File: vector-interleaved-load-i8-stride-3.ll

package info (click to toggle)
llvm-toolchain-19 1%3A19.1.7-3
links: PTS, VCS
area: main
in suites: forky, sid, trixie
size: 1,998,520 kB
sloc: cpp: 6,951,680; ansic: 1,486,157; asm: 913,598; python: 232,024; f90: 80,126; objc: 75,281; lisp: 37,276; pascal: 16,990; sh: 10,009; ml: 5,058; perl: 4,724; awk: 3,523; makefile: 3,167; javascript: 2,504; xml: 892; fortran: 664; cs: 573
file content (2748 lines) | stat: -rw-r--r-- 171,062 bytes
parent folder | download | duplicates (3)
; 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_i8_stride3_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
; SSE-LABEL: load_i8_stride3_vf2:
; SSE:       # %bb.0:
; SSE-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
; SSE-NEXT:    pxor %xmm1, %xmm1
; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[0,3,2,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm1, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm2, %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm0, %xmm0
; SSE-NEXT:    movd %xmm1, %eax
; SSE-NEXT:    movw %ax, (%rsi)
; SSE-NEXT:    movd %xmm2, %eax
; SSE-NEXT:    movw %ax, (%rdx)
; SSE-NEXT:    movd %xmm0, %eax
; SSE-NEXT:    movw %ax, (%rcx)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i8_stride3_vf2:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX-NEXT:    vpextrw $0, %xmm1, (%rsi)
; AVX-NEXT:    vpextrw $0, %xmm2, (%rdx)
; AVX-NEXT:    vpextrw $0, %xmm0, (%rcx)
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i8_stride3_vf2:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX2-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-NEXT:    vpextrw $0, %xmm1, (%rsi)
; AVX2-NEXT:    vpextrw $0, %xmm2, (%rdx)
; AVX2-NEXT:    vpextrw $0, %xmm0, (%rcx)
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i8_stride3_vf2:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX2-FP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FP-NEXT:    vpextrw $0, %xmm1, (%rsi)
; AVX2-FP-NEXT:    vpextrw $0, %xmm2, (%rdx)
; AVX2-FP-NEXT:    vpextrw $0, %xmm0, (%rcx)
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i8_stride3_vf2:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX2-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FCP-NEXT:    vpextrw $0, %xmm1, (%rsi)
; AVX2-FCP-NEXT:    vpextrw $0, %xmm2, (%rdx)
; AVX2-FCP-NEXT:    vpextrw $0, %xmm0, (%rcx)
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i8_stride3_vf2:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX512-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512-NEXT:    vpextrw $0, %xmm1, (%rsi)
; AVX512-NEXT:    vpextrw $0, %xmm2, (%rdx)
; AVX512-NEXT:    vpextrw $0, %xmm0, (%rcx)
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i8_stride3_vf2:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT:    vpextrw $0, %xmm1, (%rsi)
; AVX512-FCP-NEXT:    vpextrw $0, %xmm2, (%rdx)
; AVX512-FCP-NEXT:    vpextrw $0, %xmm0, (%rcx)
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i8_stride3_vf2:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-NEXT:    vpextrw $0, %xmm1, (%rsi)
; AVX512DQ-NEXT:    vpextrw $0, %xmm2, (%rdx)
; AVX512DQ-NEXT:    vpextrw $0, %xmm0, (%rcx)
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i8_stride3_vf2:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT:    vpextrw $0, %xmm1, (%rsi)
; AVX512DQ-FCP-NEXT:    vpextrw $0, %xmm2, (%rdx)
; AVX512DQ-FCP-NEXT:    vpextrw $0, %xmm0, (%rcx)
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i8_stride3_vf2:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX512BW-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT:    vpextrw $0, %xmm1, (%rsi)
; AVX512BW-NEXT:    vpextrw $0, %xmm2, (%rdx)
; AVX512BW-NEXT:    vpextrw $0, %xmm0, (%rcx)
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i8_stride3_vf2:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-FCP-NEXT:    vpextrw $0, %xmm1, (%rsi)
; AVX512BW-FCP-NEXT:    vpextrw $0, %xmm2, (%rdx)
; AVX512BW-FCP-NEXT:    vpextrw $0, %xmm0, (%rcx)
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i8_stride3_vf2:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX512DQ-BW-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-NEXT:    vpextrw $0, %xmm1, (%rsi)
; AVX512DQ-BW-NEXT:    vpextrw $0, %xmm2, (%rdx)
; AVX512DQ-BW-NEXT:    vpextrw $0, %xmm0, (%rcx)
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i8_stride3_vf2:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-FCP-NEXT:    vpextrw $0, %xmm1, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vpextrw $0, %xmm2, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vpextrw $0, %xmm0, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <6 x i8>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <6 x i8> %wide.vec, <6 x i8> poison, <2 x i32> <i32 0, i32 3>
  %strided.vec1 = shufflevector <6 x i8> %wide.vec, <6 x i8> poison, <2 x i32> <i32 1, i32 4>
  %strided.vec2 = shufflevector <6 x i8> %wide.vec, <6 x i8> poison, <2 x i32> <i32 2, i32 5>
  store <2 x i8> %strided.vec0, ptr %out.vec0, align 64
  store <2 x i8> %strided.vec1, ptr %out.vec1, align 64
  store <2 x i8> %strided.vec2, ptr %out.vec2, align 64
  ret void
}

define void @load_i8_stride3_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
; SSE-LABEL: load_i8_stride3_vf4:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,65535,65535,65535,65535]
; SSE-NEXT:    pxor %xmm2, %xmm2
; SSE-NEXT:    movdqa %xmm0, %xmm3
; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
; SSE-NEXT:    movdqa %xmm0, %xmm2
; SSE-NEXT:    pand %xmm1, %xmm2
; SSE-NEXT:    pandn %xmm3, %xmm1
; SSE-NEXT:    por %xmm2, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
; SSE-NEXT:    packuswb %xmm1, %xmm1
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm3[0,3,2,3,4,5,6,7]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm0[0,0]
; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm0[2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,2,3,0,4,5,6,7]
; SSE-NEXT:    packuswb %xmm3, %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE-NEXT:    packuswb %xmm0, %xmm0
; SSE-NEXT:    movd %xmm1, (%rsi)
; SSE-NEXT:    movd %xmm3, (%rdx)
; SSE-NEXT:    movd %xmm0, (%rcx)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i8_stride3_vf4:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovdqa (%rdi), %xmm0
; AVX-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,6,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,7,10,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX-NEXT:    vmovd %xmm1, (%rsi)
; AVX-NEXT:    vmovd %xmm2, (%rdx)
; AVX-NEXT:    vmovd %xmm0, (%rcx)
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i8_stride3_vf4:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,6,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,7,10,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-NEXT:    vmovd %xmm1, (%rsi)
; AVX2-NEXT:    vmovd %xmm2, (%rdx)
; AVX2-NEXT:    vmovd %xmm0, (%rcx)
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i8_stride3_vf4:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-FP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,6,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,7,10,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FP-NEXT:    vmovd %xmm1, (%rsi)
; AVX2-FP-NEXT:    vmovd %xmm2, (%rdx)
; AVX2-FP-NEXT:    vmovd %xmm0, (%rcx)
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i8_stride3_vf4:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,6,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,7,10,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX2-FCP-NEXT:    vmovd %xmm1, (%rsi)
; AVX2-FCP-NEXT:    vmovd %xmm2, (%rdx)
; AVX2-FCP-NEXT:    vmovd %xmm0, (%rcx)
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i8_stride3_vf4:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,6,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,7,10,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512-NEXT:    vmovd %xmm1, (%rsi)
; AVX512-NEXT:    vmovd %xmm2, (%rdx)
; AVX512-NEXT:    vmovd %xmm0, (%rcx)
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i8_stride3_vf4:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,6,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,7,10,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT:    vmovd %xmm1, (%rsi)
; AVX512-FCP-NEXT:    vmovd %xmm2, (%rdx)
; AVX512-FCP-NEXT:    vmovd %xmm0, (%rcx)
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i8_stride3_vf4:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,6,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,7,10,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-NEXT:    vmovd %xmm1, (%rsi)
; AVX512DQ-NEXT:    vmovd %xmm2, (%rdx)
; AVX512DQ-NEXT:    vmovd %xmm0, (%rcx)
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i8_stride3_vf4:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,6,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,7,10,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT:    vmovd %xmm1, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovd %xmm2, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovd %xmm0, (%rcx)
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i8_stride3_vf4:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,6,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,7,10,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT:    vmovd %xmm1, (%rsi)
; AVX512BW-NEXT:    vmovd %xmm2, (%rdx)
; AVX512BW-NEXT:    vmovd %xmm0, (%rcx)
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i8_stride3_vf4:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,6,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,7,10,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512BW-FCP-NEXT:    vmovd %xmm1, (%rsi)
; AVX512BW-FCP-NEXT:    vmovd %xmm2, (%rdx)
; AVX512BW-FCP-NEXT:    vmovd %xmm0, (%rcx)
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i8_stride3_vf4:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-BW-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,6,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,7,10,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-NEXT:    vmovd %xmm1, (%rsi)
; AVX512DQ-BW-NEXT:    vmovd %xmm2, (%rdx)
; AVX512DQ-BW-NEXT:    vmovd %xmm0, (%rcx)
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i8_stride3_vf4:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,3,6,9,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = xmm0[1,4,7,10,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-FCP-NEXT:    vmovd %xmm1, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovd %xmm2, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovd %xmm0, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <12 x i8>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <12 x i8> %wide.vec, <12 x i8> poison, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
  %strided.vec1 = shufflevector <12 x i8> %wide.vec, <12 x i8> poison, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
  %strided.vec2 = shufflevector <12 x i8> %wide.vec, <12 x i8> poison, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
  store <4 x i8> %strided.vec0, ptr %out.vec0, align 64
  store <4 x i8> %strided.vec1, ptr %out.vec1, align 64
  store <4 x i8> %strided.vec2, ptr %out.vec2, align 64
  ret void
}

define void @load_i8_stride3_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
; SSE-LABEL: load_i8_stride3_vf8:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    movdqa 16(%rdi), %xmm1
; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,0,255,255,0,255,255,255,255,255,255,255,255,255,255]
; SSE-NEXT:    movdqa %xmm0, %xmm2
; SSE-NEXT:    pand %xmm4, %xmm2
; SSE-NEXT:    pandn %xmm1, %xmm4
; SSE-NEXT:    por %xmm2, %xmm4
; SSE-NEXT:    pxor %xmm2, %xmm2
; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3],xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7]
; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,0,65535,65535,0]
; SSE-NEXT:    pand %xmm5, %xmm4
; SSE-NEXT:    movdqa %xmm0, %xmm3
; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
; SSE-NEXT:    pandn %xmm3, %xmm5
; SSE-NEXT:    por %xmm4, %xmm5
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[0,2,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,5,4,7]
; SSE-NEXT:    packuswb %xmm4, %xmm4
; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [0,255,255,0,255,255,0,255,255,255,255,255,255,255,255,255]
; SSE-NEXT:    movdqa %xmm0, %xmm6
; SSE-NEXT:    pand %xmm5, %xmm6
; SSE-NEXT:    pandn %xmm1, %xmm5
; SSE-NEXT:    por %xmm6, %xmm5
; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3],xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,0,65535,65535,0,65535,65535]
; SSE-NEXT:    pand %xmm6, %xmm5
; SSE-NEXT:    pandn %xmm3, %xmm6
; SSE-NEXT:    por %xmm5, %xmm6
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm6[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,2,3,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,6,7,4]
; SSE-NEXT:    packuswb %xmm5, %xmm5
; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,0,255,255,0,255,255,0,255,255,255,255,255,255,255,255]
; SSE-NEXT:    pand %xmm6, %xmm0
; SSE-NEXT:    pandn %xmm1, %xmm6
; SSE-NEXT:    por %xmm0, %xmm6
; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1],xmm6[2],xmm2[2],xmm6[3],xmm2[3],xmm6[4],xmm2[4],xmm6[5],xmm2[5],xmm6[6],xmm2[6],xmm6[7],xmm2[7]
; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [0,65535,65535,0,65535,65535,0,65535]
; SSE-NEXT:    pand %xmm0, %xmm6
; SSE-NEXT:    pandn %xmm3, %xmm0
; SSE-NEXT:    por %xmm6, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
; SSE-NEXT:    packuswb %xmm0, %xmm0
; SSE-NEXT:    movq %xmm4, (%rsi)
; SSE-NEXT:    movq %xmm5, (%rdx)
; SSE-NEXT:    movq %xmm0, (%rcx)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i8_stride3_vf8:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovdqa (%rdi), %xmm0
; AVX-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX-NEXT:    vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u]
; AVX-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX-NEXT:    vpor %xmm2, %xmm3, %xmm2
; AVX-NEXT:    vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u]
; AVX-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX-NEXT:    vpor %xmm3, %xmm4, %xmm3
; AVX-NEXT:    vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u]
; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
; AVX-NEXT:    vmovq %xmm2, (%rsi)
; AVX-NEXT:    vmovq %xmm3, (%rdx)
; AVX-NEXT:    vmovq %xmm0, (%rcx)
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i8_stride3_vf8:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-NEXT:    vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u]
; AVX2-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX2-NEXT:    vpor %xmm2, %xmm3, %xmm2
