; 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,FALLBACK0 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx | FileCheck %s --check-prefixes=AVX,AVX1,AVX1-ONLY,FALLBACK1 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX1,AVX2,AVX2-ONLY,AVX2-SLOW,FALLBACK2 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX1,AVX2,AVX2-ONLY,AVX2-FAST,FALLBACK3 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX1,AVX2,AVX2-ONLY,AVX2-FAST-PERLANE,FALLBACK4 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512F,AVX512F-ONLY,AVX512-SLOW,AVX512F-SLOW,AVX512F-ONLY-SLOW,FALLBACK5 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512F,AVX512F-ONLY,AVX512-FAST,AVX512F-FAST,AVX512F-ONLY-FAST,FALLBACK6 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512F,AVX512DQ-ONLY,AVX512-SLOW,AVX512F-SLOW,AVX512DQ-SLOW,FALLBACK7 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512F,AVX512DQ-ONLY,AVX512-FAST,AVX512F-FAST,AVX512DQ-FAST,FALLBACK8 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512bw | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512BW,AVX512BW-ONLY,AVX512-SLOW,AVX512BW-SLOW,AVX512BW-ONLY-SLOW,FALLBACK9 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512bw,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512BW,AVX512BW-ONLY,AVX512-FAST,AVX512BW-FAST,AVX512BW-ONLY-FAST,FALLBACK10 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq,+avx512bw | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512BW,AVX512DQBW-ONLY,AVX512-SLOW,AVX512BW-SLOW,AVX512DQBW-SLOW,FALLBACK11 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl,+avx512dq,+avx512bw,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX2,AVX512,AVX512BW,AVX512DQBW-ONLY,AVX512-FAST,AVX512BW-FAST,AVX512DQBW-FAST,FALLBACK12 ; These patterns are produced by LoopVectorizer for interleaved loads. define void @load_i32_stride7_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind { ; SSE-LABEL: load_i32_stride7_vf2: ; SSE: # %bb.0: ; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; SSE-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; SSE-NEXT: movdqa (%rdi), %xmm0 ; SSE-NEXT: movdqa 16(%rdi), %xmm1 ; SSE-NEXT: movdqa 32(%rdi), %xmm2 ; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[3,3,3,3] ; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[1,1,1,1] ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm0[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1] ; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1] ; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1] ; SSE-NEXT: punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm2[2],xmm6[3],xmm2[3] ; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[3,3,3,3] ; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[1,1,1,1] ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm1[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1] ; SSE-NEXT: movdqa 48(%rdi), %xmm2 ; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1] ; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1] ; SSE-NEXT: movq %xmm0, (%rsi) ; SSE-NEXT: movq %xmm4, (%rdx) ; SSE-NEXT: movq %xmm5, (%rcx) ; SSE-NEXT: movq %xmm6, (%r8) ; SSE-NEXT: movq %xmm1, (%r9) ; SSE-NEXT: movq %xmm3, (%r10) ; SSE-NEXT: movq %xmm7, (%rax) ; SSE-NEXT: retq ; ; AVX1-ONLY-LABEL: load_i32_stride7_vf2: ; AVX1-ONLY: # %bb.0: ; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm3 ; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm4 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm3[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm2[0],xmm5[1],xmm2[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm4[0],xmm2[1],xmm4[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm6[1,0,2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm2[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],xmm4[1],xmm7[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,2,2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm1[0,0],ymm0[1,0],ymm1[4,4],ymm0[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,0],ymm0[2,0],ymm1[5,4],ymm0[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX1-ONLY-NEXT: vmovlps %xmm5, (%rsi) ; AVX1-ONLY-NEXT: vmovlps %xmm6, (%rdx) ; AVX1-ONLY-NEXT: vmovlps %xmm7, (%rcx) ; AVX1-ONLY-NEXT: vmovlps %xmm2, (%r8) ; AVX1-ONLY-NEXT: vmovlps %xmm3, (%r9) ; AVX1-ONLY-NEXT: vmovlps %xmm4, (%r10) ; AVX1-ONLY-NEXT: vmovlps %xmm0, (%rax) ; AVX1-ONLY-NEXT: vzeroupper ; AVX1-ONLY-NEXT: retq ; ; AVX2-ONLY-LABEL: load_i32_stride7_vf2: ; AVX2-ONLY: # %bb.0: ; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm0 ; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm1 ; AVX2-ONLY-NEXT: vbroadcastss 28(%rdi), %xmm2 ; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm3 ; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm4 ; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1] ; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm4[0],xmm3[1],xmm4[2,3] ; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[1,0,2,3] ; AVX2-ONLY-NEXT: vbroadcastss 8(%rdi), %xmm6 ; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3] ; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3] ; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3] ; AVX2-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [4,3,4,3] ; AVX2-ONLY-NEXT: # xmm4 = mem[0,0] ; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5,6,7] ; AVX2-ONLY-NEXT: vpermps %ymm7, %ymm4, %ymm4 ; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7] ; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,2,3,5,4,6,7] ; AVX2-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7 ; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7] ; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7] ; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX2-ONLY-NEXT: vmovlps %xmm2, (%rsi) ; AVX2-ONLY-NEXT: vmovlps %xmm5, (%rdx) ; AVX2-ONLY-NEXT: vmovlps %xmm6, (%rcx) ; AVX2-ONLY-NEXT: vmovlps %xmm3, (%r8) ; AVX2-ONLY-NEXT: vmovlps %xmm4, (%r9) ; AVX2-ONLY-NEXT: vmovlps %xmm7, (%r10) ; AVX2-ONLY-NEXT: vmovlps %xmm0, (%rax) ; AVX2-ONLY-NEXT: vzeroupper ; AVX2-ONLY-NEXT: retq ; ; AVX512-SLOW-LABEL: load_i32_stride7_vf2: ; AVX512-SLOW: # %bb.0: ; AVX512-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX512-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX512-SLOW-NEXT: vmovdqa (%rdi), %xmm0 ; AVX512-SLOW-NEXT: vmovdqa 32(%rdi), %xmm1 ; AVX512-SLOW-NEXT: vpinsrd $1, 28(%rdi), %xmm0, %xmm2 ; AVX512-SLOW-NEXT: vmovd %xmm1, %r11d ; AVX512-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[1,1,1,1] ; AVX512-SLOW-NEXT: vpinsrd $1, %r11d, %xmm3, %xmm3 ; AVX512-SLOW-NEXT: vpbroadcastd 8(%rdi), %xmm4 ; AVX512-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0],xmm1[1],xmm4[2,3] ; AVX512-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] ; AVX512-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3] ; AVX512-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm1 = [4,11,4,11] ; AVX512-SLOW-NEXT: vmovdqa 32(%rdi), %ymm5 ; AVX512-SLOW-NEXT: vmovdqa (%rdi), %ymm6 ; AVX512-SLOW-NEXT: vpermi2d %ymm5, %ymm6, %ymm1 ; AVX512-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm5[0],ymm6[1],ymm5[2,3,4],ymm6[5],ymm5[6,7] ; AVX512-SLOW-NEXT: vpshufd {{.*#+}} ymm7 = ymm7[1,0,2,3,5,4,6,7] ; AVX512-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm7 ; AVX512-SLOW-NEXT: vpshufd {{.*#+}} ymm6 = ymm6[2,3,2,3,6,7,6,7] ; AVX512-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0],ymm5[1],ymm6[2,3,4],ymm5[5],ymm6[6,7] ; AVX512-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5 ; AVX512-SLOW-NEXT: vmovq %xmm2, (%rsi) ; AVX512-SLOW-NEXT: vmovq %xmm3, (%rdx) ; AVX512-SLOW-NEXT: vmovq %xmm4, (%rcx) ; AVX512-SLOW-NEXT: vmovq %xmm0, (%r8) ; AVX512-SLOW-NEXT: vmovq %xmm1, (%r9) ; AVX512-SLOW-NEXT: vmovq %xmm7, (%r10) ; AVX512-SLOW-NEXT: vmovq %xmm5, (%rax) ; AVX512-SLOW-NEXT: vzeroupper ; AVX512-SLOW-NEXT: retq ; ; AVX512-FAST-LABEL: load_i32_stride7_vf2: ; AVX512-FAST: # %bb.0: ; AVX512-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX512-FAST-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX512-FAST-NEXT: vmovdqa (%rdi), %xmm0 ; AVX512-FAST-NEXT: vmovdqa 32(%rdi), %xmm1 ; AVX512-FAST-NEXT: vpinsrd $1, 28(%rdi), %xmm0, %xmm2 ; AVX512-FAST-NEXT: vpbroadcastq {{.*#+}} xmm3 = [1,4,1,4] ; AVX512-FAST-NEXT: vpermi2d %xmm1, %xmm0, %xmm3 ; AVX512-FAST-NEXT: vpbroadcastd 8(%rdi), %xmm4 ; AVX512-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0],xmm1[1],xmm4[2,3] ; AVX512-FAST-NEXT: vpbroadcastq {{.*#+}} xmm5 = [7,2,7,2] ; AVX512-FAST-NEXT: vpermi2d %xmm0, %xmm1, %xmm5 ; AVX512-FAST-NEXT: vpbroadcastq {{.*#+}} xmm0 = [4,11,4,11] ; AVX512-FAST-NEXT: vmovdqa 32(%rdi), %ymm1 ; AVX512-FAST-NEXT: vmovdqa (%rdi), %ymm6 ; AVX512-FAST-NEXT: vpermi2d %ymm1, %ymm6, %ymm0 ; AVX512-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm7 = [13,4,6,7,13,4,6,7] ; AVX512-FAST-NEXT: # ymm7 = mem[0,1,0,1] ; AVX512-FAST-NEXT: vpermi2d %ymm6, %ymm1, %ymm7 ; AVX512-FAST-NEXT: vextracti128 $1, %ymm7, %xmm7 ; AVX512-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm8 = [6,13,6,7,6,13,6,7] ; AVX512-FAST-NEXT: # ymm8 = mem[0,1,0,1] ; AVX512-FAST-NEXT: vpermi2d %ymm1, %ymm6, %ymm8 ; AVX512-FAST-NEXT: vextracti128 $1, %ymm8, %xmm1 ; AVX512-FAST-NEXT: vmovq %xmm2, (%rsi) ; AVX512-FAST-NEXT: vmovq %xmm3, (%rdx) ; AVX512-FAST-NEXT: vmovq %xmm4, (%rcx) ; AVX512-FAST-NEXT: vmovq %xmm5, (%r8) ; AVX512-FAST-NEXT: vmovq %xmm0, (%r9) ; AVX512-FAST-NEXT: vmovq %xmm7, (%r10) ; AVX512-FAST-NEXT: vmovq %xmm1, (%rax) ; AVX512-FAST-NEXT: vzeroupper ; AVX512-FAST-NEXT: retq %wide.vec = load <14 