--- /dev/null
+; 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=AVX1
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX2,AVX2-SLOW
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX2,AVX2-FAST-ALL
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX2,AVX2-FAST-PERLANE
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512bw,+avx512vl | FileCheck %s --check-prefixes=AVX512
+
+; These patterns are produced by LoopVectorizer for interleaved loads.
+
+define void @load_i32_stride3_vf2(<6 x i32>* %in.vec, <2 x i32>* %out.vec0, <2 x i32>* %out.vec1, <2 x i32>* %out.vec2) nounwind {
+; SSE-LABEL: load_i32_stride3_vf2:
+; SSE: # %bb.0:
+; SSE-NEXT: movdqa (%rdi), %xmm0
+; SSE-NEXT: movdqa 16(%rdi), %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[0,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: movq %xmm2, (%rsi)
+; SSE-NEXT: movq %xmm3, (%rdx)
+; SSE-NEXT: movq %xmm0, (%rcx)
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: load_i32_stride3_vf2:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovaps (%rdi), %xmm0
+; AVX1-NEXT: vmovaps 16(%rdi), %xmm1
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[0,3,2,3]
+; AVX1-NEXT: vblendps {{.*#+}} xmm3 = xmm1[0],xmm0[1],xmm1[2,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[1,0,2,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
+; AVX1-NEXT: vmovlps %xmm2, (%rsi)
+; AVX1-NEXT: vmovlps %xmm3, (%rdx)
+; AVX1-NEXT: vmovlps %xmm0, (%rcx)
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: load_i32_stride3_vf2:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vmovaps (%rdi), %xmm0
+; AVX2-NEXT: vmovaps 16(%rdi), %xmm1
+; AVX2-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[0,3,2,3]
+; AVX2-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
+; AVX2-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,0,2,3]
+; AVX2-NEXT: vbroadcastss 8(%rdi), %xmm3
+; AVX2-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2,3]
+; AVX2-NEXT: vmovlps %xmm2, (%rsi)
+; AVX2-NEXT: vmovlps %xmm0, (%rdx)
+; AVX2-NEXT: vmovlps %xmm1, (%rcx)
+; AVX2-NEXT: retq
+;
+; AVX512-LABEL: load_i32_stride3_vf2:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovaps (%rdi), %xmm0
+; AVX512-NEXT: vmovaps 16(%rdi), %xmm1
+; AVX512-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[0,3,2,3]
+; AVX512-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
+; AVX512-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,0,2,3]
+; AVX512-NEXT: vbroadcastss 8(%rdi), %xmm3
+; AVX512-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2,3]
+; AVX512-NEXT: vmovlps %xmm2, (%rsi)
+; AVX512-NEXT: vmovlps %xmm0, (%rdx)
+; AVX512-NEXT: vmovlps %xmm1, (%rcx)
+; AVX512-NEXT: retq
+ %wide.vec = load <6 x i32>, <6 x i32>* %in.vec, align 32
+
+ %strided.vec0 = shufflevector <6 x i32> %wide.vec, <6 x i32> poison, <2 x i32> <i32 0, i32 3>
+ %strided.vec1 = shufflevector <6 x i32> %wide.vec, <6 x i32> poison, <2 x i32> <i32 1, i32 4>
+ %strided.vec2 = shufflevector <6 x i32> %wide.vec, <6 x i32> poison, <2 x i32> <i32 2, i32 5>
+
+ store <2 x i32> %strided.vec0, <2 x i32>* %out.vec0, align 32
+ store <2 x i32> %strided.vec1, <2 x i32>* %out.vec1, align 32
+ store <2 x i32> %strided.vec2, <2 x i32>* %out.vec2, align 32
+
+ ret void
+}
+
+define void @load_i32_stride3_vf4(<12 x i32>* %in.vec, <4 x i32>* %out.vec0, <4 x i32>* %out.vec1, <4 x i32>* %out.vec2) nounwind {
+; SSE-LABEL: load_i32_stride3_vf4:
+; SSE: # %bb.0:
+; SSE-NEXT: movdqa (%rdi), %xmm0
+; SSE-NEXT: movaps 16(%rdi), %xmm1
+; SSE-NEXT: movaps 32(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: movaps %xmm2, %xmm4
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,1],xmm1[3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[2,3,2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,0]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm1[1,1,1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,3],xmm2[1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,3],xmm1[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm4[2,0]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm2[0,3]
+; SSE-NEXT: movaps %xmm3, (%rsi)
+; SSE-NEXT: movaps %xmm0, (%rdx)
+; SSE-NEXT: movaps %xmm5, (%rcx)
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: load_i32_stride3_vf4:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovaps (%rdi), %xmm0
+; AVX1-NEXT: vmovaps 16(%rdi), %xmm1
+; AVX1-NEXT: vmovaps 32(%rdi), %xmm2
+; AVX1-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0],xmm2[1],xmm0[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2],xmm3[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[0,3,2,1]
+; AVX1-NEXT: vmovaps 32(%rdi), %xmm4
+; AVX1-NEXT: vblendps {{.*#+}} xmm4 = mem[0,1],xmm4[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} xmm4 = xmm1[0],xmm4[1,2],xmm1[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm4 = xmm4[1,0,3,2]
+; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm2[0,1],xmm0[2],xmm2[3]
+; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; AVX1-NEXT: vmovaps %xmm3, (%rsi)
+; AVX1-NEXT: vmovaps %xmm4, (%rdx)
+; AVX1-NEXT: vmovaps %xmm0, (%rcx)
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: load_i32_stride3_vf4:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-NEXT: vmovaps 32(%rdi), %ymm1
+; AVX2-NEXT: vmovaps {{.*#+}} xmm2 = [0,3,6,1]
+; AVX2-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6,7]
+; AVX2-NEXT: vpermps %ymm3, %ymm2, %ymm2
+; AVX2-NEXT: vmovaps {{.*#+}} xmm3 = [1,4,7,2]
+; AVX2-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
+; AVX2-NEXT: vpermps %ymm4, %ymm3, %ymm3
+; AVX2-NEXT: vmovaps {{.*#+}} xmm4 = [2,5,0,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
+; AVX2-NEXT: vpermps %ymm0, %ymm4, %ymm0
+; AVX2-NEXT: vmovaps %xmm2, (%rsi)
+; AVX2-NEXT: vmovaps %xmm3, (%rdx)
+; AVX2-NEXT: vmovaps %xmm0, (%rcx)
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+;
+; AVX512-LABEL: load_i32_stride3_vf4:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqa (%rdi), %ymm0
+; AVX512-NEXT: vmovdqa 32(%rdi), %ymm1
+; AVX512-NEXT: vmovdqa {{.*#+}} xmm2 = [0,3,6,9]
+; AVX512-NEXT: vpermi2d %ymm1, %ymm0, %ymm2
+; AVX512-NEXT: vmovdqa {{.*#+}} xmm3 = [1,4,7,10]
+; AVX512-NEXT: vpermi2d %ymm1, %ymm0, %ymm3
+; AVX512-NEXT: vmovdqa {{.*#+}} xmm4 = [2,5,8,11]
+; AVX512-NEXT: vpermi2d %ymm1, %ymm0, %ymm4
+; AVX512-NEXT: vmovdqa %xmm2, (%rsi)
+; AVX512-NEXT: vmovdqa %xmm3, (%rdx)
+; AVX512-NEXT: vmovdqa %xmm4, (%rcx)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %wide.vec = load <12 x i32>, <12 x i32>* %in.vec, align 32
+
+ %strided.vec0 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
+ %strided.vec1 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
+ %strided.vec2 = shufflevector <12 x i32> %wide.vec, <12 x i32> poison, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
+
+ store <4 x i32> %strided.vec0, <4 x i32>* %out.vec0, align 32
+ store <4 x i32> %strided.vec1, <4 x i32>* %out.vec1, align 32
+ store <4 x i32> %strided.vec2, <4 x i32>* %out.vec2, align 32
+
+ ret void
+}
+
+define void @load_i32_stride3_vf8(<24 x i32>* %in.vec, <8 x i32>* %out.vec0, <8 x i32>* %out.vec1, <8 x i32>* %out.vec2) nounwind {
+; SSE-LABEL: load_i32_stride3_vf8:
+; SSE: # %bb.0:
+; SSE-NEXT: movaps 80(%rdi), %xmm8
+; SSE-NEXT: movaps 64(%rdi), %xmm3
+; SSE-NEXT: movdqa (%rdi), %xmm1
+; SSE-NEXT: movaps 16(%rdi), %xmm5
+; SSE-NEXT: movaps 32(%rdi), %xmm10
