llvm/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.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_stride3_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
; SSE-LABEL: load_i64_stride3_vf2:
; SSE:       # %bb.0:
; SSE-NEXT:    movapd (%rdi), %xmm0
; SSE-NEXT:    movapd 16(%rdi), %xmm1
; SSE-NEXT:    movapd 32(%rdi), %xmm2
; SSE-NEXT:    movapd %xmm1, %xmm3
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
; SSE-NEXT:    shufpd {{.*#+}} xmm0 = xmm0[1],xmm2[0]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT:    movapd %xmm3, (%rsi)
; SSE-NEXT:    movapd %xmm0, (%rdx)
; SSE-NEXT:    movapd %xmm2, (%rcx)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i64_stride3_vf2:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovdqa (%rdi), %xmm0
; AVX-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX-NEXT:    vpblendw {{.*#+}} xmm3 = xmm0[0,1,2,3],xmm1[4,5,6,7]
; AVX-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
; AVX-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],mem[4,5,6,7]
; AVX-NEXT:    vmovdqa %xmm3, (%rsi)
; AVX-NEXT:    vmovdqa %xmm0, (%rdx)
; AVX-NEXT:    vmovdqa %xmm1, (%rcx)
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i64_stride3_vf2:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX2-NEXT:    vpblendd {{.*#+}} xmm3 = xmm0[0,1],xmm1[2,3]
; AVX2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
; AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
; AVX2-NEXT:    vmovdqa %xmm3, (%rsi)
; AVX2-NEXT:    vmovdqa %xmm0, (%rdx)
; AVX2-NEXT:    vmovdqa %xmm1, (%rcx)
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i64_stride3_vf2:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-FP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-FP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX2-FP-NEXT:    vpblendd {{.*#+}} xmm3 = xmm0[0,1],xmm1[2,3]
; AVX2-FP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
; AVX2-FP-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
; AVX2-FP-NEXT:    vmovdqa %xmm3, (%rsi)
; AVX2-FP-NEXT:    vmovdqa %xmm0, (%rdx)
; AVX2-FP-NEXT:    vmovdqa %xmm1, (%rcx)
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i64_stride3_vf2:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-FCP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX2-FCP-NEXT:    vpblendd {{.*#+}} xmm3 = xmm0[0,1],xmm1[2,3]
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
; AVX2-FCP-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vmovdqa %xmm3, (%rsi)
; AVX2-FCP-NEXT:    vmovdqa %xmm0, (%rdx)
; AVX2-FCP-NEXT:    vmovdqa %xmm1, (%rcx)
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i64_stride3_vf2:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vpermpd {{.*#+}} zmm0 = mem[0,3,2,3,4,7,6,7]
; AVX512-NEXT:    vmovdqa 32(%rdi), %xmm1
; AVX512-NEXT:    vpalignr {{.*#+}} xmm2 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
; AVX512-NEXT:    vpblendd {{.*#+}} xmm1 = mem[0,1],xmm1[2,3]
; AVX512-NEXT:    vmovaps %xmm0, (%rsi)
; AVX512-NEXT:    vmovdqa %xmm2, (%rdx)
; AVX512-NEXT:    vmovdqa %xmm1, (%rcx)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i64_stride3_vf2:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovaps {{.*#+}} xmm0 = [1,4]
; AVX512-FCP-NEXT:    vmovaps (%rdi), %zmm1
; AVX512-FCP-NEXT:    vpermpd %zmm1, %zmm0, %zmm0
; AVX512-FCP-NEXT:    vpermpd {{.*#+}} zmm1 = zmm1[0,3,2,3,4,7,6,7]
; AVX512-FCP-NEXT:    vmovaps 16(%rdi), %xmm2
; AVX512-FCP-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
; AVX512-FCP-NEXT:    vmovaps %xmm1, (%rsi)
; AVX512-FCP-NEXT:    vmovaps %xmm0, (%rdx)
; AVX512-FCP-NEXT:    vmovaps %xmm2, (%rcx)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i64_stride3_vf2:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vpermpd {{.*#+}} zmm0 = mem[0,3,2,3,4,7,6,7]
; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %xmm1
; AVX512DQ-NEXT:    vpalignr {{.*#+}} xmm2 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
; AVX512DQ-NEXT:    vpblendd {{.*#+}} xmm1 = mem[0,1],xmm1[2,3]
; AVX512DQ-NEXT:    vmovaps %xmm0, (%rsi)
; AVX512DQ-NEXT:    vmovdqa %xmm2, (%rdx)
; AVX512DQ-NEXT:    vmovdqa %xmm1, (%rcx)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride3_vf2:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovaps {{.*#+}} xmm0 = [1,4]
; AVX512DQ-FCP-NEXT:    vmovaps (%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vpermpd %zmm1, %zmm0, %zmm0
; AVX512DQ-FCP-NEXT:    vpermpd {{.*#+}} zmm1 = zmm1[0,3,2,3,4,7,6,7]
; AVX512DQ-FCP-NEXT:    vmovaps 16(%rdi), %xmm2
; AVX512DQ-FCP-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
; AVX512DQ-FCP-NEXT:    vmovaps %xmm1, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovaps %xmm0, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovaps %xmm2, (%rcx)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i64_stride3_vf2:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vpermpd {{.*#+}} zmm0 = mem[0,3,2,3,4,7,6,7]
; AVX512BW-NEXT:    vmovdqa 32(%rdi), %xmm1
; AVX512BW-NEXT:    vpalignr {{.*#+}} xmm2 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
; AVX512BW-NEXT:    vpblendd {{.*#+}} xmm1 = mem[0,1],xmm1[2,3]
; AVX512BW-NEXT:    vmovaps %xmm0, (%rsi)
; AVX512BW-NEXT:    vmovdqa %xmm2, (%rdx)
; AVX512BW-NEXT:    vmovdqa %xmm1, (%rcx)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i64_stride3_vf2:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovaps {{.*#+}} xmm0 = [1,4]
; AVX512BW-FCP-NEXT:    vmovaps (%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vpermpd %zmm1, %zmm0, %zmm0
; AVX512BW-FCP-NEXT:    vpermpd {{.*#+}} zmm1 = zmm1[0,3,2,3,4,7,6,7]
; AVX512BW-FCP-NEXT:    vmovaps 16(%rdi), %xmm2
; AVX512BW-FCP-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
; AVX512BW-FCP-NEXT:    vmovaps %xmm1, (%rsi)
; AVX512BW-FCP-NEXT:    vmovaps %xmm0, (%rdx)
; AVX512BW-FCP-NEXT:    vmovaps %xmm2, (%rcx)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i64_stride3_vf2:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vpermpd {{.*#+}} zmm0 = mem[0,3,2,3,4,7,6,7]
; AVX512DQ-BW-NEXT:    vmovdqa 32(%rdi), %xmm1
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} xmm2 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
; AVX512DQ-BW-NEXT:    vpblendd {{.*#+}} xmm1 = mem[0,1],xmm1[2,3]
; AVX512DQ-BW-NEXT:    vmovaps %xmm0, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa %xmm2, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa %xmm1, (%rcx)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride3_vf2:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovaps {{.*#+}} xmm0 = [1,4]
; AVX512DQ-BW-FCP-NEXT:    vmovaps (%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermpd %zmm1, %zmm0, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vpermpd {{.*#+}} zmm1 = zmm1[0,3,2,3,4,7,6,7]
; AVX512DQ-BW-FCP-NEXT:    vmovaps 16(%rdi), %xmm2
; AVX512DQ-BW-FCP-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
; AVX512DQ-BW-FCP-NEXT:    vmovaps %xmm1, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovaps %xmm0, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovaps %xmm2, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <6 x i64>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <6 x i64> %wide.vec, <6 x i64> poison, <2 x i32> <i32 0, i32 3>
  %strided.vec1 = shufflevector <6 x i64> %wide.vec, <6 x i64> poison, <2 x i32> <i32 1, i32 4>
  %strided.vec2 = shufflevector <6 x i64> %wide.vec, <6 x i64> poison, <2 x i32> <i32 2, i32 5>
  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
  ret void
}

define void @load_i64_stride3_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
; SSE-LABEL: load_i64_stride3_vf4:
; SSE:       # %bb.0:
; SSE-NEXT:    movapd 80(%rdi), %xmm0
; SSE-NEXT:    movapd (%rdi), %xmm1
; SSE-NEXT:    movapd 16(%rdi), %xmm2
; SSE-NEXT:    movapd 32(%rdi), %xmm3
; SSE-NEXT:    movapd 48(%rdi), %xmm4
; SSE-NEXT:    movapd 64(%rdi), %xmm5
; SSE-NEXT:    movapd %xmm5, %xmm6
; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
; SSE-NEXT:    movapd %xmm2, %xmm7
; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1]
; SSE-NEXT:    shufpd {{.*#+}} xmm4 = xmm4[1],xmm0[0]
; SSE-NEXT:    shufpd {{.*#+}} xmm1 = xmm1[1],xmm3[0]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
; SSE-NEXT:    movapd %xmm6, 16(%rsi)
; SSE-NEXT:    movapd %xmm7, (%rsi)
; SSE-NEXT:    movapd %xmm4, 16(%rdx)
; SSE-NEXT:    movapd %xmm1, (%rdx)
; SSE-NEXT:    movapd %xmm0, 16(%rcx)
; SSE-NEXT:    movapd %xmm3, (%rcx)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i64_stride3_vf4:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovapd 32(%rdi), %ymm0
; AVX-NEXT:    vmovaps 16(%rdi), %xmm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm2 = mem[0,1],ymm0[2,3]
; AVX-NEXT:    vinsertf128 $1, 64(%rdi), %ymm1, %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm3 = ymm2[0],ymm1[1],ymm2[2],ymm1[3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[1],ymm0[0],ymm2[3],ymm0[2]
; AVX-NEXT:    vbroadcastsd 80(%rdi), %ymm4
; AVX-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX-NEXT:    vmovapd %ymm3, (%rsi)
; AVX-NEXT:    vmovapd %ymm2, (%rdx)
; AVX-NEXT:    vmovapd %ymm0, (%rcx)
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i64_stride3_vf4:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm0
; AVX2-NEXT:    vmovaps (%rdi), %ymm1
; AVX2-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm2
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = ymm1[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 80(%rdi), %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovaps 16(%rdi), %xmm1
; AVX2-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-NEXT:    vmovaps %ymm2, (%rsi)
; AVX2-NEXT:    vmovaps %ymm0, (%rdx)
; AVX2-NEXT:    vmovaps %ymm1, (%rcx)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i64_stride3_vf4:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm1
; AVX2-FP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm2
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm1[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 80(%rdi), %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovaps 16(%rdi), %xmm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps %ymm2, (%rsi)
; AVX2-FP-NEXT:    vmovaps %ymm0, (%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm1, (%rcx)
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i64_stride3_vf4:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm1
; AVX2-FCP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm2
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm3 = ymm1[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 80(%rdi), %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovaps 16(%rdi), %xmm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm3 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps %ymm2, (%rsi)
; AVX2-FCP-NEXT:    vmovaps %ymm0, (%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm1, (%rcx)
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i64_stride3_vf4:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512-NEXT:    vpmovsxbq {{.*#+}} ymm2 = [0,3,6,9]
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2
; AVX512-NEXT:    vpmovsxbq {{.*#+}} ymm3 = [1,4,7,10]
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512-NEXT:    vpmovsxbq {{.*#+}} ymm4 = [2,5,8,11]
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512-NEXT:    vmovdqa %ymm3, (%rdx)
; AVX512-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i64_stride3_vf4:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm2 = [0,3,6,9]
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm3 = [1,4,7,10]
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm4 = [2,5,8,11]
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512-FCP-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa %ymm3, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i64_stride3_vf4:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} ymm2 = [0,3,6,9]
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} ymm3 = [1,4,7,10]
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} ymm4 = [2,5,8,11]
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512DQ-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512DQ-NEXT:    vmovdqa %ymm3, (%rdx)
; AVX512DQ-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride3_vf4:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm2 = [0,3,6,9]
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm3 = [1,4,7,10]
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm4 = [2,5,8,11]
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm3, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i64_stride3_vf4:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} ymm2 = [0,3,6,9]
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} ymm3 = [1,4,7,10]
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} ymm4 = [2,5,8,11]
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512BW-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512BW-NEXT:    vmovdqa %ymm3, (%rdx)
; AVX512BW-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i64_stride3_vf4:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm2 = [0,3,6,9]
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm3 = [1,4,7,10]
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm4 = [2,5,8,11]
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512BW-FCP-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa %ymm3, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i64_stride3_vf4:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} ymm2 = [0,3,6,9]
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} ymm3 = [1,4,7,10]
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} ymm4 = [2,5,8,11]
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512DQ-BW-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa %ymm3, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride3_vf4:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm2 = [0,3,6,9]
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm3 = [1,4,7,10]
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} ymm4 = [2,5,8,11]
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm2, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm3, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm4, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <12 x i64>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <12 x i64> %wide.vec, <12 x i64> poison, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
  %strided.vec1 = shufflevector <12 x i64> %wide.vec, <12 x i64> poison, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
  %strided.vec2 = shufflevector <12 x i64> %wide.vec, <12 x i64> poison, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
  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
  ret void
}

define void @load_i64_stride3_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
; SSE-LABEL: load_i64_stride3_vf8:
; SSE:       # %bb.0:
; SSE-NEXT:    movapd 128(%rdi), %xmm2
; SSE-NEXT:    movapd 176(%rdi), %xmm1
; SSE-NEXT:    movapd 80(%rdi), %xmm0
; SSE-NEXT:    movapd 96(%rdi), %xmm3
; SSE-NEXT:    movapd 112(%rdi), %xmm8
; SSE-NEXT:    movapd 144(%rdi), %xmm5
; SSE-NEXT:    movapd 160(%rdi), %xmm9
; SSE-NEXT:    movapd (%rdi), %xmm6
; SSE-NEXT:    movapd 16(%rdi), %xmm10
; SSE-NEXT:    movapd 32(%rdi), %xmm4
; SSE-NEXT:    movapd 48(%rdi), %xmm7
; SSE-NEXT:    movapd 64(%rdi), %xmm11
; SSE-NEXT:    movapd %xmm11, %xmm12
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm7[0],xmm12[1]
; SSE-NEXT:    movapd %xmm9, %xmm13
; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm5[0],xmm13[1]
; SSE-NEXT:    movapd %xmm8, %xmm14
; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm3[0],xmm14[1]
; SSE-NEXT:    movapd %xmm10, %xmm15
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm6[0],xmm15[1]
; SSE-NEXT:    shufpd {{.*#+}} xmm7 = xmm7[1],xmm0[0]
; SSE-NEXT:    shufpd {{.*#+}} xmm5 = xmm5[1],xmm1[0]
; SSE-NEXT:    shufpd {{.*#+}} xmm3 = xmm3[1],xmm2[0]
; SSE-NEXT:    shufpd {{.*#+}} xmm6 = xmm6[1],xmm4[0]
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm11[0],xmm0[1]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm9[0],xmm1[1]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm8[0],xmm2[1]
; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm10[0],xmm4[1]
; SSE-NEXT:    movapd %xmm14, 32(%rsi)
; SSE-NEXT:    movapd %xmm13, 48(%rsi)
; SSE-NEXT:    movapd %xmm15, (%rsi)
; SSE-NEXT:    movapd %xmm12, 16(%rsi)
; SSE-NEXT:    movapd %xmm3, 32(%rdx)
; SSE-NEXT:    movapd %xmm5, 48(%rdx)
; SSE-NEXT:    movapd %xmm6, (%rdx)
; SSE-NEXT:    movapd %xmm7, 16(%rdx)
; SSE-NEXT:    movapd %xmm2, 32(%rcx)
; SSE-NEXT:    movapd %xmm1, 48(%rcx)
; SSE-NEXT:    movapd %xmm4, (%rcx)
; SSE-NEXT:    movapd %xmm0, 16(%rcx)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i64_stride3_vf8:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovapd 128(%rdi), %ymm0
; AVX-NEXT:    vmovapd 32(%rdi), %ymm1
; AVX-NEXT:    vmovaps 16(%rdi), %xmm2
; AVX-NEXT:    vblendpd {{.*#+}} ymm3 = mem[0,1],ymm1[2,3]
; AVX-NEXT:    vinsertf128 $1, 64(%rdi), %ymm2, %ymm2
; AVX-NEXT:    vblendpd {{.*#+}} ymm4 = ymm3[0],ymm2[1],ymm3[2],ymm2[3]
; AVX-NEXT:    vmovaps 112(%rdi), %xmm5
; AVX-NEXT:    vblendpd {{.*#+}} ymm6 = mem[0,1],ymm0[2,3]
; AVX-NEXT:    vinsertf128 $1, 160(%rdi), %ymm5, %ymm5
; AVX-NEXT:    vblendpd {{.*#+}} ymm7 = ymm6[0],ymm5[1],ymm6[2],ymm5[3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm3 = ymm3[1],ymm1[0],ymm3[3],ymm1[2]
; AVX-NEXT:    vbroadcastsd 80(%rdi), %ymm8
; AVX-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm6 = ymm6[1],ymm0[0],ymm6[3],ymm0[2]
; AVX-NEXT:    vbroadcastsd 176(%rdi), %ymm8
; AVX-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm5[0],ymm0[1],ymm5[2],ymm0[3]
; AVX-NEXT:    vmovapd %ymm7, 32(%rsi)
; AVX-NEXT:    vmovapd %ymm4, (%rsi)
; AVX-NEXT:    vmovapd %ymm6, 32(%rdx)
; AVX-NEXT:    vmovapd %ymm3, (%rdx)
; AVX-NEXT:    vmovapd %ymm0, 32(%rcx)
; AVX-NEXT:    vmovapd %ymm1, (%rcx)
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i64_stride3_vf8:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovaps 128(%rdi), %ymm0
; AVX2-NEXT:    vmovaps 96(%rdi), %ymm1
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm2
; AVX2-NEXT:    vmovaps (%rdi), %ymm3
; AVX2-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm4
; AVX2-NEXT:    vpermpd {{.*#+}} ymm5 = ymm3[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
; AVX2-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm5
; AVX2-NEXT:    vpermpd {{.*#+}} ymm6 = ymm1[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 80(%rdi), %ymm3
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 176(%rdi), %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovaps 16(%rdi), %xmm1
; AVX2-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm3 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-NEXT:    vmovaps 112(%rdi), %xmm3
; AVX2-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm6 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm6[4,5,6,7]
; AVX2-NEXT:    vmovaps %ymm5, 32(%rsi)
; AVX2-NEXT:    vmovaps %ymm4, (%rsi)
; AVX2-NEXT:    vmovaps %ymm0, 32(%rdx)
; AVX2-NEXT:    vmovaps %ymm2, (%rdx)
; AVX2-NEXT:    vmovaps %ymm3, 32(%rcx)
; AVX2-NEXT:    vmovaps %ymm1, (%rcx)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i64_stride3_vf8:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovaps 96(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm3
; AVX2-FP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm4
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm5 = ymm3[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm5
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm6 = ymm1[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 80(%rdi), %ymm3
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 176(%rdi), %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovaps 16(%rdi), %xmm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm3 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 112(%rdi), %xmm3
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm6 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps %ymm5, 32(%rsi)
; AVX2-FP-NEXT:    vmovaps %ymm4, (%rsi)
; AVX2-FP-NEXT:    vmovaps %ymm0, 32(%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm2, (%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm3, 32(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm1, (%rcx)
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i64_stride3_vf8:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovaps 96(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm3
; AVX2-FCP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm4
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm5 = ymm3[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm5
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm6 = ymm1[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 80(%rdi), %ymm3
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 176(%rdi), %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovaps 16(%rdi), %xmm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm3 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 112(%rdi), %xmm3
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm6 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps %ymm5, 32(%rsi)
; AVX2-FCP-NEXT:    vmovaps %ymm4, (%rsi)
; AVX2-FCP-NEXT:    vmovaps %ymm0, 32(%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm2, (%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm3, 32(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm1, (%rcx)
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i64_stride3_vf8:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [0,3,6,9,12,15,0,0]
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm4 = [0,1,2,3,4,5,10,13]
; AVX512-NEXT:    vpermi2q %zmm2, %zmm3, %zmm4
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [1,4,7,10,13,0,0,0]
; AVX512-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,8,11,14]
; AVX512-NEXT:    vpermi2q %zmm2, %zmm3, %zmm5
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [10,13,0,3,6,0,0,0]
; AVX512-NEXT:    vpermi2q %zmm0, %zmm1, %zmm3
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm0 = [0,1,2,3,4,9,12,15]
; AVX512-NEXT:    vpermi2q %zmm2, %zmm3, %zmm0
; AVX512-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm0, (%rcx)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i64_stride3_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:    vpmovsxbq {{.*#+}} zmm3 = [0,3,6,9,12,15,0,0]
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm4 = [0,1,2,3,4,5,10,13]
; AVX512-FCP-NEXT:    vpermi2q %zmm2, %zmm3, %zmm4
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [1,4,7,10,13,0,0,0]
; AVX512-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,8,11,14]
; AVX512-FCP-NEXT:    vpermi2q %zmm2, %zmm3, %zmm5
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [10,13,0,3,6,0,0,0]
; AVX512-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm3
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm0 = [0,1,2,3,4,9,12,15]
; AVX512-FCP-NEXT:    vpermi2q %zmm2, %zmm3, %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, (%rcx)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i64_stride3_vf8:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [0,3,6,9,12,15,0,0]
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm4 = [0,1,2,3,4,5,10,13]
; AVX512DQ-NEXT:    vpermi2q %zmm2, %zmm3, %zmm4
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [1,4,7,10,13,0,0,0]
; AVX512DQ-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,8,11,14]
; AVX512DQ-NEXT:    vpermi2q %zmm2, %zmm3, %zmm5
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [10,13,0,3,6,0,0,0]
; AVX512DQ-NEXT:    vpermi2q %zmm0, %zmm1, %zmm3
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm0 = [0,1,2,3,4,9,12,15]
; AVX512DQ-NEXT:    vpermi2q %zmm2, %zmm3, %zmm0
; AVX512DQ-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, (%rcx)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride3_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:    vpmovsxbq {{.*#+}} zmm3 = [0,3,6,9,12,15,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm4 = [0,1,2,3,4,5,10,13]
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm2, %zmm3, %zmm4
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [1,4,7,10,13,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,8,11,14]
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm2, %zmm3, %zmm5
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [10,13,0,3,6,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm3
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm0 = [0,1,2,3,4,9,12,15]
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm2, %zmm3, %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, (%rcx)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i64_stride3_vf8:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm2
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [0,3,6,9,12,15,0,0]
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm4 = [0,1,2,3,4,5,10,13]
; AVX512BW-NEXT:    vpermi2q %zmm2, %zmm3, %zmm4
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [1,4,7,10,13,0,0,0]
; AVX512BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,8,11,14]
; AVX512BW-NEXT:    vpermi2q %zmm2, %zmm3, %zmm5
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [10,13,0,3,6,0,0,0]
; AVX512BW-NEXT:    vpermi2q %zmm0, %zmm1, %zmm3
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm0 = [0,1,2,3,4,9,12,15]
; AVX512BW-NEXT:    vpermi2q %zmm2, %zmm3, %zmm0
; AVX512BW-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm0, (%rcx)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i64_stride3_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:    vpmovsxbq {{.*#+}} zmm3 = [0,3,6,9,12,15,0,0]
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm4 = [0,1,2,3,4,5,10,13]
; AVX512BW-FCP-NEXT:    vpermi2q %zmm2, %zmm3, %zmm4
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [1,4,7,10,13,0,0,0]
; AVX512BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,8,11,14]
; AVX512BW-FCP-NEXT:    vpermi2q %zmm2, %zmm3, %zmm5
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [10,13,0,3,6,0,0,0]
; AVX512BW-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm3
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm0 = [0,1,2,3,4,9,12,15]
; AVX512BW-FCP-NEXT:    vpermi2q %zmm2, %zmm3, %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, (%rcx)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i64_stride3_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:    vpmovsxbq {{.*#+}} zmm3 = [0,3,6,9,12,15,0,0]
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm4 = [0,1,2,3,4,5,10,13]
; AVX512DQ-BW-NEXT:    vpermi2q %zmm2, %zmm3, %zmm4
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [1,4,7,10,13,0,0,0]
; AVX512DQ-BW-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,8,11,14]
; AVX512DQ-BW-NEXT:    vpermi2q %zmm2, %zmm3, %zmm5
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [10,13,0,3,6,0,0,0]
; AVX512DQ-BW-NEXT:    vpermi2q %zmm0, %zmm1, %zmm3
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm0 = [0,1,2,3,4,9,12,15]