x i32>, ptr %in.vec, align 64 %strided.vec0 = shufflevector <14 x i32> %wide.vec, <14 x i32> poison, <2 x i32> %strided.vec1 = shufflevector <14 x i32> %wide.vec, <14 x i32> poison, <2 x i32> %strided.vec2 = shufflevector <14 x i32> %wide.vec, <14 x i32> poison, <2 x i32> %strided.vec3 = shufflevector <14 x i32> %wide.vec, <14 x i32> poison, <2 x i32> %strided.vec4 = shufflevector <14 x i32> %wide.vec, <14 x i32> poison, <2 x i32> %strided.vec5 = shufflevector <14 x i32> %wide.vec, <14 x i32> poison, <2 x i32> %strided.vec6 = shufflevector <14 x i32> %wide.vec, <14 x i32> poison, <2 x i32> store <2 x i32> %strided.vec0, ptr %out.vec0, align 64 store <2 x i32> %strided.vec1, ptr %out.vec1, align 64 store <2 x i32> %strided.vec2, ptr %out.vec2, align 64 store <2 x i32> %strided.vec3, ptr %out.vec3, align 64 store <2 x i32> %strided.vec4, ptr %out.vec4, align 64 store <2 x i32> %strided.vec5, ptr %out.vec5, align 64 store <2 x i32> %strided.vec6, ptr %out.vec6, align 64 ret void } define void @load_i32_stride7_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind { ; SSE-LABEL: load_i32_stride7_vf4: ; SSE: # %bb.0: ; SSE-NEXT: movdqa 96(%rdi), %xmm1 ; SSE-NEXT: movdqa 64(%rdi), %xmm0 ; SSE-NEXT: movdqa 80(%rdi), %xmm2 ; SSE-NEXT: movdqa (%rdi), %xmm5 ; SSE-NEXT: movdqa 16(%rdi), %xmm3 ; SSE-NEXT: movdqa 32(%rdi), %xmm4 ; SSE-NEXT: movdqa 48(%rdi), %xmm6 ; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm3[3,3,3,3] ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm5[1,1,1,1] ; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm5[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm5[2,2,3,3] ; SSE-NEXT: movdqa %xmm5, %xmm11 ; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm10[0],xmm11[1],xmm10[1] ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm6[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm11[0],xmm5[1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1] ; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm2[2,2,2,2] ; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm3[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm6[0],xmm10[1],xmm6[1] ; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm6[1,1,1,1] ; SSE-NEXT: punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm11[2],xmm6[3],xmm11[3] ; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm7[0],xmm6[1] ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm4[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm7[0],xmm8[1],xmm7[1] ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm2[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm7[0],xmm11[1],xmm7[1] ; SSE-NEXT: movsd {{.*#+}} xmm11 = xmm8[0],xmm11[1] ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm1[0,0,1,1] ; SSE-NEXT: movdqa %xmm0, %xmm8 ; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm7[0],xmm8[1],xmm7[1] ; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rdi ; SSE-NEXT: punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm4[2],xmm9[3],xmm4[3] ; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm9[0],xmm8[1] ; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[3,3,3,3] ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm3[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1] ; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm3[0],xmm4[1] ; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[2,2,2,2] ; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm10[0],xmm0[1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm12[0],xmm7[1],xmm12[1] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm7[0],xmm2[1] ; SSE-NEXT: movapd %xmm5, (%rsi) ; SSE-NEXT: movapd %xmm6, (%rdx) ; SSE-NEXT: movapd %xmm11, (%rcx) ; SSE-NEXT: movapd %xmm8, (%r8) ; SSE-NEXT: movapd %xmm4, (%r9) ; SSE-NEXT: movapd %xmm0, (%rdi) ; SSE-NEXT: movapd %xmm2, (%rax) ; SSE-NEXT: retq ; ; AVX1-ONLY-LABEL: load_i32_stride7_vf4: ; AVX1-ONLY: # %bb.0: ; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2 ; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm3 ; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm4 ; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm5 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm6 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm6[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm7[3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm4[0],xmm3[1],xmm4[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm8[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm8 = xmm8[0,1,2],xmm6[2] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm5[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1,2],xmm6[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm3[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm9[0],xmm4[1],xmm9[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1],xmm6[2,3] ; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm9 