; AVX2-NEXT:    vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u]
; AVX2-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX2-NEXT:    vpor %xmm3, %xmm4, %xmm3
; AVX2-NEXT:    vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u]
; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT:    vmovq %xmm2, (%rsi)
; AVX2-NEXT:    vmovq %xmm3, (%rdx)
; AVX2-NEXT:    vmovq %xmm0, (%rcx)
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i8_stride3_vf8:
; AVX2-FP:       # %bb.0:
; AVX2-FP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-FP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-FP-NEXT:    vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u]
; AVX2-FP-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX2-FP-NEXT:    vpor %xmm2, %xmm3, %xmm2
; AVX2-FP-NEXT:    vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u]
; AVX2-FP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX2-FP-NEXT:    vpor %xmm3, %xmm4, %xmm3
; AVX2-FP-NEXT:    vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u]
; AVX2-FP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX2-FP-NEXT:    vpor %xmm1, %xmm0, %xmm0
; AVX2-FP-NEXT:    vmovq %xmm2, (%rsi)
; AVX2-FP-NEXT:    vmovq %xmm3, (%rdx)
; AVX2-FP-NEXT:    vmovq %xmm0, (%rcx)
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i8_stride3_vf8:
; AVX2-FCP:       # %bb.0:
; AVX2-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u]
; AVX2-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX2-FCP-NEXT:    vpor %xmm2, %xmm3, %xmm2
; AVX2-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u]
; AVX2-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX2-FCP-NEXT:    vpor %xmm3, %xmm4, %xmm3
; AVX2-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u]
; AVX2-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX2-FCP-NEXT:    vpor %xmm1, %xmm0, %xmm0
; AVX2-FCP-NEXT:    vmovq %xmm2, (%rsi)
; AVX2-FCP-NEXT:    vmovq %xmm3, (%rdx)
; AVX2-FCP-NEXT:    vmovq %xmm0, (%rcx)
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i8_stride3_vf8:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512-NEXT:    vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u]
; AVX512-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512-NEXT:    vpor %xmm2, %xmm3, %xmm2
; AVX512-NEXT:    vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u]
; AVX512-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512-NEXT:    vpor %xmm3, %xmm4, %xmm3
; AVX512-NEXT:    vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u]
; AVX512-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512-NEXT:    vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT:    vmovq %xmm2, (%rsi)
; AVX512-NEXT:    vmovq %xmm3, (%rdx)
; AVX512-NEXT:    vmovq %xmm0, (%rcx)
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i8_stride3_vf8:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT:    vpor %xmm2, %xmm3, %xmm2
; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT:    vpor %xmm3, %xmm4, %xmm3
; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512-FCP-NEXT:    vpor %xmm1, %xmm0, %xmm0
; AVX512-FCP-NEXT:    vmovq %xmm2, (%rsi)
; AVX512-FCP-NEXT:    vmovq %xmm3, (%rdx)
; AVX512-FCP-NEXT:    vmovq %xmm0, (%rcx)
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i8_stride3_vf8:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u]
; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512DQ-NEXT:    vpor %xmm2, %xmm3, %xmm2
; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u]
; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512DQ-NEXT:    vpor %xmm3, %xmm4, %xmm3
; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u]
; AVX512DQ-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512DQ-NEXT:    vpor %xmm1, %xmm0, %xmm0
; AVX512DQ-NEXT:    vmovq %xmm2, (%rsi)
; AVX512DQ-NEXT:    vmovq %xmm3, (%rdx)
; AVX512DQ-NEXT:    vmovq %xmm0, (%rcx)
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i8_stride3_vf8:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT:    vpor %xmm2, %xmm3, %xmm2
; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT:    vpor %xmm3, %xmm4, %xmm3
; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512DQ-FCP-NEXT:    vpor %xmm1, %xmm0, %xmm0
; AVX512DQ-FCP-NEXT:    vmovq %xmm2, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovq %xmm3, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovq %xmm0, (%rcx)
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i8_stride3_vf8:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512BW-NEXT:    vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512BW-NEXT:    vpor %xmm2, %xmm3, %xmm2
; AVX512BW-NEXT:    vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512BW-NEXT:    vpor %xmm3, %xmm4, %xmm3
; AVX512BW-NEXT:    vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u]
; AVX512BW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512BW-NEXT:    vpor %xmm1, %xmm0, %xmm0
; AVX512BW-NEXT:    vmovq %xmm2, (%rsi)
; AVX512BW-NEXT:    vmovq %xmm3, (%rdx)
; AVX512BW-NEXT:    vmovq %xmm0, (%rcx)
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i8_stride3_vf8:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512BW-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u]
; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512BW-FCP-NEXT:    vpor %xmm2, %xmm3, %xmm2
; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u]
; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512BW-FCP-NEXT:    vpor %xmm3, %xmm4, %xmm3
; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u]
; AVX512BW-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512BW-FCP-NEXT:    vpor %xmm1, %xmm0, %xmm0
; AVX512BW-FCP-NEXT:    vmovq %xmm2, (%rsi)
; AVX512BW-FCP-NEXT:    vmovq %xmm3, (%rdx)
; AVX512BW-FCP-NEXT:    vmovq %xmm0, (%rcx)
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i8_stride3_vf8:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-BW-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-BW-NEXT:    vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512DQ-BW-NEXT:    vpor %xmm2, %xmm3, %xmm2
; AVX512DQ-BW-NEXT:    vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512DQ-BW-NEXT:    vpor %xmm3, %xmm4, %xmm3
; AVX512DQ-BW-NEXT:    vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512DQ-BW-NEXT:    vpor %xmm1, %xmm0, %xmm0
; AVX512DQ-BW-NEXT:    vmovq %xmm2, (%rsi)
; AVX512DQ-BW-NEXT:    vmovq %xmm3, (%rdx)
; AVX512DQ-BW-NEXT:    vmovq %xmm0, (%rcx)
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i8_stride3_vf8:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512DQ-BW-FCP-NEXT:    vpor %xmm2, %xmm3, %xmm2
; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} xmm4 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512DQ-BW-FCP-NEXT:    vpor %xmm3, %xmm4, %xmm3
; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u]
; AVX512DQ-BW-FCP-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u]
; AVX512DQ-BW-FCP-NEXT:    vpor %xmm1, %xmm0, %xmm0
; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm2, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm3, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovq %xmm0, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <24 x i8>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <24 x i8> %wide.vec, <24 x i8> 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 i8> %wide.vec, <24 x i8> 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 i8> %wide.vec, <24 x i8> poison, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
  store <8 x i8> %strided.vec0, ptr %out.vec0, align 64
  store <8 x i8> %strided.vec1, ptr %out.vec1, align 64
  store <8 x i8> %strided.vec2, ptr %out.vec2, align 64
  ret void
}

define void @load_i8_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
; SSE-LABEL: load_i8_stride3_vf16:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa (%rdi), %xmm5
; SSE-NEXT:    movdqa 16(%rdi), %xmm4
; SSE-NEXT:    movdqa 32(%rdi), %xmm0
; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255]
; SSE-NEXT:    movdqa %xmm3, %xmm1
; SSE-NEXT:    pandn %xmm4, %xmm1
; SSE-NEXT:    movdqa %xmm5, %xmm2
; SSE-NEXT:    pand %xmm3, %xmm2
; SSE-NEXT:    por %xmm1, %xmm2
; SSE-NEXT:    pxor %xmm6, %xmm6
; SSE-NEXT:    movdqa %xmm2, %xmm7
; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm6[8],xmm7[9],xmm6[9],xmm7[10],xmm6[10],xmm7[11],xmm6[11],xmm7[12],xmm6[12],xmm7[13],xmm6[13],xmm7[14],xmm6[14],xmm7[15],xmm6[15]
; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,0,65535,65535,0]
; SSE-NEXT:    movdqa %xmm1, %xmm8
; SSE-NEXT:    pandn %xmm7, %xmm8
; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
; SSE-NEXT:    pand %xmm1, %xmm2
; SSE-NEXT:    por %xmm8, %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm2[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm7[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
; SSE-NEXT:    packuswb %xmm2, %xmm10
; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; SSE-NEXT:    pand %xmm2, %xmm10
; SSE-NEXT:    movdqa %xmm0, %xmm7
; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm6[8],xmm7[9],xmm6[9],xmm7[10],xmm6[10],xmm7[11],xmm6[11],xmm7[12],xmm6[12],xmm7[13],xmm6[13],xmm7[14],xmm6[14],xmm7[15],xmm6[15]
; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,0,65535,65535,0,65535,65535]
; SSE-NEXT:    movdqa %xmm9, %xmm8
; SSE-NEXT:    pandn %xmm7, %xmm8
; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
; SSE-NEXT:    movdqa %xmm0, %xmm11
; SSE-NEXT:    pand %xmm9, %xmm11
; SSE-NEXT:    por %xmm8, %xmm11
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm11[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,7,6,5]
; SSE-NEXT:    packuswb %xmm8, %xmm11
; SSE-NEXT:    movdqa %xmm2, %xmm8
; SSE-NEXT:    pandn %xmm11, %xmm8
; SSE-NEXT:    por %xmm10, %xmm8
; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,0,255,255,0,255,255,0,255,255,0,255,255,0,255,255]
; SSE-NEXT:    movdqa %xmm4, %xmm11
; SSE-NEXT:    pand %xmm10, %xmm11
; SSE-NEXT:    pandn %xmm5, %xmm10
; SSE-NEXT:    por %xmm11, %xmm10
; SSE-NEXT:    movdqa %xmm10, %xmm11
; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm6[8],xmm11[9],xmm6[9],xmm11[10],xmm6[10],xmm11[11],xmm6[11],xmm11[12],xmm6[12],xmm11[13],xmm6[13],xmm11[14],xmm6[14],xmm11[15],xmm6[15]
; SSE-NEXT:    movdqa %xmm9, %xmm12
; SSE-NEXT:    pandn %xmm11, %xmm12
; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm6[0],xmm10[1],xmm6[1],xmm10[2],xmm6[2],xmm10[3],xmm6[3],xmm10[4],xmm6[4],xmm10[5],xmm6[5],xmm10[6],xmm6[6],xmm10[7],xmm6[7]
; SSE-NEXT:    pand %xmm9, %xmm10
; SSE-NEXT:    por %xmm12, %xmm10
; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[1,2,3,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,6,7,4]
; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,7,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[1,2,3,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm11, %xmm10
; SSE-NEXT:    pand %xmm2, %xmm10
; SSE-NEXT:    movdqa %xmm7, %xmm11
; SSE-NEXT:    pand %xmm9, %xmm11
; SSE-NEXT:    pandn %xmm0, %xmm9
; SSE-NEXT:    por %xmm11, %xmm9
; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,6,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,5,6,7,4]
; SSE-NEXT:    packuswb %xmm9, %xmm9
; SSE-NEXT:    pandn %xmm9, %xmm2
; SSE-NEXT:    por %xmm10, %xmm2
; SSE-NEXT:    pand %xmm3, %xmm4
; SSE-NEXT:    pandn %xmm5, %xmm3
; SSE-NEXT:    por %xmm4, %xmm3
; SSE-NEXT:    movdqa %xmm3, %xmm4
; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm6[8],xmm4[9],xmm6[9],xmm4[10],xmm6[10],xmm4[11],xmm6[11],xmm4[12],xmm6[12],xmm4[13],xmm6[13],xmm4[14],xmm6[14],xmm4[15],xmm6[15]
; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [0,65535,65535,0,65535,65535,0,65535]
; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3],xmm3[4],xmm6[4],xmm3[5],xmm6[5],xmm3[6],xmm6[6],xmm3[7],xmm6[7]
; SSE-NEXT:    pand %xmm5, %xmm3
; SSE-NEXT:    pandn %xmm4, %xmm5
; SSE-NEXT:    por %xmm3, %xmm5
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[3,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,2,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm4, %xmm3
; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,65535,0,0,0]
; SSE-NEXT:    pand %xmm4, %xmm3
; SSE-NEXT:    pand %xmm1, %xmm0
; SSE-NEXT:    pandn %xmm7, %xmm1
; SSE-NEXT:    por %xmm0, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
; SSE-NEXT:    packuswb %xmm0, %xmm0
; SSE-NEXT:    pandn %xmm0, %xmm4
; SSE-NEXT:    por %xmm3, %xmm4
; SSE-NEXT:    movdqa %xmm8, (%rsi)
; SSE-NEXT:    movdqa %xmm2, (%rdx)
; SSE-NEXT:    movdqa %xmm4, (%rcx)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i8_stride3_vf16:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovdqa (%rdi), %xmm0
; AVX-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
; AVX-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
; AVX-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
; AVX-NEXT:    vpalignr {{.*#+}} xmm3 = xmm2[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpalignr {{.*#+}} xmm2 = xmm3[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpmovsxdq {{.*#+}} xmm4 = [18446744073709551615,16777215]
; AVX-NEXT:    vpblendvb %xmm4, %xmm0, %xmm1, %xmm1
; AVX-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX-NEXT:    vmovdqa %xmm1, (%rdx)
; AVX-NEXT:    vmovdqa %xmm2, (%rcx)
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i8_stride3_vf16:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX2-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
; AVX2-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
; AVX2-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
; AVX2-NEXT:    vpalignr {{.*#+}} xmm3 = xmm2[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-NEXT:    vpalignr {{.*#+}} xmm2 = xmm3[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-NEXT:    vpmovsxdq {{.*#+}} xmm4 = [18446744073709551615,16777215]
; AVX2-NEXT:    vpblendvb %xmm4, %xmm0, %xmm1, %xmm1
; AVX2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX2-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX2-NEXT:    vmovdqa %xmm1, (%rdx)
; AVX2-NEXT:    vmovdqa %xmm2, (%rcx)
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i8_stride3_vf16:
; 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:    vmovdqa {{.*#+}} xmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX2-FP-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
; AVX2-FP-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
; AVX2-FP-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
; AVX2-FP-NEXT:    vpalignr {{.*#+}} xmm3 = xmm2[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-FP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-FP-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-FP-NEXT:    vpalignr {{.*#+}} xmm2 = xmm3[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-FP-NEXT:    vpmovsxdq {{.*#+}} xmm4 = [18446744073709551615,16777215]