+; SSE-NEXT: movdqa 48(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm1, %xmm11
+; SSE-NEXT: movaps %xmm10, %xmm7
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[2,1],xmm5[3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,0],xmm5[0,0]
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm5[1,1,1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,3],xmm10[1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,3],xmm5[0,2]
+; SSE-NEXT: movdqa %xmm2, %xmm5
+; SSE-NEXT: movaps %xmm8, %xmm4
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,1],xmm3[3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm2[2,3,2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm3[0,0]
+; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm3[1,1,1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,3],xmm8[1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[2,0]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,2],xmm7[2,0]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm10[0,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm8[0,3]
+; SSE-NEXT: movaps %xmm5, 16(%rsi)
+; SSE-NEXT: movaps %xmm11, (%rsi)
+; SSE-NEXT: movaps %xmm2, 16(%rdx)
+; SSE-NEXT: movaps %xmm1, (%rdx)
+; SSE-NEXT: movaps %xmm6, 16(%rcx)
+; SSE-NEXT: movaps %xmm0, (%rcx)
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: load_i32_stride3_vf8:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovaps (%rdi), %ymm0
+; AVX1-NEXT: vmovaps 32(%rdi), %ymm1
+; AVX1-NEXT: vmovaps 80(%rdi), %xmm2
+; AVX1-NEXT: vmovaps 64(%rdi), %xmm3
+; AVX1-NEXT: vinsertps {{.*#+}} xmm4 = zero,zero,xmm3[2],xmm2[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX1-NEXT: vextractf128 $1, %ymm5, %xmm6
+; AVX1-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2],xmm5[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[0,3,2,1]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[0,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm5, %ymm5
+; AVX1-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-NEXT: vblendps {{.*#+}} xmm5 = xmm3[0,1],xmm2[2],xmm3[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[0,0,3,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm6, %xmm7
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0],xmm6[1,2],xmm7[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[1,0,3,2]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[1,1,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6,7]
+; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm3[0,1],xmm2[0,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-NEXT: vmovaps %ymm4, (%rsi)
+; AVX1-NEXT: vmovaps %ymm5, (%rdx)
+; AVX1-NEXT: vmovaps %ymm0, (%rcx)
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-SLOW-LABEL: load_i32_stride3_vf8:
+; AVX2-SLOW: # %bb.0:
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm2
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm3 = [21474836482,21474836482,21474836482,21474836482]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm3, %ymm3
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm5 = <0,3,6,1,4,7,u,u>
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm5, %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm4 = [0,0,3,6,0,0,3,6]
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm4, %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm6 = <1,4,7,2,5,u,u,u>
+; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm6, %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm1 = <2,5,0,3,6,u,u,u>
+; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm2[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rsi)
+; AVX2-SLOW-NEXT: vmovaps %ymm4, (%rdx)
+; AVX2-SLOW-NEXT: vmovaps %ymm0, (%rcx)
+; AVX2-SLOW-NEXT: vzeroupper
+; AVX2-SLOW-NEXT: retq
+;
+; AVX2-FAST-ALL-LABEL: load_i32_stride3_vf8:
+; AVX2-FAST-ALL: # %bb.0:
+; AVX2-FAST-ALL-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-FAST-ALL-NEXT: vmovaps 32(%rdi), %ymm1
+; AVX2-FAST-ALL-NEXT: vmovaps 64(%rdi), %ymm2
+; AVX2-FAST-ALL-NEXT: vbroadcastsd {{.*#+}} ymm3 = [21474836482,21474836482,21474836482,21474836482]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm2, %ymm3, %ymm3
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm5 = <0,3,6,1,4,7,u,u>
+; AVX2-FAST-ALL-NEXT: vpermps %ymm4, %ymm5, %ymm4
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-FAST-ALL-NEXT: vbroadcastf128 {{.*#+}} ymm4 = [0,0,3,6,0,0,3,6]
+; AVX2-FAST-ALL-NEXT: # ymm4 = mem[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm2, %ymm4, %ymm4
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm6 = <1,4,7,2,5,u,u,u>
+; AVX2-FAST-ALL-NEXT: vpermps %ymm5, %ymm6, %ymm5
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm5 = <u,u,u,u,u,1,4,7>
+; AVX2-FAST-ALL-NEXT: vpermps %ymm2, %ymm5, %ymm2
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm1 = <2,5,0,3,6,u,u,u>
+; AVX2-FAST-ALL-NEXT: vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm3, (%rsi)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm4, (%rdx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, (%rcx)
+; AVX2-FAST-ALL-NEXT: vzeroupper
+; AVX2-FAST-ALL-NEXT: retq
+;
+; AVX2-FAST-PERLANE-LABEL: load_i32_stride3_vf8:
+; AVX2-FAST-PERLANE: # %bb.0:
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm3 = [21474836482,21474836482,21474836482,21474836482]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm3, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm5 = <0,3,6,1,4,7,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm5, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm4 = [0,0,3,6,0,0,3,6]
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm4, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm6 = <1,4,7,2,5,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm6, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm1 = <2,5,0,3,6,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm1, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm2[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, (%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, (%rcx)
+; AVX2-FAST-PERLANE-NEXT: vzeroupper
+; AVX2-FAST-PERLANE-NEXT: retq
+;
+; AVX512-LABEL: load_i32_stride3_vf8:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm0
+; AVX512-NEXT: vmovdqa 64(%rdi), %ymm1
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm2 = [0,3,6,9,12,15,18,21]
+; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm2
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm3 = [1,4,7,10,13,16,19,22]
+; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm3
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm4 = [2,5,8,11,14,17,20,23]
+; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm4
+; AVX512-NEXT: vmovdqa %ymm2, (%rsi)
+; AVX512-NEXT: vmovdqa %ymm3, (%rdx)
+; AVX512-NEXT: vmovdqa %ymm4, (%rcx)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %wide.vec = load <24 x i32>, <24 x i32>* %in.vec, align 32
+
+ %strided.vec0 = shufflevector <24 x i32> %wide.vec, <24 x i32> 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 i32> %wide.vec, <24 x i32> 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 i32> %wide.vec, <24 x i32> poison, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
+
+ store <8 x i32> %strided.vec0, <8 x i32>* %out.vec0, align 32
+ store <8 x i32> %strided.vec1, <8 x i32>* %out.vec1, align 32
+ store <8 x i32> %strided.vec2, <8 x i32>* %out.vec2, align 32
+
+ ret void
+}
+
+define void @load_i32_stride3_vf16(<48 x i32>* %in.vec, <16 x i32>* %out.vec0, <16 x i32>* %out.vec1, <16 x i32>* %out.vec2) nounwind {