; AVX512DQ-BW-NEXT:    vpermi2q %zmm2, %zmm3, %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, (%rcx)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride3_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:    vpmovsxbq {{.*#+}} zmm3 = [0,3,6,9,12,15,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm4 = [0,1,2,3,4,5,10,13]
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm2, %zmm3, %zmm4
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [1,4,7,10,13,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm1, %zmm0, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,8,11,14]
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm2, %zmm3, %zmm5
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm3 = [10,13,0,3,6,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm0, %zmm1, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm0 = [0,1,2,3,4,9,12,15]
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm2, %zmm3, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm4, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <24 x i64>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <24 x i64> %wide.vec, <24 x i64> poison, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
  %strided.vec1 = shufflevector <24 x i64> %wide.vec, <24 x i64> poison, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
  %strided.vec2 = shufflevector <24 x i64> %wide.vec, <24 x i64> poison, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
  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
  ret void
}

define void @load_i64_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
; SSE-LABEL: load_i64_stride3_vf16:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $24, %rsp
; SSE-NEXT:    movapd 128(%rdi), %xmm0
; SSE-NEXT:    movapd 176(%rdi), %xmm1
; SSE-NEXT:    movapd 224(%rdi), %xmm4
; SSE-NEXT:    movapd 272(%rdi), %xmm3
; SSE-NEXT:    movapd 80(%rdi), %xmm2
; SSE-NEXT:    movapd 96(%rdi), %xmm5
; SSE-NEXT:    movapd 112(%rdi), %xmm12
; SSE-NEXT:    movapd 144(%rdi), %xmm6
; SSE-NEXT:    movapd 160(%rdi), %xmm14
; SSE-NEXT:    movapd 192(%rdi), %xmm7
; SSE-NEXT:    movapd 208(%rdi), %xmm11
; SSE-NEXT:    movapd 240(%rdi), %xmm10
; SSE-NEXT:    movapd 256(%rdi), %xmm13
; SSE-NEXT:    movapd 48(%rdi), %xmm9
; SSE-NEXT:    movapd 64(%rdi), %xmm15
; SSE-NEXT:    movapd %xmm15, %xmm8
; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm9[0],xmm8[1]
; SSE-NEXT:    movapd %xmm8, (%rsp) # 16-byte Spill
; SSE-NEXT:    shufpd {{.*#+}} xmm9 = xmm9[1],xmm2[0]
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm15[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd %xmm14, %xmm15
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm6[0],xmm15[1]
; SSE-NEXT:    shufpd {{.*#+}} xmm6 = xmm6[1],xmm1[0]
; SSE-NEXT:    movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm14[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd %xmm12, %xmm14
; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm5[0],xmm14[1]
; SSE-NEXT:    shufpd {{.*#+}} xmm5 = xmm5[1],xmm0[0]
; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm12[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd %xmm13, %xmm12
; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm10[0],xmm12[1]
; SSE-NEXT:    shufpd {{.*#+}} xmm10 = xmm10[1],xmm3[0]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm13[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd %xmm11, %xmm8
; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm7[0],xmm8[1]
; SSE-NEXT:    shufpd {{.*#+}} xmm7 = xmm7[1],xmm4[0]
; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm11[0],xmm4[1]
; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 336(%rdi), %xmm13
; SSE-NEXT:    movapd 352(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm7
; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm13[0],xmm7[1]
; SSE-NEXT:    movapd 368(%rdi), %xmm11
; SSE-NEXT:    shufpd {{.*#+}} xmm13 = xmm13[1],xmm11[0]
; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm0[0],xmm11[1]
; SSE-NEXT:    movapd 288(%rdi), %xmm0
; SSE-NEXT:    movapd 304(%rdi), %xmm2
; SSE-NEXT:    movapd %xmm2, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd 320(%rdi), %xmm6
; SSE-NEXT:    shufpd {{.*#+}} xmm0 = xmm0[1],xmm6[0]
; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm2[0],xmm6[1]
; SSE-NEXT:    movapd (%rdi), %xmm2
; SSE-NEXT:    movapd 16(%rdi), %xmm5
; SSE-NEXT:    movapd %xmm5, %xmm3
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
; SSE-NEXT:    movapd 32(%rdi), %xmm4
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm4[0]
; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm5[0],xmm4[1]
; SSE-NEXT:    movapd %xmm1, 96(%rsi)
; SSE-NEXT:    movapd %xmm14, 32(%rsi)
; SSE-NEXT:    movapd %xmm7, 112(%rsi)
; SSE-NEXT:    movapd %xmm15, 48(%rsi)
; SSE-NEXT:    movapd %xmm8, 64(%rsi)
; SSE-NEXT:    movapd %xmm3, (%rsi)
; SSE-NEXT:    movapd %xmm12, 80(%rsi)
; SSE-NEXT:    movaps (%rsp), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 16(%rsi)
; SSE-NEXT:    movapd %xmm0, 96(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%rdx)
; SSE-NEXT:    movapd %xmm13, 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:    movapd %xmm2, (%rdx)
; SSE-NEXT:    movapd %xmm10, 80(%rdx)
; SSE-NEXT:    movapd %xmm9, 16(%rdx)
; SSE-NEXT:    movapd %xmm6, 96(%rcx)
; SSE-NEXT:    movapd %xmm11, 112(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 64(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 80(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 32(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 48(%rcx)
; SSE-NEXT:    movapd %xmm4, (%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rcx)
; SSE-NEXT:    addq $24, %rsp
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i64_stride3_vf16:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovapd 32(%rdi), %ymm1
; AVX-NEXT:    vmovapd 224(%rdi), %ymm0
; AVX-NEXT:    vmovapd 320(%rdi), %ymm2
; AVX-NEXT:    vmovapd 128(%rdi), %ymm5
; AVX-NEXT:    vblendpd {{.*#+}} ymm7 = mem[0,1],ymm5[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm8 = mem[0,1],ymm2[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm3 = mem[0,1],ymm0[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm9 = mem[0,1],ymm1[2,3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm4 = ymm7[1],ymm5[0],ymm7[3],ymm5[2]
; AVX-NEXT:    vbroadcastsd 176(%rdi), %ymm6
; AVX-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm6 = ymm8[1],ymm2[0],ymm8[3],ymm2[2]
; AVX-NEXT:    vbroadcastsd 368(%rdi), %ymm10
; AVX-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1,2],ymm10[3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm10 = ymm9[1],ymm1[0],ymm9[3],ymm1[2]
; AVX-NEXT:    vbroadcastsd 80(%rdi), %ymm11
; AVX-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm11 = ymm3[1],ymm0[0],ymm3[3],ymm0[2]
; AVX-NEXT:    vbroadcastsd 272(%rdi), %ymm12
; AVX-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3]
; AVX-NEXT:    vmovaps 112(%rdi), %xmm12
; AVX-NEXT:    vinsertf128 $1, 160(%rdi), %ymm12, %ymm12
; AVX-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0],ymm12[1],ymm7[2],ymm12[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm5 = ymm12[0],ymm5[1],ymm12[2],ymm5[3]
; AVX-NEXT:    vmovaps 304(%rdi), %xmm12
; AVX-NEXT:    vinsertf128 $1, 352(%rdi), %ymm12, %ymm12
; AVX-NEXT:    vblendpd {{.*#+}} ymm8 = ymm8[0],ymm12[1],ymm8[2],ymm12[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm2 = ymm12[0],ymm2[1],ymm12[2],ymm2[3]
; AVX-NEXT:    vmovaps 16(%rdi), %xmm12
; AVX-NEXT:    vinsertf128 $1, 64(%rdi), %ymm12, %ymm12
; AVX-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0],ymm12[1],ymm9[2],ymm12[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm1 = ymm12[0],ymm1[1],ymm12[2],ymm1[3]
; AVX-NEXT:    vmovaps 208(%rdi), %xmm12
; AVX-NEXT:    vinsertf128 $1, 256(%rdi), %ymm12, %ymm12
; AVX-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0],ymm12[1],ymm3[2],ymm12[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm12[0],ymm0[1],ymm12[2],ymm0[3]
; AVX-NEXT:    vmovapd %ymm9, (%rsi)
; AVX-NEXT:    vmovapd %ymm3, 64(%rsi)
; AVX-NEXT:    vmovapd %ymm8, 96(%rsi)
; AVX-NEXT:    vmovapd %ymm7, 32(%rsi)
; AVX-NEXT:    vmovapd %ymm11, 64(%rdx)
; AVX-NEXT:    vmovapd %ymm10, (%rdx)
; AVX-NEXT:    vmovapd %ymm6, 96(%rdx)
; AVX-NEXT:    vmovapd %ymm4, 32(%rdx)
; AVX-NEXT:    vmovapd %ymm0, 64(%rcx)
; AVX-NEXT:    vmovapd %ymm1, (%rcx)
; AVX-NEXT:    vmovapd %ymm2, 96(%rcx)
; AVX-NEXT:    vmovapd %ymm5, 32(%rcx)
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i64_stride3_vf16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm8
; AVX2-NEXT:    vmovaps (%rdi), %ymm9
; AVX2-NEXT:    vmovaps 224(%rdi), %ymm6
; AVX2-NEXT:    vmovaps 192(%rdi), %ymm7
; AVX2-NEXT:    vmovaps 320(%rdi), %ymm5
; AVX2-NEXT:    vmovaps 288(%rdi), %ymm10
; AVX2-NEXT:    vmovaps 128(%rdi), %ymm3
; AVX2-NEXT:    vmovaps 96(%rdi), %ymm11
; AVX2-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm11[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm1
; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm10[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vinsertf128 $1, 256(%rdi), %ymm0, %ymm2
; AVX2-NEXT:    vpermpd {{.*#+}} ymm4 = ymm7[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
; AVX2-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm4
; AVX2-NEXT:    vpermpd {{.*#+}} ymm12 = ymm9[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm8[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3,4,5],ymm4[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm11[2,3],ymm3[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 176(%rdi), %ymm11
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 368(%rdi), %ymm10
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm10[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 80(%rdi), %ymm9
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3],ymm6[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 272(%rdi), %ymm7
; AVX2-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
; AVX2-NEXT:    vmovaps 112(%rdi), %xmm7
; AVX2-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm9 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm9[4,5,6,7]
; AVX2-NEXT:    vmovaps 304(%rdi), %xmm9
; AVX2-NEXT:    vblendps {{.*#+}} xmm9 = xmm9[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm10 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
; AVX2-NEXT:    vmovaps 16(%rdi), %xmm10
; AVX2-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm11 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
; AVX2-NEXT:    vmovaps 208(%rdi), %xmm11
; AVX2-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm12 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
; AVX2-NEXT:    vmovaps %ymm4, (%rsi)
; AVX2-NEXT:    vmovaps %ymm2, 64(%rsi)
; AVX2-NEXT:    vmovaps %ymm1, 96(%rsi)
; AVX2-NEXT:    vmovaps %ymm0, 32(%rsi)
; AVX2-NEXT:    vmovaps %ymm6, 64(%rdx)
; AVX2-NEXT:    vmovaps %ymm8, (%rdx)
; AVX2-NEXT:    vmovaps %ymm5, 96(%rdx)
; AVX2-NEXT:    vmovaps %ymm3, 32(%rdx)
; AVX2-NEXT:    vmovaps %ymm11, 64(%rcx)
; AVX2-NEXT:    vmovaps %ymm10, (%rcx)
; AVX2-NEXT:    vmovaps %ymm9, 96(%rcx)
; AVX2-NEXT:    vmovaps %ymm7, 32(%rcx)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i64_stride3_vf16:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm8
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm9
; AVX2-FP-NEXT:    vmovaps 224(%rdi), %ymm6
; AVX2-FP-NEXT:    vmovaps 192(%rdi), %ymm7
; AVX2-FP-NEXT:    vmovaps 320(%rdi), %ymm5
; AVX2-FP-NEXT:    vmovaps 288(%rdi), %ymm10
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovaps 96(%rdi), %ymm11
; AVX2-FP-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm11[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm1
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm10[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 256(%rdi), %ymm0, %ymm2
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm4 = ymm7[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm4
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm12 = ymm9[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm11[2,3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 176(%rdi), %ymm11
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 368(%rdi), %ymm10
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 80(%rdi), %ymm9
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 272(%rdi), %ymm7
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FP-NEXT:    vmovaps 112(%rdi), %xmm7
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm9 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 304(%rdi), %xmm9
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm9 = xmm9[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm10 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 16(%rdi), %xmm10
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm11 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 208(%rdi), %xmm11
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm12 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps %ymm4, (%rsi)
; AVX2-FP-NEXT:    vmovaps %ymm2, 64(%rsi)
; AVX2-FP-NEXT:    vmovaps %ymm1, 96(%rsi)
; AVX2-FP-NEXT:    vmovaps %ymm0, 32(%rsi)
; AVX2-FP-NEXT:    vmovaps %ymm6, 64(%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm8, (%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm5, 96(%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm3, 32(%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm11, 64(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm10, (%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm9, 96(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm7, 32(%rcx)
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i64_stride3_vf16:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm8
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm9
; AVX2-FCP-NEXT:    vmovaps 224(%rdi), %ymm6
; AVX2-FCP-NEXT:    vmovaps 192(%rdi), %ymm7
; AVX2-FCP-NEXT:    vmovaps 320(%rdi), %ymm5
; AVX2-FCP-NEXT:    vmovaps 288(%rdi), %ymm10
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovaps 96(%rdi), %ymm11
; AVX2-FCP-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm11[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm1
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm2 = ymm10[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 256(%rdi), %ymm0, %ymm2
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm4 = ymm7[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm4
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm12 = ymm9[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm11[2,3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 176(%rdi), %ymm11
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm10[2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm5 = ymm5[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 368(%rdi), %ymm10
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 80(%rdi), %ymm9
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3],ymm6[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 272(%rdi), %ymm7
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FCP-NEXT:    vmovaps 112(%rdi), %xmm7
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm9 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 304(%rdi), %xmm9
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm9 = xmm9[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm10 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 16(%rdi), %xmm10
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm11 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 208(%rdi), %xmm11
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm11 = xmm11[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm12 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps %ymm4, (%rsi)
; AVX2-FCP-NEXT:    vmovaps %ymm2, 64(%rsi)
; AVX2-FCP-NEXT:    vmovaps %ymm1, 96(%rsi)
; AVX2-FCP-NEXT:    vmovaps %ymm0, 32(%rsi)
; AVX2-FCP-NEXT:    vmovaps %ymm6, 64(%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm8, (%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm5, 96(%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm3, 32(%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm11, 64(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm10, (%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm9, 96(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm7, 32(%rcx)
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i64_stride3_vf16:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512-NEXT:    vmovdqa64 128(%rdi), %zmm4
; AVX512-NEXT:    vmovdqa64 192(%rdi), %zmm5
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm6 = [0,3,6,9,12,15,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512-NEXT:    vpermt2q %zmm1, %zmm6, %zmm7
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [0,1,2,3,4,5,10,13]
; AVX512-NEXT:    vpermt2q %zmm0, %zmm8, %zmm7
; AVX512-NEXT:    vpermi2q %zmm3, %zmm2, %zmm6
; AVX512-NEXT:    vpermt2q %zmm4, %zmm8, %zmm6
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [1,4,7,10,13,0,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm9
; AVX512-NEXT:    vpermt2q %zmm1, %zmm8, %zmm9
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [0,1,2,3,4,8,11,14]
; AVX512-NEXT:    vpermt2q %zmm0, %zmm10, %zmm9
; AVX512-NEXT:    vpermi2q %zmm3, %zmm2, %zmm8
; AVX512-NEXT:    vpermt2q %zmm4, %zmm10, %zmm8
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [10,13,0,3,6,0,0,0]
; AVX512-NEXT:    vpermt2q %zmm5, %zmm10, %zmm1
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,9,12,15]
; AVX512-NEXT:    vpermt2q %zmm0, %zmm5, %zmm1
; AVX512-NEXT:    vpermt2q %zmm2, %zmm10, %zmm3
; AVX512-NEXT:    vpermt2q %zmm4, %zmm5, %zmm3
; AVX512-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm6, (%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm9, 64(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm8, (%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i64_stride3_vf16:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm4
; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm5
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm6 = [0,3,6,9,12,15,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm1, %zmm6, %zmm7
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [0,1,2,3,4,5,10,13]
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm8, %zmm7
; AVX512-FCP-NEXT:    vpermi2q %zmm3, %zmm2, %zmm6
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm8, %zmm6
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [1,4,7,10,13,0,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, %zmm9
; AVX512-FCP-NEXT:    vpermt2q %zmm1, %zmm8, %zmm9
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [0,1,2,3,4,8,11,14]
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm10, %zmm9
; AVX512-FCP-NEXT:    vpermi2q %zmm3, %zmm2, %zmm8
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm10, %zmm8
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [10,13,0,3,6,0,0,0]
; AVX512-FCP-NEXT:    vpermt2q %zmm5, %zmm10, %zmm1
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,9,12,15]
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm5, %zmm1
; AVX512-FCP-NEXT:    vpermt2q %zmm2, %zmm10, %zmm3
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm5, %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm6, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm9, 64(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm8, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i64_stride3_vf16:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %zmm4
; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %zmm5
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm6 = [0,3,6,9,12,15,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm1, %zmm6, %zmm7
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [0,1,2,3,4,5,10,13]
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm8, %zmm7
; AVX512DQ-NEXT:    vpermi2q %zmm3, %zmm2, %zmm6
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm8, %zmm6
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [1,4,7,10,13,0,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, %zmm9
; AVX512DQ-NEXT:    vpermt2q %zmm1, %zmm8, %zmm9
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [0,1,2,3,4,8,11,14]
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm10, %zmm9
; AVX512DQ-NEXT:    vpermi2q %zmm3, %zmm2, %zmm8
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm10, %zmm8
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [10,13,0,3,6,0,0,0]
; AVX512DQ-NEXT:    vpermt2q %zmm5, %zmm10, %zmm1
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,9,12,15]
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm5, %zmm1
; AVX512DQ-NEXT:    vpermt2q %zmm2, %zmm10, %zmm3
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm5, %zmm3
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm6, (%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm9, 64(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm8, (%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride3_vf16:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm5
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm6 = [0,3,6,9,12,15,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm1, %zmm6, %zmm7
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [0,1,2,3,4,5,10,13]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm8, %zmm7
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm3, %zmm2, %zmm6
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm8, %zmm6
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [1,4,7,10,13,0,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, %zmm9
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm1, %zmm8, %zmm9
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [0,1,2,3,4,8,11,14]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm10, %zmm9
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm3, %zmm2, %zmm8
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm10, %zmm8
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [10,13,0,3,6,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm5, %zmm10, %zmm1
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,9,12,15]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm5, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm2, %zmm10, %zmm3
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm5, %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm6, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm9, 64(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm8, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i64_stride3_vf16:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm4
; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm5
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm6 = [0,3,6,9,12,15,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm6, %zmm7
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [0,1,2,3,4,5,10,13]
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm8, %zmm7
; AVX512BW-NEXT:    vpermi2q %zmm3, %zmm2, %zmm6
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm8, %zmm6
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [1,4,7,10,13,0,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm9
; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm8, %zmm9
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [0,1,2,3,4,8,11,14]
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm10, %zmm9
; AVX512BW-NEXT:    vpermi2q %zmm3, %zmm2, %zmm8
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm10, %zmm8
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [10,13,0,3,6,0,0,0]
; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm10, %zmm1
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,9,12,15]
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm5, %zmm1
; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm10, %zmm3
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm5, %zmm3
; AVX512BW-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm6, (%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm9, 64(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm8, (%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i64_stride3_vf16:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm4
; AVX512BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm5
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm6 = [0,3,6,9,12,15,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm1, %zmm6, %zmm7
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [0,1,2,3,4,5,10,13]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm8, %zmm7
; AVX512BW-FCP-NEXT:    vpermi2q %zmm3, %zmm2, %zmm6
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm8, %zmm6
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [1,4,7,10,13,0,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm9
; AVX512BW-FCP-NEXT:    vpermt2q %zmm1, %zmm8, %zmm9
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [0,1,2,3,4,8,11,14]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm10, %zmm9
; AVX512BW-FCP-NEXT:    vpermi2q %zmm3, %zmm2, %zmm8
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm10, %zmm8
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [10,13,0,3,6,0,0,0]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm5, %zmm10, %zmm1
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,9,12,15]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm5, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2q %zmm2, %zmm10, %zmm3
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm5, %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm6, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm9, 64(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm8, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i64_stride3_vf16:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 128(%rdi), %zmm4
; AVX512DQ-BW-NEXT:    vmovdqa64 192(%rdi), %zmm5
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm6 = [0,3,6,9,12,15,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm1, %zmm6, %zmm7
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [0,1,2,3,4,5,10,13]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm8, %zmm7
; AVX512DQ-BW-NEXT:    vpermi2q %zmm3, %zmm2, %zmm6
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm8, %zmm6
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [1,4,7,10,13,0,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, %zmm9
; AVX512DQ-BW-NEXT:    vpermt2q %zmm1, %zmm8, %zmm9
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [0,1,2,3,4,8,11,14]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm10, %zmm9
; AVX512DQ-BW-NEXT:    vpermi2q %zmm3, %zmm2, %zmm8
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm10, %zmm8
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [10,13,0,3,6,0,0,0]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm5, %zmm10, %zmm1
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,9,12,15]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm5, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2q %zmm2, %zmm10, %zmm3
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm5, %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm6, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm9, 64(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm8, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride3_vf16:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm5
