llvm/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.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_i64_stride4_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: load_i64_stride4_vf2:
; SSE:       # %bb.0:
; SSE-NEXT:    movaps (%rdi), %xmm0
; SSE-NEXT:    movaps 16(%rdi), %xmm1
; SSE-NEXT:    movaps 32(%rdi), %xmm2
; SSE-NEXT:    movaps 48(%rdi), %xmm3
; SSE-NEXT:    movaps %xmm0, %xmm4
; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm2[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
; SSE-NEXT:    movaps %xmm4, (%rsi)
; SSE-NEXT:    movaps %xmm0, (%rdx)
; SSE-NEXT:    movaps %xmm2, (%rcx)
; SSE-NEXT:    movaps %xmm1, (%r8)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i64_stride4_vf2:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovaps 32(%rdi), %xmm0
; AVX-NEXT:    vmovaps (%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps 48(%rdi), %xmm1
; AVX-NEXT:    vmovaps 16(%rdi), %xmm3
; AVX-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm1[0]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm1[1]
; AVX-NEXT:    vmovaps %xmm2, (%rsi)
; AVX-NEXT:    vmovaps %xmm0, (%rdx)
; AVX-NEXT:    vmovaps %xmm4, (%rcx)
; AVX-NEXT:    vmovaps %xmm1, (%r8)
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i64_stride4_vf2:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovaps (%rdi), %xmm0
; AVX2-NEXT:    vmovaps 32(%rdi), %xmm1
; AVX2-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX2-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-NEXT:    vmovaps (%rdi), %ymm3
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm1[1],ymm3[3],ymm1[3]
; AVX2-NEXT:    vmovaps %xmm2, (%rsi)
; AVX2-NEXT:    vmovaps %xmm0, (%rdx)
; AVX2-NEXT:    vextractf128 $1, %ymm4, (%rcx)
; AVX2-NEXT:    vextractf128 $1, %ymm1, (%r8)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i64_stride4_vf2:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    vmovaps (%rdi), %xmm0
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %xmm1
; AVX2-FP-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm3
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm1[1],ymm3[3],ymm1[3]
; AVX2-FP-NEXT:    vmovaps %xmm2, (%rsi)
; AVX2-FP-NEXT:    vmovaps %xmm0, (%rdx)
; AVX2-FP-NEXT:    vextractf128 $1, %ymm4, (%rcx)
; AVX2-FP-NEXT:    vextractf128 $1, %ymm1, (%r8)
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i64_stride4_vf2:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    vmovaps (%rdi), %xmm0
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %xmm1
; AVX2-FCP-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm3
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm1[1],ymm3[3],ymm1[3]
; AVX2-FCP-NEXT:    vmovaps %xmm2, (%rsi)
; AVX2-FCP-NEXT:    vmovaps %xmm0, (%rdx)
; AVX2-FCP-NEXT:    vextractf128 $1, %ymm4, (%rcx)
; AVX2-FCP-NEXT:    vextractf128 $1, %ymm1, (%r8)
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i64_stride4_vf2:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovaps (%rdi), %xmm0
; AVX512-NEXT:    vmovaps 32(%rdi), %xmm1
; AVX512-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX512-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX512-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX512-NEXT:    vmovaps (%rdi), %ymm3
; AVX512-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
; AVX512-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm1[1],ymm3[3],ymm1[3]
; AVX512-NEXT:    vmovaps %xmm2, (%rsi)
; AVX512-NEXT:    vmovaps %xmm0, (%rdx)
; AVX512-NEXT:    vextractf128 $1, %ymm4, (%rcx)
; AVX512-NEXT:    vextractf128 $1, %ymm1, (%r8)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i64_stride4_vf2:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovaps {{.*#+}} xmm0 = [0,0,0,0,0,0,0,0,4,0,0,0,0,0,0,0]
; AVX512-FCP-NEXT:    vpermpd (%rdi), %zmm0, %zmm0
; AVX512-FCP-NEXT:    vmovaps (%rdi), %xmm1
; AVX512-FCP-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
; AVX512-FCP-NEXT:    vmovaps (%rdi), %ymm2
; AVX512-FCP-NEXT:    vmovaps 32(%rdi), %ymm3
; AVX512-FCP-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX512-FCP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
; AVX512-FCP-NEXT:    vmovaps %xmm0, (%rsi)
; AVX512-FCP-NEXT:    vmovaps %xmm1, (%rdx)
; AVX512-FCP-NEXT:    vextractf128 $1, %ymm4, (%rcx)
; AVX512-FCP-NEXT:    vextractf128 $1, %ymm2, (%r8)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i64_stride4_vf2:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovaps (%rdi), %xmm0
; AVX512DQ-NEXT:    vmovaps 32(%rdi), %xmm1
; AVX512DQ-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX512DQ-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX512DQ-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX512DQ-NEXT:    vmovaps (%rdi), %ymm3
; AVX512DQ-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
; AVX512DQ-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm1[1],ymm3[3],ymm1[3]
; AVX512DQ-NEXT:    vmovaps %xmm2, (%rsi)
; AVX512DQ-NEXT:    vmovaps %xmm0, (%rdx)
; AVX512DQ-NEXT:    vextractf128 $1, %ymm4, (%rcx)
; AVX512DQ-NEXT:    vextractf128 $1, %ymm1, (%r8)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride4_vf2:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovaps {{.*#+}} xmm0 = [0,0,0,0,0,0,0,0,4,0,0,0,0,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermpd (%rdi), %zmm0, %zmm0
; AVX512DQ-FCP-NEXT:    vmovaps (%rdi), %xmm1
; AVX512DQ-FCP-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
; AVX512DQ-FCP-NEXT:    vmovaps (%rdi), %ymm2
; AVX512DQ-FCP-NEXT:    vmovaps 32(%rdi), %ymm3
; AVX512DQ-FCP-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX512DQ-FCP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
; AVX512DQ-FCP-NEXT:    vmovaps %xmm0, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovaps %xmm1, (%rdx)
; AVX512DQ-FCP-NEXT:    vextractf128 $1, %ymm4, (%rcx)
; AVX512DQ-FCP-NEXT:    vextractf128 $1, %ymm2, (%r8)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i64_stride4_vf2:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovaps (%rdi), %xmm0
; AVX512BW-NEXT:    vmovaps 32(%rdi), %xmm1
; AVX512BW-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX512BW-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX512BW-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX512BW-NEXT:    vmovaps (%rdi), %ymm3
; AVX512BW-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
; AVX512BW-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm1[1],ymm3[3],ymm1[3]
; AVX512BW-NEXT:    vmovaps %xmm2, (%rsi)
; AVX512BW-NEXT:    vmovaps %xmm0, (%rdx)
; AVX512BW-NEXT:    vextractf128 $1, %ymm4, (%rcx)
; AVX512BW-NEXT:    vextractf128 $1, %ymm1, (%r8)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i64_stride4_vf2:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovaps {{.*#+}} xmm0 = [0,0,0,0,0,0,0,0,4,0,0,0,0,0,0,0]
; AVX512BW-FCP-NEXT:    vpermpd (%rdi), %zmm0, %zmm0
; AVX512BW-FCP-NEXT:    vmovaps (%rdi), %xmm1
; AVX512BW-FCP-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
; AVX512BW-FCP-NEXT:    vmovaps (%rdi), %ymm2
; AVX512BW-FCP-NEXT:    vmovaps 32(%rdi), %ymm3
; AVX512BW-FCP-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX512BW-FCP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
; AVX512BW-FCP-NEXT:    vmovaps %xmm0, (%rsi)
; AVX512BW-FCP-NEXT:    vmovaps %xmm1, (%rdx)
; AVX512BW-FCP-NEXT:    vextractf128 $1, %ymm4, (%rcx)
; AVX512BW-FCP-NEXT:    vextractf128 $1, %ymm2, (%r8)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i64_stride4_vf2:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovaps (%rdi), %xmm0
; AVX512DQ-BW-NEXT:    vmovaps 32(%rdi), %xmm1
; AVX512DQ-BW-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX512DQ-BW-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX512DQ-BW-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX512DQ-BW-NEXT:    vmovaps (%rdi), %ymm3
; AVX512DQ-BW-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
; AVX512DQ-BW-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm1[1],ymm3[3],ymm1[3]
; AVX512DQ-BW-NEXT:    vmovaps %xmm2, (%rsi)
; AVX512DQ-BW-NEXT:    vmovaps %xmm0, (%rdx)
; AVX512DQ-BW-NEXT:    vextractf128 $1, %ymm4, (%rcx)
; AVX512DQ-BW-NEXT:    vextractf128 $1, %ymm1, (%r8)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride4_vf2:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovaps {{.*#+}} xmm0 = [0,0,0,0,0,0,0,0,4,0,0,0,0,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermpd (%rdi), %zmm0, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovaps (%rdi), %xmm1
; AVX512DQ-BW-FCP-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
; AVX512DQ-BW-FCP-NEXT:    vmovaps (%rdi), %ymm2
; AVX512DQ-BW-FCP-NEXT:    vmovaps 32(%rdi), %ymm3
; AVX512DQ-BW-FCP-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX512DQ-BW-FCP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
; AVX512DQ-BW-FCP-NEXT:    vmovaps %xmm0, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovaps %xmm1, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vextractf128 $1, %ymm4, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vextractf128 $1, %ymm2, (%r8)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <8 x i64>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <8 x i64> %wide.vec, <8 x i64> poison, <2 x i32> <i32 0, i32 4>
  %strided.vec1 = shufflevector <8 x i64> %wide.vec, <8 x i64> poison, <2 x i32> <i32 1, i32 5>
  %strided.vec2 = shufflevector <8 x i64> %wide.vec, <8 x i64> poison, <2 x i32> <i32 2, i32 6>
  %strided.vec3 = shufflevector <8 x i64> %wide.vec, <8 x i64> poison, <2 x i32> <i32 3, i32 7>
  store <2 x i64> %strided.vec0, ptr %out.vec0, align 64
  store <2 x i64> %strided.vec1, ptr %out.vec1, align 64
  store <2 x i64> %strided.vec2, ptr %out.vec2, align 64
  store <2 x i64> %strided.vec3, ptr %out.vec3, align 64
  ret void
}

define void @load_i64_stride4_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: load_i64_stride4_vf4:
; SSE:       # %bb.0:
; SSE-NEXT:    movaps 112(%rdi), %xmm0
; SSE-NEXT:    movaps 80(%rdi), %xmm1
; SSE-NEXT:    movaps (%rdi), %xmm2
; SSE-NEXT:    movaps 16(%rdi), %xmm3
; SSE-NEXT:    movaps 32(%rdi), %xmm4
; SSE-NEXT:    movaps 48(%rdi), %xmm5
; SSE-NEXT:    movaps 96(%rdi), %xmm6
; SSE-NEXT:    movaps 64(%rdi), %xmm7
; SSE-NEXT:    movaps %xmm7, %xmm8
; SSE-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm6[0]
; SSE-NEXT:    movaps %xmm2, %xmm9
; SSE-NEXT:    movlhps {{.*#+}} xmm9 = xmm9[0],xmm4[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm6[1]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm4[1]
; SSE-NEXT:    movaps %xmm1, %xmm4
; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
; SSE-NEXT:    movaps %xmm3, %xmm6
; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm5[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm5[1]
; SSE-NEXT:    movaps %xmm8, 16(%rsi)
; SSE-NEXT:    movaps %xmm9, (%rsi)
; SSE-NEXT:    movaps %xmm7, 16(%rdx)
; SSE-NEXT:    movaps %xmm2, (%rdx)
; SSE-NEXT:    movaps %xmm4, 16(%rcx)
; SSE-NEXT:    movaps %xmm6, (%rcx)
; SSE-NEXT:    movaps %xmm1, 16(%r8)
; SSE-NEXT:    movaps %xmm3, (%r8)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i64_stride4_vf4:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovaps (%rdi), %ymm0
; AVX-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm2
; AVX-NEXT:    vinsertf128 $1, 96(%rdi), %ymm1, %ymm3
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],mem[2,3]
; AVX-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],mem[2,3]
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX-NEXT:    vmovaps %ymm4, (%rsi)
; AVX-NEXT:    vmovaps %ymm2, (%rdx)
; AVX-NEXT:    vmovaps %ymm5, (%rcx)
; AVX-NEXT:    vmovaps %ymm0, (%r8)
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i64_stride4_vf4:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovaps (%rdi), %ymm0
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-NEXT:    vmovaps 64(%rdi), %ymm2
; AVX2-NEXT:    vmovaps 96(%rdi), %ymm3
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm0[0,1],ymm2[0,1]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm1[0,1],ymm3[0,1]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-NEXT:    vmovaps %ymm2, (%rsi)
; AVX2-NEXT:    vmovaps %ymm4, (%rdx)
; AVX2-NEXT:    vmovaps %ymm3, (%rcx)
; AVX2-NEXT:    vmovaps %ymm0, (%r8)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i64_stride4_vf4:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm0
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovaps 64(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovaps 96(%rdi), %ymm3
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm0[0,1],ymm2[0,1]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm1[0,1],ymm3[0,1]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-FP-NEXT:    vmovaps %ymm2, (%rsi)
; AVX2-FP-NEXT:    vmovaps %ymm4, (%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm3, (%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm0, (%r8)
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i64_stride4_vf4:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovaps 64(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovaps 96(%rdi), %ymm3
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm0[0,1],ymm2[0,1]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm1[0,1],ymm3[0,1]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-FCP-NEXT:    vmovaps %ymm2, (%rsi)
; AVX2-FCP-NEXT:    vmovaps %ymm4, (%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm3, (%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm0, (%r8)
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i64_stride4_vf4:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa (%rdi), %ymm0
; AVX512-NEXT:    vmovdqa 32(%rdi), %ymm1
; AVX512-NEXT:    vmovdqa 64(%rdi), %ymm2
; AVX512-NEXT:    vmovdqa 96(%rdi), %ymm3
; AVX512-NEXT:    vperm2i128 {{.*#+}} ymm4 = ymm0[0,1],ymm2[0,1]
; AVX512-NEXT:    vperm2i128 {{.*#+}} ymm5 = ymm1[0,1],ymm3[0,1]
; AVX512-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX512-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
; AVX512-NEXT:    vpunpcklqdq {{.*#+}} ymm2 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX512-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX512-NEXT:    vpunpckhqdq {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX512-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX512-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512-NEXT:    vmovdqa %ymm4, (%rdx)
; AVX512-NEXT:    vmovdqa %ymm3, (%rcx)
; AVX512-NEXT:    vmovdqa %ymm0, (%r8)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i64_stride4_vf4:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovdqa (%rdi), %ymm0
; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %ymm1
; AVX512-FCP-NEXT:    vmovdqa 64(%rdi), %ymm2
; AVX512-FCP-NEXT:    vmovdqa 96(%rdi), %ymm3
; AVX512-FCP-NEXT:    vperm2i128 {{.*#+}} ymm4 = ymm0[0,1],ymm2[0,1]
; AVX512-FCP-NEXT:    vperm2i128 {{.*#+}} ymm5 = ymm1[0,1],ymm3[0,1]
; AVX512-FCP-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX512-FCP-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
; AVX512-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm2 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX512-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX512-FCP-NEXT:    vpunpckhqdq {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX512-FCP-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX512-FCP-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa %ymm4, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa %ymm3, (%rcx)
; AVX512-FCP-NEXT:    vmovdqa %ymm0, (%r8)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i64_stride4_vf4:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa (%rdi), %ymm0
; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %ymm1
; AVX512DQ-NEXT:    vmovdqa 64(%rdi), %ymm2
; AVX512DQ-NEXT:    vmovdqa 96(%rdi), %ymm3
; AVX512DQ-NEXT:    vperm2i128 {{.*#+}} ymm4 = ymm0[0,1],ymm2[0,1]
; AVX512DQ-NEXT:    vperm2i128 {{.*#+}} ymm5 = ymm1[0,1],ymm3[0,1]
; AVX512DQ-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX512DQ-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
; AVX512DQ-NEXT:    vpunpcklqdq {{.*#+}} ymm2 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX512DQ-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX512DQ-NEXT:    vpunpckhqdq {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX512DQ-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX512DQ-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512DQ-NEXT:    vmovdqa %ymm4, (%rdx)
; AVX512DQ-NEXT:    vmovdqa %ymm3, (%rcx)
; AVX512DQ-NEXT:    vmovdqa %ymm0, (%r8)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride4_vf4:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %ymm0
; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %ymm1
; AVX512DQ-FCP-NEXT:    vmovdqa 64(%rdi), %ymm2
; AVX512DQ-FCP-NEXT:    vmovdqa 96(%rdi), %ymm3
; AVX512DQ-FCP-NEXT:    vperm2i128 {{.*#+}} ymm4 = ymm0[0,1],ymm2[0,1]
; AVX512DQ-FCP-NEXT:    vperm2i128 {{.*#+}} ymm5 = ymm1[0,1],ymm3[0,1]
; AVX512DQ-FCP-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX512DQ-FCP-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
; AVX512DQ-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm2 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX512DQ-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX512DQ-FCP-NEXT:    vpunpckhqdq {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX512DQ-FCP-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm4, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm3, (%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, (%r8)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i64_stride4_vf4:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa (%rdi), %ymm0
; AVX512BW-NEXT:    vmovdqa 32(%rdi), %ymm1
; AVX512BW-NEXT:    vmovdqa 64(%rdi), %ymm2
; AVX512BW-NEXT:    vmovdqa 96(%rdi), %ymm3
; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm4 = ymm0[0,1],ymm2[0,1]
; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm5 = ymm1[0,1],ymm3[0,1]
; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX512BW-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm2 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX512BW-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX512BW-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX512BW-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512BW-NEXT:    vmovdqa %ymm4, (%rdx)
; AVX512BW-NEXT:    vmovdqa %ymm3, (%rcx)
; AVX512BW-NEXT:    vmovdqa %ymm0, (%r8)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i64_stride4_vf4:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovdqa (%rdi), %ymm0
; AVX512BW-FCP-NEXT:    vmovdqa 32(%rdi), %ymm1
; AVX512BW-FCP-NEXT:    vmovdqa 64(%rdi), %ymm2
; AVX512BW-FCP-NEXT:    vmovdqa 96(%rdi), %ymm3
; AVX512BW-FCP-NEXT:    vperm2i128 {{.*#+}} ymm4 = ymm0[0,1],ymm2[0,1]
; AVX512BW-FCP-NEXT:    vperm2i128 {{.*#+}} ymm5 = ymm1[0,1],ymm3[0,1]
; AVX512BW-FCP-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX512BW-FCP-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
; AVX512BW-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm2 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX512BW-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX512BW-FCP-NEXT:    vpunpckhqdq {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX512BW-FCP-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX512BW-FCP-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa %ymm4, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa %ymm3, (%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa %ymm0, (%r8)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i64_stride4_vf4:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovdqa (%rdi), %ymm0
; AVX512DQ-BW-NEXT:    vmovdqa 32(%rdi), %ymm1
; AVX512DQ-BW-NEXT:    vmovdqa 64(%rdi), %ymm2
; AVX512DQ-BW-NEXT:    vmovdqa 96(%rdi), %ymm3
; AVX512DQ-BW-NEXT:    vperm2i128 {{.*#+}} ymm4 = ymm0[0,1],ymm2[0,1]
; AVX512DQ-BW-NEXT:    vperm2i128 {{.*#+}} ymm5 = ymm1[0,1],ymm3[0,1]
; AVX512DQ-BW-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX512DQ-BW-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
; AVX512DQ-BW-NEXT:    vpunpcklqdq {{.*#+}} ymm2 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX512DQ-BW-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX512DQ-BW-NEXT:    vpunpckhqdq {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX512DQ-BW-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX512DQ-BW-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa %ymm4, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa %ymm3, (%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa %ymm0, (%r8)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride4_vf4:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovdqa (%rdi), %ymm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 32(%rdi), %ymm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 64(%rdi), %ymm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 96(%rdi), %ymm3
; AVX512DQ-BW-FCP-NEXT:    vperm2i128 {{.*#+}} ymm4 = ymm0[0,1],ymm2[0,1]
; AVX512DQ-BW-FCP-NEXT:    vperm2i128 {{.*#+}} ymm5 = ymm1[0,1],ymm3[0,1]
; AVX512DQ-BW-FCP-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX512DQ-BW-FCP-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
; AVX512DQ-BW-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm2 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX512DQ-BW-FCP-NEXT:    vpunpcklqdq {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX512DQ-BW-FCP-NEXT:    vpunpckhqdq {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX512DQ-BW-FCP-NEXT:    vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm4, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm3, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm0, (%r8)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <16 x i64>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
  %strided.vec1 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
  %strided.vec2 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
  %strided.vec3 = shufflevector <16 x i64> %wide.vec, <16 x i64> poison, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
  store <4 x i64> %strided.vec0, ptr %out.vec0, align 64
  store <4 x i64> %strided.vec1, ptr %out.vec1, align 64
  store <4 x i64> %strided.vec2, ptr %out.vec2, align 64
  store <4 x i64> %strided.vec3, ptr %out.vec3, align 64
  ret void
}

define void @load_i64_stride4_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: load_i64_stride4_vf8:
; SSE:       # %bb.0:
; SSE-NEXT:    movaps 240(%rdi), %xmm5
; SSE-NEXT:    movaps 208(%rdi), %xmm3
; SSE-NEXT:    movaps 176(%rdi), %xmm9
; SSE-NEXT:    movaps 144(%rdi), %xmm1
; SSE-NEXT:    movaps 112(%rdi), %xmm11
; SSE-NEXT:    movaps 80(%rdi), %xmm2
; SSE-NEXT:    movaps (%rdi), %xmm4
; SSE-NEXT:    movaps 16(%rdi), %xmm0
; SSE-NEXT:    movaps 32(%rdi), %xmm13
; SSE-NEXT:    movaps 224(%rdi), %xmm14
; SSE-NEXT:    movaps 192(%rdi), %xmm7
; SSE-NEXT:    movaps 160(%rdi), %xmm15
; SSE-NEXT:    movaps 128(%rdi), %xmm6
; SSE-NEXT:    movaps 96(%rdi), %xmm12
; SSE-NEXT:    movaps 64(%rdi), %xmm8
; SSE-NEXT:    movaps %xmm8, %xmm10
; SSE-NEXT:    movlhps {{.*#+}} xmm10 = xmm10[0],xmm12[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm12[1]
; SSE-NEXT:    movaps %xmm6, %xmm12
; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm15[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm15[1]
; SSE-NEXT:    movaps %xmm7, %xmm15
; SSE-NEXT:    movlhps {{.*#+}} xmm15 = xmm15[0],xmm14[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm14[1]
; SSE-NEXT:    movaps %xmm4, %xmm14
; SSE-NEXT:    movlhps {{.*#+}} xmm14 = xmm14[0],xmm13[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm13[1]
; SSE-NEXT:    movaps %xmm2, %xmm13
; SSE-NEXT:    movlhps {{.*#+}} xmm13 = xmm13[0],xmm11[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm11[1]
; SSE-NEXT:    movaps %xmm1, %xmm11
; SSE-NEXT:    movlhps {{.*#+}} xmm11 = xmm11[0],xmm9[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm9[1]
; SSE-NEXT:    movaps %xmm3, %xmm9
; SSE-NEXT:    movlhps {{.*#+}} xmm9 = xmm9[0],xmm5[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm5[1]
; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 48(%rdi), %xmm5
; SSE-NEXT:    movaps %xmm0, %xmm3
; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm5[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm5[1]
; SSE-NEXT:    movaps %xmm15, 48(%rsi)
; SSE-NEXT:    movaps %xmm14, (%rsi)
; SSE-NEXT:    movaps %xmm12, 32(%rsi)
; SSE-NEXT:    movaps %xmm10, 16(%rsi)
; SSE-NEXT:    movaps %xmm7, 48(%rdx)
; SSE-NEXT:    movaps %xmm4, (%rdx)
; SSE-NEXT:    movaps %xmm6, 32(%rdx)
; SSE-NEXT:    movaps %xmm8, 16(%rdx)
; SSE-NEXT:    movaps %xmm9, 48(%rcx)
; SSE-NEXT:    movaps %xmm11, 32(%rcx)
; SSE-NEXT:    movaps %xmm13, 16(%rcx)
; SSE-NEXT:    movaps %xmm3, (%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT:    movaps %xmm3, 48(%r8)
; SSE-NEXT:    movaps %xmm1, 32(%r8)
; SSE-NEXT:    movaps %xmm2, 16(%r8)
; SSE-NEXT:    movaps %xmm0, (%r8)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i64_stride4_vf8:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovaps 224(%rdi), %ymm3
; AVX-NEXT:    vmovaps 192(%rdi), %ymm2
; AVX-NEXT:    vmovaps 96(%rdi), %ymm4
; AVX-NEXT:    vmovaps 64(%rdi), %ymm5
; AVX-NEXT:    vmovaps 96(%rdi), %xmm6
; AVX-NEXT:    vmovaps 64(%rdi), %xmm7
; AVX-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm7[0],xmm6[0]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 32(%rdi), %xmm8
; AVX-NEXT:    vmovaps (%rdi), %xmm9
; AVX-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm9[0],xmm8[0]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 224(%rdi), %xmm10
; AVX-NEXT:    vmovaps 192(%rdi), %xmm11
; AVX-NEXT:    vmovlhps {{.*#+}} xmm12 = xmm11[0],xmm10[0]
; AVX-NEXT:    vmovaps 160(%rdi), %xmm13
; AVX-NEXT:    vmovaps 128(%rdi), %xmm14
; AVX-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm14[0],xmm13[0]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm7[1],xmm6[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm9[1],xmm8[1]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm11[1],xmm10[1]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm9 = xmm14[1],xmm13[1]
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX-NEXT:    vmovaps 48(%rdi), %xmm11
; AVX-NEXT:    vmovaps 16(%rdi), %xmm13
; AVX-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm13[0],xmm11[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm10 = ymm14[0,1,2,3],ymm10[4,5,6,7]
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX-NEXT:    vmovaps 176(%rdi), %xmm1
; AVX-NEXT:    vmovaps 144(%rdi), %xmm0
; AVX-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm0[0],xmm1[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm14[4,5,6,7]
; AVX-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm13[1],xmm11[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
; AVX-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX-NEXT:    vmovaps %xmm15, 32(%rsi)
; AVX-NEXT:    vmovaps %xmm12, 48(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, (%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 16(%rsi)
; AVX-NEXT:    vmovaps %xmm9, 32(%rdx)
; AVX-NEXT:    vmovaps %xmm8, 48(%rdx)
; AVX-NEXT:    vmovaps %xmm7, (%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 16(%rdx)
; AVX-NEXT:    vmovaps %ymm6, 32(%rcx)
; AVX-NEXT:    vmovaps %ymm10, (%rcx)
; AVX-NEXT:    vmovaps %ymm0, 32(%r8)
; AVX-NEXT:    vmovaps %ymm4, (%r8)
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i64_stride4_vf8:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovaps 160(%rdi), %ymm0
; AVX2-NEXT:    vmovaps 128(%rdi), %ymm1
; AVX2-NEXT:    vmovaps 224(%rdi), %ymm2
; AVX2-NEXT:    vmovaps 192(%rdi), %ymm3
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm4
; AVX2-NEXT:    vmovaps (%rdi), %ymm5
; AVX2-NEXT:    vmovaps 96(%rdi), %ymm6
; AVX2-NEXT:    vmovaps 64(%rdi), %ymm7
; AVX2-NEXT:    vmovaps (%rdi), %xmm8
; AVX2-NEXT:    vmovaps 32(%rdi), %xmm9
; AVX2-NEXT:    vinsertf128 $1, 96(%rdi), %ymm9, %ymm9
; AVX2-NEXT:    vinsertf128 $1, 64(%rdi), %ymm8, %ymm8
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm8[0],ymm9[0],ymm8[2],ymm9[2]
; AVX2-NEXT:    vmovaps 160(%rdi), %xmm11
; AVX2-NEXT:    vinsertf128 $1, 224(%rdi), %ymm11, %ymm11
; AVX2-NEXT:    vmovaps 128(%rdi), %xmm12
; AVX2-NEXT:    vinsertf128 $1, 192(%rdi), %ymm12, %ymm12
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm12[0],ymm11[0],ymm12[2],ymm11[2]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm8[1],ymm9[1],ymm8[3],ymm9[3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm9 = ymm12[1],ymm11[1],ymm12[3],ymm11[3]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm11 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm12[2,3],ymm11[2,3]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm14[2,3],ymm12[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm6 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm6[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX2-NEXT:    vmovaps %ymm13, 32(%rsi)
; AVX2-NEXT:    vmovaps %ymm10, (%rsi)
; AVX2-NEXT:    vmovaps %ymm9, 32(%rdx)
; AVX2-NEXT:    vmovaps %ymm8, (%rdx)
; AVX2-NEXT:    vmovaps %ymm12, 32(%rcx)
; AVX2-NEXT:    vmovaps %ymm11, (%rcx)
; AVX2-NEXT:    vmovaps %ymm0, 32(%r8)
; AVX2-NEXT:    vmovaps %ymm4, (%r8)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i64_stride4_vf8:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    vmovaps 160(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovaps 224(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovaps 192(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm4
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm5
; AVX2-FP-NEXT:    vmovaps 96(%rdi), %ymm6
; AVX2-FP-NEXT:    vmovaps 64(%rdi), %ymm7
; AVX2-FP-NEXT:    vmovaps (%rdi), %xmm8
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %xmm9
; AVX2-FP-NEXT:    vinsertf128 $1, 96(%rdi), %ymm9, %ymm9
; AVX2-FP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm8, %ymm8
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm8[0],ymm9[0],ymm8[2],ymm9[2]
; AVX2-FP-NEXT:    vmovaps 160(%rdi), %xmm11
; AVX2-FP-NEXT:    vinsertf128 $1, 224(%rdi), %ymm11, %ymm11
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %xmm12
; AVX2-FP-NEXT:    vinsertf128 $1, 192(%rdi), %ymm12, %ymm12
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm12[0],ymm11[0],ymm12[2],ymm11[2]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm8[1],ymm9[1],ymm8[3],ymm9[3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm9 = ymm12[1],ymm11[1],ymm12[3],ymm11[3]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm11 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm12[2,3],ymm11[2,3]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm14[2,3],ymm12[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm6 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm6[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX2-FP-NEXT:    vmovaps %ymm13, 32(%rsi)
; AVX2-FP-NEXT:    vmovaps %ymm10, (%rsi)
; AVX2-FP-NEXT:    vmovaps %ymm9, 32(%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm8, (%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm12, 32(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm11, (%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm0, 32(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm4, (%r8)
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i64_stride4_vf8:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    vmovaps 160(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovaps 224(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovaps 192(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm4
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm5
; AVX2-FCP-NEXT:    vmovaps 96(%rdi), %ymm6
; AVX2-FCP-NEXT:    vmovaps 64(%rdi), %ymm7
; AVX2-FCP-NEXT:    vmovaps (%rdi), %xmm8
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %xmm9
; AVX2-FCP-NEXT:    vinsertf128 $1, 96(%rdi), %ymm9, %ymm9
; AVX2-FCP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm8, %ymm8
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm8[0],ymm9[0],ymm8[2],ymm9[2]
; AVX2-FCP-NEXT:    vmovaps 160(%rdi), %xmm11
; AVX2-FCP-NEXT:    vinsertf128 $1, 224(%rdi), %ymm11, %ymm11
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %xmm12
; AVX2-FCP-NEXT:    vinsertf128 $1, 192(%rdi), %ymm12, %ymm12
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm12[0],ymm11[0],ymm12[2],ymm11[2]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm8[1],ymm9[1],ymm8[3],ymm9[3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm9 = ymm12[1],ymm11[1],ymm12[3],ymm11[3]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm11 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm12[2,3],ymm11[2,3]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm14[2,3],ymm12[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm6 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm6[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
; AVX2-FCP-NEXT:    vmovaps %ymm13, 32(%rsi)
; AVX2-FCP-NEXT:    vmovaps %ymm10, (%rsi)
; AVX2-FCP-NEXT:    vmovaps %ymm9, 32(%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm8, (%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm12, 32(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm11, (%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm0, 32(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm4, (%r8)
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i64_stride4_vf8:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [0,4,8,12,0,4,8,12]
; AVX512-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512-NEXT:    vpermt2q %zmm3, %zmm4, %zmm5
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7]
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [1,5,9,13,1,5,9,13]
; AVX512-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm6
; AVX512-NEXT:    vpermt2q %zmm3, %zmm5, %zmm6
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm5
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7]
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [2,6,10,14,2,6,10,14]
; AVX512-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512-NEXT:    vpermt2q %zmm3, %zmm6, %zmm7
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm6
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7]
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [3,7,11,15,3,7,11,15]
; AVX512-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermt2q %zmm3, %zmm7, %zmm2
; AVX512-NEXT:    vpermt2q %zmm1, %zmm7, %zmm0
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm6, (%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i64_stride4_vf8:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [0,4,8,12,0,4,8,12]
; AVX512-FCP-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm4, %zmm5
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7]
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [1,5,9,13,1,5,9,13]
; AVX512-FCP-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm6
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm5, %zmm6
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm5
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7]
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [2,6,10,14,2,6,10,14]
; AVX512-FCP-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm6, %zmm7
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm6
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7]
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [3,7,11,15,3,7,11,15]
; AVX512-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm7, %zmm2
; AVX512-FCP-NEXT:    vpermt2q %zmm1, %zmm7, %zmm0
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm6, (%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i64_stride4_vf8:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [0,4,8,12,0,4,8,12]
; AVX512DQ-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm4, %zmm5
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7]
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [1,5,9,13,1,5,9,13]
; AVX512DQ-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm6
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm5, %zmm6
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm5
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7]
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [2,6,10,14,2,6,10,14]
; AVX512DQ-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm6, %zmm7
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm6
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [3,7,11,15,3,7,11,15]
; AVX512DQ-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm7, %zmm2
; AVX512DQ-NEXT:    vpermt2q %zmm1, %zmm7, %zmm0
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm6, (%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride4_vf8:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [0,4,8,12,0,4,8,12]
; AVX512DQ-FCP-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm4, %zmm5
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [1,5,9,13,1,5,9,13]
; AVX512DQ-FCP-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm6
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm5, %zmm6
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm5
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [2,6,10,14,2,6,10,14]
; AVX512DQ-FCP-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm6, %zmm7
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm6
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [3,7,11,15,3,7,11,15]
; AVX512DQ-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm7, %zmm2
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm1, %zmm7, %zmm0
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm6, (%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i64_stride4_vf8:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [0,4,8,12,0,4,8,12]
; AVX512BW-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm4, %zmm5
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [1,5,9,13,1,5,9,13]
; AVX512BW-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm6
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm5, %zmm6
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm5
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [2,6,10,14,2,6,10,14]
; AVX512BW-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm6, %zmm7
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm6
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [3,7,11,15,3,7,11,15]
; AVX512BW-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm7, %zmm2
; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm7, %zmm0
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm6, (%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i64_stride4_vf8:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [0,4,8,12,0,4,8,12]
; AVX512BW-FCP-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm4, %zmm5
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7]
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [1,5,9,13,1,5,9,13]
; AVX512BW-FCP-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm6
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm5, %zmm6
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm5
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7]
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [2,6,10,14,2,6,10,14]
; AVX512BW-FCP-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm6, %zmm7
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm6
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7]
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [3,7,11,15,3,7,11,15]
; AVX512BW-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm7, %zmm2
; AVX512BW-FCP-NEXT:    vpermt2q %zmm1, %zmm7, %zmm0
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm6, (%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i64_stride4_vf8:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [0,4,8,12,0,4,8,12]
; AVX512DQ-BW-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm4, %zmm5
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7]
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [1,5,9,13,1,5,9,13]
; AVX512DQ-BW-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm6
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm5, %zmm6
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm5
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7]
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [2,6,10,14,2,6,10,14]
; AVX512DQ-BW-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm6, %zmm7
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm6