; AVX2-FP-NEXT:    vpblendvb %xmm4, %xmm0, %xmm1, %xmm1
; AVX2-FP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-FP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX2-FP-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX2-FP-NEXT:    vmovdqa %xmm1, (%rdx)
; AVX2-FP-NEXT:    vmovdqa %xmm2, (%rcx)
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i8_stride3_vf16:
; 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:    vmovdqa {{.*#+}} xmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX2-FCP-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
; AVX2-FCP-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
; AVX2-FCP-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} xmm3 = xmm2[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} xmm2 = xmm3[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-FCP-NEXT:    vpmovsxdq {{.*#+}} xmm4 = [18446744073709551615,16777215]
; AVX2-FCP-NEXT:    vpblendvb %xmm4, %xmm0, %xmm1, %xmm1
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7,8,9,10]
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX2-FCP-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX2-FCP-NEXT:    vmovdqa %xmm1, (%rdx)
; AVX2-FCP-NEXT:    vmovdqa %xmm2, (%rcx)
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i8_stride3_vf16:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
; AVX512-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
; AVX512-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
; AVX512-NEXT:    vpalignr {{.*#+}} xmm3 = xmm2[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX512-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10]
; AVX512-NEXT:    vpalignr {{.*#+}} xmm2 = xmm3[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
; AVX512-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7,8,9,10]
; AVX512-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX512-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX512-NEXT:    vmovdqa %xmm1, (%rdx)
; AVX512-NEXT:    vmovdqa %xmm2, (%rcx)
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i8_stride3_vf16:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512-FCP-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512-FCP-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
; AVX512-FCP-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
; AVX512-FCP-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} xmm3 = xmm2[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} xmm2 = xmm3[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7,8,9,10]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX512-FCP-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa %xmm1, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa %xmm2, (%rcx)
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i8_stride3_vf16:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512DQ-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
; AVX512DQ-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
; AVX512DQ-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
; AVX512DQ-NEXT:    vpalignr {{.*#+}} xmm3 = xmm2[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} xmm2 = xmm3[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
; AVX512DQ-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX512DQ-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX512DQ-NEXT:    vmovdqa %xmm1, (%rdx)
; AVX512DQ-NEXT:    vmovdqa %xmm2, (%rcx)
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i8_stride3_vf16:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-FCP-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512DQ-FCP-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
; AVX512DQ-FCP-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
; AVX512DQ-FCP-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} xmm3 = xmm2[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} xmm2 = xmm3[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-FCP-NEXT:    vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm1
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX512DQ-FCP-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa %xmm1, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa %xmm2, (%rcx)
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i8_stride3_vf16:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512BW-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512BW-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
; AVX512BW-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
; AVX512BW-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
; AVX512BW-NEXT:    vpalignr {{.*#+}} xmm3 = xmm2[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX512BW-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512BW-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10]
; AVX512BW-NEXT:    movw $-2048, %ax # imm = 0xF800
; AVX512BW-NEXT:    kmovd %eax, %k1
; AVX512BW-NEXT:    vpblendmb %xmm1, %xmm0, %xmm2 {%k1}
; AVX512BW-NEXT:    vpalignr {{.*#+}} xmm1 = xmm3[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512BW-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7,8,9,10]
; AVX512BW-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX512BW-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX512BW-NEXT:    vmovdqa %xmm2, (%rdx)
; AVX512BW-NEXT:    vmovdqa %xmm1, (%rcx)
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i8_stride3_vf16:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512BW-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512BW-FCP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512BW-FCP-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512BW-FCP-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
; AVX512BW-FCP-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
; AVX512BW-FCP-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} xmm3 = xmm2[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10]
; AVX512BW-FCP-NEXT:    movw $-2048, %ax # imm = 0xF800
; AVX512BW-FCP-NEXT:    kmovd %eax, %k1
; AVX512BW-FCP-NEXT:    vpblendmb %xmm1, %xmm0, %xmm2 {%k1}
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} xmm1 = xmm3[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7,8,9,10]
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX512BW-FCP-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa %xmm2, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa %xmm1, (%rcx)
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i8_stride3_vf16:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-BW-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-BW-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-BW-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512DQ-BW-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
; AVX512DQ-BW-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
; AVX512DQ-BW-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} xmm3 = xmm2[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-BW-NEXT:    movw $-2048, %ax # imm = 0xF800
; AVX512DQ-BW-NEXT:    kmovd %eax, %k1
; AVX512DQ-BW-NEXT:    vpblendmb %xmm1, %xmm0, %xmm2 {%k1}
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} xmm1 = xmm3[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX512DQ-BW-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa %xmm2, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa %xmm1, (%rcx)
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i8_stride3_vf16:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
; AVX512DQ-BW-FCP-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} xmm3 = xmm2[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-BW-FCP-NEXT:    movw $-2048, %ax # imm = 0xF800
; AVX512DQ-BW-FCP-NEXT:    kmovd %eax, %k1
; AVX512DQ-BW-FCP-NEXT:    vpblendmb %xmm1, %xmm0, %xmm2 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} xmm1 = xmm3[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7,8,9,10]
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm2, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %xmm1, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <48 x i8>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <48 x i8> %wide.vec, <48 x i8> 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 i8> %wide.vec, <48 x i8> 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 i8> %wide.vec, <48 x i8> 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 i8> %strided.vec0, ptr %out.vec0, align 64
  store <16 x i8> %strided.vec1, ptr %out.vec1, align 64
  store <16 x i8> %strided.vec2, ptr %out.vec2, align 64
  ret void
}

define void @load_i8_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
; SSE-LABEL: load_i8_stride3_vf32:
; SSE:       # %bb.0:
; SSE-NEXT:    movdqa 64(%rdi), %xmm2
; SSE-NEXT:    movdqa (%rdi), %xmm6
; SSE-NEXT:    movdqa 16(%rdi), %xmm4
; SSE-NEXT:    movdqa 32(%rdi), %xmm8
; SSE-NEXT:    movdqa 48(%rdi), %xmm12
; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255]
; SSE-NEXT:    movdqa %xmm13, %xmm7
; SSE-NEXT:    pandn %xmm4, %xmm7
; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,0,255,255,0,255,255,0,255,255,0,255,255,0,255,255]
; SSE-NEXT:    movdqa %xmm5, %xmm14
; SSE-NEXT:    pandn %xmm6, %xmm14
; SSE-NEXT:    movdqa %xmm13, %xmm0
; SSE-NEXT:    pandn %xmm6, %xmm0
; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm6, %xmm0
; SSE-NEXT:    pand %xmm13, %xmm0
; SSE-NEXT:    por %xmm7, %xmm0
; SSE-NEXT:    pxor %xmm9, %xmm9
; SSE-NEXT:    movdqa %xmm0, %xmm1
; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,0,65535,65535,0,65535,65535,0]
; SSE-NEXT:    movdqa %xmm6, %xmm3
; SSE-NEXT:    pandn %xmm1, %xmm3
; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
; SSE-NEXT:    pand %xmm6, %xmm0
; SSE-NEXT:    por %xmm3, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
; SSE-NEXT:    packuswb %xmm1, %xmm0
; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; SSE-NEXT:    pand %xmm7, %xmm0
; SSE-NEXT:    movdqa %xmm8, %xmm3
; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm9[8],xmm3[9],xmm9[9],xmm3[10],xmm9[10],xmm3[11],xmm9[11],xmm3[12],xmm9[12],xmm3[13],xmm9[13],xmm3[14],xmm9[14],xmm3[15],xmm9[15]
; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,65535,0,65535,65535,0,65535,65535]
; SSE-NEXT:    movdqa %xmm15, %xmm1
; SSE-NEXT:    pandn %xmm3, %xmm1
; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1],xmm8[2],xmm9[2],xmm8[3],xmm9[3],xmm8[4],xmm9[4],xmm8[5],xmm9[5],xmm8[6],xmm9[6],xmm8[7],xmm9[7]
; SSE-NEXT:    movdqa %xmm8, %xmm10
; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pand %xmm15, %xmm10
; SSE-NEXT:    por %xmm1, %xmm10
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
; SSE-NEXT:    packuswb %xmm1, %xmm1
; SSE-NEXT:    movdqa %xmm7, %xmm10
; SSE-NEXT:    pandn %xmm1, %xmm10
; SSE-NEXT:    por %xmm0, %xmm10
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm13, %xmm0
; SSE-NEXT:    pandn %xmm2, %xmm0
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm12, %xmm1
; SSE-NEXT:    pand %xmm13, %xmm1
; SSE-NEXT:    por %xmm0, %xmm1
; SSE-NEXT:    movdqa %xmm1, %xmm0
; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
; SSE-NEXT:    movdqa %xmm6, %xmm10
; SSE-NEXT:    pandn %xmm0, %xmm10
; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
; SSE-NEXT:    pand %xmm6, %xmm1
; SSE-NEXT:    por %xmm10, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
; SSE-NEXT:    packuswb %xmm0, %xmm1
; SSE-NEXT:    movdqa 80(%rdi), %xmm10
; SSE-NEXT:    movdqa %xmm10, %xmm12
; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm9[8],xmm12[9],xmm9[9],xmm12[10],xmm9[10],xmm12[11],xmm9[11],xmm12[12],xmm9[12],xmm12[13],xmm9[13],xmm12[14],xmm9[14],xmm12[15],xmm9[15]
; SSE-NEXT:    movdqa %xmm15, %xmm0
; SSE-NEXT:    pandn %xmm12, %xmm0
; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3],xmm10[4],xmm9[4],xmm10[5],xmm9[5],xmm10[6],xmm9[6],xmm10[7],xmm9[7]
; SSE-NEXT:    movdqa %xmm10, %xmm11
; SSE-NEXT:    pand %xmm15, %xmm11
; SSE-NEXT:    por %xmm0, %xmm11
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
; SSE-NEXT:    packuswb %xmm0, %xmm0
; SSE-NEXT:    movdqa %xmm7, %xmm11
; SSE-NEXT:    pandn %xmm0, %xmm11
; SSE-NEXT:    pand %xmm7, %xmm1
; SSE-NEXT:    por %xmm1, %xmm11
; SSE-NEXT:    movdqa %xmm4, %xmm0
; SSE-NEXT:    pand %xmm5, %xmm0
; SSE-NEXT:    por %xmm14, %xmm0
; SSE-NEXT:    movdqa %xmm0, %xmm1
; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
; SSE-NEXT:    movdqa %xmm15, %xmm14
; SSE-NEXT:    pandn %xmm1, %xmm14
; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
; SSE-NEXT:    pand %xmm15, %xmm0
; SSE-NEXT:    por %xmm14, %xmm0
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm0[0,1,2,3,5,6,7,4]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,7,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm0, %xmm14
; SSE-NEXT:    movdqa %xmm15, %xmm0
; SSE-NEXT:    pandn %xmm8, %xmm0
; SSE-NEXT:    movdqa %xmm6, %xmm1
; SSE-NEXT:    pandn %xmm3, %xmm1
; SSE-NEXT:    pand %xmm15, %xmm3
; SSE-NEXT:    por %xmm0, %xmm3
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
; SSE-NEXT:    packuswb %xmm0, %xmm0
; SSE-NEXT:    movdqa %xmm7, %xmm3
; SSE-NEXT:    pandn %xmm0, %xmm3
; SSE-NEXT:    pand %xmm7, %xmm14
; SSE-NEXT:    por %xmm14, %xmm3
; SSE-NEXT:    pand %xmm5, %xmm2
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT:    pandn %xmm8, %xmm5
; SSE-NEXT:    por %xmm2, %xmm5
; SSE-NEXT:    movdqa %xmm5, %xmm0
; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
; SSE-NEXT:    movdqa %xmm15, %xmm14
; SSE-NEXT:    pandn %xmm0, %xmm14
; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3],xmm5[4],xmm9[4],xmm5[5],xmm9[5],xmm5[6],xmm9[6],xmm5[7],xmm9[7]
; SSE-NEXT:    pand %xmm15, %xmm5
; SSE-NEXT:    por %xmm14, %xmm5
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,2,3,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,6,7,4]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm0, %xmm5
; SSE-NEXT:    movdqa %xmm12, %xmm0
; SSE-NEXT:    pand %xmm15, %xmm0
; SSE-NEXT:    pandn %xmm10, %xmm15
; SSE-NEXT:    por %xmm0, %xmm15
; SSE-NEXT:    pand %xmm7, %xmm5
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm15[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
; SSE-NEXT:    packuswb %xmm0, %xmm0
; SSE-NEXT:    pandn %xmm0, %xmm7
; SSE-NEXT:    por %xmm5, %xmm7
; SSE-NEXT:    pand %xmm13, %xmm4
; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    movdqa %xmm4, %xmm0
; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [0,65535,65535,0,65535,65535,0,65535]
; SSE-NEXT:    movdqa %xmm5, %xmm2
; SSE-NEXT:    pandn %xmm0, %xmm2
; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
; SSE-NEXT:    pand %xmm5, %xmm4
; SSE-NEXT:    por %xmm2, %xmm4
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[3,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm0, %xmm4
; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,65535,0,0,0]
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    pand %xmm6, %xmm0
; SSE-NEXT:    por %xmm1, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
; SSE-NEXT:    packuswb %xmm0, %xmm1
; SSE-NEXT:    movdqa %xmm2, %xmm0
; SSE-NEXT:    pandn %xmm1, %xmm0
; SSE-NEXT:    pand %xmm2, %xmm4
; SSE-NEXT:    por %xmm4, %xmm0
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    pand %xmm13, %xmm1
; SSE-NEXT:    pandn %xmm8, %xmm13
; SSE-NEXT:    por %xmm1, %xmm13
; SSE-NEXT:    movdqa %xmm13, %xmm1
; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm9[0],xmm13[1],xmm9[1],xmm13[2],xmm9[2],xmm13[3],xmm9[3],xmm13[4],xmm9[4],xmm13[5],xmm9[5],xmm13[6],xmm9[6],xmm13[7],xmm9[7]
; SSE-NEXT:    pand %xmm5, %xmm13
; SSE-NEXT:    pandn %xmm1, %xmm5
; SSE-NEXT:    por %xmm13, %xmm5
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[3,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm1, %xmm4
; SSE-NEXT:    pand %xmm6, %xmm10
; SSE-NEXT:    pandn %xmm12, %xmm6
; SSE-NEXT:    por %xmm10, %xmm6
; SSE-NEXT:    pand %xmm2, %xmm4
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT:    packuswb %xmm1, %xmm1
; SSE-NEXT:    pandn %xmm1, %xmm2
; SSE-NEXT:    por %xmm4, %xmm2
; SSE-NEXT:    movdqa %xmm11, 16(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, (%rsi)
; SSE-NEXT:    movdqa %xmm7, 16(%rdx)
; SSE-NEXT:    movdqa %xmm3, (%rdx)
; SSE-NEXT:    movdqa %xmm2, 16(%rcx)
; SSE-NEXT:    movdqa %xmm0, (%rcx)
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i8_stride3_vf32:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovdqa (%rdi), %xmm0
; AVX-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX-NEXT:    vmovdqa 48(%rdi), %xmm3
; AVX-NEXT:    vmovdqa 64(%rdi), %xmm4
; AVX-NEXT:    vmovdqa 80(%rdi), %xmm5
; AVX-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX-NEXT:    vpshufb %xmm6, %xmm0, %xmm0
; AVX-NEXT:    vpshufb %xmm6, %xmm3, %xmm3
; AVX-NEXT:    vpshufb %xmm6, %xmm1, %xmm1
; AVX-NEXT:    vpshufb %xmm6, %xmm4, %xmm4
; AVX-NEXT:    vpshufb %xmm6, %xmm2, %xmm2
; AVX-NEXT:    vpshufb %xmm6, %xmm5, %xmm5
; AVX-NEXT:    vpalignr {{.*#+}} xmm6 = xmm5[11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpalignr {{.*#+}} xmm7 = xmm2[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpalignr {{.*#+}} xmm3 = xmm3[11,12,13,14,15],xmm4[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm8
; AVX-NEXT:    vpalignr {{.*#+}} xmm4 = xmm4[11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpalignr {{.*#+}} xmm1 = xmm1[11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vinsertf128 $1, %xmm4, %ymm1, %ymm2
; AVX-NEXT:    vpalignr {{.*#+}} xmm1 = xmm7[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpalignr {{.*#+}} xmm4 = xmm6[11,12,13,14,15],xmm4[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vbroadcastf128 {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; AVX-NEXT:    # ymm5 = mem[0,1,0,1]
; AVX-NEXT:    vandnps %ymm2, %ymm5, %ymm2
; AVX-NEXT:    vandps %ymm5, %ymm8, %ymm5
; AVX-NEXT:    vorps %ymm2, %ymm5, %ymm2
; AVX-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX-NEXT:    vpalignr {{.*#+}} xmm3 = xmm3[11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpalignr {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9]
; AVX-NEXT:    vmovdqa %xmm3, 16(%rsi)
; AVX-NEXT:    vmovdqa %xmm0, (%rsi)
; AVX-NEXT:    vmovaps %ymm2, (%rdx)
; AVX-NEXT:    vmovdqa %xmm4, 16(%rcx)
; AVX-NEXT:    vmovdqa %xmm1, (%rcx)
; AVX-NEXT:    vzeroupper
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i8_stride3_vf32:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX2-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX2-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX2-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX2-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX2-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX2-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
; AVX2-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
; AVX2-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
; AVX2-NEXT:    vpalignr {{.*#+}} ymm3 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-NEXT:    vpalignr {{.*#+}} ymm2 = ymm3[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-NEXT:    vbroadcasti128 {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; AVX2-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX2-NEXT:    vpblendvb %ymm4, %ymm0, %ymm1, %ymm1
; AVX2-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX2-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX2-NEXT:    vmovdqa %ymm1, (%rdx)
; AVX2-NEXT:    vmovdqa %ymm2, (%rcx)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i8_stride3_vf32:
; 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:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX2-FP-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX2-FP-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX2-FP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX2-FP-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
; AVX2-FP-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
; AVX2-FP-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
; AVX2-FP-NEXT:    vpalignr {{.*#+}} ymm3 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FP-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FP-NEXT:    vpalignr {{.*#+}} ymm2 = ymm3[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FP-NEXT:    vbroadcasti128 {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; AVX2-FP-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpblendvb %ymm4, %ymm0, %ymm1, %ymm1
; AVX2-FP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX2-FP-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX2-FP-NEXT:    vmovdqa %ymm1, (%rdx)
; AVX2-FP-NEXT:    vmovdqa %ymm2, (%rcx)
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i8_stride3_vf32:
; 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:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX2-FCP-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX2-FCP-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX2-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
; AVX2-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
; AVX2-FCP-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} ymm3 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} ymm2 = ymm3[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; AVX2-FCP-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpblendvb %ymm4, %ymm0, %ymm1, %ymm1
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX2-FCP-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX2-FCP-NEXT:    vmovdqa %ymm1, (%rdx)
; AVX2-FCP-NEXT:    vmovdqa %ymm2, (%rcx)
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i8_stride3_vf32:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX512-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
; AVX512-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
; AVX512-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
; AVX512-NEXT:    vpalignr {{.*#+}} ymm3 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vpalignr {{.*#+}} ymm2 = ymm3[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vbroadcasti128 {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; AVX512-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX512-NEXT:    vpternlogq $202, %ymm1, %ymm0, %ymm4
; AVX512-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX512-NEXT:    vmovdqa %ymm4, (%rdx)
; AVX512-NEXT:    vmovdqa %ymm2, (%rcx)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i8_stride3_vf32:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512-FCP-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512-FCP-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512-FCP-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX512-FCP-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
; AVX512-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
; AVX512-FCP-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm3 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm2 = ymm3[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; AVX512-FCP-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX512-FCP-NEXT:    vpternlogq $202, %ymm1, %ymm0, %ymm4
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512-FCP-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa %ymm4, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa %ymm2, (%rcx)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i8_stride3_vf32:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512DQ-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512DQ-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512DQ-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512DQ-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm3 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm2 = ymm3[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vbroadcasti128 {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; AVX512DQ-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX512DQ-NEXT:    vpternlogq $202, %ymm1, %ymm0, %ymm4
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX512DQ-NEXT:    vmovdqa %ymm4, (%rdx)
; AVX512DQ-NEXT:    vmovdqa %ymm2, (%rcx)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i8_stride3_vf32:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-FCP-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512DQ-FCP-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512DQ-FCP-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512DQ-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
; AVX512DQ-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
; AVX512DQ-FCP-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm3 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm2 = ymm3[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; AVX512DQ-FCP-NEXT:    # ymm4 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT:    vpternlogq $202, %ymm1, %ymm0, %ymm4
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm4, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm2, (%rcx)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i8_stride3_vf32:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512BW-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512BW-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512BW-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512BW-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512BW-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512BW-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX512BW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
; AVX512BW-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
; AVX512BW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm3 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX512BW-NEXT:    movl $-134154240, %eax # imm = 0xF800F800
; AVX512BW-NEXT:    kmovd %eax, %k1
; AVX512BW-NEXT:    vpblendmb %ymm1, %ymm0, %ymm2 {%k1}
; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm3[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX512BW-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512BW-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX512BW-NEXT:    vmovdqa %ymm2, (%rdx)
; AVX512BW-NEXT:    vmovdqa %ymm1, (%rcx)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i8_stride3_vf32:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512BW-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512BW-FCP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512BW-FCP-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512BW-FCP-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512BW-FCP-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX512BW-FCP-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
; AVX512BW-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
; AVX512BW-FCP-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} ymm3 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX512BW-FCP-NEXT:    movl $-134154240, %eax # imm = 0xF800F800
; AVX512BW-FCP-NEXT:    kmovd %eax, %k1
; AVX512BW-FCP-NEXT:    vpblendmb %ymm1, %ymm0, %ymm2 {%k1}
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} ymm1 = ymm3[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512BW-FCP-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa %ymm2, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa %ymm1, (%rcx)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i8_stride3_vf32:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-BW-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-BW-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-BW-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512DQ-BW-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512DQ-BW-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512DQ-BW-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512DQ-BW-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX512DQ-BW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
; AVX512DQ-BW-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
; AVX512DQ-BW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} ymm3 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-BW-NEXT:    movl $-134154240, %eax # imm = 0xF800F800
; AVX512DQ-BW-NEXT:    kmovd %eax, %k1
; AVX512DQ-BW-NEXT:    vpblendmb %ymm1, %ymm0, %ymm2 {%k1}
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} ymm1 = ymm3[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512DQ-BW-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa %ymm2, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa %ymm1, (%rcx)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i8_stride3_vf32:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512DQ-BW-FCP-NEXT:    # ymm3 = mem[0,1,0,1]
; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
; AVX512DQ-BW-FCP-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} ymm3 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-BW-FCP-NEXT:    movl $-134154240, %eax # imm = 0xF800F800
; AVX512DQ-BW-FCP-NEXT:    kmovd %eax, %k1
; AVX512DQ-BW-FCP-NEXT:    vpblendmb %ymm1, %ymm0, %ymm2 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} ymm1 = ymm3[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm2, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa %ymm1, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <96 x i8>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <96 x i8> %wide.vec, <96 x i8> 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 i8> %wide.vec, <96 x i8> 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 i8> %wide.vec, <96 x i8> 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 i8> %strided.vec0, ptr %out.vec0, align 64