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm6 = [0,3,6,9,12,15,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm1, %zmm6, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [0,1,2,3,4,5,10,13]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm8, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm3, %zmm2, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm8, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm8 = [1,4,7,10,13,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm9
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm1, %zmm8, %zmm9
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [0,1,2,3,4,8,11,14]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm10, %zmm9
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm3, %zmm2, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm10, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm10 = [10,13,0,3,6,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm5, %zmm10, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm5 = [0,1,2,3,4,9,12,15]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm5, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm2, %zmm10, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm5, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, 64(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm6, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm9, 64(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm8, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <48 x i64>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <48 x i64> %wide.vec, <48 x i64> poison, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
  %strided.vec1 = shufflevector <48 x i64> %wide.vec, <48 x i64> poison, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
  %strided.vec2 = shufflevector <48 x i64> %wide.vec, <48 x i64> poison, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
  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
  ret void
}

define void @load_i64_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
; SSE-LABEL: load_i64_stride3_vf32:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $408, %rsp # imm = 0x198
; SSE-NEXT:    movapd 224(%rdi), %xmm6
; SSE-NEXT:    movapd 272(%rdi), %xmm9
; SSE-NEXT:    movapd 128(%rdi), %xmm5
; SSE-NEXT:    movapd 176(%rdi), %xmm8
; SSE-NEXT:    movapd 80(%rdi), %xmm7
; SSE-NEXT:    movapd 96(%rdi), %xmm10
; SSE-NEXT:    movapd 112(%rdi), %xmm0
; SSE-NEXT:    movapd 144(%rdi), %xmm11
; SSE-NEXT:    movapd 160(%rdi), %xmm1
; SSE-NEXT:    movapd 192(%rdi), %xmm12
; SSE-NEXT:    movapd 208(%rdi), %xmm2
; SSE-NEXT:    movapd 240(%rdi), %xmm13
; SSE-NEXT:    movapd 256(%rdi), %xmm3
; SSE-NEXT:    movapd 48(%rdi), %xmm14
; SSE-NEXT:    movapd 64(%rdi), %xmm4
; SSE-NEXT:    movapd %xmm4, %xmm15
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm14[0],xmm15[1]
; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufpd {{.*#+}} xmm14 = xmm14[1],xmm7[0]
; SSE-NEXT:    movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm4[0],xmm7[1]
; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd %xmm0, %xmm4
; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm10[0],xmm4[1]
; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufpd {{.*#+}} xmm10 = xmm10[1],xmm5[0]
; SSE-NEXT:    movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm0[0],xmm5[1]
; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd %xmm1, %xmm0
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm11[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufpd {{.*#+}} xmm11 = xmm11[1],xmm8[0]
; SSE-NEXT:    movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
; SSE-NEXT:    movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd %xmm2, %xmm0
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm12[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufpd {{.*#+}} xmm12 = xmm12[1],xmm6[0]
; SSE-NEXT:    movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm2[0],xmm6[1]
; SSE-NEXT:    movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd %xmm3, %xmm0
; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm13[0],xmm0[1]
; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufpd {{.*#+}} xmm13 = xmm13[1],xmm9[0]
; SSE-NEXT:    movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm3[0],xmm9[1]
; SSE-NEXT:    movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 288(%rdi), %xmm2
; SSE-NEXT:    movapd 304(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 320(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 336(%rdi), %xmm2
; SSE-NEXT:    movapd 352(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 368(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 384(%rdi), %xmm2
; SSE-NEXT:    movapd 400(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 416(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 432(%rdi), %xmm2
; SSE-NEXT:    movapd 448(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 464(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, (%rsp) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 480(%rdi), %xmm2
; SSE-NEXT:    movapd 496(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 512(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 528(%rdi), %xmm15
; SSE-NEXT:    movapd 544(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm15[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 560(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm15 = xmm15[1],xmm1[0]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 576(%rdi), %xmm12
; SSE-NEXT:    movapd 592(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm14
; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm12[0],xmm14[1]
; SSE-NEXT:    movapd 608(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm12 = xmm12[1],xmm1[0]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 624(%rdi), %xmm8
; SSE-NEXT:    movapd 640(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm11
; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm8[0],xmm11[1]
; SSE-NEXT:    movapd 656(%rdi), %xmm13
; SSE-NEXT:    shufpd {{.*#+}} xmm8 = xmm8[1],xmm13[0]
; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
; SSE-NEXT:    movapd 672(%rdi), %xmm6
; SSE-NEXT:    movapd 688(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm5
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm6[0],xmm5[1]
; SSE-NEXT:    movapd 704(%rdi), %xmm10
; SSE-NEXT:    shufpd {{.*#+}} xmm6 = xmm6[1],xmm10[0]
; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
; SSE-NEXT:    movapd 720(%rdi), %xmm4
; SSE-NEXT:    movapd 736(%rdi), %xmm2
; SSE-NEXT:    movapd %xmm2, %xmm3
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm4[0],xmm3[1]
; SSE-NEXT:    movapd 752(%rdi), %xmm7
; SSE-NEXT:    shufpd {{.*#+}} xmm4 = xmm4[1],xmm7[0]
; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm2[0],xmm7[1]
; SSE-NEXT:    movapd (%rdi), %xmm2
; SSE-NEXT:    movapd 16(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd 32(%rdi), %xmm9
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm9[0]
; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
; SSE-NEXT:    movapd %xmm5, 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:    movapd %xmm3, 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:    movapd %xmm14, 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:    movapd %xmm1, (%rsi)
; SSE-NEXT:    movapd %xmm11, 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:    movapd %xmm6, 224(%rdx)
; SSE-NEXT:    movapd %xmm4, 240(%rdx)
; SSE-NEXT:    movapd %xmm12, 192(%rdx)
; SSE-NEXT:    movapd %xmm8, 208(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 160(%rdx)
; SSE-NEXT:    movapd %xmm15, 176(%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 128(%rdx)
; SSE-NEXT:    movaps (%rsp), %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:    movapd %xmm2, (%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 16(%rdx)
; SSE-NEXT:    movapd %xmm7, 240(%rcx)
; SSE-NEXT:    movapd %xmm10, 224(%rcx)
; SSE-NEXT:    movapd %xmm13, 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:    movapd %xmm9, (%rcx)
; SSE-NEXT:    addq $408, %rsp # imm = 0x198
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i64_stride3_vf32:
; AVX:       # %bb.0:
; AVX-NEXT:    subq $232, %rsp
; AVX-NEXT:    vmovapd 32(%rdi), %ymm7
; AVX-NEXT:    vmovapd 704(%rdi), %ymm13
; AVX-NEXT:    vmovapd 512(%rdi), %ymm9
; AVX-NEXT:    vmovapd 320(%rdi), %ymm8
; AVX-NEXT:    vmovapd 128(%rdi), %ymm10
; AVX-NEXT:    vblendpd {{.*#+}} ymm12 = mem[0,1],ymm10[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm11 = mem[0,1],ymm8[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm3 = mem[0,1],ymm9[2,3]
; AVX-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendpd {{.*#+}} ymm2 = mem[0,1],ymm13[2,3]
; AVX-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendpd {{.*#+}} ymm15 = mem[0,1],ymm7[2,3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm12[1],ymm10[0],ymm12[3],ymm10[2]
; AVX-NEXT:    vbroadcastsd 176(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm11[1],ymm8[0],ymm11[3],ymm8[2]
; AVX-NEXT:    vbroadcastsd 368(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm3[1],ymm9[0],ymm3[3],ymm9[2]
; AVX-NEXT:    vbroadcastsd 560(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm2[1],ymm13[0],ymm2[3],ymm13[2]
; AVX-NEXT:    vbroadcastsd 752(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm15[1],ymm7[0],ymm15[3],ymm7[2]
; AVX-NEXT:    vbroadcastsd 80(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 224(%rdi), %ymm3
; AVX-NEXT:    vblendpd {{.*#+}} ymm4 = mem[0,1],ymm3[2,3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm1 = ymm4[1],ymm3[0],ymm4[3],ymm3[2]
; AVX-NEXT:    vbroadcastsd 272(%rdi), %ymm2
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm2[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 416(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm6 = mem[0,1],ymm1[2,3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm2 = ymm6[1],ymm1[0],ymm6[3],ymm1[2]
; AVX-NEXT:    vbroadcastsd 464(%rdi), %ymm5
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0,1,2],ymm5[3]
; AVX-NEXT:    vmovupd %ymm0, (%rsp) # 32-byte Spill
; AVX-NEXT:    vmovapd 608(%rdi), %ymm2
; AVX-NEXT:    vblendpd {{.*#+}} ymm5 = mem[0,1],ymm2[2,3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm5[1],ymm2[0],ymm5[3],ymm2[2]
; AVX-NEXT:    vbroadcastsd 656(%rdi), %ymm14
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm14[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 112(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX-NEXT:    vblendpd {{.*#+}} ymm12 = ymm12[0],ymm0[1],ymm12[2],ymm0[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm14 = ymm0[0],ymm10[1],ymm0[2],ymm10[3]
; AVX-NEXT:    vmovaps 16(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm0
; AVX-NEXT:    vblendpd {{.*#+}} ymm15 = ymm15[0],ymm0[1],ymm15[2],ymm0[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2],ymm7[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 304(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
; AVX-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0],ymm0[1],ymm11[2],ymm0[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm7 = ymm8[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm10 = ymm0[0],ymm7[1],ymm0[2],ymm7[3]
; AVX-NEXT:    vmovaps 208(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 256(%rdi), %ymm0, %ymm0
; AVX-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0],ymm0[1],ymm4[2],ymm0[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2],ymm3[3]
; AVX-NEXT:    vmovaps 496(%rdi), %xmm3
; AVX-NEXT:    vinsertf128 $1, 544(%rdi), %ymm3, %ymm3
; AVX-NEXT:    vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm7 # 32-byte Folded Reload
; AVX-NEXT:    # ymm7 = mem[0],ymm3[1],mem[2],ymm3[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm8 = ymm9[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0],ymm8[1],ymm3[2],ymm8[3]
; AVX-NEXT:    vmovaps 400(%rdi), %xmm8
; AVX-NEXT:    vinsertf128 $1, 448(%rdi), %ymm8, %ymm8
; AVX-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0],ymm8[1],ymm6[2],ymm8[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm1 = ymm8[0],ymm1[1],ymm8[2],ymm1[3]
; AVX-NEXT:    vmovaps 688(%rdi), %xmm8
; AVX-NEXT:    vinsertf128 $1, 736(%rdi), %ymm8, %ymm8
; AVX-NEXT:    vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm9 # 32-byte Folded Reload
; AVX-NEXT:    # ymm9 = mem[0],ymm8[1],mem[2],ymm8[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm13 = ymm13[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm8 = ymm8[0],ymm13[1],ymm8[2],ymm13[3]
; AVX-NEXT:    vmovaps 592(%rdi), %xmm13
; AVX-NEXT:    vinsertf128 $1, 640(%rdi), %ymm13, %ymm13
; AVX-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0],ymm13[1],ymm5[2],ymm13[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm2 = ymm13[0],ymm2[1],ymm13[2],ymm2[3]
; AVX-NEXT:    vmovapd %ymm5, 192(%rsi)
; AVX-NEXT:    vmovapd %ymm6, 128(%rsi)
; AVX-NEXT:    vmovapd %ymm4, 64(%rsi)
; AVX-NEXT:    vmovapd %ymm15, (%rsi)
; AVX-NEXT:    vmovapd %ymm9, 224(%rsi)
; AVX-NEXT:    vmovapd %ymm7, 160(%rsi)
; AVX-NEXT:    vmovapd %ymm11, 96(%rsi)
; AVX-NEXT:    vmovapd %ymm12, 32(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm4, 192(%rdx)
; AVX-NEXT:    vmovups (%rsp), %ymm4 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm4, 128(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm4, 64(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm4, (%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm4, 224(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm4, 160(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm4, 96(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm4, 32(%rdx)
; AVX-NEXT:    vmovapd %ymm2, 192(%rcx)
; AVX-NEXT:    vmovapd %ymm8, 224(%rcx)
; AVX-NEXT:    vmovapd %ymm1, 128(%rcx)
; AVX-NEXT:    vmovapd %ymm3, 160(%rcx)
; AVX-NEXT:    vmovapd %ymm0, 64(%rcx)
; AVX-NEXT:    vmovapd %ymm10, 96(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, (%rcx)
; AVX-NEXT:    vmovapd %ymm14, 32(%rcx)
; AVX-NEXT:    addq $232, %rsp
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i64_stride3_vf32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    subq $232, %rsp
; 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 704(%rdi), %ymm11
; AVX2-NEXT:    vmovaps 672(%rdi), %ymm14
; AVX2-NEXT:    vmovaps 512(%rdi), %ymm10
; AVX2-NEXT:    vmovaps 480(%rdi), %ymm15
; AVX2-NEXT:    vmovaps 320(%rdi), %ymm9
; AVX2-NEXT:    vmovaps 288(%rdi), %ymm8
; AVX2-NEXT:    vmovaps 128(%rdi), %ymm7
; AVX2-NEXT:    vmovaps 96(%rdi), %ymm6
; AVX2-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm6[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm8[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm9[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vinsertf128 $1, 544(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm15[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vinsertf128 $1, 736(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm14[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm11[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm3[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 224(%rdi), %ymm5
; AVX2-NEXT:    vmovaps 192(%rdi), %ymm4
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm4[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 256(%rdi), %ymm0, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 416(%rdi), %ymm3
; AVX2-NEXT:    vmovaps 384(%rdi), %ymm2
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm2[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 448(%rdi), %ymm0, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 608(%rdi), %ymm1
; AVX2-NEXT:    vmovaps 576(%rdi), %ymm0
; AVX2-NEXT:    vpermpd {{.*#+}} ymm12 = ymm0[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 640(%rdi), %ymm0, %ymm13
; AVX2-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm13[6,7]
; AVX2-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 176(%rdi), %ymm7
; AVX2-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
; AVX2-NEXT:    vmovups %ymm6, (%rsp) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm8[2,3],ymm9[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 368(%rdi), %ymm8
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm7[0,1,2,3,4,5],ymm8[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1],ymm15[2,3],ymm10[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 560(%rdi), %ymm9
; AVX2-NEXT:    vblendps {{.*#+}} ymm9 = ymm7[0,1,2,3,4,5],ymm9[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm14[2,3],ymm11[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 752(%rdi), %ymm10
; AVX2-NEXT:    vblendps {{.*#+}} ymm11 = ymm7[0,1,2,3,4,5],ymm10[6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm7 = ymm6[0,1],mem[2,3],ymm6[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 80(%rdi), %ymm10
; AVX2-NEXT:    vblendps {{.*#+}} ymm12 = ymm7[0,1,2,3,4,5],ymm10[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 272(%rdi), %ymm5
; AVX2-NEXT:    vblendps {{.*#+}} ymm13 = ymm4[0,1,2,3,4,5],ymm5[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 464(%rdi), %ymm3
; AVX2-NEXT:    vblendps {{.*#+}} ymm15 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 656(%rdi), %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovaps 112(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 16(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 304(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 208(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm4 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
; AVX2-NEXT:    vmovaps 496(%rdi), %xmm4
; AVX2-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm5 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-NEXT:    vmovaps 400(%rdi), %xmm5
; AVX2-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm7 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm7[4,5,6,7]
; AVX2-NEXT:    vmovaps 688(%rdi), %xmm7
; AVX2-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm10 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm10[4,5,6,7]
; AVX2-NEXT:    vmovaps 592(%rdi), %xmm10
; AVX2-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm6 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm6[4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm10, 192(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm10, 128(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm10, 64(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm10, (%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm10, 224(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm10, 160(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm10, 96(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm10, 32(%rsi)
; AVX2-NEXT:    vmovaps %ymm2, 192(%rdx)
; AVX2-NEXT:    vmovaps %ymm15, 128(%rdx)
; AVX2-NEXT:    vmovaps %ymm13, 64(%rdx)
; AVX2-NEXT:    vmovaps %ymm12, (%rdx)
; AVX2-NEXT:    vmovaps %ymm11, 224(%rdx)
; AVX2-NEXT:    vmovaps %ymm9, 160(%rdx)
; AVX2-NEXT:    vmovaps %ymm8, 96(%rdx)
; AVX2-NEXT:    vmovups (%rsp), %ymm2 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm2, 32(%rdx)
; AVX2-NEXT:    vmovaps %ymm6, 192(%rcx)
; AVX2-NEXT:    vmovaps %ymm7, 224(%rcx)
; AVX2-NEXT:    vmovaps %ymm5, 128(%rcx)
; AVX2-NEXT:    vmovaps %ymm4, 160(%rcx)
; AVX2-NEXT:    vmovaps %ymm0, 64(%rcx)
; AVX2-NEXT:    vmovaps %ymm1, 96(%rcx)
; AVX2-NEXT:    vmovaps %ymm3, (%rcx)
; AVX2-NEXT:    vmovaps %ymm14, 32(%rcx)
; AVX2-NEXT:    addq $232, %rsp
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i64_stride3_vf32:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    subq $232, %rsp
; 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 704(%rdi), %ymm11
; AVX2-FP-NEXT:    vmovaps 672(%rdi), %ymm14
; AVX2-FP-NEXT:    vmovaps 512(%rdi), %ymm10
; AVX2-FP-NEXT:    vmovaps 480(%rdi), %ymm15
; AVX2-FP-NEXT:    vmovaps 320(%rdi), %ymm9
; AVX2-FP-NEXT:    vmovaps 288(%rdi), %ymm8
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %ymm7
; AVX2-FP-NEXT:    vmovaps 96(%rdi), %ymm6
; AVX2-FP-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm6[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm8[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vinsertf128 $1, 544(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm15[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vinsertf128 $1, 736(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm14[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm11[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm3[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 224(%rdi), %ymm5
; AVX2-FP-NEXT:    vmovaps 192(%rdi), %ymm4
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm4[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 256(%rdi), %ymm0, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 416(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovaps 384(%rdi), %ymm2
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm2[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 448(%rdi), %ymm0, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 608(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovaps 576(%rdi), %ymm0
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm12 = ymm0[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 640(%rdi), %ymm0, %ymm13
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm13[6,7]
; AVX2-FP-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 176(%rdi), %ymm7
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FP-NEXT:    vmovups %ymm6, (%rsp) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm8[2,3],ymm9[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 368(%rdi), %ymm8
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm7[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1],ymm15[2,3],ymm10[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 560(%rdi), %ymm9
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm9 = ymm7[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm14[2,3],ymm11[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 752(%rdi), %ymm10
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm11 = ymm7[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm7 = ymm6[0,1],mem[2,3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 80(%rdi), %ymm10
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm12 = ymm7[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 272(%rdi), %ymm5
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm13 = ymm4[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 464(%rdi), %ymm3
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm15 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 656(%rdi), %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovaps 112(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 16(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 304(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 208(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm4 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 496(%rdi), %xmm4
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm5 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 400(%rdi), %xmm5
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm7 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 688(%rdi), %xmm7
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm10 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 592(%rdi), %xmm10
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm6 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm10, 192(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm10, 128(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm10, 64(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm10, (%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm10, 224(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm10, 160(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm10, 96(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm10, 32(%rsi)
; AVX2-FP-NEXT:    vmovaps %ymm2, 192(%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm15, 128(%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm13, 64(%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm12, (%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm11, 224(%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm9, 160(%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm8, 96(%rdx)
; AVX2-FP-NEXT:    vmovups (%rsp), %ymm2 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm2, 32(%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm6, 192(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm7, 224(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm5, 128(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm4, 160(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm0, 64(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm1, 96(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm3, (%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm14, 32(%rcx)
; AVX2-FP-NEXT:    addq $232, %rsp
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i64_stride3_vf32:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    subq $232, %rsp
; 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 704(%rdi), %ymm11
; AVX2-FCP-NEXT:    vmovaps 672(%rdi), %ymm14
; AVX2-FCP-NEXT:    vmovaps 512(%rdi), %ymm10
; AVX2-FCP-NEXT:    vmovaps 480(%rdi), %ymm15
; AVX2-FCP-NEXT:    vmovaps 320(%rdi), %ymm9
; AVX2-FCP-NEXT:    vmovaps 288(%rdi), %ymm8
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %ymm7
; AVX2-FCP-NEXT:    vmovaps 96(%rdi), %ymm6
; AVX2-FCP-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm6[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm8[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vinsertf128 $1, 544(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm15[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vinsertf128 $1, 736(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm14[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm11[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm3[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 224(%rdi), %ymm5
; AVX2-FCP-NEXT:    vmovaps 192(%rdi), %ymm4
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm4[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 256(%rdi), %ymm0, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 416(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovaps 384(%rdi), %ymm2
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm2[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 448(%rdi), %ymm0, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 608(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovaps 576(%rdi), %ymm0
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm12 = ymm0[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 640(%rdi), %ymm0, %ymm13
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm13[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 176(%rdi), %ymm7