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [3,7,11,15,3,7,11,15]
; AVX512DQ-BW-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm7, %zmm2
; AVX512DQ-BW-NEXT:    vpermt2q %zmm1, %zmm7, %zmm0
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm6, (%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride4_vf8:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm4 = [0,4,8,12,0,4,8,12]
; AVX512DQ-BW-FCP-NEXT:    # zmm4 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm4, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm5 = [1,5,9,13,1,5,9,13]
; AVX512DQ-BW-FCP-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm5, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm6 = [2,6,10,14,2,6,10,14]
; AVX512DQ-BW-FCP-NEXT:    # zmm6 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm6, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm7 = [3,7,11,15,3,7,11,15]
; AVX512DQ-BW-FCP-NEXT:    # zmm7 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm7, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm1, %zmm7, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm6, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <32 x i64>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
  %strided.vec1 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
  %strided.vec2 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
  %strided.vec3 = shufflevector <32 x i64> %wide.vec, <32 x i64> poison, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
  store <8 x i64> %strided.vec0, ptr %out.vec0, align 64
  store <8 x i64> %strided.vec1, ptr %out.vec1, align 64
  store <8 x i64> %strided.vec2, ptr %out.vec2, align 64
  store <8 x i64> %strided.vec3, ptr %out.vec3, align 64
  ret void
}

define void @load_i64_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: load_i64_stride4_vf16:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $152, %rsp
; SSE-NEXT:    movaps (%rdi), %xmm7
; SSE-NEXT:    movaps 416(%rdi), %xmm0
; SSE-NEXT:    movaps 384(%rdi), %xmm8
; SSE-NEXT:    movaps 160(%rdi), %xmm1
; SSE-NEXT:    movaps 128(%rdi), %xmm9
; SSE-NEXT:    movaps 480(%rdi), %xmm2
; SSE-NEXT:    movaps 448(%rdi), %xmm10
; SSE-NEXT:    movaps 224(%rdi), %xmm3
; SSE-NEXT:    movaps 192(%rdi), %xmm11
; SSE-NEXT:    movaps 288(%rdi), %xmm4
; SSE-NEXT:    movaps 256(%rdi), %xmm13
; SSE-NEXT:    movaps 352(%rdi), %xmm5
; SSE-NEXT:    movaps 320(%rdi), %xmm12
; SSE-NEXT:    movaps 96(%rdi), %xmm6
; SSE-NEXT:    movaps 64(%rdi), %xmm14
; SSE-NEXT:    movaps %xmm14, %xmm15
; SSE-NEXT:    movlhps {{.*#+}} xmm15 = xmm15[0],xmm6[0]
; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm6[1]
; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm12, %xmm6
; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm5[0]
; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm5[1]
; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm13, %xmm5
; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm4[0]
; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm4[1]
; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm11, %xmm4
; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm3[0]
; SSE-NEXT:    movaps %xmm4, (%rsp) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm3[1]
; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm10, %xmm3
; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm2[1]
; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm9, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm1[1]
; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm8, %xmm1
; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 32(%rdi), %xmm0
; SSE-NEXT:    movaps %xmm7, %xmm1
; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm0[1]
; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 112(%rdi), %xmm0
; SSE-NEXT:    movaps 80(%rdi), %xmm15
; SSE-NEXT:    movaps %xmm15, %xmm1
; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm0[1]
; SSE-NEXT:    movaps 176(%rdi), %xmm0
; SSE-NEXT:    movaps 144(%rdi), %xmm11
; SSE-NEXT:    movaps %xmm11, %xmm14
; SSE-NEXT:    movlhps {{.*#+}} xmm14 = xmm14[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm0[1]
; SSE-NEXT:    movaps 240(%rdi), %xmm0
; SSE-NEXT:    movaps 208(%rdi), %xmm12
; SSE-NEXT:    movaps %xmm12, %xmm13
; SSE-NEXT:    movlhps {{.*#+}} xmm13 = xmm13[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm0[1]
; SSE-NEXT:    movaps 304(%rdi), %xmm0
; SSE-NEXT:    movaps 272(%rdi), %xmm7
; SSE-NEXT:    movaps %xmm7, %xmm10
; SSE-NEXT:    movlhps {{.*#+}} xmm10 = xmm10[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm0[1]
; SSE-NEXT:    movaps 368(%rdi), %xmm0
; SSE-NEXT:    movaps 336(%rdi), %xmm8
; SSE-NEXT:    movaps %xmm8, %xmm9
; SSE-NEXT:    movlhps {{.*#+}} xmm9 = xmm9[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
; SSE-NEXT:    movaps 432(%rdi), %xmm0
; SSE-NEXT:    movaps 400(%rdi), %xmm5
; SSE-NEXT:    movaps %xmm5, %xmm6
; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm0[1]
; SSE-NEXT:    movaps 496(%rdi), %xmm1
; SSE-NEXT:    movaps 464(%rdi), %xmm2
; SSE-NEXT:    movaps %xmm2, %xmm4
; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm1[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm1[1]
; SSE-NEXT:    movaps 16(%rdi), %xmm1
; SSE-NEXT:    movaps 48(%rdi), %xmm0
; SSE-NEXT:    movaps %xmm1, %xmm3
; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[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 (%rsp), %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 %xmm6, 96(%rcx)
; SSE-NEXT:    movaps %xmm14, 32(%rcx)
; SSE-NEXT:    movaps %xmm4, 112(%rcx)
; SSE-NEXT:    movaps %xmm13, 48(%rcx)
; SSE-NEXT:    movaps %xmm10, 64(%rcx)
; SSE-NEXT:    movaps %xmm3, (%rcx)
; SSE-NEXT:    movaps %xmm9, 80(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rcx)
; SSE-NEXT:    movaps %xmm2, 112(%r8)
; SSE-NEXT:    movaps %xmm5, 96(%r8)
; SSE-NEXT:    movaps %xmm8, 80(%r8)
; SSE-NEXT:    movaps %xmm7, 64(%r8)
; SSE-NEXT:    movaps %xmm12, 48(%r8)
; SSE-NEXT:    movaps %xmm11, 32(%r8)
; SSE-NEXT:    movaps %xmm15, 16(%r8)
; SSE-NEXT:    movaps %xmm1, (%r8)
; SSE-NEXT:    addq $152, %rsp
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i64_stride4_vf16:
; AVX:       # %bb.0:
; AVX-NEXT:    subq $296, %rsp # imm = 0x128
; AVX-NEXT:    vmovaps 224(%rdi), %xmm0
; AVX-NEXT:    vmovaps 192(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 160(%rdi), %xmm2
; AVX-NEXT:    vmovaps 128(%rdi), %xmm3
; AVX-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
; AVX-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 416(%rdi), %xmm4
; AVX-NEXT:    vmovaps 384(%rdi), %xmm8
; AVX-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm8[0],xmm4[0]
; AVX-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 480(%rdi), %xmm7
; AVX-NEXT:    vmovaps 448(%rdi), %xmm9
; AVX-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm9[0],xmm7[0]
; AVX-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 352(%rdi), %xmm10
; AVX-NEXT:    vmovaps 320(%rdi), %xmm12
; AVX-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm12[0],xmm10[0]
; AVX-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 288(%rdi), %xmm13
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 256(%rdi), %xmm0
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm2[1]
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm13[0]
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm9[1],xmm7[1]
; AVX-NEXT:    vmovaps %xmm1, (%rsp) # 16-byte Spill
; AVX-NEXT:    vmovaps 32(%rdi), %xmm1
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm8[1],xmm4[1]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 96(%rdi), %xmm2
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm12[1],xmm10[1]
; AVX-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 64(%rdi), %xmm3
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm13[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm3[0],xmm2[0]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps (%rdi), %xmm0
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 224(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 192(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX-NEXT:    vmovaps 176(%rdi), %xmm14
; AVX-NEXT:    vmovaps 144(%rdi), %xmm13
; AVX-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm13[0],xmm14[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 480(%rdi), %ymm12
; AVX-NEXT:    vmovaps 448(%rdi), %ymm11
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
; AVX-NEXT:    vmovaps 432(%rdi), %xmm10
; AVX-NEXT:    vmovaps 400(%rdi), %xmm9
; AVX-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm9[0],xmm10[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 96(%rdi), %ymm8
; AVX-NEXT:    vmovaps 64(%rdi), %ymm7
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
; AVX-NEXT:    vmovaps 48(%rdi), %xmm6
; AVX-NEXT:    vmovaps 16(%rdi), %xmm4
; AVX-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm4[0],xmm6[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 352(%rdi), %ymm3
; AVX-NEXT:    vmovaps 320(%rdi), %ymm2
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX-NEXT:    vmovaps 304(%rdi), %xmm1
; AVX-NEXT:    vmovaps 272(%rdi), %xmm0
; AVX-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm0[0],xmm1[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm5 = ymm15[0,1,2,3],ymm5[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm13 = xmm13[1],xmm14[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm15[4,5,6,7]
; AVX-NEXT:    vunpckhpd {{.*#+}} ymm11 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm9 = xmm9[1],xmm10[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm11[4,5,6,7]
; AVX-NEXT:    vunpckhpd {{.*#+}} ymm7 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm4[1],xmm6[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm7[4,5,6,7]
; AVX-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 16(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, (%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 64(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 80(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 112(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 96(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 32(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 48(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, (%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 16(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 64(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 80(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 96(%rdx)
; AVX-NEXT:    vmovaps (%rsp), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 112(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 32(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 48(%rdx)
; AVX-NEXT:    vmovaps %ymm5, 64(%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, 96(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, 32(%rcx)
; AVX-NEXT:    vmovaps %ymm0, 64(%r8)
; AVX-NEXT:    vmovaps %ymm4, (%r8)
; AVX-NEXT:    vmovaps %ymm9, 96(%r8)
; AVX-NEXT:    vmovaps %ymm13, 32(%r8)
; AVX-NEXT:    addq $296, %rsp # imm = 0x128
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i64_stride4_vf16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    subq $360, %rsp # imm = 0x168
; AVX2-NEXT:    vmovaps 416(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 384(%rdi), %ymm4
; AVX2-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 160(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 128(%rdi), %ymm6
; AVX2-NEXT:    vmovups %ymm6, {{[-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), %ymm5
; AVX2-NEXT:    vmovaps 160(%rdi), %xmm2
; AVX2-NEXT:    vinsertf128 $1, 224(%rdi), %ymm2, %ymm9
; AVX2-NEXT:    vmovaps 128(%rdi), %xmm2
; AVX2-NEXT:    vinsertf128 $1, 192(%rdi), %ymm2, %ymm10
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm9[0],ymm10[2],ymm9[2]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 416(%rdi), %xmm7
; AVX2-NEXT:    vinsertf128 $1, 480(%rdi), %ymm7, %ymm11
; AVX2-NEXT:    vmovaps 384(%rdi), %xmm7
; AVX2-NEXT:    vinsertf128 $1, 448(%rdi), %ymm7, %ymm12
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm12[0],ymm11[0],ymm12[2],ymm11[2]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps (%rdi), %xmm8
; AVX2-NEXT:    vmovaps 32(%rdi), %xmm13
; AVX2-NEXT:    vinsertf128 $1, 96(%rdi), %ymm13, %ymm13
; AVX2-NEXT:    vinsertf128 $1, 64(%rdi), %ymm8, %ymm15
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm15[0],ymm13[0],ymm15[2],ymm13[2]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 288(%rdi), %xmm14
; AVX2-NEXT:    vinsertf128 $1, 352(%rdi), %ymm14, %ymm1
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm9[1],ymm10[3],ymm9[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 256(%rdi), %xmm10
; AVX2-NEXT:    vinsertf128 $1, 320(%rdi), %ymm10, %ymm0
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm12[1],ymm11[1],ymm12[3],ymm11[3]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm15[1],ymm13[1],ymm15[3],ymm13[3]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; 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:    vmovaps %ymm5, %ymm10
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm14[0],ymm5[2],ymm14[2]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm6[0],ymm13[0],ymm6[2],ymm13[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 480(%rdi), %ymm11
; AVX2-NEXT:    vmovaps 448(%rdi), %ymm8
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm8[0],ymm11[0],ymm8[2],ymm11[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm2[2,3],ymm15[2,3]
; AVX2-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm9
; AVX2-NEXT:    vmovaps (%rdi), %ymm6
; AVX2-NEXT:    vmovaps 96(%rdi), %ymm5
; AVX2-NEXT:    vmovaps 64(%rdi), %ymm4
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm6[0],ymm9[0],ymm6[2],ymm9[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm7[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovaps 288(%rdi), %ymm7
; AVX2-NEXT:    vmovaps 256(%rdi), %ymm3
; AVX2-NEXT:    vmovaps 352(%rdi), %ymm1
; AVX2-NEXT:    vmovaps 320(%rdi), %ymm0
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm3[0],ymm7[0],ymm3[2],ymm7[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm12[2,3],ymm2[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm12 = ymm10[1],ymm14[1],ymm10[3],ymm14[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm13 = ymm10[1],ymm13[1],ymm10[3],ymm13[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm13[2,3],ymm12[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm8[1],ymm11[1],ymm8[3],ymm11[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm10 = ymm10[1],mem[1],ymm10[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm10[2,3],ymm8[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm6[1],ymm9[1],ymm6[3],ymm9[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm5[2,3],ymm4[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm7[1],ymm3[3],ymm7[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; 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, 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, 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, 96(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 32(%rdx)
; AVX2-NEXT:    vmovaps %ymm2, 64(%rcx)
; AVX2-NEXT:    vmovaps %ymm15, (%rcx)
; AVX2-NEXT:    vmovups (%rsp), %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:    vmovaps %ymm0, 64(%r8)
; AVX2-NEXT:    vmovaps %ymm4, (%r8)
; AVX2-NEXT:    vmovaps %ymm8, 96(%r8)
; AVX2-NEXT:    vmovaps %ymm12, 32(%r8)
; AVX2-NEXT:    addq $360, %rsp # imm = 0x168
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i64_stride4_vf16:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    subq $360, %rsp # imm = 0x168
; AVX2-FP-NEXT:    vmovaps 416(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 384(%rdi), %ymm4
; AVX2-FP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 160(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %ymm6
; AVX2-FP-NEXT:    vmovups %ymm6, {{[-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), %ymm5
; AVX2-FP-NEXT:    vmovaps 160(%rdi), %xmm2
; AVX2-FP-NEXT:    vinsertf128 $1, 224(%rdi), %ymm2, %ymm9
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %xmm2
; AVX2-FP-NEXT:    vinsertf128 $1, 192(%rdi), %ymm2, %ymm10
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm9[0],ymm10[2],ymm9[2]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 416(%rdi), %xmm7
; AVX2-FP-NEXT:    vinsertf128 $1, 480(%rdi), %ymm7, %ymm11
; AVX2-FP-NEXT:    vmovaps 384(%rdi), %xmm7
; AVX2-FP-NEXT:    vinsertf128 $1, 448(%rdi), %ymm7, %ymm12
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm12[0],ymm11[0],ymm12[2],ymm11[2]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps (%rdi), %xmm8
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %xmm13
; AVX2-FP-NEXT:    vinsertf128 $1, 96(%rdi), %ymm13, %ymm13
; AVX2-FP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm8, %ymm15
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm15[0],ymm13[0],ymm15[2],ymm13[2]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 288(%rdi), %xmm14
; AVX2-FP-NEXT:    vinsertf128 $1, 352(%rdi), %ymm14, %ymm1
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm9[1],ymm10[3],ymm9[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 256(%rdi), %xmm10
; AVX2-FP-NEXT:    vinsertf128 $1, 320(%rdi), %ymm10, %ymm0
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm12[1],ymm11[1],ymm12[3],ymm11[3]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm15[1],ymm13[1],ymm15[3],ymm13[3]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; 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:    vmovaps %ymm5, %ymm10
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm14[0],ymm5[2],ymm14[2]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm6[0],ymm13[0],ymm6[2],ymm13[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 480(%rdi), %ymm11
; AVX2-FP-NEXT:    vmovaps 448(%rdi), %ymm8
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm8[0],ymm11[0],ymm8[2],ymm11[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm2[2,3],ymm15[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm9
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm6
; AVX2-FP-NEXT:    vmovaps 96(%rdi), %ymm5
; AVX2-FP-NEXT:    vmovaps 64(%rdi), %ymm4
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm6[0],ymm9[0],ymm6[2],ymm9[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm7[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovaps 288(%rdi), %ymm7
; AVX2-FP-NEXT:    vmovaps 256(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovaps 352(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovaps 320(%rdi), %ymm0
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm3[0],ymm7[0],ymm3[2],ymm7[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm12[2,3],ymm2[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm12 = ymm10[1],ymm14[1],ymm10[3],ymm14[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm13 = ymm10[1],ymm13[1],ymm10[3],ymm13[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm13[2,3],ymm12[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm8[1],ymm11[1],ymm8[3],ymm11[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm10 = ymm10[1],mem[1],ymm10[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm10[2,3],ymm8[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm6[1],ymm9[1],ymm6[3],ymm9[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm5[2,3],ymm4[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm7[1],ymm3[3],ymm7[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; 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, 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, 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, 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:    vmovaps %ymm2, 64(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm15, (%rcx)
; AVX2-FP-NEXT:    vmovups (%rsp), %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:    vmovaps %ymm0, 64(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm4, (%r8)
; AVX2-FP-NEXT:    vmovaps %ymm8, 96(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm12, 32(%r8)
; AVX2-FP-NEXT:    addq $360, %rsp # imm = 0x168
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i64_stride4_vf16:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    subq $360, %rsp # imm = 0x168
; AVX2-FCP-NEXT:    vmovaps 416(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 384(%rdi), %ymm4
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 160(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %ymm6
; AVX2-FCP-NEXT:    vmovups %ymm6, {{[-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), %ymm5
; AVX2-FCP-NEXT:    vmovaps 160(%rdi), %xmm2
; AVX2-FCP-NEXT:    vinsertf128 $1, 224(%rdi), %ymm2, %ymm9
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %xmm2
; AVX2-FCP-NEXT:    vinsertf128 $1, 192(%rdi), %ymm2, %ymm10
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm9[0],ymm10[2],ymm9[2]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 416(%rdi), %xmm7
; AVX2-FCP-NEXT:    vinsertf128 $1, 480(%rdi), %ymm7, %ymm11
; AVX2-FCP-NEXT:    vmovaps 384(%rdi), %xmm7
; AVX2-FCP-NEXT:    vinsertf128 $1, 448(%rdi), %ymm7, %ymm12
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm12[0],ymm11[0],ymm12[2],ymm11[2]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps (%rdi), %xmm8
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %xmm13
; AVX2-FCP-NEXT:    vinsertf128 $1, 96(%rdi), %ymm13, %ymm13
; AVX2-FCP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm8, %ymm15
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm15[0],ymm13[0],ymm15[2],ymm13[2]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 288(%rdi), %xmm14
; AVX2-FCP-NEXT:    vinsertf128 $1, 352(%rdi), %ymm14, %ymm1
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm9[1],ymm10[3],ymm9[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 256(%rdi), %xmm10
; AVX2-FCP-NEXT:    vinsertf128 $1, 320(%rdi), %ymm10, %ymm0
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm12[1],ymm11[1],ymm12[3],ymm11[3]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm15[1],ymm13[1],ymm15[3],ymm13[3]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm5, %ymm10
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm14[0],ymm5[2],ymm14[2]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm6[0],ymm13[0],ymm6[2],ymm13[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 480(%rdi), %ymm11
; AVX2-FCP-NEXT:    vmovaps 448(%rdi), %ymm8
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm8[0],ymm11[0],ymm8[2],ymm11[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm2[2,3],ymm15[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm9
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm6
; AVX2-FCP-NEXT:    vmovaps 96(%rdi), %ymm5
; AVX2-FCP-NEXT:    vmovaps 64(%rdi), %ymm4
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm6[0],ymm9[0],ymm6[2],ymm9[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm7[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovaps 288(%rdi), %ymm7
; AVX2-FCP-NEXT:    vmovaps 256(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovaps 352(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovaps 320(%rdi), %ymm0
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm3[0],ymm7[0],ymm3[2],ymm7[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm12[2,3],ymm2[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm12 = ymm10[1],ymm14[1],ymm10[3],ymm14[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm13 = ymm10[1],ymm13[1],ymm10[3],ymm13[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm13[2,3],ymm12[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm8[1],ymm11[1],ymm8[3],ymm11[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm10 = ymm10[1],mem[1],ymm10[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm10[2,3],ymm8[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm6[1],ymm9[1],ymm6[3],ymm9[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm5[2,3],ymm4[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm7[1],ymm3[3],ymm7[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; 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, 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, 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, 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:    vmovaps %ymm2, 64(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm15, (%rcx)
; AVX2-FCP-NEXT:    vmovups (%rsp), %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:    vmovaps %ymm0, 64(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm4, (%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm8, 96(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm12, 32(%r8)
; AVX2-FCP-NEXT:    addq $360, %rsp # imm = 0x168
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i64_stride4_vf16:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512-NEXT:    vmovdqa64 320(%rdi), %zmm4
; AVX512-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512-NEXT:    vmovdqa64 448(%rdi), %zmm6
; AVX512-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,0,4,8,12]
; AVX512-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512-NEXT:    vpermt2q %zmm6, %zmm8, %zmm9
; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm10
; AVX512-NEXT:    vpermt2q %zmm4, %zmm8, %zmm10
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512-NEXT:    vpermt2q %zmm3, %zmm8, %zmm10
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm8
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,1,5,9,13]
; AVX512-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm11
; AVX512-NEXT:    vpermt2q %zmm6, %zmm10, %zmm11
; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm12
; AVX512-NEXT:    vpermt2q %zmm4, %zmm10, %zmm12
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm12
; AVX512-NEXT:    vpermt2q %zmm3, %zmm10, %zmm12
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm10
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,2,6,10,14]
; AVX512-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512-NEXT:    vpermt2q %zmm6, %zmm12, %zmm13
; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm14
; AVX512-NEXT:    vpermt2q %zmm4, %zmm12, %zmm14
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm14
; AVX512-NEXT:    vpermt2q %zmm3, %zmm12, %zmm14
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm12
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,3,7,11,15]
; AVX512-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermt2q %zmm6, %zmm14, %zmm7
; AVX512-NEXT:    vpermt2q %zmm4, %zmm14, %zmm5
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512-NEXT:    vpermt2q %zmm3, %zmm14, %zmm2
; AVX512-NEXT:    vpermt2q %zmm1, %zmm14, %zmm0
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm9, 64(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm11, 64(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm10, (%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm12, (%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm4, 64(%r8)
; AVX512-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i64_stride4_vf16:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm4
; AVX512-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm6
; AVX512-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,0,4,8,12]
; AVX512-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512-FCP-NEXT:    vpermt2q %zmm6, %zmm8, %zmm9
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, %zmm10
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm8, %zmm10
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm8, %zmm10
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm8
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,1,5,9,13]
; AVX512-FCP-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm11
; AVX512-FCP-NEXT:    vpermt2q %zmm6, %zmm10, %zmm11
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, %zmm12
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm10, %zmm12
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm12
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm10, %zmm12
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm10
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,2,6,10,14]
; AVX512-FCP-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512-FCP-NEXT:    vpermt2q %zmm6, %zmm12, %zmm13
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, %zmm14
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm12, %zmm14
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm14
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm12, %zmm14
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm12
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,3,7,11,15]
; AVX512-FCP-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermt2q %zmm6, %zmm14, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm14, %zmm5
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm14, %zmm2
; AVX512-FCP-NEXT:    vpermt2q %zmm1, %zmm14, %zmm0
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm9, 64(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, 64(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm10, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm12, (%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm4, 64(%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i64_stride4_vf16:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %zmm4
; AVX512DQ-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512DQ-NEXT:    vmovdqa64 448(%rdi), %zmm6
; AVX512DQ-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,0,4,8,12]
; AVX512DQ-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512DQ-NEXT:    vpermt2q %zmm6, %zmm8, %zmm9
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, %zmm10
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm8, %zmm10
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm8, %zmm10
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm8
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,1,5,9,13]
; AVX512DQ-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm11
; AVX512DQ-NEXT:    vpermt2q %zmm6, %zmm10, %zmm11
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, %zmm12
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm10, %zmm12
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm12
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm10, %zmm12
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm10
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,2,6,10,14]
; AVX512DQ-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512DQ-NEXT:    vpermt2q %zmm6, %zmm12, %zmm13
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, %zmm14
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm12, %zmm14
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm14
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm12, %zmm14
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm12
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,3,7,11,15]
; AVX512DQ-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermt2q %zmm6, %zmm14, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm14, %zmm5
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm14, %zmm2
; AVX512DQ-NEXT:    vpermt2q %zmm1, %zmm14, %zmm0
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm9, 64(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, 64(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm10, (%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm12, (%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm4, 64(%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride4_vf16:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512DQ-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm6
; AVX512DQ-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,0,4,8,12]
; AVX512DQ-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm6, %zmm8, %zmm9
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, %zmm10
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm8, %zmm10
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm8, %zmm10
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm8
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,1,5,9,13]
; AVX512DQ-FCP-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm11
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm6, %zmm10, %zmm11
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, %zmm12
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm10, %zmm12
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm12
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm10, %zmm12
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm10
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,2,6,10,14]
; AVX512DQ-FCP-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm6, %zmm12, %zmm13
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, %zmm14
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm12, %zmm14
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm14
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm12, %zmm14
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm12
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,3,7,11,15]
; AVX512DQ-FCP-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm6, %zmm14, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm14, %zmm5
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm14, %zmm2
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm1, %zmm14, %zmm0
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm9, 64(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, 64(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm10, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm12, (%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm4, 64(%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i64_stride4_vf16:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm4
; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm6
; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,0,4,8,12]
; AVX512BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm8, %zmm9
; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm10
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm8, %zmm10
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm8, %zmm10
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm8
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,1,5,9,13]
; AVX512BW-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm11
; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm10, %zmm11
; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm12
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm10, %zmm12
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm12
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm10, %zmm12
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm10
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,2,6,10,14]
; AVX512BW-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm12, %zmm13
; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm14
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm12, %zmm14
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm14
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm12, %zmm14
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm12
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,3,7,11,15]
; AVX512BW-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm14, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm14, %zmm5
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm14, %zmm2
; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm14, %zmm0
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm9, 64(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm11, 64(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm10, (%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm12, (%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm4, 64(%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i64_stride4_vf16:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm4
; AVX512BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm6
; AVX512BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,0,4,8,12]
; AVX512BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512BW-FCP-NEXT:    vpermt2q %zmm6, %zmm8, %zmm9
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm10
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm8, %zmm10
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm8, %zmm10
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm8
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,1,5,9,13]
; AVX512BW-FCP-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm11
; AVX512BW-FCP-NEXT:    vpermt2q %zmm6, %zmm10, %zmm11
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm12
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm10, %zmm12
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm12
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm10, %zmm12
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm10
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,2,6,10,14]
; AVX512BW-FCP-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512BW-FCP-NEXT:    vpermt2q %zmm6, %zmm12, %zmm13
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm14
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm12, %zmm14
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm14
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm12, %zmm14
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm12
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,3,7,11,15]
; AVX512BW-FCP-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm6, %zmm14, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm14, %zmm5
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm14, %zmm2
; AVX512BW-FCP-NEXT:    vpermt2q %zmm1, %zmm14, %zmm0
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm9, 64(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, 64(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm10, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm12, (%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, 64(%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i64_stride4_vf16:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 320(%rdi), %zmm4
; AVX512DQ-BW-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512DQ-BW-NEXT:    vmovdqa64 448(%rdi), %zmm6
; AVX512DQ-BW-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,0,4,8,12]
; AVX512DQ-BW-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512DQ-BW-NEXT:    vpermt2q %zmm6, %zmm8, %zmm9
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, %zmm10
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm8, %zmm10
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm8, %zmm10
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm8
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,1,5,9,13]
; AVX512DQ-BW-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm11
; AVX512DQ-BW-NEXT:    vpermt2q %zmm6, %zmm10, %zmm11
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, %zmm12
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm10, %zmm12
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm12
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm10, %zmm12
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm10
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,2,6,10,14]
; AVX512DQ-BW-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512DQ-BW-NEXT:    vpermt2q %zmm6, %zmm12, %zmm13
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, %zmm14
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm12, %zmm14
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm14
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm12, %zmm14
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm12
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,3,7,11,15]
; AVX512DQ-BW-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm6, %zmm14, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm14, %zmm5
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm14, %zmm2
; AVX512DQ-BW-NEXT:    vpermt2q %zmm1, %zmm14, %zmm0
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm9, 64(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, 64(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm10, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm12, (%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm4, 64(%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride4_vf16:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm6
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,0,4,8,12]