  store <32 x i8> %strided.vec1, ptr %out.vec1, align 64
  store <32 x i8> %strided.vec2, ptr %out.vec2, align 64
  ret void
}

define void @load_i8_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
; SSE-LABEL: load_i8_stride3_vf64:
; SSE:       # %bb.0:
; SSE-NEXT:    subq $168, %rsp
; SSE-NEXT:    movdqa 80(%rdi), %xmm9
; SSE-NEXT:    movdqa (%rdi), %xmm0
; SSE-NEXT:    movdqa 16(%rdi), %xmm10
; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 48(%rdi), %xmm5
; SSE-NEXT:    movdqa 64(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255]
; SSE-NEXT:    movdqa %xmm2, %xmm6
; SSE-NEXT:    pandn %xmm1, %xmm6
; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,0,255,255,0,255,255,0,255,255,0,255,255,0,255,255]
; SSE-NEXT:    movdqa %xmm4, %xmm1
; SSE-NEXT:    pandn %xmm5, %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm2, %xmm1
; SSE-NEXT:    pandn %xmm5, %xmm1
; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pand %xmm2, %xmm5
; SSE-NEXT:    movdqa %xmm2, %xmm11
; SSE-NEXT:    por %xmm6, %xmm5
; SSE-NEXT:    pxor %xmm8, %xmm8
; SSE-NEXT:    movdqa %xmm5, %xmm1
; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm8[8],xmm1[9],xmm8[9],xmm1[10],xmm8[10],xmm1[11],xmm8[11],xmm1[12],xmm8[12],xmm1[13],xmm8[13],xmm1[14],xmm8[14],xmm1[15],xmm8[15]
; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,0,65535,65535,0,65535,65535,0]
; SSE-NEXT:    movdqa %xmm6, %xmm2
; SSE-NEXT:    pandn %xmm1, %xmm2
; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1],xmm5[2],xmm8[2],xmm5[3],xmm8[3],xmm5[4],xmm8[4],xmm5[5],xmm8[5],xmm5[6],xmm8[6],xmm5[7],xmm8[7]
; SSE-NEXT:    pand %xmm6, %xmm5
; SSE-NEXT:    por %xmm2, %xmm5
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,2,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
; SSE-NEXT:    packuswb %xmm1, %xmm2
; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; SSE-NEXT:    pand %xmm7, %xmm2
; SSE-NEXT:    movdqa %xmm9, %xmm13
; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm8[8],xmm13[9],xmm8[9],xmm13[10],xmm8[10],xmm13[11],xmm8[11],xmm13[12],xmm8[12],xmm13[13],xmm8[13],xmm13[14],xmm8[14],xmm13[15],xmm8[15]
; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,65535,0,65535,65535,0,65535,65535]
; SSE-NEXT:    movdqa %xmm12, %xmm1
; SSE-NEXT:    pandn %xmm13, %xmm1
; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3],xmm9[4],xmm8[4],xmm9[5],xmm8[5],xmm9[6],xmm8[6],xmm9[7],xmm8[7]
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pand %xmm12, %xmm9
; SSE-NEXT:    por %xmm1, %xmm9
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
; SSE-NEXT:    packuswb %xmm1, %xmm1
; SSE-NEXT:    movdqa %xmm7, %xmm3
; SSE-NEXT:    pandn %xmm1, %xmm3
; SSE-NEXT:    por %xmm2, %xmm3
; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm11, %xmm1
; SSE-NEXT:    pandn %xmm10, %xmm1
; SSE-NEXT:    movdqa %xmm4, %xmm2
; SSE-NEXT:    pandn %xmm0, %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm11, %xmm2
; SSE-NEXT:    pandn %xmm0, %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pand %xmm11, %xmm0
; SSE-NEXT:    por %xmm1, %xmm0
; SSE-NEXT:    movdqa %xmm0, %xmm1
; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm8[8],xmm1[9],xmm8[9],xmm1[10],xmm8[10],xmm1[11],xmm8[11],xmm1[12],xmm8[12],xmm1[13],xmm8[13],xmm1[14],xmm8[14],xmm1[15],xmm8[15]
; SSE-NEXT:    movdqa %xmm6, %xmm2
; SSE-NEXT:    pandn %xmm1, %xmm2
; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
; SSE-NEXT:    pand %xmm6, %xmm0
; SSE-NEXT:    por %xmm2, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
; SSE-NEXT:    packuswb %xmm1, %xmm0
; SSE-NEXT:    movdqa 32(%rdi), %xmm15
; SSE-NEXT:    movdqa %xmm15, %xmm10
; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm8[8],xmm10[9],xmm8[9],xmm10[10],xmm8[10],xmm10[11],xmm8[11],xmm10[12],xmm8[12],xmm10[13],xmm8[13],xmm10[14],xmm8[14],xmm10[15],xmm8[15]
; SSE-NEXT:    movdqa %xmm12, %xmm1
; SSE-NEXT:    pandn %xmm10, %xmm1
; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3],xmm15[4],xmm8[4],xmm15[5],xmm8[5],xmm15[6],xmm8[6],xmm15[7],xmm8[7]
; SSE-NEXT:    movdqa %xmm15, %xmm2
; SSE-NEXT:    pand %xmm12, %xmm2
; SSE-NEXT:    por %xmm1, %xmm2
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
; SSE-NEXT:    packuswb %xmm1, %xmm1
; SSE-NEXT:    movdqa %xmm7, %xmm2
; SSE-NEXT:    pandn %xmm1, %xmm2
; SSE-NEXT:    pand %xmm7, %xmm0
; SSE-NEXT:    por %xmm0, %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 160(%rdi), %xmm14
; SSE-NEXT:    movdqa %xmm11, %xmm0
; SSE-NEXT:    pandn %xmm14, %xmm0
; SSE-NEXT:    movdqa 144(%rdi), %xmm1
; SSE-NEXT:    movdqa %xmm4, %xmm2
; SSE-NEXT:    pandn %xmm1, %xmm2
; SSE-NEXT:    movdqa %xmm2, (%rsp) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm11, %xmm5
; SSE-NEXT:    movdqa %xmm11, %xmm2
; SSE-NEXT:    pandn %xmm1, %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pand %xmm11, %xmm1
; SSE-NEXT:    por %xmm0, %xmm1
; SSE-NEXT:    movdqa %xmm1, %xmm0
; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
; SSE-NEXT:    movdqa %xmm6, %xmm2
; SSE-NEXT:    pandn %xmm0, %xmm2
; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
; SSE-NEXT:    pand %xmm6, %xmm1
; SSE-NEXT:    por %xmm2, %xmm1
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
; SSE-NEXT:    packuswb %xmm0, %xmm9
; SSE-NEXT:    movdqa 176(%rdi), %xmm3
; SSE-NEXT:    movdqa %xmm3, %xmm1
; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm8[8],xmm1[9],xmm8[9],xmm1[10],xmm8[10],xmm1[11],xmm8[11],xmm1[12],xmm8[12],xmm1[13],xmm8[13],xmm1[14],xmm8[14],xmm1[15],xmm8[15]
; SSE-NEXT:    movdqa %xmm12, %xmm0
; SSE-NEXT:    pandn %xmm1, %xmm0
; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3],xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
; SSE-NEXT:    movdqa %xmm3, %xmm11
; SSE-NEXT:    pand %xmm12, %xmm11
; SSE-NEXT:    por %xmm0, %xmm11
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
; SSE-NEXT:    packuswb %xmm0, %xmm0
; SSE-NEXT:    movdqa %xmm7, %xmm2
; SSE-NEXT:    pandn %xmm0, %xmm2
; SSE-NEXT:    pand %xmm7, %xmm9
; SSE-NEXT:    por %xmm9, %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa 112(%rdi), %xmm9
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm5, %xmm0
; SSE-NEXT:    pandn %xmm9, %xmm0
; SSE-NEXT:    movdqa 96(%rdi), %xmm9
; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pand %xmm5, %xmm9
; SSE-NEXT:    por %xmm0, %xmm9
; SSE-NEXT:    movdqa %xmm9, %xmm0
; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
; SSE-NEXT:    movdqa %xmm6, %xmm11
; SSE-NEXT:    pandn %xmm0, %xmm11
; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3],xmm9[4],xmm8[4],xmm9[5],xmm8[5],xmm9[6],xmm8[6],xmm9[7],xmm8[7]
; SSE-NEXT:    pand %xmm6, %xmm9
; SSE-NEXT:    por %xmm11, %xmm9
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,2,1,3]
; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[0,3,2,1,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
; SSE-NEXT:    packuswb %xmm0, %xmm9
; SSE-NEXT:    movdqa 128(%rdi), %xmm2
; SSE-NEXT:    movdqa %xmm2, %xmm11
; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm8[8],xmm11[9],xmm8[9],xmm11[10],xmm8[10],xmm11[11],xmm8[11],xmm11[12],xmm8[12],xmm11[13],xmm8[13],xmm11[14],xmm8[14],xmm11[15],xmm8[15]
; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm12, %xmm0
; SSE-NEXT:    pandn %xmm11, %xmm0
; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3],xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa %xmm2, %xmm11
; SSE-NEXT:    pand %xmm12, %xmm11
; SSE-NEXT:    por %xmm0, %xmm11
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
; SSE-NEXT:    packuswb %xmm0, %xmm0
; SSE-NEXT:    movdqa %xmm7, %xmm2
; SSE-NEXT:    pandn %xmm0, %xmm2
; SSE-NEXT:    pand %xmm7, %xmm9
; SSE-NEXT:    por %xmm9, %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    pand %xmm4, %xmm0
; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    movdqa %xmm0, %xmm9
; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm8[8],xmm9[9],xmm8[9],xmm9[10],xmm8[10],xmm9[11],xmm8[11],xmm9[12],xmm8[12],xmm9[13],xmm8[13],xmm9[14],xmm8[14],xmm9[15],xmm8[15]
; SSE-NEXT:    movdqa %xmm12, %xmm11
; SSE-NEXT:    pandn %xmm9, %xmm11
; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
; SSE-NEXT:    pand %xmm12, %xmm0
; SSE-NEXT:    por %xmm11, %xmm0
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,7,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[1,2,3,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm9, %xmm0
; SSE-NEXT:    movdqa %xmm12, %xmm9
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT:    pandn %xmm5, %xmm9
; SSE-NEXT:    movdqa %xmm6, %xmm2
; SSE-NEXT:    pandn %xmm13, %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    pand %xmm12, %xmm13
; SSE-NEXT:    por %xmm9, %xmm13
; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm13[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,6,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,5,6,7,4]
; SSE-NEXT:    packuswb %xmm9, %xmm9
; SSE-NEXT:    movdqa %xmm7, %xmm2
; SSE-NEXT:    pandn %xmm9, %xmm2
; SSE-NEXT:    pand %xmm7, %xmm0
; SSE-NEXT:    por %xmm0, %xmm2
; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    pand %xmm4, %xmm0
; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    movdqa %xmm0, %xmm9
; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm8[8],xmm9[9],xmm8[9],xmm9[10],xmm8[10],xmm9[11],xmm8[11],xmm9[12],xmm8[12],xmm9[13],xmm8[13],xmm9[14],xmm8[14],xmm9[15],xmm8[15]
; SSE-NEXT:    movdqa %xmm12, %xmm11
; SSE-NEXT:    pandn %xmm9, %xmm11
; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
; SSE-NEXT:    pand %xmm12, %xmm0
; SSE-NEXT:    por %xmm11, %xmm0
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,7,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[1,2,3,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm9, %xmm0
; SSE-NEXT:    movdqa %xmm12, %xmm11
; SSE-NEXT:    pandn %xmm15, %xmm11
; SSE-NEXT:    movdqa %xmm6, %xmm9
; SSE-NEXT:    pandn %xmm10, %xmm9
; SSE-NEXT:    pand %xmm12, %xmm10
; SSE-NEXT:    por %xmm11, %xmm10
; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,6,7,4]
; SSE-NEXT:    packuswb %xmm10, %xmm11
; SSE-NEXT:    movdqa %xmm7, %xmm10
; SSE-NEXT:    pandn %xmm11, %xmm10
; SSE-NEXT:    pand %xmm7, %xmm0
; SSE-NEXT:    por %xmm0, %xmm10
; SSE-NEXT:    movdqa %xmm14, %xmm0
; SSE-NEXT:    pand %xmm4, %xmm0
; SSE-NEXT:    por (%rsp), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    movdqa %xmm0, %xmm11
; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm8[8],xmm11[9],xmm8[9],xmm11[10],xmm8[10],xmm11[11],xmm8[11],xmm11[12],xmm8[12],xmm11[13],xmm8[13],xmm11[14],xmm8[14],xmm11[15],xmm8[15]
; SSE-NEXT:    movdqa %xmm12, %xmm13
; SSE-NEXT:    pandn %xmm11, %xmm13
; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
; SSE-NEXT:    pand %xmm12, %xmm0
; SSE-NEXT:    por %xmm13, %xmm0
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,7,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[1,2,3,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm11, %xmm0
; SSE-NEXT:    movdqa %xmm12, %xmm13
; SSE-NEXT:    pandn %xmm3, %xmm13
; SSE-NEXT:    movdqa %xmm6, %xmm11
; SSE-NEXT:    pandn %xmm1, %xmm11
; SSE-NEXT:    pand %xmm12, %xmm1
; SSE-NEXT:    por %xmm13, %xmm1
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
; SSE-NEXT:    packuswb %xmm1, %xmm1
; SSE-NEXT:    movdqa %xmm7, %xmm13
; SSE-NEXT:    pandn %xmm1, %xmm13
; SSE-NEXT:    pand %xmm7, %xmm0
; SSE-NEXT:    por %xmm0, %xmm13
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    pand %xmm4, %xmm0
; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT:    por %xmm0, %xmm4
; SSE-NEXT:    movdqa %xmm4, %xmm0
; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
; SSE-NEXT:    movdqa %xmm12, %xmm1
; SSE-NEXT:    pandn %xmm0, %xmm1
; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3],xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
; SSE-NEXT:    pand %xmm12, %xmm4
; SSE-NEXT:    por %xmm1, %xmm4
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm4[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,1]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm0, %xmm1
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    pand %xmm12, %xmm0
; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; SSE-NEXT:    por %xmm0, %xmm12
; SSE-NEXT:    pand %xmm7, %xmm1
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm12[0,3,2,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
; SSE-NEXT:    packuswb %xmm0, %xmm0
; SSE-NEXT:    pandn %xmm0, %xmm7
; SSE-NEXT:    por %xmm1, %xmm7
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255]
; SSE-NEXT:    pand %xmm2, %xmm0
; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    movdqa %xmm0, %xmm4
; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm8[8],xmm4[9],xmm8[9],xmm4[10],xmm8[10],xmm4[11],xmm8[11],xmm4[12],xmm8[12],xmm4[13],xmm8[13],xmm4[14],xmm8[14],xmm4[15],xmm8[15]
; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [0,65535,65535,0,65535,65535,0,65535]
; SSE-NEXT:    movdqa %xmm12, %xmm1
; SSE-NEXT:    pandn %xmm4, %xmm1
; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
; SSE-NEXT:    pand %xmm12, %xmm0
; SSE-NEXT:    por %xmm1, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm1, %xmm0
; SSE-NEXT:    movdqa %xmm5, %xmm1
; SSE-NEXT:    pand %xmm6, %xmm1
; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT:    packuswb %xmm1, %xmm1
; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,65535,0,0,0]
; SSE-NEXT:    movdqa %xmm4, %xmm5
; SSE-NEXT:    pandn %xmm1, %xmm5
; SSE-NEXT:    pand %xmm4, %xmm0
; SSE-NEXT:    por %xmm0, %xmm5
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT:    pand %xmm2, %xmm0
; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT:    movdqa %xmm0, %xmm2
; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
; SSE-NEXT:    movdqa %xmm12, %xmm1
; SSE-NEXT:    pandn %xmm2, %xmm1
; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
; SSE-NEXT:    pand %xmm12, %xmm0
; SSE-NEXT:    por %xmm1, %xmm0
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm0, %xmm1
; SSE-NEXT:    pand %xmm6, %xmm15
; SSE-NEXT:    por %xmm9, %xmm15
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
; SSE-NEXT:    packuswb %xmm0, %xmm0
; SSE-NEXT:    movdqa %xmm4, %xmm9
; SSE-NEXT:    pandn %xmm0, %xmm9
; SSE-NEXT:    pand %xmm4, %xmm1
; SSE-NEXT:    por %xmm1, %xmm9
; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255]
; SSE-NEXT:    pand %xmm2, %xmm14
; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; SSE-NEXT:    movdqa %xmm14, %xmm0
; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
; SSE-NEXT:    movdqa %xmm12, %xmm1
; SSE-NEXT:    pandn %xmm0, %xmm1
; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm8[0],xmm14[1],xmm8[1],xmm14[2],xmm8[2],xmm14[3],xmm8[3],xmm14[4],xmm8[4],xmm14[5],xmm8[5],xmm14[6],xmm8[6],xmm14[7],xmm8[7]
; SSE-NEXT:    pand %xmm12, %xmm14