+; SSE-LABEL: load_i32_stride3_vf16:
+; SSE: # %bb.0:
+; SSE-NEXT: movaps 96(%rdi), %xmm10
+; SSE-NEXT: movaps 128(%rdi), %xmm6
+; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 112(%rdi), %xmm13
+; SSE-NEXT: movaps 144(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 176(%rdi), %xmm7
+; SSE-NEXT: movaps 160(%rdi), %xmm5
+; SSE-NEXT: movaps (%rdi), %xmm15
+; SSE-NEXT: movaps 16(%rdi), %xmm9
+; SSE-NEXT: movaps 32(%rdi), %xmm8
+; SSE-NEXT: movaps 48(%rdi), %xmm11
+; SSE-NEXT: movaps 80(%rdi), %xmm12
+; SSE-NEXT: movaps 64(%rdi), %xmm3
+; SSE-NEXT: movaps %xmm3, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm12[1,1]
+; SSE-NEXT: movaps %xmm11, %xmm4
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm5, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm7[1,1]
+; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm1, %xmm4
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm9, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm8[1,1]
+; SSE-NEXT: movaps %xmm15, %xmm1
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm13, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm6[1,1]
+; SSE-NEXT: movaps %xmm10, %xmm14
+; SSE-NEXT: movaps %xmm10, %xmm1
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm12, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,1],xmm3[3,3]
+; SSE-NEXT: movaps %xmm11, %xmm10
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[1,0],xmm3[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,2],xmm0[2,0]
+; SSE-NEXT: movaps %xmm7, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,1],xmm5[3,3]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: movaps %xmm7, %xmm4
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm5[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,2],xmm0[2,0]
+; SSE-NEXT: movaps %xmm6, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,1],xmm13[3,3]
+; SSE-NEXT: movaps %xmm1, %xmm2
+; SSE-NEXT: movaps %xmm1, %xmm6
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm13[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2],xmm0[2,0]
+; SSE-NEXT: movaps %xmm8, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,1],xmm9[3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[2,3,2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[1,0],xmm9[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[0,2],xmm0[2,0]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm11[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm12[0,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm7[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = xmm5[0,1],mem[0,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm8[0,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
+; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = xmm6[0,1],mem[0,3]
+; SSE-NEXT: movaps %xmm14, 32(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, (%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 48(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 16(%rsi)
+; SSE-NEXT: movaps %xmm2, 32(%rdx)
+; SSE-NEXT: movaps %xmm15, (%rdx)
+; SSE-NEXT: movaps %xmm4, 48(%rdx)
+; SSE-NEXT: movaps %xmm10, 16(%rdx)
+; SSE-NEXT: movaps %xmm6, 32(%rcx)
+; SSE-NEXT: movaps %xmm1, (%rcx)
+; SSE-NEXT: movaps %xmm5, 48(%rcx)
+; SSE-NEXT: movaps %xmm3, 16(%rcx)
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: load_i32_stride3_vf16:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovaps (%rdi), %ymm5
+; AVX1-NEXT: vmovaps 32(%rdi), %ymm6
+; AVX1-NEXT: vmovaps 128(%rdi), %ymm11
+; AVX1-NEXT: vmovaps 96(%rdi), %ymm12
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0],ymm11[1],ymm12[2,3],ymm11[4],ymm12[5,6],ymm11[7]
+; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2],xmm2[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[0,3,2,1]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm3
+; AVX1-NEXT: vmovaps 176(%rdi), %xmm13
+; AVX1-NEXT: vmovaps 160(%rdi), %xmm4
+; AVX1-NEXT: vinsertps {{.*#+}} xmm7 = zero,zero,xmm4[2],xmm13[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-NEXT: vblendps {{.*#+}} ymm9 = ymm3[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-NEXT: vmovaps 80(%rdi), %xmm3
+; AVX1-NEXT: vmovaps 64(%rdi), %xmm7
+; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = zero,zero,xmm7[2],xmm3[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vblendps {{.*#+}} ymm8 = ymm5[0],ymm6[1],ymm5[2,3],ymm6[4],ymm5[5,6],ymm6[7]
+; AVX1-NEXT: vextractf128 $1, %ymm8, %xmm1
+; AVX1-NEXT: vblendps {{.*#+}} xmm2 = xmm8[0,1],xmm1[2],xmm8[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[0,3,2,1]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm7[0,1],xmm3[2],xmm7[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,0,3,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm6[2],ymm5[3,4],ymm6[5],ymm5[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0],xmm1[1,2],xmm2[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[1,0,3,2]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1],ymm11[2],ymm12[3,4],ymm11[5],ymm12[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,0,3,2]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: vblendps {{.*#+}} xmm1 = xmm4[0,1],xmm13[2],xmm4[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,0,3,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm7[0,1],xmm3[0,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm5[2],ymm6[3,4],ymm5[5],ymm6[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[2,1,0,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[2,3,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm11[0,1],ymm12[2],ymm11[3,4],ymm12[5],ymm11[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[2,1,0,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[2,3,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
+; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm4[0,1],xmm13[0,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-NEXT: vmovaps %ymm10, (%rsi)
+; AVX1-NEXT: vmovaps %ymm9, 32(%rsi)
+; AVX1-NEXT: vmovaps %ymm0, 32(%rdx)
+; AVX1-NEXT: vmovaps %ymm8, (%rdx)
+; AVX1-NEXT: vmovaps %ymm2, 32(%rcx)
+; AVX1-NEXT: vmovaps %ymm1, (%rcx)
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-SLOW-LABEL: load_i32_stride3_vf16:
+; AVX2-SLOW: # %bb.0:
+; AVX2-SLOW-NEXT: vmovaps 128(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovaps 160(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm2
+; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm3
+; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %ymm5
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm6 = [21474836482,21474836482,21474836482,21474836482]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm6, %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm2[0],ymm3[1],ymm2[2,3],ymm3[4],ymm2[5,6],ymm3[7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm9 = <0,3,6,1,4,7,u,u>
+; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm9, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm6, %ymm6
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm5[0],ymm1[1],ymm5[2,3],ymm1[4],ymm5[5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm9, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm8 = [0,0,3,6,0,0,3,6]
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm8, %ymm9
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm11 = <1,4,7,2,5,u,u,u>
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm11, %ymm10