; AVX512DQ-BW-FCP-NEXT:    # zmm8 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm9
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm6, %zmm8, %zmm9
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm10
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm8, %zmm10
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm8, %zmm10
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,1,5,9,13]
; AVX512DQ-BW-FCP-NEXT:    # zmm10 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm11
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm6, %zmm10, %zmm11
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm12
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm10, %zmm12
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm12
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm10, %zmm12
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm10
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,2,6,10,14]
; AVX512DQ-BW-FCP-NEXT:    # zmm12 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm6, %zmm12, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm14
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm12, %zmm14
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm14
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm12, %zmm14
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm12
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,3,7,11,15]
; AVX512DQ-BW-FCP-NEXT:    # zmm14 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm6, %zmm14, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm14, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm14, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm1, %zmm14, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm9, 64(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm8, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, 64(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm10, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm12, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, 64(%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <64 x i64>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
  %strided.vec1 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
  %strided.vec2 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
  %strided.vec3 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
  store <16 x i64> %strided.vec0, ptr %out.vec0, align 64
  store <16 x i64> %strided.vec1, ptr %out.vec1, align 64
  store <16 x i64> %strided.vec2, ptr %out.vec2, align 64
  store <16 x i64> %strided.vec3, ptr %out.vec3, align 64
  ret void
}

define void @load_i64_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: load_i64_stride4_vf32:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $664, %rsp # imm = 0x298
; SSE-NEXT:    movaps 416(%rdi), %xmm0
; SSE-NEXT:    movaps 384(%rdi), %xmm9
; SSE-NEXT:    movaps 160(%rdi), %xmm3
; SSE-NEXT:    movaps 128(%rdi), %xmm8
; SSE-NEXT:    movaps 480(%rdi), %xmm1
; SSE-NEXT:    movaps 448(%rdi), %xmm11
; SSE-NEXT:    movaps 224(%rdi), %xmm5
; SSE-NEXT:    movaps 192(%rdi), %xmm10
; SSE-NEXT:    movaps 288(%rdi), %xmm4
; SSE-NEXT:    movaps 256(%rdi), %xmm12
; SSE-NEXT:    movaps 608(%rdi), %xmm2
; SSE-NEXT:    movaps 352(%rdi), %xmm6
; SSE-NEXT:    movaps 320(%rdi), %xmm14
; SSE-NEXT:    movaps 96(%rdi), %xmm7
; SSE-NEXT:    movaps 64(%rdi), %xmm13
; SSE-NEXT:    movaps %xmm13, %xmm15
; SSE-NEXT:    movlhps {{.*#+}} xmm15 = xmm15[0],xmm7[0]
; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm7[1]
; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm10, %xmm7
; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm5[0]
; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm5[1]
; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm8, %xmm5
; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm3[0]
; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm3[1]
; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm14, %xmm3
; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm6[0]
; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm6[1]
; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm12, %xmm3
; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm4[0]
; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm4[1]
; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm11, %xmm3
; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm1[1]
; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm9, %xmm1
; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm0[1]
; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 576(%rdi), %xmm0
; SSE-NEXT:    movaps %xmm0, %xmm1
; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 544(%rdi), %xmm0
; SSE-NEXT:    movaps 512(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 736(%rdi), %xmm0
; SSE-NEXT:    movaps 704(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 672(%rdi), %xmm0
; SSE-NEXT:    movaps 640(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 864(%rdi), %xmm0
; SSE-NEXT:    movaps 832(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 800(%rdi), %xmm0
; SSE-NEXT:    movaps 768(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 992(%rdi), %xmm0
; SSE-NEXT:    movaps 960(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 928(%rdi), %xmm0
; SSE-NEXT:    movaps 896(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps (%rdi), %xmm1
; SSE-NEXT:    movaps 32(%rdi), %xmm0
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 112(%rdi), %xmm0
; SSE-NEXT:    movaps 80(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 176(%rdi), %xmm0
; SSE-NEXT:    movaps 144(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 240(%rdi), %xmm0
; SSE-NEXT:    movaps 208(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 304(%rdi), %xmm0
; SSE-NEXT:    movaps 272(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 368(%rdi), %xmm0
; SSE-NEXT:    movaps 336(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 432(%rdi), %xmm0
; SSE-NEXT:    movaps 400(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
; SSE-NEXT:    movaps 496(%rdi), %xmm0
; SSE-NEXT:    movaps 464(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 560(%rdi), %xmm0
; SSE-NEXT:    movaps 528(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 624(%rdi), %xmm0
; SSE-NEXT:    movaps 592(%rdi), %xmm14
; SSE-NEXT:    movaps %xmm14, %xmm1
; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm0[1]
; SSE-NEXT:    movaps 688(%rdi), %xmm0
; SSE-NEXT:    movaps 656(%rdi), %xmm13
; SSE-NEXT:    movaps %xmm13, %xmm15
; SSE-NEXT:    movlhps {{.*#+}} xmm15 = xmm15[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm0[1]
; SSE-NEXT:    movaps 752(%rdi), %xmm0
; SSE-NEXT:    movaps 720(%rdi), %xmm11
; SSE-NEXT:    movaps %xmm11, %xmm12
; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm0[1]
; SSE-NEXT:    movaps 816(%rdi), %xmm0
; SSE-NEXT:    movaps 784(%rdi), %xmm9
; SSE-NEXT:    movaps %xmm9, %xmm10
; SSE-NEXT:    movlhps {{.*#+}} xmm10 = xmm10[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm0[1]
; SSE-NEXT:    movaps 880(%rdi), %xmm0
; SSE-NEXT:    movaps 848(%rdi), %xmm7
; SSE-NEXT:    movaps %xmm7, %xmm8
; SSE-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm0[1]
; SSE-NEXT:    movaps 944(%rdi), %xmm0
; SSE-NEXT:    movaps 912(%rdi), %xmm3
; SSE-NEXT:    movaps %xmm3, %xmm5
; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm0[1]
; SSE-NEXT:    movaps 1008(%rdi), %xmm4
; SSE-NEXT:    movaps 976(%rdi), %xmm2
; SSE-NEXT:    movaps %xmm2, %xmm1
; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm4[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm4[1]
; SSE-NEXT:    movaps 16(%rdi), %xmm4
; SSE-NEXT:    movaps 48(%rdi), %xmm0
; SSE-NEXT:    movaps %xmm4, %xmm6
; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm0[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 %xmm1, 240(%rcx)
; SSE-NEXT:    movaps %xmm5, 224(%rcx)
; SSE-NEXT:    movaps %xmm8, 208(%rcx)
; SSE-NEXT:    movaps %xmm10, 192(%rcx)
; SSE-NEXT:    movaps %xmm12, 176(%rcx)
; SSE-NEXT:    movaps %xmm15, 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 %xmm6, (%rcx)
; SSE-NEXT:    movaps %xmm2, 240(%r8)
; SSE-NEXT:    movaps %xmm3, 224(%r8)
; SSE-NEXT:    movaps %xmm7, 208(%r8)
; SSE-NEXT:    movaps %xmm9, 192(%r8)
; SSE-NEXT:    movaps %xmm11, 176(%r8)
; SSE-NEXT:    movaps %xmm13, 160(%r8)
; SSE-NEXT:    movaps %xmm14, 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 (%rsp), %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 {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%r8)
; SSE-NEXT:    movaps %xmm4, (%r8)
; SSE-NEXT:    addq $664, %rsp # imm = 0x298
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i64_stride4_vf32:
; AVX:       # %bb.0:
; AVX-NEXT:    subq $1112, %rsp # imm = 0x458
; AVX-NEXT:    vmovaps 480(%rdi), %xmm0
; AVX-NEXT:    vmovaps 448(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 736(%rdi), %xmm3
; AVX-NEXT:    vmovaps 704(%rdi), %xmm5
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm5[0],xmm3[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 224(%rdi), %xmm2
; AVX-NEXT:    vmovaps 192(%rdi), %xmm4
; AVX-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm2[0]
; AVX-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 672(%rdi), %xmm6
; AVX-NEXT:    vmovaps 640(%rdi), %xmm7
; AVX-NEXT:    vmovaps 928(%rdi), %xmm8
; AVX-NEXT:    vmovaps 896(%rdi), %xmm9
; AVX-NEXT:    vmovaps 992(%rdi), %xmm10
; AVX-NEXT:    vmovaps 960(%rdi), %xmm11
; AVX-NEXT:    vmovlhps {{.*#+}} xmm12 = xmm11[0],xmm10[0]
; AVX-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm10 = xmm11[1],xmm10[1]
; AVX-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm10 = xmm9[0],xmm8[0]
; AVX-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm9[1],xmm8[1]
; AVX-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm6[0]
; AVX-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm5[1],xmm3[1]
; AVX-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 160(%rdi), %xmm3
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm7[1],xmm6[1]
; AVX-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 128(%rdi), %xmm5
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm4[1],xmm2[1]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm5[0],xmm3[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm5[1],xmm3[1]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 416(%rdi), %xmm2
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 384(%rdi), %xmm0
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 864(%rdi), %xmm0
; AVX-NEXT:    vmovaps 832(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 800(%rdi), %xmm0
; AVX-NEXT:    vmovaps 768(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 96(%rdi), %xmm0
; AVX-NEXT:    vmovaps 64(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 32(%rdi), %xmm0
; AVX-NEXT:    vmovaps (%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 608(%rdi), %xmm0
; AVX-NEXT:    vmovaps 576(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 544(%rdi), %xmm0
; AVX-NEXT:    vmovaps 512(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 352(%rdi), %xmm0
; AVX-NEXT:    vmovaps 320(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 288(%rdi), %xmm0
; AVX-NEXT:    vmovaps 256(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 224(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 192(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vmovaps 176(%rdi), %xmm2
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 144(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; 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:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vmovaps 432(%rdi), %xmm2
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 400(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 736(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 704(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vmovaps 688(%rdi), %xmm2
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 656(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 992(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 960(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vmovaps 944(%rdi), %xmm2
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 912(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; 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:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vmovaps 48(%rdi), %xmm14
; AVX-NEXT:    vmovaps 16(%rdi), %xmm13
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm13[0],xmm14[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 352(%rdi), %ymm12
; AVX-NEXT:    vmovaps 320(%rdi), %ymm11
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
; AVX-NEXT:    vmovaps 304(%rdi), %xmm10
; AVX-NEXT:    vmovaps 272(%rdi), %xmm9
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm10[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 608(%rdi), %ymm8
; AVX-NEXT:    vmovaps 576(%rdi), %ymm7
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
; AVX-NEXT:    vmovaps 560(%rdi), %xmm6
; AVX-NEXT:    vmovaps 528(%rdi), %xmm5
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm5[0],xmm6[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 864(%rdi), %ymm4
; AVX-NEXT:    vmovaps 832(%rdi), %ymm3
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
; AVX-NEXT:    vmovaps 816(%rdi), %xmm2
; AVX-NEXT:    vmovaps 784(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm1[0],xmm2[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
; AVX-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
; AVX-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
; AVX-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
; AVX-NEXT:    # xmm15 = mem[0,1],xmm15[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm13 = xmm13[1],xmm14[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm9 = xmm9[1],xmm10[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm5[1],xmm6[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 208(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 192(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 128(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 64(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, (%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 144(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 80(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 16(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 240(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 224(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 32(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 160(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 96(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 48(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 176(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 112(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 64(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 80(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 128(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 144(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, (%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 16(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 192(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 208(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 96(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 112(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 32(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 48(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 160(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 176(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 224(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm1, 240(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, 192(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, 128(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, 64(%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, 224(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, 160(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, 96(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm1, 32(%rcx)
; AVX-NEXT:    vmovaps %ymm2, 192(%r8)
; AVX-NEXT:    vmovaps %ymm0, 128(%r8)
; AVX-NEXT:    vmovaps %ymm9, 64(%r8)
; AVX-NEXT:    vmovaps %ymm13, (%r8)
; AVX-NEXT:    vmovaps %ymm15, 224(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 160(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 96(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 32(%r8)
; AVX-NEXT:    addq $1112, %rsp # imm = 0x458
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i64_stride4_vf32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    subq $1224, %rsp # imm = 0x4C8
; AVX2-NEXT:    vmovaps 160(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 224(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vmovaps 128(%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 192(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 416(%rdi), %xmm2
; AVX2-NEXT:    vinsertf128 $1, 480(%rdi), %ymm2, %ymm2
; AVX2-NEXT:    vmovaps 384(%rdi), %xmm3
; AVX2-NEXT:    vinsertf128 $1, 448(%rdi), %ymm3, %ymm3
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 672(%rdi), %xmm4
; AVX2-NEXT:    vinsertf128 $1, 736(%rdi), %ymm4, %ymm4
; AVX2-NEXT:    vmovaps 640(%rdi), %xmm5
; AVX2-NEXT:    vinsertf128 $1, 704(%rdi), %ymm5, %ymm5
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 928(%rdi), %xmm6
; AVX2-NEXT:    vinsertf128 $1, 992(%rdi), %ymm6, %ymm6
; AVX2-NEXT:    vmovaps 896(%rdi), %xmm7
; AVX2-NEXT:    vinsertf128 $1, 960(%rdi), %ymm7, %ymm7
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps (%rdi), %xmm9
; AVX2-NEXT:    vmovaps 32(%rdi), %xmm8
; AVX2-NEXT:    vinsertf128 $1, 96(%rdi), %ymm8, %ymm8
; AVX2-NEXT:    vinsertf128 $1, 64(%rdi), %ymm9, %ymm9
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
; AVX2-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 288(%rdi), %xmm10
; AVX2-NEXT:    vinsertf128 $1, 352(%rdi), %ymm10, %ymm10
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 256(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 320(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm10[0],ymm0[2],ymm10[2]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 544(%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 608(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 512(%rdi), %xmm2
; AVX2-NEXT:    vinsertf128 $1, 576(%rdi), %ymm2, %ymm2
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm10[1],ymm0[3],ymm10[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 800(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 864(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 768(%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 832(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-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:    vmovaps 128(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 224(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 192(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 416(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 384(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-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:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 672(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 640(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 736(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 704(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-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:    vmovaps 896(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 992(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 960(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps (%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-NEXT:    vmovaps 96(%rdi), %ymm12
; AVX2-NEXT:    vmovaps 64(%rdi), %ymm11
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 288(%rdi), %ymm10
; AVX2-NEXT:    vmovaps 256(%rdi), %ymm9
; AVX2-NEXT:    vmovaps 352(%rdi), %ymm8
; AVX2-NEXT:    vmovaps 320(%rdi), %ymm7
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm9[0],ymm10[0],ymm9[2],ymm10[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 544(%rdi), %ymm6
; AVX2-NEXT:    vmovaps 512(%rdi), %ymm15
; AVX2-NEXT:    vmovaps 608(%rdi), %ymm5
; AVX2-NEXT:    vmovaps 576(%rdi), %ymm3
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm5[0],ymm3[2],ymm5[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm15[0],ymm6[0],ymm15[2],ymm6[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 800(%rdi), %ymm4
; AVX2-NEXT:    vmovaps 768(%rdi), %ymm14
; AVX2-NEXT:    vmovaps 864(%rdi), %ymm2
; AVX2-NEXT:    vmovaps 832(%rdi), %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm14[0],ymm4[0],ymm14[2],ymm4[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm11 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
; AVX2-NEXT:    vmovups (%rsp), %ymm12 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm12[2,3],ymm11[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm7 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm9[1],ymm10[1],ymm9[3],ymm10[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm8[2,3],ymm7[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm5[1],ymm3[3],ymm5[3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm15[1],ymm6[1],ymm15[3],ymm6[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm5[2,3],ymm3[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm14[1],ymm4[1],ymm14[3],ymm4[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 192(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 128(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 64(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, (%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 224(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 160(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 96(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 32(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 192(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 128(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 64(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, (%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 224(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 160(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 96(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 32(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 192(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 128(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 64(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, (%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 224(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 160(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 96(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 32(%rcx)
; AVX2-NEXT:    vmovaps %ymm1, 192(%r8)
; AVX2-NEXT:    vmovaps %ymm3, 128(%r8)
; AVX2-NEXT:    vmovaps %ymm7, 64(%r8)
; AVX2-NEXT:    vmovaps %ymm11, (%r8)
; AVX2-NEXT:    vmovaps %ymm0, 224(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 160(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 96(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 32(%r8)
; AVX2-NEXT:    addq $1224, %rsp # imm = 0x4C8
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i64_stride4_vf32:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    subq $1224, %rsp # imm = 0x4C8
; AVX2-FP-NEXT:    vmovaps 160(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 224(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 192(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 416(%rdi), %xmm2
; AVX2-FP-NEXT:    vinsertf128 $1, 480(%rdi), %ymm2, %ymm2
; AVX2-FP-NEXT:    vmovaps 384(%rdi), %xmm3
; AVX2-FP-NEXT:    vinsertf128 $1, 448(%rdi), %ymm3, %ymm3
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 672(%rdi), %xmm4
; AVX2-FP-NEXT:    vinsertf128 $1, 736(%rdi), %ymm4, %ymm4
; AVX2-FP-NEXT:    vmovaps 640(%rdi), %xmm5
; AVX2-FP-NEXT:    vinsertf128 $1, 704(%rdi), %ymm5, %ymm5
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-FP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 928(%rdi), %xmm6
; AVX2-FP-NEXT:    vinsertf128 $1, 992(%rdi), %ymm6, %ymm6
; AVX2-FP-NEXT:    vmovaps 896(%rdi), %xmm7
; AVX2-FP-NEXT:    vinsertf128 $1, 960(%rdi), %ymm7, %ymm7
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-FP-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps (%rdi), %xmm9
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %xmm8
; AVX2-FP-NEXT:    vinsertf128 $1, 96(%rdi), %ymm8, %ymm8
; AVX2-FP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm9, %ymm9
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
; AVX2-FP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 288(%rdi), %xmm10
; AVX2-FP-NEXT:    vinsertf128 $1, 352(%rdi), %ymm10, %ymm10
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 256(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 320(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm10[0],ymm0[2],ymm10[2]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 544(%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 608(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 512(%rdi), %xmm2
; AVX2-FP-NEXT:    vinsertf128 $1, 576(%rdi), %ymm2, %ymm2
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm10[1],ymm0[3],ymm10[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 800(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 864(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 768(%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 832(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-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:    vmovaps 128(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; 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), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 416(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 384(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-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:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 672(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 640(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 736(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 704(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-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:    vmovaps 896(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 992(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 960(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 96(%rdi), %ymm12
; AVX2-FP-NEXT:    vmovaps 64(%rdi), %ymm11
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 288(%rdi), %ymm10
; AVX2-FP-NEXT:    vmovaps 256(%rdi), %ymm9
; AVX2-FP-NEXT:    vmovaps 352(%rdi), %ymm8
; AVX2-FP-NEXT:    vmovaps 320(%rdi), %ymm7
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm9[0],ymm10[0],ymm9[2],ymm10[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 544(%rdi), %ymm6
; AVX2-FP-NEXT:    vmovaps 512(%rdi), %ymm15
; AVX2-FP-NEXT:    vmovaps 608(%rdi), %ymm5
; AVX2-FP-NEXT:    vmovaps 576(%rdi), %ymm3
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm5[0],ymm3[2],ymm5[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm15[0],ymm6[0],ymm15[2],ymm6[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 800(%rdi), %ymm4
; AVX2-FP-NEXT:    vmovaps 768(%rdi), %ymm14
; AVX2-FP-NEXT:    vmovaps 864(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovaps 832(%rdi), %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm14[0],ymm4[0],ymm14[2],ymm4[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm11 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
; AVX2-FP-NEXT:    vmovups (%rsp), %ymm12 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm12[2,3],ymm11[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm7 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm9[1],ymm10[1],ymm9[3],ymm10[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm8[2,3],ymm7[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm5[1],ymm3[3],ymm5[3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm15[1],ymm6[1],ymm15[3],ymm6[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm5[2,3],ymm3[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm14[1],ymm4[1],ymm14[3],ymm4[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 192(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 128(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 64(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, (%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 224(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 160(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 96(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 32(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 192(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 128(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 64(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, (%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 224(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 160(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 96(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 32(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 192(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 128(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 64(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, (%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 224(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 160(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 96(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 32(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm1, 192(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm3, 128(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm7, 64(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm11, (%r8)
; AVX2-FP-NEXT:    vmovaps %ymm0, 224(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 160(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 96(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 32(%r8)
; AVX2-FP-NEXT:    addq $1224, %rsp # imm = 0x4C8
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i64_stride4_vf32:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    subq $1224, %rsp # imm = 0x4C8
; AVX2-FCP-NEXT:    vmovaps 160(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 224(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 192(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 416(%rdi), %xmm2
; AVX2-FCP-NEXT:    vinsertf128 $1, 480(%rdi), %ymm2, %ymm2
; AVX2-FCP-NEXT:    vmovaps 384(%rdi), %xmm3
; AVX2-FCP-NEXT:    vinsertf128 $1, 448(%rdi), %ymm3, %ymm3
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 672(%rdi), %xmm4
; AVX2-FCP-NEXT:    vinsertf128 $1, 736(%rdi), %ymm4, %ymm4
; AVX2-FCP-NEXT:    vmovaps 640(%rdi), %xmm5
; AVX2-FCP-NEXT:    vinsertf128 $1, 704(%rdi), %ymm5, %ymm5
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-FCP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 928(%rdi), %xmm6
; AVX2-FCP-NEXT:    vinsertf128 $1, 992(%rdi), %ymm6, %ymm6
; AVX2-FCP-NEXT:    vmovaps 896(%rdi), %xmm7
; AVX2-FCP-NEXT:    vinsertf128 $1, 960(%rdi), %ymm7, %ymm7
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-FCP-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps (%rdi), %xmm9
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %xmm8
; AVX2-FCP-NEXT:    vinsertf128 $1, 96(%rdi), %ymm8, %ymm8
; AVX2-FCP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm9, %ymm9
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
; AVX2-FCP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 288(%rdi), %xmm10
; AVX2-FCP-NEXT:    vinsertf128 $1, 352(%rdi), %ymm10, %ymm10
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 256(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 320(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm10[0],ymm0[2],ymm10[2]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 544(%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 608(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 512(%rdi), %xmm2
; AVX2-FCP-NEXT:    vinsertf128 $1, 576(%rdi), %ymm2, %ymm2
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm10[1],ymm0[3],ymm10[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 800(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 864(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 768(%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 832(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-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:    vmovaps 128(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 224(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 192(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 416(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 384(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-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:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 672(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 640(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 736(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 704(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-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:    vmovaps 896(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 992(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 960(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 96(%rdi), %ymm12
; AVX2-FCP-NEXT:    vmovaps 64(%rdi), %ymm11
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 288(%rdi), %ymm10
; AVX2-FCP-NEXT:    vmovaps 256(%rdi), %ymm9
; AVX2-FCP-NEXT:    vmovaps 352(%rdi), %ymm8
; AVX2-FCP-NEXT:    vmovaps 320(%rdi), %ymm7
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm9[0],ymm10[0],ymm9[2],ymm10[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 544(%rdi), %ymm6
; AVX2-FCP-NEXT:    vmovaps 512(%rdi), %ymm15
; AVX2-FCP-NEXT:    vmovaps 608(%rdi), %ymm5
; AVX2-FCP-NEXT:    vmovaps 576(%rdi), %ymm3
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm5[0],ymm3[2],ymm5[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm15[0],ymm6[0],ymm15[2],ymm6[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 800(%rdi), %ymm4
; AVX2-FCP-NEXT:    vmovaps 768(%rdi), %ymm14
; AVX2-FCP-NEXT:    vmovaps 864(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovaps 832(%rdi), %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm14[0],ymm4[0],ymm14[2],ymm4[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm11 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
; AVX2-FCP-NEXT:    vmovups (%rsp), %ymm12 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm12[2,3],ymm11[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm7 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm9[1],ymm10[1],ymm9[3],ymm10[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm8[2,3],ymm7[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm5[1],ymm3[3],ymm5[3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm15[1],ymm6[1],ymm15[3],ymm6[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm5[2,3],ymm3[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm14[1],ymm4[1],ymm14[3],ymm4[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 192(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 128(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 64(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, (%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 224(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 160(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 96(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 32(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 192(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 128(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 64(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, (%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 224(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 160(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 96(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 32(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 192(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 128(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 64(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, (%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 224(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 160(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 96(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 32(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm1, 192(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm3, 128(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm7, 64(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm11, (%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm0, 224(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 160(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 96(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 32(%r8)
; AVX2-FCP-NEXT:    addq $1224, %rsp # imm = 0x4C8
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i64_stride4_vf32:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512-NEXT:    vmovdqa64 832(%rdi), %zmm4
; AVX512-NEXT:    vmovdqa64 768(%rdi), %zmm5
; AVX512-NEXT:    vmovdqa64 960(%rdi), %zmm10
; AVX512-NEXT:    vmovdqa64 896(%rdi), %zmm7
; AVX512-NEXT:    vmovdqa64 320(%rdi), %zmm14
; AVX512-NEXT:    vmovdqa64 256(%rdi), %zmm15
; AVX512-NEXT:    vmovdqa64 448(%rdi), %zmm17
; AVX512-NEXT:    vmovdqa64 384(%rdi), %zmm16
; AVX512-NEXT:    vmovdqa64 576(%rdi), %zmm8
; AVX512-NEXT:    vmovdqa64 512(%rdi), %zmm9
; AVX512-NEXT:    vmovdqa64 704(%rdi), %zmm12
; AVX512-NEXT:    vmovdqa64 640(%rdi), %zmm11
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,4,8,12,0,4,8,12]
; AVX512-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm6
; AVX512-NEXT:    vpermt2q %zmm12, %zmm19, %zmm6
; AVX512-NEXT:    vmovdqa64 %zmm9, %zmm13
; AVX512-NEXT:    vpermt2q %zmm8, %zmm19, %zmm13
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm13[0,1,2,3],zmm6[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm16, %zmm13
; AVX512-NEXT:    vpermt2q %zmm17, %zmm19, %zmm13
; AVX512-NEXT:    vmovdqa64 %zmm15, %zmm18
; AVX512-NEXT:    vpermt2q %zmm14, %zmm19, %zmm18
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm18[0,1,2,3],zmm13[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512-NEXT:    vpermt2q %zmm10, %zmm19, %zmm18
; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512-NEXT:    vpermt2q %zmm4, %zmm19, %zmm20
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm18 = zmm20[0,1,2,3],zmm18[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm20
; AVX512-NEXT:    vpermt2q %zmm3, %zmm19, %zmm20
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm19
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm20[4,5,6,7]
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [1,5,9,13,1,5,9,13]
; AVX512-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm20
; AVX512-NEXT:    vpermt2q %zmm12, %zmm21, %zmm20
; AVX512-NEXT:    vmovdqa64 %zmm9, %zmm22
; AVX512-NEXT:    vpermt2q %zmm8, %zmm21, %zmm22
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm20 = zmm22[0,1,2,3],zmm20[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm16, %zmm22
; AVX512-NEXT:    vpermt2q %zmm17, %zmm21, %zmm22
; AVX512-NEXT:    vmovdqa64 %zmm15, %zmm23
; AVX512-NEXT:    vpermt2q %zmm14, %zmm21, %zmm23
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm22 = zmm23[0,1,2,3],zmm22[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm23
; AVX512-NEXT:    vpermt2q %zmm10, %zmm21, %zmm23
; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm24
; AVX512-NEXT:    vpermt2q %zmm4, %zmm21, %zmm24
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm24[0,1,2,3],zmm23[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm24
; AVX512-NEXT:    vpermt2q %zmm3, %zmm21, %zmm24
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm21
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm24[4,5,6,7]
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm24 = [2,6,10,14,2,6,10,14]
; AVX512-NEXT:    # zmm24 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm25
; AVX512-NEXT:    vpermt2q %zmm12, %zmm24, %zmm25
; AVX512-NEXT:    vmovdqa64 %zmm9, %zmm26
; AVX512-NEXT:    vpermt2q %zmm8, %zmm24, %zmm26
