    vpermt2d %zmm12, %zmm14, %zmm15
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,6,12,18,24,30,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, %zmm8
; AVX512DQ-NEXT:    vpermt2d %zmm1, %zmm7, %zmm8
; AVX512DQ-NEXT:    movb $56, %dil
; AVX512DQ-NEXT:    kmovw %edi, %k2
; AVX512DQ-NEXT:    vmovdqa64 %zmm15, %zmm8 {%k2}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512DQ-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm9, %zmm16
; AVX512DQ-NEXT:    vpermt2d %zmm10, %zmm15, %zmm16
; AVX512DQ-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512DQ-NEXT:    kmovw %edi, %k1
; AVX512DQ-NEXT:    vmovdqa32 %zmm16, %zmm8 {%k1}
; AVX512DQ-NEXT:    vpermi2d %zmm6, %zmm4, %zmm15
; AVX512DQ-NEXT:    vpermi2d %zmm2, %zmm5, %zmm14
; AVX512DQ-NEXT:    vpermi2d %zmm0, %zmm3, %zmm7
; AVX512DQ-NEXT:    vmovdqa64 %zmm14, %zmm7 {%k2}
; AVX512DQ-NEXT:    vmovdqa32 %zmm15, %zmm7 {%k1}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm16 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512DQ-NEXT:    # zmm16 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm13, %zmm17
; AVX512DQ-NEXT:    vpermt2d %zmm12, %zmm16, %zmm17
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [1,7,13,19,25,31,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, %zmm15
; AVX512DQ-NEXT:    vpermt2d %zmm1, %zmm14, %zmm15
; AVX512DQ-NEXT:    vmovdqa64 %zmm17, %zmm15 {%k2}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512DQ-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm9, %zmm18
; AVX512DQ-NEXT:    vpermt2d %zmm10, %zmm17, %zmm18
; AVX512DQ-NEXT:    vmovdqa32 %zmm18, %zmm15 {%k1}
; AVX512DQ-NEXT:    vpermi2d %zmm6, %zmm4, %zmm17
; AVX512DQ-NEXT:    vpermi2d %zmm2, %zmm5, %zmm16
; AVX512DQ-NEXT:    vpermi2d %zmm0, %zmm3, %zmm14
; AVX512DQ-NEXT:    vmovdqa64 %zmm16, %zmm14 {%k2}
; AVX512DQ-NEXT:    vmovdqa32 %zmm17, %zmm14 {%k1}
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm18 = [2,8,14,20,26,0,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512DQ-NEXT:    vpermt2d %zmm1, %zmm18, %zmm19
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512DQ-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm12, %zmm16
; AVX512DQ-NEXT:    vpermt2d %zmm13, %zmm17, %zmm16
; AVX512DQ-NEXT:    movw $31, %di
; AVX512DQ-NEXT:    kmovw %edi, %k2
; AVX512DQ-NEXT:    vmovdqa32 %zmm19, %zmm16 {%k2}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512DQ-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm10, %zmm20
; AVX512DQ-NEXT:    vpermt2d %zmm9, %zmm19, %zmm20
; AVX512DQ-NEXT:    vmovdqa32 %zmm20, %zmm16 {%k1}
; AVX512DQ-NEXT:    vpermi2d %zmm4, %zmm6, %zmm19
; AVX512DQ-NEXT:    vpermi2d %zmm5, %zmm2, %zmm17
; AVX512DQ-NEXT:    vpermi2d %zmm0, %zmm3, %zmm18
; AVX512DQ-NEXT:    vmovdqa32 %zmm18, %zmm17 {%k2}
; AVX512DQ-NEXT:    vmovdqa32 %zmm19, %zmm17 {%k1}
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm20 = [3,9,15,21,27,0,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512DQ-NEXT:    vpermt2d %zmm1, %zmm20, %zmm21
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512DQ-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm12, %zmm18
; AVX512DQ-NEXT:    vpermt2d %zmm13, %zmm19, %zmm18
; AVX512DQ-NEXT:    vmovdqa32 %zmm21, %zmm18 {%k2}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512DQ-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm10, %zmm22
; AVX512DQ-NEXT:    vpermt2d %zmm9, %zmm21, %zmm22
; AVX512DQ-NEXT:    vmovdqa32 %zmm22, %zmm18 {%k1}
; AVX512DQ-NEXT:    vpermi2d %zmm4, %zmm6, %zmm21
; AVX512DQ-NEXT:    vpermi2d %zmm5, %zmm2, %zmm19
; AVX512DQ-NEXT:    vpermi2d %zmm0, %zmm3, %zmm20
; AVX512DQ-NEXT:    vmovdqa32 %zmm20, %zmm19 {%k2}
; AVX512DQ-NEXT:    vmovdqa32 %zmm21, %zmm19 {%k1}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512DQ-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm12, %zmm21
; AVX512DQ-NEXT:    vpermt2d %zmm13, %zmm20, %zmm21
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm22 = [20,26,0,6,12,0,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm1, %zmm23
; AVX512DQ-NEXT:    vpermt2d %zmm11, %zmm22, %zmm23
; AVX512DQ-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512DQ-NEXT:    kmovw %edi, %k1
; AVX512DQ-NEXT:    vmovdqa32 %zmm21, %zmm23 {%k1}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512DQ-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm9, %zmm24
; AVX512DQ-NEXT:    vpermt2d %zmm10, %zmm21, %zmm24
; AVX512DQ-NEXT:    movb $-32, %dil
; AVX512DQ-NEXT:    kmovw %edi, %k2
; AVX512DQ-NEXT:    vmovdqa64 %zmm24, %zmm23 {%k2}
; AVX512DQ-NEXT:    vpermi2d %zmm6, %zmm4, %zmm21
; AVX512DQ-NEXT:    vpermi2d %zmm5, %zmm2, %zmm20
; AVX512DQ-NEXT:    vpermi2d %zmm3, %zmm0, %zmm22
; AVX512DQ-NEXT:    vmovdqa32 %zmm20, %zmm22 {%k1}
; AVX512DQ-NEXT:    vmovdqa64 %zmm21, %zmm22 {%k2}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512DQ-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermt2d %zmm13, %zmm20, %zmm12
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [21,27,1,7,13,0,0,0]
; AVX512DQ-NEXT:    vpermt2d %zmm11, %zmm13, %zmm1
; AVX512DQ-NEXT:    vmovdqa32 %zmm12, %zmm1 {%k1}
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm11 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512DQ-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermt2d %zmm10, %zmm11, %zmm9
; AVX512DQ-NEXT:    vmovdqa64 %zmm9, %zmm1 {%k2}
; AVX512DQ-NEXT:    vpermt2d %zmm6, %zmm11, %zmm4
; AVX512DQ-NEXT:    vpermt2d %zmm5, %zmm20, %zmm2
; AVX512DQ-NEXT:    vpermt2d %zmm3, %zmm13, %zmm0
; AVX512DQ-NEXT:    vmovdqa32 %zmm2, %zmm0 {%k1}
; AVX512DQ-NEXT:    vmovdqa64 %zmm4, %zmm0 {%k2}
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm14, 64(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm15, (%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm17, 64(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm16, (%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm19, 64(%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm18, (%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm22, 64(%r9)
; AVX512DQ-NEXT:    vmovdqa64 %zmm23, (%r9)
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, 64(%rax)
; AVX512DQ-NEXT:    vmovdqa64 %zmm1, (%rax)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i32_stride6_vf32:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512DQ-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm6
; AVX512DQ-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm10
; AVX512DQ-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm9
; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %zmm11
; AVX512DQ-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm12
; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm13
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512DQ-FCP-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm13, %zmm15
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm12, %zmm14, %zmm15
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,6,12,18,24,30,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, %zmm8
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm1, %zmm7, %zmm8
; AVX512DQ-FCP-NEXT:    movb $56, %dil
; AVX512DQ-FCP-NEXT:    kmovw %edi, %k2
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm15, %zmm8 {%k2}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512DQ-FCP-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm9, %zmm16
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm10, %zmm15, %zmm16
; AVX512DQ-FCP-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512DQ-FCP-NEXT:    kmovw %edi, %k1
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm16, %zmm8 {%k1}
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm6, %zmm4, %zmm15
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm2, %zmm5, %zmm14
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm14, %zmm7 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm15, %zmm7 {%k1}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm16 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512DQ-FCP-NEXT:    # zmm16 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm13, %zmm17
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm12, %zmm16, %zmm17
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [1,7,13,19,25,31,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, %zmm15
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm1, %zmm14, %zmm15
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm17, %zmm15 {%k2}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512DQ-FCP-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm9, %zmm18
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm10, %zmm17, %zmm18
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm18, %zmm15 {%k1}
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm6, %zmm4, %zmm17
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm2, %zmm5, %zmm16
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm14
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm16, %zmm14 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm17, %zmm14 {%k1}
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm18 = [2,8,14,20,26,0,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm1, %zmm18, %zmm19
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512DQ-FCP-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm12, %zmm16
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm13, %zmm17, %zmm16
; AVX512DQ-FCP-NEXT:    movw $31, %di
; AVX512DQ-FCP-NEXT:    kmovw %edi, %k2
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm19, %zmm16 {%k2}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512DQ-FCP-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm10, %zmm20
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm9, %zmm19, %zmm20
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm20, %zmm16 {%k1}
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm4, %zmm6, %zmm19
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm5, %zmm2, %zmm17
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm18
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm18, %zmm17 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm19, %zmm17 {%k1}
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm20 = [3,9,15,21,27,0,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm1, %zmm20, %zmm21
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512DQ-FCP-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm12, %zmm18
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm13, %zmm19, %zmm18
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm21, %zmm18 {%k2}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512DQ-FCP-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm10, %zmm22
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm9, %zmm21, %zmm22
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm22, %zmm18 {%k1}
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm4, %zmm6, %zmm21
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm5, %zmm2, %zmm19
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm20
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm20, %zmm19 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm21, %zmm19 {%k1}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512DQ-FCP-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm12, %zmm21
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm13, %zmm20, %zmm21
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm22 = [20,26,0,6,12,0,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, %zmm23
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm11, %zmm22, %zmm23
; AVX512DQ-FCP-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512DQ-FCP-NEXT:    kmovw %edi, %k1
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm21, %zmm23 {%k1}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512DQ-FCP-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm9, %zmm24
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm10, %zmm21, %zmm24
; AVX512DQ-FCP-NEXT:    movb $-32, %dil
; AVX512DQ-FCP-NEXT:    kmovw %edi, %k2
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm24, %zmm23 {%k2}
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm6, %zmm4, %zmm21
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm5, %zmm2, %zmm20
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm3, %zmm0, %zmm22
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm20, %zmm22 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm21, %zmm22 {%k2}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512DQ-FCP-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm13, %zmm20, %zmm12
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [21,27,1,7,13,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm11, %zmm13, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm12, %zmm1 {%k1}
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm11 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512DQ-FCP-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm10, %zmm11, %zmm9
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm9, %zmm1 {%k2}
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm6, %zmm11, %zmm4
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm5, %zmm20, %zmm2
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm3, %zmm13, %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm2, %zmm0 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm4, %zmm0 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm14, 64(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm15, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm17, 64(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm16, (%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm19, 64(%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm18, (%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm22, 64(%r9)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm23, (%r9)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, 64(%rax)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, (%rax)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i32_stride6_vf32:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm3
; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm6
; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm10
; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm9
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm11
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm12
; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm13
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512BW-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm15
; AVX512BW-NEXT:    vpermt2d %zmm12, %zmm14, %zmm15
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,6,12,18,24,30,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm8
; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm7, %zmm8
; AVX512BW-NEXT:    movb $56, %dil
; AVX512BW-NEXT:    kmovd %edi, %k2
; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm8 {%k2}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512BW-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm16
; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm15, %zmm16
; AVX512BW-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512BW-NEXT:    kmovd %edi, %k1
; AVX512BW-NEXT:    vmovdqa32 %zmm16, %zmm8 {%k1}
; AVX512BW-NEXT:    vpermi2d %zmm6, %zmm4, %zmm15
; AVX512BW-NEXT:    vpermi2d %zmm2, %zmm5, %zmm14
; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm3, %zmm7
; AVX512BW-NEXT:    vmovdqa64 %zmm14, %zmm7 {%k2}
; AVX512BW-NEXT:    vmovdqa32 %zmm15, %zmm7 {%k1}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm16 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512BW-NEXT:    # zmm16 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm17
; AVX512BW-NEXT:    vpermt2d %zmm12, %zmm16, %zmm17
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [1,7,13,19,25,31,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm15
; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm14, %zmm15
; AVX512BW-NEXT:    vmovdqa64 %zmm17, %zmm15 {%k2}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512BW-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm18
; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm17, %zmm18
; AVX512BW-NEXT:    vmovdqa32 %zmm18, %zmm15 {%k1}
; AVX512BW-NEXT:    vpermi2d %zmm6, %zmm4, %zmm17
; AVX512BW-NEXT:    vpermi2d %zmm2, %zmm5, %zmm16
; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm3, %zmm14
; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm14 {%k2}
; AVX512BW-NEXT:    vmovdqa32 %zmm17, %zmm14 {%k1}
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm18 = [2,8,14,20,26,0,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm18, %zmm19
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512BW-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm16
; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm17, %zmm16
; AVX512BW-NEXT:    movw $31, %di
; AVX512BW-NEXT:    kmovd %edi, %k2
; AVX512BW-NEXT:    vmovdqa32 %zmm19, %zmm16 {%k2}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512BW-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm20
; AVX512BW-NEXT:    vpermt2d %zmm9, %zmm19, %zmm20
; AVX512BW-NEXT:    vmovdqa32 %zmm20, %zmm16 {%k1}
; AVX512BW-NEXT:    vpermi2d %zmm4, %zmm6, %zmm19
; AVX512BW-NEXT:    vpermi2d %zmm5, %zmm2, %zmm17
; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm3, %zmm18
; AVX512BW-NEXT:    vmovdqa32 %zmm18, %zmm17 {%k2}
; AVX512BW-NEXT:    vmovdqa32 %zmm19, %zmm17 {%k1}
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm20 = [3,9,15,21,27,0,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm20, %zmm21
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512BW-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm18
; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm19, %zmm18
; AVX512BW-NEXT:    vmovdqa32 %zmm21, %zmm18 {%k2}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512BW-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm22
; AVX512BW-NEXT:    vpermt2d %zmm9, %zmm21, %zmm22
; AVX512BW-NEXT:    vmovdqa32 %zmm22, %zmm18 {%k1}
; AVX512BW-NEXT:    vpermi2d %zmm4, %zmm6, %zmm21
; AVX512BW-NEXT:    vpermi2d %zmm5, %zmm2, %zmm19
; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm3, %zmm20
; AVX512BW-NEXT:    vmovdqa32 %zmm20, %zmm19 {%k2}
; AVX512BW-NEXT:    vmovdqa32 %zmm21, %zmm19 {%k1}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512BW-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm12, %zmm21
; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm20, %zmm21
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm22 = [20,26,0,6,12,0,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm23
; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm22, %zmm23
; AVX512BW-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512BW-NEXT:    kmovd %edi, %k1
; AVX512BW-NEXT:    vmovdqa32 %zmm21, %zmm23 {%k1}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512BW-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm24
; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm21, %zmm24
; AVX512BW-NEXT:    movb $-32, %dil
; AVX512BW-NEXT:    kmovd %edi, %k2
; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm23 {%k2}
; AVX512BW-NEXT:    vpermi2d %zmm6, %zmm4, %zmm21
; AVX512BW-NEXT:    vpermi2d %zmm5, %zmm2, %zmm20
; AVX512BW-NEXT:    vpermi2d %zmm3, %zmm0, %zmm22
; AVX512BW-NEXT:    vmovdqa32 %zmm20, %zmm22 {%k1}
; AVX512BW-NEXT:    vmovdqa64 %zmm21, %zmm22 {%k2}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512BW-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermt2d %zmm13, %zmm20, %zmm12
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [21,27,1,7,13,0,0,0]
; AVX512BW-NEXT:    vpermt2d %zmm11, %zmm13, %zmm1
; AVX512BW-NEXT:    vmovdqa32 %zmm12, %zmm1 {%k1}
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm11 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512BW-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm11, %zmm9
; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm1 {%k2}
; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm11, %zmm4
; AVX512BW-NEXT:    vpermt2d %zmm5, %zmm20, %zmm2
; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm13, %zmm0
; AVX512BW-NEXT:    vmovdqa32 %zmm2, %zmm0 {%k1}
; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm0 {%k2}
; AVX512BW-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm14, 64(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm15, (%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm17, 64(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm16, (%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm19, 64(%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm18, (%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm22, 64(%r9)
; AVX512BW-NEXT:    vmovdqa64 %zmm23, (%r9)
; AVX512BW-NEXT:    vmovdqa64 %zmm0, 64(%rax)
; AVX512BW-NEXT:    vmovdqa64 %zmm1, (%rax)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i32_stride6_vf32:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512BW-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm6
; AVX512BW-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm10
; AVX512BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm9
; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm11
; AVX512BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm12
; AVX512BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm13
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512BW-FCP-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm13, %zmm15
; AVX512BW-FCP-NEXT:    vpermt2d %zmm12, %zmm14, %zmm15
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,6,12,18,24,30,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm8
; AVX512BW-FCP-NEXT:    vpermt2d %zmm1, %zmm7, %zmm8
; AVX512BW-FCP-NEXT:    movb $56, %dil
; AVX512BW-FCP-NEXT:    kmovd %edi, %k2
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm15, %zmm8 {%k2}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512BW-FCP-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm16
; AVX512BW-FCP-NEXT:    vpermt2d %zmm10, %zmm15, %zmm16
; AVX512BW-FCP-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512BW-FCP-NEXT:    kmovd %edi, %k1
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm16, %zmm8 {%k1}
; AVX512BW-FCP-NEXT:    vpermi2d %zmm6, %zmm4, %zmm15
; AVX512BW-FCP-NEXT:    vpermi2d %zmm2, %zmm5, %zmm14
; AVX512BW-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm14, %zmm7 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm15, %zmm7 {%k1}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm16 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512BW-FCP-NEXT:    # zmm16 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm13, %zmm17
; AVX512BW-FCP-NEXT:    vpermt2d %zmm12, %zmm16, %zmm17
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [1,7,13,19,25,31,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm15
; AVX512BW-FCP-NEXT:    vpermt2d %zmm1, %zmm14, %zmm15
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm17, %zmm15 {%k2}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512BW-FCP-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm18
; AVX512BW-FCP-NEXT:    vpermt2d %zmm10, %zmm17, %zmm18
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm18, %zmm15 {%k1}
; AVX512BW-FCP-NEXT:    vpermi2d %zmm6, %zmm4, %zmm17
; AVX512BW-FCP-NEXT:    vpermi2d %zmm2, %zmm5, %zmm16
; AVX512BW-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm14
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm16, %zmm14 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm17, %zmm14 {%k1}
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm18 = [2,8,14,20,26,0,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512BW-FCP-NEXT:    vpermt2d %zmm1, %zmm18, %zmm19
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512BW-FCP-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm12, %zmm16
; AVX512BW-FCP-NEXT:    vpermt2d %zmm13, %zmm17, %zmm16
; AVX512BW-FCP-NEXT:    movw $31, %di
; AVX512BW-FCP-NEXT:    kmovd %edi, %k2
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm19, %zmm16 {%k2}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512BW-FCP-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm10, %zmm20
; AVX512BW-FCP-NEXT:    vpermt2d %zmm9, %zmm19, %zmm20
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm20, %zmm16 {%k1}
; AVX512BW-FCP-NEXT:    vpermi2d %zmm4, %zmm6, %zmm19
; AVX512BW-FCP-NEXT:    vpermi2d %zmm5, %zmm2, %zmm17
; AVX512BW-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm18
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm18, %zmm17 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm19, %zmm17 {%k1}
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm20 = [3,9,15,21,27,0,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512BW-FCP-NEXT:    vpermt2d %zmm1, %zmm20, %zmm21
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512BW-FCP-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm12, %zmm18
; AVX512BW-FCP-NEXT:    vpermt2d %zmm13, %zmm19, %zmm18
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm21, %zmm18 {%k2}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512BW-FCP-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm10, %zmm22
; AVX512BW-FCP-NEXT:    vpermt2d %zmm9, %zmm21, %zmm22
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm22, %zmm18 {%k1}
; AVX512BW-FCP-NEXT:    vpermi2d %zmm4, %zmm6, %zmm21
; AVX512BW-FCP-NEXT:    vpermi2d %zmm5, %zmm2, %zmm19
; AVX512BW-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm20
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm20, %zmm19 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm21, %zmm19 {%k1}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512BW-FCP-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm12, %zmm21
; AVX512BW-FCP-NEXT:    vpermt2d %zmm13, %zmm20, %zmm21
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm22 = [20,26,0,6,12,0,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm23
; AVX512BW-FCP-NEXT:    vpermt2d %zmm11, %zmm22, %zmm23
; AVX512BW-FCP-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512BW-FCP-NEXT:    kmovd %edi, %k1
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm21, %zmm23 {%k1}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512BW-FCP-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm24
; AVX512BW-FCP-NEXT:    vpermt2d %zmm10, %zmm21, %zmm24
; AVX512BW-FCP-NEXT:    movb $-32, %dil
; AVX512BW-FCP-NEXT:    kmovd %edi, %k2
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm24, %zmm23 {%k2}
; AVX512BW-FCP-NEXT:    vpermi2d %zmm6, %zmm4, %zmm21
; AVX512BW-FCP-NEXT:    vpermi2d %zmm5, %zmm2, %zmm20
; AVX512BW-FCP-NEXT:    vpermi2d %zmm3, %zmm0, %zmm22
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm20, %zmm22 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm21, %zmm22 {%k2}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512BW-FCP-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermt2d %zmm13, %zmm20, %zmm12
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [21,27,1,7,13,0,0,0]
; AVX512BW-FCP-NEXT:    vpermt2d %zmm11, %zmm13, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm12, %zmm1 {%k1}
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm11 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512BW-FCP-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermt2d %zmm10, %zmm11, %zmm9
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm1 {%k2}
; AVX512BW-FCP-NEXT:    vpermt2d %zmm6, %zmm11, %zmm4
; AVX512BW-FCP-NEXT:    vpermt2d %zmm5, %zmm20, %zmm2
; AVX512BW-FCP-NEXT:    vpermt2d %zmm3, %zmm13, %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm2, %zmm0 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm0 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm14, 64(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm15, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm17, 64(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm16, (%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm19, 64(%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm18, (%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm22, 64(%r9)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm23, (%r9)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, 64(%rax)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, (%rax)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i32_stride6_vf32:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-BW-NEXT:    vmovdqa64 448(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 384(%rdi), %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512DQ-BW-NEXT:    vmovdqa64 704(%rdi), %zmm6
; AVX512DQ-BW-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512DQ-BW-NEXT:    vmovdqa64 320(%rdi), %zmm10
; AVX512DQ-BW-NEXT:    vmovdqa64 256(%rdi), %zmm9
; AVX512DQ-BW-NEXT:    vmovdqa64 (%rdi), %zmm11
; AVX512DQ-BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 128(%rdi), %zmm12
; AVX512DQ-BW-NEXT:    vmovdqa64 192(%rdi), %zmm13
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512DQ-BW-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm13, %zmm15
; AVX512DQ-BW-NEXT:    vpermt2d %zmm12, %zmm14, %zmm15
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,6,12,18,24,30,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, %zmm8
; AVX512DQ-BW-NEXT:    vpermt2d %zmm1, %zmm7, %zmm8
; AVX512DQ-BW-NEXT:    movb $56, %dil
; AVX512DQ-BW-NEXT:    kmovd %edi, %k2
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm15, %zmm8 {%k2}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512DQ-BW-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm9, %zmm16
; AVX512DQ-BW-NEXT:    vpermt2d %zmm10, %zmm15, %zmm16
; AVX512DQ-BW-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512DQ-BW-NEXT:    kmovd %edi, %k1
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm16, %zmm8 {%k1}
; AVX512DQ-BW-NEXT:    vpermi2d %zmm6, %zmm4, %zmm15
; AVX512DQ-BW-NEXT:    vpermi2d %zmm2, %zmm5, %zmm14
; AVX512DQ-BW-NEXT:    vpermi2d %zmm0, %zmm3, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm14, %zmm7 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm15, %zmm7 {%k1}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm16 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512DQ-BW-NEXT:    # zmm16 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm13, %zmm17
; AVX512DQ-BW-NEXT:    vpermt2d %zmm12, %zmm16, %zmm17
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [1,7,13,19,25,31,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, %zmm15