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm6, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm8[2,3],ymm9[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 368(%rdi), %ymm8
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm7[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm10[0,1],ymm15[2,3],ymm10[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 560(%rdi), %ymm9
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm9 = ymm7[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm14[2,3],ymm11[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 752(%rdi), %ymm10
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm11 = ymm7[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm7 = ymm6[0,1],mem[2,3],ymm6[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm7 = ymm7[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 80(%rdi), %ymm10
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm12 = ymm7[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 272(%rdi), %ymm5
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm13 = ymm4[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 464(%rdi), %ymm3
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm15 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 656(%rdi), %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovaps 112(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 16(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 304(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 208(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm4 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 496(%rdi), %xmm4
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm5 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 400(%rdi), %xmm5
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm7 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 688(%rdi), %xmm7
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm7 = xmm7[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm10 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 592(%rdi), %xmm10
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm6 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm10, 192(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm10, 128(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm10, 64(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm10, (%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm10, 224(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm10, 160(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm10, 96(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm10, 32(%rsi)
; AVX2-FCP-NEXT:    vmovaps %ymm2, 192(%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm15, 128(%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm13, 64(%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm12, (%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm11, 224(%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm9, 160(%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm8, 96(%rdx)
; AVX2-FCP-NEXT:    vmovups (%rsp), %ymm2 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm2, 32(%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm6, 192(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm7, 224(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm5, 128(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm4, 160(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm0, 64(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm1, 96(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm3, (%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm14, 32(%rcx)
; AVX2-FCP-NEXT:    addq $232, %rsp
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i64_stride3_vf32:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512-NEXT:    vmovdqa64 640(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512-NEXT:    vmovdqa64 512(%rdi), %zmm6
; AVX512-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512-NEXT:    vmovdqa64 320(%rdi), %zmm8
; AVX512-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm9
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512-NEXT:    vmovdqa64 128(%rdi), %zmm10
; AVX512-NEXT:    vmovdqa64 192(%rdi), %zmm11
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm12 = [0,3,6,9,12,15,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm13
; AVX512-NEXT:    vpermt2q %zmm1, %zmm12, %zmm13
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [0,1,2,3,4,5,10,13]
; AVX512-NEXT:    vpermt2q %zmm8, %zmm14, %zmm13
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm15
; AVX512-NEXT:    vpermt2q %zmm2, %zmm12, %zmm15
; AVX512-NEXT:    vpermt2q %zmm6, %zmm14, %zmm15
; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm16
; AVX512-NEXT:    vpermt2q %zmm0, %zmm12, %zmm16
; AVX512-NEXT:    vpermt2q %zmm4, %zmm14, %zmm16
; AVX512-NEXT:    vpermi2q %zmm3, %zmm9, %zmm12
; AVX512-NEXT:    vpermt2q %zmm10, %zmm14, %zmm12
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [1,4,7,10,13,0,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm7, %zmm17
; AVX512-NEXT:    vpermt2q %zmm2, %zmm14, %zmm17
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [0,1,2,3,4,8,11,14]
; AVX512-NEXT:    vpermt2q %zmm6, %zmm18, %zmm17
; AVX512-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512-NEXT:    vpermt2q %zmm1, %zmm14, %zmm19
; AVX512-NEXT:    vpermt2q %zmm8, %zmm18, %zmm19
; AVX512-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512-NEXT:    vpermt2q %zmm0, %zmm14, %zmm20
; AVX512-NEXT:    vpermt2q %zmm4, %zmm18, %zmm20
; AVX512-NEXT:    vpermi2q %zmm3, %zmm9, %zmm14
; AVX512-NEXT:    vpermt2q %zmm10, %zmm18, %zmm14
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [10,13,0,3,6,0,0,0]
; AVX512-NEXT:    vpermt2q %zmm11, %zmm18, %zmm1
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,1,2,3,4,9,12,15]
; AVX512-NEXT:    vpermt2q %zmm8, %zmm11, %zmm1
; AVX512-NEXT:    vpermt2q %zmm5, %zmm18, %zmm0
; AVX512-NEXT:    vpermt2q %zmm4, %zmm11, %zmm0
; AVX512-NEXT:    vpermt2q %zmm7, %zmm18, %zmm2
; AVX512-NEXT:    vpermt2q %zmm6, %zmm11, %zmm2
; AVX512-NEXT:    vpermt2q %zmm9, %zmm18, %zmm3
; AVX512-NEXT:    vpermt2q %zmm10, %zmm11, %zmm3
; AVX512-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm12, (%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm20, 192(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm14, (%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm17, 128(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm2, 128(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm0, 192(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i64_stride3_vf32:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm6
; AVX512-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm8
; AVX512-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %zmm9
; AVX512-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm10
; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm11
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm12 = [0,3,6,9,12,15,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, %zmm13
; AVX512-FCP-NEXT:    vpermt2q %zmm1, %zmm12, %zmm13
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [0,1,2,3,4,5,10,13]
; AVX512-FCP-NEXT:    vpermt2q %zmm8, %zmm14, %zmm13
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm15
; AVX512-FCP-NEXT:    vpermt2q %zmm2, %zmm12, %zmm15
; AVX512-FCP-NEXT:    vpermt2q %zmm6, %zmm14, %zmm15
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, %zmm16
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm12, %zmm16
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm14, %zmm16
; AVX512-FCP-NEXT:    vpermi2q %zmm3, %zmm9, %zmm12
; AVX512-FCP-NEXT:    vpermt2q %zmm10, %zmm14, %zmm12
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [1,4,7,10,13,0,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, %zmm17
; AVX512-FCP-NEXT:    vpermt2q %zmm2, %zmm14, %zmm17
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [0,1,2,3,4,8,11,14]
; AVX512-FCP-NEXT:    vpermt2q %zmm6, %zmm18, %zmm17
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512-FCP-NEXT:    vpermt2q %zmm1, %zmm14, %zmm19
; AVX512-FCP-NEXT:    vpermt2q %zmm8, %zmm18, %zmm19
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm14, %zmm20
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm18, %zmm20
; AVX512-FCP-NEXT:    vpermi2q %zmm3, %zmm9, %zmm14
; AVX512-FCP-NEXT:    vpermt2q %zmm10, %zmm18, %zmm14
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [10,13,0,3,6,0,0,0]
; AVX512-FCP-NEXT:    vpermt2q %zmm11, %zmm18, %zmm1
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,1,2,3,4,9,12,15]
; AVX512-FCP-NEXT:    vpermt2q %zmm8, %zmm11, %zmm1
; AVX512-FCP-NEXT:    vpermt2q %zmm5, %zmm18, %zmm0
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm11, %zmm0
; AVX512-FCP-NEXT:    vpermt2q %zmm7, %zmm18, %zmm2
; AVX512-FCP-NEXT:    vpermt2q %zmm6, %zmm11, %zmm2
; AVX512-FCP-NEXT:    vpermt2q %zmm9, %zmm18, %zmm3
; AVX512-FCP-NEXT:    vpermt2q %zmm10, %zmm11, %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm12, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm20, 192(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm14, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm17, 128(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm2, 128(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm0, 192(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i64_stride3_vf32:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512DQ-NEXT:    vmovdqa64 640(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512DQ-NEXT:    vmovdqa64 512(%rdi), %zmm6
; AVX512DQ-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512DQ-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %zmm8
; AVX512DQ-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm9
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %zmm10
; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %zmm11
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm12 = [0,3,6,9,12,15,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, %zmm13
; AVX512DQ-NEXT:    vpermt2q %zmm1, %zmm12, %zmm13
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [0,1,2,3,4,5,10,13]
; AVX512DQ-NEXT:    vpermt2q %zmm8, %zmm14, %zmm13
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm15
; AVX512DQ-NEXT:    vpermt2q %zmm2, %zmm12, %zmm15
; AVX512DQ-NEXT:    vpermt2q %zmm6, %zmm14, %zmm15
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, %zmm16
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm12, %zmm16
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm14, %zmm16
; AVX512DQ-NEXT:    vpermi2q %zmm3, %zmm9, %zmm12
; AVX512DQ-NEXT:    vpermt2q %zmm10, %zmm14, %zmm12
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [1,4,7,10,13,0,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, %zmm17
; AVX512DQ-NEXT:    vpermt2q %zmm2, %zmm14, %zmm17
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [0,1,2,3,4,8,11,14]
; AVX512DQ-NEXT:    vpermt2q %zmm6, %zmm18, %zmm17
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512DQ-NEXT:    vpermt2q %zmm1, %zmm14, %zmm19
; AVX512DQ-NEXT:    vpermt2q %zmm8, %zmm18, %zmm19
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm14, %zmm20
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm18, %zmm20
; AVX512DQ-NEXT:    vpermi2q %zmm3, %zmm9, %zmm14
; AVX512DQ-NEXT:    vpermt2q %zmm10, %zmm18, %zmm14
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [10,13,0,3,6,0,0,0]
; AVX512DQ-NEXT:    vpermt2q %zmm11, %zmm18, %zmm1
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,1,2,3,4,9,12,15]
; AVX512DQ-NEXT:    vpermt2q %zmm8, %zmm11, %zmm1
; AVX512DQ-NEXT:    vpermt2q %zmm5, %zmm18, %zmm0
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm11, %zmm0
; AVX512DQ-NEXT:    vpermt2q %zmm7, %zmm18, %zmm2
; AVX512DQ-NEXT:    vpermt2q %zmm6, %zmm11, %zmm2
; AVX512DQ-NEXT:    vpermt2q %zmm9, %zmm18, %zmm3
; AVX512DQ-NEXT:    vpermt2q %zmm10, %zmm11, %zmm3
; AVX512DQ-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm12, (%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm20, 192(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm14, (%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm17, 128(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm2, 128(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm0, 192(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride3_vf32:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512DQ-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm6
; AVX512DQ-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm8
; AVX512DQ-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %zmm9
; AVX512DQ-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm10
; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm11
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm12 = [0,3,6,9,12,15,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, %zmm13
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm1, %zmm12, %zmm13
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [0,1,2,3,4,5,10,13]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm8, %zmm14, %zmm13
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm15
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm2, %zmm12, %zmm15
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm6, %zmm14, %zmm15
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, %zmm16
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm12, %zmm16
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm14, %zmm16
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm3, %zmm9, %zmm12
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm10, %zmm14, %zmm12
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [1,4,7,10,13,0,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, %zmm17
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm2, %zmm14, %zmm17
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [0,1,2,3,4,8,11,14]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm6, %zmm18, %zmm17
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm1, %zmm14, %zmm19
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm8, %zmm18, %zmm19
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm14, %zmm20
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm18, %zmm20
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm3, %zmm9, %zmm14
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm10, %zmm18, %zmm14
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [10,13,0,3,6,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm11, %zmm18, %zmm1
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,1,2,3,4,9,12,15]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm8, %zmm11, %zmm1
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm5, %zmm18, %zmm0
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm11, %zmm0
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm7, %zmm18, %zmm2
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm6, %zmm11, %zmm2
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm9, %zmm18, %zmm3
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm10, %zmm11, %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm12, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm20, 192(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm14, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm17, 128(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm2, 128(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm0, 192(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i64_stride3_vf32:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm6
; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm8
; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm9
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm10
; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm11
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm12 = [0,3,6,9,12,15,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm13
; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm12, %zmm13
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [0,1,2,3,4,5,10,13]
; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm14, %zmm13
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm15
; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm12, %zmm15
; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm14, %zmm15
; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm16
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm12, %zmm16
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm14, %zmm16
; AVX512BW-NEXT:    vpermi2q %zmm3, %zmm9, %zmm12
; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm14, %zmm12
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [1,4,7,10,13,0,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm7, %zmm17
; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm14, %zmm17
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [0,1,2,3,4,8,11,14]
; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm18, %zmm17
; AVX512BW-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm14, %zmm19
; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm18, %zmm19
; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm14, %zmm20
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm18, %zmm20
; AVX512BW-NEXT:    vpermi2q %zmm3, %zmm9, %zmm14
; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm18, %zmm14
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [10,13,0,3,6,0,0,0]
; AVX512BW-NEXT:    vpermt2q %zmm11, %zmm18, %zmm1
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,1,2,3,4,9,12,15]
; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm11, %zmm1
; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm18, %zmm0
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm11, %zmm0
; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm18, %zmm2
; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm11, %zmm2
; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm18, %zmm3
; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm11, %zmm3
; AVX512BW-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm12, (%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm20, 192(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm14, (%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm17, 128(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm2, 128(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm0, 192(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i64_stride3_vf32:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512BW-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512BW-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm6
; AVX512BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm8
; AVX512BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm9
; AVX512BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm10
; AVX512BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm11
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm12 = [0,3,6,9,12,15,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm13
; AVX512BW-FCP-NEXT:    vpermt2q %zmm1, %zmm12, %zmm13
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [0,1,2,3,4,5,10,13]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm8, %zmm14, %zmm13
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm15
; AVX512BW-FCP-NEXT:    vpermt2q %zmm2, %zmm12, %zmm15
; AVX512BW-FCP-NEXT:    vpermt2q %zmm6, %zmm14, %zmm15
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm16
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm12, %zmm16
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm14, %zmm16
; AVX512BW-FCP-NEXT:    vpermi2q %zmm3, %zmm9, %zmm12
; AVX512BW-FCP-NEXT:    vpermt2q %zmm10, %zmm14, %zmm12
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [1,4,7,10,13,0,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm17
; AVX512BW-FCP-NEXT:    vpermt2q %zmm2, %zmm14, %zmm17
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [0,1,2,3,4,8,11,14]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm6, %zmm18, %zmm17
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512BW-FCP-NEXT:    vpermt2q %zmm1, %zmm14, %zmm19
; AVX512BW-FCP-NEXT:    vpermt2q %zmm8, %zmm18, %zmm19
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm14, %zmm20
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm18, %zmm20
; AVX512BW-FCP-NEXT:    vpermi2q %zmm3, %zmm9, %zmm14
; AVX512BW-FCP-NEXT:    vpermt2q %zmm10, %zmm18, %zmm14
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [10,13,0,3,6,0,0,0]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm11, %zmm18, %zmm1
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,1,2,3,4,9,12,15]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm8, %zmm11, %zmm1
; AVX512BW-FCP-NEXT:    vpermt2q %zmm5, %zmm18, %zmm0
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm11, %zmm0
; AVX512BW-FCP-NEXT:    vpermt2q %zmm7, %zmm18, %zmm2
; AVX512BW-FCP-NEXT:    vpermt2q %zmm6, %zmm11, %zmm2
; AVX512BW-FCP-NEXT:    vpermt2q %zmm9, %zmm18, %zmm3
; AVX512BW-FCP-NEXT:    vpermt2q %zmm10, %zmm11, %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm12, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm20, 192(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm14, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm17, 128(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, 128(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, 192(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i64_stride3_vf32:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512DQ-BW-NEXT:    vmovdqa64 640(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512DQ-BW-NEXT:    vmovdqa64 512(%rdi), %zmm6
; AVX512DQ-BW-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512DQ-BW-NEXT:    vmovdqa64 320(%rdi), %zmm8
; AVX512DQ-BW-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 (%rdi), %zmm9
; AVX512DQ-BW-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 128(%rdi), %zmm10
; AVX512DQ-BW-NEXT:    vmovdqa64 192(%rdi), %zmm11
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm12 = [0,3,6,9,12,15,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, %zmm13
; AVX512DQ-BW-NEXT:    vpermt2q %zmm1, %zmm12, %zmm13
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [0,1,2,3,4,5,10,13]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm8, %zmm14, %zmm13
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm15
; AVX512DQ-BW-NEXT:    vpermt2q %zmm2, %zmm12, %zmm15
; AVX512DQ-BW-NEXT:    vpermt2q %zmm6, %zmm14, %zmm15
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, %zmm16
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm12, %zmm16
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm14, %zmm16
; AVX512DQ-BW-NEXT:    vpermi2q %zmm3, %zmm9, %zmm12
; AVX512DQ-BW-NEXT:    vpermt2q %zmm10, %zmm14, %zmm12
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [1,4,7,10,13,0,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, %zmm17
; AVX512DQ-BW-NEXT:    vpermt2q %zmm2, %zmm14, %zmm17
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [0,1,2,3,4,8,11,14]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm6, %zmm18, %zmm17
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512DQ-BW-NEXT:    vpermt2q %zmm1, %zmm14, %zmm19
; AVX512DQ-BW-NEXT:    vpermt2q %zmm8, %zmm18, %zmm19
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm14, %zmm20
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm18, %zmm20
; AVX512DQ-BW-NEXT:    vpermi2q %zmm3, %zmm9, %zmm14
; AVX512DQ-BW-NEXT:    vpermt2q %zmm10, %zmm18, %zmm14
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [10,13,0,3,6,0,0,0]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm11, %zmm18, %zmm1
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,1,2,3,4,9,12,15]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm8, %zmm11, %zmm1
; AVX512DQ-BW-NEXT:    vpermt2q %zmm5, %zmm18, %zmm0
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm11, %zmm0
; AVX512DQ-BW-NEXT:    vpermt2q %zmm7, %zmm18, %zmm2
; AVX512DQ-BW-NEXT:    vpermt2q %zmm6, %zmm11, %zmm2
; AVX512DQ-BW-NEXT:    vpermt2q %zmm9, %zmm18, %zmm3
; AVX512DQ-BW-NEXT:    vpermt2q %zmm10, %zmm11, %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm12, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm20, 192(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm14, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm17, 128(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, 128(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, 192(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride3_vf32:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm5
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm6
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm8
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm9
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm10
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm11
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm12 = [0,3,6,9,12,15,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm1, %zmm12, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [0,1,2,3,4,5,10,13]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm8, %zmm14, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm2, %zmm12, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm6, %zmm14, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm12, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm14, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm3, %zmm9, %zmm12
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm10, %zmm14, %zmm12
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm14 = [1,4,7,10,13,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm2, %zmm14, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [0,1,2,3,4,8,11,14]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm6, %zmm18, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, %zmm19
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm1, %zmm14, %zmm19
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm8, %zmm18, %zmm19
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm5, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm14, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm18, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm3, %zmm9, %zmm14
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm10, %zmm18, %zmm14
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm18 = [10,13,0,3,6,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm11, %zmm18, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,1,2,3,4,9,12,15]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm8, %zmm11, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm5, %zmm18, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm11, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm7, %zmm18, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm6, %zmm11, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm9, %zmm18, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm10, %zmm11, %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm13, 64(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm12, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm20, 192(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm14, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm19, 64(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm17, 128(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, 128(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, 192(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm3, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm1, 64(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <96 x i64>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <96 x i64> %wide.vec, <96 x i64> poison, <32 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45, i32 48, i32 51, i32 54, i32 57, i32 60, i32 63, i32 66, i32 69, i32 72, i32 75, i32 78, i32 81, i32 84, i32 87, i32 90, i32 93>