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm25 = zmm26[0,1,2,3],zmm25[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm16, %zmm26
; AVX512-NEXT:    vpermt2q %zmm17, %zmm24, %zmm26
; AVX512-NEXT:    vmovdqa64 %zmm15, %zmm27
; AVX512-NEXT:    vpermt2q %zmm14, %zmm24, %zmm27
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm27[0,1,2,3],zmm26[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm27
; AVX512-NEXT:    vpermt2q %zmm10, %zmm24, %zmm27
; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm28
; AVX512-NEXT:    vpermt2q %zmm4, %zmm24, %zmm28
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm27 = zmm28[0,1,2,3],zmm27[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm28
; AVX512-NEXT:    vpermt2q %zmm3, %zmm24, %zmm28
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm24
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm24 = zmm24[0,1,2,3],zmm28[4,5,6,7]
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm28 = [3,7,11,15,3,7,11,15]
; AVX512-NEXT:    # zmm28 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermt2q %zmm17, %zmm28, %zmm16
; AVX512-NEXT:    vpermt2q %zmm14, %zmm28, %zmm15
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm15[0,1,2,3],zmm16[4,5,6,7]
; AVX512-NEXT:    vpermt2q %zmm10, %zmm28, %zmm7
; AVX512-NEXT:    vpermt2q %zmm4, %zmm28, %zmm5
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512-NEXT:    vpermt2q %zmm12, %zmm28, %zmm11
; AVX512-NEXT:    vpermt2q %zmm8, %zmm28, %zmm9
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm9[0,1,2,3],zmm11[4,5,6,7]
; AVX512-NEXT:    vpermt2q %zmm3, %zmm28, %zmm2
; AVX512-NEXT:    vpermt2q %zmm1, %zmm28, %zmm0
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm18, 192(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm19, (%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm6, 128(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm23, 192(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm22, 64(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm20, 128(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm27, 192(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm26, 64(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm5, 128(%r8)
; AVX512-NEXT:    vmovdqa64 %zmm4, 192(%r8)
; AVX512-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512-NEXT:    vmovdqa64 %zmm14, 64(%r8)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i64_stride4_vf32:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm4
; AVX512-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm5
; AVX512-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm10
; AVX512-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm7
; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm14
; AVX512-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm15
; AVX512-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm17
; AVX512-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm16
; AVX512-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm8
; AVX512-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm9
; AVX512-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm12
; AVX512-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm11
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,4,8,12,0,4,8,12]
; AVX512-FCP-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, %zmm6
; AVX512-FCP-NEXT:    vpermt2q %zmm12, %zmm19, %zmm6
; AVX512-FCP-NEXT:    vmovdqa64 %zmm9, %zmm13
; AVX512-FCP-NEXT:    vpermt2q %zmm8, %zmm19, %zmm13
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm13[0,1,2,3],zmm6[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm16, %zmm13
; AVX512-FCP-NEXT:    vpermt2q %zmm17, %zmm19, %zmm13
; AVX512-FCP-NEXT:    vmovdqa64 %zmm15, %zmm18
; AVX512-FCP-NEXT:    vpermt2q %zmm14, %zmm19, %zmm18
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm18[0,1,2,3],zmm13[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512-FCP-NEXT:    vpermt2q %zmm10, %zmm19, %zmm18
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm19, %zmm20
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm18 = zmm20[0,1,2,3],zmm18[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm20
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm19, %zmm20
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm19
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm20[4,5,6,7]
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [1,5,9,13,1,5,9,13]
; AVX512-FCP-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, %zmm20
; AVX512-FCP-NEXT:    vpermt2q %zmm12, %zmm21, %zmm20
; AVX512-FCP-NEXT:    vmovdqa64 %zmm9, %zmm22
; AVX512-FCP-NEXT:    vpermt2q %zmm8, %zmm21, %zmm22
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm20 = zmm22[0,1,2,3],zmm20[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm16, %zmm22
; AVX512-FCP-NEXT:    vpermt2q %zmm17, %zmm21, %zmm22
; AVX512-FCP-NEXT:    vmovdqa64 %zmm15, %zmm23
; AVX512-FCP-NEXT:    vpermt2q %zmm14, %zmm21, %zmm23
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm22 = zmm23[0,1,2,3],zmm22[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm23
; AVX512-FCP-NEXT:    vpermt2q %zmm10, %zmm21, %zmm23
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, %zmm24
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm21, %zmm24
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm24[0,1,2,3],zmm23[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm24
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm21, %zmm24
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm21
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm24[4,5,6,7]
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm24 = [2,6,10,14,2,6,10,14]
; AVX512-FCP-NEXT:    # zmm24 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, %zmm25
; AVX512-FCP-NEXT:    vpermt2q %zmm12, %zmm24, %zmm25
; AVX512-FCP-NEXT:    vmovdqa64 %zmm9, %zmm26
; AVX512-FCP-NEXT:    vpermt2q %zmm8, %zmm24, %zmm26
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm25 = zmm26[0,1,2,3],zmm25[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm16, %zmm26
; AVX512-FCP-NEXT:    vpermt2q %zmm17, %zmm24, %zmm26
; AVX512-FCP-NEXT:    vmovdqa64 %zmm15, %zmm27
; AVX512-FCP-NEXT:    vpermt2q %zmm14, %zmm24, %zmm27
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm27[0,1,2,3],zmm26[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm27
; AVX512-FCP-NEXT:    vpermt2q %zmm10, %zmm24, %zmm27
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, %zmm28
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm24, %zmm28
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm27 = zmm28[0,1,2,3],zmm27[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm28
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm24, %zmm28
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm24
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm24 = zmm24[0,1,2,3],zmm28[4,5,6,7]
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm28 = [3,7,11,15,3,7,11,15]
; AVX512-FCP-NEXT:    # zmm28 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermt2q %zmm17, %zmm28, %zmm16
; AVX512-FCP-NEXT:    vpermt2q %zmm14, %zmm28, %zmm15
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm15[0,1,2,3],zmm16[4,5,6,7]
; AVX512-FCP-NEXT:    vpermt2q %zmm10, %zmm28, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm28, %zmm5
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512-FCP-NEXT:    vpermt2q %zmm12, %zmm28, %zmm11
; AVX512-FCP-NEXT:    vpermt2q %zmm8, %zmm28, %zmm9
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm9[0,1,2,3],zmm11[4,5,6,7]
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm28, %zmm2
; AVX512-FCP-NEXT:    vpermt2q %zmm1, %zmm28, %zmm0
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm18, 192(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm19, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm6, 128(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm23, 192(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm22, 64(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm20, 128(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm27, 192(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm26, 64(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, 128(%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm4, 192(%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm14, 64(%r8)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i64_stride4_vf32:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-NEXT:    vmovdqa64 832(%rdi), %zmm4
; AVX512DQ-NEXT:    vmovdqa64 768(%rdi), %zmm5
; AVX512DQ-NEXT:    vmovdqa64 960(%rdi), %zmm10
; AVX512DQ-NEXT:    vmovdqa64 896(%rdi), %zmm7
; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %zmm14
; AVX512DQ-NEXT:    vmovdqa64 256(%rdi), %zmm15
; AVX512DQ-NEXT:    vmovdqa64 448(%rdi), %zmm17
; AVX512DQ-NEXT:    vmovdqa64 384(%rdi), %zmm16
; AVX512DQ-NEXT:    vmovdqa64 576(%rdi), %zmm8
; AVX512DQ-NEXT:    vmovdqa64 512(%rdi), %zmm9
; AVX512DQ-NEXT:    vmovdqa64 704(%rdi), %zmm12
; AVX512DQ-NEXT:    vmovdqa64 640(%rdi), %zmm11
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,4,8,12,0,4,8,12]
; AVX512DQ-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, %zmm6
; AVX512DQ-NEXT:    vpermt2q %zmm12, %zmm19, %zmm6
; AVX512DQ-NEXT:    vmovdqa64 %zmm9, %zmm13
; AVX512DQ-NEXT:    vpermt2q %zmm8, %zmm19, %zmm13
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm13[0,1,2,3],zmm6[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm16, %zmm13
; AVX512DQ-NEXT:    vpermt2q %zmm17, %zmm19, %zmm13
; AVX512DQ-NEXT:    vmovdqa64 %zmm15, %zmm18
; AVX512DQ-NEXT:    vpermt2q %zmm14, %zmm19, %zmm18
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm18[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512DQ-NEXT:    vpermt2q %zmm10, %zmm19, %zmm18
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm19, %zmm20
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm18 = zmm20[0,1,2,3],zmm18[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm20
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm19, %zmm20
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm19
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm20[4,5,6,7]
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [1,5,9,13,1,5,9,13]
; AVX512DQ-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, %zmm20
; AVX512DQ-NEXT:    vpermt2q %zmm12, %zmm21, %zmm20
; AVX512DQ-NEXT:    vmovdqa64 %zmm9, %zmm22
; AVX512DQ-NEXT:    vpermt2q %zmm8, %zmm21, %zmm22
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm20 = zmm22[0,1,2,3],zmm20[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm16, %zmm22
; AVX512DQ-NEXT:    vpermt2q %zmm17, %zmm21, %zmm22
; AVX512DQ-NEXT:    vmovdqa64 %zmm15, %zmm23
; AVX512DQ-NEXT:    vpermt2q %zmm14, %zmm21, %zmm23
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm22 = zmm23[0,1,2,3],zmm22[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm23
; AVX512DQ-NEXT:    vpermt2q %zmm10, %zmm21, %zmm23
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, %zmm24
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm21, %zmm24
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm24[0,1,2,3],zmm23[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm24
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm21, %zmm24
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm21
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm24[4,5,6,7]
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm24 = [2,6,10,14,2,6,10,14]
; AVX512DQ-NEXT:    # zmm24 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, %zmm25
; AVX512DQ-NEXT:    vpermt2q %zmm12, %zmm24, %zmm25
; AVX512DQ-NEXT:    vmovdqa64 %zmm9, %zmm26
; AVX512DQ-NEXT:    vpermt2q %zmm8, %zmm24, %zmm26
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm25 = zmm26[0,1,2,3],zmm25[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm16, %zmm26
; AVX512DQ-NEXT:    vpermt2q %zmm17, %zmm24, %zmm26
; AVX512DQ-NEXT:    vmovdqa64 %zmm15, %zmm27
; AVX512DQ-NEXT:    vpermt2q %zmm14, %zmm24, %zmm27
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm27[0,1,2,3],zmm26[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm27
; AVX512DQ-NEXT:    vpermt2q %zmm10, %zmm24, %zmm27
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, %zmm28
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm24, %zmm28
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm27 = zmm28[0,1,2,3],zmm27[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm28
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm24, %zmm28
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm24
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm24 = zmm24[0,1,2,3],zmm28[4,5,6,7]
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm28 = [3,7,11,15,3,7,11,15]
; AVX512DQ-NEXT:    # zmm28 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermt2q %zmm17, %zmm28, %zmm16
; AVX512DQ-NEXT:    vpermt2q %zmm14, %zmm28, %zmm15
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm15[0,1,2,3],zmm16[4,5,6,7]
; AVX512DQ-NEXT:    vpermt2q %zmm10, %zmm28, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm28, %zmm5
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-NEXT:    vpermt2q %zmm12, %zmm28, %zmm11
; AVX512DQ-NEXT:    vpermt2q %zmm8, %zmm28, %zmm9
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm9[0,1,2,3],zmm11[4,5,6,7]
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm28, %zmm2
; AVX512DQ-NEXT:    vpermt2q %zmm1, %zmm28, %zmm0
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm18, 192(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm19, (%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm6, 128(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm23, 192(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm22, 64(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm20, 128(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm27, 192(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm26, 64(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, 128(%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm4, 192(%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm14, 64(%r8)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride4_vf32:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm5
; AVX512DQ-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm10
; AVX512DQ-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm14
; AVX512DQ-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm15
; AVX512DQ-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm17
; AVX512DQ-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm16
; AVX512DQ-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm8
; AVX512DQ-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm9
; AVX512DQ-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm12
; AVX512DQ-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm11
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,4,8,12,0,4,8,12]
; AVX512DQ-FCP-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, %zmm6
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm12, %zmm19, %zmm6
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm9, %zmm13
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm8, %zmm19, %zmm13
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm13[0,1,2,3],zmm6[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm16, %zmm13
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm17, %zmm19, %zmm13
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm15, %zmm18
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm14, %zmm19, %zmm18
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm18[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm10, %zmm19, %zmm18
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm19, %zmm20
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm18 = zmm20[0,1,2,3],zmm18[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm20
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm19, %zmm20
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm19
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm20[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [1,5,9,13,1,5,9,13]
; AVX512DQ-FCP-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, %zmm20
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm12, %zmm21, %zmm20
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm9, %zmm22
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm8, %zmm21, %zmm22
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm20 = zmm22[0,1,2,3],zmm20[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm16, %zmm22
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm17, %zmm21, %zmm22
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm15, %zmm23
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm14, %zmm21, %zmm23
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm22 = zmm23[0,1,2,3],zmm22[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm23
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm10, %zmm21, %zmm23
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, %zmm24
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm21, %zmm24
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm24[0,1,2,3],zmm23[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm24
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm21, %zmm24
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm21
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm24[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm24 = [2,6,10,14,2,6,10,14]
; AVX512DQ-FCP-NEXT:    # zmm24 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, %zmm25
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm12, %zmm24, %zmm25
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm9, %zmm26
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm8, %zmm24, %zmm26
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm25 = zmm26[0,1,2,3],zmm25[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm16, %zmm26
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm17, %zmm24, %zmm26
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm15, %zmm27
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm14, %zmm24, %zmm27
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm27[0,1,2,3],zmm26[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm27
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm10, %zmm24, %zmm27
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, %zmm28
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm24, %zmm28
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm27 = zmm28[0,1,2,3],zmm27[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm28
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm24, %zmm28
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm24
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm24 = zmm24[0,1,2,3],zmm28[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm28 = [3,7,11,15,3,7,11,15]
; AVX512DQ-FCP-NEXT:    # zmm28 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm17, %zmm28, %zmm16
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm14, %zmm28, %zmm15
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm15[0,1,2,3],zmm16[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm10, %zmm28, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm28, %zmm5
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm12, %zmm28, %zmm11
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm8, %zmm28, %zmm9
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm9[0,1,2,3],zmm11[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm28, %zmm2
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm1, %zmm28, %zmm0
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm18, 192(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm19, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm6, 128(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm23, 192(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm22, 64(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm20, 128(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm27, 192(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm26, 64(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, 128(%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm4, 192(%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm14, 64(%r8)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i64_stride4_vf32:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm4
; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm5
; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm10
; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm7
; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm14
; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm15
; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm17
; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm16
; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm8
; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm9
; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm12
; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm11
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,4,8,12,0,4,8,12]
; AVX512BW-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm6
; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm19, %zmm6
; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm13
; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm19, %zmm13
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm13[0,1,2,3],zmm6[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm13
; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm19, %zmm13
; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm18
; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm19, %zmm18
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm18[0,1,2,3],zmm13[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm19, %zmm18
; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm19, %zmm20
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm18 = zmm20[0,1,2,3],zmm18[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm20
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm19, %zmm20
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm19
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm20[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [1,5,9,13,1,5,9,13]
; AVX512BW-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm20
; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm21, %zmm20
; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm22
; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm21, %zmm22
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm20 = zmm22[0,1,2,3],zmm20[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm22
; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm21, %zmm22
; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm23
; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm21, %zmm23
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm22 = zmm23[0,1,2,3],zmm22[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm23
; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm21, %zmm23
; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm24
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm21, %zmm24
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm24[0,1,2,3],zmm23[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm24
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm21, %zmm24
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm21
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm24[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm24 = [2,6,10,14,2,6,10,14]
; AVX512BW-NEXT:    # zmm24 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm25
; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm24, %zmm25
; AVX512BW-NEXT:    vmovdqa64 %zmm9, %zmm26
; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm24, %zmm26
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm25 = zmm26[0,1,2,3],zmm25[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm16, %zmm26
; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm24, %zmm26
; AVX512BW-NEXT:    vmovdqa64 %zmm15, %zmm27
; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm24, %zmm27
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm27[0,1,2,3],zmm26[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm27
; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm24, %zmm27
; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm28
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm24, %zmm28
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm27 = zmm28[0,1,2,3],zmm27[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm28
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm24, %zmm28
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm24
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm24 = zmm24[0,1,2,3],zmm28[4,5,6,7]
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm28 = [3,7,11,15,3,7,11,15]
; AVX512BW-NEXT:    # zmm28 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermt2q %zmm17, %zmm28, %zmm16
; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm28, %zmm15
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm15[0,1,2,3],zmm16[4,5,6,7]
; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm28, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm28, %zmm5
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm28, %zmm11
; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm28, %zmm9
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm9[0,1,2,3],zmm11[4,5,6,7]
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm28, %zmm2
; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm28, %zmm0
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm18, 192(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm19, (%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm6, 128(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm23, 192(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm22, 64(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm20, 128(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm27, 192(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm26, 64(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm5, 128(%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm4, 192(%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm14, 64(%r8)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i64_stride4_vf32:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm4
; AVX512BW-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm5
; AVX512BW-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm10
; AVX512BW-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm7
; AVX512BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm14
; AVX512BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm15
; AVX512BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm17
; AVX512BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm16
; AVX512BW-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm8
; AVX512BW-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm9
; AVX512BW-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm12
; AVX512BW-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm11
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,4,8,12,0,4,8,12]
; AVX512BW-FCP-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm6
; AVX512BW-FCP-NEXT:    vpermt2q %zmm12, %zmm19, %zmm6
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm13
; AVX512BW-FCP-NEXT:    vpermt2q %zmm8, %zmm19, %zmm13
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm13[0,1,2,3],zmm6[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm16, %zmm13
; AVX512BW-FCP-NEXT:    vpermt2q %zmm17, %zmm19, %zmm13
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm15, %zmm18
; AVX512BW-FCP-NEXT:    vpermt2q %zmm14, %zmm19, %zmm18
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm18[0,1,2,3],zmm13[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512BW-FCP-NEXT:    vpermt2q %zmm10, %zmm19, %zmm18
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm19, %zmm20
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm18 = zmm20[0,1,2,3],zmm18[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm20
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm19, %zmm20
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm19
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm20[4,5,6,7]
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [1,5,9,13,1,5,9,13]
; AVX512BW-FCP-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm20
; AVX512BW-FCP-NEXT:    vpermt2q %zmm12, %zmm21, %zmm20
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm22
; AVX512BW-FCP-NEXT:    vpermt2q %zmm8, %zmm21, %zmm22
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm20 = zmm22[0,1,2,3],zmm20[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm16, %zmm22
; AVX512BW-FCP-NEXT:    vpermt2q %zmm17, %zmm21, %zmm22
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm15, %zmm23
; AVX512BW-FCP-NEXT:    vpermt2q %zmm14, %zmm21, %zmm23
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm22 = zmm23[0,1,2,3],zmm22[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm23
; AVX512BW-FCP-NEXT:    vpermt2q %zmm10, %zmm21, %zmm23
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm24
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm21, %zmm24
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm24[0,1,2,3],zmm23[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm24
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm21, %zmm24
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm21
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm24[4,5,6,7]
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm24 = [2,6,10,14,2,6,10,14]
; AVX512BW-FCP-NEXT:    # zmm24 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm25
; AVX512BW-FCP-NEXT:    vpermt2q %zmm12, %zmm24, %zmm25
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm26
; AVX512BW-FCP-NEXT:    vpermt2q %zmm8, %zmm24, %zmm26
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm25 = zmm26[0,1,2,3],zmm25[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm16, %zmm26
; AVX512BW-FCP-NEXT:    vpermt2q %zmm17, %zmm24, %zmm26
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm15, %zmm27
; AVX512BW-FCP-NEXT:    vpermt2q %zmm14, %zmm24, %zmm27
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm27[0,1,2,3],zmm26[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm27
; AVX512BW-FCP-NEXT:    vpermt2q %zmm10, %zmm24, %zmm27
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm28
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm24, %zmm28
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm27 = zmm28[0,1,2,3],zmm27[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm28
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm24, %zmm28
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm24
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm24 = zmm24[0,1,2,3],zmm28[4,5,6,7]
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm28 = [3,7,11,15,3,7,11,15]
; AVX512BW-FCP-NEXT:    # zmm28 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm17, %zmm28, %zmm16
; AVX512BW-FCP-NEXT:    vpermt2q %zmm14, %zmm28, %zmm15
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm15[0,1,2,3],zmm16[4,5,6,7]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm10, %zmm28, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm28, %zmm5
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm12, %zmm28, %zmm11
; AVX512BW-FCP-NEXT:    vpermt2q %zmm8, %zmm28, %zmm9
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm9[0,1,2,3],zmm11[4,5,6,7]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm28, %zmm2
; AVX512BW-FCP-NEXT:    vpermt2q %zmm1, %zmm28, %zmm0
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm18, 192(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm19, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm6, 128(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm23, 192(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm22, 64(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm20, 128(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm27, 192(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm26, 64(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, 128(%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, 192(%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm14, 64(%r8)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i64_stride4_vf32:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 832(%rdi), %zmm4
; AVX512DQ-BW-NEXT:    vmovdqa64 768(%rdi), %zmm5
; AVX512DQ-BW-NEXT:    vmovdqa64 960(%rdi), %zmm10
; AVX512DQ-BW-NEXT:    vmovdqa64 896(%rdi), %zmm7
; AVX512DQ-BW-NEXT:    vmovdqa64 320(%rdi), %zmm14
; AVX512DQ-BW-NEXT:    vmovdqa64 256(%rdi), %zmm15
; AVX512DQ-BW-NEXT:    vmovdqa64 448(%rdi), %zmm17
; AVX512DQ-BW-NEXT:    vmovdqa64 384(%rdi), %zmm16
; AVX512DQ-BW-NEXT:    vmovdqa64 576(%rdi), %zmm8
; AVX512DQ-BW-NEXT:    vmovdqa64 512(%rdi), %zmm9
; AVX512DQ-BW-NEXT:    vmovdqa64 704(%rdi), %zmm12
; AVX512DQ-BW-NEXT:    vmovdqa64 640(%rdi), %zmm11
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,4,8,12,0,4,8,12]
; AVX512DQ-BW-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, %zmm6
; AVX512DQ-BW-NEXT:    vpermt2q %zmm12, %zmm19, %zmm6
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm9, %zmm13
; AVX512DQ-BW-NEXT:    vpermt2q %zmm8, %zmm19, %zmm13
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm13[0,1,2,3],zmm6[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm16, %zmm13
; AVX512DQ-BW-NEXT:    vpermt2q %zmm17, %zmm19, %zmm13
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm15, %zmm18
; AVX512DQ-BW-NEXT:    vpermt2q %zmm14, %zmm19, %zmm18
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm18[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512DQ-BW-NEXT:    vpermt2q %zmm10, %zmm19, %zmm18
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm19, %zmm20
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm18 = zmm20[0,1,2,3],zmm18[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm20
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm19, %zmm20
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm19
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm20[4,5,6,7]
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [1,5,9,13,1,5,9,13]
; AVX512DQ-BW-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, %zmm20
; AVX512DQ-BW-NEXT:    vpermt2q %zmm12, %zmm21, %zmm20
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm9, %zmm22
; AVX512DQ-BW-NEXT:    vpermt2q %zmm8, %zmm21, %zmm22
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm20 = zmm22[0,1,2,3],zmm20[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm16, %zmm22
; AVX512DQ-BW-NEXT:    vpermt2q %zmm17, %zmm21, %zmm22
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm15, %zmm23
; AVX512DQ-BW-NEXT:    vpermt2q %zmm14, %zmm21, %zmm23
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm22 = zmm23[0,1,2,3],zmm22[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm23
; AVX512DQ-BW-NEXT:    vpermt2q %zmm10, %zmm21, %zmm23
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, %zmm24
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm21, %zmm24
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm24[0,1,2,3],zmm23[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm24
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm21, %zmm24
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm21
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm24[4,5,6,7]
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm24 = [2,6,10,14,2,6,10,14]
; AVX512DQ-BW-NEXT:    # zmm24 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, %zmm25
; AVX512DQ-BW-NEXT:    vpermt2q %zmm12, %zmm24, %zmm25
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm9, %zmm26
; AVX512DQ-BW-NEXT:    vpermt2q %zmm8, %zmm24, %zmm26
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm25 = zmm26[0,1,2,3],zmm25[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm16, %zmm26
; AVX512DQ-BW-NEXT:    vpermt2q %zmm17, %zmm24, %zmm26
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm15, %zmm27
; AVX512DQ-BW-NEXT:    vpermt2q %zmm14, %zmm24, %zmm27
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm27[0,1,2,3],zmm26[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm27
; AVX512DQ-BW-NEXT:    vpermt2q %zmm10, %zmm24, %zmm27
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, %zmm28
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm24, %zmm28
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm27 = zmm28[0,1,2,3],zmm27[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm28
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm24, %zmm28
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm24
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm24 = zmm24[0,1,2,3],zmm28[4,5,6,7]
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm28 = [3,7,11,15,3,7,11,15]
; AVX512DQ-BW-NEXT:    # zmm28 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm17, %zmm28, %zmm16
; AVX512DQ-BW-NEXT:    vpermt2q %zmm14, %zmm28, %zmm15
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm15[0,1,2,3],zmm16[4,5,6,7]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm10, %zmm28, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm28, %zmm5
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm12, %zmm28, %zmm11
; AVX512DQ-BW-NEXT:    vpermt2q %zmm8, %zmm28, %zmm9
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm9[0,1,2,3],zmm11[4,5,6,7]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm28, %zmm2
; AVX512DQ-BW-NEXT:    vpermt2q %zmm1, %zmm28, %zmm0
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm18, 192(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm19, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm6, 128(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm23, 192(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm22, 64(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm20, 128(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm27, 192(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm26, 64(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, 128(%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm4, 192(%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm14, 64(%r8)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride4_vf32:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm5
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm10
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm14
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm15
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm17
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm16
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm8
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm9
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm12
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm11
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm19 = [0,4,8,12,0,4,8,12]
; AVX512DQ-BW-FCP-NEXT:    # zmm19 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm12, %zmm19, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm8, %zmm19, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm13[0,1,2,3],zmm6[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm16, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm17, %zmm19, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm15, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm14, %zmm19, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm13 = zmm18[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm10, %zmm19, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm19, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm18 = zmm20[0,1,2,3],zmm18[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm19, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm19
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm20[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm21 = [1,5,9,13,1,5,9,13]
; AVX512DQ-BW-FCP-NEXT:    # zmm21 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm12, %zmm21, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm22
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm8, %zmm21, %zmm22
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm20 = zmm22[0,1,2,3],zmm20[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm16, %zmm22
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm17, %zmm21, %zmm22
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm15, %zmm23
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm14, %zmm21, %zmm23