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm8, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm5[0,1],ymm1[2],ymm5[3,4],ymm1[5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm11, %ymm10
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm3 = <2,5,0,3,6,u,u,u>
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm3, %ymm2
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm4 = ymm4[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm5[2],ymm1[3,4],ymm5[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm3, %ymm1
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%rsi)
+; AVX2-SLOW-NEXT: vmovaps %ymm7, (%rsi)
+; AVX2-SLOW-NEXT: vmovaps %ymm8, 32(%rdx)
+; AVX2-SLOW-NEXT: vmovaps %ymm9, (%rdx)
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rcx)
+; AVX2-SLOW-NEXT: vzeroupper
+; AVX2-SLOW-NEXT: retq
+;
+; AVX2-FAST-ALL-LABEL: load_i32_stride3_vf16:
+; AVX2-FAST-ALL: # %bb.0:
+; AVX2-FAST-ALL-NEXT: vmovaps 128(%rdi), %ymm0
+; AVX2-FAST-ALL-NEXT: vmovaps 160(%rdi), %ymm1
+; AVX2-FAST-ALL-NEXT: vmovaps (%rdi), %ymm2
+; AVX2-FAST-ALL-NEXT: vmovaps 32(%rdi), %ymm3
+; AVX2-FAST-ALL-NEXT: vmovaps 64(%rdi), %ymm4
+; AVX2-FAST-ALL-NEXT: vmovaps 96(%rdi), %ymm5
+; AVX2-FAST-ALL-NEXT: vbroadcastsd {{.*#+}} ymm6 = [21474836482,21474836482,21474836482,21474836482]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm4, %ymm6, %ymm7
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm8 = ymm2[0],ymm3[1],ymm2[2,3],ymm3[4],ymm2[5,6],ymm3[7]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm9 = <0,3,6,1,4,7,u,u>
+; AVX2-FAST-ALL-NEXT: vpermps %ymm8, %ymm9, %ymm8
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm1, %ymm6, %ymm6
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm8 = ymm5[0],ymm0[1],ymm5[2,3],ymm0[4],ymm5[5,6],ymm0[7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm8, %ymm9, %ymm8
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-FAST-ALL-NEXT: vbroadcastf128 {{.*#+}} ymm8 = [0,0,3,6,0,0,3,6]
+; AVX2-FAST-ALL-NEXT: # ymm8 = mem[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm4, %ymm8, %ymm9
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm11 = <1,4,7,2,5,u,u,u>
+; AVX2-FAST-ALL-NEXT: vpermps %ymm10, %ymm11, %ymm10
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm1, %ymm8, %ymm8
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm10 = ymm5[0,1],ymm0[2],ymm5[3,4],ymm0[5],ymm5[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm10, %ymm11, %ymm10
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm10 = [0,1,0,3,0,1,4,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm4, %ymm10, %ymm4
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm3 = <2,5,0,3,6,u,u,u>
+; AVX2-FAST-ALL-NEXT: vpermps %ymm2, %ymm3, %ymm2
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm1, %ymm10, %ymm1
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm5[2],ymm0[3,4],ymm5[5],ymm0[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm0, %ymm3, %ymm0
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm6, 32(%rsi)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm7, (%rsi)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm8, 32(%rdx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm9, (%rdx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm2, (%rcx)
+; AVX2-FAST-ALL-NEXT: vzeroupper
+; AVX2-FAST-ALL-NEXT: retq
+;
+; AVX2-FAST-PERLANE-LABEL: load_i32_stride3_vf16:
+; AVX2-FAST-PERLANE: # %bb.0:
+; AVX2-FAST-PERLANE-NEXT: vmovaps 128(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps 160(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm6 = [21474836482,21474836482,21474836482,21474836482]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm6, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm2[0],ymm3[1],ymm2[2,3],ymm3[4],ymm2[5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm9 = <0,3,6,1,4,7,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm9, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm6, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm5[0],ymm1[1],ymm5[2,3],ymm1[4],ymm5[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm9, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm8 = [0,0,3,6,0,0,3,6]
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm8, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm11 = <1,4,7,2,5,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm11, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm8, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm5[0,1],ymm1[2],ymm5[3,4],ymm1[5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm11, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm3 = <2,5,0,3,6,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm3, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm4 = ymm4[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm5[2],ymm1[3,4],ymm5[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm3, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm7, (%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, 32(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, (%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rcx)
+; AVX2-FAST-PERLANE-NEXT: vzeroupper
+; AVX2-FAST-PERLANE-NEXT: retq
+;
+; AVX512-LABEL: load_i32_stride3_vf16:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm0
+; AVX512-NEXT: vmovdqu64 64(%rdi), %zmm1
+; AVX512-NEXT: vmovdqu64 128(%rdi), %zmm2
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm3 = <0,3,6,9,12,15,18,21,24,27,30,u,u,u,u,u>
+; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm3
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm4 = [0,1,2,3,4,5,6,7,8,9,10,17,20,23,26,29]
+; AVX512-NEXT: vpermi2d %zmm2, %zmm3, %zmm4
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm3 = <17,20,23,26,29,0,3,6,9,12,15,u,u,u,u,u>
+; AVX512-NEXT: vpermi2d %zmm0, %zmm1, %zmm3
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm5 = [0,1,2,3,4,5,6,7,8,9,10,18,21,24,27,30]
+; AVX512-NEXT: vpermi2d %zmm2, %zmm3, %zmm5
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm3 = <2,5,8,11,14,17,20,23,26,29,u,u,u,u,u,u>
+; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm3
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,1,2,3,4,5,6,7,8,9,16,19,22,25,28,31]
+; AVX512-NEXT: vpermi2d %zmm2, %zmm3, %zmm0
+; AVX512-NEXT: vmovdqu64 %zmm4, (%rsi)
+; AVX512-NEXT: vmovdqu64 %zmm5, (%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm0, (%rcx)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %wide.vec = load <48 x i32>, <48 x i32>* %in.vec, align 32
+
+ %strided.vec0 = shufflevector <48 x i32> %wide.vec, <48 x i32> 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 i32> %wide.vec, <48 x i32> 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 i32> %wide.vec, <48 x i32> 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 i32> %strided.vec0, <16 x i32>* %out.vec0, align 32
+ store <16 x i32> %strided.vec1, <16 x i32>* %out.vec1, align 32
+ store <16 x i32> %strided.vec2, <16 x i32>* %out.vec2, align 32
+
+ ret void
+}
--- /dev/null
+; 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=AVX1
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX2,AVX2-SLOW
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX2,AVX2-FAST-ALL
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX2,AVX2-FAST-PERLANE
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512bw,+avx512vl | FileCheck %s --check-prefixes=AVX512
+
+; These patterns are produced by LoopVectorizer for interleaved stores.