  %strided.vec1 = shufflevector <96 x i64> %wide.vec, <96 x i64> poison, <32 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46, i32 49, i32 52, i32 55, i32 58, i32 61, i32 64, i32 67, i32 70, i32 73, i32 76, i32 79, i32 82, i32 85, i32 88, i32 91, i32 94>
  %strided.vec2 = shufflevector <96 x i64> %wide.vec, <96 x i64> poison, <32 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47, i32 50, i32 53, i32 56, i32 59, i32 62, i32 65, i32 68, i32 71, i32 74, i32 77, i32 80, i32 83, i32 86, i32 89, i32 92, i32 95>
  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
  ret void
}

define void @load_i64_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
; SSE-LABEL: load_i64_stride3_vf64:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $1176, %rsp # imm = 0x498
; SSE-NEXT:    movapd 272(%rdi), %xmm9
; SSE-NEXT:    movapd 224(%rdi), %xmm8
; SSE-NEXT:    movapd 176(%rdi), %xmm7
; SSE-NEXT:    movapd 128(%rdi), %xmm6
; SSE-NEXT:    movapd 80(%rdi), %xmm5
; SSE-NEXT:    movapd 240(%rdi), %xmm10
; SSE-NEXT:    movapd 256(%rdi), %xmm0
; SSE-NEXT:    movapd 192(%rdi), %xmm11
; SSE-NEXT:    movapd 208(%rdi), %xmm1
; SSE-NEXT:    movapd 144(%rdi), %xmm12
; SSE-NEXT:    movapd 160(%rdi), %xmm2
; SSE-NEXT:    movapd 96(%rdi), %xmm13
; SSE-NEXT:    movapd 112(%rdi), %xmm3
; SSE-NEXT:    movapd 48(%rdi), %xmm14
; SSE-NEXT:    movapd 64(%rdi), %xmm4
; SSE-NEXT:    movapd %xmm4, %xmm15
; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm14[0],xmm15[1]
; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufpd {{.*#+}} xmm14 = xmm14[1],xmm5[0]
; SSE-NEXT:    movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd %xmm3, %xmm4
; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm13[0],xmm4[1]
; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufpd {{.*#+}} xmm13 = xmm13[1],xmm6[0]
; SSE-NEXT:    movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm3[0],xmm6[1]
; SSE-NEXT:    movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd %xmm2, %xmm3
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm12[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufpd {{.*#+}} xmm12 = xmm12[1],xmm7[0]
; SSE-NEXT:    movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm2[0],xmm7[1]
; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd %xmm1, %xmm2
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm11[0],xmm2[1]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufpd {{.*#+}} xmm11 = xmm11[1],xmm8[0]
; SSE-NEXT:    movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
; SSE-NEXT:    movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm10[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    shufpd {{.*#+}} xmm10 = xmm10[1],xmm9[0]
; SSE-NEXT:    movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
; SSE-NEXT:    movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 288(%rdi), %xmm2
; SSE-NEXT:    movapd 304(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 320(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 336(%rdi), %xmm2
; SSE-NEXT:    movapd 352(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 368(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 384(%rdi), %xmm2
; SSE-NEXT:    movapd 400(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 416(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 432(%rdi), %xmm2
; SSE-NEXT:    movapd 448(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 464(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 480(%rdi), %xmm2
; SSE-NEXT:    movapd 496(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 512(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 528(%rdi), %xmm2
; SSE-NEXT:    movapd 544(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 560(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 576(%rdi), %xmm2
; SSE-NEXT:    movapd 592(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 608(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 624(%rdi), %xmm2
; SSE-NEXT:    movapd 640(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 656(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 672(%rdi), %xmm2
; SSE-NEXT:    movapd 688(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 704(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 720(%rdi), %xmm2
; SSE-NEXT:    movapd 736(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 752(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 768(%rdi), %xmm2
; SSE-NEXT:    movapd 784(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 800(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 816(%rdi), %xmm2
; SSE-NEXT:    movapd 832(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 848(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 864(%rdi), %xmm2
; SSE-NEXT:    movapd 880(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 896(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 912(%rdi), %xmm2
; SSE-NEXT:    movapd 928(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 944(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 960(%rdi), %xmm2
; SSE-NEXT:    movapd 976(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 992(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1008(%rdi), %xmm2
; SSE-NEXT:    movapd 1024(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1040(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1056(%rdi), %xmm2
; SSE-NEXT:    movapd 1072(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1088(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1104(%rdi), %xmm2
; SSE-NEXT:    movapd 1120(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1136(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1152(%rdi), %xmm2
; SSE-NEXT:    movapd 1168(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1184(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1200(%rdi), %xmm2
; SSE-NEXT:    movapd 1216(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1232(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1248(%rdi), %xmm2
; SSE-NEXT:    movapd 1264(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm14
; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm2[0],xmm14[1]
; SSE-NEXT:    movapd 1280(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1296(%rdi), %xmm15
; SSE-NEXT:    movapd 1312(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm11
; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm15[0],xmm11[1]
; SSE-NEXT:    movapd 1328(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm15 = xmm15[1],xmm1[0]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1344(%rdi), %xmm12
; SSE-NEXT:    movapd 1360(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm7
; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm12[0],xmm7[1]
; SSE-NEXT:    movapd 1376(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm12 = xmm12[1],xmm1[0]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, (%rsp) # 16-byte Spill
; SSE-NEXT:    movapd 1392(%rdi), %xmm10
; SSE-NEXT:    movapd 1408(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm2
; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm10[0],xmm2[1]
; SSE-NEXT:    movapd 1424(%rdi), %xmm1
; SSE-NEXT:    shufpd {{.*#+}} xmm10 = xmm10[1],xmm1[0]
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1440(%rdi), %xmm9
; SSE-NEXT:    movapd 1456(%rdi), %xmm0
; SSE-NEXT:    movapd %xmm0, %xmm1
; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm9[0],xmm1[1]
; SSE-NEXT:    movapd 1472(%rdi), %xmm3
; SSE-NEXT:    shufpd {{.*#+}} xmm9 = xmm9[1],xmm3[0]
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movapd 1488(%rdi), %xmm0
; SSE-NEXT:    movapd 1504(%rdi), %xmm8
; SSE-NEXT:    movapd %xmm8, %xmm3
; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
; SSE-NEXT:    movapd 1520(%rdi), %xmm13
; SSE-NEXT:    shufpd {{.*#+}} xmm0 = xmm0[1],xmm13[0]
; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm8[0],xmm13[1]
; SSE-NEXT:    movapd (%rdi), %xmm8
; SSE-NEXT:    movapd 16(%rdi), %xmm5
; SSE-NEXT:    movapd %xmm5, %xmm6
; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm8[0],xmm6[1]
; SSE-NEXT:    movapd 32(%rdi), %xmm4
; SSE-NEXT:    shufpd {{.*#+}} xmm8 = xmm8[1],xmm4[0]
; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm5[0],xmm4[1]
; SSE-NEXT:    movapd %xmm3, 496(%rsi)
; SSE-NEXT:    movapd %xmm1, 480(%rsi)
; SSE-NEXT:    movapd %xmm2, 464(%rsi)
; SSE-NEXT:    movapd %xmm7, 448(%rsi)
; SSE-NEXT:    movapd %xmm11, 432(%rsi)
; SSE-NEXT:    movapd %xmm14, 416(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 400(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 384(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 368(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 352(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 336(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 320(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 304(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 288(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 272(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 256(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 240(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 224(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 208(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 192(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 176(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 160(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 144(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 128(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 112(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 96(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 80(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 64(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 48(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 32(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 16(%rsi)
; SSE-NEXT:    movapd %xmm6, (%rsi)
; SSE-NEXT:    movapd %xmm0, 496(%rdx)
; SSE-NEXT:    movapd %xmm9, 480(%rdx)
; SSE-NEXT:    movapd %xmm10, 464(%rdx)
; SSE-NEXT:    movapd %xmm12, 448(%rdx)
; SSE-NEXT:    movapd %xmm15, 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:    movapd %xmm8, (%rdx)
; SSE-NEXT:    movapd %xmm13, 496(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 480(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 464(%rcx)
; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 448(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 432(%rcx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movaps %xmm0, 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:    movapd %xmm4, (%rcx)
; SSE-NEXT:    addq $1176, %rsp # imm = 0x498
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i64_stride3_vf64:
; AVX:       # %bb.0:
; AVX-NEXT:    subq $1096, %rsp # imm = 0x448
; AVX-NEXT:    vmovapd 896(%rdi), %ymm0
; AVX-NEXT:    vmovapd 704(%rdi), %ymm1
; AVX-NEXT:    vmovapd 512(%rdi), %ymm2
; AVX-NEXT:    vmovapd 320(%rdi), %ymm3
; AVX-NEXT:    vmovapd 128(%rdi), %ymm4
; AVX-NEXT:    vblendpd {{.*#+}} ymm7 = mem[0,1],ymm4[2,3]
; AVX-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd %ymm4, %ymm6
; AVX-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendpd {{.*#+}} ymm8 = mem[0,1],ymm3[2,3]
; AVX-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd %ymm3, %ymm5
; AVX-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendpd {{.*#+}} ymm9 = mem[0,1],ymm2[2,3]
; AVX-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd %ymm2, %ymm4
; AVX-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendpd {{.*#+}} ymm10 = mem[0,1],ymm1[2,3]
; AVX-NEXT:    vmovupd %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd %ymm1, %ymm3
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendpd {{.*#+}} ymm11 = mem[0,1],ymm0[2,3]
; AVX-NEXT:    vmovupd %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd %ymm0, %ymm2
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm7[1],ymm6[0],ymm7[3],ymm6[2]
; AVX-NEXT:    vbroadcastsd 176(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm8[1],ymm5[0],ymm8[3],ymm5[2]
; AVX-NEXT:    vbroadcastsd 368(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm9[1],ymm4[0],ymm9[3],ymm4[2]
; AVX-NEXT:    vbroadcastsd 560(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm10[1],ymm3[0],ymm10[3],ymm3[2]
; AVX-NEXT:    vbroadcastsd 752(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm11[1],ymm2[0],ymm11[3],ymm2[2]
; AVX-NEXT:    vbroadcastsd 944(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 1088(%rdi), %ymm9
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = mem[0,1],ymm9[2,3]
; AVX-NEXT:    vmovupd %ymm0, (%rsp) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm9[0],ymm0[3],ymm9[2]
; AVX-NEXT:    vbroadcastsd 1136(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 1280(%rdi), %ymm8
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = mem[0,1],ymm8[2,3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm8[0],ymm0[3],ymm8[2]
; AVX-NEXT:    vbroadcastsd 1328(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 1472(%rdi), %ymm7
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = mem[0,1],ymm7[2,3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm7[0],ymm0[3],ymm7[2]
; AVX-NEXT:    vbroadcastsd 1520(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 32(%rdi), %ymm0
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendpd {{.*#+}} ymm1 = mem[0,1],ymm0[2,3]
; AVX-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[1],ymm0[0],ymm1[3],ymm0[2]
; AVX-NEXT:    vbroadcastsd 80(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 224(%rdi), %ymm13
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = mem[0,1],ymm13[2,3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm13[0],ymm0[3],ymm13[2]
; AVX-NEXT:    vbroadcastsd 272(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 416(%rdi), %ymm12
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = mem[0,1],ymm12[2,3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm12[0],ymm0[3],ymm12[2]
; AVX-NEXT:    vbroadcastsd 464(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 608(%rdi), %ymm10
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = mem[0,1],ymm10[2,3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm10[0],ymm0[3],ymm10[2]
; AVX-NEXT:    vbroadcastsd 656(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 800(%rdi), %ymm5
; AVX-NEXT:    vblendpd {{.*#+}} ymm14 = mem[0,1],ymm5[2,3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm14[1],ymm5[0],ymm14[3],ymm5[2]
; AVX-NEXT:    vbroadcastsd 848(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 992(%rdi), %ymm4
; AVX-NEXT:    vblendpd {{.*#+}} ymm11 = mem[0,1],ymm4[2,3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm11[1],ymm4[0],ymm11[3],ymm4[2]
; AVX-NEXT:    vbroadcastsd 1040(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 1184(%rdi), %ymm2
; AVX-NEXT:    vblendpd {{.*#+}} ymm6 = mem[0,1],ymm2[2,3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm6[1],ymm2[0],ymm6[3],ymm2[2]
; AVX-NEXT:    vbroadcastsd 1232(%rdi), %ymm3
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovapd 1376(%rdi), %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm3 = mem[0,1],ymm1[2,3]
; AVX-NEXT:    vshufpd {{.*#+}} ymm0 = ymm3[1],ymm1[0],ymm3[3],ymm1[2]
; AVX-NEXT:    vbroadcastsd 1424(%rdi), %ymm15
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 112(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],mem[4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm15[2,3],ymm0[4,5],ymm15[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 304(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
; AVX-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],mem[4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm15[2,3],ymm0[4,5],ymm15[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 496(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 544(%rdi), %ymm0, %ymm0
; AVX-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],mem[4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm15[2,3],ymm0[4,5],ymm15[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 688(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 736(%rdi), %ymm0, %ymm0
; AVX-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],mem[4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm15[2,3],ymm0[4,5],ymm15[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 880(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 928(%rdi), %ymm0, %ymm0
; AVX-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX-NEXT:    vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],mem[4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm15[2,3],ymm0[4,5],ymm15[6,7]
; AVX-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1072(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 1120(%rdi), %ymm0, %ymm0
; AVX-NEXT:    vblendpd $5, (%rsp), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = mem[0],ymm0[1],mem[2],ymm0[3]
; AVX-NEXT:    vmovupd %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm9[1],ymm0[2],ymm9[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1264(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 1312(%rdi), %ymm0, %ymm0
; AVX-NEXT:    vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX-NEXT:    # ymm15 = mem[0],ymm0[1],mem[2],ymm0[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm8 = ymm8[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm8[1],ymm0[2],ymm8[3]
; AVX-NEXT:    vmovupd %ymm0, (%rsp) # 32-byte Spill
; AVX-NEXT:    vmovaps 1456(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 1504(%rdi), %ymm0, %ymm0
; AVX-NEXT:    vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
; AVX-NEXT:    # ymm8 = mem[0],ymm0[1],mem[2],ymm0[3]
; AVX-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm7[1],ymm0[2],ymm7[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1360(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 1408(%rdi), %ymm0, %ymm0
; AVX-NEXT:    vblendpd {{.*#+}} ymm7 = ymm3[0],ymm0[1],ymm3[2],ymm0[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 1168(%rdi), %xmm0
; AVX-NEXT:    vinsertf128 $1, 1216(%rdi), %ymm0, %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm3 = ymm6[0],ymm1[1],ymm6[2],ymm1[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm2[1],ymm1[2],ymm2[3]
; AVX-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX-NEXT:    vmovaps 976(%rdi), %xmm1
; AVX-NEXT:    vinsertf128 $1, 1024(%rdi), %ymm1, %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm2 = ymm11[0],ymm1[1],ymm11[2],ymm1[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm6 = ymm1[0],ymm4[1],ymm1[2],ymm4[3]
; AVX-NEXT:    vmovaps 784(%rdi), %xmm1
; AVX-NEXT:    vinsertf128 $1, 832(%rdi), %ymm1, %ymm1
; AVX-NEXT:    vblendpd {{.*#+}} ymm14 = ymm14[0],ymm1[1],ymm14[2],ymm1[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm4 = ymm5[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2],ymm4[3]
; AVX-NEXT:    vmovaps 592(%rdi), %xmm4
; AVX-NEXT:    vinsertf128 $1, 640(%rdi), %ymm4, %ymm4
; AVX-NEXT:    vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
; AVX-NEXT:    # ymm5 = mem[0],ymm4[1],mem[2],ymm4[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm9 = ymm10[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm10 = ymm4[0],ymm9[1],ymm4[2],ymm9[3]
; AVX-NEXT:    vmovaps 400(%rdi), %xmm4
; AVX-NEXT:    vinsertf128 $1, 448(%rdi), %ymm4, %ymm4
; AVX-NEXT:    vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm9 # 32-byte Folded Reload
; AVX-NEXT:    # ymm9 = mem[0],ymm4[1],mem[2],ymm4[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm11 = ymm12[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2],ymm11[3]
; AVX-NEXT:    vmovaps 208(%rdi), %xmm11
; AVX-NEXT:    vinsertf128 $1, 256(%rdi), %ymm11, %ymm11
; AVX-NEXT:    vblendpd $5, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm12 # 32-byte Folded Reload
; AVX-NEXT:    # ymm12 = mem[0],ymm11[1],mem[2],ymm11[3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm13 = ymm13[0,1],mem[2,3]
; AVX-NEXT:    vblendpd {{.*#+}} ymm13 = ymm11[0],ymm13[1],ymm11[2],ymm13[3]
; AVX-NEXT:    vmovaps 16(%rdi), %xmm11
; AVX-NEXT:    vinsertf128 $1, 64(%rdi), %ymm11, %ymm11
; AVX-NEXT:    vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
; AVX-NEXT:    # ymm0 = mem[0,1],ymm11[2,3],mem[4,5],ymm11[6,7]
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],mem[4,5,6,7]
; AVX-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm8[2,3],ymm11[4,5],ymm8[6,7]
; AVX-NEXT:    vmovapd %ymm7, 448(%rsi)
; AVX-NEXT:    vmovapd %ymm3, 384(%rsi)
; AVX-NEXT:    vmovapd %ymm2, 320(%rsi)
; AVX-NEXT:    vmovapd %ymm14, 256(%rsi)
; AVX-NEXT:    vmovapd %ymm5, 192(%rsi)
; AVX-NEXT:    vmovapd %ymm9, 128(%rsi)
; AVX-NEXT:    vmovapd %ymm12, 64(%rsi)
; AVX-NEXT:    vmovaps %ymm0, (%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 480(%rsi)
; AVX-NEXT:    vmovapd %ymm15, 416(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 352(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 288(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 224(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 160(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 96(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 32(%rsi)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 448(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 384(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 320(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 256(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 192(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 128(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 64(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, (%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 480(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 416(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 352(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 288(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 224(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 160(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 96(%rdx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 32(%rdx)
; AVX-NEXT:    vmovaps %ymm11, (%rcx)
; AVX-NEXT:    vmovapd %ymm13, 64(%rcx)
; AVX-NEXT:    vmovapd %ymm4, 128(%rcx)
; AVX-NEXT:    vmovapd %ymm10, 192(%rcx)
; AVX-NEXT:    vmovapd %ymm1, 256(%rcx)
; AVX-NEXT:    vmovapd %ymm6, 320(%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, 448(%rcx)
; AVX-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX-NEXT:    vmovaps %ymm0, 480(%rcx)
; AVX-NEXT:    vmovups (%rsp), %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:    addq $1096, %rsp # imm = 0x448
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i64_stride3_vf64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    subq $968, %rsp # imm = 0x3C8
; AVX2-NEXT:    vmovaps 896(%rdi), %ymm2
; AVX2-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 864(%rdi), %ymm3
; AVX2-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 704(%rdi), %ymm4
; AVX2-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 672(%rdi), %ymm5
; AVX2-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 512(%rdi), %ymm6
; AVX2-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 480(%rdi), %ymm7
; AVX2-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 320(%rdi), %ymm8
; AVX2-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 288(%rdi), %ymm9
; AVX2-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 128(%rdi), %ymm10
; AVX2-NEXT:    vmovups %ymm10, {{[-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:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm9[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vinsertf128 $1, 544(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm7[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vinsertf128 $1, 736(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm5[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vinsertf128 $1, 928(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm3[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1088(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1056(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 1120(%rdi), %ymm0, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1280(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1248(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 1312(%rdi), %ymm0, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1472(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1440(%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 1504(%rdi), %ymm0, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps (%rdi), %ymm0
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 224(%rdi), %ymm13
; AVX2-NEXT:    vmovaps 192(%rdi), %ymm12
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm12[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm13[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 256(%rdi), %ymm0, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 416(%rdi), %ymm11
; AVX2-NEXT:    vmovaps 384(%rdi), %ymm10
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm10[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm11[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 448(%rdi), %ymm0, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 608(%rdi), %ymm9
; AVX2-NEXT:    vmovaps 576(%rdi), %ymm8
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm8[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 640(%rdi), %ymm0, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 800(%rdi), %ymm7
; AVX2-NEXT:    vmovaps 768(%rdi), %ymm6
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm6[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 832(%rdi), %ymm0, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 992(%rdi), %ymm5
; AVX2-NEXT:    vmovaps 960(%rdi), %ymm4
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm4[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 1024(%rdi), %ymm0, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1184(%rdi), %ymm3
; AVX2-NEXT:    vmovaps 1152(%rdi), %ymm2
; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm2[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 1216(%rdi), %ymm0, %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 1376(%rdi), %ymm1
; AVX2-NEXT:    vmovaps 1344(%rdi), %ymm0
; AVX2-NEXT:    vpermpd {{.*#+}} ymm15 = ymm0[0,3,2,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vinsertf128 $1, 1408(%rdi), %ymm0, %ymm14
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 176(%rdi), %ymm15