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm22 = zmm23[0,1,2,3],zmm22[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm23
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm10, %zmm21, %zmm23
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm24
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm21, %zmm24
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm24[0,1,2,3],zmm23[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm24
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm21, %zmm24
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm21
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm24[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm24 = [2,6,10,14,2,6,10,14]
; AVX512DQ-BW-FCP-NEXT:    # zmm24 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm25
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm12, %zmm24, %zmm25
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm9, %zmm26
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm8, %zmm24, %zmm26
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm25 = zmm26[0,1,2,3],zmm25[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm16, %zmm26
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm17, %zmm24, %zmm26
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm15, %zmm27
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm14, %zmm24, %zmm27
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm27[0,1,2,3],zmm26[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm27
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm10, %zmm24, %zmm27
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm28
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm24, %zmm28
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm27 = zmm28[0,1,2,3],zmm27[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm28
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm24, %zmm28
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm24
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm24 = zmm24[0,1,2,3],zmm28[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm28 = [3,7,11,15,3,7,11,15]
; AVX512DQ-BW-FCP-NEXT:    # zmm28 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm17, %zmm28, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm14, %zmm28, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm15[0,1,2,3],zmm16[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm10, %zmm28, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm28, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm12, %zmm28, %zmm11
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm8, %zmm28, %zmm9
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm9[0,1,2,3],zmm11[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm28, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm1, %zmm28, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm18, 192(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm19, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm6, 128(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm23, 192(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm22, 64(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm20, 128(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm27, 192(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm24, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm26, 64(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, 128(%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, 192(%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, (%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm14, 64(%r8)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <128 x i64>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <32 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60, i32 64, i32 68, i32 72, i32 76, i32 80, i32 84, i32 88, i32 92, i32 96, i32 100, i32 104, i32 108, i32 112, i32 116, i32 120, i32 124>
  %strided.vec1 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <32 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61, i32 65, i32 69, i32 73, i32 77, i32 81, i32 85, i32 89, i32 93, i32 97, i32 101, i32 105, i32 109, i32 113, i32 117, i32 121, i32 125>
  %strided.vec2 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <32 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62, i32 66, i32 70, i32 74, i32 78, i32 82, i32 86, i32 90, i32 94, i32 98, i32 102, i32 106, i32 110, i32 114, i32 118, i32 122, i32 126>
  %strided.vec3 = shufflevector <128 x i64> %wide.vec, <128 x i64> poison, <32 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63, i32 67, i32 71, i32 75, i32 79, i32 83, i32 87, i32 91, i32 95, i32 99, i32 103, i32 107, i32 111, i32 115, i32 119, i32 123, i32 127>
  store <32 x i64> %strided.vec0, ptr %out.vec0, align 64
  store <32 x i64> %strided.vec1, ptr %out.vec1, align 64
  store <32 x i64> %strided.vec2, ptr %out.vec2, align 64
  store <32 x i64> %strided.vec3, ptr %out.vec3, align 64
  ret void
}

define void @load_i64_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
; SSE-LABEL: load_i64_stride4_vf64:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $1688, %rsp # imm = 0x698
; SSE-NEXT:    movaps 544(%rdi), %xmm0
; SSE-NEXT:    movaps 480(%rdi), %xmm1
; SSE-NEXT:    movaps 448(%rdi), %xmm8
; SSE-NEXT:    movaps 416(%rdi), %xmm2
; SSE-NEXT:    movaps 384(%rdi), %xmm9
; SSE-NEXT:    movaps 352(%rdi), %xmm3
; SSE-NEXT:    movaps 320(%rdi), %xmm10
; SSE-NEXT:    movaps 288(%rdi), %xmm4
; SSE-NEXT:    movaps 256(%rdi), %xmm11
; SSE-NEXT:    movaps 224(%rdi), %xmm5
; SSE-NEXT:    movaps 192(%rdi), %xmm12
; SSE-NEXT:    movaps 160(%rdi), %xmm6
; SSE-NEXT:    movaps 128(%rdi), %xmm13
; SSE-NEXT:    movaps 96(%rdi), %xmm7
; SSE-NEXT:    movaps 64(%rdi), %xmm14
; SSE-NEXT:    movaps %xmm14, %xmm15
; SSE-NEXT:    movlhps {{.*#+}} xmm15 = xmm15[0],xmm7[0]
; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm7[1]
; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm13, %xmm7
; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm6[0]
; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm6[1]
; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm12, %xmm6
; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm5[0]
; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm5[1]
; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm11, %xmm5
; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm4[0]
; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm4[1]
; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm10, %xmm4
; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm3[0]
; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm3[1]
; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm9, %xmm3
; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm2[1]
; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps %xmm8, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm1[1]
; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 512(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 608(%rdi), %xmm0
; SSE-NEXT:    movaps 576(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 672(%rdi), %xmm0
; SSE-NEXT:    movaps 640(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 736(%rdi), %xmm0
; SSE-NEXT:    movaps 704(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 800(%rdi), %xmm0
; SSE-NEXT:    movaps 768(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 864(%rdi), %xmm0
; SSE-NEXT:    movaps 832(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 928(%rdi), %xmm0
; SSE-NEXT:    movaps 896(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 992(%rdi), %xmm0
; SSE-NEXT:    movaps 960(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1056(%rdi), %xmm0
; SSE-NEXT:    movaps 1024(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1120(%rdi), %xmm0
; SSE-NEXT:    movaps 1088(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1184(%rdi), %xmm0
; SSE-NEXT:    movaps 1152(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1248(%rdi), %xmm0
; SSE-NEXT:    movaps 1216(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1312(%rdi), %xmm0
; SSE-NEXT:    movaps 1280(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1376(%rdi), %xmm0
; SSE-NEXT:    movaps 1344(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1440(%rdi), %xmm0
; SSE-NEXT:    movaps 1408(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1504(%rdi), %xmm0
; SSE-NEXT:    movaps 1472(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1568(%rdi), %xmm0
; SSE-NEXT:    movaps 1536(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1632(%rdi), %xmm0
; SSE-NEXT:    movaps 1600(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1696(%rdi), %xmm0
; SSE-NEXT:    movaps 1664(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1760(%rdi), %xmm0
; SSE-NEXT:    movaps 1728(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1824(%rdi), %xmm0
; SSE-NEXT:    movaps 1792(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1888(%rdi), %xmm0
; SSE-NEXT:    movaps 1856(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1952(%rdi), %xmm0
; SSE-NEXT:    movaps 1920(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 2016(%rdi), %xmm0
; SSE-NEXT:    movaps 1984(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps (%rdi), %xmm1
; SSE-NEXT:    movaps 32(%rdi), %xmm0
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 112(%rdi), %xmm0
; SSE-NEXT:    movaps 80(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 176(%rdi), %xmm0
; SSE-NEXT:    movaps 144(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 240(%rdi), %xmm0
; SSE-NEXT:    movaps 208(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 304(%rdi), %xmm0
; SSE-NEXT:    movaps 272(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 368(%rdi), %xmm0
; SSE-NEXT:    movaps 336(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 432(%rdi), %xmm0
; SSE-NEXT:    movaps 400(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 496(%rdi), %xmm0
; SSE-NEXT:    movaps 464(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 560(%rdi), %xmm0
; SSE-NEXT:    movaps 528(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 624(%rdi), %xmm0
; SSE-NEXT:    movaps 592(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 688(%rdi), %xmm0
; SSE-NEXT:    movaps 656(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 752(%rdi), %xmm0
; SSE-NEXT:    movaps 720(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 816(%rdi), %xmm0
; SSE-NEXT:    movaps 784(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 880(%rdi), %xmm0
; SSE-NEXT:    movaps 848(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 944(%rdi), %xmm0
; SSE-NEXT:    movaps 912(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1008(%rdi), %xmm0
; SSE-NEXT:    movaps 976(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1072(%rdi), %xmm0
; SSE-NEXT:    movaps 1040(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1136(%rdi), %xmm0
; SSE-NEXT:    movaps 1104(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1200(%rdi), %xmm0
; SSE-NEXT:    movaps 1168(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1264(%rdi), %xmm0
; SSE-NEXT:    movaps 1232(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1328(%rdi), %xmm0
; SSE-NEXT:    movaps 1296(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1392(%rdi), %xmm0
; SSE-NEXT:    movaps 1360(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1456(%rdi), %xmm0
; SSE-NEXT:    movaps 1424(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1520(%rdi), %xmm0
; SSE-NEXT:    movaps 1488(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
; SSE-NEXT:    movaps 1584(%rdi), %xmm0
; SSE-NEXT:    movaps 1552(%rdi), %xmm1
; SSE-NEXT:    movaps %xmm1, %xmm2
; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movaps 1648(%rdi), %xmm0
; SSE-NEXT:    movaps 1616(%rdi), %xmm15
; SSE-NEXT:    movaps %xmm15, %xmm1
; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm0[1]
; SSE-NEXT:    movaps 1712(%rdi), %xmm0
; SSE-NEXT:    movaps 1680(%rdi), %xmm13
; SSE-NEXT:    movaps %xmm13, %xmm14
; SSE-NEXT:    movlhps {{.*#+}} xmm14 = xmm14[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm0[1]
; SSE-NEXT:    movaps 1776(%rdi), %xmm0
; SSE-NEXT:    movaps 1744(%rdi), %xmm12
; SSE-NEXT:    movaps %xmm12, %xmm11
; SSE-NEXT:    movlhps {{.*#+}} xmm11 = xmm11[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm0[1]
; SSE-NEXT:    movaps 1840(%rdi), %xmm0
; SSE-NEXT:    movaps 1808(%rdi), %xmm9
; SSE-NEXT:    movaps %xmm9, %xmm7
; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm0[1]
; SSE-NEXT:    movaps 1904(%rdi), %xmm0
; SSE-NEXT:    movaps 1872(%rdi), %xmm6
; SSE-NEXT:    movaps %xmm6, %xmm5
; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm0[1]
; SSE-NEXT:    movaps 1968(%rdi), %xmm0
; SSE-NEXT:    movaps 1936(%rdi), %xmm4
; SSE-NEXT:    movaps %xmm4, %xmm3
; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm0[1]
; SSE-NEXT:    movaps 2032(%rdi), %xmm8
; SSE-NEXT:    movaps 2000(%rdi), %xmm2
; SSE-NEXT:    movaps %xmm2, %xmm1
; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm8[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm8[1]
; SSE-NEXT:    movaps 16(%rdi), %xmm10
; SSE-NEXT:    movaps 48(%rdi), %xmm0
; SSE-NEXT:    movaps %xmm10, %xmm8
; SSE-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm0[0]
; SSE-NEXT:    unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm0[1]
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 496(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 480(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 464(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 448(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 432(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 416(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 400(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 384(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 368(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 352(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 336(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 320(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 304(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 288(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 272(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 256(%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, 224(%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, 192(%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, 160(%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, 128(%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, 96(%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, 64(%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, 32(%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, (%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 496(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 480(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 464(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 448(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 432(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 416(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 400(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 384(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 368(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 352(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 336(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 320(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 304(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 288(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 272(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 256(%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, 224(%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, 192(%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, 160(%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, 128(%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, 96(%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, 64(%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, 32(%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, (%rdx)
; SSE-NEXT:    movaps %xmm1, 496(%rcx)
; SSE-NEXT:    movaps %xmm3, 480(%rcx)
; SSE-NEXT:    movaps %xmm5, 464(%rcx)
; SSE-NEXT:    movaps %xmm7, 448(%rcx)
; SSE-NEXT:    movaps %xmm11, 432(%rcx)
; SSE-NEXT:    movaps %xmm14, 416(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 400(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 384(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 368(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 352(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 336(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 320(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 304(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 288(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 272(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 256(%rcx)
; 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 %xmm8, (%rcx)
; SSE-NEXT:    movaps %xmm2, 496(%r8)
; SSE-NEXT:    movaps %xmm4, 480(%r8)
; SSE-NEXT:    movaps %xmm6, 464(%r8)
; SSE-NEXT:    movaps %xmm9, 448(%r8)
; SSE-NEXT:    movaps %xmm12, 432(%r8)
; SSE-NEXT:    movaps %xmm13, 416(%r8)
; SSE-NEXT:    movaps %xmm15, 400(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 384(%r8)
; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 368(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 352(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 336(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 320(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 304(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 288(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 272(%r8)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 256(%r8)
; 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 {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%r8)
; SSE-NEXT:    movaps %xmm10, (%r8)
; SSE-NEXT:    addq $1688, %rsp # imm = 0x698
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i64_stride4_vf64:
; AVX:       # %bb.0:
; AVX-NEXT:    subq $2728, %rsp # imm = 0xAA8
; AVX-NEXT:    vmovaps 1248(%rdi), %xmm0
; AVX-NEXT:    vmovaps 1216(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1504(%rdi), %xmm2
; AVX-NEXT:    vmovaps 1472(%rdi), %xmm3
; AVX-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
; AVX-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1760(%rdi), %xmm4
; AVX-NEXT:    vmovaps 1728(%rdi), %xmm5
; AVX-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
; AVX-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1440(%rdi), %xmm6
; AVX-NEXT:    vmovaps 1696(%rdi), %xmm7
; AVX-NEXT:    vmovaps 1952(%rdi), %xmm8
; AVX-NEXT:    vmovaps 1920(%rdi), %xmm9
; AVX-NEXT:    vmovaps 2016(%rdi), %xmm10
; AVX-NEXT:    vmovaps 1984(%rdi), %xmm11
; AVX-NEXT:    vmovlhps {{.*#+}} xmm12 = xmm11[0],xmm10[0]
; AVX-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm10 = xmm11[1],xmm10[1]
; AVX-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm10 = xmm9[0],xmm8[0]
; AVX-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm9[1],xmm8[1]
; AVX-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1664(%rdi), %xmm8
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
; AVX-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm8[0],xmm7[0]
; AVX-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm8[1],xmm7[1]
; AVX-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1408(%rdi), %xmm4
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm6[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm4[1],xmm6[1]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1184(%rdi), %xmm2
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1152(%rdi), %xmm0
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 992(%rdi), %xmm0
; AVX-NEXT:    vmovaps 960(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 928(%rdi), %xmm0
; AVX-NEXT:    vmovaps 896(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 736(%rdi), %xmm0
; AVX-NEXT:    vmovaps 704(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 672(%rdi), %xmm0
; AVX-NEXT:    vmovaps 640(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 224(%rdi), %xmm0
; AVX-NEXT:    vmovaps 192(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 160(%rdi), %xmm0
; AVX-NEXT:    vmovaps 128(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 480(%rdi), %xmm0
; AVX-NEXT:    vmovaps 448(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 416(%rdi), %xmm0
; AVX-NEXT:    vmovaps 384(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1888(%rdi), %xmm0
; AVX-NEXT:    vmovaps 1856(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1824(%rdi), %xmm0
; AVX-NEXT:    vmovaps 1792(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1632(%rdi), %xmm0
; AVX-NEXT:    vmovaps 1600(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1568(%rdi), %xmm0
; AVX-NEXT:    vmovaps 1536(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1376(%rdi), %xmm0
; AVX-NEXT:    vmovaps 1344(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1312(%rdi), %xmm0
; AVX-NEXT:    vmovaps 1280(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 864(%rdi), %xmm0
; AVX-NEXT:    vmovaps 832(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 800(%rdi), %xmm0
; AVX-NEXT:    vmovaps 768(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 96(%rdi), %xmm0
; AVX-NEXT:    vmovaps 64(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 32(%rdi), %xmm0
; AVX-NEXT:    vmovaps (%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 352(%rdi), %xmm0
; AVX-NEXT:    vmovaps 320(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 288(%rdi), %xmm0
; AVX-NEXT:    vmovaps 256(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1120(%rdi), %xmm0
; AVX-NEXT:    vmovaps 1088(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1056(%rdi), %xmm0
; AVX-NEXT:    vmovaps 1024(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 608(%rdi), %xmm0
; AVX-NEXT:    vmovaps 576(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 544(%rdi), %xmm0
; AVX-NEXT:    vmovaps 512(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 224(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 192(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vmovaps 176(%rdi), %xmm2
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 144(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; 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:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vmovaps 432(%rdi), %xmm2
; AVX-NEXT:    vmovaps %xmm2, (%rsp) # 16-byte Spill
; AVX-NEXT:    vmovaps 400(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 736(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 704(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vmovaps 688(%rdi), %xmm2
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 656(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 992(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 960(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vmovaps 944(%rdi), %xmm2
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 912(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1248(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-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:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vmovaps 1200(%rdi), %xmm2
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1168(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1504(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1472(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vmovaps 1456(%rdi), %xmm2
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1424(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1760(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1728(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vmovaps 1712(%rdi), %xmm2
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1680(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 2016(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1984(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX-NEXT:    vmovaps 1968(%rdi), %xmm2
; AVX-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 1936(%rdi), %xmm1
; AVX-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 96(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 64(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX-NEXT:    vmovaps 48(%rdi), %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 16(%rdi), %xmm8
; AVX-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm8[0],xmm0[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 352(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 320(%rdi), %ymm1
; AVX-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX-NEXT:    vmovaps 304(%rdi), %xmm0
; AVX-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovaps 272(%rdi), %xmm1
; AVX-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm0[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 608(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 576(%rdi), %ymm2
; AVX-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
; AVX-NEXT:    vmovaps 560(%rdi), %xmm10
; AVX-NEXT:    vmovaps 528(%rdi), %xmm2
; AVX-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm2[0],xmm10[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 864(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 832(%rdi), %ymm3
; AVX-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
; AVX-NEXT:    vmovaps 816(%rdi), %xmm11
; AVX-NEXT:    vmovaps 784(%rdi), %xmm3
; AVX-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm3[0],xmm11[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm4[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1120(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1088(%rdi), %ymm4
; AVX-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm4[0],ymm0[0],ymm4[2],ymm0[2]
; AVX-NEXT:    vmovaps 1072(%rdi), %xmm12
; AVX-NEXT:    vmovaps 1040(%rdi), %xmm4
; AVX-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm12[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm5[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1376(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1344(%rdi), %ymm5
; AVX-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm5[0],ymm0[0],ymm5[2],ymm0[2]
; AVX-NEXT:    vmovaps 1328(%rdi), %xmm13
; AVX-NEXT:    vmovaps 1296(%rdi), %xmm5
; AVX-NEXT:    vmovlhps {{.*#+}} xmm7 = xmm5[0],xmm13[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1632(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1600(%rdi), %ymm6
; AVX-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm6[0],ymm0[0],ymm6[2],ymm0[2]
; AVX-NEXT:    vmovaps 1584(%rdi), %xmm14
; AVX-NEXT:    vmovaps 1552(%rdi), %xmm6
; AVX-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm6[0],xmm14[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm7[4,5,6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1888(%rdi), %ymm7
; AVX-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1856(%rdi), %ymm0
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm7[0],ymm0[2],ymm7[2]
; AVX-NEXT:    vmovaps 1840(%rdi), %xmm15
; AVX-NEXT:    vmovaps 1808(%rdi), %xmm7
; AVX-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm7[0],xmm15[0]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
; AVX-NEXT:    # xmm8 = xmm8[1],mem[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm9 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
; AVX-NEXT:    # xmm8 = mem[0,1],xmm8[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX-NEXT:    # xmm1 = xmm1[1],mem[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vmovaps (%rsp), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX-NEXT:    # xmm1 = mem[0,1],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm10[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX-NEXT:    # xmm1 = mem[0,1],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm11[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX-NEXT:    # xmm1 = mem[0,1],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm12[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX-NEXT:    # xmm1 = mem[0,1],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm13[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX-NEXT:    # xmm1 = mem[0,1],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm6[1],xmm14[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX-NEXT:    # xmm1 = mem[0,1],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm15[1]
; AVX-NEXT:    vblendps {{.*#+}} ymm12 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX-NEXT:    vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX-NEXT:    # xmm1 = mem[0,1],xmm1[2,3]
; AVX-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 464(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 448(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 256(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 384(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 320(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 192(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 128(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 64(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, (%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 272(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 400(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 336(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 208(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 144(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 80(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 16(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 496(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 480(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 32(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 416(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 352(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 288(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 224(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 160(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 96(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 48(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 432(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 368(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 304(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 240(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 176(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 112(%rsi)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 128(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 144(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 256(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 272(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 64(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 80(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, (%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 16(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 192(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 208(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 320(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 336(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 384(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 400(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 448(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 464(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 96(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 112(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 32(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 48(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 160(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 176(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 224(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 240(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 288(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 304(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 352(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 368(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 416(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 432(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 480(%rdx)
; AVX-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vmovaps %xmm0, 496(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 448(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 384(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 320(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 256(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 192(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 128(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 64(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, (%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 480(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 416(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 352(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 288(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 224(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 160(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 96(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 32(%rcx)
; AVX-NEXT:    vmovaps %ymm7, 480(%r8)
; AVX-NEXT:    vmovaps %ymm12, 448(%r8)
; AVX-NEXT:    vmovaps %ymm6, 416(%r8)
; AVX-NEXT:    vmovaps %ymm11, 384(%r8)
; AVX-NEXT:    vmovaps %ymm5, 352(%r8)
; AVX-NEXT:    vmovaps %ymm10, 320(%r8)
; AVX-NEXT:    vmovaps %ymm4, 288(%r8)
; AVX-NEXT:    vmovaps %ymm3, 256(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 224(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 192(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 160(%r8)
; AVX-NEXT:    vmovaps %ymm2, 128(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 96(%r8)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 64(%r8)
; AVX-NEXT:    vmovaps %ymm8, 32(%r8)
; AVX-NEXT:    vmovaps %ymm9, (%r8)
; AVX-NEXT:    addq $2728, %rsp # imm = 0xAA8
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i64_stride4_vf64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    subq $3016, %rsp # imm = 0xBC8
; AVX2-NEXT:    vmovaps 160(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 224(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vmovaps 128(%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 192(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 416(%rdi), %xmm2
; AVX2-NEXT:    vinsertf128 $1, 480(%rdi), %ymm2, %ymm2
; AVX2-NEXT:    vmovaps 384(%rdi), %xmm3
; AVX2-NEXT:    vinsertf128 $1, 448(%rdi), %ymm3, %ymm3
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 672(%rdi), %xmm4
; AVX2-NEXT:    vinsertf128 $1, 736(%rdi), %ymm4, %ymm4
; AVX2-NEXT:    vmovaps 640(%rdi), %xmm5
; AVX2-NEXT:    vinsertf128 $1, 704(%rdi), %ymm5, %ymm5
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 928(%rdi), %xmm6
; AVX2-NEXT:    vinsertf128 $1, 992(%rdi), %ymm6, %ymm6
; AVX2-NEXT:    vmovaps 896(%rdi), %xmm7
; AVX2-NEXT:    vinsertf128 $1, 960(%rdi), %ymm7, %ymm7
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1184(%rdi), %xmm8
; AVX2-NEXT:    vinsertf128 $1, 1248(%rdi), %ymm8, %ymm8
; AVX2-NEXT:    vmovaps 1152(%rdi), %xmm9
; AVX2-NEXT:    vinsertf128 $1, 1216(%rdi), %ymm9, %ymm9
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
; AVX2-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1440(%rdi), %xmm10
; AVX2-NEXT:    vinsertf128 $1, 1504(%rdi), %ymm10, %ymm10
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1408(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 1472(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm10[0],ymm0[2],ymm10[2]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1696(%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 1760(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1664(%rdi), %xmm2
; AVX2-NEXT:    vinsertf128 $1, 1728(%rdi), %ymm2, %ymm2
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm10[1],ymm0[3],ymm10[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1952(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 2016(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1920(%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 1984(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 32(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 96(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vmovaps (%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 64(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 288(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vmovaps 256(%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 320(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 544(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 608(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vmovaps 512(%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 576(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 800(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 864(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vmovaps 768(%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 832(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1056(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 1120(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vmovaps 1024(%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 1088(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1312(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 1376(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vmovaps 1280(%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 1344(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1568(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 1632(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vmovaps 1536(%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 1600(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1824(%rdi), %xmm0
; AVX2-NEXT:    vinsertf128 $1, 1888(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vmovaps 1792(%rdi), %xmm1
; AVX2-NEXT:    vinsertf128 $1, 1856(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 160(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
; AVX2-NEXT:    vmovaps 128(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 224(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 192(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 416(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 384(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-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:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 672(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 640(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 736(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 704(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-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:    vmovaps 896(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 992(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 960(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1184(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1152(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1248(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1216(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1440(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1408(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1504(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1472(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1696(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1664(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1760(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1728(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1952(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1920(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 2016(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1984(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps (%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 96(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 64(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-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), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 352(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 320(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-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:    vmovaps 512(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 608(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 576(%rdi), %ymm15