+
+define void @store_i32_stride3_vf2(<2 x i32>* %in.vecptr0, <2 x i32>* %in.vecptr1, <2 x i32>* %in.vecptr2, <6 x i32>* %out.vec) nounwind {
+; SSE-LABEL: store_i32_stride3_vf2:
+; SSE: # %bb.0:
+; SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
+; SSE-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
+; SSE-NEXT: movq {{.*#+}} xmm2 = mem[0],zero
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[1,1,1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[1,3]
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE-NEXT: movq %xmm1, 16(%rcx)
+; SSE-NEXT: movaps %xmm0, (%rcx)
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: store_i32_stride3_vf2:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX1-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm2
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,1,0,3,7,5,4,7]
+; AVX1-NEXT: vpermilps {{.*#+}} ymm1 = ymm2[0,2,u,1,u,5,u,u]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3],ymm0[4],ymm1[5,6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vmovlps %xmm1, 16(%rcx)
+; AVX1-NEXT: vmovaps %xmm0, (%rcx)
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: store_i32_stride3_vf2:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX2-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX2-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
+; AVX2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT: vmovaps {{.*#+}} ymm1 = <0,2,4,1,3,5,u,u>
+; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm0
+; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vmovlps %xmm1, 16(%rcx)
+; AVX2-NEXT: vmovaps %xmm0, (%rcx)
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+;
+; AVX512-LABEL: store_i32_stride3_vf2:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX512-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
+; AVX512-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX512-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
+; AVX512-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX512-NEXT: vmovaps {{.*#+}} ymm1 = <0,2,4,1,3,5,u,u>
+; AVX512-NEXT: vpermps %ymm0, %ymm1, %ymm0
+; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX512-NEXT: vmovlps %xmm1, 16(%rcx)
+; AVX512-NEXT: vmovaps %xmm0, (%rcx)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %in.vec0 = load <2 x i32>, <2 x i32>* %in.vecptr0, align 32
+ %in.vec1 = load <2 x i32>, <2 x i32>* %in.vecptr1, align 32
+ %in.vec2 = load <2 x i32>, <2 x i32>* %in.vecptr2, align 32
+
+ %concat01 = shufflevector <2 x i32> %in.vec0, <2 x i32> %in.vec1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %concat2u = shufflevector <2 x i32> %in.vec2, <2 x i32> poison, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
+ %concat012 = shufflevector <4 x i32> %concat01, <4 x i32> %concat2u, <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
+ %interleaved.vec = shufflevector <6 x i32> %concat012, <6 x i32> poison, <6 x i32> <i32 0, i32 2, i32 4, i32 1, i32 3, i32 5>
+
+ store <6 x i32> %interleaved.vec, <6 x i32>* %out.vec, align 32
+
+ ret void
+}
+
+define void @store_i32_stride3_vf4(<4 x i32>* %in.vecptr0, <4 x i32>* %in.vecptr1, <4 x i32>* %in.vecptr2, <12 x i32>* %out.vec) nounwind {
+; SSE-LABEL: store_i32_stride3_vf4:
+; SSE: # %bb.0:
+; SSE-NEXT: movaps (%rdi), %xmm0
+; SSE-NEXT: movaps (%rsi), %xmm1
+; SSE-NEXT: movaps (%rdx), %xmm2
+; SSE-NEXT: movaps %xmm2, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[1,3]
+; SSE-NEXT: movaps %xmm0, %xmm4
+; SSE-NEXT: unpcklps {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[0,2]
+; SSE-NEXT: movaps %xmm1, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,3],xmm2[3,3]
+; SSE-NEXT: movaps %xmm2, %xmm5
+; SSE-NEXT: unpckhps {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,2]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm1[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, 16(%rcx)
+; SSE-NEXT: movaps %xmm5, 32(%rcx)
+; SSE-NEXT: movaps %xmm4, (%rcx)
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: store_i32_stride3_vf4:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovaps (%rdi), %xmm0
+; AVX1-NEXT: vmovaps (%rsi), %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm2
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm3
+; AVX1-NEXT: vmovsldup {{.*#+}} ymm3 = ymm3[0,0,2,2,4,4,6,6]
+; AVX1-NEXT: vpermilps {{.*#+}} ymm2 = ymm2[0,u,u,1,5,u,u,6]
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3,4,5],ymm3[6],ymm2[7]
+; AVX1-NEXT: vbroadcastsd (%rdx), %ymm3
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; AVX1-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
+; AVX1-NEXT: vmovaps %xmm0, 32(%rcx)
+; AVX1-NEXT: vmovaps %ymm2, (%rcx)
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-SLOW-LABEL: store_i32_stride3_vf4:
+; AVX2-SLOW: # %bb.0:
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovaps (%rsi), %xmm1
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm3 = <0,4,u,1,5,u,2,6>
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm3, %ymm2
+; AVX2-SLOW-NEXT: vbroadcastsd (%rdx), %ymm3
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
+; AVX2-SLOW-NEXT: vmovaps %xmm0, 32(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rcx)
+; AVX2-SLOW-NEXT: vzeroupper
+; AVX2-SLOW-NEXT: retq
+;
+; AVX2-FAST-ALL-LABEL: store_i32_stride3_vf4:
+; AVX2-FAST-ALL: # %bb.0:
+; AVX2-FAST-ALL-NEXT: vmovaps (%rdi), %xmm0
+; AVX2-FAST-ALL-NEXT: vinsertf128 $1, (%rsi), %ymm0, %ymm0
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm1 = <u,3,7,u,u,u,u,u>
+; AVX2-FAST-ALL-NEXT: vpermps %ymm0, %ymm1, %ymm1
+; AVX2-FAST-ALL-NEXT: vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0],xmm1[1,2],xmm2[3]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm2 = <0,4,u,1,5,u,2,6>
+; AVX2-FAST-ALL-NEXT: vpermps %ymm0, %ymm2, %ymm0
+; AVX2-FAST-ALL-NEXT: vbroadcastsd (%rdx), %ymm2
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2],ymm0[3,4],ymm2[5],ymm0[6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, (%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %xmm1, 32(%rcx)
+; AVX2-FAST-ALL-NEXT: vzeroupper
+; AVX2-FAST-ALL-NEXT: retq
+;
+; AVX2-FAST-PERLANE-LABEL: store_i32_stride3_vf4:
+; AVX2-FAST-PERLANE: # %bb.0:
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rsi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm3 = <0,4,u,1,5,u,2,6>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm3, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd (%rdx), %ymm3
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rcx)
+; AVX2-FAST-PERLANE-NEXT: vzeroupper
+; AVX2-FAST-PERLANE-NEXT: retq
+;
+; AVX512-LABEL: store_i32_stride3_vf4:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovaps (%rdi), %xmm0
+; AVX512-NEXT: vmovaps (%rdx), %xmm1
+; AVX512-NEXT: vinsertf128 $1, (%rsi), %ymm0, %ymm0
+; AVX512-NEXT: vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512-NEXT: vmovaps {{.*#+}} zmm1 = <0,4,8,1,5,9,2,6,10,3,7,11,u,u,u,u>
+; AVX512-NEXT: vpermps %zmm0, %zmm1, %zmm0
+; AVX512-NEXT: vextractf32x4 $2, %zmm0, 32(%rcx)
+; AVX512-NEXT: vmovaps %ymm0, (%rcx)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %in.vec0 = load <4 x i32>, <4 x i32>* %in.vecptr0, align 32
+ %in.vec1 = load <4 x i32>, <4 x i32>* %in.vecptr1, align 32
+ %in.vec2 = load <4 x i32>, <4 x i32>* %in.vecptr2, align 32
+
+ %concat01 = shufflevector <4 x i32> %in.vec0, <4 x i32> %in.vec1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ %concat2u = shufflevector <4 x i32> %in.vec2, <4 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
+ %concat012 = shufflevector <8 x i32> %concat01, <8 x i32> %concat2u, <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
+ %interleaved.vec = shufflevector <12 x i32> %concat012, <12 x i32> poison, <12 x i32> <i32 0, i32 4, i32 8, i32 1, i32 5, i32 9, i32 2, i32 6, i32 10, i32 3, i32 7, i32 11>
+
+ store <12 x i32> %interleaved.vec, <12 x i32>* %out.vec, align 32
+
+ ret void
+}
+