; AVX512DQ-BW-NEXT:    vpermt2d %zmm1, %zmm14, %zmm15
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm17, %zmm15 {%k2}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512DQ-BW-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm9, %zmm18
; AVX512DQ-BW-NEXT:    vpermt2d %zmm10, %zmm17, %zmm18
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm18, %zmm15 {%k1}
; AVX512DQ-BW-NEXT:    vpermi2d %zmm6, %zmm4, %zmm17
; AVX512DQ-BW-NEXT:    vpermi2d %zmm2, %zmm5, %zmm16
; AVX512DQ-BW-NEXT:    vpermi2d %zmm0, %zmm3, %zmm14
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm16, %zmm14 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm17, %zmm14 {%k1}
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm18 = [2,8,14,20,26,0,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512DQ-BW-NEXT:    vpermt2d %zmm1, %zmm18, %zmm19
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512DQ-BW-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm12, %zmm16
; AVX512DQ-BW-NEXT:    vpermt2d %zmm13, %zmm17, %zmm16
; AVX512DQ-BW-NEXT:    movw $31, %di
; AVX512DQ-BW-NEXT:    kmovd %edi, %k2
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm19, %zmm16 {%k2}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512DQ-BW-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm10, %zmm20
; AVX512DQ-BW-NEXT:    vpermt2d %zmm9, %zmm19, %zmm20
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm20, %zmm16 {%k1}
; AVX512DQ-BW-NEXT:    vpermi2d %zmm4, %zmm6, %zmm19
; AVX512DQ-BW-NEXT:    vpermi2d %zmm5, %zmm2, %zmm17
; AVX512DQ-BW-NEXT:    vpermi2d %zmm0, %zmm3, %zmm18
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm18, %zmm17 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm19, %zmm17 {%k1}
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm20 = [3,9,15,21,27,0,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512DQ-BW-NEXT:    vpermt2d %zmm1, %zmm20, %zmm21
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512DQ-BW-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm12, %zmm18
; AVX512DQ-BW-NEXT:    vpermt2d %zmm13, %zmm19, %zmm18
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm21, %zmm18 {%k2}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512DQ-BW-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm10, %zmm22
; AVX512DQ-BW-NEXT:    vpermt2d %zmm9, %zmm21, %zmm22
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm22, %zmm18 {%k1}
; AVX512DQ-BW-NEXT:    vpermi2d %zmm4, %zmm6, %zmm21
; AVX512DQ-BW-NEXT:    vpermi2d %zmm5, %zmm2, %zmm19
; AVX512DQ-BW-NEXT:    vpermi2d %zmm0, %zmm3, %zmm20
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm20, %zmm19 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm21, %zmm19 {%k1}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512DQ-BW-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm12, %zmm21
; AVX512DQ-BW-NEXT:    vpermt2d %zmm13, %zmm20, %zmm21
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm22 = [20,26,0,6,12,0,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, %zmm23
; AVX512DQ-BW-NEXT:    vpermt2d %zmm11, %zmm22, %zmm23
; AVX512DQ-BW-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512DQ-BW-NEXT:    kmovd %edi, %k1
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm21, %zmm23 {%k1}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512DQ-BW-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm9, %zmm24
; AVX512DQ-BW-NEXT:    vpermt2d %zmm10, %zmm21, %zmm24
; AVX512DQ-BW-NEXT:    movb $-32, %dil
; AVX512DQ-BW-NEXT:    kmovd %edi, %k2
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm24, %zmm23 {%k2}
; AVX512DQ-BW-NEXT:    vpermi2d %zmm6, %zmm4, %zmm21
; AVX512DQ-BW-NEXT:    vpermi2d %zmm5, %zmm2, %zmm20
; AVX512DQ-BW-NEXT:    vpermi2d %zmm3, %zmm0, %zmm22
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm20, %zmm22 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm21, %zmm22 {%k2}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512DQ-BW-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermt2d %zmm13, %zmm20, %zmm12
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [21,27,1,7,13,0,0,0]
; AVX512DQ-BW-NEXT:    vpermt2d %zmm11, %zmm13, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm12, %zmm1 {%k1}
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm11 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512DQ-BW-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermt2d %zmm10, %zmm11, %zmm9
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm9, %zmm1 {%k2}
; AVX512DQ-BW-NEXT:    vpermt2d %zmm6, %zmm11, %zmm4
; AVX512DQ-BW-NEXT:    vpermt2d %zmm5, %zmm20, %zmm2
; AVX512DQ-BW-NEXT:    vpermt2d %zmm3, %zmm13, %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm2, %zmm0 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm4, %zmm0 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm14, 64(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm15, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm17, 64(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm16, (%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm19, 64(%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm18, (%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm22, 64(%r9)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm23, (%r9)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, 64(%rax)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, (%rax)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i32_stride6_vf32:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm6
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm10
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm9
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm11
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm12
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm13
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512DQ-BW-FCP-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm13, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm12, %zmm14, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm7 = [0,6,12,18,24,30,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm1, %zmm7, %zmm8
; AVX512DQ-BW-FCP-NEXT:    movb $56, %dil
; AVX512DQ-BW-FCP-NEXT:    kmovd %edi, %k2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm15, %zmm8 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512DQ-BW-FCP-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm10, %zmm15, %zmm16
; AVX512DQ-BW-FCP-NEXT:    movw $-2048, %di # imm = 0xF800
; AVX512DQ-BW-FCP-NEXT:    kmovd %edi, %k1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm16, %zmm8 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm6, %zmm4, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm2, %zmm5, %zmm14
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm14, %zmm7 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm15, %zmm7 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm16 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512DQ-BW-FCP-NEXT:    # zmm16 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm13, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm12, %zmm16, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [1,7,13,19,25,31,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm1, %zmm14, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm17, %zmm15 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512DQ-BW-FCP-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm10, %zmm17, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm18, %zmm15 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm6, %zmm4, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm2, %zmm5, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm14
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm16, %zmm14 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm17, %zmm14 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm18 = [2,8,14,20,26,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm1, %zmm18, %zmm19
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm17 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512DQ-BW-FCP-NEXT:    # zmm17 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm12, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm13, %zmm17, %zmm16
; AVX512DQ-BW-FCP-NEXT:    movw $31, %di
; AVX512DQ-BW-FCP-NEXT:    kmovd %edi, %k2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm19, %zmm16 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512DQ-BW-FCP-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm10, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm9, %zmm19, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm20, %zmm16 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm4, %zmm6, %zmm19
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm5, %zmm2, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm18, %zmm17 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm19, %zmm17 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm20 = [3,9,15,21,27,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm1, %zmm20, %zmm21
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512DQ-BW-FCP-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm12, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm13, %zmm19, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm21, %zmm18 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512DQ-BW-FCP-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm10, %zmm22
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm9, %zmm21, %zmm22
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm22, %zmm18 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm4, %zmm6, %zmm21
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm5, %zmm2, %zmm19
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm0, %zmm3, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm20, %zmm19 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm21, %zmm19 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512DQ-BW-FCP-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm12, %zmm21
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm13, %zmm20, %zmm21
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm22 = [20,26,0,6,12,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm23
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm11, %zmm22, %zmm23
; AVX512DQ-BW-FCP-NEXT:    movw $992, %di # imm = 0x3E0
; AVX512DQ-BW-FCP-NEXT:    kmovd %edi, %k1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm21, %zmm23 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512DQ-BW-FCP-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm24
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm10, %zmm21, %zmm24
; AVX512DQ-BW-FCP-NEXT:    movb $-32, %dil
; AVX512DQ-BW-FCP-NEXT:    kmovd %edi, %k2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm24, %zmm23 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm6, %zmm4, %zmm21
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm5, %zmm2, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm3, %zmm0, %zmm22
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm20, %zmm22 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm21, %zmm22 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm20 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512DQ-BW-FCP-NEXT:    # zmm20 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm13, %zmm20, %zmm12
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm13 = [21,27,1,7,13,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm11, %zmm13, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm12, %zmm1 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm11 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512DQ-BW-FCP-NEXT:    # zmm11 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm10, %zmm11, %zmm9
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm1 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm6, %zmm11, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm5, %zmm20, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm3, %zmm13, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm2, %zmm0 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm0 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm14, 64(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm15, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm17, 64(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm16, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm19, 64(%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm18, (%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm22, 64(%r9)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm23, (%r9)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, 64(%rax)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm1, (%rax)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <192 x i32>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <32 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90, i32 96, i32 102, i32 108, i32 114, i32 120, i32 126, i32 132, i32 138, i32 144, i32 150, i32 156, i32 162, i32 168, i32 174, i32 180, i32 186>
  %strided.vec1 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <32 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91, i32 97, i32 103, i32 109, i32 115, i32 121, i32 127, i32 133, i32 139, i32 145, i32 151, i32 157, i32 163, i32 169, i32 175, i32 181, i32 187>
  %strided.vec2 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <32 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92, i32 98, i32 104, i32 110, i32 116, i32 122, i32 128, i32 134, i32 140, i32 146, i32 152, i32 158, i32 164, i32 170, i32 176, i32 182, i32 188>
  %strided.vec3 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <32 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93, i32 99, i32 105, i32 111, i32 117, i32 123, i32 129, i32 135, i32 141, i32 147, i32 153, i32 159, i32 165, i32 171, i32 177, i32 183, i32 189>
  %strided.vec4 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <32 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94, i32 100, i32 106, i32 112, i32 118, i32 124, i32 130, i32 136, i32 142, i32 148, i32 154, i32 160, i32 166, i32 172, i32 178, i32 184, i32 190>
  %strided.vec5 = shufflevector <192 x i32> %wide.vec, <192 x i32> poison, <32 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95, i32 101, i32 107, i32 113, i32 119, i32 125, i32 131, i32 137, i32 143, i32 149, i32 155, i32 161, i32 167, i32 173, i32 179, i32 185, i32 191>
  store <32 x i32> %strided.vec0, ptr %out.vec0, align 64
  store <32 x i32> %strided.vec1, ptr %out.vec1, align 64
  store <32 x i32> %strided.vec2, ptr %out.vec2, align 64
  store <32 x i32> %strided.vec3, ptr %out.vec3, align 64
  store <32 x i32> %strided.vec4, ptr %out.vec4, align 64
  store <32 x i32> %strided.vec5, ptr %out.vec5, align 64
  ret void
}

define void @load_i32_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
; SSE-LABEL: load_i32_stride6_vf64:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $2184, %rsp # imm = 0x888
; SSE-NEXT:    movdqa 912(%rdi), %xmm7
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 928(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 864(%rdi), %xmm8
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 880(%rdi), %xmm4
; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 528(%rdi), %xmm9
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 544(%rdi), %xmm5
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 480(%rdi), %xmm10
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 496(%rdi), %xmm6
; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 144(%rdi), %xmm11
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 160(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 96(%rdi), %xmm12
; SSE-NEXT:    movdqa %xmm12, (%rsp) # 16-byte Spill
; SSE-NEXT:    movdqa 112(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT:    movdqa %xmm12, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
; SSE-NEXT:    movdqa %xmm10, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
; SSE-NEXT:    movdqa %xmm8, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1248(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1264(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT:    movdqa %xmm2, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 1296(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1312(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa (%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-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:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 64(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 48(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 384(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-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:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 432(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 448(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 768(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-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:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 816(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 832(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1152(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1168(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 1200(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1216(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 288(%rdi), %xmm2
; SSE-NEXT:    movdqa 304(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT:    movdqa %xmm2, %xmm1
; SSE-NEXT:    movdqa %xmm2, %xmm15
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 336(%rdi), %xmm7
; SSE-NEXT:    movdqa 352(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[0,0,1,1]
; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 672(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 688(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 720(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 736(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1056(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1072(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT:    movdqa %xmm2, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 1104(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1120(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1440(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1456(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    movdqa 1488(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1504(%rdi), %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 192(%rdi), %xmm5
; SSE-NEXT:    movdqa 208(%rdi), %xmm6
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,3,2,3]
; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm5, %xmm3
; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movdqa 240(%rdi), %xmm2
; SSE-NEXT:    movdqa 256(%rdi), %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,2,3,3]
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm2[0,0,1,1]
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm3[0],xmm12[1]
; SSE-NEXT:    movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 576(%rdi), %xmm10
; SSE-NEXT:    movdqa 592(%rdi), %xmm14
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm14[2,3,2,3]
; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm10, %xmm4
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; SSE-NEXT:    movdqa 624(%rdi), %xmm11
; SSE-NEXT:    movdqa 640(%rdi), %xmm13
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[2,2,3,3]
; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm11[0,0,1,1]
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1]
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm4[0],xmm12[1]
; SSE-NEXT:    movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 960(%rdi), %xmm4
; SSE-NEXT:    movdqa %xmm4, {{[-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:    pshufd {{.*#+}} xmm3 = xmm0[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; SSE-NEXT:    movdqa 1008(%rdi), %xmm8
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1024(%rdi), %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,2,3,3]
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm8[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1]
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm4[0],xmm12[1]
; SSE-NEXT:    movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1344(%rdi), %xmm9
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1360(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
; SSE-NEXT:    movdqa %xmm9, %xmm4
; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; SSE-NEXT:    movdqa 1392(%rdi), %xmm9
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 1408(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,2,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm9[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1]
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm4[0],xmm12[1]
; SSE-NEXT:    movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, (%rsp), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm9, %xmm12
; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm3[0],xmm12[1]
; SSE-NEXT:    movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm8, %xmm12
; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm3[0],xmm12[1]
; SSE-NEXT:    movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm15[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm3[0],xmm7[1]
; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm3[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm15[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm12, %xmm7
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm3[0],xmm7[1]
; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm2, %xmm5
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm3[0],xmm5[1]
; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm3[0],xmm5[1]
; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm14[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[2,3,2,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm4[0],xmm11[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm3[0],xmm11[1]
; SSE-NEXT:    movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm11, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm13, %xmm1
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm14, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    movdqa %xmm5, %xmm0
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
; SSE-NEXT:    movdqa 80(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
; SSE-NEXT:    movdqa 32(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
; SSE-NEXT:    movdqa 176(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd $238, (%rsp), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
; SSE-NEXT:    movdqa 128(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    movdqa 272(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
; SSE-NEXT:    movdqa 224(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    movdqa 368(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
; SSE-NEXT:    movdqa 320(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
; SSE-NEXT:    movdqa 464(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[2,3,2,3]
; SSE-NEXT:    movdqa 416(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
; SSE-NEXT:    movdqa 560(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm15[2,3,2,3]
; SSE-NEXT:    movdqa 512(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    movdqa 656(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
; SSE-NEXT:    movdqa 608(%rdi), %xmm9
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1]
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    movdqa 752(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm12[2,3,2,3]
; SSE-NEXT:    movdqa 704(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    movdqa 848(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[2,3,2,3]
; SSE-NEXT:    movdqa 800(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
; SSE-NEXT:    movdqa 944(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
; SSE-NEXT:    movdqa 896(%rdi), %xmm2
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; SSE-NEXT:    movdqa %xmm2, %xmm8
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
; SSE-NEXT:    movdqa 1040(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[2,3,2,3]
; SSE-NEXT:    movdqa 992(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3]
; SSE-NEXT:    movdqa 1136(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
; SSE-NEXT:    movdqa 1088(%rdi), %xmm6
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1]
; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
; SSE-NEXT:    movdqa 1232(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[2,3,2,3]
; SSE-NEXT:    movdqa 1184(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    movdqa 1328(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
; SSE-NEXT:    movdqa 1280(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    movdqa 1424(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
; SSE-NEXT:    movdqa 1376(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    movdqa 1520(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
; SSE-NEXT:    movdqa 1472(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    # xmm4 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $255, (%rsp), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, (%rsp) # 16-byte Spill
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[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:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[3,3,3,3]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[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:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[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:    pshufd {{.*#+}} xmm0 = xmm10[3,3,3,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; SSE-NEXT:    # xmm15 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1]
; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; SSE-NEXT:    # xmm11 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm1[0],xmm11[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; SSE-NEXT:    # xmm9 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT:    # xmm8 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
; SSE-NEXT:    # xmm13 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm1[0],xmm13[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT:    # xmm7 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; SSE-NEXT:    # xmm5 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT:    # xmm3 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[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:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
; SSE-NEXT:    movapd %xmm15, %xmm4
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
; SSE-NEXT:    movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
; SSE-NEXT:    movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
; SSE-NEXT:    movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
; SSE-NEXT:    movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
; SSE-NEXT:    movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm1[0],xmm12[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm0[0],xmm12[1]
; SSE-NEXT:    movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm1[0],xmm12[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm0[0],xmm12[1]
; SSE-NEXT:    movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[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:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm1[0],xmm12[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm0[0],xmm12[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:    pshufd {{.*#+}} xmm1 = xmm6[3,3,3,3]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 224(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 160(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 96(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 240(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 176(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 112(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 192(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 128(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 64(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 208(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 144(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 80(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 224(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 240(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 192(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 208(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 160(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 176(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 128(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 144(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 96(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 112(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 64(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 80(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 240(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 224(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 208(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 192(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 176(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 160(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 144(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 128(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 112(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 96(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 80(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 64(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 240(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 224(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 208(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 192(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 176(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 160(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 144(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 128(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 112(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 96(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 80(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 64(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%r8)
; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%r8)
; SSE-NEXT:    movapd %xmm2, 240(%r9)
; SSE-NEXT:    movapd %xmm3, 224(%r9)
; SSE-NEXT:    movapd %xmm5, 208(%r9)
; SSE-NEXT:    movapd %xmm7, 192(%r9)
; SSE-NEXT:    movapd %xmm13, 176(%r9)
; SSE-NEXT:    movapd %xmm8, 160(%r9)
; SSE-NEXT:    movapd %xmm9, 144(%r9)
; SSE-NEXT:    movapd %xmm11, 128(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 112(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 96(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 80(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 64(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%r9)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, (%r9)
; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT:    movapd %xmm14, 240(%rax)
; SSE-NEXT:    movapd %xmm12, 224(%rax)
; SSE-NEXT:    movapd %xmm15, 208(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 192(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 176(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 160(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 144(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 128(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 112(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 96(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 80(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 64(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%rax)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rax)
; SSE-NEXT:    movapd %xmm4, (%rax)
; SSE-NEXT:    addq $2184, %rsp # imm = 0x888
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i32_stride6_vf64:
; AVX:       # %bb.0:
; AVX-NEXT:    subq $2584, %rsp # imm = 0xA18
; AVX-NEXT:    vmovaps 608(%rdi), %ymm6
; AVX-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 672(%rdi), %ymm2
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 640(%rdi), %ymm3
; AVX-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 352(%rdi), %ymm4
; AVX-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 320(%rdi), %ymm5
; AVX-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 224(%rdi), %ymm7
; AVX-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 192(%rdi), %ymm8
; AVX-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 288(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 256(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 288(%rdi), %ymm0, %ymm9
; AVX-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm9[0,0],ymm1[6,4],ymm9[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm9[2,2],ymm0[6,4],ymm9[6,6]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm7[4,5],ymm8[6,7]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm9
; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm9[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm9[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm4[0,1]
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm5[0],ymm1[3],ymm5[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 672(%rdi), %ymm3, %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[2,0],ymm1[0,0],ymm2[6,4],ymm1[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,2],ymm0[6,4],ymm1[6,6]
; AVX-NEXT:    vmovaps 576(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vmovapd 736(%rdi), %ymm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 704(%rdi), %ymm2
; AVX-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1056(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1024(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 1056(%rdi), %ymm0, %ymm2
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm2[0,0],ymm1[6,4],ymm2[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm2[2,2],ymm0[6,4],ymm2[6,6]
; AVX-NEXT:    vmovaps 992(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 960(%rdi), %ymm2
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm13
; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm13[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vmovapd 1120(%rdi), %ymm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 1088(%rdi), %ymm2
; AVX-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1440(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1408(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 1440(%rdi), %ymm0, %ymm11
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm11[0,0],ymm1[6,4],ymm11[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm11[2,2],ymm0[6,4],ymm11[6,6]
; AVX-NEXT:    vmovaps 1376(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1344(%rdi), %ymm2
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm10
; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm10[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm10[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vmovapd 1504(%rdi), %ymm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 1472(%rdi), %ymm2
; AVX-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 96(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 64(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 96(%rdi), %ymm0, %ymm8
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm8[0,0],ymm1[6,4],ymm8[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm8[2,2],ymm0[6,4],ymm8[6,6]
; AVX-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps (%rdi), %ymm2
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm7
; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm7[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vmovapd 160(%rdi), %ymm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 128(%rdi), %ymm2
; AVX-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 480(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 448(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 480(%rdi), %ymm0, %ymm6
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm6[0,0],ymm1[6,4],ymm6[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm6[2,2],ymm0[6,4],ymm6[6,6]
; AVX-NEXT:    vmovaps 416(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 384(%rdi), %ymm2
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm4
; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm4[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vmovapd 544(%rdi), %ymm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 512(%rdi), %ymm2
; AVX-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 864(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 832(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 864(%rdi), %ymm0, %ymm3
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm3[0,0],ymm1[6,4],ymm3[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm3[2,2],ymm0[6,4],ymm3[6,6]
; AVX-NEXT:    vmovaps 800(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 768(%rdi), %ymm2
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vextractf128 $1, %ymm1, %xmm2
; AVX-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vmovapd 928(%rdi), %ymm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 896(%rdi), %ymm5
; AVX-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm1[0,1]
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm5[0],ymm1[3],ymm5[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1248(%rdi), %ymm5
; AVX-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1216(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vinsertf128 $1, 1248(%rdi), %ymm0, %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm5[2,0],ymm1[0,0],ymm5[6,4],ymm1[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm15 = ymm0[2,0],ymm1[2,2],ymm0[6,4],ymm1[6,6]
; AVX-NEXT:    vmovaps 1184(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1152(%rdi), %ymm5
; AVX-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5],ymm5[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm5
; AVX-NEXT:    vblendps {{.*#+}} xmm14 = xmm0[0,1],xmm5[2,3]
; AVX-NEXT:    vshufps {{.*#+}} xmm14 = xmm14[0,2],xmm5[0,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX-NEXT:    vmovapd 1312(%rdi), %ymm12
; AVX-NEXT:    vmovupd %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 1280(%rdi), %ymm0
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm12[0,1]
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm15 = ymm1[0],ymm0[0],ymm1[3],ymm0[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm14 = ymm12[3,0],ymm0[1,0],ymm12[7,4],ymm0[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0],ymm0[2,3],ymm14[6,4],ymm0[6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[1,0],xmm9[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[0,2],xmm9[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2],ymm14[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm15 = ymm0[3,1],ymm12[1,3],ymm0[7,5],ymm12[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm9 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm14 = ymm9[3,0],ymm0[1,0],ymm9[7,4],ymm0[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0],ymm0[2,3],ymm14[6,4],ymm0[6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[1,0],xmm1[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm15 = xmm15[0,2],xmm1[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2],ymm14[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = ymm0[3,1],mem[1,3],ymm0[7,5],mem[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm14 = ymm0[3,0],ymm1[1,0],ymm0[7,4],ymm1[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,0],ymm1[2,3],ymm14[6,4],ymm1[6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} xmm15 = xmm0[1,0],xmm13[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm13 = xmm15[0,2],xmm13[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm14 = ymm0[3,1],ymm15[1,3],ymm0[7,5],ymm15[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm14[6,7]
; AVX-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm13 = ymm14[3,0],ymm11[1,0],ymm14[7,4],ymm11[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm11 = ymm13[2,0],ymm11[2,3],ymm13[6,4],ymm11[6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} xmm13 = xmm0[1,0],xmm10[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm10 = xmm13[0,2],xmm10[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm11 = ymm0[3,1],ymm13[1,3],ymm0[7,5],ymm13[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
; AVX-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm10 = ymm11[3,0],ymm8[1,0],ymm11[7,4],ymm8[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm8 = ymm10[2,0],ymm8[2,3],ymm10[6,4],ymm8[6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} xmm10 = xmm0[1,0],xmm7[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm7 = xmm10[0,2],xmm7[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
; AVX-NEXT:    # ymm8 = ymm0[3,1],mem[1,3],ymm0[7,5],mem[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
; AVX-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm7 = ymm8[3,0],ymm6[1,0],ymm8[7,4],ymm6[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm6 = ymm7[2,0],ymm6[2,3],ymm7[6,4],ymm6[6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} xmm7 = xmm0[1,0],xmm4[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm4 = xmm7[0,2],xmm4[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm6 = ymm0[3,1],ymm7[1,3],ymm0[7,5],ymm7[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7]
; AVX-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm4 = ymm6[3,0],ymm3[1,0],ymm6[7,4],ymm3[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm4[2,0],ymm3[2,3],ymm4[6,4],ymm3[6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} xmm4 = xmm0[1,0],xmm2[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm4[0,2],xmm2[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm0[3,1],ymm4[1,3],ymm0[7,5],ymm4[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,0],ymm0[1,0],ymm3[7,4],ymm0[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm0[2,3],ymm2[6,4],ymm0[6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[1,0],xmm5[3,0]
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm2[0,2],xmm5[1,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm1 # 32-byte Folded Reload
; AVX-NEXT:    # ymm1 = ymm2[3,1],mem[1,3],ymm2[7,5],mem[5,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,0],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
; AVX-NEXT:    # ymm12 = ymm12[0,1,2,3],mem[4,5],ymm12[6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm12[2,3,0,1]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm12[2,0],ymm1[4,4],ymm12[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm9[2,1],mem[2,0],ymm9[6,5],mem[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,0],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm2[2,0],ymm1[4,4],ymm2[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,0],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = ymm15[0,1,2,3],mem[4,5],ymm15[6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm2[2,0],ymm1[4,4],ymm2[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm14[2,1],mem[2,0],ymm14[6,5],mem[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,0],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm10 # 32-byte Folded Reload
; AVX-NEXT:    # ymm10 = ymm13[0,1,2,3],mem[4,5],ymm13[6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm10[2,3,0,1]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm10[2,0],ymm1[4,4],ymm10[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm11[2,1],mem[2,0],ymm11[6,5],mem[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,0],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm2[2,0],ymm1[4,4],ymm2[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm8[2,1],mem[2,0],ymm8[6,5],mem[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm2, %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,0],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm9 # 32-byte Folded Reload
; AVX-NEXT:    # ymm9 = ymm7[0,1,2,3],mem[4,5],ymm7[6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm9[2,3,0,1]
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm9[2,0],ymm1[4,4],ymm9[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
; AVX-NEXT:    # ymm8 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm6[2,1],mem[2,0],ymm6[6,5],mem[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm8, %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm8[2,0],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm6 # 32-byte Folded Reload
; AVX-NEXT:    # ymm6 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm6[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm13[0,0],ymm6[2,0],ymm13[4,4],ymm6[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX-NEXT:    # ymm5 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovaps %ymm3, %ymm4
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm4[2,1],ymm3[2,0],ymm4[6,5],ymm3[6,4]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm2[2,3,0,1]
; AVX-NEXT:    vextractf128 $1, %ymm5, %xmm7
; AVX-NEXT:    vshufps {{.*#+}} xmm15 = xmm5[2,0],xmm7[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2],ymm0[3,4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm15[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm11 = ymm2[0,0],ymm15[2,0],ymm2[4,4],ymm15[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm11[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm12[3,1],ymm0[4,5],ymm12[7,5]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm11 # 16-byte Folded Reload
; AVX-NEXT:    # xmm11 = xmm1[3,1],mem[3,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm12 = ymm1[3,1],ymm14[2,1],ymm1[7,5],ymm14[6,5]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4],ymm0[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[0,1],mem[3,1],ymm0[4,5],mem[7,5]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm11 # 16-byte Folded Reload
; AVX-NEXT:    # xmm11 = xmm1[3,1],mem[3,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX-NEXT:    vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
; AVX-NEXT:    # ymm12 = ymm12[3,1],mem[2,1],ymm12[7,5],mem[6,5]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4],ymm0[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
; AVX-NEXT:    # ymm1 = ymm0[0,1],mem[3,1],ymm0[4,5],mem[7,5]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm11 # 16-byte Folded Reload
; AVX-NEXT:    # xmm11 = xmm0[3,1],mem[3,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX-NEXT:    # ymm12 = ymm0[3,1],mem[2,1],ymm0[7,5],mem[6,5]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm10[3,1],ymm0[4,5],ymm10[7,5]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm10 # 16-byte Folded Reload
; AVX-NEXT:    # xmm10 = xmm1[3,1],mem[3,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX-NEXT:    vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
; AVX-NEXT:    # ymm11 = ymm11[3,1],mem[2,1],ymm11[7,5],mem[6,5]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm11[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3,4],ymm0[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm2[0,1],ymm15[3,1],ymm2[4,5],ymm15[7,5]
; AVX-NEXT:    vshufps {{.*#+}} xmm1 = xmm5[3,1],xmm7[3,3]
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm4[3,1],ymm3[2,1],ymm4[7,5],ymm3[6,5]
; AVX-NEXT:    vmovaps %ymm4, %ymm15
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm13[0,1],ymm6[3,1],ymm13[4,5],ymm6[7,5]
; AVX-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm1 # 16-byte Folded Reload
; AVX-NEXT:    # xmm1 = xmm8[3,1],mem[3,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm3[3,1],ymm6[2,1],ymm3[7,5],ymm6[6,5]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm9[3,1],ymm0[4,5],ymm9[7,5]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX-NEXT:    # xmm1 = xmm1[3,1],mem[3,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm4[3,1],ymm8[2,1],ymm4[7,5],ymm8[6,5]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vshufps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[0,1],mem[3,1],ymm0[4,5],mem[7,5]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX-NEXT:    # xmm1 = xmm1[3,1],mem[3,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm5[3,1],ymm7[2,1],ymm5[7,5],ymm7[6,5]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovaps 32(%rdi), %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX-NEXT:    vmovaps 16(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX-NEXT:    vmovapd 80(%rdi), %xmm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm7[0],ymm1[2],ymm7[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,0],ymm5[4,5],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm2[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm10[0,0],ymm2[6,4],ymm10[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm10[0,2],ymm1[2,0],ymm10[4,6],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 224(%rdi), %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX-NEXT:    vmovaps 208(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX-NEXT:    vmovapd 272(%rdi), %xmm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm14[0],ymm1[2],ymm14[2]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,0],ymm5[4,5],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm2[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm7[0,0],ymm2[6,4],ymm7[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm7[0,2],ymm1[2,0],ymm7[4,6],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX-NEXT:    # ymm2 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 416(%rdi), %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX-NEXT:    vmovaps 400(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX-NEXT:    vmovapd 464(%rdi), %xmm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm8[0],ymm1[2],ymm8[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,0],ymm4[4,5],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm2[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm5[0,0],ymm2[6,4],ymm5[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm5[0,2],ymm1[2,0],ymm5[4,6],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX-NEXT:    # ymm12 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovaps 608(%rdi), %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX-NEXT:    vmovaps 592(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX-NEXT:    vmovapd 656(%rdi), %xmm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX-NEXT:    # ymm1 = ymm1[1],mem[0],ymm1[2],mem[2]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,0],ymm2[4,5],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm12[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm12[2,0],ymm4[0,0],ymm12[6,4],ymm4[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm4[0,2],ymm1[2,0],ymm4[4,6],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
; AVX-NEXT:    # ymm11 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovaps 800(%rdi), %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX-NEXT:    vmovaps 784(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX-NEXT:    vmovapd 848(%rdi), %xmm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm6[0],ymm1[2],ymm6[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,0],ymm3[4,5],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm11[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm11[2,0],ymm3[0,0],ymm11[6,4],ymm3[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm1[2,0],ymm3[4,6],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX-NEXT:    # ymm14 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovaps 992(%rdi), %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX-NEXT:    vmovaps 976(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX-NEXT:    vmovapd 1040(%rdi), %xmm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX-NEXT:    # ymm1 = ymm1[1],mem[0],ymm1[2],mem[2]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,0],ymm2[4,5],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm14[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm14[2,0],ymm2[0,0],ymm14[6,4],ymm2[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm2[0,2],ymm1[2,0],ymm2[4,6],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
; AVX-NEXT:    # ymm8 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovaps 1184(%rdi), %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX-NEXT:    vmovaps 1168(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX-NEXT:    vmovapd 1232(%rdi), %xmm1
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX-NEXT:    # ymm1 = ymm1[1],mem[0],ymm1[2],mem[2]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm15[0,1],ymm1[2,0],ymm15[4,5],ymm1[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm6 = ymm8[2,0],ymm1[0,0],ymm8[6,4],ymm1[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm6 = ymm1[0,2],ymm6[2,0],ymm1[4,6],ymm6[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm6[5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX-NEXT:    # ymm6 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovaps 1376(%rdi), %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX-NEXT:    vmovaps 1360(%rdi), %xmm9
; AVX-NEXT:    vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm9[0],xmm0[1],xmm9[2,3]
; AVX-NEXT:    vmovapd 1424(%rdi), %xmm9
; AVX-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = ymm9[1],mem[0],ymm9[2],mem[2]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm15 = ymm9[0,1],ymm15[2,0],ymm9[4,5],ymm15[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm15 = ymm0[0,1],ymm15[2,3,4,5,6,7]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm6[2,3,0,1]
; AVX-NEXT:    vshufps {{.*#+}} ymm13 = ymm6[2,0],ymm0[0,0],ymm6[6,4],ymm0[4,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm13 = ymm0[0,2],ymm13[2,0],ymm0[4,6],ymm13[6,4]
; AVX-NEXT:    vblendps {{.*#+}} ymm9 = ymm15[0,1,2,3,4],ymm13[5,6,7]
; AVX-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm13 = ymm9[3,0],ymm10[1,0],ymm9[7,4],ymm10[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm9 = ymm10[0,3],ymm13[2,0],ymm10[4,7],ymm13[6,4]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm13 # 16-byte Folded Reload
; AVX-NEXT:    # xmm13 = xmm10[0,1],mem[2,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = ymm10[3,1],mem[1,3],ymm10[7,5],mem[5,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm15 = ymm10[1,1],ymm15[2,0],ymm10[5,5],ymm15[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm9 = ymm13[0,1,2,3,4],ymm9[5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm13 = ymm10[3,0],ymm7[1,0],ymm10[7,4],ymm7[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[0,3],ymm13[2,0],ymm7[4,7],ymm13[6,4]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm13 # 16-byte Folded Reload
; AVX-NEXT:    # xmm13 = xmm10[0,1],mem[2,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = ymm10[3,1],mem[1,3],ymm10[7,5],mem[5,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm15 = ymm10[1,1],ymm15[2,0],ymm10[5,5],ymm15[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm7 = ymm13[0,1,2,3,4],ymm7[5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm13 = ymm10[3,0],ymm5[1,0],ymm10[7,4],ymm5[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[0,3],ymm13[2,0],ymm5[4,7],ymm13[6,4]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm13 # 16-byte Folded Reload
; AVX-NEXT:    # xmm13 = xmm10[0,1],mem[2,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = ymm10[3,1],mem[1,3],ymm10[7,5],mem[5,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm15 = ymm10[1,1],ymm15[2,0],ymm10[5,5],ymm15[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm13 = xmm13[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm5 = ymm13[0,1,2,3,4],ymm5[5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[3,0],ymm4[1,0],ymm12[7,4],ymm4[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[0,3],ymm12[2,0],ymm4[4,7],ymm12[6,4]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm12 # 16-byte Folded Reload
; AVX-NEXT:    # xmm12 = xmm10[0,1],mem[2,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm13 # 32-byte Folded Reload
; AVX-NEXT:    # ymm13 = ymm10[3,1],mem[1,3],ymm10[7,5],mem[5,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm13 = ymm10[1,1],ymm13[2,0],ymm10[5,5],ymm13[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm12 = xmm12[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm13[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3,4],ymm4[5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[3,0],ymm3[1,0],ymm11[7,4],ymm3[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[0,3],ymm11[2,0],ymm3[4,7],ymm11[6,4]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm11 # 16-byte Folded Reload
; AVX-NEXT:    # xmm11 = xmm10[0,1],mem[2,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm12 # 32-byte Folded Reload
; AVX-NEXT:    # ymm12 = ymm10[3,1],mem[1,3],ymm10[7,5],mem[5,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm12 = ymm10[1,1],ymm12[2,0],ymm10[5,5],ymm12[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm11 = xmm11[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm12[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3,4],ymm3[5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm10 = ymm14[3,0],ymm2[1,0],ymm14[7,4],ymm2[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,3],ymm10[2,0],ymm2[4,7],ymm10[6,4]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
; AVX-NEXT:    # xmm10 = xmm10[0,1],mem[2,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
; AVX-NEXT:    # ymm11 = ymm11[3,1],mem[1,3],ymm11[7,5],mem[5,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm11 = ymm12[1,1],ymm11[2,0],ymm12[5,5],ymm11[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm10 = xmm10[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm11[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3,4],ymm2[5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[3,0],ymm1[1,0],ymm8[7,4],ymm1[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,3],ymm8[2,0],ymm1[4,7],ymm8[6,4]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
; AVX-NEXT:    # xmm8 = xmm8[0,1],mem[2,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX-NEXT:    # ymm10 = ymm10[3,1],mem[1,3],ymm10[7,5],mem[5,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm10 = ymm11[1,1],ymm10[2,0],ymm11[5,5],ymm10[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm8 = xmm8[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm10[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3,4],ymm1[5,6,7]
; AVX-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[3,0],ymm0[1,0],ymm6[7,4],ymm0[5,4]
; AVX-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm6[2,0],ymm0[4,7],ymm6[6,4]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
; AVX-NEXT:    # xmm6 = xmm6[0,1],mem[2,3]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX-NEXT:    vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX-NEXT:    # ymm8 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX-NEXT:    vshufps {{.*#+}} ymm8 = ymm10[1,1],ymm8[2,0],ymm10[5,5],ymm8[6,4]
; AVX-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[1,3,2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm8[2,3,4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm0[5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 192(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 128(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 64(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, (%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 224(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 160(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 96(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 32(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 192(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 128(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 64(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, (%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 224(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 160(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 96(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 32(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 192(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 128(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 64(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, (%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 224(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 160(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 96(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 32(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, (%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 64(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 128(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 192(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 224(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 160(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 96(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 32(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 224(%r9)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 192(%r9)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 160(%r9)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 128(%r9)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 96(%r9)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 64(%r9)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, 32(%r9)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm6, (%r9)