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm15[0],ymm0[0],ymm15[2],ymm0[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 800(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-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:    vmovaps 864(%rdi), %ymm14
; AVX2-NEXT:    vmovaps 832(%rdi), %ymm11
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm14[0],ymm11[2],ymm14[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1056(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1024(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1120(%rdi), %ymm10
; AVX2-NEXT:    vmovaps 1088(%rdi), %ymm8
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm8[0],ymm10[0],ymm8[2],ymm10[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1312(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1280(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1376(%rdi), %ymm7
; AVX2-NEXT:    vmovaps 1344(%rdi), %ymm5
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm7[0],ymm5[2],ymm7[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1568(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1536(%rdi), %ymm9
; AVX2-NEXT:    vmovaps 1632(%rdi), %ymm4
; AVX2-NEXT:    vmovaps 1600(%rdi), %ymm3
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm9[0],ymm1[0],ymm9[2],ymm1[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm2[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1824(%rdi), %ymm12
; AVX2-NEXT:    vmovaps 1792(%rdi), %ymm6
; AVX2-NEXT:    vmovaps 1888(%rdi), %ymm2
; AVX2-NEXT:    vmovaps 1856(%rdi), %ymm1
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm6[0],ymm12[0],ymm6[2],ymm12[2]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd (%rsp), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm15[1],mem[1],ymm15[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm14[1],ymm11[3],ymm14[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm11 = ymm11[1],mem[1],ymm11[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm11 = ymm11[1],mem[1],ymm11[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm11[2,3],ymm0[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm8[1],ymm10[1],ymm8[3],ymm10[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm8 = ymm8[1],mem[1],ymm8[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm8[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm8 = ymm8[1],mem[1],ymm8[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm8[2,3],ymm0[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm7[1],ymm5[3],ymm7[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm5[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm5[2,3],ymm0[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = ymm9[1],mem[1],ymm9[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm3[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm3[2,3],ymm0[2,3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
; AVX2-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm6[1],ymm12[1],ymm6[3],ymm12[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
; AVX2-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 448(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 384(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 320(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 256(%rsi)
; 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, 480(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 416(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 352(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 288(%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, 448(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 384(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 320(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 256(%rdx)
; 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, 480(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 416(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 352(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 288(%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, 448(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 384(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 320(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 256(%rcx)
; 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, 480(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 416(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 352(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm1, 288(%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:    vmovaps %ymm0, 480(%r8)
; AVX2-NEXT:    vmovaps %ymm6, 448(%r8)
; AVX2-NEXT:    vmovaps %ymm5, 416(%r8)
; AVX2-NEXT:    vmovaps %ymm7, 384(%r8)
; AVX2-NEXT:    vmovaps %ymm13, 352(%r8)
; AVX2-NEXT:    vmovaps %ymm10, 320(%r8)
; AVX2-NEXT:    vmovaps %ymm11, 288(%r8)
; AVX2-NEXT:    vmovaps %ymm14, 256(%r8)
; AVX2-NEXT:    vmovaps %ymm15, 224(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 192(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 160(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 128(%r8)
; AVX2-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 96(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 64(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 32(%r8)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, (%r8)
; AVX2-NEXT:    addq $3016, %rsp # imm = 0xBC8
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i64_stride4_vf64:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    subq $3016, %rsp # imm = 0xBC8
; AVX2-FP-NEXT:    vmovaps 160(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 224(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 192(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 416(%rdi), %xmm2
; AVX2-FP-NEXT:    vinsertf128 $1, 480(%rdi), %ymm2, %ymm2
; AVX2-FP-NEXT:    vmovaps 384(%rdi), %xmm3
; AVX2-FP-NEXT:    vinsertf128 $1, 448(%rdi), %ymm3, %ymm3
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 672(%rdi), %xmm4
; AVX2-FP-NEXT:    vinsertf128 $1, 736(%rdi), %ymm4, %ymm4
; AVX2-FP-NEXT:    vmovaps 640(%rdi), %xmm5
; AVX2-FP-NEXT:    vinsertf128 $1, 704(%rdi), %ymm5, %ymm5
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-FP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 928(%rdi), %xmm6
; AVX2-FP-NEXT:    vinsertf128 $1, 992(%rdi), %ymm6, %ymm6
; AVX2-FP-NEXT:    vmovaps 896(%rdi), %xmm7
; AVX2-FP-NEXT:    vinsertf128 $1, 960(%rdi), %ymm7, %ymm7
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-FP-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1184(%rdi), %xmm8
; AVX2-FP-NEXT:    vinsertf128 $1, 1248(%rdi), %ymm8, %ymm8
; AVX2-FP-NEXT:    vmovaps 1152(%rdi), %xmm9
; AVX2-FP-NEXT:    vinsertf128 $1, 1216(%rdi), %ymm9, %ymm9
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
; AVX2-FP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1440(%rdi), %xmm10
; AVX2-FP-NEXT:    vinsertf128 $1, 1504(%rdi), %ymm10, %ymm10
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1408(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 1472(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm10[0],ymm0[2],ymm10[2]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1696(%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 1760(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1664(%rdi), %xmm2
; AVX2-FP-NEXT:    vinsertf128 $1, 1728(%rdi), %ymm2, %ymm2
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm10[1],ymm0[3],ymm10[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1952(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 2016(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1920(%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 1984(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 96(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vmovaps (%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 288(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vmovaps 256(%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 320(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 544(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 608(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vmovaps 512(%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 576(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 800(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 864(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vmovaps 768(%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 832(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1056(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 1120(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vmovaps 1024(%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 1088(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1312(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 1376(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vmovaps 1280(%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 1344(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1568(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 1632(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vmovaps 1536(%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 1600(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1824(%rdi), %xmm0
; AVX2-FP-NEXT:    vinsertf128 $1, 1888(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vmovaps 1792(%rdi), %xmm1
; AVX2-FP-NEXT:    vinsertf128 $1, 1856(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 160(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; 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), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 416(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 384(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-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:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 672(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 640(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 736(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 704(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-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:    vmovaps 896(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 992(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 960(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1184(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1152(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; 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), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1440(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1408(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1504(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1472(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1696(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1664(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1760(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1728(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1952(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1920(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 2016(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1984(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 96(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 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:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-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), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 352(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 320(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-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:    vmovaps 512(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 608(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 576(%rdi), %ymm15
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm15[0],ymm0[0],ymm15[2],ymm0[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 800(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-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:    vmovaps 864(%rdi), %ymm14
; AVX2-FP-NEXT:    vmovaps 832(%rdi), %ymm11
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm14[0],ymm11[2],ymm14[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1056(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1024(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1120(%rdi), %ymm10
; AVX2-FP-NEXT:    vmovaps 1088(%rdi), %ymm8
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm8[0],ymm10[0],ymm8[2],ymm10[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1312(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1280(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1376(%rdi), %ymm7
; AVX2-FP-NEXT:    vmovaps 1344(%rdi), %ymm5
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm7[0],ymm5[2],ymm7[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1568(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1536(%rdi), %ymm9
; AVX2-FP-NEXT:    vmovaps 1632(%rdi), %ymm4
; AVX2-FP-NEXT:    vmovaps 1600(%rdi), %ymm3
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm9[0],ymm1[0],ymm9[2],ymm1[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm2[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1824(%rdi), %ymm12
; AVX2-FP-NEXT:    vmovaps 1792(%rdi), %ymm6
; AVX2-FP-NEXT:    vmovaps 1888(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovaps 1856(%rdi), %ymm1
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-FP-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm6[0],ymm12[0],ymm6[2],ymm12[2]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd (%rsp), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm15[1],mem[1],ymm15[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm14[1],ymm11[3],ymm14[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm11 = ymm11[1],mem[1],ymm11[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm11 = ymm11[1],mem[1],ymm11[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm11[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm8[1],ymm10[1],ymm8[3],ymm10[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm8 = ymm8[1],mem[1],ymm8[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm8[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm8 = ymm8[1],mem[1],ymm8[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm8[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm7[1],ymm5[3],ymm7[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm5[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm5[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = ymm9[1],mem[1],ymm9[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm3[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm3[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
; AVX2-FP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm6[1],ymm12[1],ymm6[3],ymm12[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
; AVX2-FP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 448(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 384(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 320(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 256(%rsi)
; 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, 480(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 416(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 352(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 288(%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, 448(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 384(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 320(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 256(%rdx)
; 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, 480(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 416(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 352(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 288(%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, 448(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 384(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 320(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 256(%rcx)
; 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, 480(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 416(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 352(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm1, 288(%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:    vmovaps %ymm0, 480(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm6, 448(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm5, 416(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm7, 384(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm13, 352(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm10, 320(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm11, 288(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm14, 256(%r8)
; AVX2-FP-NEXT:    vmovaps %ymm15, 224(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 192(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 160(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 128(%r8)
; AVX2-FP-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 96(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 64(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 32(%r8)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, (%r8)
; AVX2-FP-NEXT:    addq $3016, %rsp # imm = 0xBC8
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i64_stride4_vf64:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    subq $3016, %rsp # imm = 0xBC8
; AVX2-FCP-NEXT:    vmovaps 160(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 224(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 192(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 416(%rdi), %xmm2
; AVX2-FCP-NEXT:    vinsertf128 $1, 480(%rdi), %ymm2, %ymm2
; AVX2-FCP-NEXT:    vmovaps 384(%rdi), %xmm3
; AVX2-FCP-NEXT:    vinsertf128 $1, 448(%rdi), %ymm3, %ymm3
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 672(%rdi), %xmm4
; AVX2-FCP-NEXT:    vinsertf128 $1, 736(%rdi), %ymm4, %ymm4
; AVX2-FCP-NEXT:    vmovaps 640(%rdi), %xmm5
; AVX2-FCP-NEXT:    vinsertf128 $1, 704(%rdi), %ymm5, %ymm5
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
; AVX2-FCP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 928(%rdi), %xmm6
; AVX2-FCP-NEXT:    vinsertf128 $1, 992(%rdi), %ymm6, %ymm6
; AVX2-FCP-NEXT:    vmovaps 896(%rdi), %xmm7
; AVX2-FCP-NEXT:    vinsertf128 $1, 960(%rdi), %ymm7, %ymm7
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
; AVX2-FCP-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1184(%rdi), %xmm8
; AVX2-FCP-NEXT:    vinsertf128 $1, 1248(%rdi), %ymm8, %ymm8
; AVX2-FCP-NEXT:    vmovaps 1152(%rdi), %xmm9
; AVX2-FCP-NEXT:    vinsertf128 $1, 1216(%rdi), %ymm9, %ymm9
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
; AVX2-FCP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1440(%rdi), %xmm10
; AVX2-FCP-NEXT:    vinsertf128 $1, 1504(%rdi), %ymm10, %ymm10
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1408(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 1472(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm10[0],ymm0[2],ymm10[2]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1696(%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 1760(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1664(%rdi), %xmm2
; AVX2-FCP-NEXT:    vinsertf128 $1, 1728(%rdi), %ymm2, %ymm2
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm10[1],ymm0[3],ymm10[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1952(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 2016(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1920(%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 1984(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 96(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vmovaps (%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 288(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vmovaps 256(%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 320(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 544(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 608(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vmovaps 512(%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 576(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 800(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 864(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vmovaps 768(%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 832(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1056(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 1120(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vmovaps 1024(%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 1088(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1312(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 1376(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vmovaps 1280(%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 1344(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1568(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 1632(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vmovaps 1536(%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 1600(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1824(%rdi), %xmm0
; AVX2-FCP-NEXT:    vinsertf128 $1, 1888(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vmovaps 1792(%rdi), %xmm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 1856(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 160(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 224(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 192(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 416(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 384(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-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:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 672(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 640(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 736(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 704(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-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:    vmovaps 896(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 992(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 960(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1184(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1152(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; 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), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1440(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1408(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1504(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1472(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1696(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1664(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1760(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1728(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1952(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1920(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 2016(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1984(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-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:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-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), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 352(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 320(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-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:    vmovaps 512(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 608(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 576(%rdi), %ymm15
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm15[0],ymm0[0],ymm15[2],ymm0[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 800(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-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:    vmovaps 864(%rdi), %ymm14
; AVX2-FCP-NEXT:    vmovaps 832(%rdi), %ymm11
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm14[0],ymm11[2],ymm14[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1056(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1024(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1120(%rdi), %ymm10
; AVX2-FCP-NEXT:    vmovaps 1088(%rdi), %ymm8
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm8[0],ymm10[0],ymm8[2],ymm10[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-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:    vmovaps 1280(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1376(%rdi), %ymm7
; AVX2-FCP-NEXT:    vmovaps 1344(%rdi), %ymm5
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm7[0],ymm5[2],ymm7[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1568(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1536(%rdi), %ymm9
; AVX2-FCP-NEXT:    vmovaps 1632(%rdi), %ymm4
; AVX2-FCP-NEXT:    vmovaps 1600(%rdi), %ymm3
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm9[0],ymm1[0],ymm9[2],ymm1[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm2[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1824(%rdi), %ymm12
; AVX2-FCP-NEXT:    vmovaps 1792(%rdi), %ymm6
; AVX2-FCP-NEXT:    vmovaps 1888(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovaps 1856(%rdi), %ymm1
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-FCP-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm6[0],ymm12[0],ymm6[2],ymm12[2]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd (%rsp), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm15[1],mem[1],ymm15[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm14[1],ymm11[3],ymm14[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm11 = ymm11[1],mem[1],ymm11[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm0[2,3]
; 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:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm11 = ymm11[1],mem[1],ymm11[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm11[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm8[1],ymm10[1],ymm8[3],ymm10[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm8 = ymm8[1],mem[1],ymm8[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm8[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm8 = ymm8[1],mem[1],ymm8[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm8[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm7[1],ymm5[3],ymm7[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm5[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm5[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm3 = ymm9[1],mem[1],ymm9[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm3[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm3[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
; AVX2-FCP-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm6[1],ymm12[1],ymm6[3],ymm12[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
; AVX2-FCP-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 448(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 384(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 320(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 256(%rsi)
; 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, 480(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 416(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 352(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 288(%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, 448(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 384(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 320(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 256(%rdx)
; 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, 480(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 416(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 352(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 288(%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, 448(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 384(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 320(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 256(%rcx)
; 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, 480(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 416(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 352(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm1, 288(%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:    vmovaps %ymm0, 480(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm6, 448(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm5, 416(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm7, 384(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm13, 352(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm10, 320(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm11, 288(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm14, 256(%r8)
; AVX2-FCP-NEXT:    vmovaps %ymm15, 224(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 192(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 160(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 128(%r8)
; AVX2-FCP-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 96(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 64(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 32(%r8)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, (%r8)
; AVX2-FCP-NEXT:    addq $3016, %rsp # imm = 0xBC8
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i64_stride4_vf64:
; AVX512:       # %bb.0:
; AVX512-NEXT:    subq $2056, %rsp # imm = 0x808
; AVX512-NEXT:    vmovdqa64 1472(%rdi), %zmm24
; AVX512-NEXT:    vmovdqa64 1408(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 832(%rdi), %zmm26
; AVX512-NEXT:    vmovdqa64 768(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 960(%rdi), %zmm28
; AVX512-NEXT:    vmovdqa64 896(%rdi), %zmm2
; AVX512-NEXT:    vmovdqa64 576(%rdi), %zmm25
; AVX512-NEXT:    vmovdqa64 512(%rdi), %zmm3
; AVX512-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512-NEXT:    vmovdqa64 320(%rdi), %zmm30
; AVX512-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512-NEXT:    vmovdqa64 448(%rdi), %zmm29
; AVX512-NEXT:    vmovdqa64 384(%rdi), %zmm6
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [0,4,8,12,0,4,8,12]
; AVX512-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512-NEXT:    vpermt2q %zmm29, %zmm9, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512-NEXT:    vpermt2q %zmm30, %zmm9, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, (%rsp) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512-NEXT:    vpermt2q %zmm27, %zmm9, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512-NEXT:    vpermt2q %zmm25, %zmm9, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512-NEXT:    vpermt2q %zmm28, %zmm9, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512-NEXT:    vpermt2q %zmm26, %zmm9, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512-NEXT:    vpermt2q %zmm24, %zmm9, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [1,5,9,13,1,5,9,13]
; AVX512-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512-NEXT:    vpermt2q %zmm29, %zmm15, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512-NEXT:    vpermt2q %zmm30, %zmm15, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512-NEXT:    vpermt2q %zmm28, %zmm15, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512-NEXT:    vpermt2q %zmm26, %zmm15, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512-NEXT:    vpermt2q %zmm27, %zmm15, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512-NEXT:    vpermt2q %zmm25, %zmm15, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm22 = [2,6,10,14,2,6,10,14]
; AVX512-NEXT:    # zmm22 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512-NEXT:    vpermt2q %zmm29, %zmm22, %zmm7
; AVX512-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [3,7,11,15,3,7,11,15]
; AVX512-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpermt2q %zmm29, %zmm31, %zmm6
; AVX512-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm6
; AVX512-NEXT:    vpermt2q %zmm30, %zmm22, %zmm6
; AVX512-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2q %zmm30, %zmm31, %zmm5
; AVX512-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512-NEXT:    vpermt2q %zmm28, %zmm22, %zmm5
; AVX512-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2q %zmm28, %zmm31, %zmm2
; AVX512-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512-NEXT:    vpermt2q %zmm26, %zmm22, %zmm2
; AVX512-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2q %zmm26, %zmm31, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm4, %zmm1
; AVX512-NEXT:    vpermt2q %zmm27, %zmm22, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2q %zmm27, %zmm31, %zmm4
; AVX512-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm3, %zmm1
; AVX512-NEXT:    vpermt2q %zmm25, %zmm22, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2q %zmm25, %zmm31, %zmm3
; AVX512-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512-NEXT:    vpermt2q %zmm24, %zmm15, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512-NEXT:    vpermt2q %zmm24, %zmm22, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2q %zmm24, %zmm31, %zmm0
; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 1344(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 1280(%rdi), %zmm24
; AVX512-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512-NEXT:    vpermt2q %zmm0, %zmm9, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm24
; AVX512-NEXT:    vmovdqa64 1216(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 1152(%rdi), %zmm28
; AVX512-NEXT:    vmovdqa64 %zmm28, %zmm30
; AVX512-NEXT:    vpermt2q %zmm0, %zmm9, %zmm30
; AVX512-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm28
; AVX512-NEXT:    vmovdqa64 1088(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 1024(%rdi), %zmm23
; AVX512-NEXT:    vmovdqa64 %zmm23, %zmm26
; AVX512-NEXT:    vpermt2q %zmm0, %zmm9, %zmm26
; AVX512-NEXT:    vmovdqa64 %zmm23, %zmm29
; AVX512-NEXT:    vpermt2q %zmm0, %zmm15, %zmm29
; AVX512-NEXT:    vmovdqa64 %zmm23, %zmm1
; AVX512-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512-NEXT:    vmovdqa64 1984(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 1920(%rdi), %zmm13
; AVX512-NEXT:    vmovdqa64 %zmm13, %zmm20
; AVX512-NEXT:    vpermt2q %zmm0, %zmm9, %zmm20
; AVX512-NEXT:    vmovdqa64 %zmm13, %zmm25
; AVX512-NEXT:    vpermt2q %zmm0, %zmm15, %zmm25
; AVX512-NEXT:    vmovdqa64 %zmm13, %zmm27
; AVX512-NEXT:    vpermt2q %zmm0, %zmm22, %zmm27
; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm13
; AVX512-NEXT:    vmovdqa64 1856(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 1792(%rdi), %zmm11
; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm18
; AVX512-NEXT:    vpermt2q %zmm0, %zmm9, %zmm18
; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512-NEXT:    vpermt2q %zmm0, %zmm15, %zmm19
; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512-NEXT:    vpermt2q %zmm0, %zmm22, %zmm21
; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm11
; AVX512-NEXT:    vmovdqa64 1728(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 1664(%rdi), %zmm8
; AVX512-NEXT:    vmovdqa64 %zmm8, %zmm14
; AVX512-NEXT:    vpermt2q %zmm0, %zmm9, %zmm14
; AVX512-NEXT:    vmovdqa64 %zmm8, %zmm16
; AVX512-NEXT:    vpermt2q %zmm0, %zmm15, %zmm16
; AVX512-NEXT:    vmovdqa64 %zmm8, %zmm17
; AVX512-NEXT:    vpermt2q %zmm0, %zmm22, %zmm17
; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm8
; AVX512-NEXT:    vmovdqa64 1600(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 1536(%rdi), %zmm3
; AVX512-NEXT:    vmovdqa64 %zmm3, %zmm6
; AVX512-NEXT:    vpermt2q %zmm0, %zmm9, %zmm6
; AVX512-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512-NEXT:    vpermt2q %zmm0, %zmm15, %zmm7
; AVX512-NEXT:    vmovdqa64 %zmm3, %zmm12
; AVX512-NEXT:    vpermt2q %zmm0, %zmm22, %zmm12
; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm3
; AVX512-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512-NEXT:    vmovdqa64 192(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm4
; AVX512-NEXT:    vpermt2q %zmm0, %zmm9, %zmm4
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512-NEXT:    vpermt2q %zmm0, %zmm15, %zmm5
; AVX512-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512-NEXT:    vpermt2q %zmm0, %zmm22, %zmm10
; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm2
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm0
; AVX512-NEXT:    vpermi2q %zmm0, %zmm1, %zmm9
; AVX512-NEXT:    vpermi2q %zmm0, %zmm1, %zmm15
; AVX512-NEXT:    vpermi2q %zmm0, %zmm1, %zmm22
; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm1
; AVX512-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm31 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm31 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm26[0,1,2,3],zmm30[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm30 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm30 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm14[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm18[0,1,2,3],zmm20[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm9[0,1,2,3],zmm4[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm7[0,1,2,3],zmm16[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm16 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm16 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm18 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm18 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm20 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm20 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm29, %zmm29 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm29 = zmm29[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm25[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm15[0,1,2,3],zmm5[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm17[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm15 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm15 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm17 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm17 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm25 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm25 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm4 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm4 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm27[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm22[0,1,2,3],zmm10[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm22 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm22 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm27 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm27 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm7 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm7 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm24, %zmm24 # 64-byte Folded Reload
; AVX512-NEXT:    # zmm24 = zmm24[0,1,2,3],mem[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm23[0,1,2,3],zmm28[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm11[0,1,2,3],zmm13[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm8[4,5,6,7]
; AVX512-NEXT:    vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,2,3],zmm2[4,5,6,7]
; AVX512-NEXT:    vmovdqa64 %zmm14, 448(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm6, 384(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm30, 320(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm26, 256(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm31, 192(%rsi)
; AVX512-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512-NEXT:    vmovaps %zmm2, 128(%rsi)
; AVX512-NEXT:    vmovups (%rsp), %zmm2 # 64-byte Reload
; AVX512-NEXT:    vmovaps %zmm2, 64(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm9, (%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm19, 448(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm29, 256(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm20, 320(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm18, 128(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm16, 192(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512-NEXT:    vmovaps %zmm2, 64(%rdx)