; SSE-NEXT:    por %xmm1, %xmm14
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[3,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm0, %xmm1
; SSE-NEXT:    pand %xmm6, %xmm3
; SSE-NEXT:    por %xmm11, %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
; SSE-NEXT:    packuswb %xmm0, %xmm3
; SSE-NEXT:    movdqa %xmm4, %xmm0
; SSE-NEXT:    pandn %xmm3, %xmm0
; SSE-NEXT:    pand %xmm4, %xmm1
; SSE-NEXT:    por %xmm1, %xmm0
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    pand %xmm2, %xmm1
; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT:    por %xmm1, %xmm2
; SSE-NEXT:    movdqa %xmm2, %xmm1
; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm8[8],xmm1[9],xmm8[9],xmm1[10],xmm8[10],xmm1[11],xmm8[11],xmm1[12],xmm8[12],xmm1[13],xmm8[13],xmm1[14],xmm8[14],xmm1[15],xmm8[15]
; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3],xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
; SSE-NEXT:    pand %xmm12, %xmm2
; SSE-NEXT:    pandn %xmm1, %xmm12
; SSE-NEXT:    por %xmm2, %xmm12
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm12[3,1,2,0]
; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,5]
; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,1,2,0]
; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,4,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT:    packuswb %xmm1, %xmm3
; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    pand %xmm6, %xmm1
; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT:    por %xmm1, %xmm6
; SSE-NEXT:    pand %xmm4, %xmm3
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT:    packuswb %xmm1, %xmm1
; SSE-NEXT:    pandn %xmm1, %xmm4
; SSE-NEXT:    por %xmm3, %xmm4
; 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, 48(%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, (%rsi)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 16(%rsi)
; SSE-NEXT:    movdqa %xmm7, 32(%rdx)
; SSE-NEXT:    movdqa %xmm13, 48(%rdx)
; SSE-NEXT:    movdqa %xmm10, (%rdx)
; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT:    movaps %xmm1, 16(%rdx)
; SSE-NEXT:    movdqa %xmm4, 32(%rcx)
; SSE-NEXT:    movdqa %xmm0, 48(%rcx)
; SSE-NEXT:    movdqa %xmm9, (%rcx)
; SSE-NEXT:    movdqa %xmm5, 16(%rcx)
; SSE-NEXT:    addq $168, %rsp
; SSE-NEXT:    retq
;
; AVX-LABEL: load_i8_stride3_vf64:
; AVX:       # %bb.0:
; AVX-NEXT:    vmovdqa (%rdi), %xmm6
; AVX-NEXT:    vmovdqa 16(%rdi), %xmm0
; AVX-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovdqa 32(%rdi), %xmm4
; AVX-NEXT:    vmovdqa 48(%rdi), %xmm7
; AVX-NEXT:    vmovdqa 64(%rdi), %xmm2
; AVX-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovdqa 96(%rdi), %xmm11
; AVX-NEXT:    vmovdqa 112(%rdi), %xmm3
; AVX-NEXT:    vmovdqa 144(%rdi), %xmm10
; AVX-NEXT:    vmovdqa 160(%rdi), %xmm1
; AVX-NEXT:    vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vmovdqa {{.*#+}} xmm12 = [128,128,128,128,128,0,3,6,9,12,15,2,5,8,11,14]
; AVX-NEXT:    vpshufb %xmm12, %xmm6, %xmm5
; AVX-NEXT:    vpshufb %xmm12, %xmm10, %xmm8
; AVX-NEXT:    vpshufb %xmm12, %xmm11, %xmm9
; AVX-NEXT:    vpshufb %xmm12, %xmm7, %xmm13
; AVX-NEXT:    vmovdqa {{.*#+}} xmm14 = [1,4,7,10,13,128,128,128,128,128,128,u,u,u,u,u]
; AVX-NEXT:    vpshufb %xmm14, %xmm6, %xmm6
; AVX-NEXT:    vmovdqa {{.*#+}} xmm15 = [128,128,128,128,128,0,3,6,9,12,15,u,u,u,u,u]
; AVX-NEXT:    vpshufb %xmm15, %xmm0, %xmm12
; AVX-NEXT:    vpor %xmm6, %xmm12, %xmm6
; AVX-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vpshufb %xmm14, %xmm10, %xmm10
; AVX-NEXT:    vpshufb %xmm15, %xmm1, %xmm12
; AVX-NEXT:    vpor %xmm10, %xmm12, %xmm6
; AVX-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vpshufb %xmm14, %xmm11, %xmm11
; AVX-NEXT:    vpshufb %xmm15, %xmm3, %xmm12
; AVX-NEXT:    vpor %xmm11, %xmm12, %xmm6
; AVX-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vpshufb %xmm14, %xmm7, %xmm7
; AVX-NEXT:    vpshufb %xmm15, %xmm2, %xmm12
; AVX-NEXT:    vpor %xmm7, %xmm12, %xmm6
; AVX-NEXT:    vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT:    vpshufb %xmm14, %xmm0, %xmm12
; AVX-NEXT:    vpshufb %xmm15, %xmm4, %xmm6
; AVX-NEXT:    vpor %xmm6, %xmm12, %xmm6
; AVX-NEXT:    vmovdqa 176(%rdi), %xmm12
; AVX-NEXT:    vpshufb %xmm14, %xmm1, %xmm0
; AVX-NEXT:    vpshufb %xmm15, %xmm12, %xmm10
; AVX-NEXT:    vpor %xmm0, %xmm10, %xmm1
; AVX-NEXT:    vmovdqa 128(%rdi), %xmm10
; AVX-NEXT:    vpshufb %xmm14, %xmm3, %xmm0
; AVX-NEXT:    vpshufb %xmm15, %xmm10, %xmm11
; AVX-NEXT:    vpor %xmm0, %xmm11, %xmm0
; AVX-NEXT:    vpshufb %xmm14, %xmm2, %xmm11
; AVX-NEXT:    vmovdqa 80(%rdi), %xmm14
; AVX-NEXT:    vpshufb %xmm15, %xmm14, %xmm15
; AVX-NEXT:    vpor %xmm11, %xmm15, %xmm11
; AVX-NEXT:    vmovdqa {{.*#+}} xmm15 = [1,4,7,10,13,128,128,128,128,128,128,128,128,128,128,128]
; AVX-NEXT:    vpshufb %xmm15, %xmm14, %xmm7
; AVX-NEXT:    vpor %xmm7, %xmm13, %xmm7
; AVX-NEXT:    vpalignr {{.*#+}} xmm13 = xmm13[11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpshufb %xmm15, %xmm10, %xmm11
; AVX-NEXT:    vpor %xmm11, %xmm9, %xmm11
; AVX-NEXT:    vpalignr {{.*#+}} xmm9 = xmm9[11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpshufb %xmm15, %xmm12, %xmm0
; AVX-NEXT:    vpor %xmm0, %xmm8, %xmm2
; AVX-NEXT:    vpalignr {{.*#+}} xmm8 = xmm8[11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vpshufb %xmm15, %xmm4, %xmm1
; AVX-NEXT:    vpor %xmm1, %xmm5, %xmm1
; AVX-NEXT:    vpalignr {{.*#+}} xmm5 = xmm5[11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7,8,9,10]
; AVX-NEXT:    vmovdqa {{.*#+}} xmm6 = [0,1,2,3,4,5,6,7,8,9,10,128,128,128,128,128]
; AVX-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX-NEXT:    vpshufb %xmm6, %xmm0, %xmm15
; AVX-NEXT:    vmovdqa {{.*#+}} xmm0 = [128,128,128,128,128,128,128,128,128,128,128,2,5,8,11,14]
; AVX-NEXT:    vpshufb %xmm0, %xmm14, %xmm14
; AVX-NEXT:    vpor %xmm14, %xmm15, %xmm14
; AVX-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    vpshufb %xmm6, %xmm15, %xmm15
; AVX-NEXT:    vpshufb %xmm0, %xmm10, %xmm10
; AVX-NEXT:    vpor %xmm10, %xmm15, %xmm10
; AVX-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    vpshufb %xmm6, %xmm15, %xmm15
; AVX-NEXT:    vpshufb %xmm0, %xmm12, %xmm12
; AVX-NEXT:    vpor %xmm12, %xmm15, %xmm12
; AVX-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    vpshufb %xmm6, %xmm15, %xmm6
; AVX-NEXT:    vpshufb %xmm0, %xmm4, %xmm0
; AVX-NEXT:    vpor %xmm0, %xmm6, %xmm0
; AVX-NEXT:    vmovdqa {{.*#+}} xmm4 = [5,6,7,8,9,10,128,128,128,128,128,0,1,2,3,4]
; AVX-NEXT:    vpshufb %xmm4, %xmm7, %xmm6
; AVX-NEXT:    vmovdqa {{.*#+}} xmm7 = [128,128,128,128,128,128,2,5,8,11,14,128,128,128,128,128]
; AVX-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX-NEXT:    vpshufb %xmm7, %xmm15, %xmm15
; AVX-NEXT:    vpor %xmm6, %xmm15, %xmm6
; AVX-NEXT:    vpshufb %xmm4, %xmm11, %xmm11
; AVX-NEXT:    vpshufb %xmm7, %xmm3, %xmm15
; AVX-NEXT:    vpor %xmm15, %xmm11, %xmm11
; AVX-NEXT:    vpshufb %xmm4, %xmm2, %xmm2
; AVX-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX-NEXT:    vpshufb %xmm7, %xmm3, %xmm3
; AVX-NEXT:    vpor %xmm3, %xmm2, %xmm2
; AVX-NEXT:    vpshufb %xmm4, %xmm1, %xmm1
; AVX-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX-NEXT:    vpshufb %xmm7, %xmm3, %xmm3
; AVX-NEXT:    vpor %xmm3, %xmm1, %xmm1
; AVX-NEXT:    vmovdqa %xmm1, (%rsi)
; AVX-NEXT:    vmovdqa %xmm2, 48(%rsi)
; AVX-NEXT:    vmovdqa %xmm11, 32(%rsi)
; AVX-NEXT:    vmovdqa %xmm6, 16(%rsi)
; AVX-NEXT:    vmovdqa %xmm0, (%rdx)
; AVX-NEXT:    vmovdqa %xmm12, 48(%rdx)
; AVX-NEXT:    vmovdqa %xmm10, 32(%rdx)
; AVX-NEXT:    vmovdqa %xmm14, 16(%rdx)
; AVX-NEXT:    vmovdqa %xmm5, (%rcx)
; AVX-NEXT:    vmovdqa %xmm8, 48(%rcx)
; AVX-NEXT:    vmovdqa %xmm9, 32(%rcx)
; AVX-NEXT:    vmovdqa %xmm13, 16(%rcx)
; AVX-NEXT:    retq
;
; AVX2-LABEL: load_i8_stride3_vf64:
; AVX2:       # %bb.0:
; AVX2-NEXT:    vmovdqa (%rdi), %xmm0
; AVX2-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX2-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX2-NEXT:    vmovdqa 96(%rdi), %xmm3
; AVX2-NEXT:    vmovdqa 112(%rdi), %xmm4
; AVX2-NEXT:    vmovdqa 128(%rdi), %xmm5
; AVX2-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm6
; AVX2-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm0
; AVX2-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm1
; AVX2-NEXT:    vinserti128 $1, 144(%rdi), %ymm3, %ymm2
; AVX2-NEXT:    vinserti128 $1, 160(%rdi), %ymm4, %ymm3
; AVX2-NEXT:    vinserti128 $1, 176(%rdi), %ymm5, %ymm4
; AVX2-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [255,0,255,255,0,255,255,0,255,255,0,255,255,0,255,255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255,255]
; AVX2-NEXT:    # ymm5 = mem[0,1,0,1]
; AVX2-NEXT:    vpblendvb %ymm5, %ymm6, %ymm1, %ymm7
; AVX2-NEXT:    vbroadcasti128 {{.*#+}} ymm8 = [1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14]
; AVX2-NEXT:    # ymm8 = mem[0,1,0,1]
; AVX2-NEXT:    vpshufb %ymm8, %ymm7, %ymm7
; AVX2-NEXT:    vpblendvb %ymm5, %ymm2, %ymm4, %ymm5
; AVX2-NEXT:    vpshufb %ymm8, %ymm5, %ymm5
; AVX2-NEXT:    vbroadcasti128 {{.*#+}} ymm8 = [255,0,0,255,0,0,255,0,0,255,0,0,255,0,0,255,255,0,0,255,0,0,255,0,0,255,0,0,255,0,0,255]
; AVX2-NEXT:    # ymm8 = mem[0,1,0,1]
; AVX2-NEXT:    vpblendvb %ymm8, %ymm0, %ymm6, %ymm6
; AVX2-NEXT:    vpblendvb %ymm8, %ymm3, %ymm2, %ymm2
; AVX2-NEXT:    vpblendvb %ymm8, %ymm1, %ymm0, %ymm9
; AVX2-NEXT:    vbroadcasti128 {{.*#+}} ymm10 = [1,4,7,10,13,0,3,6,9,12,15,0,0,0,0,0,1,4,7,10,13,0,3,6,9,12,15,0,0,0,0,0]
; AVX2-NEXT:    # ymm10 = mem[0,1,0,1]
; AVX2-NEXT:    vpshufb %ymm10, %ymm9, %ymm9
; AVX2-NEXT:    vpblendvb %ymm8, %ymm4, %ymm3, %ymm8
; AVX2-NEXT:    vpshufb %ymm10, %ymm8, %ymm8
; AVX2-NEXT:    vpalignr {{.*#+}} ymm8 = ymm5[11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7,8,9,10],ymm5[27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-NEXT:    vpalignr {{.*#+}} ymm9 = ymm7[11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7,8,9,10],ymm7[27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-NEXT:    vmovdqa {{.*#+}} ymm10 = [1,4,7,10,13,0,3,6,9,12,15,128,128,128,128,128,17,20,23,26,29,16,19,22,25,28,31,128,128,128,128,128]
; AVX2-NEXT:    vpshufb %ymm10, %ymm2, %ymm2
; AVX2-NEXT:    vmovdqa {{.*#+}} ymm11 = [128,128,128,128,128,128,128,128,128,128,128,2,5,8,11,14,128,128,128,128,128,128,128,128,128,128,128,18,21,24,27,30]
; AVX2-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
; AVX2-NEXT:    vpor %ymm4, %ymm2, %ymm2
; AVX2-NEXT:    vpshufb %ymm10, %ymm6, %ymm4
; AVX2-NEXT:    vpshufb %ymm11, %ymm1, %ymm1
; AVX2-NEXT:    vpor %ymm1, %ymm4, %ymm1
; AVX2-NEXT:    vpalignr {{.*#+}} ymm4 = ymm5[5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,21,22,23,24,25,26,27,28,29,30,31,16,17,18,19,20]
; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm5 = [8,11,14,0,0,0,2,5,8,11,14,0,0,0,2,5,8,11,14,0,0,0,2,5,8,11,14,0,0,0,2,5]
; AVX2-NEXT:    vpshufb %ymm5, %ymm3, %ymm3
; AVX2-NEXT:    vbroadcasti128 {{.*#+}} ymm6 = [255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255]
; AVX2-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX2-NEXT:    vpblendvb %ymm6, %ymm4, %ymm3, %ymm3
; AVX2-NEXT:    vpalignr {{.*#+}} ymm4 = ymm7[5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,21,22,23,24,25,26,27,28,29,30,31,16,17,18,19,20]
; AVX2-NEXT:    vpshufb %ymm5, %ymm0, %ymm0
; AVX2-NEXT:    vpblendvb %ymm6, %ymm4, %ymm0, %ymm0
; AVX2-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX2-NEXT:    vmovdqa %ymm3, 32(%rsi)
; AVX2-NEXT:    vmovdqa %ymm1, (%rdx)
; AVX2-NEXT:    vmovdqa %ymm2, 32(%rdx)
; AVX2-NEXT:    vmovdqa %ymm9, (%rcx)
; AVX2-NEXT:    vmovdqa %ymm8, 32(%rcx)
; AVX2-NEXT:    vzeroupper
; AVX2-NEXT:    retq
;
; AVX2-FP-LABEL: load_i8_stride3_vf64:
; 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:    vmovdqa 96(%rdi), %xmm3
; AVX2-FP-NEXT:    vmovdqa 112(%rdi), %xmm4
; AVX2-FP-NEXT:    vmovdqa 128(%rdi), %xmm5
; AVX2-FP-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm6
; AVX2-FP-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm0
; AVX2-FP-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm1
; AVX2-FP-NEXT:    vinserti128 $1, 144(%rdi), %ymm3, %ymm2
; AVX2-FP-NEXT:    vinserti128 $1, 160(%rdi), %ymm4, %ymm3
; AVX2-FP-NEXT:    vinserti128 $1, 176(%rdi), %ymm5, %ymm4
; AVX2-FP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [255,0,255,255,0,255,255,0,255,255,0,255,255,0,255,255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255,255]
; AVX2-FP-NEXT:    # ymm5 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpblendvb %ymm5, %ymm6, %ymm1, %ymm7
; AVX2-FP-NEXT:    vbroadcasti128 {{.*#+}} ymm8 = [1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14]
; AVX2-FP-NEXT:    # ymm8 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpshufb %ymm8, %ymm7, %ymm7
; AVX2-FP-NEXT:    vpblendvb %ymm5, %ymm2, %ymm4, %ymm5
; AVX2-FP-NEXT:    vpshufb %ymm8, %ymm5, %ymm5
; AVX2-FP-NEXT:    vbroadcasti128 {{.*#+}} ymm8 = [255,0,0,255,0,0,255,0,0,255,0,0,255,0,0,255,255,0,0,255,0,0,255,0,0,255,0,0,255,0,0,255]
; AVX2-FP-NEXT:    # ymm8 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpblendvb %ymm8, %ymm0, %ymm6, %ymm6
; AVX2-FP-NEXT:    vpblendvb %ymm8, %ymm3, %ymm2, %ymm2
; AVX2-FP-NEXT:    vpblendvb %ymm8, %ymm1, %ymm0, %ymm9
; AVX2-FP-NEXT:    vbroadcasti128 {{.*#+}} ymm10 = [1,4,7,10,13,0,3,6,9,12,15,0,0,0,0,0,1,4,7,10,13,0,3,6,9,12,15,0,0,0,0,0]
; AVX2-FP-NEXT:    # ymm10 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpshufb %ymm10, %ymm9, %ymm9
; AVX2-FP-NEXT:    vpblendvb %ymm8, %ymm4, %ymm3, %ymm8
; AVX2-FP-NEXT:    vpshufb %ymm10, %ymm8, %ymm8
; AVX2-FP-NEXT:    vpalignr {{.*#+}} ymm8 = ymm5[11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7,8,9,10],ymm5[27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FP-NEXT:    vpalignr {{.*#+}} ymm9 = ymm7[11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7,8,9,10],ymm7[27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FP-NEXT:    vmovdqa {{.*#+}} ymm10 = [1,4,7,10,13,0,3,6,9,12,15,128,128,128,128,128,17,20,23,26,29,16,19,22,25,28,31,128,128,128,128,128]
; AVX2-FP-NEXT:    vpshufb %ymm10, %ymm2, %ymm2
; AVX2-FP-NEXT:    vmovdqa {{.*#+}} ymm11 = [128,128,128,128,128,128,128,128,128,128,128,2,5,8,11,14,128,128,128,128,128,128,128,128,128,128,128,18,21,24,27,30]
; AVX2-FP-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
; AVX2-FP-NEXT:    vpor %ymm4, %ymm2, %ymm2
; AVX2-FP-NEXT:    vpshufb %ymm10, %ymm6, %ymm4
; AVX2-FP-NEXT:    vpshufb %ymm11, %ymm1, %ymm1
; AVX2-FP-NEXT:    vpor %ymm1, %ymm4, %ymm1
; AVX2-FP-NEXT:    vpalignr {{.*#+}} ymm4 = ymm5[5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,21,22,23,24,25,26,27,28,29,30,31,16,17,18,19,20]
; AVX2-FP-NEXT:    vpbroadcastq {{.*#+}} ymm5 = [8,11,14,0,0,0,2,5,8,11,14,0,0,0,2,5,8,11,14,0,0,0,2,5,8,11,14,0,0,0,2,5]
; AVX2-FP-NEXT:    vpshufb %ymm5, %ymm3, %ymm3