; AVX-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX-NEXT:    vmovaps %ymm0, 224(%rax)
; AVX-NEXT:    vmovaps %ymm1, 192(%rax)
; AVX-NEXT:    vmovaps %ymm2, 160(%rax)
; AVX-NEXT:    vmovaps %ymm3, 128(%rax)
; AVX-NEXT:    vmovaps %ymm4, 96(%rax)
; AVX-NEXT:    vmovaps %ymm5, 64(%rax)
; AVX-NEXT:    vmovaps %ymm7, 32(%rax)
; AVX-NEXT:    vmovaps %ymm9, (%rax)
; AVX-NEXT:    addq $2584, %rsp # imm = 0xA18
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i32_stride6_vf64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    subq $2568, %rsp # imm = 0xA08
; AVX2-NEXT:    vmovaps 672(%rdi), %ymm4
; AVX2-NEXT:    vmovaps 640(%rdi), %ymm5
; AVX2-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 608(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 320(%rdi), %ymm6
; AVX2-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 352(%rdi), %ymm7
; AVX2-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 288(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 256(%rdi), %ymm8
; AVX2-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 224(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 192(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps {{.*#+}} xmm9 = [0,6,4,u]
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-NEXT:    vpermps %ymm14, %ymm9, %ymm0
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[0,1],ymm2[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm15[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm6 = [4,2,4,2,4,2,4,2]
; AVX2-NEXT:    vpermps %ymm1, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm5[0,1],ymm4[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],ymm4[6,7]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 576(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm0, %ymm9, %ymm0
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-NEXT:    vmovaps 704(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 736(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm1, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1056(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1024(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 992(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 960(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm0, %ymm9, %ymm0
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-NEXT:    vmovaps 1088(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1120(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm1, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1440(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1408(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1376(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1344(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm0, %ymm9, %ymm0
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-NEXT:    vmovaps 1472(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1504(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm1, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 96(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-NEXT:    vmovaps 64(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps (%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-NEXT:    vpermps %ymm13, %ymm9, %ymm0
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-NEXT:    vmovaps 128(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 160(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm12 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm12, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 480(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 448(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovaps 416(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 384(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-NEXT:    vpermps %ymm10, %ymm9, %ymm0
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm11[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-NEXT:    vmovaps 512(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 544(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm8, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 864(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 832(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovaps 800(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 768(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-NEXT:    vpermps %ymm4, %ymm9, %ymm0
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-NEXT:    vmovaps 896(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 928(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm3, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1184(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1152(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-NEXT:    vpermps %ymm2, %ymm9, %ymm0
; AVX2-NEXT:    vmovaps 1248(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1216(%rdi), %ymm5
; AVX2-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[0,1],ymm1[0,1]
; AVX2-NEXT:    vblendps {{.*#+}} ymm9 = ymm5[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm5 = ymm9[0,2,2,2,4,6,6,6]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6,7]
; AVX2-NEXT:    vmovaps 1280(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1312(%rdi), %ymm5
; AVX2-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm5, %ymm6, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps {{.*#+}} xmm0 = [1,7,5,u]
; AVX2-NEXT:    vpermps %ymm14, %ymm0, %ymm1
; AVX2-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm15 = ymm1[0,1,2],ymm15[3,4,5,6,7]
; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm1 = [5,3,5,3,5,3,5,3]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm13, %ymm0, %ymm13
; AVX2-NEXT:    vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = mem[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm12, %ymm1, %ymm12
; AVX2-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
; AVX2-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm10, %ymm0, %ymm10
; AVX2-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm8, %ymm1, %ymm8
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1,2,3,4,5],ymm8[6,7]
; AVX2-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm4, %ymm0, %ymm4
; AVX2-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm7[3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm3, %ymm1, %ymm3
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm2, %ymm0, %ymm0
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[1,3,2,3,5,7,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm5, %ymm1, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1],ymm11[2,3],ymm12[4,5],ymm11[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1],ymm7[2,3],ymm9[4,5],ymm7[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm0[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm5[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm13[0,1],mem[2,3],ymm13[4,5],mem[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm5 = ymm14[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm15 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm5 = ymm15[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm5[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm8[0,1],ymm10[2,3],ymm8[4,5],ymm10[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm5 = ymm2[0,2,0,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm5 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm5[0,0,2,0,4,4,6,4]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vpermilps $247, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = mem[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vpermilps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = mem[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm7[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm9[1],ymm0[2,3,4],ymm9[5],ymm0[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vpermilps $247, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = mem[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vpermilps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = mem[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm4[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm9[1],ymm0[2,3,4],ymm9[5],ymm0[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vpermilps $247, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = mem[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vpermilps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = mem[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm7[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vpermilps $247, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = mem[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vpermilps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = mem[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm10[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1],ymm0[2,3,4],ymm8[5],ymm0[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm5[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vmovaps %ymm13, %ymm5
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1],ymm0[2,3,4],ymm13[5],ymm0[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm14[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm15[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm13[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vpermilps $247, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = mem[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vpermilps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = mem[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups (%rsp), %ymm14 # 32-byte Reload
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm14[3,3,3,3,7,7,7,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4],ymm3[5],ymm0[6,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-NEXT:    vpermilps $247, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = mem[3,1,3,3,7,5,7,7]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-NEXT:    vpermilps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = mem[0,1,3,1,4,5,7,5]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm14[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 80(%rdi), %xmm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vpermps %ymm3, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = [0,2,0,6,0,2,0,6]
; AVX2-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX2-NEXT:    vpermps %ymm2, %ymm3, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 272(%rdi), %xmm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vpermps %ymm14, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm2, %ymm3, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm12[0,1,2,3],ymm13[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm12 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 464(%rdi), %xmm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vpermps %ymm12, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm2, %ymm3, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm12 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 656(%rdi), %xmm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vpermps %ymm12, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm2, %ymm3, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm11[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm5 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 848(%rdi), %xmm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vpermps %ymm5, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermps %ymm2, %ymm3, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm15 = ymm9[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovaps 1040(%rdi), %xmm13
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm13[2,3],ymm15[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vpermps %ymm14, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm12 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm12 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm12, %ymm3, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm11 = ymm8[0,1,2,3],ymm10[4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm10 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm10 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovaps 1232(%rdi), %xmm9
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1],ymm9[2,3],ymm11[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vpermps %ymm10, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm8 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm8 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm8, %ymm3, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm4[0,1,2,3],ymm7[4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm5 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-NEXT:    vmovaps 1424(%rdi), %xmm4
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm4[2,3],ymm7[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-NEXT:    vpermps %ymm5, %ymm6, %ymm2
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm2, %ymm3, %ymm3
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = mem[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm0[2,3,4,5,6,7]
; AVX2-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
; AVX2-NEXT:    # ymm0 = mem[0,1,0,1]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm6[5,6,7]
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermilps $85, (%rsp), %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = mem[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3,4,5,6,7]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm6[5,6,7]
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = mem[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3,4,5,6,7]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm6[5,6,7]
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = mem[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3,4,5,6,7]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm6[5,6,7]
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm6 = mem[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm6 = ymm6[0,1,2],mem[3],ymm6[4,5,6,7]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3,4,5,6,7]
; AVX2-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX2-NEXT:    vblendps {{.*#+}} ymm6 = ymm3[0,1,2,3,4],ymm6[5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm15[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3],ymm3[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm14, %ymm1, %ymm13
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm13[0,1],ymm3[2,3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm12, %ymm0, %ymm12
; AVX2-NEXT:    vblendps {{.*#+}} ymm12 = ymm3[0,1,2,3,4],ymm12[5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm11[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm9[3],ymm3[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm10, %ymm1, %ymm9
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm9[0,1],ymm3[2,3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm8, %ymm0, %ymm8
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm3[0,1,2,3,4],ymm8[5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm7[1,1,1,1,5,5,5,5]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3],ymm3[4,5,6,7]
; AVX2-NEXT:    vpermps %ymm5, %ymm1, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
; AVX2-NEXT:    vpermps %ymm2, %ymm0, %ymm0
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 192(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 128(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 64(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, (%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 224(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 160(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 96(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 32(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 192(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 128(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 64(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, (%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 224(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 160(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 96(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 32(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 192(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 128(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 64(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, (%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 224(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 160(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 96(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 32(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, (%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 64(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 128(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 192(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 224(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 160(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 96(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 32(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 224(%r9)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 192(%r9)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 160(%r9)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 128(%r9)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 96(%r9)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 64(%r9)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 32(%r9)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, (%r9)
; AVX2-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-NEXT:    vmovaps %ymm0, 224(%rax)
; AVX2-NEXT:    vmovaps %ymm8, 192(%rax)
; AVX2-NEXT:    vmovaps %ymm12, 160(%rax)
; AVX2-NEXT:    vmovaps %ymm6, 128(%rax)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 96(%rax)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 64(%rax)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 32(%rax)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, (%rax)
; AVX2-NEXT:    addq $2568, %rsp # imm = 0xA08
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i32_stride6_vf64:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    subq $2568, %rsp # imm = 0xA08
; AVX2-FP-NEXT:    vmovaps 672(%rdi), %ymm4
; AVX2-FP-NEXT:    vmovaps 640(%rdi), %ymm5
; AVX2-FP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 608(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 320(%rdi), %ymm6
; AVX2-FP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 352(%rdi), %ymm7
; AVX2-FP-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 288(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 256(%rdi), %ymm8
; AVX2-FP-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 224(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 192(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps {{.*#+}} xmm9 = [0,6,4,u]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vpermps %ymm14, %ymm9, %ymm0
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[0,1],ymm2[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm15[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vbroadcastsd {{.*#+}} ymm6 = [4,2,4,2,4,2,4,2]
; AVX2-FP-NEXT:    vpermps %ymm1, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm5[0,1],ymm4[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 576(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm0, %ymm9, %ymm0
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 704(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 736(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm1, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1056(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1024(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 992(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 960(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm0, %ymm9, %ymm0
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 1088(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1120(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm1, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1440(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1408(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1376(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1344(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm0, %ymm9, %ymm0
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 1472(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1504(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm1, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 96(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 64(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vpermps %ymm13, %ymm9, %ymm0
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 160(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm12 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm12, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 480(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 448(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovaps 416(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 384(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vpermps %ymm10, %ymm9, %ymm0
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm11[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 512(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 544(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm8, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 864(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 832(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovaps 800(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 768(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vpermps %ymm4, %ymm9, %ymm0
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 896(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 928(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm3, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1184(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1152(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm9, %ymm0
; AVX2-FP-NEXT:    vmovaps 1248(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1216(%rdi), %ymm5
; AVX2-FP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[0,1],ymm1[0,1]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm9 = ymm5[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm5 = ymm9[0,2,2,2,4,6,6,6]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 1280(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1312(%rdi), %ymm5
; AVX2-FP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm5, %ymm6, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps {{.*#+}} xmm0 = [1,7,5,u]
; AVX2-FP-NEXT:    vpermps %ymm14, %ymm0, %ymm1
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm15 = ymm15[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm15 = ymm1[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FP-NEXT:    vbroadcastsd {{.*#+}} ymm1 = [5,3,5,3,5,3,5,3]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm13, %ymm0, %ymm13
; AVX2-FP-NEXT:    vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = mem[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm12, %ymm1, %ymm12
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
; AVX2-FP-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm10, %ymm0, %ymm10
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm11 = ymm11[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm8, %ymm1, %ymm8
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FP-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm4, %ymm0, %ymm4
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm3, %ymm1, %ymm3
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm0, %ymm0
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm9[1,3,2,3,5,7,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm5, %ymm1, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1],ymm11[2,3],ymm12[4,5],ymm11[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1],ymm7[2,3],ymm9[4,5],ymm7[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm5 = ymm0[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm13[0,1],mem[2,3],ymm13[4,5],mem[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm5 = ymm14[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm15 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm5 = ymm15[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm8[0,1],ymm10[2,3],ymm8[4,5],ymm10[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm5 = ymm2[0,2,0,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm3 = ymm2[2,0,2,3,6,4,6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm5 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm5[0,0,2,0,4,4,6,4]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vpermilps $247, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = mem[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermilps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = mem[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm7[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm9[1],ymm0[2,3,4],ymm9[5],ymm0[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vpermilps $247, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = mem[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermilps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = mem[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm4[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm9[1],ymm0[2,3,4],ymm9[5],ymm0[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vpermilps $247, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = mem[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermilps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = mem[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm7[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vpermilps $247, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = mem[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermilps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = mem[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm10[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1],ymm0[2,3,4],ymm8[5],ymm0[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm5[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm11[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vmovaps %ymm13, %ymm5
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm13[1],ymm0[2,3,4],ymm13[5],ymm0[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm14[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm15[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm13[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vpermilps $247, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = mem[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermilps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = mem[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups (%rsp), %ymm14 # 32-byte Reload
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm14[3,3,3,3,7,7,7,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3,4],ymm3[5],ymm0[6,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,3]
; AVX2-FP-NEXT:    vpermilps $247, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = mem[3,1,3,3,7,5,7,7]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FP-NEXT:    vpermilps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = mem[0,1,3,1,4,5,7,5]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm14[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 80(%rdi), %xmm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vpermps %ymm3, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vbroadcastf128 {{.*#+}} ymm3 = [0,2,0,6,0,2,0,6]
; AVX2-FP-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm3, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 272(%rdi), %xmm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vpermps %ymm14, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm3, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm12[0,1,2,3],ymm13[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm12 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 464(%rdi), %xmm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vpermps %ymm12, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm3, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm12 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 656(%rdi), %xmm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vpermps %ymm12, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm3, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm11[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm5 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 848(%rdi), %xmm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vpermps %ymm5, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm3, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm15 = ymm9[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 1040(%rdi), %xmm13