; AVX512-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512-NEXT:    vmovaps %zmm2, 384(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm21, 448(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm0, 256(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm4, 320(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm17, 192(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm10, (%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm15, 64(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm12, 384(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm3, 384(%r8)
; AVX512-NEXT:    vmovdqa64 %zmm11, 448(%r8)
; AVX512-NEXT:    vmovdqa64 %zmm23, 256(%r8)
; AVX512-NEXT:    vmovdqa64 %zmm24, 320(%r8)
; AVX512-NEXT:    vmovdqa64 %zmm7, 128(%r8)
; AVX512-NEXT:    vmovdqa64 %zmm27, 192(%r8)
; AVX512-NEXT:    vmovdqa64 %zmm1, (%r8)
; AVX512-NEXT:    vmovdqa64 %zmm22, 64(%r8)
; AVX512-NEXT:    addq $2056, %rsp # imm = 0x808
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i64_stride4_vf64:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    subq $2056, %rsp # imm = 0x808
; AVX512-FCP-NEXT:    vmovdqa64 1472(%rdi), %zmm24
; AVX512-FCP-NEXT:    vmovdqa64 1408(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm26
; AVX512-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm28
; AVX512-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm25
; AVX512-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm30
; AVX512-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm29
; AVX512-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm6
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [0,4,8,12,0,4,8,12]
; AVX512-FCP-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm29, %zmm9, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm30, %zmm9, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, (%rsp) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm27, %zmm9, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm25, %zmm9, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm28, %zmm9, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm26, %zmm9, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm24, %zmm9, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [1,5,9,13,1,5,9,13]
; AVX512-FCP-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm29, %zmm15, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm30, %zmm15, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm28, %zmm15, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm26, %zmm15, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm27, %zmm15, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm25, %zmm15, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm22 = [2,6,10,14,2,6,10,14]
; AVX512-FCP-NEXT:    # zmm22 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm29, %zmm22, %zmm7
; AVX512-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [3,7,11,15,3,7,11,15]
; AVX512-FCP-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpermt2q %zmm29, %zmm31, %zmm6
; AVX512-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, %zmm6
; AVX512-FCP-NEXT:    vpermt2q %zmm30, %zmm22, %zmm6
; AVX512-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2q %zmm30, %zmm31, %zmm5
; AVX512-FCP-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512-FCP-NEXT:    vpermt2q %zmm28, %zmm22, %zmm5
; AVX512-FCP-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2q %zmm28, %zmm31, %zmm2
; AVX512-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512-FCP-NEXT:    vpermt2q %zmm26, %zmm22, %zmm2
; AVX512-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2q %zmm26, %zmm31, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm4, %zmm1
; AVX512-FCP-NEXT:    vpermt2q %zmm27, %zmm22, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2q %zmm27, %zmm31, %zmm4
; AVX512-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, %zmm1
; AVX512-FCP-NEXT:    vpermt2q %zmm25, %zmm22, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm3
; AVX512-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512-FCP-NEXT:    vpermt2q %zmm24, %zmm15, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512-FCP-NEXT:    vpermt2q %zmm24, %zmm22, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2q %zmm24, %zmm31, %zmm0
; AVX512-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 1344(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 1280(%rdi), %zmm24
; AVX512-FCP-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm24
; AVX512-FCP-NEXT:    vmovdqa64 1216(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 1152(%rdi), %zmm28
; AVX512-FCP-NEXT:    vmovdqa64 %zmm28, %zmm30
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm30
; AVX512-FCP-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm28
; AVX512-FCP-NEXT:    vmovdqa64 1088(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 1024(%rdi), %zmm23
; AVX512-FCP-NEXT:    vmovdqa64 %zmm23, %zmm26
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm26
; AVX512-FCP-NEXT:    vmovdqa64 %zmm23, %zmm29
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm29
; AVX512-FCP-NEXT:    vmovdqa64 %zmm23, %zmm1
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512-FCP-NEXT:    vmovdqa64 1984(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 1920(%rdi), %zmm13
; AVX512-FCP-NEXT:    vmovdqa64 %zmm13, %zmm20
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm20
; AVX512-FCP-NEXT:    vmovdqa64 %zmm13, %zmm25
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm25
; AVX512-FCP-NEXT:    vmovdqa64 %zmm13, %zmm27
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm27
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm13
; AVX512-FCP-NEXT:    vmovdqa64 1856(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 1792(%rdi), %zmm11
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, %zmm18
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm18
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm19
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm21
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm11
; AVX512-FCP-NEXT:    vmovdqa64 1728(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 1664(%rdi), %zmm8
; AVX512-FCP-NEXT:    vmovdqa64 %zmm8, %zmm14
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm14
; AVX512-FCP-NEXT:    vmovdqa64 %zmm8, %zmm16
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm16
; AVX512-FCP-NEXT:    vmovdqa64 %zmm8, %zmm17
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm17
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm8
; AVX512-FCP-NEXT:    vmovdqa64 1600(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 1536(%rdi), %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, %zmm6
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm6
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm7
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, %zmm12
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm12
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm4
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm4
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm5
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm10
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm0
; AVX512-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm9
; AVX512-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm15
; AVX512-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm22
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm1
; AVX512-FCP-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm31 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm31 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm26[0,1,2,3],zmm30[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm30 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm30 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm14[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm18[0,1,2,3],zmm20[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm9[0,1,2,3],zmm4[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm7[0,1,2,3],zmm16[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm16 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm16 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm18 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm18 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm20 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm20 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm29, %zmm29 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm29 = zmm29[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm25[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm15[0,1,2,3],zmm5[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm17[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm15 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm15 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm17 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm17 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm25 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm25 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm4 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm4 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm27[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm22[0,1,2,3],zmm10[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm22 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm22 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm27 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm27 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm7 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm7 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm24, %zmm24 # 64-byte Folded Reload
; AVX512-FCP-NEXT:    # zmm24 = zmm24[0,1,2,3],mem[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm23[0,1,2,3],zmm28[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm11[0,1,2,3],zmm13[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm8[4,5,6,7]
; AVX512-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,2,3],zmm2[4,5,6,7]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm14, 448(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm6, 384(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm30, 320(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm26, 256(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm31, 192(%rsi)
; AVX512-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovaps %zmm2, 128(%rsi)
; AVX512-FCP-NEXT:    vmovups (%rsp), %zmm2 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovaps %zmm2, 64(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm9, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm19, 448(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm29, 256(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm20, 320(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm18, 128(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm16, 192(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovaps %zmm2, 64(%rdx)
; AVX512-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovaps %zmm2, 384(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm21, 448(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, 256(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm4, 320(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm17, 192(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm10, (%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm15, 64(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm12, 384(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, 384(%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, 448(%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm23, 256(%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm24, 320(%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, 128(%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm27, 192(%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, (%r8)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm22, 64(%r8)
; AVX512-FCP-NEXT:    addq $2056, %rsp # imm = 0x808
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i64_stride4_vf64:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    subq $2056, %rsp # imm = 0x808
; AVX512DQ-NEXT:    vmovdqa64 1472(%rdi), %zmm24
; AVX512DQ-NEXT:    vmovdqa64 1408(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 832(%rdi), %zmm26
; AVX512DQ-NEXT:    vmovdqa64 768(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 960(%rdi), %zmm28
; AVX512DQ-NEXT:    vmovdqa64 896(%rdi), %zmm2
; AVX512DQ-NEXT:    vmovdqa64 576(%rdi), %zmm25
; AVX512DQ-NEXT:    vmovdqa64 512(%rdi), %zmm3
; AVX512DQ-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512DQ-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %zmm30
; AVX512DQ-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512DQ-NEXT:    vmovdqa64 448(%rdi), %zmm29
; AVX512DQ-NEXT:    vmovdqa64 384(%rdi), %zmm6
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [0,4,8,12,0,4,8,12]
; AVX512DQ-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm29, %zmm9, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm30, %zmm9, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, (%rsp) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm27, %zmm9, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm25, %zmm9, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm28, %zmm9, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm26, %zmm9, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm24, %zmm9, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [1,5,9,13,1,5,9,13]
; AVX512DQ-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm29, %zmm15, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm30, %zmm15, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm28, %zmm15, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm26, %zmm15, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm27, %zmm15, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm25, %zmm15, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm22 = [2,6,10,14,2,6,10,14]
; AVX512DQ-NEXT:    # zmm22 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm29, %zmm22, %zmm7
; AVX512DQ-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [3,7,11,15,3,7,11,15]
; AVX512DQ-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpermt2q %zmm29, %zmm31, %zmm6
; AVX512DQ-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, %zmm6
; AVX512DQ-NEXT:    vpermt2q %zmm30, %zmm22, %zmm6
; AVX512DQ-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2q %zmm30, %zmm31, %zmm5
; AVX512DQ-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512DQ-NEXT:    vpermt2q %zmm28, %zmm22, %zmm5
; AVX512DQ-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2q %zmm28, %zmm31, %zmm2
; AVX512DQ-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-NEXT:    vpermt2q %zmm26, %zmm22, %zmm2
; AVX512DQ-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2q %zmm26, %zmm31, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm4, %zmm1
; AVX512DQ-NEXT:    vpermt2q %zmm27, %zmm22, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2q %zmm27, %zmm31, %zmm4
; AVX512DQ-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, %zmm1
; AVX512DQ-NEXT:    vpermt2q %zmm25, %zmm22, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2q %zmm25, %zmm31, %zmm3
; AVX512DQ-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-NEXT:    vpermt2q %zmm24, %zmm15, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-NEXT:    vpermt2q %zmm24, %zmm22, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2q %zmm24, %zmm31, %zmm0
; AVX512DQ-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 1344(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 1280(%rdi), %zmm24
; AVX512DQ-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm9, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm31, %zmm24
; AVX512DQ-NEXT:    vmovdqa64 1216(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 1152(%rdi), %zmm28
; AVX512DQ-NEXT:    vmovdqa64 %zmm28, %zmm30
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm9, %zmm30
; AVX512DQ-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm31, %zmm28
; AVX512DQ-NEXT:    vmovdqa64 1088(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 1024(%rdi), %zmm23
; AVX512DQ-NEXT:    vmovdqa64 %zmm23, %zmm26
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm9, %zmm26
; AVX512DQ-NEXT:    vmovdqa64 %zmm23, %zmm29
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm15, %zmm29
; AVX512DQ-NEXT:    vmovdqa64 %zmm23, %zmm1
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512DQ-NEXT:    vmovdqa64 1984(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 1920(%rdi), %zmm13
; AVX512DQ-NEXT:    vmovdqa64 %zmm13, %zmm20
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm9, %zmm20
; AVX512DQ-NEXT:    vmovdqa64 %zmm13, %zmm25
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm15, %zmm25
; AVX512DQ-NEXT:    vmovdqa64 %zmm13, %zmm27
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm22, %zmm27
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm31, %zmm13
; AVX512DQ-NEXT:    vmovdqa64 1856(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 1792(%rdi), %zmm11
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, %zmm18
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm9, %zmm18
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm15, %zmm19
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm22, %zmm21
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm31, %zmm11
; AVX512DQ-NEXT:    vmovdqa64 1728(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 1664(%rdi), %zmm8
; AVX512DQ-NEXT:    vmovdqa64 %zmm8, %zmm14
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm9, %zmm14
; AVX512DQ-NEXT:    vmovdqa64 %zmm8, %zmm16
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm15, %zmm16
; AVX512DQ-NEXT:    vmovdqa64 %zmm8, %zmm17
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm22, %zmm17
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm31, %zmm8
; AVX512DQ-NEXT:    vmovdqa64 1600(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 1536(%rdi), %zmm3
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, %zmm6
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm9, %zmm6
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm15, %zmm7
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, %zmm12
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm22, %zmm12
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm31, %zmm3
; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm4
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm9, %zmm4
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm15, %zmm5
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm22, %zmm10
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm31, %zmm2
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm0
; AVX512DQ-NEXT:    vpermi2q %zmm0, %zmm1, %zmm9
; AVX512DQ-NEXT:    vpermi2q %zmm0, %zmm1, %zmm15
; AVX512DQ-NEXT:    vpermi2q %zmm0, %zmm1, %zmm22
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm31, %zmm1
; AVX512DQ-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm31 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm31 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm26[0,1,2,3],zmm30[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm30 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm30 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm14[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm18[0,1,2,3],zmm20[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm9[0,1,2,3],zmm4[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm7[0,1,2,3],zmm16[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm16 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm16 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm18 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm18 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm20 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm20 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm29, %zmm29 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm29 = zmm29[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm25[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm15[0,1,2,3],zmm5[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm17[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm15 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm15 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm17 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm17 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm25 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm25 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm4 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm4 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm27[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm22[0,1,2,3],zmm10[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm22 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm22 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm27 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm27 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm7 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm7 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm24, %zmm24 # 64-byte Folded Reload
; AVX512DQ-NEXT:    # zmm24 = zmm24[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm23[0,1,2,3],zmm28[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm11[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm8[4,5,6,7]
; AVX512DQ-NEXT:    vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-NEXT:    vmovdqa64 %zmm14, 448(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm6, 384(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm30, 320(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm26, 256(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm31, 192(%rsi)
; AVX512DQ-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-NEXT:    vmovaps %zmm2, 128(%rsi)
; AVX512DQ-NEXT:    vmovups (%rsp), %zmm2 # 64-byte Reload
; AVX512DQ-NEXT:    vmovaps %zmm2, 64(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm9, (%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm19, 448(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm29, 256(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm20, 320(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm18, 128(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm16, 192(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-NEXT:    vmovaps %zmm2, 64(%rdx)
; AVX512DQ-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-NEXT:    vmovaps %zmm2, 384(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm21, 448(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, 256(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm4, 320(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm17, 192(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm10, (%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm15, 64(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm12, 384(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, 384(%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, 448(%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm23, 256(%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm24, 320(%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, 128(%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm27, 192(%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm1, (%r8)
; AVX512DQ-NEXT:    vmovdqa64 %zmm22, 64(%r8)
; AVX512DQ-NEXT:    addq $2056, %rsp # imm = 0x808
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride4_vf64:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    subq $2056, %rsp # imm = 0x808
; AVX512DQ-FCP-NEXT:    vmovdqa64 1472(%rdi), %zmm24
; AVX512DQ-FCP-NEXT:    vmovdqa64 1408(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm26
; AVX512DQ-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm28
; AVX512DQ-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm25
; AVX512DQ-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512DQ-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm30
; AVX512DQ-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512DQ-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm29
; AVX512DQ-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm6
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [0,4,8,12,0,4,8,12]
; AVX512DQ-FCP-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm29, %zmm9, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm30, %zmm9, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, (%rsp) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm27, %zmm9, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm25, %zmm9, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm28, %zmm9, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm26, %zmm9, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm24, %zmm9, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [1,5,9,13,1,5,9,13]
; AVX512DQ-FCP-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm29, %zmm15, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm30, %zmm15, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm28, %zmm15, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm26, %zmm15, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm27, %zmm15, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm25, %zmm15, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm22 = [2,6,10,14,2,6,10,14]
; AVX512DQ-FCP-NEXT:    # zmm22 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm29, %zmm22, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [3,7,11,15,3,7,11,15]
; AVX512DQ-FCP-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm29, %zmm31, %zmm6
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, %zmm6
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm30, %zmm22, %zmm6
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm30, %zmm31, %zmm5
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm28, %zmm22, %zmm5
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm28, %zmm31, %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm26, %zmm22, %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm26, %zmm31, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm4, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm27, %zmm22, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm27, %zmm31, %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm25, %zmm22, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm24, %zmm15, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm24, %zmm22, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm24, %zmm31, %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 1344(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 1280(%rdi), %zmm24
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm24
; AVX512DQ-FCP-NEXT:    vmovdqa64 1216(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 1152(%rdi), %zmm28
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm28, %zmm30
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm30
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm28
; AVX512DQ-FCP-NEXT:    vmovdqa64 1088(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 1024(%rdi), %zmm23
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm23, %zmm26
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm26
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm23, %zmm29
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm29
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm23, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512DQ-FCP-NEXT:    vmovdqa64 1984(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 1920(%rdi), %zmm13
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm13, %zmm20
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm20
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm13, %zmm25
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm25
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm13, %zmm27
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm27
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm13
; AVX512DQ-FCP-NEXT:    vmovdqa64 1856(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 1792(%rdi), %zmm11
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, %zmm18
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm18
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm19
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm21
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm11
; AVX512DQ-FCP-NEXT:    vmovdqa64 1728(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 1664(%rdi), %zmm8
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm8, %zmm14
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm14
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm8, %zmm16
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm16
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm8, %zmm17
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm17
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm8
; AVX512DQ-FCP-NEXT:    vmovdqa64 1600(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 1536(%rdi), %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, %zmm6
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm6
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, %zmm12
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm12
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm4
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm5
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm10
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm9
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm15
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm22
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm31 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm31 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm26[0,1,2,3],zmm30[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm30 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm30 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm14[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm18[0,1,2,3],zmm20[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm9[0,1,2,3],zmm4[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm7[0,1,2,3],zmm16[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm16 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm16 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm18 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm18 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm20 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm20 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm29, %zmm29 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm29 = zmm29[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm25[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm15[0,1,2,3],zmm5[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm17[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm15 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm15 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm17 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm17 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm25 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm25 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm4 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm4 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm27[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm22[0,1,2,3],zmm10[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm22 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm22 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm27 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm27 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm7 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm7 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm24, %zmm24 # 64-byte Folded Reload
; AVX512DQ-FCP-NEXT:    # zmm24 = zmm24[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm23[0,1,2,3],zmm28[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm11[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm8[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm14, 448(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm6, 384(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm30, 320(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm26, 256(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm31, 192(%rsi)
; AVX512DQ-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovaps %zmm2, 128(%rsi)
; AVX512DQ-FCP-NEXT:    vmovups (%rsp), %zmm2 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovaps %zmm2, 64(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm9, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm19, 448(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm29, 256(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm20, 320(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm18, 128(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm16, 192(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovaps %zmm2, 64(%rdx)
; AVX512DQ-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovaps %zmm2, 384(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm21, 448(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, 256(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm4, 320(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm17, 192(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm10, (%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm15, 64(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm12, 384(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, 384(%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, 448(%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm23, 256(%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm24, 320(%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, 128(%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm27, 192(%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, (%r8)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm22, 64(%r8)
; AVX512DQ-FCP-NEXT:    addq $2056, %rsp # imm = 0x808
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i64_stride4_vf64:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    subq $2056, %rsp # imm = 0x808
; AVX512BW-NEXT:    vmovdqa64 1472(%rdi), %zmm24
; AVX512BW-NEXT:    vmovdqa64 1408(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm26
; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm28
; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm25
; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm3
; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm30
; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm29
; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm6
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [0,4,8,12,0,4,8,12]
; AVX512BW-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm9, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm9, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, (%rsp) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm9, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm9, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm9, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm9, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm9, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [1,5,9,13,1,5,9,13]
; AVX512BW-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm15, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm15, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm15, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm15, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm15, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm15, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm22 = [2,6,10,14,2,6,10,14]
; AVX512BW-NEXT:    # zmm22 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm22, %zmm7
; AVX512BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [3,7,11,15,3,7,11,15]
; AVX512BW-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm31, %zmm6
; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm6
; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm22, %zmm6
; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm31, %zmm5
; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm22, %zmm5
; AVX512BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm31, %zmm2
; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm22, %zmm2
; AVX512BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm31, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm1
; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm22, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm31, %zmm4
; AVX512BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm1
; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm22, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm31, %zmm3
; AVX512BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm15, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm22, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2q %zmm24, %zmm31, %zmm0
; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 1344(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 1280(%rdi), %zmm24
; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm24
; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %zmm28
; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm30
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm30
; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm28
; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 1024(%rdi), %zmm23
; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm26
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm26
; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm29
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm29
; AVX512BW-NEXT:    vmovdqa64 %zmm23, %zmm1
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512BW-NEXT:    vmovdqa64 1984(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 1920(%rdi), %zmm13
; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm20
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm20
; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm25
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm25
; AVX512BW-NEXT:    vmovdqa64 %zmm13, %zmm27
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm27
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm13
; AVX512BW-NEXT:    vmovdqa64 1856(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 1792(%rdi), %zmm11
; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm18
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm18
; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm19
; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm21
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm11
; AVX512BW-NEXT:    vmovdqa64 1728(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 1664(%rdi), %zmm8
; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm14
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm14
; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm16
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm16
; AVX512BW-NEXT:    vmovdqa64 %zmm8, %zmm17
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm17
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm8
; AVX512BW-NEXT:    vmovdqa64 1600(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 1536(%rdi), %zmm3
; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm6
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm6
; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm7
; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm12
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm12
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm3
; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm4
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm4
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm5
; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm10
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm2
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm0
; AVX512BW-NEXT:    vpermi2q %zmm0, %zmm1, %zmm9
; AVX512BW-NEXT:    vpermi2q %zmm0, %zmm1, %zmm15
; AVX512BW-NEXT:    vpermi2q %zmm0, %zmm1, %zmm22
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm1