; AVX2-FP-NEXT:    vbroadcasti128 {{.*#+}} ymm6 = [255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255]
; AVX2-FP-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX2-FP-NEXT:    vpblendvb %ymm6, %ymm4, %ymm3, %ymm3
; AVX2-FP-NEXT:    vpalignr {{.*#+}} ymm4 = ymm7[5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,21,22,23,24,25,26,27,28,29,30,31,16,17,18,19,20]
; AVX2-FP-NEXT:    vpshufb %ymm5, %ymm0, %ymm0
; AVX2-FP-NEXT:    vpblendvb %ymm6, %ymm4, %ymm0, %ymm0
; AVX2-FP-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX2-FP-NEXT:    vmovdqa %ymm3, 32(%rsi)
; AVX2-FP-NEXT:    vmovdqa %ymm1, (%rdx)
; AVX2-FP-NEXT:    vmovdqa %ymm2, 32(%rdx)
; AVX2-FP-NEXT:    vmovdqa %ymm9, (%rcx)
; AVX2-FP-NEXT:    vmovdqa %ymm8, 32(%rcx)
; AVX2-FP-NEXT:    vzeroupper
; AVX2-FP-NEXT:    retq
;
; AVX2-FCP-LABEL: load_i8_stride3_vf64:
; 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:    vmovdqa 96(%rdi), %xmm3
; AVX2-FCP-NEXT:    vmovdqa 112(%rdi), %xmm4
; AVX2-FCP-NEXT:    vmovdqa 128(%rdi), %xmm5
; AVX2-FCP-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm6
; AVX2-FCP-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm0
; AVX2-FCP-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm1
; AVX2-FCP-NEXT:    vinserti128 $1, 144(%rdi), %ymm3, %ymm2
; AVX2-FCP-NEXT:    vinserti128 $1, 160(%rdi), %ymm4, %ymm3
; AVX2-FCP-NEXT:    vinserti128 $1, 176(%rdi), %ymm5, %ymm4
; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm5 = [255,0,255,255,0,255,255,0,255,255,0,255,255,0,255,255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255,255]
; AVX2-FCP-NEXT:    # ymm5 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpblendvb %ymm5, %ymm6, %ymm1, %ymm7
; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm8 = [1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14]
; AVX2-FCP-NEXT:    # ymm8 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpshufb %ymm8, %ymm7, %ymm7
; AVX2-FCP-NEXT:    vpblendvb %ymm5, %ymm2, %ymm4, %ymm5
; AVX2-FCP-NEXT:    vpshufb %ymm8, %ymm5, %ymm5
; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm8 = [255,0,0,255,0,0,255,0,0,255,0,0,255,0,0,255,255,0,0,255,0,0,255,0,0,255,0,0,255,0,0,255]
; AVX2-FCP-NEXT:    # ymm8 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpblendvb %ymm8, %ymm0, %ymm6, %ymm6
; AVX2-FCP-NEXT:    vpblendvb %ymm8, %ymm3, %ymm2, %ymm2
; AVX2-FCP-NEXT:    vpblendvb %ymm8, %ymm1, %ymm0, %ymm9
; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm10 = [1,4,7,10,13,0,3,6,9,12,15,0,0,0,0,0,1,4,7,10,13,0,3,6,9,12,15,0,0,0,0,0]
; AVX2-FCP-NEXT:    # ymm10 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpshufb %ymm10, %ymm9, %ymm9
; AVX2-FCP-NEXT:    vpblendvb %ymm8, %ymm4, %ymm3, %ymm8
; AVX2-FCP-NEXT:    vpshufb %ymm10, %ymm8, %ymm8
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} ymm8 = ymm5[11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7,8,9,10],ymm5[27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} ymm9 = ymm7[11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7,8,9,10],ymm7[27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23,24,25,26]
; AVX2-FCP-NEXT:    vmovdqa {{.*#+}} ymm10 = [1,4,7,10,13,0,3,6,9,12,15,128,128,128,128,128,17,20,23,26,29,16,19,22,25,28,31,128,128,128,128,128]
; AVX2-FCP-NEXT:    vpshufb %ymm10, %ymm2, %ymm2
; AVX2-FCP-NEXT:    vmovdqa {{.*#+}} ymm11 = [128,128,128,128,128,128,128,128,128,128,128,2,5,8,11,14,128,128,128,128,128,128,128,128,128,128,128,18,21,24,27,30]
; AVX2-FCP-NEXT:    vpshufb %ymm11, %ymm4, %ymm4
; AVX2-FCP-NEXT:    vpor %ymm4, %ymm2, %ymm2
; AVX2-FCP-NEXT:    vpshufb %ymm10, %ymm6, %ymm4
; AVX2-FCP-NEXT:    vpshufb %ymm11, %ymm1, %ymm1
; AVX2-FCP-NEXT:    vpor %ymm1, %ymm4, %ymm1
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} ymm4 = ymm5[5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,21,22,23,24,25,26,27,28,29,30,31,16,17,18,19,20]
; AVX2-FCP-NEXT:    vpbroadcastq {{.*#+}} ymm5 = [8,11,14,0,0,0,2,5,8,11,14,0,0,0,2,5,8,11,14,0,0,0,2,5,8,11,14,0,0,0,2,5]
; AVX2-FCP-NEXT:    vpshufb %ymm5, %ymm3, %ymm3
; AVX2-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm6 = [255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255]
; AVX2-FCP-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX2-FCP-NEXT:    vpblendvb %ymm6, %ymm4, %ymm3, %ymm3
; AVX2-FCP-NEXT:    vpalignr {{.*#+}} ymm4 = ymm7[5,6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,21,22,23,24,25,26,27,28,29,30,31,16,17,18,19,20]
; AVX2-FCP-NEXT:    vpshufb %ymm5, %ymm0, %ymm0
; AVX2-FCP-NEXT:    vpblendvb %ymm6, %ymm4, %ymm0, %ymm0
; AVX2-FCP-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX2-FCP-NEXT:    vmovdqa %ymm3, 32(%rsi)
; AVX2-FCP-NEXT:    vmovdqa %ymm1, (%rdx)
; AVX2-FCP-NEXT:    vmovdqa %ymm2, 32(%rdx)
; AVX2-FCP-NEXT:    vmovdqa %ymm9, (%rcx)
; AVX2-FCP-NEXT:    vmovdqa %ymm8, 32(%rcx)
; AVX2-FCP-NEXT:    vzeroupper
; AVX2-FCP-NEXT:    retq
;
; AVX512-LABEL: load_i8_stride3_vf64:
; AVX512:       # %bb.0:
; AVX512-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512-NEXT:    vmovdqa 96(%rdi), %xmm3
; AVX512-NEXT:    vmovdqa 112(%rdi), %xmm4
; AVX512-NEXT:    vmovdqa 128(%rdi), %xmm5
; AVX512-NEXT:    vinserti128 $1, 144(%rdi), %ymm3, %ymm3
; AVX512-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512-NEXT:    vinserti128 $1, 160(%rdi), %ymm4, %ymm4
; AVX512-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512-NEXT:    vinserti128 $1, 176(%rdi), %ymm5, %ymm5
; AVX512-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512-NEXT:    vbroadcasti128 {{.*#+}} ymm6 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX512-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
; AVX512-NEXT:    vpshufb %ymm6, %ymm3, %ymm3
; AVX512-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
; AVX512-NEXT:    vpshufb %ymm6, %ymm4, %ymm4
; AVX512-NEXT:    vpshufb %ymm6, %ymm2, %ymm2
; AVX512-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
; AVX512-NEXT:    vpalignr {{.*#+}} ymm6 = ymm5[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm5[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vpalignr {{.*#+}} ymm7 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vpalignr {{.*#+}} ymm3 = ymm3[11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm8
; AVX512-NEXT:    vpalignr {{.*#+}} ymm4 = ymm4[11,12,13,14,15],ymm5[0,1,2,3,4,5,6,7,8,9,10],ymm4[27,28,29,30,31],ymm5[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vinserti64x4 $1, %ymm4, %zmm1, %zmm2
; AVX512-NEXT:    vpalignr {{.*#+}} ymm1 = ymm7[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm7[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vpalignr {{.*#+}} ymm4 = ymm6[11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10],ymm6[27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vbroadcasti32x4 {{.*#+}} zmm5 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; AVX512-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512-NEXT:    vpternlogq $202, %zmm2, %zmm8, %zmm5
; AVX512-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512-NEXT:    vpalignr {{.*#+}} ymm2 = ymm3[11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-NEXT:    vpalignr {{.*#+}} ymm2 = ymm2[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512-NEXT:    vmovdqa %ymm2, 32(%rsi)
; AVX512-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX512-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512-NEXT:    vmovdqa %ymm4, 32(%rcx)
; AVX512-NEXT:    vmovdqa %ymm1, (%rcx)
; AVX512-NEXT:    vzeroupper
; AVX512-NEXT:    retq
;
; AVX512-FCP-LABEL: load_i8_stride3_vf64:
; AVX512-FCP:       # %bb.0:
; AVX512-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512-FCP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512-FCP-NEXT:    vmovdqa 96(%rdi), %xmm3
; AVX512-FCP-NEXT:    vmovdqa 112(%rdi), %xmm4
; AVX512-FCP-NEXT:    vmovdqa 128(%rdi), %xmm5
; AVX512-FCP-NEXT:    vinserti128 $1, 144(%rdi), %ymm3, %ymm3
; AVX512-FCP-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512-FCP-NEXT:    vinserti128 $1, 160(%rdi), %ymm4, %ymm4
; AVX512-FCP-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512-FCP-NEXT:    vinserti128 $1, 176(%rdi), %ymm5, %ymm5
; AVX512-FCP-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm6 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512-FCP-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm3, %ymm3
; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm4, %ymm4
; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm2, %ymm2
; AVX512-FCP-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm6 = ymm5[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm5[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm7 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm3 = ymm3[11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm8
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm4 = ymm4[11,12,13,14,15],ymm5[0,1,2,3,4,5,6,7,8,9,10],ymm4[27,28,29,30,31],ymm5[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vinserti64x4 $1, %ymm4, %zmm1, %zmm2
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm1 = ymm7[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm7[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm4 = ymm6[11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10],ymm6[27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm5 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; AVX512-FCP-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512-FCP-NEXT:    vpternlogq $202, %zmm2, %zmm8, %zmm5
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm2 = ymm3[11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26]
; AVX512-FCP-NEXT:    vpalignr {{.*#+}} ymm2 = ymm2[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512-FCP-NEXT:    vmovdqa %ymm2, 32(%rsi)
; AVX512-FCP-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX512-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512-FCP-NEXT:    vmovdqa %ymm4, 32(%rcx)
; AVX512-FCP-NEXT:    vmovdqa %ymm1, (%rcx)
; AVX512-FCP-NEXT:    vzeroupper
; AVX512-FCP-NEXT:    retq
;
; AVX512DQ-LABEL: load_i8_stride3_vf64:
; AVX512DQ:       # %bb.0:
; AVX512DQ-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-NEXT:    vmovdqa 96(%rdi), %xmm3
; AVX512DQ-NEXT:    vmovdqa 112(%rdi), %xmm4
; AVX512DQ-NEXT:    vmovdqa 128(%rdi), %xmm5
; AVX512DQ-NEXT:    vinserti128 $1, 144(%rdi), %ymm3, %ymm3
; AVX512DQ-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512DQ-NEXT:    vinserti128 $1, 160(%rdi), %ymm4, %ymm4
; AVX512DQ-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512DQ-NEXT:    vinserti128 $1, 176(%rdi), %ymm5, %ymm5
; AVX512DQ-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512DQ-NEXT:    vbroadcasti128 {{.*#+}} ymm6 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512DQ-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm3, %ymm3
; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm4, %ymm4
; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm2, %ymm2
; AVX512DQ-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm6 = ymm5[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm5[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm7 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm3 = ymm3[11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm8
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm4 = ymm4[11,12,13,14,15],ymm5[0,1,2,3,4,5,6,7,8,9,10],ymm4[27,28,29,30,31],ymm5[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vinserti64x4 $1, %ymm4, %zmm1, %zmm2
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm1 = ymm7[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm7[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm4 = ymm6[11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10],ymm6[27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vbroadcasti32x4 {{.*#+}} zmm5 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; AVX512DQ-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQ-NEXT:    vpternlogq $202, %zmm2, %zmm8, %zmm5
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm2 = ymm3[11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-NEXT:    vpalignr {{.*#+}} ymm2 = ymm2[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512DQ-NEXT:    vmovdqa %ymm2, 32(%rsi)
; AVX512DQ-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX512DQ-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-NEXT:    vmovdqa %ymm4, 32(%rcx)
; AVX512DQ-NEXT:    vmovdqa %ymm1, (%rcx)
; AVX512DQ-NEXT:    vzeroupper
; AVX512DQ-NEXT:    retq
;
; AVX512DQ-FCP-LABEL: load_i8_stride3_vf64:
; AVX512DQ-FCP:       # %bb.0:
; AVX512DQ-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-FCP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-FCP-NEXT:    vmovdqa 96(%rdi), %xmm3
; AVX512DQ-FCP-NEXT:    vmovdqa 112(%rdi), %xmm4
; AVX512DQ-FCP-NEXT:    vmovdqa 128(%rdi), %xmm5
; AVX512DQ-FCP-NEXT:    vinserti128 $1, 144(%rdi), %ymm3, %ymm3
; AVX512DQ-FCP-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512DQ-FCP-NEXT:    vinserti128 $1, 160(%rdi), %ymm4, %ymm4
; AVX512DQ-FCP-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512DQ-FCP-NEXT:    vinserti128 $1, 176(%rdi), %ymm5, %ymm5
; AVX512DQ-FCP-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512DQ-FCP-NEXT:    vbroadcasti128 {{.*#+}} ymm6 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512DQ-FCP-NEXT:    # ymm6 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm0, %ymm0
; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm3, %ymm3
; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm1, %ymm1
; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm4, %ymm4
; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm2, %ymm2
; AVX512DQ-FCP-NEXT:    vpshufb %ymm6, %ymm5, %ymm5
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm6 = ymm5[11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10],ymm5[27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm7 = ymm2[11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10],ymm2[27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm3 = ymm3[11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm8
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm4 = ymm4[11,12,13,14,15],ymm5[0,1,2,3,4,5,6,7,8,9,10],ymm4[27,28,29,30,31],ymm5[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm1 = ymm1[11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10],ymm1[27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vinserti64x4 $1, %ymm4, %zmm1, %zmm2
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm1 = ymm7[11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10],ymm7[27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm4 = ymm6[11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10],ymm6[27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm5 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