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 368(%rdi), %ymm15
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 560(%rdi), %ymm15
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 752(%rdi), %ymm15
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 944(%rdi), %ymm15
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 1136(%rdi), %ymm15
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, (%rsp), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 1328(%rdi), %ymm15
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 1520(%rdi), %ymm15
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 80(%rdi), %ymm15
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2,3],ymm13[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 272(%rdi), %ymm13
; AVX2-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm13[6,7]
; AVX2-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,3],ymm11[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 464(%rdi), %ymm11
; AVX2-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
; AVX2-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3],ymm9[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 656(%rdi), %ymm9
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
; AVX2-NEXT:    vmovups %ymm8, (%rsp) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 848(%rdi), %ymm7
; AVX2-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
; AVX2-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 1040(%rdi), %ymm5
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
; AVX2-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 1232(%rdi), %ymm3
; AVX2-NEXT:    vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
; AVX2-NEXT:    vbroadcastsd 1424(%rdi), %ymm1
; AVX2-NEXT:    vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-NEXT:    vmovaps 112(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 304(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 496(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-NEXT:    vmovaps 688(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 880(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 1072(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 1264(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 1456(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 1360(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm7 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 1168(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 976(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 784(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 592(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 400(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 208(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT:    vmovaps 16(%rdi), %xmm0
; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-NEXT:    vpermpd {{.*#+}} ymm15 = mem[0,1,0,3]
; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 448(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 384(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 320(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 256(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 192(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 128(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 64(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, (%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 480(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 416(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 352(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 288(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 224(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 160(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 96(%rsi)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm15, 32(%rsi)
; AVX2-NEXT:    vmovaps %ymm11, 448(%rdx)
; AVX2-NEXT:    vmovaps %ymm13, 384(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, 320(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, 256(%rdx)
; AVX2-NEXT:    vmovups (%rsp), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, 192(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, 128(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, 64(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, (%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, 480(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, 416(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, 352(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, 288(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, 224(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, 160(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, 96(%rdx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm11, 32(%rdx)
; AVX2-NEXT:    vmovaps %ymm0, (%rcx)
; AVX2-NEXT:    vmovaps %ymm1, 64(%rcx)
; AVX2-NEXT:    vmovaps %ymm2, 128(%rcx)
; AVX2-NEXT:    vmovaps %ymm3, 192(%rcx)
; AVX2-NEXT:    vmovaps %ymm4, 256(%rcx)
; AVX2-NEXT:    vmovaps %ymm5, 320(%rcx)
; AVX2-NEXT:    vmovaps %ymm6, 384(%rcx)
; AVX2-NEXT:    vmovaps %ymm7, 448(%rcx)
; AVX2-NEXT:    vmovaps %ymm8, 480(%rcx)
; AVX2-NEXT:    vmovaps %ymm9, 416(%rcx)
; AVX2-NEXT:    vmovaps %ymm10, 352(%rcx)
; AVX2-NEXT:    vmovaps %ymm12, 288(%rcx)
; AVX2-NEXT:    vmovaps %ymm14, 224(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 160(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 96(%rcx)
; AVX2-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-NEXT:    vmovaps %ymm0, 32(%rcx)
; AVX2-NEXT:    addq $968, %rsp # imm = 0x3C8
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i64_stride3_vf64:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    subq $968, %rsp # imm = 0x3C8
; AVX2-FP-NEXT:    vmovaps 896(%rdi), %ymm2
; AVX2-FP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 864(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 704(%rdi), %ymm4
; AVX2-FP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 672(%rdi), %ymm5
; AVX2-FP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 512(%rdi), %ymm6
; AVX2-FP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 480(%rdi), %ymm7
; AVX2-FP-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 320(%rdi), %ymm8
; AVX2-FP-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 288(%rdi), %ymm9
; AVX2-FP-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 128(%rdi), %ymm10
; AVX2-FP-NEXT:    vmovups %ymm10, {{[-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:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm9[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vinsertf128 $1, 544(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm7[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vinsertf128 $1, 736(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm5[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vinsertf128 $1, 928(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm3[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1088(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1056(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 1120(%rdi), %ymm0, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1280(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1248(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 1312(%rdi), %ymm0, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1472(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1440(%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 1504(%rdi), %ymm0, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps (%rdi), %ymm0
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 224(%rdi), %ymm13
; AVX2-FP-NEXT:    vmovaps 192(%rdi), %ymm12
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm12[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm13[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 256(%rdi), %ymm0, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 416(%rdi), %ymm11
; AVX2-FP-NEXT:    vmovaps 384(%rdi), %ymm10
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm10[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm11[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 448(%rdi), %ymm0, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 608(%rdi), %ymm9
; AVX2-FP-NEXT:    vmovaps 576(%rdi), %ymm8
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm8[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 640(%rdi), %ymm0, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 800(%rdi), %ymm7
; AVX2-FP-NEXT:    vmovaps 768(%rdi), %ymm6
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm6[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 832(%rdi), %ymm0, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 992(%rdi), %ymm5
; AVX2-FP-NEXT:    vmovaps 960(%rdi), %ymm4
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm4[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 1024(%rdi), %ymm0, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1184(%rdi), %ymm3
; AVX2-FP-NEXT:    vmovaps 1152(%rdi), %ymm2
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm2[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 1216(%rdi), %ymm0, %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 1376(%rdi), %ymm1
; AVX2-FP-NEXT:    vmovaps 1344(%rdi), %ymm0
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm15 = ymm0[0,3,2,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vinsertf128 $1, 1408(%rdi), %ymm0, %ymm14
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 176(%rdi), %ymm15
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 368(%rdi), %ymm15
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 560(%rdi), %ymm15
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 752(%rdi), %ymm15
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 944(%rdi), %ymm15
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 1136(%rdi), %ymm15
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, (%rsp), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 1328(%rdi), %ymm15
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 1520(%rdi), %ymm15
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 80(%rdi), %ymm15
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2,3],ymm13[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 272(%rdi), %ymm13
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm13[6,7]
; AVX2-FP-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,3],ymm11[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 464(%rdi), %ymm11
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
; AVX2-FP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3],ymm9[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 656(%rdi), %ymm9
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FP-NEXT:    vmovups %ymm8, (%rsp) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 848(%rdi), %ymm7
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 1040(%rdi), %ymm5
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 1232(%rdi), %ymm3
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
; AVX2-FP-NEXT:    vbroadcastsd 1424(%rdi), %ymm1
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FP-NEXT:    vmovaps 112(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 304(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 496(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FP-NEXT:    vmovaps 688(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 880(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 1072(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 1264(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 1456(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 1360(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm7 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 1168(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 976(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 784(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 592(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 400(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 208(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FP-NEXT:    vmovaps 16(%rdi), %xmm0
; AVX2-FP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FP-NEXT:    vpermpd {{.*#+}} ymm15 = mem[0,1,0,3]
; AVX2-FP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 448(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 384(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 320(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 256(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 192(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 128(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 64(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, (%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 480(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 416(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 352(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 288(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 224(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 160(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 96(%rsi)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm15, 32(%rsi)
; AVX2-FP-NEXT:    vmovaps %ymm11, 448(%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm13, 384(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, 320(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, 256(%rdx)
; AVX2-FP-NEXT:    vmovups (%rsp), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, 192(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, 128(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, 64(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, (%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, 480(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, 416(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, 352(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, 288(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, 224(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, 160(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, 96(%rdx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm11, 32(%rdx)
; AVX2-FP-NEXT:    vmovaps %ymm0, (%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm1, 64(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm2, 128(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm3, 192(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm4, 256(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm5, 320(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm6, 384(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm7, 448(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm8, 480(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm9, 416(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm10, 352(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm12, 288(%rcx)
; AVX2-FP-NEXT:    vmovaps %ymm14, 224(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 160(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 96(%rcx)
; AVX2-FP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FP-NEXT:    vmovaps %ymm0, 32(%rcx)
; AVX2-FP-NEXT:    addq $968, %rsp # imm = 0x3C8
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i64_stride3_vf64:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    subq $968, %rsp # imm = 0x3C8
; AVX2-FCP-NEXT:    vmovaps 896(%rdi), %ymm2
; AVX2-FCP-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 864(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 704(%rdi), %ymm4
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 672(%rdi), %ymm5
; AVX2-FCP-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 512(%rdi), %ymm6
; AVX2-FCP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 480(%rdi), %ymm7
; AVX2-FCP-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 320(%rdi), %ymm8
; AVX2-FCP-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 288(%rdi), %ymm9
; AVX2-FCP-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 128(%rdi), %ymm10
; AVX2-FCP-NEXT:    vmovups %ymm10, {{[-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:    vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vinsertf128 $1, 352(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm9[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vinsertf128 $1, 544(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm7[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vinsertf128 $1, 736(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm5[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vinsertf128 $1, 928(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = ymm3[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1088(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1056(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 1120(%rdi), %ymm0, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1280(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1248(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 1312(%rdi), %ymm0, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1472(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1440(%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 1504(%rdi), %ymm0, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 32(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps (%rdi), %ymm0
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 64(%rdi), %ymm0, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 224(%rdi), %ymm13
; AVX2-FCP-NEXT:    vmovaps 192(%rdi), %ymm12
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm12[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm13[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 256(%rdi), %ymm0, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 416(%rdi), %ymm11
; AVX2-FCP-NEXT:    vmovaps 384(%rdi), %ymm10
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm10[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm11[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 448(%rdi), %ymm0, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 608(%rdi), %ymm9
; AVX2-FCP-NEXT:    vmovaps 576(%rdi), %ymm8
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm8[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 640(%rdi), %ymm0, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 800(%rdi), %ymm7
; AVX2-FCP-NEXT:    vmovaps 768(%rdi), %ymm6
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm6[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 832(%rdi), %ymm0, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 992(%rdi), %ymm5
; AVX2-FCP-NEXT:    vmovaps 960(%rdi), %ymm4
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm4[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 1024(%rdi), %ymm0, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1184(%rdi), %ymm3
; AVX2-FCP-NEXT:    vmovaps 1152(%rdi), %ymm2
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm0 = ymm2[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 1216(%rdi), %ymm0, %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 1376(%rdi), %ymm1
; AVX2-FCP-NEXT:    vmovaps 1344(%rdi), %ymm0
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm15 = ymm0[0,3,2,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vinsertf128 $1, 1408(%rdi), %ymm0, %ymm14
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 176(%rdi), %ymm15
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 368(%rdi), %ymm15
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 560(%rdi), %ymm15
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 752(%rdi), %ymm15
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 944(%rdi), %ymm15
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 1136(%rdi), %ymm15
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, (%rsp), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 1328(%rdi), %ymm15
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 1520(%rdi), %ymm15
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FCP-NEXT:    vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX2-FCP-NEXT:    # ymm14 = ymm14[0,1],mem[2,3],ymm14[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm14 = ymm14[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 80(%rdi), %ymm15
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2,3],ymm13[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm12 = ymm12[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 272(%rdi), %ymm13
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm13[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,3],ymm11[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm10 = ymm10[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 464(%rdi), %ymm11
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm11[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3],ymm9[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 656(%rdi), %ymm9
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm8, (%rsp) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm6 = ymm6[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 848(%rdi), %ymm7
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 1040(%rdi), %ymm5
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FCP-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 1232(%rdi), %ymm3
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm13 = ymm2[0,1,2,3,4,5],ymm3[6,7]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
; AVX2-FCP-NEXT:    vbroadcastsd 1424(%rdi), %ymm1
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FCP-NEXT:    vmovaps 112(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 304(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 496(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FCP-NEXT:    vmovaps 688(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 880(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 1072(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 1264(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 1456(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 1360(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm7 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 1168(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 976(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 784(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 592(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 400(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 208(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm1 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FCP-NEXT:    vmovaps 16(%rdi), %xmm0
; AVX2-FCP-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
; AVX2-FCP-NEXT:    vpermpd {{.*#+}} ymm15 = mem[0,1,0,3]
; AVX2-FCP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 448(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 384(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 320(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 256(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 192(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 128(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 64(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, (%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 480(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 416(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 352(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 288(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 224(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 160(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 96(%rsi)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm15, 32(%rsi)
; AVX2-FCP-NEXT:    vmovaps %ymm11, 448(%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm13, 384(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, 320(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, 256(%rdx)
; AVX2-FCP-NEXT:    vmovups (%rsp), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, 192(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, 128(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, 64(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, (%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, 480(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, 416(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, 352(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, 288(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, 224(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, 160(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, 96(%rdx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm11, 32(%rdx)
; AVX2-FCP-NEXT:    vmovaps %ymm0, (%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm1, 64(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm2, 128(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm3, 192(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm4, 256(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm5, 320(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm6, 384(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm7, 448(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm8, 480(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm9, 416(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm10, 352(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm12, 288(%rcx)
; AVX2-FCP-NEXT:    vmovaps %ymm14, 224(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 160(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 96(%rcx)
; AVX2-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FCP-NEXT:    vmovaps %ymm0, 32(%rcx)
; AVX2-FCP-NEXT:    addq $968, %rsp # imm = 0x3C8
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i64_stride3_vf64:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa64 1472(%rdi), %zmm0
; AVX512-NEXT:    vmovdqa64 1408(%rdi), %zmm14
; AVX512-NEXT:    vmovdqa64 1344(%rdi), %zmm22
; AVX512-NEXT:    vmovdqa64 1280(%rdi), %zmm1
; AVX512-NEXT:    vmovdqa64 1216(%rdi), %zmm13
; AVX512-NEXT:    vmovdqa64 1152(%rdi), %zmm19
; AVX512-NEXT:    vmovdqa64 1088(%rdi), %zmm2
; AVX512-NEXT:    vmovdqa64 1024(%rdi), %zmm12
; AVX512-NEXT:    vmovdqa64 960(%rdi), %zmm27
; AVX512-NEXT:    vmovdqa64 896(%rdi), %zmm3
; AVX512-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512-NEXT:    vmovdqa64 768(%rdi), %zmm26
; AVX512-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512-NEXT:    vmovdqa64 640(%rdi), %zmm9
; AVX512-NEXT:    vmovdqa64 576(%rdi), %zmm29
; AVX512-NEXT:    vmovdqa64 512(%rdi), %zmm5
; AVX512-NEXT:    vmovdqa64 448(%rdi), %zmm8
; AVX512-NEXT:    vmovdqa64 384(%rdi), %zmm28
; AVX512-NEXT:    vmovdqa64 256(%rdi), %zmm7
; AVX512-NEXT:    vmovdqa64 (%rdi), %zmm25
; AVX512-NEXT:    vmovdqa64 192(%rdi), %zmm30
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,3,6,9,12,15,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm30, %zmm6
; AVX512-NEXT:    vpermt2q %zmm7, %zmm11, %zmm6
; AVX512-NEXT:    vmovdqa64 %zmm28, %zmm15
; AVX512-NEXT:    vpermt2q %zmm8, %zmm11, %zmm15
; AVX512-NEXT:    vmovdqa64 %zmm29, %zmm16
; AVX512-NEXT:    vpermt2q %zmm9, %zmm11, %zmm16
; AVX512-NEXT:    vmovdqa64 %zmm26, %zmm17
; AVX512-NEXT:    vpermt2q %zmm10, %zmm11, %zmm17