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm13[2,3],ymm15[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vpermps %ymm14, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm12 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm12 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm12, %ymm3, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm11 = ymm8[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm10 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm10 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 1232(%rdi), %xmm9
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1],ymm9[2,3],ymm11[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vpermps %ymm10, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm8 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm8 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm8, %ymm3, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm4[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm5 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 1424(%rdi), %xmm4
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm4[2,3],ymm7[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FP-NEXT:    vpermps %ymm5, %ymm6, %ymm2
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm3, %ymm3
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = mem[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
; AVX2-FP-NEXT:    # ymm0 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermilps $85, (%rsp), %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = mem[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = mem[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = mem[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm6 = mem[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm6 = ymm6[0,1,2],mem[3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm6 = ymm3[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm3 = ymm15[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm14, %ymm1, %ymm13
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm13[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm12, %ymm0, %ymm12
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm12 = ymm3[0,1,2,3,4],ymm12[5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm3 = ymm11[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm9[3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm10, %ymm1, %ymm9
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm9[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm8, %ymm0, %ymm8
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm3[0,1,2,3,4],ymm8[5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm3 = ymm7[1,1,1,1,5,5,5,5]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm5, %ymm1, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FP-NEXT:    vpermps %ymm2, %ymm0, %ymm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 192(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 128(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 64(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, (%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 224(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 160(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 96(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 32(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 192(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 128(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 64(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, (%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 224(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 160(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 96(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 32(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 192(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 128(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 64(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, (%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 224(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 160(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 96(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 32(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, (%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 64(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 128(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 192(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 224(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 160(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 96(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 32(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 224(%r9)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 192(%r9)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 160(%r9)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 128(%r9)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 96(%r9)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 64(%r9)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 32(%r9)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, (%r9)
; AVX2-FP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FP-NEXT:    vmovaps %ymm0, 224(%rax)
; AVX2-FP-NEXT:    vmovaps %ymm8, 192(%rax)
; AVX2-FP-NEXT:    vmovaps %ymm12, 160(%rax)
; AVX2-FP-NEXT:    vmovaps %ymm6, 128(%rax)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 96(%rax)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 64(%rax)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 32(%rax)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, (%rax)
; AVX2-FP-NEXT:    addq $2568, %rsp # imm = 0xA08
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i32_stride6_vf64:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    subq $2536, %rsp # imm = 0x9E8
; AVX2-FCP-NEXT:    vmovaps 672(%rdi), %ymm5
; AVX2-FCP-NEXT:    vmovaps 640(%rdi), %ymm6
; AVX2-FCP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 608(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 320(%rdi), %ymm4
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 352(%rdi), %ymm7
; AVX2-FCP-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 288(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 256(%rdi), %ymm8
; AVX2-FCP-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 224(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 192(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm12 = [0,6,4,u]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm15, %ymm12, %ymm0
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm8[0,1],ymm2[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vbroadcastsd {{.*#+}} ymm1 = [4,2,4,2,4,2,4,2]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm1, %ymm2
; AVX2-FCP-NEXT:    vmovaps %ymm1, %ymm4
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm6[0,1],ymm5[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 576(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm12, %ymm0
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm2 = ymm1[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 704(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 736(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm1, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1056(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1024(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 992(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 960(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm12, %ymm0
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 1088(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1120(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm1, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1440(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1408(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1376(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1344(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm12, %ymm0
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 1472(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1504(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm1, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 96(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 64(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm13, %ymm12, %ymm0
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 160(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm11, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 480(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 448(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovaps 416(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 384(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm9, %ymm12, %ymm0
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm2 = ymm10[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 512(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 544(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm8, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 864(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 832(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovaps 800(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 768(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm6, %ymm12, %ymm0
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm2 = ymm7[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 896(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 928(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm5, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1184(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1152(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm3, %ymm12, %ymm0
; AVX2-FCP-NEXT:    vmovaps 1248(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1216(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[0,1],ymm1[0,1]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm12 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm2 = ymm12[0,2,2,2,4,6,6,6]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 1280(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1312(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm4, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm0 = [1,7,5,u]
; AVX2-FCP-NEXT:    vpermps %ymm15, %ymm0, %ymm1
; AVX2-FCP-NEXT:    vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm15 = ymm1[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastsd {{.*#+}} ymm1 = [5,3,5,3,5,3,5,3]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm13, %ymm0, %ymm13
; AVX2-FCP-NEXT:    vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm14 = mem[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm14[3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm11, %ymm1, %ymm11
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm13[0,1,2,3,4,5],ymm11[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm9, %ymm0, %ymm9
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm8, %ymm1, %ymm8
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm9[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm6, %ymm0, %ymm6
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm5, %ymm1, %ymm5
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm3, %ymm0, %ymm0
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm3 = ymm12[1,3,2,3,5,7,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm1, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm7[0,1],mem[2,3],ymm7[4,5],mem[6,7]
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} ymm13 = [2,0,6,4,2,0,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm13, %ymm0
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} xmm12 = [2,0,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm12, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm15 = [0,0,6,4,0,0,6,4]
; AVX2-FCP-NEXT:    # ymm15 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm15, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm6[0,1],mem[2,3],ymm6[4,5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm13, %ymm0
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm12, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm15, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm13, %ymm0
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm12, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm8 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm8 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm8, %ymm15, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm4[0,1],mem[2,3],ymm4[4,5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm13, %ymm0
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm14 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm14, %ymm12, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm11 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm11 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm11, %ymm15, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm13, %ymm0
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, (%rsp), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm12, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm15, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm13, %ymm0
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm12, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm15, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm10[0,1],mem[2,3],ymm10[4,5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm13, %ymm0
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm9 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm9 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm9, %ymm12, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm15, %ymm5
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm3[0,1],mem[2,3],ymm3[4,5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm13, %ymm0
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm13 = ymm5[0,1],mem[2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm13, %ymm12, %ymm5
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm15, %ymm12
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm12[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm5 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2,3,4],ymm7[5],ymm5[6,7]
; AVX2-FCP-NEXT:    vblendps $222, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm12 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm12 = ymm5[0],mem[1,2,3,4],ymm5[5],mem[6,7]
; AVX2-FCP-NEXT:    vmovaps {{.*#+}} ymm5 = [3,1,7,5,0,u,u,u]
; AVX2-FCP-NEXT:    vpermps %ymm12, %ymm5, %ymm15
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm12 = [0,1,7,5,0,1,7,5]
; AVX2-FCP-NEXT:    # ymm12 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm7 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm15[0,1,2,3,4],ymm7[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm7 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm6[1],ymm7[2,3,4],ymm6[5],ymm7[6,7]
; AVX2-FCP-NEXT:    vblendps $222, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm7 = ymm7[0],mem[1,2,3,4],ymm7[5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm7, %ymm5, %ymm7
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm15 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm15[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm7 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm15[1],ymm7[2,3,4],ymm15[5],ymm7[6,7]
; AVX2-FCP-NEXT:    vblendps $222, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm7 = ymm7[0],mem[1,2,3,4],ymm7[5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm7, %ymm5, %ymm7
; AVX2-FCP-NEXT:    vpermps %ymm8, %ymm12, %ymm8
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm8[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm7 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm4[1],ymm7[2,3,4],ymm4[5],ymm7[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0],ymm14[1,2,3,4],ymm7[5],ymm14[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm7, %ymm5, %ymm7
; AVX2-FCP-NEXT:    vpermps %ymm11, %ymm12, %ymm6
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm6 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm3[1],ymm6[2,3,4],ymm3[5],ymm6[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0],ymm13[1,2,3,4],ymm6[5],ymm13[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm6, %ymm5, %ymm6
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm12, %ymm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm10[1],ymm0[2,3,4],ymm10[5],ymm0[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm9[1,2,3,4],ymm0[5],ymm9[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm5, %ymm0
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm12, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2,3,4],ymm7[5],ymm0[6,7]
; AVX2-FCP-NEXT:    vblendps $222, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[0],mem[1,2,3,4],ymm0[5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm0, %ymm5, %ymm0
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vpermilps {{.*#+}} xmm2 = mem[3,3,3,3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2,3,4],ymm4[5],ymm2[6,7]
; AVX2-FCP-NEXT:    vblendps $222, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm2[0],mem[1,2,3,4],ymm2[5],mem[6,7]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm5, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm4[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $15, (%rsp), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm5 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 80(%rdi), %xmm0
; AVX2-FCP-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FCP-NEXT:    vpermps %ymm5, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm5 = [0,2,0,6,0,2,0,6]
; AVX2-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm5, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm8 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 272(%rdi), %xmm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vpermps %ymm8, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm5, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm7[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm7 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 464(%rdi), %xmm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vpermps %ymm7, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm5, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm7 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 656(%rdi), %xmm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vpermps %ymm7, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm5, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = ymm10[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm6 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 848(%rdi), %xmm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vpermps %ymm6, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm5, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm15 = ymm15[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm14 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 1040(%rdi), %xmm13
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm13[2,3],ymm15[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vpermps %ymm14, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm12 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm12 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm12, %ymm5, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm11 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm11 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm10 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm10 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 1232(%rdi), %xmm9
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1],ymm9[2,3],ymm11[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vpermps %ymm10, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm8 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm8 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm8, %ymm5, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm7 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm6 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 1424(%rdi), %xmm3
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm3[2,3],ymm7[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
; AVX2-FCP-NEXT:    vpermps %ymm6, %ymm4, %ymm2
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm5, %ymm4
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = mem[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps $8, (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
; AVX2-FCP-NEXT:    # ymm0 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm4 = mem[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm4 = ymm4[0,1,2],mem[3],ymm4[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm4 = mem[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm4 = ymm4[0,1,2],mem[3],ymm4[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm4 = mem[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm4 = ymm4[0,1,2],mem[3],ymm4[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm4 = mem[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm4 = ymm4[0,1,2],mem[3],ymm4[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm5 = ymm15[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm13[3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm14, %ymm1, %ymm13
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm13[0,1],ymm5[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm12, %ymm0, %ymm12
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm12 = ymm5[0,1,2,3,4],ymm12[5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm5 = ymm11[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm9[3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm10, %ymm1, %ymm9
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm9[0,1],ymm5[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm8, %ymm0, %ymm8
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm5[0,1,2,3,4],ymm8[5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm5 = ymm7[1,1,1,1,5,5,5,5]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm6, %ymm1, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vpermps %ymm2, %ymm0, %ymm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 192(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 128(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 64(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, (%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 224(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 160(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 96(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 32(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 192(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 128(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 64(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, (%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 224(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 160(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 96(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 32(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 192(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 128(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 64(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, (%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 224(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 160(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 96(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 32(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, (%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 64(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 128(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 192(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 224(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 160(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 96(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 32(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 224(%r9)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 192(%r9)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 160(%r9)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 128(%r9)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 96(%r9)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 64(%r9)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 32(%r9)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, (%r9)
; AVX2-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX2-FCP-NEXT:    vmovaps %ymm0, 224(%rax)
; AVX2-FCP-NEXT:    vmovaps %ymm8, 192(%rax)
; AVX2-FCP-NEXT:    vmovaps %ymm12, 160(%rax)
; AVX2-FCP-NEXT:    vmovaps %ymm4, 128(%rax)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 96(%rax)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 64(%rax)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 32(%rax)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, (%rax)
; AVX2-FCP-NEXT:    addq $2536, %rsp # imm = 0x9E8
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i32_stride6_vf64:
; AVX512:       # %bb.0:
; AVX512-NEXT:    subq $2632, %rsp # imm = 0xA48
; AVX512-NEXT:    vmovdqa64 1472(%rdi), %zmm21
; AVX512-NEXT:    vmovdqa64 1408(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 1088(%rdi), %zmm20
; AVX512-NEXT:    vmovdqa64 1024(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512-NEXT:    vmovdqa64 576(%rdi), %zmm18
; AVX512-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512-NEXT:    vmovdqa64 640(%rdi), %zmm3
; AVX512-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512-NEXT:    vmovdqa64 256(%rdi), %zmm4
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512-NEXT:    vpermt2d %zmm25, %zmm6, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512-NEXT:    vpermt2d %zmm27, %zmm6, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm18, %zmm7
; AVX512-NEXT:    vpermt2d %zmm2, %zmm5, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512-NEXT:    vpermt2d %zmm20, %zmm6, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermi2d %zmm21, %zmm1, %zmm6
; AVX512-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm8
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm6
; AVX512-NEXT:    vpermt2d %zmm20, %zmm7, %zmm8
; AVX512-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512-NEXT:    vpermt2d %zmm25, %zmm7, %zmm0
; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermt2d %zmm27, %zmm0, %zmm3
; AVX512-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm4, %zmm3
; AVX512-NEXT:    vpermt2d %zmm25, %zmm7, %zmm3
; AVX512-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2d %zmm25, %zmm0, %zmm4
; AVX512-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm6, %zmm3
; AVX512-NEXT:    vpermt2d %zmm20, %zmm7, %zmm3
; AVX512-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2d %zmm21, %zmm0, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm18, %zmm1
; AVX512-NEXT:    vpermt2d %zmm2, %zmm8, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2d %zmm20, %zmm0, %zmm6
; AVX512-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512-NEXT:    vpermt2d %zmm18, %zmm31, %zmm0
; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm25 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512-NEXT:    # zmm25 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512-NEXT:    vpermt2d %zmm18, %zmm25, %zmm0
; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512-NEXT:    vpermt2d %zmm18, %zmm3, %zmm0
; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermt2d %zmm18, %zmm0, %zmm2
; AVX512-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 128(%rdi), %zmm26
; AVX512-NEXT:    vmovdqa64 192(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512-NEXT:    vpermt2d %zmm26, %zmm5, %zmm2
; AVX512-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512-NEXT:    vpermt2d %zmm26, %zmm8, %zmm2
; AVX512-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512-NEXT:    vpermt2d %zmm1, %zmm31, %zmm2
; AVX512-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512-NEXT:    vpermt2d %zmm1, %zmm25, %zmm2
; AVX512-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512-NEXT:    vpermt2d %zmm1, %zmm3, %zmm2
; AVX512-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2d %zmm1, %zmm0, %zmm26
; AVX512-NEXT:    vmovdqa64 896(%rdi), %zmm22
; AVX512-NEXT:    vmovdqa64 960(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512-NEXT:    vpermt2d %zmm22, %zmm5, %zmm2
; AVX512-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 1280(%rdi), %zmm19
; AVX512-NEXT:    vmovdqa64 1344(%rdi), %zmm2
; AVX512-NEXT:    vpermi2d %zmm19, %zmm2, %zmm5
; AVX512-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm1, %zmm4
; AVX512-NEXT:    vpermt2d %zmm22, %zmm8, %zmm4
; AVX512-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermi2d %zmm19, %zmm2, %zmm8
; AVX512-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm22, %zmm28
; AVX512-NEXT:    vpermt2d %zmm1, %zmm31, %zmm28
; AVX512-NEXT:    vpermi2d %zmm2, %zmm19, %zmm31
; AVX512-NEXT:    vmovdqa64 %zmm22, %zmm29
; AVX512-NEXT:    vpermt2d %zmm1, %zmm25, %zmm29
; AVX512-NEXT:    vpermi2d %zmm2, %zmm19, %zmm25
; AVX512-NEXT:    vmovdqa64 %zmm22, %zmm4
; AVX512-NEXT:    vpermt2d %zmm1, %zmm3, %zmm4
; AVX512-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermi2d %zmm2, %zmm19, %zmm3
; AVX512-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2d %zmm2, %zmm0, %zmm19
; AVX512-NEXT:    vpermt2d %zmm1, %zmm0, %zmm22
; AVX512-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512-NEXT:    vmovdqa64 384(%rdi), %zmm0
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [0,6,12,18,24,30,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm20
; AVX512-NEXT:    vpermt2d %zmm2, %zmm3, %zmm20
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [1,7,13,19,25,31,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm21
; AVX512-NEXT:    vpermt2d %zmm2, %zmm4, %zmm21
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm12 = [2,8,14,20,26,0,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512-NEXT:    vpermt2d %zmm2, %zmm12, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [3,9,15,21,27,0,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512-NEXT:    vpermt2d %zmm2, %zmm14, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [20,26,0,6,12,0,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm23
; AVX512-NEXT:    vpermt2d %zmm0, %zmm5, %zmm23
; AVX512-NEXT:    vpmovsxbd {{.*#+}} ymm24 = [21,27,1,7,13,0,0,0]
; AVX512-NEXT:    vpermt2d %zmm0, %zmm24, %zmm2