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm9[0],xmm5[1],xmm9[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm10[0,1,1,0] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm10[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm9[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm5[0,1,2],xmm10[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm10[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm9[0,1,2],xmm5[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm0[0,0],ymm1[1,0],ymm0[4,4],ymm1[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm10[0,1],xmm5[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1,2],xmm9[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm1[2,0],ymm0[5,4],ymm1[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm7[2,3] ; AVX1-ONLY-NEXT: vmovaps %xmm2, (%rsi) ; AVX1-ONLY-NEXT: vmovaps %xmm8, (%rdx) ; AVX1-ONLY-NEXT: vmovaps %xmm6, (%rcx) ; AVX1-ONLY-NEXT: vmovaps %xmm3, (%r8) ; AVX1-ONLY-NEXT: vmovaps %xmm4, (%r9) ; AVX1-ONLY-NEXT: vmovaps %xmm5, (%r10) ; AVX1-ONLY-NEXT: vmovaps %xmm0, (%rax) ; AVX1-ONLY-NEXT: vzeroupper ; AVX1-ONLY-NEXT: retq ; ; AVX2-SLOW-LABEL: load_i32_stride7_vf4: ; AVX2-SLOW: # %bb.0: ; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm0 ; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm2 = [0,7,6,u] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7] ; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm2, %ymm2 ; AVX2-SLOW-NEXT: vbroadcastss 84(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3] ; AVX2-SLOW-NEXT: vmovaps 80(%rdi), %xmm4 ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm3 = xmm4[2,2,2,2] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm5[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm6[0,1,2],xmm3[3] ; AVX2-SLOW-NEXT: vbroadcastss 8(%rdi), %xmm6 ; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %xmm7 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm7[1],xmm6[2,3] ; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %xmm8 ; AVX2-SLOW-NEXT: vbroadcastss %xmm8, %xmm9 ; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm4 = xmm9[2],xmm4[2],xmm9[3],xmm4[3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3] ; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %xmm6 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm9 = xmm6[0],xmm8[1],xmm6[2,3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm9 = xmm9[0,1,1,0] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm7 = xmm7[3,2,2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3] ; AVX2-SLOW-NEXT: vbroadcastss 100(%rdi), %xmm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm9 = xmm8[0,1,2],xmm9[3] ; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm10 = [4,3,4,3] ; AVX2-SLOW-NEXT: # xmm10 = mem[0,0] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5,6,7] ; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm10, %ymm10 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm8 = xmm6[0,1,2],xmm8[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm8 = xmm8[0,1,3,2] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm5, %xmm5 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm8[2,3] ; AVX2-SLOW-NEXT: vbroadcastss 80(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3] ; AVX2-SLOW-NEXT: vmovaps %xmm2, (%rsi) ; AVX2-SLOW-NEXT: vmovaps %xmm3, (%rdx) ; AVX2-SLOW-NEXT: vmovaps %xmm4, (%rcx) ; AVX2-SLOW-NEXT: vmovaps %xmm7, (%r8) ; AVX2-SLOW-NEXT: vmovaps %xmm9, (%r9) ; AVX2-SLOW-NEXT: vmovaps %xmm5, (%r10) ; AVX2-SLOW-NEXT: vmovaps %xmm0, (%rax) ; AVX2-SLOW-NEXT: vzeroupper ; AVX2-SLOW-NEXT: retq ; ; AVX2-FAST-LABEL: load_i32_stride7_vf4: ; AVX2-FAST: # %bb.0: ; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX2-FAST-NEXT: vmovaps (%rdi), %ymm0 ; AVX2-FAST-NEXT: vmovaps 32(%rdi), %ymm1 ; AVX2-FAST-NEXT: vmovaps {{.*#+}} xmm2 = [0,7,6,u] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7] ; AVX2-FAST-NEXT: vpermps %ymm3, %ymm2, %ymm2 ; AVX2-FAST-NEXT: vbroadcastss 84(%rdi), %xmm3 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3] ; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm3 = [1,0,7,0,1,0,7,0] ; AVX2-FAST-NEXT: # ymm3 = mem[0,1,0,1] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7] ; AVX2-FAST-NEXT: vpermps %ymm4, %ymm3, %ymm3 ; AVX2-FAST-NEXT: vmovaps 80(%rdi), %xmm5 ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm6 = xmm5[2,2,2,2] ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1,2],xmm6[3] ; AVX2-FAST-NEXT: vbroadcastss 8(%rdi), %xmm6 ; AVX2-FAST-NEXT: vmovaps 32(%rdi), %xmm7 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm7[1],xmm6[2,3] ; AVX2-FAST-NEXT: vmovaps 64(%rdi), %xmm8 ; AVX2-FAST-NEXT: vbroadcastss %xmm8, %xmm9 ; AVX2-FAST-NEXT: vunpckhps {{.