; AVX512-NEXT:    vmovdqa64 %zmm27, %zmm18
; AVX512-NEXT:    vpermt2q %zmm12, %zmm11, %zmm18
; AVX512-NEXT:    vmovdqa64 %zmm19, %zmm20
; AVX512-NEXT:    vpermt2q %zmm13, %zmm11, %zmm20
; AVX512-NEXT:    vmovdqa64 %zmm22, %zmm23
; AVX512-NEXT:    vpermt2q %zmm14, %zmm11, %zmm23
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm21 = [1,4,7,10,13,0,0,0]
; AVX512-NEXT:    vmovdqa64 %zmm30, %zmm24
; AVX512-NEXT:    vpermt2q %zmm7, %zmm21, %zmm24
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [10,13,0,3,6,0,0,0]
; AVX512-NEXT:    vpermt2q %zmm30, %zmm31, %zmm7
; AVX512-NEXT:    vmovdqa64 %zmm29, %zmm30
; AVX512-NEXT:    vpermt2q %zmm9, %zmm21, %zmm30
; AVX512-NEXT:    vpermt2q %zmm29, %zmm31, %zmm9
; AVX512-NEXT:    vmovdqa64 %zmm28, %zmm29
; AVX512-NEXT:    vpermt2q %zmm8, %zmm21, %zmm29
; AVX512-NEXT:    vpermt2q %zmm28, %zmm31, %zmm8
; AVX512-NEXT:    vmovdqa64 %zmm27, %zmm28
; AVX512-NEXT:    vpermt2q %zmm12, %zmm21, %zmm28
; AVX512-NEXT:    vpermt2q %zmm27, %zmm31, %zmm12
; AVX512-NEXT:    vmovdqa64 %zmm26, %zmm27
; AVX512-NEXT:    vpermt2q %zmm10, %zmm21, %zmm27
; AVX512-NEXT:    vpermt2q %zmm26, %zmm31, %zmm10
; AVX512-NEXT:    vmovdqa64 %zmm22, %zmm26
; AVX512-NEXT:    vpermt2q %zmm14, %zmm21, %zmm26
; AVX512-NEXT:    vpermt2q %zmm22, %zmm31, %zmm14
; AVX512-NEXT:    vmovdqa64 %zmm19, %zmm22
; AVX512-NEXT:    vpermt2q %zmm13, %zmm21, %zmm22
; AVX512-NEXT:    vpermt2q %zmm19, %zmm31, %zmm13
; AVX512-NEXT:    vmovdqa64 64(%rdi), %zmm19
; AVX512-NEXT:    vpermi2q %zmm19, %zmm25, %zmm11
; AVX512-NEXT:    vpermi2q %zmm19, %zmm25, %zmm21
; AVX512-NEXT:    vpermt2q %zmm25, %zmm31, %zmm19
; AVX512-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,5,10,13]
; AVX512-NEXT:    vpermt2q %zmm25, %zmm31, %zmm6
; AVX512-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-NEXT:    vpermt2q %zmm5, %zmm31, %zmm15
; AVX512-NEXT:    vpermt2q %zmm4, %zmm31, %zmm16
; AVX512-NEXT:    vpermt2q %zmm3, %zmm31, %zmm17
; AVX512-NEXT:    vpermt2q %zmm2, %zmm31, %zmm18
; AVX512-NEXT:    vpermt2q %zmm1, %zmm31, %zmm20
; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512-NEXT:    vmovdqa64 128(%rdi), %zmm6
; AVX512-NEXT:    vpermt2q %zmm6, %zmm31, %zmm11
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,8,11,14]
; AVX512-NEXT:    vpermt2q %zmm1, %zmm31, %zmm22
; AVX512-NEXT:    vpermt2q %zmm25, %zmm31, %zmm24
; AVX512-NEXT:    vpermt2q %zmm4, %zmm31, %zmm30
; AVX512-NEXT:    vpermt2q %zmm5, %zmm31, %zmm29
; AVX512-NEXT:    vpermt2q %zmm2, %zmm31, %zmm28
; AVX512-NEXT:    vpermt2q %zmm3, %zmm31, %zmm27
; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm26
; AVX512-NEXT:    vpermt2q %zmm6, %zmm31, %zmm21
; AVX512-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,9,12,15]
; AVX512-NEXT:    vpermt2q %zmm25, %zmm31, %zmm7
; AVX512-NEXT:    vpermt2q %zmm4, %zmm31, %zmm9
; AVX512-NEXT:    vpermt2q %zmm5, %zmm31, %zmm8
; AVX512-NEXT:    vpermt2q %zmm2, %zmm31, %zmm12
; AVX512-NEXT:    vpermt2q %zmm3, %zmm31, %zmm10
; AVX512-NEXT:    vpermt2q %zmm0, %zmm31, %zmm14
; AVX512-NEXT:    vpermt2q %zmm1, %zmm31, %zmm13
; AVX512-NEXT:    vpermt2q %zmm6, %zmm31, %zmm19
; AVX512-NEXT:    vmovdqa64 %zmm23, 448(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm20, 384(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm18, 320(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm17, 256(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-NEXT:    vmovaps %zmm0, 64(%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm11, (%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm26, 448(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm27, 256(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm28, 320(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm29, 128(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm30, 192(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm24, 64(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm22, 384(%rdx)
; AVX512-NEXT:    vmovdqa64 %zmm13, 384(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm14, 448(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm10, 256(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm12, 320(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm8, 128(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm9, 192(%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm19, (%rcx)
; AVX512-NEXT:    vmovdqa64 %zmm7, 64(%rcx)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i64_stride3_vf64:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovdqa64 1472(%rdi), %zmm0
; AVX512-FCP-NEXT:    vmovdqa64 1408(%rdi), %zmm14
; AVX512-FCP-NEXT:    vmovdqa64 1344(%rdi), %zmm22
; AVX512-FCP-NEXT:    vmovdqa64 1280(%rdi), %zmm1
; AVX512-FCP-NEXT:    vmovdqa64 1216(%rdi), %zmm13
; AVX512-FCP-NEXT:    vmovdqa64 1152(%rdi), %zmm19
; AVX512-FCP-NEXT:    vmovdqa64 1088(%rdi), %zmm2
; AVX512-FCP-NEXT:    vmovdqa64 1024(%rdi), %zmm12
; AVX512-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm27
; AVX512-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm3
; AVX512-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm26
; AVX512-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm9
; AVX512-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm29
; AVX512-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm5
; AVX512-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm8
; AVX512-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm28
; AVX512-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm7
; AVX512-FCP-NEXT:    vmovdqa64 (%rdi), %zmm25
; AVX512-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm30
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,3,6,9,12,15,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm30, %zmm6
; AVX512-FCP-NEXT:    vpermt2q %zmm7, %zmm11, %zmm6
; AVX512-FCP-NEXT:    vmovdqa64 %zmm28, %zmm15
; AVX512-FCP-NEXT:    vpermt2q %zmm8, %zmm11, %zmm15
; AVX512-FCP-NEXT:    vmovdqa64 %zmm29, %zmm16
; AVX512-FCP-NEXT:    vpermt2q %zmm9, %zmm11, %zmm16
; AVX512-FCP-NEXT:    vmovdqa64 %zmm26, %zmm17
; AVX512-FCP-NEXT:    vpermt2q %zmm10, %zmm11, %zmm17
; AVX512-FCP-NEXT:    vmovdqa64 %zmm27, %zmm18
; AVX512-FCP-NEXT:    vpermt2q %zmm12, %zmm11, %zmm18
; AVX512-FCP-NEXT:    vmovdqa64 %zmm19, %zmm20
; AVX512-FCP-NEXT:    vpermt2q %zmm13, %zmm11, %zmm20
; AVX512-FCP-NEXT:    vmovdqa64 %zmm22, %zmm23
; AVX512-FCP-NEXT:    vpermt2q %zmm14, %zmm11, %zmm23
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm21 = [1,4,7,10,13,0,0,0]
; AVX512-FCP-NEXT:    vmovdqa64 %zmm30, %zmm24
; AVX512-FCP-NEXT:    vpermt2q %zmm7, %zmm21, %zmm24
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [10,13,0,3,6,0,0,0]
; AVX512-FCP-NEXT:    vpermt2q %zmm30, %zmm31, %zmm7
; AVX512-FCP-NEXT:    vmovdqa64 %zmm29, %zmm30
; AVX512-FCP-NEXT:    vpermt2q %zmm9, %zmm21, %zmm30
; AVX512-FCP-NEXT:    vpermt2q %zmm29, %zmm31, %zmm9
; AVX512-FCP-NEXT:    vmovdqa64 %zmm28, %zmm29
; AVX512-FCP-NEXT:    vpermt2q %zmm8, %zmm21, %zmm29
; AVX512-FCP-NEXT:    vpermt2q %zmm28, %zmm31, %zmm8
; AVX512-FCP-NEXT:    vmovdqa64 %zmm27, %zmm28
; AVX512-FCP-NEXT:    vpermt2q %zmm12, %zmm21, %zmm28
; AVX512-FCP-NEXT:    vpermt2q %zmm27, %zmm31, %zmm12
; AVX512-FCP-NEXT:    vmovdqa64 %zmm26, %zmm27
; AVX512-FCP-NEXT:    vpermt2q %zmm10, %zmm21, %zmm27
; AVX512-FCP-NEXT:    vpermt2q %zmm26, %zmm31, %zmm10
; AVX512-FCP-NEXT:    vmovdqa64 %zmm22, %zmm26
; AVX512-FCP-NEXT:    vpermt2q %zmm14, %zmm21, %zmm26
; AVX512-FCP-NEXT:    vpermt2q %zmm22, %zmm31, %zmm14
; AVX512-FCP-NEXT:    vmovdqa64 %zmm19, %zmm22
; AVX512-FCP-NEXT:    vpermt2q %zmm13, %zmm21, %zmm22
; AVX512-FCP-NEXT:    vpermt2q %zmm19, %zmm31, %zmm13
; AVX512-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm19
; AVX512-FCP-NEXT:    vpermi2q %zmm19, %zmm25, %zmm11
; AVX512-FCP-NEXT:    vpermi2q %zmm19, %zmm25, %zmm21
; AVX512-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm19
; AVX512-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,5,10,13]
; AVX512-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm6
; AVX512-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512-FCP-NEXT:    vpermt2q %zmm5, %zmm31, %zmm15
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm31, %zmm16
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm31, %zmm17
; AVX512-FCP-NEXT:    vpermt2q %zmm2, %zmm31, %zmm18
; AVX512-FCP-NEXT:    vpermt2q %zmm1, %zmm31, %zmm20
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm6
; AVX512-FCP-NEXT:    vpermt2q %zmm6, %zmm31, %zmm11
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,8,11,14]
; AVX512-FCP-NEXT:    vpermt2q %zmm1, %zmm31, %zmm22
; AVX512-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm24
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm31, %zmm30
; AVX512-FCP-NEXT:    vpermt2q %zmm5, %zmm31, %zmm29
; AVX512-FCP-NEXT:    vpermt2q %zmm2, %zmm31, %zmm28
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm31, %zmm27
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm26
; AVX512-FCP-NEXT:    vpermt2q %zmm6, %zmm31, %zmm21
; AVX512-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,9,12,15]
; AVX512-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm7
; AVX512-FCP-NEXT:    vpermt2q %zmm4, %zmm31, %zmm9
; AVX512-FCP-NEXT:    vpermt2q %zmm5, %zmm31, %zmm8
; AVX512-FCP-NEXT:    vpermt2q %zmm2, %zmm31, %zmm12
; AVX512-FCP-NEXT:    vpermt2q %zmm3, %zmm31, %zmm10
; AVX512-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm14
; AVX512-FCP-NEXT:    vpermt2q %zmm1, %zmm31, %zmm13
; AVX512-FCP-NEXT:    vpermt2q %zmm6, %zmm31, %zmm19
; AVX512-FCP-NEXT:    vmovdqa64 %zmm23, 448(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm20, 384(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm18, 320(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm17, 256(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512-FCP-NEXT:    vmovaps %zmm0, 64(%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm11, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm26, 448(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm27, 256(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm28, 320(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm29, 128(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm30, 192(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm24, 64(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm22, 384(%rdx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm13, 384(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm14, 448(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm10, 256(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm12, 320(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm8, 128(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm9, 192(%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm19, (%rcx)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm7, 64(%rcx)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i64_stride3_vf64:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa64 1472(%rdi), %zmm0
; AVX512DQ-NEXT:    vmovdqa64 1408(%rdi), %zmm14
; AVX512DQ-NEXT:    vmovdqa64 1344(%rdi), %zmm22
; AVX512DQ-NEXT:    vmovdqa64 1280(%rdi), %zmm1
; AVX512DQ-NEXT:    vmovdqa64 1216(%rdi), %zmm13
; AVX512DQ-NEXT:    vmovdqa64 1152(%rdi), %zmm19
; AVX512DQ-NEXT:    vmovdqa64 1088(%rdi), %zmm2
; AVX512DQ-NEXT:    vmovdqa64 1024(%rdi), %zmm12
; AVX512DQ-NEXT:    vmovdqa64 960(%rdi), %zmm27
; AVX512DQ-NEXT:    vmovdqa64 896(%rdi), %zmm3
; AVX512DQ-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512DQ-NEXT:    vmovdqa64 768(%rdi), %zmm26
; AVX512DQ-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512DQ-NEXT:    vmovdqa64 640(%rdi), %zmm9
; AVX512DQ-NEXT:    vmovdqa64 576(%rdi), %zmm29
; AVX512DQ-NEXT:    vmovdqa64 512(%rdi), %zmm5
; AVX512DQ-NEXT:    vmovdqa64 448(%rdi), %zmm8
; AVX512DQ-NEXT:    vmovdqa64 384(%rdi), %zmm28
; AVX512DQ-NEXT:    vmovdqa64 256(%rdi), %zmm7
; AVX512DQ-NEXT:    vmovdqa64 (%rdi), %zmm25
; AVX512DQ-NEXT:    vmovdqa64 192(%rdi), %zmm30
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,3,6,9,12,15,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm30, %zmm6
; AVX512DQ-NEXT:    vpermt2q %zmm7, %zmm11, %zmm6
; AVX512DQ-NEXT:    vmovdqa64 %zmm28, %zmm15
; AVX512DQ-NEXT:    vpermt2q %zmm8, %zmm11, %zmm15
; AVX512DQ-NEXT:    vmovdqa64 %zmm29, %zmm16
; AVX512DQ-NEXT:    vpermt2q %zmm9, %zmm11, %zmm16
; AVX512DQ-NEXT:    vmovdqa64 %zmm26, %zmm17
; AVX512DQ-NEXT:    vpermt2q %zmm10, %zmm11, %zmm17
; AVX512DQ-NEXT:    vmovdqa64 %zmm27, %zmm18
; AVX512DQ-NEXT:    vpermt2q %zmm12, %zmm11, %zmm18
; AVX512DQ-NEXT:    vmovdqa64 %zmm19, %zmm20
; AVX512DQ-NEXT:    vpermt2q %zmm13, %zmm11, %zmm20
; AVX512DQ-NEXT:    vmovdqa64 %zmm22, %zmm23
; AVX512DQ-NEXT:    vpermt2q %zmm14, %zmm11, %zmm23
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm21 = [1,4,7,10,13,0,0,0]
; AVX512DQ-NEXT:    vmovdqa64 %zmm30, %zmm24
; AVX512DQ-NEXT:    vpermt2q %zmm7, %zmm21, %zmm24
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [10,13,0,3,6,0,0,0]
; AVX512DQ-NEXT:    vpermt2q %zmm30, %zmm31, %zmm7
; AVX512DQ-NEXT:    vmovdqa64 %zmm29, %zmm30
; AVX512DQ-NEXT:    vpermt2q %zmm9, %zmm21, %zmm30
; AVX512DQ-NEXT:    vpermt2q %zmm29, %zmm31, %zmm9
; AVX512DQ-NEXT:    vmovdqa64 %zmm28, %zmm29
; AVX512DQ-NEXT:    vpermt2q %zmm8, %zmm21, %zmm29
; AVX512DQ-NEXT:    vpermt2q %zmm28, %zmm31, %zmm8
; AVX512DQ-NEXT:    vmovdqa64 %zmm27, %zmm28
; AVX512DQ-NEXT:    vpermt2q %zmm12, %zmm21, %zmm28
; AVX512DQ-NEXT:    vpermt2q %zmm27, %zmm31, %zmm12
; AVX512DQ-NEXT:    vmovdqa64 %zmm26, %zmm27
; AVX512DQ-NEXT:    vpermt2q %zmm10, %zmm21, %zmm27
; AVX512DQ-NEXT:    vpermt2q %zmm26, %zmm31, %zmm10
; AVX512DQ-NEXT:    vmovdqa64 %zmm22, %zmm26
; AVX512DQ-NEXT:    vpermt2q %zmm14, %zmm21, %zmm26
; AVX512DQ-NEXT:    vpermt2q %zmm22, %zmm31, %zmm14
; AVX512DQ-NEXT:    vmovdqa64 %zmm19, %zmm22
; AVX512DQ-NEXT:    vpermt2q %zmm13, %zmm21, %zmm22
; AVX512DQ-NEXT:    vpermt2q %zmm19, %zmm31, %zmm13
; AVX512DQ-NEXT:    vmovdqa64 64(%rdi), %zmm19
; AVX512DQ-NEXT:    vpermi2q %zmm19, %zmm25, %zmm11
; AVX512DQ-NEXT:    vpermi2q %zmm19, %zmm25, %zmm21
; AVX512DQ-NEXT:    vpermt2q %zmm25, %zmm31, %zmm19
; AVX512DQ-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,5,10,13]
; AVX512DQ-NEXT:    vpermt2q %zmm25, %zmm31, %zmm6
; AVX512DQ-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-NEXT:    vpermt2q %zmm5, %zmm31, %zmm15
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm31, %zmm16
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm31, %zmm17
; AVX512DQ-NEXT:    vpermt2q %zmm2, %zmm31, %zmm18
; AVX512DQ-NEXT:    vpermt2q %zmm1, %zmm31, %zmm20
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512DQ-NEXT:    vmovdqa64 128(%rdi), %zmm6
; AVX512DQ-NEXT:    vpermt2q %zmm6, %zmm31, %zmm11
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,8,11,14]
; AVX512DQ-NEXT:    vpermt2q %zmm1, %zmm31, %zmm22
; AVX512DQ-NEXT:    vpermt2q %zmm25, %zmm31, %zmm24
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm31, %zmm30
; AVX512DQ-NEXT:    vpermt2q %zmm5, %zmm31, %zmm29
; AVX512DQ-NEXT:    vpermt2q %zmm2, %zmm31, %zmm28
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm31, %zmm27
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm31, %zmm26
; AVX512DQ-NEXT:    vpermt2q %zmm6, %zmm31, %zmm21
; AVX512DQ-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,9,12,15]
; AVX512DQ-NEXT:    vpermt2q %zmm25, %zmm31, %zmm7
; AVX512DQ-NEXT:    vpermt2q %zmm4, %zmm31, %zmm9
; AVX512DQ-NEXT:    vpermt2q %zmm5, %zmm31, %zmm8
; AVX512DQ-NEXT:    vpermt2q %zmm2, %zmm31, %zmm12
; AVX512DQ-NEXT:    vpermt2q %zmm3, %zmm31, %zmm10
; AVX512DQ-NEXT:    vpermt2q %zmm0, %zmm31, %zmm14
; AVX512DQ-NEXT:    vpermt2q %zmm1, %zmm31, %zmm13
; AVX512DQ-NEXT:    vpermt2q %zmm6, %zmm31, %zmm19
; AVX512DQ-NEXT:    vmovdqa64 %zmm23, 448(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm20, 384(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm18, 320(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm17, 256(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512DQ-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-NEXT:    vmovaps %zmm0, 64(%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm11, (%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm26, 448(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm27, 256(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm28, 320(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm29, 128(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm30, 192(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm24, 64(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm22, 384(%rdx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm13, 384(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm14, 448(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm10, 256(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm12, 320(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm8, 128(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm9, 192(%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm19, (%rcx)
; AVX512DQ-NEXT:    vmovdqa64 %zmm7, 64(%rcx)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i64_stride3_vf64:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovdqa64 1472(%rdi), %zmm0
; AVX512DQ-FCP-NEXT:    vmovdqa64 1408(%rdi), %zmm14
; AVX512DQ-FCP-NEXT:    vmovdqa64 1344(%rdi), %zmm22
; AVX512DQ-FCP-NEXT:    vmovdqa64 1280(%rdi), %zmm1
; AVX512DQ-FCP-NEXT:    vmovdqa64 1216(%rdi), %zmm13
; AVX512DQ-FCP-NEXT:    vmovdqa64 1152(%rdi), %zmm19
; AVX512DQ-FCP-NEXT:    vmovdqa64 1088(%rdi), %zmm2
; AVX512DQ-FCP-NEXT:    vmovdqa64 1024(%rdi), %zmm12
; AVX512DQ-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm27
; AVX512DQ-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm3
; AVX512DQ-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512DQ-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm26
; AVX512DQ-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512DQ-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm9
; AVX512DQ-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm29
; AVX512DQ-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm5
; AVX512DQ-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm8
; AVX512DQ-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm28
; AVX512DQ-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqa64 (%rdi), %zmm25
; AVX512DQ-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm30
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,3,6,9,12,15,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm30, %zmm6
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm7, %zmm11, %zmm6
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm28, %zmm15
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm8, %zmm11, %zmm15
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm29, %zmm16
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm9, %zmm11, %zmm16
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm26, %zmm17
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm10, %zmm11, %zmm17
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm27, %zmm18
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm12, %zmm11, %zmm18
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm19, %zmm20
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm13, %zmm11, %zmm20
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm22, %zmm23
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm14, %zmm11, %zmm23
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm21 = [1,4,7,10,13,0,0,0]
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm30, %zmm24
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm7, %zmm21, %zmm24
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [10,13,0,3,6,0,0,0]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm30, %zmm31, %zmm7
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm29, %zmm30
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm9, %zmm21, %zmm30
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm29, %zmm31, %zmm9
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm28, %zmm29
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm8, %zmm21, %zmm29
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm28, %zmm31, %zmm8
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm27, %zmm28
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm12, %zmm21, %zmm28
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm27, %zmm31, %zmm12
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm26, %zmm27
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm10, %zmm21, %zmm27
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm26, %zmm31, %zmm10
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm22, %zmm26
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm14, %zmm21, %zmm26
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm22, %zmm31, %zmm14
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm19, %zmm22
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm13, %zmm21, %zmm22
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm19, %zmm31, %zmm13
; AVX512DQ-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm19
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm19, %zmm25, %zmm11
; AVX512DQ-FCP-NEXT:    vpermi2q %zmm19, %zmm25, %zmm21
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm19
; AVX512DQ-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,5,10,13]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm6
; AVX512DQ-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm5, %zmm31, %zmm15
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm31, %zmm16
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm31, %zmm17
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm2, %zmm31, %zmm18
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm1, %zmm31, %zmm20
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512DQ-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm6
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm6, %zmm31, %zmm11
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,8,11,14]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm1, %zmm31, %zmm22
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm24
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm31, %zmm30
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm5, %zmm31, %zmm29
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm2, %zmm31, %zmm28
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm31, %zmm27
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm26
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm6, %zmm31, %zmm21
; AVX512DQ-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,9,12,15]
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm7
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm4, %zmm31, %zmm9
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm5, %zmm31, %zmm8
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm2, %zmm31, %zmm12
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm3, %zmm31, %zmm10
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm14
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm1, %zmm31, %zmm13
; AVX512DQ-FCP-NEXT:    vpermt2q %zmm6, %zmm31, %zmm19
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm23, 448(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm20, 384(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm18, 320(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm17, 256(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512DQ-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-FCP-NEXT:    vmovaps %zmm0, 64(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm11, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm26, 448(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm27, 256(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm28, 320(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm29, 128(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm30, 192(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm24, 64(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm22, 384(%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm13, 384(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm14, 448(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm10, 256(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm12, 320(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm8, 128(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm9, 192(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm19, (%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm7, 64(%rcx)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i64_stride3_vf64:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa64 1472(%rdi), %zmm0
; AVX512BW-NEXT:    vmovdqa64 1408(%rdi), %zmm14
; AVX512BW-NEXT:    vmovdqa64 1344(%rdi), %zmm22
; AVX512BW-NEXT:    vmovdqa64 1280(%rdi), %zmm1
; AVX512BW-NEXT:    vmovdqa64 1216(%rdi), %zmm13
; AVX512BW-NEXT:    vmovdqa64 1152(%rdi), %zmm19
; AVX512BW-NEXT:    vmovdqa64 1088(%rdi), %zmm2
; AVX512BW-NEXT:    vmovdqa64 1024(%rdi), %zmm12
; AVX512BW-NEXT:    vmovdqa64 960(%rdi), %zmm27
; AVX512BW-NEXT:    vmovdqa64 896(%rdi), %zmm3
; AVX512BW-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512BW-NEXT:    vmovdqa64 768(%rdi), %zmm26
; AVX512BW-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512BW-NEXT:    vmovdqa64 640(%rdi), %zmm9
; AVX512BW-NEXT:    vmovdqa64 576(%rdi), %zmm29
; AVX512BW-NEXT:    vmovdqa64 512(%rdi), %zmm5
; AVX512BW-NEXT:    vmovdqa64 448(%rdi), %zmm8
; AVX512BW-NEXT:    vmovdqa64 384(%rdi), %zmm28
; AVX512BW-NEXT:    vmovdqa64 256(%rdi), %zmm7
; AVX512BW-NEXT:    vmovdqa64 (%rdi), %zmm25
; AVX512BW-NEXT:    vmovdqa64 192(%rdi), %zmm30
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,3,6,9,12,15,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm6
; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm11, %zmm6
; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm15
; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm11, %zmm15
; AVX512BW-NEXT:    vmovdqa64 %zmm29, %zmm16
; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm11, %zmm16
; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm17
; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm11, %zmm17
; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm18
; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm11, %zmm18
; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm20
; AVX512BW-NEXT:    vpermt2q %zmm13, %zmm11, %zmm20
; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm23
; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm11, %zmm23
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm21 = [1,4,7,10,13,0,0,0]
; AVX512BW-NEXT:    vmovdqa64 %zmm30, %zmm24
; AVX512BW-NEXT:    vpermt2q %zmm7, %zmm21, %zmm24
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [10,13,0,3,6,0,0,0]
; AVX512BW-NEXT:    vpermt2q %zmm30, %zmm31, %zmm7
; AVX512BW-NEXT:    vmovdqa64 %zmm29, %zmm30
; AVX512BW-NEXT:    vpermt2q %zmm9, %zmm21, %zmm30
; AVX512BW-NEXT:    vpermt2q %zmm29, %zmm31, %zmm9
; AVX512BW-NEXT:    vmovdqa64 %zmm28, %zmm29
; AVX512BW-NEXT:    vpermt2q %zmm8, %zmm21, %zmm29
; AVX512BW-NEXT:    vpermt2q %zmm28, %zmm31, %zmm8
; AVX512BW-NEXT:    vmovdqa64 %zmm27, %zmm28