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm6
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm15
; AVX512-NEXT:    vpermt2d %zmm6, %zmm3, %zmm15
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm16
; AVX512-NEXT:    vpermt2d %zmm6, %zmm4, %zmm16
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm27
; AVX512-NEXT:    vpermt2d %zmm6, %zmm12, %zmm27
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm30
; AVX512-NEXT:    vpermt2d %zmm6, %zmm14, %zmm30
; AVX512-NEXT:    vmovdqa64 %zmm6, %zmm17
; AVX512-NEXT:    vpermt2d %zmm0, %zmm5, %zmm17
; AVX512-NEXT:    vpermt2d %zmm0, %zmm24, %zmm6
; AVX512-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512-NEXT:    vmovdqa64 768(%rdi), %zmm7
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm8
; AVX512-NEXT:    vpermt2d %zmm10, %zmm3, %zmm8
; AVX512-NEXT:    vmovdqa64 1216(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 1152(%rdi), %zmm0
; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512-NEXT:    vpermt2d %zmm10, %zmm4, %zmm9
; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512-NEXT:    vpermt2d %zmm10, %zmm12, %zmm13
; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm12
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512-NEXT:    vpermt2d %zmm10, %zmm14, %zmm18
; AVX512-NEXT:    vpermi2d %zmm1, %zmm0, %zmm14
; AVX512-NEXT:    vmovdqa64 %zmm10, %zmm11
; AVX512-NEXT:    vpermt2d %zmm7, %zmm5, %zmm11
; AVX512-NEXT:    vpermi2d %zmm0, %zmm1, %zmm5
; AVX512-NEXT:    vpermt2d %zmm0, %zmm24, %zmm1
; AVX512-NEXT:    vpermt2d %zmm7, %zmm24, %zmm10
; AVX512-NEXT:    movb $56, %al
; AVX512-NEXT:    kmovw %eax, %k2
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm15 {%k2}
; AVX512-NEXT:    movw $-2048, %ax # imm = 0xF800
; AVX512-NEXT:    kmovw %eax, %k1
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm15 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm20 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm20 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm8 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm3 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm9 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm21 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm16 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k1}
; AVX512-NEXT:    movw $31, %ax
; AVX512-NEXT:    kmovw %eax, %k2
; AVX512-NEXT:    vmovdqa32 %zmm13, %zmm28 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm28 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm27, %zmm7 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm24
; AVX512-NEXT:    vmovdqa32 %zmm12, %zmm31 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm31 {%k1}
; AVX512-NEXT:    vmovdqa32 %zmm18, %zmm29 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm29 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm30, %zmm12 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm12 {%k1}
; AVX512-NEXT:    vmovdqa32 %zmm14, %zmm25 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm25 {%k1}
; AVX512-NEXT:    movw $992, %ax # imm = 0x3E0
; AVX512-NEXT:    kmovw %eax, %k1
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm11 {%k1}
; AVX512-NEXT:    movb $-32, %al
; AVX512-NEXT:    kmovw %eax, %k2
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm23 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm23 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm17 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm17 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm5 {%k2}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa32 %zmm0, %zmm2 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k2}
; AVX512-NEXT:    vmovdqa32 %zmm26, %zmm6 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm6 {%k2}
; AVX512-NEXT:    vmovdqa32 %zmm19, %zmm1 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
; AVX512-NEXT:    vmovdqa32 %zmm22, %zmm10 {%k1}
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm10 {%k2}
; AVX512-NEXT:    vmovdqa64 %zmm3, 192(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm8, 128(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm20, 64(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm15, (%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm4, 192(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm16, (%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm21, 64(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm9, 128(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm31, 192(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm28, 128(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm25, 192(%r8)
; AVX512-NEXT:    vmovdqa64 %zmm12, (%r8)
; AVX512-NEXT:    vmovdqa64 %zmm7, 64(%r8)
; AVX512-NEXT:    vmovdqa64 %zmm29, 128(%r8)
; AVX512-NEXT:    vmovdqa64 %zmm5, 192(%r9)
; AVX512-NEXT:    vmovdqa64 %zmm17, (%r9)
; AVX512-NEXT:    vmovdqa64 %zmm23, 64(%r9)
; AVX512-NEXT:    vmovdqa64 %zmm11, 128(%r9)
; AVX512-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-NEXT:    vmovdqa64 %zmm10, 128(%rax)
; AVX512-NEXT:    vmovdqa64 %zmm1, 192(%rax)
; AVX512-NEXT:    vmovdqa64 %zmm6, (%rax)
; AVX512-NEXT:    vmovdqa64 %zmm2, 64(%rax)
; AVX512-NEXT:    addq $2632, %rsp # imm = 0xA48
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i32_stride6_vf64:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    subq $2632, %rsp # imm = 0xA48
; AVX512-FCP-NEXT:    vmovdqa64 1472(%rdi), %zmm21
; AVX512-FCP-NEXT:    vmovdqa64 1408(%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 1088(%rdi), %zmm20
; AVX512-FCP-NEXT:    vmovdqa64 1024(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm18
; AVX512-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm4
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512-FCP-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512-FCP-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512-FCP-NEXT:    vpermt2d %zmm25, %zmm6, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512-FCP-NEXT:    vpermt2d %zmm27, %zmm6, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm18, %zmm7
; AVX512-FCP-NEXT:    vpermt2d %zmm2, %zmm5, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512-FCP-NEXT:    vpermt2d %zmm20, %zmm6, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermi2d %zmm21, %zmm1, %zmm6
; AVX512-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm8
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm6
; AVX512-FCP-NEXT:    vpermt2d %zmm20, %zmm7, %zmm8
; AVX512-FCP-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512-FCP-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512-FCP-NEXT:    vpermt2d %zmm25, %zmm7, %zmm0
; AVX512-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512-FCP-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512-FCP-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512-FCP-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512-FCP-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512-FCP-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512-FCP-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512-FCP-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermt2d %zmm27, %zmm0, %zmm3
; AVX512-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm4, %zmm3
; AVX512-FCP-NEXT:    vpermt2d %zmm25, %zmm7, %zmm3
; AVX512-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2d %zmm25, %zmm0, %zmm4
; AVX512-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm6, %zmm3
; AVX512-FCP-NEXT:    vpermt2d %zmm20, %zmm7, %zmm3
; AVX512-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2d %zmm21, %zmm0, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm18, %zmm1
; AVX512-FCP-NEXT:    vpermt2d %zmm2, %zmm8, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2d %zmm20, %zmm0, %zmm6
; AVX512-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512-FCP-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512-FCP-NEXT:    vpermt2d %zmm18, %zmm31, %zmm0
; AVX512-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm25 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512-FCP-NEXT:    # zmm25 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512-FCP-NEXT:    vpermt2d %zmm18, %zmm25, %zmm0
; AVX512-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512-FCP-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512-FCP-NEXT:    vpermt2d %zmm18, %zmm3, %zmm0
; AVX512-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermt2d %zmm18, %zmm0, %zmm2
; AVX512-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm26
; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512-FCP-NEXT:    vpermt2d %zmm26, %zmm5, %zmm2
; AVX512-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512-FCP-NEXT:    vpermt2d %zmm26, %zmm8, %zmm2
; AVX512-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512-FCP-NEXT:    vpermt2d %zmm1, %zmm31, %zmm2
; AVX512-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512-FCP-NEXT:    vpermt2d %zmm1, %zmm25, %zmm2
; AVX512-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512-FCP-NEXT:    vpermt2d %zmm1, %zmm3, %zmm2
; AVX512-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2d %zmm1, %zmm0, %zmm26
; AVX512-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm22
; AVX512-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512-FCP-NEXT:    vpermt2d %zmm22, %zmm5, %zmm2
; AVX512-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 1280(%rdi), %zmm19
; AVX512-FCP-NEXT:    vmovdqa64 1344(%rdi), %zmm2
; AVX512-FCP-NEXT:    vpermi2d %zmm19, %zmm2, %zmm5
; AVX512-FCP-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, %zmm4
; AVX512-FCP-NEXT:    vpermt2d %zmm22, %zmm8, %zmm4
; AVX512-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermi2d %zmm19, %zmm2, %zmm8
; AVX512-FCP-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm22, %zmm28
; AVX512-FCP-NEXT:    vpermt2d %zmm1, %zmm31, %zmm28
; AVX512-FCP-NEXT:    vpermi2d %zmm2, %zmm19, %zmm31
; AVX512-FCP-NEXT:    vmovdqa64 %zmm22, %zmm29
; AVX512-FCP-NEXT:    vpermt2d %zmm1, %zmm25, %zmm29
; AVX512-FCP-NEXT:    vpermi2d %zmm2, %zmm19, %zmm25
; AVX512-FCP-NEXT:    vmovdqa64 %zmm22, %zmm4
; AVX512-FCP-NEXT:    vpermt2d %zmm1, %zmm3, %zmm4
; AVX512-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermi2d %zmm2, %zmm19, %zmm3
; AVX512-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2d %zmm2, %zmm0, %zmm19
; AVX512-FCP-NEXT:    vpermt2d %zmm1, %zmm0, %zmm22
; AVX512-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm0
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [0,6,12,18,24,30,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm20
; AVX512-FCP-NEXT:    vpermt2d %zmm2, %zmm3, %zmm20
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [1,7,13,19,25,31,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm21
; AVX512-FCP-NEXT:    vpermt2d %zmm2, %zmm4, %zmm21
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm12 = [2,8,14,20,26,0,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512-FCP-NEXT:    vpermt2d %zmm2, %zmm12, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [3,9,15,21,27,0,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512-FCP-NEXT:    vpermt2d %zmm2, %zmm14, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [20,26,0,6,12,0,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm23
; AVX512-FCP-NEXT:    vpermt2d %zmm0, %zmm5, %zmm23
; AVX512-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm24 = [21,27,1,7,13,0,0,0]
; AVX512-FCP-NEXT:    vpermt2d %zmm0, %zmm24, %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm6
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm15
; AVX512-FCP-NEXT:    vpermt2d %zmm6, %zmm3, %zmm15
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm16
; AVX512-FCP-NEXT:    vpermt2d %zmm6, %zmm4, %zmm16
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm27
; AVX512-FCP-NEXT:    vpermt2d %zmm6, %zmm12, %zmm27
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm30
; AVX512-FCP-NEXT:    vpermt2d %zmm6, %zmm14, %zmm30
; AVX512-FCP-NEXT:    vmovdqa64 %zmm6, %zmm17
; AVX512-FCP-NEXT:    vpermt2d %zmm0, %zmm5, %zmm17
; AVX512-FCP-NEXT:    vpermt2d %zmm0, %zmm24, %zmm6
; AVX512-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm7
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm8
; AVX512-FCP-NEXT:    vpermt2d %zmm10, %zmm3, %zmm8
; AVX512-FCP-NEXT:    vmovdqa64 1216(%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 1152(%rdi), %zmm0
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512-FCP-NEXT:    vpermt2d %zmm10, %zmm4, %zmm9
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512-FCP-NEXT:    vpermt2d %zmm10, %zmm12, %zmm13
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm12
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512-FCP-NEXT:    vpermt2d %zmm10, %zmm14, %zmm18
; AVX512-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm14
; AVX512-FCP-NEXT:    vmovdqa64 %zmm10, %zmm11
; AVX512-FCP-NEXT:    vpermt2d %zmm7, %zmm5, %zmm11
; AVX512-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm5
; AVX512-FCP-NEXT:    vpermt2d %zmm0, %zmm24, %zmm1
; AVX512-FCP-NEXT:    vpermt2d %zmm7, %zmm24, %zmm10
; AVX512-FCP-NEXT:    movb $56, %al
; AVX512-FCP-NEXT:    kmovw %eax, %k2
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm15 {%k2}
; AVX512-FCP-NEXT:    movw $-2048, %ax # imm = 0xF800
; AVX512-FCP-NEXT:    kmovw %eax, %k1
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm15 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm20 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm20 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm8 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm3 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm9 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm21 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm16 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k1}
; AVX512-FCP-NEXT:    movw $31, %ax
; AVX512-FCP-NEXT:    kmovw %eax, %k2
; AVX512-FCP-NEXT:    vmovdqa32 %zmm13, %zmm28 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm28 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm27, %zmm7 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm24
; AVX512-FCP-NEXT:    vmovdqa32 %zmm12, %zmm31 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm31 {%k1}
; AVX512-FCP-NEXT:    vmovdqa32 %zmm18, %zmm29 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm29 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm30, %zmm12 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm12 {%k1}
; AVX512-FCP-NEXT:    vmovdqa32 %zmm14, %zmm25 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm25 {%k1}
; AVX512-FCP-NEXT:    movw $992, %ax # imm = 0x3E0
; AVX512-FCP-NEXT:    kmovw %eax, %k1
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm11 {%k1}
; AVX512-FCP-NEXT:    movb $-32, %al
; AVX512-FCP-NEXT:    kmovw %eax, %k2
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm23 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm23 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm17 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm17 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm5 {%k2}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa32 %zmm0, %zmm2 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k2}
; AVX512-FCP-NEXT:    vmovdqa32 %zmm26, %zmm6 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm6 {%k2}
; AVX512-FCP-NEXT:    vmovdqa32 %zmm19, %zmm1 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
; AVX512-FCP-NEXT:    vmovdqa32 %zmm22, %zmm10 {%k1}
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm10 {%k2}
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, 192(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm8, 128(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm20, 64(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm15, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm4, 192(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm16, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm21, 64(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm9, 128(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm31, 192(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm28, 128(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm25, 192(%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm12, (%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, 64(%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm29, 128(%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, 192(%r9)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm17, (%r9)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm23, 64(%r9)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, 128(%r9)
; AVX512-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512-FCP-NEXT:    vmovdqa64 %zmm10, 128(%rax)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, 192(%rax)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm6, (%rax)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, 64(%rax)
; AVX512-FCP-NEXT:    addq $2632, %rsp # imm = 0xA48
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i32_stride6_vf64:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    subq $2632, %rsp # imm = 0xA48
; AVX512DQ-NEXT:    vmovdqa64 1472(%rdi), %zmm21
; AVX512DQ-NEXT:    vmovdqa64 1408(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 1088(%rdi), %zmm20
; AVX512DQ-NEXT:    vmovdqa64 1024(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512DQ-NEXT:    vmovdqa64 576(%rdi), %zmm18
; AVX512DQ-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512DQ-NEXT:    vmovdqa64 640(%rdi), %zmm3
; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512DQ-NEXT:    vmovdqa64 256(%rdi), %zmm4
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512DQ-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512DQ-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512DQ-NEXT:    vpermt2d %zmm25, %zmm6, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-NEXT:    vpermt2d %zmm27, %zmm6, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm18, %zmm7
; AVX512DQ-NEXT:    vpermt2d %zmm2, %zmm5, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512DQ-NEXT:    vpermt2d %zmm20, %zmm6, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermi2d %zmm21, %zmm1, %zmm6
; AVX512DQ-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512DQ-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm8
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm6
; AVX512DQ-NEXT:    vpermt2d %zmm20, %zmm7, %zmm8
; AVX512DQ-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512DQ-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512DQ-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512DQ-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512DQ-NEXT:    vpermt2d %zmm25, %zmm7, %zmm0
; AVX512DQ-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512DQ-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512DQ-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512DQ-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512DQ-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512DQ-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512DQ-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512DQ-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512DQ-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512DQ-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512DQ-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512DQ-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512DQ-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512DQ-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512DQ-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermt2d %zmm27, %zmm0, %zmm3
; AVX512DQ-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm4, %zmm3
; AVX512DQ-NEXT:    vpermt2d %zmm25, %zmm7, %zmm3
; AVX512DQ-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2d %zmm25, %zmm0, %zmm4
; AVX512DQ-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm6, %zmm3
; AVX512DQ-NEXT:    vpermt2d %zmm20, %zmm7, %zmm3
; AVX512DQ-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2d %zmm21, %zmm0, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm18, %zmm1
; AVX512DQ-NEXT:    vpermt2d %zmm2, %zmm8, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2d %zmm20, %zmm0, %zmm6
; AVX512DQ-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512DQ-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQ-NEXT:    vpermt2d %zmm18, %zmm31, %zmm0
; AVX512DQ-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm25 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512DQ-NEXT:    # zmm25 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQ-NEXT:    vpermt2d %zmm18, %zmm25, %zmm0
; AVX512DQ-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512DQ-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQ-NEXT:    vpermt2d %zmm18, %zmm3, %zmm0
; AVX512DQ-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512DQ-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermt2d %zmm18, %zmm0, %zmm2
; AVX512DQ-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %zmm26
; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-NEXT:    vpermt2d %zmm26, %zmm5, %zmm2
; AVX512DQ-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-NEXT:    vpermt2d %zmm26, %zmm8, %zmm2
; AVX512DQ-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQ-NEXT:    vpermt2d %zmm1, %zmm31, %zmm2
; AVX512DQ-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQ-NEXT:    vpermt2d %zmm1, %zmm25, %zmm2
; AVX512DQ-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQ-NEXT:    vpermt2d %zmm1, %zmm3, %zmm2
; AVX512DQ-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2d %zmm1, %zmm0, %zmm26
; AVX512DQ-NEXT:    vmovdqa64 896(%rdi), %zmm22
; AVX512DQ-NEXT:    vmovdqa64 960(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-NEXT:    vpermt2d %zmm22, %zmm5, %zmm2
; AVX512DQ-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 1280(%rdi), %zmm19
; AVX512DQ-NEXT:    vmovdqa64 1344(%rdi), %zmm2
; AVX512DQ-NEXT:    vpermi2d %zmm19, %zmm2, %zmm5
; AVX512DQ-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm1, %zmm4
; AVX512DQ-NEXT:    vpermt2d %zmm22, %zmm8, %zmm4
; AVX512DQ-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermi2d %zmm19, %zmm2, %zmm8
; AVX512DQ-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm22, %zmm28
; AVX512DQ-NEXT:    vpermt2d %zmm1, %zmm31, %zmm28
; AVX512DQ-NEXT:    vpermi2d %zmm2, %zmm19, %zmm31
; AVX512DQ-NEXT:    vmovdqa64 %zmm22, %zmm29
; AVX512DQ-NEXT:    vpermt2d %zmm1, %zmm25, %zmm29
; AVX512DQ-NEXT:    vpermi2d %zmm2, %zmm19, %zmm25
; AVX512DQ-NEXT:    vmovdqa64 %zmm22, %zmm4
; AVX512DQ-NEXT:    vpermt2d %zmm1, %zmm3, %zmm4
; AVX512DQ-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermi2d %zmm2, %zmm19, %zmm3
; AVX512DQ-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2d %zmm2, %zmm0, %zmm19
; AVX512DQ-NEXT:    vpermt2d %zmm1, %zmm0, %zmm22
; AVX512DQ-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512DQ-NEXT:    vmovdqa64 384(%rdi), %zmm0
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [0,6,12,18,24,30,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm20
; AVX512DQ-NEXT:    vpermt2d %zmm2, %zmm3, %zmm20
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [1,7,13,19,25,31,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm21
; AVX512DQ-NEXT:    vpermt2d %zmm2, %zmm4, %zmm21
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm12 = [2,8,14,20,26,0,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-NEXT:    vpermt2d %zmm2, %zmm12, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [3,9,15,21,27,0,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-NEXT:    vpermt2d %zmm2, %zmm14, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [20,26,0,6,12,0,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm23
; AVX512DQ-NEXT:    vpermt2d %zmm0, %zmm5, %zmm23
; AVX512DQ-NEXT:    vpmovsxbd {{.*#+}} ymm24 = [21,27,1,7,13,0,0,0]
; AVX512DQ-NEXT:    vpermt2d %zmm0, %zmm24, %zmm2
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm6
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm15
; AVX512DQ-NEXT:    vpermt2d %zmm6, %zmm3, %zmm15
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm16
; AVX512DQ-NEXT:    vpermt2d %zmm6, %zmm4, %zmm16
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm27
; AVX512DQ-NEXT:    vpermt2d %zmm6, %zmm12, %zmm27
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm30
; AVX512DQ-NEXT:    vpermt2d %zmm6, %zmm14, %zmm30
; AVX512DQ-NEXT:    vmovdqa64 %zmm6, %zmm17
; AVX512DQ-NEXT:    vpermt2d %zmm0, %zmm5, %zmm17
; AVX512DQ-NEXT:    vpermt2d %zmm0, %zmm24, %zmm6
; AVX512DQ-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512DQ-NEXT:    vmovdqa64 768(%rdi), %zmm7
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm8
; AVX512DQ-NEXT:    vpermt2d %zmm10, %zmm3, %zmm8
; AVX512DQ-NEXT:    vmovdqa64 1216(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 1152(%rdi), %zmm0
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512DQ-NEXT:    vpermt2d %zmm10, %zmm4, %zmm9
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512DQ-NEXT:    vpermt2d %zmm10, %zmm12, %zmm13
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm0, %zmm12
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512DQ-NEXT:    vpermt2d %zmm10, %zmm14, %zmm18
; AVX512DQ-NEXT:    vpermi2d %zmm1, %zmm0, %zmm14
; AVX512DQ-NEXT:    vmovdqa64 %zmm10, %zmm11
; AVX512DQ-NEXT:    vpermt2d %zmm7, %zmm5, %zmm11
; AVX512DQ-NEXT:    vpermi2d %zmm0, %zmm1, %zmm5
; AVX512DQ-NEXT:    vpermt2d %zmm0, %zmm24, %zmm1
; AVX512DQ-NEXT:    vpermt2d %zmm7, %zmm24, %zmm10
; AVX512DQ-NEXT:    movb $56, %al
; AVX512DQ-NEXT:    kmovw %eax, %k2
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm15 {%k2}
; AVX512DQ-NEXT:    movw $-2048, %ax # imm = 0xF800
; AVX512DQ-NEXT:    kmovw %eax, %k1
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm15 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm20 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm20 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm8 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm3 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm9 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm21 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm16 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k1}
; AVX512DQ-NEXT:    movw $31, %ax
; AVX512DQ-NEXT:    kmovw %eax, %k2
; AVX512DQ-NEXT:    vmovdqa32 %zmm13, %zmm28 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm28 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm27, %zmm7 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm24
; AVX512DQ-NEXT:    vmovdqa32 %zmm12, %zmm31 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm31 {%k1}
; AVX512DQ-NEXT:    vmovdqa32 %zmm18, %zmm29 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm29 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm30, %zmm12 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm12 {%k1}
; AVX512DQ-NEXT:    vmovdqa32 %zmm14, %zmm25 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm25 {%k1}
; AVX512DQ-NEXT:    movw $992, %ax # imm = 0x3E0
; AVX512DQ-NEXT:    kmovw %eax, %k1
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm11 {%k1}
; AVX512DQ-NEXT:    movb $-32, %al
; AVX512DQ-NEXT:    kmovw %eax, %k2
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm23 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm23 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm17 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm17 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm5 {%k2}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa32 %zmm0, %zmm2 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k2}
; AVX512DQ-NEXT:    vmovdqa32 %zmm26, %zmm6 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm6 {%k2}
; AVX512DQ-NEXT:    vmovdqa32 %zmm19, %zmm1 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
; AVX512DQ-NEXT:    vmovdqa32 %zmm22, %zmm10 {%k1}
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm10 {%k2}
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, 192(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm8, 128(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm20, 64(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm15, (%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm4, 192(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm16, (%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm21, 64(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm9, 128(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm31, 192(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm28, 128(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm25, 192(%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm12, (%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, 64(%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm29, 128(%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, 192(%r9)
; AVX512DQ-NEXT:    vmovdqa64 %zmm17, (%r9)
; AVX512DQ-NEXT:    vmovdqa64 %zmm23, 64(%r9)
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, 128(%r9)
; AVX512DQ-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-NEXT:    vmovdqa64 %zmm10, 128(%rax)
; AVX512DQ-NEXT:    vmovdqa64 %zmm1, 192(%rax)
; AVX512DQ-NEXT:    vmovdqa64 %zmm6, (%rax)
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, 64(%rax)
; AVX512DQ-NEXT:    addq $2632, %rsp # imm = 0xA48
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i32_stride6_vf64:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    subq $2632, %rsp # imm = 0xA48
; AVX512DQ-FCP-NEXT:    vmovdqa64 1472(%rdi), %zmm21
; AVX512DQ-FCP-NEXT:    vmovdqa64 1408(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 1088(%rdi), %zmm20
; AVX512DQ-FCP-NEXT:    vmovdqa64 1024(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm18
; AVX512DQ-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512DQ-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512DQ-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm4
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512DQ-FCP-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512DQ-FCP-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm25, %zmm6, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm27, %zmm6, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm18, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm2, %zmm5, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm20, %zmm6, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm21, %zmm1, %zmm6