; AVX512DQ-FCP-NEXT:    # zmm5 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQ-FCP-NEXT:    vpternlogq $202, %zmm2, %zmm8, %zmm5
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7,8,9,10],ymm0[27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm2 = ymm3[11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10],ymm3[27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26]
; AVX512DQ-FCP-NEXT:    vpalignr {{.*#+}} ymm2 = ymm2[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm2, 32(%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm0, (%rsi)
; AVX512DQ-FCP-NEXT:    vmovdqa64 %zmm5, (%rdx)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm4, 32(%rcx)
; AVX512DQ-FCP-NEXT:    vmovdqa %ymm1, (%rcx)
; AVX512DQ-FCP-NEXT:    vzeroupper
; AVX512DQ-FCP-NEXT:    retq
;
; AVX512BW-LABEL: load_i8_stride3_vf64:
; AVX512BW:       # %bb.0:
; AVX512BW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512BW-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512BW-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512BW-NEXT:    vmovdqa 96(%rdi), %xmm3
; AVX512BW-NEXT:    vmovdqa 112(%rdi), %xmm4
; AVX512BW-NEXT:    vmovdqa 128(%rdi), %xmm5
; AVX512BW-NEXT:    vinserti128 $1, 144(%rdi), %ymm3, %ymm3
; AVX512BW-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512BW-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
; AVX512BW-NEXT:    vinserti128 $1, 160(%rdi), %ymm4, %ymm3
; AVX512BW-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512BW-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
; AVX512BW-NEXT:    vinserti128 $1, 176(%rdi), %ymm5, %ymm3
; AVX512BW-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512BW-NEXT:    vinserti64x4 $1, %ymm3, %zmm2, %zmm2
; AVX512BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512BW-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-NEXT:    vpshufb %zmm3, %zmm0, %zmm0
; AVX512BW-NEXT:    vpshufb %zmm3, %zmm1, %zmm1
; AVX512BW-NEXT:    vpshufb %zmm3, %zmm2, %zmm2
; AVX512BW-NEXT:    vpalignr {{.*#+}} zmm3 = zmm2[11,12,13,14,15],zmm0[0,1,2,3,4,5,6,7,8,9,10],zmm2[27,28,29,30,31],zmm0[16,17,18,19,20,21,22,23,24,25,26],zmm2[43,44,45,46,47],zmm0[32,33,34,35,36,37,38,39,40,41,42],zmm2[59,60,61,62,63],zmm0[48,49,50,51,52,53,54,55,56,57,58]
; AVX512BW-NEXT:    vpalignr {{.*#+}} zmm0 = zmm0[11,12,13,14,15],zmm1[0,1,2,3,4,5,6,7,8,9,10],zmm0[27,28,29,30,31],zmm1[16,17,18,19,20,21,22,23,24,25,26],zmm0[43,44,45,46,47],zmm1[32,33,34,35,36,37,38,39,40,41,42],zmm0[59,60,61,62,63],zmm1[48,49,50,51,52,53,54,55,56,57,58]
; AVX512BW-NEXT:    vpalignr {{.*#+}} zmm1 = zmm1[11,12,13,14,15],zmm2[0,1,2,3,4,5,6,7,8,9,10],zmm1[27,28,29,30,31],zmm2[16,17,18,19,20,21,22,23,24,25,26],zmm1[43,44,45,46,47],zmm2[32,33,34,35,36,37,38,39,40,41,42],zmm1[59,60,61,62,63],zmm2[48,49,50,51,52,53,54,55,56,57,58]
; AVX512BW-NEXT:    movabsq $-576188069258921984, %rax # imm = 0xF800F800F800F800
; AVX512BW-NEXT:    kmovq %rax, %k1
; AVX512BW-NEXT:    vpblendmb %zmm1, %zmm0, %zmm2 {%k1}
; AVX512BW-NEXT:    vpalignr {{.*#+}} zmm1 = zmm3[11,12,13,14,15],zmm1[0,1,2,3,4,5,6,7,8,9,10],zmm3[27,28,29,30,31],zmm1[16,17,18,19,20,21,22,23,24,25,26],zmm3[43,44,45,46,47],zmm1[32,33,34,35,36,37,38,39,40,41,42],zmm3[59,60,61,62,63],zmm1[48,49,50,51,52,53,54,55,56,57,58]
; AVX512BW-NEXT:    vpalignr {{.*#+}} zmm0 = zmm0[11,12,13,14,15],zmm3[0,1,2,3,4,5,6,7,8,9,10],zmm0[27,28,29,30,31],zmm3[16,17,18,19,20,21,22,23,24,25,26],zmm0[43,44,45,46,47],zmm3[32,33,34,35,36,37,38,39,40,41,42],zmm0[59,60,61,62,63],zmm3[48,49,50,51,52,53,54,55,56,57,58]
; AVX512BW-NEXT:    vpalignr {{.*#+}} zmm0 = zmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25,42,43,44,45,46,47,32,33,34,35,36,37,38,39,40,41,58,59,60,61,62,63,48,49,50,51,52,53,54,55,56,57]
; AVX512BW-NEXT:    vmovdqa64 %zmm0, (%rsi)
; AVX512BW-NEXT:    vmovdqa64 %zmm2, (%rdx)
; AVX512BW-NEXT:    vmovdqa64 %zmm1, (%rcx)
; AVX512BW-NEXT:    vzeroupper
; AVX512BW-NEXT:    retq
;
; AVX512BW-FCP-LABEL: load_i8_stride3_vf64:
; AVX512BW-FCP:       # %bb.0:
; AVX512BW-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512BW-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512BW-FCP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512BW-FCP-NEXT:    vmovdqa 96(%rdi), %xmm3
; AVX512BW-FCP-NEXT:    vmovdqa 112(%rdi), %xmm4
; AVX512BW-FCP-NEXT:    vmovdqa 128(%rdi), %xmm5
; AVX512BW-FCP-NEXT:    vinserti128 $1, 144(%rdi), %ymm3, %ymm3
; AVX512BW-FCP-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512BW-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
; AVX512BW-FCP-NEXT:    vinserti128 $1, 160(%rdi), %ymm4, %ymm3
; AVX512BW-FCP-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512BW-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
; AVX512BW-FCP-NEXT:    vinserti128 $1, 176(%rdi), %ymm5, %ymm3
; AVX512BW-FCP-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512BW-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm2, %zmm2
; AVX512BW-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512BW-FCP-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512BW-FCP-NEXT:    vpshufb %zmm3, %zmm0, %zmm0
; AVX512BW-FCP-NEXT:    vpshufb %zmm3, %zmm1, %zmm1
; AVX512BW-FCP-NEXT:    vpshufb %zmm3, %zmm2, %zmm2
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} zmm3 = zmm2[11,12,13,14,15],zmm0[0,1,2,3,4,5,6,7,8,9,10],zmm2[27,28,29,30,31],zmm0[16,17,18,19,20,21,22,23,24,25,26],zmm2[43,44,45,46,47],zmm0[32,33,34,35,36,37,38,39,40,41,42],zmm2[59,60,61,62,63],zmm0[48,49,50,51,52,53,54,55,56,57,58]
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} zmm0 = zmm0[11,12,13,14,15],zmm1[0,1,2,3,4,5,6,7,8,9,10],zmm0[27,28,29,30,31],zmm1[16,17,18,19,20,21,22,23,24,25,26],zmm0[43,44,45,46,47],zmm1[32,33,34,35,36,37,38,39,40,41,42],zmm0[59,60,61,62,63],zmm1[48,49,50,51,52,53,54,55,56,57,58]
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} zmm1 = zmm1[11,12,13,14,15],zmm2[0,1,2,3,4,5,6,7,8,9,10],zmm1[27,28,29,30,31],zmm2[16,17,18,19,20,21,22,23,24,25,26],zmm1[43,44,45,46,47],zmm2[32,33,34,35,36,37,38,39,40,41,42],zmm1[59,60,61,62,63],zmm2[48,49,50,51,52,53,54,55,56,57,58]
; AVX512BW-FCP-NEXT:    movabsq $-576188069258921984, %rax # imm = 0xF800F800F800F800
; AVX512BW-FCP-NEXT:    kmovq %rax, %k1
; AVX512BW-FCP-NEXT:    vpblendmb %zmm1, %zmm0, %zmm2 {%k1}
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} zmm1 = zmm3[11,12,13,14,15],zmm1[0,1,2,3,4,5,6,7,8,9,10],zmm3[27,28,29,30,31],zmm1[16,17,18,19,20,21,22,23,24,25,26],zmm3[43,44,45,46,47],zmm1[32,33,34,35,36,37,38,39,40,41,42],zmm3[59,60,61,62,63],zmm1[48,49,50,51,52,53,54,55,56,57,58]
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} zmm0 = zmm0[11,12,13,14,15],zmm3[0,1,2,3,4,5,6,7,8,9,10],zmm0[27,28,29,30,31],zmm3[16,17,18,19,20,21,22,23,24,25,26],zmm0[43,44,45,46,47],zmm3[32,33,34,35,36,37,38,39,40,41,42],zmm0[59,60,61,62,63],zmm3[48,49,50,51,52,53,54,55,56,57,58]
; AVX512BW-FCP-NEXT:    vpalignr {{.*#+}} zmm0 = zmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25,42,43,44,45,46,47,32,33,34,35,36,37,38,39,40,41,58,59,60,61,62,63,48,49,50,51,52,53,54,55,56,57]
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm0, (%rsi)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm2, (%rdx)
; AVX512BW-FCP-NEXT:    vmovdqa64 %zmm1, (%rcx)
; AVX512BW-FCP-NEXT:    vzeroupper
; AVX512BW-FCP-NEXT:    retq
;
; AVX512DQ-BW-LABEL: load_i8_stride3_vf64:
; AVX512DQ-BW:       # %bb.0:
; AVX512DQ-BW-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-BW-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-BW-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-BW-NEXT:    vmovdqa 96(%rdi), %xmm3
; AVX512DQ-BW-NEXT:    vmovdqa 112(%rdi), %xmm4
; AVX512DQ-BW-NEXT:    vmovdqa 128(%rdi), %xmm5
; AVX512DQ-BW-NEXT:    vinserti128 $1, 144(%rdi), %ymm3, %ymm3
; AVX512DQ-BW-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512DQ-BW-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
; AVX512DQ-BW-NEXT:    vinserti128 $1, 160(%rdi), %ymm4, %ymm3
; AVX512DQ-BW-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512DQ-BW-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
; AVX512DQ-BW-NEXT:    vinserti128 $1, 176(%rdi), %ymm5, %ymm3
; AVX512DQ-BW-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512DQ-BW-NEXT:    vinserti64x4 $1, %ymm3, %zmm2, %zmm2
; AVX512DQ-BW-NEXT:    vbroadcasti32x4 {{.*#+}} zmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512DQ-BW-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQ-BW-NEXT:    vpshufb %zmm3, %zmm0, %zmm0
; AVX512DQ-BW-NEXT:    vpshufb %zmm3, %zmm1, %zmm1
; AVX512DQ-BW-NEXT:    vpshufb %zmm3, %zmm2, %zmm2
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} zmm3 = zmm2[11,12,13,14,15],zmm0[0,1,2,3,4,5,6,7,8,9,10],zmm2[27,28,29,30,31],zmm0[16,17,18,19,20,21,22,23,24,25,26],zmm2[43,44,45,46,47],zmm0[32,33,34,35,36,37,38,39,40,41,42],zmm2[59,60,61,62,63],zmm0[48,49,50,51,52,53,54,55,56,57,58]
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} zmm0 = zmm0[11,12,13,14,15],zmm1[0,1,2,3,4,5,6,7,8,9,10],zmm0[27,28,29,30,31],zmm1[16,17,18,19,20,21,22,23,24,25,26],zmm0[43,44,45,46,47],zmm1[32,33,34,35,36,37,38,39,40,41,42],zmm0[59,60,61,62,63],zmm1[48,49,50,51,52,53,54,55,56,57,58]
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} zmm1 = zmm1[11,12,13,14,15],zmm2[0,1,2,3,4,5,6,7,8,9,10],zmm1[27,28,29,30,31],zmm2[16,17,18,19,20,21,22,23,24,25,26],zmm1[43,44,45,46,47],zmm2[32,33,34,35,36,37,38,39,40,41,42],zmm1[59,60,61,62,63],zmm2[48,49,50,51,52,53,54,55,56,57,58]
; AVX512DQ-BW-NEXT:    movabsq $-576188069258921984, %rax # imm = 0xF800F800F800F800
; AVX512DQ-BW-NEXT:    kmovq %rax, %k1
; AVX512DQ-BW-NEXT:    vpblendmb %zmm1, %zmm0, %zmm2 {%k1}
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} zmm1 = zmm3[11,12,13,14,15],zmm1[0,1,2,3,4,5,6,7,8,9,10],zmm3[27,28,29,30,31],zmm1[16,17,18,19,20,21,22,23,24,25,26],zmm3[43,44,45,46,47],zmm1[32,33,34,35,36,37,38,39,40,41,42],zmm3[59,60,61,62,63],zmm1[48,49,50,51,52,53,54,55,56,57,58]
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} zmm0 = zmm0[11,12,13,14,15],zmm3[0,1,2,3,4,5,6,7,8,9,10],zmm0[27,28,29,30,31],zmm3[16,17,18,19,20,21,22,23,24,25,26],zmm0[43,44,45,46,47],zmm3[32,33,34,35,36,37,38,39,40,41,42],zmm0[59,60,61,62,63],zmm3[48,49,50,51,52,53,54,55,56,57,58]
; AVX512DQ-BW-NEXT:    vpalignr {{.*#+}} zmm0 = zmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25,42,43,44,45,46,47,32,33,34,35,36,37,38,39,40,41,58,59,60,61,62,63,48,49,50,51,52,53,54,55,56,57]
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm0, (%rsi)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm2, (%rdx)
; AVX512DQ-BW-NEXT:    vmovdqa64 %zmm1, (%rcx)
; AVX512DQ-BW-NEXT:    vzeroupper
; AVX512DQ-BW-NEXT:    retq
;
; AVX512DQ-BW-FCP-LABEL: load_i8_stride3_vf64:
; AVX512DQ-BW-FCP:       # %bb.0:
; AVX512DQ-BW-FCP-NEXT:    vmovdqa (%rdi), %xmm0
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 16(%rdi), %xmm1
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 32(%rdi), %xmm2
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 96(%rdi), %xmm3
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 112(%rdi), %xmm4
; AVX512DQ-BW-FCP-NEXT:    vmovdqa 128(%rdi), %xmm5
; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, 144(%rdi), %ymm3, %ymm3
; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, 48(%rdi), %ymm0, %ymm0
; AVX512DQ-BW-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, 160(%rdi), %ymm4, %ymm3
; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, 64(%rdi), %ymm1, %ymm1
; AVX512DQ-BW-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm1, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, 176(%rdi), %ymm5, %ymm3
; AVX512DQ-BW-FCP-NEXT:    vinserti128 $1, 80(%rdi), %ymm2, %ymm2
; AVX512DQ-BW-FCP-NEXT:    vinserti64x4 $1, %ymm3, %zmm2, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vbroadcasti32x4 {{.*#+}} zmm3 = [0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13,0,3,6,9,12,15,2,5,8,11,14,1,4,7,10,13]
; AVX512DQ-BW-FCP-NEXT:    # zmm3 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
; AVX512DQ-BW-FCP-NEXT:    vpshufb %zmm3, %zmm0, %zmm0
; AVX512DQ-BW-FCP-NEXT:    vpshufb %zmm3, %zmm1, %zmm1
; AVX512DQ-BW-FCP-NEXT:    vpshufb %zmm3, %zmm2, %zmm2
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} zmm3 = zmm2[11,12,13,14,15],zmm0[0,1,2,3,4,5,6,7,8,9,10],zmm2[27,28,29,30,31],zmm0[16,17,18,19,20,21,22,23,24,25,26],zmm2[43,44,45,46,47],zmm0[32,33,34,35,36,37,38,39,40,41,42],zmm2[59,60,61,62,63],zmm0[48,49,50,51,52,53,54,55,56,57,58]
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} zmm0 = zmm0[11,12,13,14,15],zmm1[0,1,2,3,4,5,6,7,8,9,10],zmm0[27,28,29,30,31],zmm1[16,17,18,19,20,21,22,23,24,25,26],zmm0[43,44,45,46,47],zmm1[32,33,34,35,36,37,38,39,40,41,42],zmm0[59,60,61,62,63],zmm1[48,49,50,51,52,53,54,55,56,57,58]
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} zmm1 = zmm1[11,12,13,14,15],zmm2[0,1,2,3,4,5,6,7,8,9,10],zmm1[27,28,29,30,31],zmm2[16,17,18,19,20,21,22,23,24,25,26],zmm1[43,44,45,46,47],zmm2[32,33,34,35,36,37,38,39,40,41,42],zmm1[59,60,61,62,63],zmm2[48,49,50,51,52,53,54,55,56,57,58]
; AVX512DQ-BW-FCP-NEXT:    movabsq $-576188069258921984, %rax # imm = 0xF800F800F800F800
; AVX512DQ-BW-FCP-NEXT:    kmovq %rax, %k1
; AVX512DQ-BW-FCP-NEXT:    vpblendmb %zmm1, %zmm0, %zmm2 {%k1}
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} zmm1 = zmm3[11,12,13,14,15],zmm1[0,1,2,3,4,5,6,7,8,9,10],zmm3[27,28,29,30,31],zmm1[16,17,18,19,20,21,22,23,24,25,26],zmm3[43,44,45,46,47],zmm1[32,33,34,35,36,37,38,39,40,41,42],zmm3[59,60,61,62,63],zmm1[48,49,50,51,52,53,54,55,56,57,58]
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} zmm0 = zmm0[11,12,13,14,15],zmm3[0,1,2,3,4,5,6,7,8,9,10],zmm0[27,28,29,30,31],zmm3[16,17,18,19,20,21,22,23,24,25,26],zmm0[43,44,45,46,47],zmm3[32,33,34,35,36,37,38,39,40,41,42],zmm0[59,60,61,62,63],zmm3[48,49,50,51,52,53,54,55,56,57,58]
; AVX512DQ-BW-FCP-NEXT:    vpalignr {{.*#+}} zmm0 = zmm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25,42,43,44,45,46,47,32,33,34,35,36,37,38,39,40,41,58,59,60,61,62,63,48,49,50,51,52,53,54,55,56,57]
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm0, (%rsi)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm2, (%rdx)
; AVX512DQ-BW-FCP-NEXT:    vmovdqa64 %zmm1, (%rcx)
; AVX512DQ-BW-FCP-NEXT:    vzeroupper
; AVX512DQ-BW-FCP-NEXT:    retq
  %wide.vec = load <192 x i8>, ptr %in.vec, align 64
  %strided.vec0 = shufflevector <192 x i8> %wide.vec, <192 x i8> 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 i8> %wide.vec, <192 x i8> 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 i8> %wide.vec, <192 x i8> 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 i8> %strided.vec0, ptr %out.vec0, align 64
  store <64 x i8> %strided.vec1, ptr %out.vec1, align 64
  store <64 x i8> %strided.vec2, ptr %out.vec2, align 64
  ret void
}