; AVX512BW-NEXT:    vpermt2q %zmm12, %zmm21, %zmm28
; AVX512BW-NEXT:    vpermt2q %zmm27, %zmm31, %zmm12
; AVX512BW-NEXT:    vmovdqa64 %zmm26, %zmm27
; AVX512BW-NEXT:    vpermt2q %zmm10, %zmm21, %zmm27
; AVX512BW-NEXT:    vpermt2q %zmm26, %zmm31, %zmm10
; AVX512BW-NEXT:    vmovdqa64 %zmm22, %zmm26
; AVX512BW-NEXT:    vpermt2q %zmm14, %zmm21, %zmm26
; AVX512BW-NEXT:    vpermt2q %zmm22, %zmm31, %zmm14
; AVX512BW-NEXT:    vmovdqa64 %zmm19, %zmm22
; AVX512BW-NEXT:    vpermt2q %zmm13, %zmm21, %zmm22
; AVX512BW-NEXT:    vpermt2q %zmm19, %zmm31, %zmm13
; AVX512BW-NEXT:    vmovdqa64 64(%rdi), %zmm19
; AVX512BW-NEXT:    vpermi2q %zmm19, %zmm25, %zmm11
; AVX512BW-NEXT:    vpermi2q %zmm19, %zmm25, %zmm21
; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm31, %zmm19
; AVX512BW-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,5,10,13]
; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm31, %zmm6
; AVX512BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm31, %zmm15
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm31, %zmm16
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm31, %zmm17
; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm31, %zmm18
; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm31, %zmm20
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512BW-NEXT:    vmovdqa64 128(%rdi), %zmm6
; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm31, %zmm11
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,8,11,14]
; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm31, %zmm22
; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm31, %zmm24
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm31, %zmm30
; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm31, %zmm29
; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm31, %zmm28
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm31, %zmm27
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm26
; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm31, %zmm21
; AVX512BW-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,9,12,15]
; AVX512BW-NEXT:    vpermt2q %zmm25, %zmm31, %zmm7
; AVX512BW-NEXT:    vpermt2q %zmm4, %zmm31, %zmm9
; AVX512BW-NEXT:    vpermt2q %zmm5, %zmm31, %zmm8
; AVX512BW-NEXT:    vpermt2q %zmm2, %zmm31, %zmm12
; AVX512BW-NEXT:    vpermt2q %zmm3, %zmm31, %zmm10
; AVX512BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm14
; AVX512BW-NEXT:    vpermt2q %zmm1, %zmm31, %zmm13
; AVX512BW-NEXT:    vpermt2q %zmm6, %zmm31, %zmm19
; AVX512BW-NEXT:    vmovdqa64 %zmm23, 448(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm20, 384(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm18, 320(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm17, 256(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-NEXT:    vmovaps %zmm0, 64(%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm11, (%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm26, 448(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm27, 256(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm28, 320(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm29, 128(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm30, 192(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm24, 64(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm22, 384(%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm13, 384(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm14, 448(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm10, 256(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm12, 320(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm8, 128(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm9, 192(%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm19, (%rcx)
; AVX512BW-NEXT:    vmovdqa64 %zmm7, 64(%rcx)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i64_stride3_vf64:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovdqa64 1472(%rdi), %zmm0
; AVX512BW-FCP-NEXT:    vmovdqa64 1408(%rdi), %zmm14
; AVX512BW-FCP-NEXT:    vmovdqa64 1344(%rdi), %zmm22
; AVX512BW-FCP-NEXT:    vmovdqa64 1280(%rdi), %zmm1
; AVX512BW-FCP-NEXT:    vmovdqa64 1216(%rdi), %zmm13
; AVX512BW-FCP-NEXT:    vmovdqa64 1152(%rdi), %zmm19
; AVX512BW-FCP-NEXT:    vmovdqa64 1088(%rdi), %zmm2
; AVX512BW-FCP-NEXT:    vmovdqa64 1024(%rdi), %zmm12
; AVX512BW-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm27
; AVX512BW-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm3
; AVX512BW-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512BW-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm26
; AVX512BW-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512BW-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm9
; AVX512BW-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm29
; AVX512BW-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm5
; AVX512BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm8
; AVX512BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm28
; AVX512BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm7
; AVX512BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm25
; AVX512BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm30
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,3,6,9,12,15,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm30, %zmm6
; AVX512BW-FCP-NEXT:    vpermt2q %zmm7, %zmm11, %zmm6
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm28, %zmm15
; AVX512BW-FCP-NEXT:    vpermt2q %zmm8, %zmm11, %zmm15
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm29, %zmm16
; AVX512BW-FCP-NEXT:    vpermt2q %zmm9, %zmm11, %zmm16
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm26, %zmm17
; AVX512BW-FCP-NEXT:    vpermt2q %zmm10, %zmm11, %zmm17
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm27, %zmm18
; AVX512BW-FCP-NEXT:    vpermt2q %zmm12, %zmm11, %zmm18
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm19, %zmm20
; AVX512BW-FCP-NEXT:    vpermt2q %zmm13, %zmm11, %zmm20
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm22, %zmm23
; AVX512BW-FCP-NEXT:    vpermt2q %zmm14, %zmm11, %zmm23
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm21 = [1,4,7,10,13,0,0,0]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm30, %zmm24
; AVX512BW-FCP-NEXT:    vpermt2q %zmm7, %zmm21, %zmm24
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [10,13,0,3,6,0,0,0]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm30, %zmm31, %zmm7
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm29, %zmm30
; AVX512BW-FCP-NEXT:    vpermt2q %zmm9, %zmm21, %zmm30
; AVX512BW-FCP-NEXT:    vpermt2q %zmm29, %zmm31, %zmm9
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm28, %zmm29
; AVX512BW-FCP-NEXT:    vpermt2q %zmm8, %zmm21, %zmm29
; AVX512BW-FCP-NEXT:    vpermt2q %zmm28, %zmm31, %zmm8
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm27, %zmm28
; AVX512BW-FCP-NEXT:    vpermt2q %zmm12, %zmm21, %zmm28
; AVX512BW-FCP-NEXT:    vpermt2q %zmm27, %zmm31, %zmm12
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm26, %zmm27
; AVX512BW-FCP-NEXT:    vpermt2q %zmm10, %zmm21, %zmm27
; AVX512BW-FCP-NEXT:    vpermt2q %zmm26, %zmm31, %zmm10
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm22, %zmm26
; AVX512BW-FCP-NEXT:    vpermt2q %zmm14, %zmm21, %zmm26
; AVX512BW-FCP-NEXT:    vpermt2q %zmm22, %zmm31, %zmm14
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm19, %zmm22
; AVX512BW-FCP-NEXT:    vpermt2q %zmm13, %zmm21, %zmm22
; AVX512BW-FCP-NEXT:    vpermt2q %zmm19, %zmm31, %zmm13
; AVX512BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm19
; AVX512BW-FCP-NEXT:    vpermi2q %zmm19, %zmm25, %zmm11
; AVX512BW-FCP-NEXT:    vpermi2q %zmm19, %zmm25, %zmm21
; AVX512BW-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm19
; AVX512BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,5,10,13]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm6
; AVX512BW-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512BW-FCP-NEXT:    vpermt2q %zmm5, %zmm31, %zmm15
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm31, %zmm16
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm31, %zmm17
; AVX512BW-FCP-NEXT:    vpermt2q %zmm2, %zmm31, %zmm18
; AVX512BW-FCP-NEXT:    vpermt2q %zmm1, %zmm31, %zmm20
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm6
; AVX512BW-FCP-NEXT:    vpermt2q %zmm6, %zmm31, %zmm11
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,8,11,14]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm1, %zmm31, %zmm22
; AVX512BW-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm24
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm31, %zmm30
; AVX512BW-FCP-NEXT:    vpermt2q %zmm5, %zmm31, %zmm29
; AVX512BW-FCP-NEXT:    vpermt2q %zmm2, %zmm31, %zmm28
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm31, %zmm27
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm26
; AVX512BW-FCP-NEXT:    vpermt2q %zmm6, %zmm31, %zmm21
; AVX512BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,9,12,15]
; AVX512BW-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm7
; AVX512BW-FCP-NEXT:    vpermt2q %zmm4, %zmm31, %zmm9
; AVX512BW-FCP-NEXT:    vpermt2q %zmm5, %zmm31, %zmm8
; AVX512BW-FCP-NEXT:    vpermt2q %zmm2, %zmm31, %zmm12
; AVX512BW-FCP-NEXT:    vpermt2q %zmm3, %zmm31, %zmm10
; AVX512BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm14
; AVX512BW-FCP-NEXT:    vpermt2q %zmm1, %zmm31, %zmm13
; AVX512BW-FCP-NEXT:    vpermt2q %zmm6, %zmm31, %zmm19
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm23, 448(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm20, 384(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm18, 320(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm17, 256(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512BW-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512BW-FCP-NEXT:    vmovaps %zmm0, 64(%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm11, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm26, 448(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm27, 256(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm28, 320(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm29, 128(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm30, 192(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm24, 64(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm22, 384(%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm13, 384(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm14, 448(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm10, 256(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm12, 320(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm8, 128(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm9, 192(%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm19, (%rcx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm7, 64(%rcx)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i64_stride3_vf64:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovdqa64 1472(%rdi), %zmm0
; AVX512DQ-BW-NEXT:    vmovdqa64 1408(%rdi), %zmm14
; AVX512DQ-BW-NEXT:    vmovdqa64 1344(%rdi), %zmm22
; AVX512DQ-BW-NEXT:    vmovdqa64 1280(%rdi), %zmm1
; AVX512DQ-BW-NEXT:    vmovdqa64 1216(%rdi), %zmm13
; AVX512DQ-BW-NEXT:    vmovdqa64 1152(%rdi), %zmm19
; AVX512DQ-BW-NEXT:    vmovdqa64 1088(%rdi), %zmm2
; AVX512DQ-BW-NEXT:    vmovdqa64 1024(%rdi), %zmm12
; AVX512DQ-BW-NEXT:    vmovdqa64 960(%rdi), %zmm27
; AVX512DQ-BW-NEXT:    vmovdqa64 896(%rdi), %zmm3
; AVX512DQ-BW-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512DQ-BW-NEXT:    vmovdqa64 768(%rdi), %zmm26
; AVX512DQ-BW-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512DQ-BW-NEXT:    vmovdqa64 640(%rdi), %zmm9
; AVX512DQ-BW-NEXT:    vmovdqa64 576(%rdi), %zmm29
; AVX512DQ-BW-NEXT:    vmovdqa64 512(%rdi), %zmm5
; AVX512DQ-BW-NEXT:    vmovdqa64 448(%rdi), %zmm8
; AVX512DQ-BW-NEXT:    vmovdqa64 384(%rdi), %zmm28
; AVX512DQ-BW-NEXT:    vmovdqa64 256(%rdi), %zmm7
; AVX512DQ-BW-NEXT:    vmovdqa64 (%rdi), %zmm25
; AVX512DQ-BW-NEXT:    vmovdqa64 192(%rdi), %zmm30
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,3,6,9,12,15,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm30, %zmm6
; AVX512DQ-BW-NEXT:    vpermt2q %zmm7, %zmm11, %zmm6
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm28, %zmm15
; AVX512DQ-BW-NEXT:    vpermt2q %zmm8, %zmm11, %zmm15
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm29, %zmm16
; AVX512DQ-BW-NEXT:    vpermt2q %zmm9, %zmm11, %zmm16
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm26, %zmm17
; AVX512DQ-BW-NEXT:    vpermt2q %zmm10, %zmm11, %zmm17
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm27, %zmm18
; AVX512DQ-BW-NEXT:    vpermt2q %zmm12, %zmm11, %zmm18
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm19, %zmm20
; AVX512DQ-BW-NEXT:    vpermt2q %zmm13, %zmm11, %zmm20
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm22, %zmm23
; AVX512DQ-BW-NEXT:    vpermt2q %zmm14, %zmm11, %zmm23
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm21 = [1,4,7,10,13,0,0,0]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm30, %zmm24
; AVX512DQ-BW-NEXT:    vpermt2q %zmm7, %zmm21, %zmm24
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [10,13,0,3,6,0,0,0]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm30, %zmm31, %zmm7
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm29, %zmm30
; AVX512DQ-BW-NEXT:    vpermt2q %zmm9, %zmm21, %zmm30
; AVX512DQ-BW-NEXT:    vpermt2q %zmm29, %zmm31, %zmm9
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm28, %zmm29
; AVX512DQ-BW-NEXT:    vpermt2q %zmm8, %zmm21, %zmm29
; AVX512DQ-BW-NEXT:    vpermt2q %zmm28, %zmm31, %zmm8
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm27, %zmm28
; AVX512DQ-BW-NEXT:    vpermt2q %zmm12, %zmm21, %zmm28
; AVX512DQ-BW-NEXT:    vpermt2q %zmm27, %zmm31, %zmm12
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm26, %zmm27
; AVX512DQ-BW-NEXT:    vpermt2q %zmm10, %zmm21, %zmm27
; AVX512DQ-BW-NEXT:    vpermt2q %zmm26, %zmm31, %zmm10
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm22, %zmm26
; AVX512DQ-BW-NEXT:    vpermt2q %zmm14, %zmm21, %zmm26
; AVX512DQ-BW-NEXT:    vpermt2q %zmm22, %zmm31, %zmm14
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm19, %zmm22
; AVX512DQ-BW-NEXT:    vpermt2q %zmm13, %zmm21, %zmm22
; AVX512DQ-BW-NEXT:    vpermt2q %zmm19, %zmm31, %zmm13
; AVX512DQ-BW-NEXT:    vmovdqa64 64(%rdi), %zmm19
; AVX512DQ-BW-NEXT:    vpermi2q %zmm19, %zmm25, %zmm11
; AVX512DQ-BW-NEXT:    vpermi2q %zmm19, %zmm25, %zmm21
; AVX512DQ-BW-NEXT:    vpermt2q %zmm25, %zmm31, %zmm19
; AVX512DQ-BW-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,5,10,13]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm25, %zmm31, %zmm6
; AVX512DQ-BW-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-NEXT:    vpermt2q %zmm5, %zmm31, %zmm15
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm31, %zmm16
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm31, %zmm17
; AVX512DQ-BW-NEXT:    vpermt2q %zmm2, %zmm31, %zmm18
; AVX512DQ-BW-NEXT:    vpermt2q %zmm1, %zmm31, %zmm20
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512DQ-BW-NEXT:    vmovdqa64 128(%rdi), %zmm6
; AVX512DQ-BW-NEXT:    vpermt2q %zmm6, %zmm31, %zmm11
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,8,11,14]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm1, %zmm31, %zmm22
; AVX512DQ-BW-NEXT:    vpermt2q %zmm25, %zmm31, %zmm24
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm31, %zmm30
; AVX512DQ-BW-NEXT:    vpermt2q %zmm5, %zmm31, %zmm29
; AVX512DQ-BW-NEXT:    vpermt2q %zmm2, %zmm31, %zmm28
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm31, %zmm27
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm26
; AVX512DQ-BW-NEXT:    vpermt2q %zmm6, %zmm31, %zmm21
; AVX512DQ-BW-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,9,12,15]
; AVX512DQ-BW-NEXT:    vpermt2q %zmm25, %zmm31, %zmm7
; AVX512DQ-BW-NEXT:    vpermt2q %zmm4, %zmm31, %zmm9
; AVX512DQ-BW-NEXT:    vpermt2q %zmm5, %zmm31, %zmm8
; AVX512DQ-BW-NEXT:    vpermt2q %zmm2, %zmm31, %zmm12
; AVX512DQ-BW-NEXT:    vpermt2q %zmm3, %zmm31, %zmm10
; AVX512DQ-BW-NEXT:    vpermt2q %zmm0, %zmm31, %zmm14
; AVX512DQ-BW-NEXT:    vpermt2q %zmm1, %zmm31, %zmm13
; AVX512DQ-BW-NEXT:    vpermt2q %zmm6, %zmm31, %zmm19
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm23, 448(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm20, 384(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm18, 320(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm17, 256(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512DQ-BW-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-NEXT:    vmovaps %zmm0, 64(%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm11, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm26, 448(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm27, 256(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm28, 320(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm29, 128(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm30, 192(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm24, 64(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm22, 384(%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm13, 384(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm14, 448(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm10, 256(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm12, 320(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm8, 128(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm9, 192(%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm19, (%rcx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm7, 64(%rcx)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i64_stride3_vf64:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1472(%rdi), %zmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1408(%rdi), %zmm14
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1344(%rdi), %zmm22
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1280(%rdi), %zmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1216(%rdi), %zmm13
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1152(%rdi), %zmm19
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1088(%rdi), %zmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 1024(%rdi), %zmm12
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 960(%rdi), %zmm27
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 896(%rdi), %zmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 832(%rdi), %zmm10
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 768(%rdi), %zmm26
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 704(%rdi), %zmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 640(%rdi), %zmm9
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 576(%rdi), %zmm29
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 512(%rdi), %zmm5
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 448(%rdi), %zmm8
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 384(%rdi), %zmm28
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 256(%rdi), %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 (%rdi), %zmm25
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 192(%rdi), %zmm30
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm11 = [0,3,6,9,12,15,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm30, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm7, %zmm11, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm28, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm8, %zmm11, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm29, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm9, %zmm11, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm26, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm10, %zmm11, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm27, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm12, %zmm11, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm19, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm13, %zmm11, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm22, %zmm23
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm14, %zmm11, %zmm23
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm21 = [1,4,7,10,13,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm30, %zmm24
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm7, %zmm21, %zmm24
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [10,13,0,3,6,0,0,0]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm30, %zmm31, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm29, %zmm30
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm9, %zmm21, %zmm30
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm29, %zmm31, %zmm9
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm28, %zmm29
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm8, %zmm21, %zmm29
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm28, %zmm31, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm27, %zmm28
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm12, %zmm21, %zmm28
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm27, %zmm31, %zmm12
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm26, %zmm27
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm10, %zmm21, %zmm27
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm26, %zmm31, %zmm10
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm22, %zmm26
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm14, %zmm21, %zmm26
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm22, %zmm31, %zmm14
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm19, %zmm22
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm13, %zmm21, %zmm22
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm19, %zmm31, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 64(%rdi), %zmm19
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm19, %zmm25, %zmm11
; AVX512DQ-BW-FCP-NEXT:    vpermi2q %zmm19, %zmm25, %zmm21
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm19
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 320(%rdi), %zmm25
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,5,10,13]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm6
; AVX512DQ-BW-FCP-NEXT:    vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm5, %zmm31, %zmm15
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm31, %zmm16
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm31, %zmm17
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm2, %zmm31, %zmm18
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm1, %zmm31, %zmm20
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm23
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 128(%rdi), %zmm6
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm6, %zmm31, %zmm11
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,8,11,14]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm1, %zmm31, %zmm22
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm24
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm31, %zmm30
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm5, %zmm31, %zmm29
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm2, %zmm31, %zmm28
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm31, %zmm27
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm26
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm6, %zmm31, %zmm21
; AVX512DQ-BW-FCP-NEXT:    vpmovsxbq {{.*#+}} zmm31 = [0,1,2,3,4,9,12,15]
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm25, %zmm31, %zmm7
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm4, %zmm31, %zmm9
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm5, %zmm31, %zmm8
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm2, %zmm31, %zmm12
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm3, %zmm31, %zmm10
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm0, %zmm31, %zmm14
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm1, %zmm31, %zmm13
; AVX512DQ-BW-FCP-NEXT:    vpermt2q %zmm6, %zmm31, %zmm19
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm23, 448(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm20, 384(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm18, 320(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm17, 256(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm16, 192(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm15, 128(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
; AVX512DQ-BW-FCP-NEXT:    vmovaps %zmm0, 64(%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm11, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm26, 448(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm27, 256(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm28, 320(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm29, 128(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm30, 192(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm21, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm24, 64(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm22, 384(%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm13, 384(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm14, 448(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm10, 256(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm12, 320(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm8, 128(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm9, 192(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm19, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm7, 64(%rcx)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <192 x i64>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <192 x i64> %wide.vec, <192 x i64> poison, <64 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45, i32 48, i32 51, i32 54, i32 57, i32 60, i32 63, i32 66, i32 69, i32 72, i32 75, i32 78, i32 81, i32 84, i32 87, i32 90, i32 93, i32 96, i32 99, i32 102, i32 105, i32 108, i32 111, i32 114, i32 117, i32 120, i32 123, i32 126, i32 129, i32 132, i32 135, i32 138, i32 141, i32 144, i32 147, i32 150, i32 153, i32 156, i32 159, i32 162, i32 165, i32 168, i32 171, i32 174, i32 177, i32 180, i32 183, i32 186, i32 189>
  %strided.vec1 = shufflevector <192 x i64> %wide.vec, <192 x i64> poison, <64 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46, i32 49, i32 52, i32 55, i32 58, i32 61, i32 64, i32 67, i32 70, i32 73, i32 76, i32 79, i32 82, i32 85, i32 88, i32 91, i32 94, i32 97, i32 100, i32 103, i32 106, i32 109, i32 112, i32 115, i32 118, i32 121, i32 124, i32 127, i32 130, i32 133, i32 136, i32 139, i32 142, i32 145, i32 148, i32 151, i32 154, i32 157, i32 160, i32 163, i32 166, i32 169, i32 172, i32 175, i32 178, i32 181, i32 184, i32 187, i32 190>
  %strided.vec2 = shufflevector <192 x i64> %wide.vec, <192 x i64> poison, <64 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47, i32 50, i32 53, i32 56, i32 59, i32 62, i32 65, i32 68, i32 71, i32 74, i32 77, i32 80, i32 83, i32 86, i32 89, i32 92, i32 95, i32 98, i32 101, i32 104, i32 107, i32 110, i32 113, i32 116, i32 119, i32 122, i32 125, i32 128, i32 131, i32 134, i32 137, i32 140, i32 143, i32 146, i32 149, i32 152, i32 155, i32 158, i32 161, i32 164, i32 167, i32 170, i32 173, i32 176, i32 179, i32 182, i32 185, i32 188, i32 191>
  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
  ret void
}