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512DQ-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm8
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm6
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm20, %zmm7, %zmm8
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512DQ-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm25, %zmm7, %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512DQ-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512DQ-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512DQ-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512DQ-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm27, %zmm0, %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm4, %zmm3
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm25, %zmm7, %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm25, %zmm0, %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm6, %zmm3
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm20, %zmm7, %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm21, %zmm0, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm18, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm2, %zmm8, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm20, %zmm0, %zmm6
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512DQ-FCP-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm18, %zmm31, %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm25 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512DQ-FCP-NEXT:    # zmm25 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm18, %zmm25, %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512DQ-FCP-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm18, %zmm3, %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512DQ-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm18, %zmm0, %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm26
; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm26, %zmm5, %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm26, %zmm8, %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm1, %zmm31, %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm1, %zmm25, %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm1, %zmm3, %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm1, %zmm0, %zmm26
; AVX512DQ-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm22
; AVX512DQ-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm22, %zmm5, %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 1280(%rdi), %zmm19
; AVX512DQ-FCP-NEXT:    vmovdqa64 1344(%rdi), %zmm2
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm19, %zmm2, %zmm5
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, %zmm4
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm22, %zmm8, %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm19, %zmm2, %zmm8
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm22, %zmm28
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm1, %zmm31, %zmm28
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm2, %zmm19, %zmm31
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm22, %zmm29
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm1, %zmm25, %zmm29
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm2, %zmm19, %zmm25
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm22, %zmm4
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm1, %zmm3, %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm2, %zmm19, %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm2, %zmm0, %zmm19
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm1, %zmm0, %zmm22
; AVX512DQ-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [0,6,12,18,24,30,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm20
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm2, %zmm3, %zmm20
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [1,7,13,19,25,31,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm21
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm2, %zmm4, %zmm21
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm12 = [2,8,14,20,26,0,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm2, %zmm12, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [3,9,15,21,27,0,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm2, %zmm14, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [20,26,0,6,12,0,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm23
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm0, %zmm5, %zmm23
; AVX512DQ-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm24 = [21,27,1,7,13,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm0, %zmm24, %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm6
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm15
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm6, %zmm3, %zmm15
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm16
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm6, %zmm4, %zmm16
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm27
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm6, %zmm12, %zmm27
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm30
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm6, %zmm14, %zmm30
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm6, %zmm17
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm0, %zmm5, %zmm17
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm0, %zmm24, %zmm6
; AVX512DQ-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512DQ-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm8
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm10, %zmm3, %zmm8
; AVX512DQ-FCP-NEXT:    vmovdqa64 1216(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 1152(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm10, %zmm4, %zmm9
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm10, %zmm12, %zmm13
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm12
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm10, %zmm14, %zmm18
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm14
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm10, %zmm11
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm7, %zmm5, %zmm11
; AVX512DQ-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm5
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm0, %zmm24, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2d %zmm7, %zmm24, %zmm10
; AVX512DQ-FCP-NEXT:    movb $56, %al
; AVX512DQ-FCP-NEXT:    kmovw %eax, %k2
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm15 {%k2}
; AVX512DQ-FCP-NEXT:    movw $-2048, %ax # imm = 0xF800
; AVX512DQ-FCP-NEXT:    kmovw %eax, %k1
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm15 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm20 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm20 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm8 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm3 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm9 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm21 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm16 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k1}
; AVX512DQ-FCP-NEXT:    movw $31, %ax
; AVX512DQ-FCP-NEXT:    kmovw %eax, %k2
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm13, %zmm28 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm28 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm27, %zmm7 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm24
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm12, %zmm31 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm31 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm18, %zmm29 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm29 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm30, %zmm12 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm12 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm14, %zmm25 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm25 {%k1}
; AVX512DQ-FCP-NEXT:    movw $992, %ax # imm = 0x3E0
; AVX512DQ-FCP-NEXT:    kmovw %eax, %k1
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm11 {%k1}
; AVX512DQ-FCP-NEXT:    movb $-32, %al
; AVX512DQ-FCP-NEXT:    kmovw %eax, %k2
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm23 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm23 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm17 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm17 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm5 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm0, %zmm2 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm26, %zmm6 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm6 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm19, %zmm1 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqa32 %zmm22, %zmm10 {%k1}
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm10 {%k2}
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, 192(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm8, 128(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm20, 64(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm15, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm4, 192(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm16, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm21, 64(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm9, 128(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm31, 192(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm28, 128(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm25, 192(%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm12, (%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, 64(%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm29, 128(%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, 192(%r9)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm17, (%r9)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm23, 64(%r9)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, 128(%r9)
; AVX512DQ-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm10, 128(%rax)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, 192(%rax)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm6, (%rax)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, 64(%rax)
; AVX512DQ-FCP-NEXT:    addq $2632, %rsp # imm = 0xA48
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i32_stride6_vf64:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    subq $2632, %rsp # imm = 0xA48
; AVX512BW-NEXT:    vmovdqa64 1472(%rdi), %zmm21
; AVX512BW-NEXT:    vmovdqa64 1408(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %zmm20
; AVX512BW-NEXT:    vmovdqa64 1024(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm18
; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm3
; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm4
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512BW-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512BW-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512BW-NEXT:    vpermt2d %zmm25, %zmm6, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm6, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm7
; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm5, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm6, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermi2d %zmm21, %zmm1, %zmm6
; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512BW-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm8
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm6
; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm7, %zmm8
; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512BW-NEXT:    vpermt2d %zmm25, %zmm7, %zmm0
; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512BW-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512BW-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512BW-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512BW-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermt2d %zmm27, %zmm0, %zmm3
; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm3
; AVX512BW-NEXT:    vpermt2d %zmm25, %zmm7, %zmm3
; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2d %zmm25, %zmm0, %zmm4
; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm3
; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm7, %zmm3
; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2d %zmm21, %zmm0, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm18, %zmm1
; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm8, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2d %zmm20, %zmm0, %zmm6
; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512BW-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm31, %zmm0
; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm25 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512BW-NEXT:    # zmm25 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm25, %zmm0
; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512BW-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm3, %zmm0
; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermt2d %zmm18, %zmm0, %zmm2
; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm26
; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512BW-NEXT:    vpermt2d %zmm26, %zmm5, %zmm2
; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512BW-NEXT:    vpermt2d %zmm26, %zmm8, %zmm2
; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm31, %zmm2
; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm25, %zmm2
; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm3, %zmm2
; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm0, %zmm26
; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm22
; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512BW-NEXT:    vpermt2d %zmm22, %zmm5, %zmm2
; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 1280(%rdi), %zmm19
; AVX512BW-NEXT:    vmovdqa64 1344(%rdi), %zmm2
; AVX512BW-NEXT:    vpermi2d %zmm19, %zmm2, %zmm5
; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm4
; AVX512BW-NEXT:    vpermt2d %zmm22, %zmm8, %zmm4
; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermi2d %zmm19, %zmm2, %zmm8
; AVX512BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm28
; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm31, %zmm28
; AVX512BW-NEXT:    vpermi2d %zmm2, %zmm19, %zmm31
; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm29
; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm25, %zmm29
; AVX512BW-NEXT:    vpermi2d %zmm2, %zmm19, %zmm25
; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm4
; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm3, %zmm4
; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermi2d %zmm2, %zmm19, %zmm3
; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm0, %zmm19
; AVX512BW-NEXT:    vpermt2d %zmm1, %zmm0, %zmm22
; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm0
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [0,6,12,18,24,30,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm20
; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm3, %zmm20
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [1,7,13,19,25,31,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm21
; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm4, %zmm21
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm12 = [2,8,14,20,26,0,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm12, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [3,9,15,21,27,0,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512BW-NEXT:    vpermt2d %zmm2, %zmm14, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [20,26,0,6,12,0,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm23
; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm5, %zmm23
; AVX512BW-NEXT:    vpmovsxbd {{.*#+}} ymm24 = [21,27,1,7,13,0,0,0]
; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm24, %zmm2
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm6
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm15
; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm3, %zmm15
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm16
; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm4, %zmm16
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm27
; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm12, %zmm27
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm30
; AVX512BW-NEXT:    vpermt2d %zmm6, %zmm14, %zmm30
; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm17
; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm5, %zmm17
; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm24, %zmm6
; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm7
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm8
; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm3, %zmm8
; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %zmm0
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm4, %zmm9
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm12, %zmm13
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm12
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512BW-NEXT:    vpermt2d %zmm10, %zmm14, %zmm18
; AVX512BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm14
; AVX512BW-NEXT:    vmovdqa64 %zmm10, %zmm11
; AVX512BW-NEXT:    vpermt2d %zmm7, %zmm5, %zmm11
; AVX512BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm5
; AVX512BW-NEXT:    vpermt2d %zmm0, %zmm24, %zmm1
; AVX512BW-NEXT:    vpermt2d %zmm7, %zmm24, %zmm10
; AVX512BW-NEXT:    movb $56, %al
; AVX512BW-NEXT:    kmovd %eax, %k2
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm15 {%k2}
; AVX512BW-NEXT:    movw $-2048, %ax # imm = 0xF800
; AVX512BW-NEXT:    kmovd %eax, %k1
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm15 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm20 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm20 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm8 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm3 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm9 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm21 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm16 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k1}
; AVX512BW-NEXT:    movw $31, %ax
; AVX512BW-NEXT:    kmovd %eax, %k2
; AVX512BW-NEXT:    vmovdqa32 %zmm13, %zmm28 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm28 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm27, %zmm7 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm24
; AVX512BW-NEXT:    vmovdqa32 %zmm12, %zmm31 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm31 {%k1}
; AVX512BW-NEXT:    vmovdqa32 %zmm18, %zmm29 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm29 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm30, %zmm12 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm12 {%k1}
; AVX512BW-NEXT:    vmovdqa32 %zmm14, %zmm25 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm25 {%k1}
; AVX512BW-NEXT:    movw $992, %ax # imm = 0x3E0
; AVX512BW-NEXT:    kmovd %eax, %k1
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm11 {%k1}
; AVX512BW-NEXT:    movb $-32, %al
; AVX512BW-NEXT:    kmovd %eax, %k2
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm23 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm23 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm17 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm17 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm5 {%k2}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa32 %zmm0, %zmm2 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k2}
; AVX512BW-NEXT:    vmovdqa32 %zmm26, %zmm6 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm6 {%k2}
; AVX512BW-NEXT:    vmovdqa32 %zmm19, %zmm1 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
; AVX512BW-NEXT:    vmovdqa32 %zmm22, %zmm10 {%k1}
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm10 {%k2}
; AVX512BW-NEXT:    vmovdqa64 %zmm3, 192(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm8, 128(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm20, 64(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm15, (%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm4, 192(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm16, (%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm21, 64(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm9, 128(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm31, 192(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm28, 128(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm25, 192(%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm12, (%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm7, 64(%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm29, 128(%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm5, 192(%r9)
; AVX512BW-NEXT:    vmovdqa64 %zmm17, (%r9)
; AVX512BW-NEXT:    vmovdqa64 %zmm23, 64(%r9)
; AVX512BW-NEXT:    vmovdqa64 %zmm11, 128(%r9)
; AVX512BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-NEXT:    vmovdqa64 %zmm10, 128(%rax)
; AVX512BW-NEXT:    vmovdqa64 %zmm1, 192(%rax)
; AVX512BW-NEXT:    vmovdqa64 %zmm6, (%rax)
; AVX512BW-NEXT:    vmovdqa64 %zmm2, 64(%rax)
; AVX512BW-NEXT:    addq $2632, %rsp # imm = 0xA48
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i32_stride6_vf64:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    subq $2632, %rsp # imm = 0xA48
; AVX512BW-FCP-NEXT:    vmovdqa64 1472(%rdi), %zmm21
; AVX512BW-FCP-NEXT:    vmovdqa64 1408(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 1088(%rdi), %zmm20
; AVX512BW-FCP-NEXT:    vmovdqa64 1024(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm18
; AVX512BW-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512BW-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm4
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512BW-FCP-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512BW-FCP-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2d %zmm25, %zmm6, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2d %zmm27, %zmm6, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm18, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2d %zmm2, %zmm5, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2d %zmm20, %zmm6, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermi2d %zmm21, %zmm1, %zmm6
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512BW-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm8
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm6
; AVX512BW-FCP-NEXT:    vpermt2d %zmm20, %zmm7, %zmm8
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512BW-FCP-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512BW-FCP-NEXT:    vpermt2d %zmm25, %zmm7, %zmm0
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512BW-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512BW-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512BW-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512BW-FCP-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512BW-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermt2d %zmm27, %zmm0, %zmm3
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm3
; AVX512BW-FCP-NEXT:    vpermt2d %zmm25, %zmm7, %zmm3
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2d %zmm25, %zmm0, %zmm4
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm6, %zmm3
; AVX512BW-FCP-NEXT:    vpermt2d %zmm20, %zmm7, %zmm3
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2d %zmm21, %zmm0, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm18, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2d %zmm2, %zmm8, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2d %zmm20, %zmm0, %zmm6
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512BW-FCP-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512BW-FCP-NEXT:    vpermt2d %zmm18, %zmm31, %zmm0
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm25 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512BW-FCP-NEXT:    # zmm25 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512BW-FCP-NEXT:    vpermt2d %zmm18, %zmm25, %zmm0
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512BW-FCP-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512BW-FCP-NEXT:    vpermt2d %zmm18, %zmm3, %zmm0
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512BW-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermt2d %zmm18, %zmm0, %zmm2
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm26
; AVX512BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512BW-FCP-NEXT:    vpermt2d %zmm26, %zmm5, %zmm2
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512BW-FCP-NEXT:    vpermt2d %zmm26, %zmm8, %zmm2
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512BW-FCP-NEXT:    vpermt2d %zmm1, %zmm31, %zmm2
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512BW-FCP-NEXT:    vpermt2d %zmm1, %zmm25, %zmm2
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512BW-FCP-NEXT:    vpermt2d %zmm1, %zmm3, %zmm2
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2d %zmm1, %zmm0, %zmm26
; AVX512BW-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm22
; AVX512BW-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512BW-FCP-NEXT:    vpermt2d %zmm22, %zmm5, %zmm2
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 1280(%rdi), %zmm19
; AVX512BW-FCP-NEXT:    vmovdqa64 1344(%rdi), %zmm2
; AVX512BW-FCP-NEXT:    vpermi2d %zmm19, %zmm2, %zmm5
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm4
; AVX512BW-FCP-NEXT:    vpermt2d %zmm22, %zmm8, %zmm4
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermi2d %zmm19, %zmm2, %zmm8
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm22, %zmm28
; AVX512BW-FCP-NEXT:    vpermt2d %zmm1, %zmm31, %zmm28
; AVX512BW-FCP-NEXT:    vpermi2d %zmm2, %zmm19, %zmm31
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm22, %zmm29
; AVX512BW-FCP-NEXT:    vpermt2d %zmm1, %zmm25, %zmm29
; AVX512BW-FCP-NEXT:    vpermi2d %zmm2, %zmm19, %zmm25
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm22, %zmm4
; AVX512BW-FCP-NEXT:    vpermt2d %zmm1, %zmm3, %zmm4
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermi2d %zmm2, %zmm19, %zmm3
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2d %zmm2, %zmm0, %zmm19
; AVX512BW-FCP-NEXT:    vpermt2d %zmm1, %zmm0, %zmm22
; AVX512BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [0,6,12,18,24,30,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm20
; AVX512BW-FCP-NEXT:    vpermt2d %zmm2, %zmm3, %zmm20
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [1,7,13,19,25,31,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm21
; AVX512BW-FCP-NEXT:    vpermt2d %zmm2, %zmm4, %zmm21
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm12 = [2,8,14,20,26,0,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2d %zmm2, %zmm12, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [3,9,15,21,27,0,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2d %zmm2, %zmm14, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [20,26,0,6,12,0,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm23
; AVX512BW-FCP-NEXT:    vpermt2d %zmm0, %zmm5, %zmm23
; AVX512BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm24 = [21,27,1,7,13,0,0,0]
; AVX512BW-FCP-NEXT:    vpermt2d %zmm0, %zmm24, %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm6
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm15
; AVX512BW-FCP-NEXT:    vpermt2d %zmm6, %zmm3, %zmm15
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm16
; AVX512BW-FCP-NEXT:    vpermt2d %zmm6, %zmm4, %zmm16
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm27
; AVX512BW-FCP-NEXT:    vpermt2d %zmm6, %zmm12, %zmm27
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm30
; AVX512BW-FCP-NEXT:    vpermt2d %zmm6, %zmm14, %zmm30
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm6, %zmm17
; AVX512BW-FCP-NEXT:    vpermt2d %zmm0, %zmm5, %zmm17
; AVX512BW-FCP-NEXT:    vpermt2d %zmm0, %zmm24, %zmm6
; AVX512BW-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512BW-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm7
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm8
; AVX512BW-FCP-NEXT:    vpermt2d %zmm10, %zmm3, %zmm8
; AVX512BW-FCP-NEXT:    vmovdqa64 1216(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 1152(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512BW-FCP-NEXT:    vpermt2d %zmm10, %zmm4, %zmm9
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512BW-FCP-NEXT:    vpermt2d %zmm10, %zmm12, %zmm13
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm12
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512BW-FCP-NEXT:    vpermt2d %zmm10, %zmm14, %zmm18
; AVX512BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm14
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm10, %zmm11
; AVX512BW-FCP-NEXT:    vpermt2d %zmm7, %zmm5, %zmm11
; AVX512BW-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm5
; AVX512BW-FCP-NEXT:    vpermt2d %zmm0, %zmm24, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2d %zmm7, %zmm24, %zmm10
; AVX512BW-FCP-NEXT:    movb $56, %al
; AVX512BW-FCP-NEXT:    kmovd %eax, %k2
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm15 {%k2}
; AVX512BW-FCP-NEXT:    movw $-2048, %ax # imm = 0xF800
; AVX512BW-FCP-NEXT:    kmovd %eax, %k1
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm15 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm20 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm20 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm8 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm3 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm9 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm21 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm16 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k1}
; AVX512BW-FCP-NEXT:    movw $31, %ax
; AVX512BW-FCP-NEXT:    kmovd %eax, %k2
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm13, %zmm28 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm28 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm27, %zmm7 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm24
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm12, %zmm31 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm31 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm18, %zmm29 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm29 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm30, %zmm12 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm12 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm14, %zmm25 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm25 {%k1}