+define void @store_i32_stride3_vf8(<8 x i32>* %in.vecptr0, <8 x i32>* %in.vecptr1, <8 x i32>* %in.vecptr2, <24 x i32>* %out.vec) nounwind {
+; SSE-LABEL: store_i32_stride3_vf8:
+; SSE: # %bb.0:
+; SSE-NEXT: movaps (%rdi), %xmm1
+; SSE-NEXT: movaps 16(%rdi), %xmm0
+; SSE-NEXT: movaps (%rsi), %xmm8
+; SSE-NEXT: movaps 16(%rsi), %xmm5
+; SSE-NEXT: movaps (%rdx), %xmm4
+; SSE-NEXT: movaps 16(%rdx), %xmm6
+; SSE-NEXT: movaps %xmm5, %xmm7
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT: movaps %xmm6, %xmm2
+; SSE-NEXT: unpckhps {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,3],xmm7[0,2]
+; SSE-NEXT: movaps %xmm0, %xmm7
+; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm5[1]
+; SSE-NEXT: movaps %xmm6, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm5[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm7[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[1,3]
+; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm6[0,2]
+; SSE-NEXT: movaps %xmm8, %xmm5
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[3,3],xmm4[3,3]
+; SSE-NEXT: movaps %xmm4, %xmm6
+; SSE-NEXT: unpckhps {{.*#+}} xmm6 = xmm6[2],xmm1[2],xmm6[3],xmm1[3]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,3],xmm5[0,2]
+; SSE-NEXT: movaps %xmm1, %xmm5
+; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm8[1]
+; SSE-NEXT: movaps %xmm4, %xmm7
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[1,0],xmm8[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[2,0],xmm5[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm1[1,3]
+; SSE-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[0,2]
+; SSE-NEXT: movaps %xmm1, (%rcx)
+; SSE-NEXT: movaps %xmm7, 16(%rcx)
+; SSE-NEXT: movaps %xmm6, 32(%rcx)
+; SSE-NEXT: movaps %xmm0, 48(%rcx)
+; SSE-NEXT: movaps %xmm3, 64(%rcx)
+; SSE-NEXT: movaps %xmm2, 80(%rcx)
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: store_i32_stride3_vf8:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovaps (%rsi), %xmm0
+; AVX1-NEXT: vmovaps (%rdi), %xmm1
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm1[1],xmm0[1]
+; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm0[1,1],xmm2[0,2]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT: vbroadcastsd (%rdx), %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX1-NEXT: vmovaps 16(%rdx), %xmm1
+; AVX1-NEXT: vmovaps 16(%rsi), %xmm2
+; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm2[3,0],xmm1[3,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm1[2,1],xmm3[0,2]
+; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],xmm2[1,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-NEXT: vbroadcastsd 24(%rdi), %ymm2
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm2 = mem[1,0,2,2]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm3 = mem[1,1,2,2]
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5,6],ymm2[7]
+; AVX1-NEXT: vpermilps {{.*#+}} ymm3 = mem[0,0,3,3,4,4,7,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX1-NEXT: vmovaps %ymm2, 32(%rcx)
+; AVX1-NEXT: vmovaps %ymm1, 64(%rcx)
+; AVX1-NEXT: vmovaps %ymm0, (%rcx)
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-SLOW-LABEL: store_i32_stride3_vf8:
+; AVX2-SLOW: # %bb.0:
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovaps (%rsi), %ymm1
+; AVX2-SLOW-NEXT: vmovaps (%rdx), %ymm2
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm3 = mem[1,0,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm0[0,0,2,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7]
+; AVX2-SLOW-NEXT: vbroadcastsd (%rdx), %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm2[2,1,3,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm1[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vbroadcastsd 24(%rdi), %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[1,1,2,2]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3],ymm0[4],ymm2[5,6],ymm0[7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 64(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rcx)
+; AVX2-SLOW-NEXT: vzeroupper
+; AVX2-SLOW-NEXT: retq
+;
+; AVX2-FAST-ALL-LABEL: store_i32_stride3_vf8:
+; AVX2-FAST-ALL: # %bb.0:
+; AVX2-FAST-ALL-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-FAST-ALL-NEXT: vmovaps (%rsi), %ymm1
+; AVX2-FAST-ALL-NEXT: vmovaps (%rdx), %ymm2
+; AVX2-FAST-ALL-NEXT: vbroadcastf128 {{.*#+}} ymm3 = [1,0,2,2,1,0,2,2]
+; AVX2-FAST-ALL-NEXT: # ymm3 = mem[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm1, %ymm3, %ymm3
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm4 = ymm0[0,0,2,1]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd (%rdx), %ymm4
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7]
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm4 = ymm2[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0],ymm0[1],ymm4[2,3],ymm0[4],ymm4[5,6],ymm0[7]
+; AVX2-FAST-ALL-NEXT: vpermilps {{.*#+}} ymm4 = ymm1[0,0,3,3,4,4,7,7]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2],ymm0[3,4],ymm4[5],ymm0[6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm4 = <5,u,u,6,u,u,7,u>
+; AVX2-FAST-ALL-NEXT: vpermps %ymm1, %ymm4, %ymm1
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[2,1,3,3]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd 24(%rdi), %ymm2
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm1, 64(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm3, (%rcx)
+; AVX2-FAST-ALL-NEXT: vzeroupper
+; AVX2-FAST-ALL-NEXT: retq
+;
+; AVX2-FAST-PERLANE-LABEL: store_i32_stride3_vf8:
+; AVX2-FAST-PERLANE: # %bb.0:
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rsi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdx), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm3 = mem[1,0,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm0[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd (%rdx), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm2[2,1,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm1[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 24(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3],ymm0[4],ymm2[5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 64(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rcx)
+; AVX2-FAST-PERLANE-NEXT: vzeroupper
+; AVX2-FAST-PERLANE-NEXT: retq
+;
+; AVX512-LABEL: store_i32_stride3_vf8:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqa (%rdi), %ymm0
+; AVX512-NEXT: vmovdqa (%rdx), %ymm1
+; AVX512-NEXT: vinserti64x4 $1, (%rsi), %zmm0, %zmm0
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm2 = [13,21,6,14,22,7,15,23]
+; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm2
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,8,16,1,9,17,2,10,18,3,11,19,4,12,20,5]
+; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm3
+; AVX512-NEXT: vmovdqu64 %zmm3, (%rcx)
+; AVX512-NEXT: vmovdqa %ymm2, 64(%rcx)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %in.vec0 = load <8 x i32>, <8 x i32>* %in.vecptr0, align 32
+ %in.vec1 = load <8 x i32>, <8 x i32>* %in.vecptr1, align 32
+ %in.vec2 = load <8 x i32>, <8 x i32>* %in.vecptr2, align 32
+
+ %concat01 = shufflevector <8 x i32> %in.vec0, <8 x i32> %in.vec1, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ %concat2u = shufflevector <8 x i32> %in.vec2, <8 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+ %concat012 = shufflevector <16 x i32> %concat01, <16 x i32> %concat2u, <24 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+ %interleaved.vec = shufflevector <24 x i32> %concat012, <24 x i32> poison, <24 x i32> <i32 0, i32 8, i32 16, i32 1, i32 9, i32 17, i32 2, i32 10, i32 18, i32 3, i32 11, i32 19, i32 4, i32 12, i32 20, i32 5, i32 13, i32 21, i32 6, i32 14, i32 22, i32 7, i32 15, i32 23>
+
+ store <24 x i32> %interleaved.vec, <24 x i32>* %out.vec, align 32
+
+ ret void
+}
+
+define void @store_i32_stride3_vf16(<16 x i32>* %in.vecptr0, <16 x i32>* %in.vecptr1, <16 x i32>* %in.vecptr2, <48 x i32>* %out.vec) nounwind {
+; SSE-LABEL: store_i32_stride3_vf16:
+; SSE: # %bb.0:
+; SSE-NEXT: movaps (%rdi), %xmm3
+; SSE-NEXT: movaps 16(%rdi), %xmm2
+; SSE-NEXT: movaps 32(%rdi), %xmm1
+; SSE-NEXT: movaps 48(%rdi), %xmm15
+; SSE-NEXT: movaps (%rsi), %xmm8
+; SSE-NEXT: movaps 16(%rsi), %xmm10
+; SSE-NEXT: movaps 32(%rsi), %xmm14
+; SSE-NEXT: movaps 48(%rsi), %xmm0
+; SSE-NEXT: movaps (%rdx), %xmm12
+; SSE-NEXT: movaps 16(%rdx), %xmm4
+; SSE-NEXT: movaps 32(%rdx), %xmm5
+; SSE-NEXT: movaps 48(%rdx), %xmm6
+; SSE-NEXT: movaps %xmm0, %xmm7
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT: movaps %xmm6, %xmm9
+; SSE-NEXT: unpckhps {{.*#+}} xmm9 = xmm9[2],xmm15[2],xmm9[3],xmm15[3]
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,3],xmm7[0,2]
+; SSE-NEXT: movaps %xmm15, %xmm7
+; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm0[1]
+; SSE-NEXT: movaps %xmm6, %xmm11
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[1,0],xmm0[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[2,0],xmm7[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm15[1,3]
+; SSE-NEXT: unpcklps {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[0,1],xmm6[0,2]
+; SSE-NEXT: movaps %xmm14, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
+; SSE-NEXT: movaps %xmm5, %xmm13
+; SSE-NEXT: unpckhps {{.*#+}} xmm13 = xmm13[2],xmm1[2],xmm13[3],xmm1[3]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm1, %xmm0
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm14[1]
+; SSE-NEXT: movaps %xmm5, %xmm7
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[1,0],xmm14[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[2,0],xmm0[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm1[1,3]
+; SSE-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm5[0,2]
+; SSE-NEXT: movaps %xmm10, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3],xmm4[3,3]
+; SSE-NEXT: movaps %xmm4, %xmm14
+; SSE-NEXT: unpckhps {{.*#+}} xmm14 = xmm14[2],xmm2[2],xmm14[3],xmm2[3]
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, %xmm0
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm10[1]
+; SSE-NEXT: movaps %xmm4, %xmm5
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[1,0],xmm10[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm0[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[1,3]
+; SSE-NEXT: unpcklps {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm4[0,2]
+; SSE-NEXT: movaps %xmm8, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3],xmm12[3,3]
+; SSE-NEXT: movaps %xmm12, %xmm4
+; SSE-NEXT: unpckhps {{.*#+}} xmm4 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm3, %xmm0
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm8[1]
+; SSE-NEXT: movaps %xmm12, %xmm6
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[1,0],xmm8[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[2,0],xmm0[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[0,1],xmm3[1,3]
+; SSE-NEXT: unpcklps {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm12[0,2]
+; SSE-NEXT: movaps %xmm3, (%rcx)
+; SSE-NEXT: movaps %xmm6, 16(%rcx)
+; SSE-NEXT: movaps %xmm4, 32(%rcx)
+; SSE-NEXT: movaps %xmm2, 48(%rcx)
+; SSE-NEXT: movaps %xmm5, 64(%rcx)
+; SSE-NEXT: movaps %xmm14, 80(%rcx)
+; SSE-NEXT: movaps %xmm1, 96(%rcx)
+; SSE-NEXT: movaps %xmm7, 112(%rcx)
+; SSE-NEXT: movaps %xmm13, 128(%rcx)
+; SSE-NEXT: movaps %xmm15, 144(%rcx)
+; SSE-NEXT: movaps %xmm11, 160(%rcx)
+; SSE-NEXT: movaps %xmm9, 176(%rcx)
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: store_i32_stride3_vf16:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovaps (%rsi), %xmm0
+; AVX1-NEXT: vmovaps (%rdi), %xmm1
+; AVX1-NEXT: vmovaps 32(%rdi), %xmm2
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm1[1],xmm0[1]
+; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm0[1,1],xmm3[0,2]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-NEXT: vbroadcastsd (%rdx), %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX1-NEXT: vmovaps 48(%rdx), %xmm1
+; AVX1-NEXT: vmovaps 48(%rsi), %xmm3
+; AVX1-NEXT: vshufps {{.*#+}} xmm4 = xmm3[3,0],xmm1[3,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm4 = xmm1[2,1],xmm4[0,2]
+; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],xmm3[1,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[2,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
+; AVX1-NEXT: vbroadcastsd 56(%rdi), %ymm3
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7]
+; AVX1-NEXT: vmovaps 32(%rsi), %xmm3
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm2[1],xmm3[1]
+; AVX1-NEXT: vshufps {{.*#+}} xmm4 = xmm3[1,1],xmm4[0,2]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm3[2,0],xmm2[2,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; AVX1-NEXT: vbroadcastsd 32(%rdx), %ymm3
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX1-NEXT: vmovaps 16(%rdx), %xmm3
+; AVX1-NEXT: vmovaps 16(%rsi), %xmm4
+; AVX1-NEXT: vshufps {{.*#+}} xmm5 = xmm4[3,0],xmm3[3,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm5 = xmm3[2,1],xmm5[0,2]
+; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,0],xmm4[1,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm3[2,0],xmm4[2,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm3
+; AVX1-NEXT: vbroadcastsd 24(%rdi), %ymm4
+; AVX1-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm4 = mem[1,0,2,2]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm5 = mem[1,1,2,2]
+; AVX1-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5,6],ymm4[7]
+; AVX1-NEXT: vpermilps {{.*#+}} ymm5 = mem[0,0,3,3,4,4,7,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm5 = mem[1,0,2,2]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm6 = mem[1,1,2,2]
+; AVX1-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0],ymm5[1],ymm6[2,3],ymm5[4],ymm6[5,6],ymm5[7]
+; AVX1-NEXT: vpermilps {{.*#+}} ymm6 = mem[0,0,3,3,4,4,7,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2],ymm5[3,4],ymm6[5],ymm5[6,7]
+; AVX1-NEXT: vmovaps %ymm5, 32(%rcx)
+; AVX1-NEXT: vmovaps %ymm4, 128(%rcx)
+; AVX1-NEXT: vmovaps %ymm3, 64(%rcx)
+; AVX1-NEXT: vmovaps %ymm2, 96(%rcx)
+; AVX1-NEXT: vmovaps %ymm1, 160(%rcx)
+; AVX1-NEXT: vmovaps %ymm0, (%rcx)
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-SLOW-LABEL: store_i32_stride3_vf16:
+; AVX2-SLOW: # %bb.0:
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vmovaps (%rsi), %ymm0
+; AVX2-SLOW-NEXT: vmovaps 32(%rsi), %ymm5
+; AVX2-SLOW-NEXT: vmovaps (%rdx), %ymm3
+; AVX2-SLOW-NEXT: vmovaps 32(%rdx), %ymm6
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm2 = mem[1,0,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm1[0,0,2,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0],ymm2[1],ymm7[2,3],ymm2[4],ymm7[5,6],ymm2[7]
+; AVX2-SLOW-NEXT: vbroadcastsd (%rdx), %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm6[2,1,3,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm8 = ymm5[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[2,2,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1],ymm8[2,3],ymm7[4],ymm8[5,6],ymm7[7]
+; AVX2-SLOW-NEXT: vbroadcastsd 56(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm8[2],ymm7[3,4],ymm8[5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm8 = mem[1,0,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm9 = ymm4[0,0,2,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0],ymm8[1],ymm9[2,3],ymm8[4],ymm9[5,6],ymm8[7]
+; AVX2-SLOW-NEXT: vbroadcastsd 32(%rdx), %ymm9
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4],ymm9[5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm9 = ymm3[2,1,3,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm10 = ymm0[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0],ymm9[1],ymm10[2,3],ymm9[4],ymm10[5,6],ymm9[7]
+; AVX2-SLOW-NEXT: vbroadcastsd 24(%rdi), %ymm10