*#+}} xmm5 = xmm9[2],xmm5[2],xmm9[3],xmm5[3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3] ; AVX2-FAST-NEXT: vmovaps 96(%rdi), %xmm6 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm9 = xmm6[0],xmm8[1],xmm6[2,3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm9 = xmm9[0,1,1,0] ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1,2],mem[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm7 = xmm7[3,2,2,3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3] ; AVX2-FAST-NEXT: vbroadcastss 100(%rdi), %xmm9 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm9 = xmm8[0,1,2],xmm9[3] ; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm10 = [4,3,4,3] ; AVX2-FAST-NEXT: # xmm10 = mem[0,0] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5,6,7] ; AVX2-FAST-NEXT: vpermps %ymm11, %ymm10, %ymm10 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm8 = xmm6[0,1,2],xmm8[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm8 = xmm8[0,1,3,2] ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm4[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm4, %xmm4 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm8[2,3] ; AVX2-FAST-NEXT: vbroadcastss 80(%rdi), %ymm8 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3] ; AVX2-FAST-NEXT: vmovaps %xmm2, (%rsi) ; AVX2-FAST-NEXT: vmovaps %xmm3, (%rdx) ; AVX2-FAST-NEXT: vmovaps %xmm5, (%rcx) ; AVX2-FAST-NEXT: vmovaps %xmm7, (%r8) ; AVX2-FAST-NEXT: vmovaps %xmm9, (%r9) ; AVX2-FAST-NEXT: vmovaps %xmm4, (%r10) ; AVX2-FAST-NEXT: vmovaps %xmm0, (%rax) ; AVX2-FAST-NEXT: vzeroupper ; AVX2-FAST-NEXT: retq ; ; AVX2-FAST-PERLANE-LABEL: load_i32_stride7_vf4: ; AVX2-FAST-PERLANE: # %bb.0: ; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm2 = [0,7,6,u] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm2, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 84(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3] ; AVX2-FAST-PERLANE-NEXT: vmovaps 80(%rdi), %xmm4 ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm3 = xmm4[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm5[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm6[0,1,2],xmm3[3] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 8(%rdi), %xmm6 ; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %xmm7 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm7[1],xmm6[2,3] ; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %xmm8 ; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm8, %xmm9 ; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm4 = xmm9[2],xmm4[2],xmm9[3],xmm4[3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3] ; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %xmm6 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm9 = xmm6[0],xmm8[1],xmm6[2,3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm9 = xmm9[0,1,1,0] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm7 = xmm7[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 100(%rdi), %xmm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm9 = xmm8[0,1,2],xmm9[3] ; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm10 = [4,3,4,3] ; AVX2-FAST-PERLANE-NEXT: # xmm10 = mem[0,0] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm10, %ymm10 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm8 = xmm6[0,1,2],xmm8[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm8 = xmm8[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm5, %xmm5 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm8[2,3] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 80(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3] ; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm2, (%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm3, (%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm4, (%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm7, (%r8) ; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm9, (%r9) ; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm5, (%r10) ; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, (%rax) ; AVX2-FAST-PERLANE-NEXT: vzeroupper ; AVX2-FAST-PERLANE-NEXT: retq ; ; AVX512-LABEL: load_i32_stride7_vf4: ; AVX512: # %bb.0: ; AVX512-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX512-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0 ; AVX512-NEXT: vmovdqa64 64(%rdi), %zmm1 ; AVX512-NEXT: vmovdqa {{.*#+}} xmm2 = [0,7,14,21] ; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm2 ; AVX512-NEXT: vmovdqa {{.*#+}} xmm3 = [1,8,15,22] ; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm3 ; AVX512-NEXT: vmovdqa {{.*#+}} xmm4 = [2,9,16,23] ; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm4 ; AVX512-NEXT: vmovdqa {{.