; AVX512BW-FCP-NEXT:    movw $992, %ax # imm = 0x3E0
; AVX512BW-FCP-NEXT:    kmovd %eax, %k1
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm11 {%k1}
; AVX512BW-FCP-NEXT:    movb $-32, %al
; AVX512BW-FCP-NEXT:    kmovd %eax, %k2
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm23 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm23 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm17 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm17 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm5 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm2 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm26, %zmm6 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm6 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm19, %zmm1 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqa32 %zmm22, %zmm10 {%k1}
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm10 {%k2}
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, 192(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm8, 128(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm20, 64(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm15, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, 192(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm16, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm21, 64(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm9, 128(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm31, 192(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm28, 128(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm25, 192(%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm12, (%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, 64(%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm29, 128(%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, 192(%r9)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm17, (%r9)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm23, 64(%r9)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, 128(%r9)
; AVX512BW-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm10, 128(%rax)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, 192(%rax)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm6, (%rax)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, 64(%rax)
; AVX512BW-FCP-NEXT:    addq $2632, %rsp # imm = 0xA48
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i32_stride6_vf64:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    subq $2632, %rsp # imm = 0xA48
; AVX512DQ-BW-NEXT:    vmovdqa64 1472(%rdi), %zmm21
; AVX512DQ-BW-NEXT:    vmovdqa64 1408(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 1088(%rdi), %zmm20
; AVX512DQ-BW-NEXT:    vmovdqa64 1024(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 576(%rdi), %zmm18
; AVX512DQ-BW-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512DQ-BW-NEXT:    vmovdqa64 640(%rdi), %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512DQ-BW-NEXT:    vmovdqa64 256(%rdi), %zmm4
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512DQ-BW-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512DQ-BW-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2d %zmm25, %zmm6, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2d %zmm27, %zmm6, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm18, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2d %zmm2, %zmm5, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2d %zmm20, %zmm6, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermi2d %zmm21, %zmm1, %zmm6
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512DQ-BW-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm8
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm6
; AVX512DQ-BW-NEXT:    vpermt2d %zmm20, %zmm7, %zmm8
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512DQ-BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512DQ-BW-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512DQ-BW-NEXT:    vpermt2d %zmm25, %zmm7, %zmm0
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512DQ-BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512DQ-BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512DQ-BW-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512DQ-BW-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512DQ-BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermt2d %zmm27, %zmm0, %zmm3
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm4, %zmm3
; AVX512DQ-BW-NEXT:    vpermt2d %zmm25, %zmm7, %zmm3
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2d %zmm25, %zmm0, %zmm4
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm6, %zmm3
; AVX512DQ-BW-NEXT:    vpermt2d %zmm20, %zmm7, %zmm3
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2d %zmm21, %zmm0, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm18, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2d %zmm2, %zmm8, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2d %zmm20, %zmm0, %zmm6
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512DQ-BW-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQ-BW-NEXT:    vpermt2d %zmm18, %zmm31, %zmm0
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm25 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512DQ-BW-NEXT:    # zmm25 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQ-BW-NEXT:    vpermt2d %zmm18, %zmm25, %zmm0
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512DQ-BW-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQ-BW-NEXT:    vpermt2d %zmm18, %zmm3, %zmm0
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512DQ-BW-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermt2d %zmm18, %zmm0, %zmm2
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 128(%rdi), %zmm26
; AVX512DQ-BW-NEXT:    vmovdqa64 192(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-BW-NEXT:    vpermt2d %zmm26, %zmm5, %zmm2
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-BW-NEXT:    vpermt2d %zmm26, %zmm8, %zmm2
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQ-BW-NEXT:    vpermt2d %zmm1, %zmm31, %zmm2
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQ-BW-NEXT:    vpermt2d %zmm1, %zmm25, %zmm2
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQ-BW-NEXT:    vpermt2d %zmm1, %zmm3, %zmm2
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2d %zmm1, %zmm0, %zmm26
; AVX512DQ-BW-NEXT:    vmovdqa64 896(%rdi), %zmm22
; AVX512DQ-BW-NEXT:    vmovdqa64 960(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-BW-NEXT:    vpermt2d %zmm22, %zmm5, %zmm2
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 1280(%rdi), %zmm19
; AVX512DQ-BW-NEXT:    vmovdqa64 1344(%rdi), %zmm2
; AVX512DQ-BW-NEXT:    vpermi2d %zmm19, %zmm2, %zmm5
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, %zmm4
; AVX512DQ-BW-NEXT:    vpermt2d %zmm22, %zmm8, %zmm4
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermi2d %zmm19, %zmm2, %zmm8
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm22, %zmm28
; AVX512DQ-BW-NEXT:    vpermt2d %zmm1, %zmm31, %zmm28
; AVX512DQ-BW-NEXT:    vpermi2d %zmm2, %zmm19, %zmm31
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm22, %zmm29
; AVX512DQ-BW-NEXT:    vpermt2d %zmm1, %zmm25, %zmm29
; AVX512DQ-BW-NEXT:    vpermi2d %zmm2, %zmm19, %zmm25
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm22, %zmm4
; AVX512DQ-BW-NEXT:    vpermt2d %zmm1, %zmm3, %zmm4
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermi2d %zmm2, %zmm19, %zmm3
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2d %zmm2, %zmm0, %zmm19
; AVX512DQ-BW-NEXT:    vpermt2d %zmm1, %zmm0, %zmm22
; AVX512DQ-BW-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 384(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [0,6,12,18,24,30,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm20
; AVX512DQ-BW-NEXT:    vpermt2d %zmm2, %zmm3, %zmm20
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [1,7,13,19,25,31,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm21
; AVX512DQ-BW-NEXT:    vpermt2d %zmm2, %zmm4, %zmm21
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm12 = [2,8,14,20,26,0,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2d %zmm2, %zmm12, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [3,9,15,21,27,0,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2d %zmm2, %zmm14, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [20,26,0,6,12,0,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm23
; AVX512DQ-BW-NEXT:    vpermt2d %zmm0, %zmm5, %zmm23
; AVX512DQ-BW-NEXT:    vpmovsxbd {{.*#+}} ymm24 = [21,27,1,7,13,0,0,0]
; AVX512DQ-BW-NEXT:    vpermt2d %zmm0, %zmm24, %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 64(%rdi), %zmm6
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm15
; AVX512DQ-BW-NEXT:    vpermt2d %zmm6, %zmm3, %zmm15
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm16
; AVX512DQ-BW-NEXT:    vpermt2d %zmm6, %zmm4, %zmm16
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm27
; AVX512DQ-BW-NEXT:    vpermt2d %zmm6, %zmm12, %zmm27
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm30
; AVX512DQ-BW-NEXT:    vpermt2d %zmm6, %zmm14, %zmm30
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm6, %zmm17
; AVX512DQ-BW-NEXT:    vpermt2d %zmm0, %zmm5, %zmm17
; AVX512DQ-BW-NEXT:    vpermt2d %zmm0, %zmm24, %zmm6
; AVX512DQ-BW-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512DQ-BW-NEXT:    vmovdqa64 768(%rdi), %zmm7
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm8
; AVX512DQ-BW-NEXT:    vpermt2d %zmm10, %zmm3, %zmm8
; AVX512DQ-BW-NEXT:    vmovdqa64 1216(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 1152(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512DQ-BW-NEXT:    vpermt2d %zmm10, %zmm4, %zmm9
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512DQ-BW-NEXT:    vpermt2d %zmm10, %zmm12, %zmm13
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm12
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512DQ-BW-NEXT:    vpermt2d %zmm10, %zmm14, %zmm18
; AVX512DQ-BW-NEXT:    vpermi2d %zmm1, %zmm0, %zmm14
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm10, %zmm11
; AVX512DQ-BW-NEXT:    vpermt2d %zmm7, %zmm5, %zmm11
; AVX512DQ-BW-NEXT:    vpermi2d %zmm0, %zmm1, %zmm5
; AVX512DQ-BW-NEXT:    vpermt2d %zmm0, %zmm24, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2d %zmm7, %zmm24, %zmm10
; AVX512DQ-BW-NEXT:    movb $56, %al
; AVX512DQ-BW-NEXT:    kmovd %eax, %k2
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm15 {%k2}
; AVX512DQ-BW-NEXT:    movw $-2048, %ax # imm = 0xF800
; AVX512DQ-BW-NEXT:    kmovd %eax, %k1
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm15 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm20 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm20 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm8 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm3 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm9 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm21 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm16 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k1}
; AVX512DQ-BW-NEXT:    movw $31, %ax
; AVX512DQ-BW-NEXT:    kmovd %eax, %k2
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm13, %zmm28 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm28 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm27, %zmm7 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm24
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm12, %zmm31 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm31 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm18, %zmm29 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm29 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm30, %zmm12 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm12 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm14, %zmm25 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm25 {%k1}
; AVX512DQ-BW-NEXT:    movw $992, %ax # imm = 0x3E0
; AVX512DQ-BW-NEXT:    kmovd %eax, %k1
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm11 {%k1}
; AVX512DQ-BW-NEXT:    movb $-32, %al
; AVX512DQ-BW-NEXT:    kmovd %eax, %k2
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm23 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm23 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm17 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm17 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm5 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm0, %zmm2 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm26, %zmm6 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm6 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm19, %zmm1 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqa32 %zmm22, %zmm10 {%k1}
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm10 {%k2}
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, 192(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm8, 128(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm20, 64(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm15, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm4, 192(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm16, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm21, 64(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm9, 128(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm31, 192(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm28, 128(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm25, 192(%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm12, (%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, 64(%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm29, 128(%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, 192(%r9)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm17, (%r9)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm23, 64(%r9)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, 128(%r9)
; AVX512DQ-BW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm10, 128(%rax)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, 192(%rax)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm6, (%rax)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, 64(%rax)
; AVX512DQ-BW-NEXT:    addq $2632, %rsp # imm = 0xA48
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i32_stride6_vf64:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    subq $2632, %rsp # imm = 0xA48
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1472(%rdi), %zmm21
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1408(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1088(%rdi), %zmm20
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1024(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm18
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm4
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [0,6,12,0,0,0,20,26,0,6,12,0,0,0,20,26]
; AVX512DQ-BW-FCP-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [0,0,0,2,8,14,20,26,0,0,0,2,8,14,20,26]
; AVX512DQ-BW-FCP-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm25, %zmm6, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm27, %zmm6, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm18, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm2, %zmm5, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm20, %zmm6, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm21, %zmm1, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,3,9,15,21,27,0,0,0,3,9,15,21,27]
; AVX512DQ-BW-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm20, %zmm7, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [1,7,13,0,0,0,21,27,1,7,13,0,0,0,21,27]
; AVX512DQ-BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm25, %zmm7, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,20,26,0,6,12,0,0,0,20,26,0,6,12]
; AVX512DQ-BW-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,0,21,27,1,7,13,0,0,0,21,27,1,7,13]
; AVX512DQ-BW-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm20, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm6, %zmm0, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm27, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm3, %zmm0, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm25, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm4, %zmm0, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm21, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,6,12,18,24,30,0,0,0,6,12,18,24,30]
; AVX512DQ-BW-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm27, %zmm7, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,1,7,13,19,25,31,0,0,1,7,13,19,25,31]
; AVX512DQ-BW-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm27, %zmm0, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm25, %zmm7, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm25, %zmm0, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm6, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm20, %zmm7, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm21, %zmm1, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm21, %zmm0, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm18, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm2, %zmm8, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm20, %zmm0, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [18,24,30,0,0,0,6,12,18,24,30,0,0,0,6,12]
; AVX512DQ-BW-FCP-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm18, %zmm31, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm25 = [19,25,31,0,0,1,7,13,19,25,31,0,0,1,7,13]
; AVX512DQ-BW-FCP-NEXT:    # zmm25 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm18, %zmm25, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm3 = [20,26,0,0,0,2,8,14,20,26,0,0,0,2,8,14]
; AVX512DQ-BW-FCP-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm18, %zmm3, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm0 = [21,27,0,0,0,3,9,15,21,27,0,0,0,3,9,15]
; AVX512DQ-BW-FCP-NEXT:    # zmm0 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm18, %zmm0, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm26
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm26, %zmm5, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm26, %zmm8, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm1, %zmm31, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm1, %zmm25, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm26, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm1, %zmm3, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm1, %zmm0, %zmm26
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm22
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm22, %zmm5, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1280(%rdi), %zmm19
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1344(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm19, %zmm2, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm22, %zmm8, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm19, %zmm2, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm22, %zmm28
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm1, %zmm31, %zmm28
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm2, %zmm19, %zmm31
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm22, %zmm29
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm1, %zmm25, %zmm29
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm2, %zmm19, %zmm25
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm22, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm1, %zmm3, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm2, %zmm19, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm2, %zmm0, %zmm19
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm1, %zmm0, %zmm22
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm3 = [0,6,12,18,24,30,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm2, %zmm3, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm4 = [1,7,13,19,25,31,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm21
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm2, %zmm4, %zmm21
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm12 = [2,8,14,20,26,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm2, %zmm12, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm14 = [3,9,15,21,27,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm2, %zmm14, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, (%rsp) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm5 = [20,26,0,6,12,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm23
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm0, %zmm5, %zmm23
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbd {{.*#+}} ymm24 = [21,27,1,7,13,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm0, %zmm24, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm6
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm6, %zmm3, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm6, %zmm4, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm27
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm6, %zmm12, %zmm27
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm30
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm6, %zmm14, %zmm30
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm6, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm0, %zmm5, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm0, %zmm24, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm10, %zmm3, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1216(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1152(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm10, %zmm4, %zmm9
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm10, %zmm12, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm12
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm10, %zmm14, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm1, %zmm0, %zmm14
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm10, %zmm11
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm7, %zmm5, %zmm11
; AVX512DQ-BW-FCP-NEXT:    vpermi2d %zmm0, %zmm1, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm0, %zmm24, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2d %zmm7, %zmm24, %zmm10
; AVX512DQ-BW-FCP-NEXT:    movb $56, %al
; AVX512DQ-BW-FCP-NEXT:    kmovd %eax, %k2
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm15 {%k2}
; AVX512DQ-BW-FCP-NEXT:    movw $-2048, %ax # imm = 0xF800
; AVX512DQ-BW-FCP-NEXT:    kmovd %eax, %k1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm15 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm20 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm20 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm8 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm8 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm3 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm3 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm9 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm9 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm21 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm21 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm16 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm16 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm4 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm4 {%k1}
; AVX512DQ-BW-FCP-NEXT:    movw $31, %ax
; AVX512DQ-BW-FCP-NEXT:    kmovd %eax, %k2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm13, %zmm28 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm28 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm27, %zmm7 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm24
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm12, %zmm31 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm31 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm18, %zmm29 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm29 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm7 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm12 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm30, %zmm12 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm12 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm14, %zmm25 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm25 {%k1}
; AVX512DQ-BW-FCP-NEXT:    movw $992, %ax # imm = 0x3E0
; AVX512DQ-BW-FCP-NEXT:    kmovd %eax, %k1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm11 {%k1}
; AVX512DQ-BW-FCP-NEXT:    movb $-32, %al
; AVX512DQ-BW-FCP-NEXT:    kmovd %eax, %k2
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm11 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm23 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm23 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm17 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm17 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm5 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm5 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm0, %zmm2 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm26, %zmm6 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm6 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm19, %zmm1 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa32 %zmm22, %zmm10 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm10 {%k2}
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm3, 192(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm8, 128(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm20, 64(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm15, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, 192(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm16, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm21, 64(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm9, 128(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm31, 192(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm28, 128(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm25, 192(%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm12, (%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, 64(%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm29, 128(%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, 192(%r9)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm17, (%r9)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm23, 64(%r9)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, 128(%r9)
; AVX512DQ-BW-FCP-NEXT:    movq {{[0-9]+}}(%rsp), %rax
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm10, 128(%rax)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm1, 192(%rax)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm6, (%rax)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, 64(%rax)
; AVX512DQ-BW-FCP-NEXT:    addq $2632, %rsp # imm = 0xA48
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <384 x i32>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <384 x i32> %wide.vec, <384 x i32> poison, <64 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90, i32 96, i32 102, i32 108, i32 114, i32 120, i32 126, i32 132, i32 138, i32 144, i32 150, i32 156, i32 162, i32 168, i32 174, i32 180, i32 186, i32 192, i32 198, i32 204, i32 210, i32 216, i32 222, i32 228, i32 234, i32 240, i32 246, i32 252, i32 258, i32 264, i32 270, i32 276, i32 282, i32 288, i32 294, i32 300, i32 306, i32 312, i32 318, i32 324, i32 330, i32 336, i32 342, i32 348, i32 354, i32 360, i32 366, i32 372, i32 378>
  %strided.vec1 = shufflevector <384 x i32> %wide.vec, <384 x i32> poison, <64 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91, i32 97, i32 103, i32 109, i32 115, i32 121, i32 127, i32 133, i32 139, i32 145, i32 151, i32 157, i32 163, i32 169, i32 175, i32 181, i32 187, i32 193, i32 199, i32 205, i32 211, i32 217, i32 223, i32 229, i32 235, i32 241, i32 247, i32 253, i32 259, i32 265, i32 271, i32 277, i32 283, i32 289, i32 295, i32 301, i32 307, i32 313, i32 319, i32 325, i32 331, i32 337, i32 343, i32 349, i32 355, i32 361, i32 367, i32 373, i32 379>
  %strided.vec2 = shufflevector <384 x i32> %wide.vec, <384 x i32> poison, <64 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92, i32 98, i32 104, i32 110, i32 116, i32 122, i32 128, i32 134, i32 140, i32 146, i32 152, i32 158, i32 164, i32 170, i32 176, i32 182, i32 188, i32 194, i32 200, i32 206, i32 212, i32 218, i32 224, i32 230, i32 236, i32 242, i32 248, i32 254, i32 260, i32 266, i32 272, i32 278, i32 284, i32 290, i32 296, i32 302, i32 308, i32 314, i32 320, i32 326, i32 332, i32 338, i32 344, i32 350, i32 356, i32 362, i32 368, i32 374, i32 380>
  %strided.vec3 = shufflevector <384 x i32> %wide.vec, <384 x i32> poison, <64 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93, i32 99, i32 105, i32 111, i32 117, i32 123, i32 129, i32 135, i32 141, i32 147, i32 153, i32 159, i32 165, i32 171, i32 177, i32 183, i32 189, i32 195, i32 201, i32 207, i32 213, i32 219, i32 225, i32 231, i32 237, i32 243, i32 249, i32 255, i32 261, i32 267, i32 273, i32 279, i32 285, i32 291, i32 297, i32 303, i32 309, i32 315, i32 321, i32 327, i32 333, i32 339, i32 345, i32 351, i32 357, i32 363, i32 369, i32 375, i32 381>
  %strided.vec4 = shufflevector <384 x i32> %wide.vec, <384 x i32> poison, <64 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94, i32 100, i32 106, i32 112, i32 118, i32 124, i32 130, i32 136, i32 142, i32 148, i32 154, i32 160, i32 166, i32 172, i32 178, i32 184, i32 190, i32 196, i32 202, i32 208, i32 214, i32 220, i32 226, i32 232, i32 238, i32 244, i32 250, i32 256, i32 262, i32 268, i32 274, i32 280, i32 286, i32 292, i32 298, i32 304, i32 310, i32 316, i32 322, i32 328, i32 334, i32 340, i32 346, i32 352, i32 358, i32 364, i32 370, i32 376, i32 382>
  %strided.vec5 = shufflevector <384 x i32> %wide.vec, <384 x i32> poison, <64 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95, i32 101, i32 107, i32 113, i32 119, i32 125, i32 131, i32 137, i32 143, i32 149, i32 155, i32 161, i32 167, i32 173, i32 179, i32 185, i32 191, i32 197, i32 203, i32 209, i32 215, i32 221, i32 227, i32 233, i32 239, i32 245, i32 251, i32 257, i32 263, i32 269, i32 275, i32 281, i32 287, i32 293, i32 299, i32 305, i32 311, i32 317, i32 323, i32 329, i32 335, i32 341, i32 347, i32 353, i32 359, i32 365, i32 371, i32 377, i32 383>
  store <64 x i32> %strided.vec0, ptr %out.vec0, align 64
  store <64 x i32> %strided.vec1, ptr %out.vec1, align 64
  store <64 x i32> %strided.vec2, ptr %out.vec2, align 64
  store <64 x i32> %strided.vec3, ptr %out.vec3, align 64
  store <64 x i32> %strided.vec4, ptr %out.vec4, align 64
  store <64 x i32> %strided.vec5, ptr %out.vec5, align 64
  ret void
}