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm10[2],ymm9[3,4],ymm10[5],ymm9[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[1,1,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[1,1,2,2]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0],ymm4[1],ymm6[2,3],ymm4[4],ymm6[5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[1,1,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[1,1,2,2]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0],ymm1[1],ymm3[2,3],ymm1[4],ymm3[5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 128(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm9, 64(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm8, 96(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm7, 160(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rcx)
+; AVX2-SLOW-NEXT: vzeroupper
+; AVX2-SLOW-NEXT: retq
+;
+; AVX2-FAST-ALL-LABEL: store_i32_stride3_vf16:
+; AVX2-FAST-ALL: # %bb.0:
+; AVX2-FAST-ALL-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-FAST-ALL-NEXT: vmovaps 32(%rdi), %ymm1
+; AVX2-FAST-ALL-NEXT: vmovaps (%rsi), %ymm2
+; AVX2-FAST-ALL-NEXT: vmovaps 32(%rsi), %ymm3
+; AVX2-FAST-ALL-NEXT: vmovaps (%rdx), %ymm4
+; AVX2-FAST-ALL-NEXT: vmovaps 32(%rdx), %ymm5
+; AVX2-FAST-ALL-NEXT: vbroadcastf128 {{.*#+}} ymm6 = [1,0,2,2,1,0,2,2]
+; AVX2-FAST-ALL-NEXT: # ymm6 = mem[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm2, %ymm6, %ymm7
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm8 = ymm0[0,0,2,1]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1],ymm8[2,3],ymm7[4],ymm8[5,6],ymm7[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd (%rdx), %ymm8
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm8[2],ymm7[3,4],ymm8[5],ymm7[6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm8 = [5,6,5,6,5,6,7,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm3, %ymm8, %ymm9
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm10 = ymm5[2,1,3,3]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1],ymm9[2,3],ymm10[4],ymm9[5,6],ymm10[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd 56(%rdi), %ymm10
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm10[2],ymm9[3,4],ymm10[5],ymm9[6,7]
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm10 = ymm1[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm10[1],ymm5[2,3],ymm10[4],ymm5[5,6],ymm10[7]
+; AVX2-FAST-ALL-NEXT: vpermilps {{.*#+}} ymm10 = ymm3[0,0,3,3,4,4,7,7]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm10[2],ymm5[3,4],ymm10[5],ymm5[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm3, %ymm6, %ymm3
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd 32(%rdx), %ymm3
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7]
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm3 = ymm4[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3],ymm0[4],ymm3[5,6],ymm0[7]
+; AVX2-FAST-ALL-NEXT: vpermilps {{.*#+}} ymm3 = ymm2[0,0,3,3,4,4,7,7]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2],ymm0[3,4],ymm3[5],ymm0[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm3 = ymm4[2,1,3,3]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3],ymm3[4],ymm2[5,6],ymm3[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd 24(%rdi), %ymm3
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm2, 64(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm1, 96(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm5, 128(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm9, 160(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm7, (%rcx)
+; AVX2-FAST-ALL-NEXT: vzeroupper
+; AVX2-FAST-ALL-NEXT: retq
+;
+; AVX2-FAST-PERLANE-LABEL: store_i32_stride3_vf16:
+; AVX2-FAST-PERLANE: # %bb.0:
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rsi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rsi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdx), %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdx), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm2 = mem[1,0,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm1[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0],ymm2[1],ymm7[2,3],ymm2[4],ymm7[5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd (%rdx), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm7[2],ymm2[3,4],ymm7[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm6[2,1,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm8 = ymm5[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1],ymm8[2,3],ymm7[4],ymm8[5,6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 56(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm8[2],ymm7[3,4],ymm8[5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm8 = mem[1,0,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm9 = ymm4[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0],ymm8[1],ymm9[2,3],ymm8[4],ymm9[5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 32(%rdx), %ymm9
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4],ymm9[5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm9 = ymm3[2,1,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm10 = ymm0[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0],ymm9[1],ymm10[2,3],ymm9[4],ymm10[5,6],ymm9[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 24(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm10[2],ymm9[3,4],ymm10[5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0],ymm4[1],ymm6[2,3],ymm4[4],ymm6[5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0],ymm1[1],ymm3[2,3],ymm1[4],ymm3[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 128(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 64(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, 96(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm7, 160(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rcx)
+; AVX2-FAST-PERLANE-NEXT: vzeroupper
+; AVX2-FAST-PERLANE-NEXT: retq
+;
+; AVX512-LABEL: store_i32_stride3_vf16:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm0
+; AVX512-NEXT: vmovdqu64 (%rsi), %zmm1
+; AVX512-NEXT: vmovdqu64 (%rdx), %zmm2
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm3 = <0,16,u,1,17,u,2,18,u,3,19,u,4,20,u,5>
+; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm3
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm4 = [0,1,16,3,4,17,6,7,18,9,10,19,12,13,20,15]
+; AVX512-NEXT: vpermi2d %zmm2, %zmm3, %zmm4
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm3 = <5,21,u,6,22,u,7,23,u,8,24,u,9,25,u,10>
+; AVX512-NEXT: vpermi2d %zmm2, %zmm1, %zmm3
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm5 = [0,1,22,3,4,23,6,7,24,9,10,25,12,13,26,15]
+; AVX512-NEXT: vpermi2d %zmm0, %zmm3, %zmm5
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm3 = <10,27,u,11,28,u,12,29,u,13,30,u,14,31,u,15>
+; AVX512-NEXT: vpermi2d %zmm0, %zmm2, %zmm3
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,1,27,3,4,28,6,7,29,9,10,30,12,13,31,15]
+; AVX512-NEXT: vpermi2d %zmm1, %zmm3, %zmm0
+; AVX512-NEXT: vmovdqu64 %zmm0, 128(%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm5, 64(%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm4, (%rcx)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %in.vec0 = load <16 x i32>, <16 x i32>* %in.vecptr0, align 32
+ %in.vec1 = load <16 x i32>, <16 x i32>* %in.vecptr1, align 32
+ %in.vec2 = load <16 x i32>, <16 x i32>* %in.vecptr2, align 32
+
+ %concat01 = shufflevector <16 x i32> %in.vec0, <16 x i32> %in.vec1, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+ %concat2u = shufflevector <16 x i32> %in.vec2, <16 x i32> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+ %concat012 = shufflevector <32 x i32> %concat01, <32 x i32> %concat2u, <48 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+ %interleaved.vec = shufflevector <48 x i32> %concat012, <48 x i32> poison, <48 x i32> <i32 0, i32 16, i32 32, i32 1, i32 17, i32 33, i32 2, i32 18, i32 34, i32 3, i32 19, i32 35, i32 4, i32 20, i32 36, i32 5, i32 21, i32 37, i32 6, i32 22, i32 38, i32 7, i32 23, i32 39, i32 8, i32 24, i32 40, i32 9, i32 25, i32 41, i32 10, i32 26, i32 42, i32 11, i32 27, i32 43, i32 12, i32 28, i32 44, i32 13, i32 29, i32 45, i32 14, i32 30, i32 46, i32 15, i32 31, i32 47>
+
+ store <48 x i32> %interleaved.vec, <48 x i32>* %out.vec, align 32
+
+ ret void
+}