; AVX512BW-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm31 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm31 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm26[0,1,2,3],zmm30[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm30 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm30 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm14[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm18[0,1,2,3],zmm20[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm9[0,1,2,3],zmm4[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm7[0,1,2,3],zmm16[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm16 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm16 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm18 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm18 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm20 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm20 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm29, %zmm29 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm29 = zmm29[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm25[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm15[0,1,2,3],zmm5[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm17[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm15 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm15 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm17 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm17 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm25 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm25 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm4 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm4 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm27[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm22[0,1,2,3],zmm10[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm22 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm22 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm27 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm27 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm7 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm7 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm24, %zmm24 # 64-byte Folded Reload
; AVX512BW-NEXT:    # zmm24 = zmm24[0,1,2,3],mem[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm23[0,1,2,3],zmm28[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm11[0,1,2,3],zmm13[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm8[4,5,6,7]
; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,2,3],zmm2[4,5,6,7]
; AVX512BW-NEXT:    vmovdqa64 %zmm14, 448(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm6, 384(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm30, 320(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm26, 256(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm31, 192(%rsi)
; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-NEXT:    vmovaps %zmm2, 128(%rsi)
; AVX512BW-NEXT:    vmovups (%rsp), %zmm2 # 64-byte Reload
; AVX512BW-NEXT:    vmovaps %zmm2, 64(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm9, (%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm19, 448(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm29, 256(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm20, 320(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm18, 128(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm16, 192(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-NEXT:    vmovaps %zmm2, 64(%rdx)
; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-NEXT:    vmovaps %zmm2, 384(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm21, 448(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm0, 256(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm4, 320(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm17, 192(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm10, (%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm15, 64(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm12, 384(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm3, 384(%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm11, 448(%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm23, 256(%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm24, 320(%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm7, 128(%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm27, 192(%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm1, (%r8)
; AVX512BW-NEXT:    vmovdqa64 %zmm22, 64(%r8)
; AVX512BW-NEXT:    addq $2056, %rsp # imm = 0x808
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i64_stride4_vf64:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    subq $2056, %rsp # imm = 0x808
; AVX512BW-FCP-NEXT:    vmovdqa64 1472(%rdi), %zmm24
; AVX512BW-FCP-NEXT:    vmovdqa64 1408(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm26
; AVX512BW-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm28
; AVX512BW-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm25
; AVX512BW-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512BW-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm30
; AVX512BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm29
; AVX512BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm6
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [0,4,8,12,0,4,8,12]
; AVX512BW-FCP-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm29, %zmm9, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm30, %zmm9, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, (%rsp) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm27, %zmm9, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm25, %zmm9, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm28, %zmm9, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm26, %zmm9, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm24, %zmm9, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [1,5,9,13,1,5,9,13]
; AVX512BW-FCP-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm29, %zmm15, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm30, %zmm15, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm28, %zmm15, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm26, %zmm15, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm27, %zmm15, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm25, %zmm15, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm22 = [2,6,10,14,2,6,10,14]
; AVX512BW-FCP-NEXT:    # zmm22 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm29, %zmm22, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [3,7,11,15,3,7,11,15]
; AVX512BW-FCP-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm29, %zmm31, %zmm6
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm6
; AVX512BW-FCP-NEXT:    vpermt2q %zmm30, %zmm22, %zmm6
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2q %zmm30, %zmm31, %zmm5
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512BW-FCP-NEXT:    vpermt2q %zmm28, %zmm22, %zmm5
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2q %zmm28, %zmm31, %zmm2
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512BW-FCP-NEXT:    vpermt2q %zmm26, %zmm22, %zmm2
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2q %zmm26, %zmm31, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2q %zmm27, %zmm22, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2q %zmm27, %zmm31, %zmm4
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2q %zmm25, %zmm22, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm3
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2q %zmm24, %zmm15, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2q %zmm24, %zmm22, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2q %zmm24, %zmm31, %zmm0
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 1344(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 1280(%rdi), %zmm24
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm24
; AVX512BW-FCP-NEXT:    vmovdqa64 1216(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 1152(%rdi), %zmm28
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm28, %zmm30
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm30
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm28
; AVX512BW-FCP-NEXT:    vmovdqa64 1088(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 1024(%rdi), %zmm23
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm23, %zmm26
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm26
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm23, %zmm29
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm29
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm23, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512BW-FCP-NEXT:    vmovdqa64 1984(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 1920(%rdi), %zmm13
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm13, %zmm20
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm20
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm13, %zmm25
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm25
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm13, %zmm27
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm27
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm13
; AVX512BW-FCP-NEXT:    vmovdqa64 1856(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 1792(%rdi), %zmm11
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm18
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm18
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm19
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm21
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm11
; AVX512BW-FCP-NEXT:    vmovdqa64 1728(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 1664(%rdi), %zmm8
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm8, %zmm14
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm14
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm8, %zmm16
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm16
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm8, %zmm17
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm17
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm8
; AVX512BW-FCP-NEXT:    vmovdqa64 1600(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 1536(%rdi), %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm6
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm6
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm12
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm12
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm4
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm4
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm5
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm10
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm9
; AVX512BW-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm15
; AVX512BW-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm22
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm1
; AVX512BW-FCP-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm31 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm31 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm26[0,1,2,3],zmm30[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm30 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm30 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm14[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm18[0,1,2,3],zmm20[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm9[0,1,2,3],zmm4[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm7[0,1,2,3],zmm16[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm16 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm16 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm18 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm18 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm20 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm20 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm29, %zmm29 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm29 = zmm29[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm25[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm15[0,1,2,3],zmm5[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm17[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm15 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm15 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm17 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm17 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm25 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm25 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm4 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm4 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm27[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm22[0,1,2,3],zmm10[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm22 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm22 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm27 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm27 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm7 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm7 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm24, %zmm24 # 64-byte Folded Reload
; AVX512BW-FCP-NEXT:    # zmm24 = zmm24[0,1,2,3],mem[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm23[0,1,2,3],zmm28[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm11[0,1,2,3],zmm13[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm8[4,5,6,7]
; AVX512BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,2,3],zmm2[4,5,6,7]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm14, 448(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm6, 384(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm30, 320(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm26, 256(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm31, 192(%rsi)
; AVX512BW-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovaps %zmm2, 128(%rsi)
; AVX512BW-FCP-NEXT:    vmovups (%rsp), %zmm2 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovaps %zmm2, 64(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm9, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm19, 448(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm29, 256(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm20, 320(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm18, 128(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm16, 192(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512BW-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovaps %zmm2, 64(%rdx)
; AVX512BW-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovaps %zmm2, 384(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm21, 448(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, 256(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, 320(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm17, 192(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm10, (%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm15, 64(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm12, 384(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, 384(%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, 448(%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm23, 256(%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm24, 320(%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, 128(%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm27, 192(%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, (%r8)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm22, 64(%r8)
; AVX512BW-FCP-NEXT:    addq $2056, %rsp # imm = 0x808
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i64_stride4_vf64:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    subq $2056, %rsp # imm = 0x808
; AVX512DQ-BW-NEXT:    vmovdqa64 1472(%rdi), %zmm24
; AVX512DQ-BW-NEXT:    vmovdqa64 1408(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 832(%rdi), %zmm26
; AVX512DQ-BW-NEXT:    vmovdqa64 768(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 960(%rdi), %zmm28
; AVX512DQ-BW-NEXT:    vmovdqa64 896(%rdi), %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 576(%rdi), %zmm25
; AVX512DQ-BW-NEXT:    vmovdqa64 512(%rdi), %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512DQ-BW-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512DQ-BW-NEXT:    vmovdqa64 320(%rdi), %zmm30
; AVX512DQ-BW-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512DQ-BW-NEXT:    vmovdqa64 448(%rdi), %zmm29
; AVX512DQ-BW-NEXT:    vmovdqa64 384(%rdi), %zmm6
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [0,4,8,12,0,4,8,12]
; AVX512DQ-BW-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm29, %zmm9, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm30, %zmm9, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, (%rsp) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm27, %zmm9, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm25, %zmm9, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm28, %zmm9, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm26, %zmm9, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm24, %zmm9, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [1,5,9,13,1,5,9,13]
; AVX512DQ-BW-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm29, %zmm15, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm30, %zmm15, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm28, %zmm15, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm26, %zmm15, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm27, %zmm15, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm25, %zmm15, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm22 = [2,6,10,14,2,6,10,14]
; AVX512DQ-BW-NEXT:    # zmm22 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm29, %zmm22, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [3,7,11,15,3,7,11,15]
; AVX512DQ-BW-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm29, %zmm31, %zmm6
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, %zmm6
; AVX512DQ-BW-NEXT:    vpermt2q %zmm30, %zmm22, %zmm6
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2q %zmm30, %zmm31, %zmm5
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512DQ-BW-NEXT:    vpermt2q %zmm28, %zmm22, %zmm5
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2q %zmm28, %zmm31, %zmm2
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, %zmm2
; AVX512DQ-BW-NEXT:    vpermt2q %zmm26, %zmm22, %zmm2
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2q %zmm26, %zmm31, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm4, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2q %zmm27, %zmm22, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2q %zmm27, %zmm31, %zmm4
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2q %zmm25, %zmm22, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2q %zmm25, %zmm31, %zmm3
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2q %zmm24, %zmm15, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2q %zmm24, %zmm22, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2q %zmm24, %zmm31, %zmm0
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 1344(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 1280(%rdi), %zmm24
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm24
; AVX512DQ-BW-NEXT:    vmovdqa64 1216(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 1152(%rdi), %zmm28
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm28, %zmm30
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm30
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm28
; AVX512DQ-BW-NEXT:    vmovdqa64 1088(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 1024(%rdi), %zmm23
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm23, %zmm26
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm26
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm23, %zmm29
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm29
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm23, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512DQ-BW-NEXT:    vmovdqa64 1984(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 1920(%rdi), %zmm13
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm13, %zmm20
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm20
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm13, %zmm25
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm25
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm13, %zmm27
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm27
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm13
; AVX512DQ-BW-NEXT:    vmovdqa64 1856(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 1792(%rdi), %zmm11
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, %zmm18
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm18
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm19
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm21
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm11
; AVX512DQ-BW-NEXT:    vmovdqa64 1728(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 1664(%rdi), %zmm8
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm8, %zmm14
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm14
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm8, %zmm16
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm16
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm8, %zmm17
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm17
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm8
; AVX512DQ-BW-NEXT:    vmovdqa64 1600(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 1536(%rdi), %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, %zmm6
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm6
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, %zmm12
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm12
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 192(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm4
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm9, %zmm4
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm15, %zmm5
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm22, %zmm10
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 (%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 64(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vpermi2q %zmm0, %zmm1, %zmm9
; AVX512DQ-BW-NEXT:    vpermi2q %zmm0, %zmm1, %zmm15
; AVX512DQ-BW-NEXT:    vpermi2q %zmm0, %zmm1, %zmm22
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm1
; AVX512DQ-BW-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm31 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm31 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm26[0,1,2,3],zmm30[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm30 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm30 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm14[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm18[0,1,2,3],zmm20[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm9[0,1,2,3],zmm4[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm7[0,1,2,3],zmm16[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm16 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm16 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm18 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm18 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm20 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm20 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm29, %zmm29 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm29 = zmm29[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm25[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm15[0,1,2,3],zmm5[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm17[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm15 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm15 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm17 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm17 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm25 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm25 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm4 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm4 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm27[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm22[0,1,2,3],zmm10[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm22 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm22 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm27 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm27 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm7 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm7 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm24, %zmm24 # 64-byte Folded Reload
; AVX512DQ-BW-NEXT:    # zmm24 = zmm24[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm23[0,1,2,3],zmm28[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm11[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm8[4,5,6,7]
; AVX512DQ-BW-NEXT:    vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm14, 448(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm6, 384(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm30, 320(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm26, 256(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm31, 192(%rsi)
; AVX512DQ-BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovaps %zmm2, 128(%rsi)
; AVX512DQ-BW-NEXT:    vmovups (%rsp), %zmm2 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovaps %zmm2, 64(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm9, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm19, 448(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm29, 256(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm20, 320(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm18, 128(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm16, 192(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovaps %zmm2, 64(%rdx)
; AVX512DQ-BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovaps %zmm2, 384(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm21, 448(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, 256(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm4, 320(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm17, 192(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm10, (%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm15, 64(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm12, 384(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, 384(%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, 448(%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm23, 256(%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm24, 320(%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, 128(%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm27, 192(%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, (%r8)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm22, 64(%r8)
; AVX512DQ-BW-NEXT:    addq $2056, %rsp # imm = 0x808
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride4_vf64:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    subq $2056, %rsp # imm = 0x808
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1472(%rdi), %zmm24
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1408(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm26
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm28
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm25
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm27
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm30
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm5
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm29
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm6
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm9 = [0,4,8,12,0,4,8,12]
; AVX512DQ-BW-FCP-NEXT:    # zmm9 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm29, %zmm9, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm30, %zmm9, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, (%rsp) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm27, %zmm9, %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:    vpermt2q %zmm25, %zmm9, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm28, %zmm9, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm26, %zmm9, %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:    vpermt2q %zmm24, %zmm9, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm15 = [1,5,9,13,1,5,9,13]
; AVX512DQ-BW-FCP-NEXT:    # zmm15 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm29, %zmm15, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm30, %zmm15, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm28, %zmm15, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm1, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm26, %zmm15, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm27, %zmm15, %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:    vpermt2q %zmm25, %zmm15, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm22 = [2,6,10,14,2,6,10,14]
; AVX512DQ-BW-FCP-NEXT:    # zmm22 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm6, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm29, %zmm22, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti64x4 {{.*#+}} zmm31 = [3,7,11,15,3,7,11,15]
; AVX512DQ-BW-FCP-NEXT:    # zmm31 = mem[0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm29, %zmm31, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm30, %zmm22, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm30, %zmm31, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm28, %zmm22, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm28, %zmm31, %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:    vpermt2q %zmm26, %zmm22, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm26, %zmm31, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm27, %zmm22, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm27, %zmm31, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm25, %zmm22, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm24, %zmm15, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm24, %zmm22, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm24, %zmm31, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1344(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1280(%rdi), %zmm24
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm24, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm24
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1216(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1152(%rdi), %zmm28
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm28, %zmm30
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm30
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm28, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm28
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1088(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1024(%rdi), %zmm23
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm23, %zmm26
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm26
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm23, %zmm29
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm29
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm23, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1984(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1920(%rdi), %zmm13
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm13, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm13, %zmm25
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm25
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm13, %zmm27
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm27
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1856(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1792(%rdi), %zmm11
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm19
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm21
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm21
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm11
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1728(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1664(%rdi), %zmm8
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm8, %zmm14
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm14
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm8, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm8, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1600(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1536(%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm3, %zmm12
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm12
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm9, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm15, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, %zmm10
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm22, %zmm10
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm9
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm22
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 (%rsp), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm0, (%rsp) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm31 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm31 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm26 = zmm26[0,1,2,3],zmm30[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm30 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm30 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm14[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm14 = zmm18[0,1,2,3],zmm20[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm9 = zmm9[0,1,2,3],zmm4[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm7[0,1,2,3],zmm16[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm16 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm16 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm18 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm18 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm20 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm20 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm29, %zmm29 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm29 = zmm29[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm19 = zmm19[0,1,2,3],zmm25[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm5 = zmm15[0,1,2,3],zmm5[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm17[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm15 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm15 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm17 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm17 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm25 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm25 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm4 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm4 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm0 = zmm0[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm21 = zmm21[0,1,2,3],zmm27[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm10 = zmm22[0,1,2,3],zmm10[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm22 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm22 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm27 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm27 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm7 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm7, %zmm7 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm7 = zmm7[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 $228, {{[-0-9]+}}(%r{{[sb]}}p), %zmm24, %zmm24 # 64-byte Folded Reload
; AVX512DQ-BW-FCP-NEXT:    # zmm24 = zmm24[0,1,2,3],mem[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm23 = zmm23[0,1,2,3],zmm28[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm11 = zmm11[0,1,2,3],zmm13[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm8[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,2,3],zmm2[4,5,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm14, 448(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm6, 384(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm30, 320(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm26, 256(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm31, 192(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovaps %zmm2, 128(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovups (%rsp), %zmm2 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovaps %zmm2, 64(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm9, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm19, 448(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm29, 256(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm20, 320(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm18, 128(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm16, 192(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovaps %zmm2, 64(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm2 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovaps %zmm2, 384(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm21, 448(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, 256(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, 320(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm25, 128(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm17, 192(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm10, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm15, 64(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm12, 384(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm3, 384(%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, 448(%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm23, 256(%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm24, 320(%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, 128(%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm27, 192(%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm1, (%r8)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm22, 64(%r8)
; AVX512DQ-BW-FCP-NEXT:    addq $2056, %rsp # imm = 0x808
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <256 x i64>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <64 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60, i32 64, i32 68, i32 72, i32 76, i32 80, i32 84, i32 88, i32 92, i32 96, i32 100, i32 104, i32 108, i32 112, i32 116, i32 120, i32 124, i32 128, i32 132, i32 136, i32 140, i32 144, i32 148, i32 152, i32 156, i32 160, i32 164, i32 168, i32 172, i32 176, i32 180, i32 184, i32 188, i32 192, i32 196, i32 200, i32 204, i32 208, i32 212, i32 216, i32 220, i32 224, i32 228, i32 232, i32 236, i32 240, i32 244, i32 248, i32 252>
  %strided.vec1 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <64 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61, i32 65, i32 69, i32 73, i32 77, i32 81, i32 85, i32 89, i32 93, i32 97, i32 101, i32 105, i32 109, i32 113, i32 117, i32 121, i32 125, i32 129, i32 133, i32 137, i32 141, i32 145, i32 149, i32 153, i32 157, i32 161, i32 165, i32 169, i32 173, i32 177, i32 181, i32 185, i32 189, i32 193, i32 197, i32 201, i32 205, i32 209, i32 213, i32 217, i32 221, i32 225, i32 229, i32 233, i32 237, i32 241, i32 245, i32 249, i32 253>
  %strided.vec2 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <64 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62, i32 66, i32 70, i32 74, i32 78, i32 82, i32 86, i32 90, i32 94, i32 98, i32 102, i32 106, i32 110, i32 114, i32 118, i32 122, i32 126, i32 130, i32 134, i32 138, i32 142, i32 146, i32 150, i32 154, i32 158, i32 162, i32 166, i32 170, i32 174, i32 178, i32 182, i32 186, i32 190, i32 194, i32 198, i32 202, i32 206, i32 210, i32 214, i32 218, i32 222, i32 226, i32 230, i32 234, i32 238, i32 242, i32 246, i32 250, i32 254>
  %strided.vec3 = shufflevector <256 x i64> %wide.vec, <256 x i64> poison, <64 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63, i32 67, i32 71, i32 75, i32 79, i32 83, i32 87, i32 91, i32 95, i32 99, i32 103, i32 107, i32 111, i32 115, i32 119, i32 123, i32 127, i32 131, i32 135, i32 139, i32 143, i32 147, i32 151, i32 155, i32 159, i32 163, i32 167, i32 171, i32 175, i32 179, i32 183, i32 187, i32 191, i32 195, i32 199, i32 203, i32 207, i32 211, i32 215, i32 219, i32 223, i32 227, i32 231, i32 235, i32 239, i32 243, i32 247, i32 251, i32 255>
  store <64 x i64> %strided.vec0, ptr %out.vec0, align 64
  store <64 x i64> %strided.vec1, ptr %out.vec1, align 64
  store <64 x i64> %strided.vec2, ptr %out.vec2, align 64
  store <64 x i64> %strided.vec3, ptr %out.vec3, align 64
  ret void
}