*#+}} xmm5 = [3,10,17,24] ; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm5 ; AVX512-NEXT: vmovdqa {{.*#+}} xmm6 = [4,11,18,25] ; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm6 ; AVX512-NEXT: vmovdqa {{.*#+}} xmm7 = [5,12,19,26] ; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm7 ; AVX512-NEXT: vmovdqa {{.*#+}} xmm8 = [6,13,20,27] ; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm8 ; AVX512-NEXT: vmovdqa %xmm2, (%rsi) ; AVX512-NEXT: vmovdqa %xmm3, (%rdx) ; AVX512-NEXT: vmovdqa %xmm4, (%rcx) ; AVX512-NEXT: vmovdqa %xmm5, (%r8) ; AVX512-NEXT: vmovdqa %xmm6, (%r9) ; AVX512-NEXT: vmovdqa %xmm7, (%r10) ; AVX512-NEXT: vmovdqa %xmm8, (%rax) ; AVX512-NEXT: vzeroupper ; AVX512-NEXT: retq %wide.vec = load <28 x i32>, ptr %in.vec, align 64 %strided.vec0 = shufflevector <28 x i32> %wide.vec, <28 x i32> poison, <4 x i32> %strided.vec1 = shufflevector <28 x i32> %wide.vec, <28 x i32> poison, <4 x i32> %strided.vec2 = shufflevector <28 x i32> %wide.vec, <28 x i32> poison, <4 x i32> %strided.vec3 = shufflevector <28 x i32> %wide.vec, <28 x i32> poison, <4 x i32> %strided.vec4 = shufflevector <28 x i32> %wide.vec, <28 x i32> poison, <4 x i32> %strided.vec5 = shufflevector <28 x i32> %wide.vec, <28 x i32> poison, <4 x i32> %strided.vec6 = shufflevector <28 x i32> %wide.vec, <28 x i32> poison, <4 x i32> store <4 x i32> %strided.vec0, ptr %out.vec0, align 64 store <4 x i32> %strided.vec1, ptr %out.vec1, align 64 store <4 x i32> %strided.vec2, ptr %out.vec2, align 64 store <4 x i32> %strided.vec3, ptr %out.vec3, align 64 store <4 x i32> %strided.vec4, ptr %out.vec4, align 64 store <4 x i32> %strided.vec5, ptr %out.vec5, align 64 store <4 x i32> %strided.vec6, ptr %out.vec6, align 64 ret void } define void @load_i32_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind { ; SSE-LABEL: load_i32_stride7_vf8: ; SSE: # %bb.0: ; SSE-NEXT: subq $24, %rsp ; SSE-NEXT: movdqa 144(%rdi), %xmm9 ; SSE-NEXT: movdqa 80(%rdi), %xmm5 ; SSE-NEXT: movdqa (%rdi), %xmm12 ; SSE-NEXT: movdqa 16(%rdi), %xmm7 ; SSE-NEXT: movdqa 48(%rdi), %xmm6 ; SSE-NEXT: movdqa 192(%rdi), %xmm8 ; SSE-NEXT: movdqa 160(%rdi), %xmm10 ; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 112(%rdi), %xmm15 ; SSE-NEXT: movdqa 128(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[1,1,1,1] ; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm15[2,3,2,3] ; SSE-NEXT: movdqa %xmm15, %xmm3 ; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, (%rsp) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3] ; SSE-NEXT: movdqa %xmm7, %xmm11 ; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm12[1,1,1,1] ; SSE-NEXT: movdqa %xmm12, %xmm4 ; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1] ; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,2,2] ; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa %xmm10, %xmm4 ; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3] ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1] ; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,2,2] ; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa %xmm6, %xmm1 ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: movdqa 32(%rdi), %xmm4 ; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] ; SSE-NEXT: movdqa 176(%rdi), %xmm10 ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm10[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm2[0],xmm13[1] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm4[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1] ; SSE-NEXT: movdqa 64(%rdi), %xmm14 ; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm5[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm14[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1] ; SSE-NEXT: movdqa 208(%rdi), %xmm3 ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,0,1,1] ; SSE-NEXT: movdqa %xmm10, %xmm7 ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[2,2,3,3] ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm9[2],xmm1[3],xmm9[3] ; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1] ; SSE-NEXT: movdqa 96(%rdi), %xmm5 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[0,0,1,1] ; SSE-NEXT: movdqa %xmm14, %xmm15 ; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3] ; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3] ; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload ; SSE-NEXT: movdqa %xmm6, %xmm12 ; SSE-NEXT: punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm10[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm3[0],xmm9[1],xmm3[1] ; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm12[0],xmm9[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[3,3,3,3] ; SSE-NEXT: movdqa %xmm11, %xmm12 ; SSE-NEXT: movdqa %xmm11, %xmm4 ; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm14[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm4[0],xmm2[1] ; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm3[2,2,2,2] ; SSE-NEXT: punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm4[2],xmm10[3],xmm4[3] ; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm6[1,1,1,1] ; SSE-NEXT: movdqa %xmm6, %xmm11 ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm4[0],xmm10[1] ; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm5[2,2,2,2] ; SSE-NEXT: punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm4[2],xmm14[3],xmm4[3] ; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm12[1,1,1,1] ; SSE-NEXT: movdqa %xmm12, %xmm6 ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm4[0],xmm14[1] ; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm1[1,1,1,1] ; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm11[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1] ; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload ; SSE-NEXT: # xmm4 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1] ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm12[0],xmm4[1] ; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,1,1] ; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm6[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload ; SSE-NEXT: # xmm3 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm12[0],xmm3[1] ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, (%rsi) ; SSE-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 16(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, (%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%rdx) ; SSE-NEXT: movapd %xmm8, (%rcx) ; SSE-NEXT: movapd %xmm13, 16(%rcx) ; SSE-NEXT: movapd %xmm15, (%r8) ; SSE-NEXT: movapd %xmm7, 16(%r8) ; SSE-NEXT: movapd %xmm2, (%r9) ; SSE-NEXT: movapd %xmm9, 16(%r9) ; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; SSE-NEXT: movapd %xmm14, (%rax) ; SSE-NEXT: movapd %xmm10, 16(%rax) ; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; SSE-NEXT: movapd %xmm3, (%rax) ; SSE-NEXT: movapd %xmm4, 16(%rax) ; SSE-NEXT: addq $24, %rsp ; SSE-NEXT: retq ; ; AVX1-ONLY-LABEL: load_i32_stride7_vf8: ; AVX1-ONLY: # %bb.0: ; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm4 ; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm8 ; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm11 ; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm12 ; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm12[0],ymm2[0],ymm12[2],ymm2[2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5 ; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm13 ; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm9 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm13[0,1],xmm5[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[0,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm5 ; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm6 ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm6[1],xmm5[1] ; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm10 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm7 = zero,xmm7[1,2],xmm10[1] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm7[5,6,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm12[1,1],ymm11[2,2],ymm12[5,5],ymm11[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm7[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm9[0],xmm13[1],xmm9[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm14[0,1,2],ymm7[3,4,5,6,7] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm8[2,3],ymm4[0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm8[0,0],ymm14[3,3],ymm8[4,4],ymm14[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm14 = zero,xmm14[1,2],xmm10[2] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm13[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm9[1],xmm14[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm2[3,1],ymm11[0,3],ymm2[7,5],ymm11[4,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm12[2,1],ymm15[2,0],ymm12[6,5],ymm15[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm14[0,1],ymm15[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm8[0],ymm4[0],ymm8[2],ymm4[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm14[0,1,2],xmm10[3] ; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm14 ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm15[0,1,2,3,4],ymm10[5,6,7] ; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm15 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm11[1,0],ymm12[0,0],ymm11[5,4],ymm12[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm12[3,1],ymm11[0,2],ymm12[7,5],ymm11[4,6] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm9[0,1,2],xmm13[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm12 = xmm12[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1],ymm11[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm14[0,1],ymm4[1,3],ymm14[4,5],ymm4[5,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm8[0,2],ymm12[2,0],ymm8[4,6],ymm12[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm12[5,6,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm14[1,0],ymm4[2,0],ymm14[5,4],ymm4[6,4] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm8[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[3,0],ymm13[0,0],ymm8[7,4],ymm13[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[2,0],ymm12[2,0],ymm8[6,4],ymm12[6,4] ; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm12 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm12[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm15[0,1,2],xmm13[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm9[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = mem[0],xmm9[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm9[0,1],xmm13[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm14[2,1],ymm4[3,3],ymm14[6,5],ymm4[7,7] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm5[0],xmm6[1],xmm5[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm9[1,0],ymm4[2,0],ymm9[5,4],ymm4[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm12[0,1,2],xmm15[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm0[0,0],ymm1[1,0],ymm0[4,4],ymm1[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm13[0,1],xmm9[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm9[0,1,2,3],ymm4[4,5,6,7] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm14[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm14[3,0],ymm9[0,0],ymm14[7,4],ymm9[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm6[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[0,1],ymm9[2,0],ymm5[4,5],ymm9[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm12[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm1[2,0],ymm0[5,4],ymm1[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rsi) ; AVX1-ONLY-NEXT: vmovaps %ymm7, (%rdx) ; AVX1-ONLY-NEXT: vmovaps %ymm10, (%rcx) ; AVX1-ONLY-NEXT: vmovaps %ymm11, (%r8) ; AVX1-ONLY-NEXT: vmovaps %ymm8, (%r9) ; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rax) ; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rax) ; AVX1-ONLY-NEXT: vzeroupper ; AVX1-ONLY-NEXT: retq ; ; AVX2-SLOW-LABEL: load_i32_stride7_vf8: ; AVX2-SLOW: # %bb.0: ; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm9 ; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm4 ; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %ymm5 ; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm10 ; AVX2-SLOW-NEXT: vpbroadcastq 80(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = [0,7,6,u] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7] ; AVX2-SLOW-NEXT: vpermd %ymm6, %ymm3, %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %xmm6 ; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm3[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6 ; AVX2-SLOW-NEXT: vpbroadcastd 196(%rdi), %ymm7 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm6[5,6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm7 = ymm4[12,13,14,15],ymm5[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm5[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm7 = ymm7[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5,6],ymm6[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm10[0,1],ymm9[2,3],ymm10[4,5],ymm9[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm7 = ymm7[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm7 = ymm7[2,2,2,2] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm11 = ymm8[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2],ymm7[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 80(%rdi), %xmm7 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm11 = ymm10[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2],ymm7[3],ymm11[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 8(%rdi), %xmm11 ; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %xmm12 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm11 = xmm11[0],xmm12[1],xmm11[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm11 = ymm5[0],ymm4[0],ymm5[2],ymm4[2] ; AVX2-SLOW-NEXT: vpbroadcastd 204(%rdi), %ymm13 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,6],ymm13[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm11[5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm11 = xmm12[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[3,2,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm9 = ymm10[0],ymm9[1],ymm10[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm9 = ymm9[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm9 = ymm11[0,1],ymm9[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm5[0,2],ymm4[1,3],ymm5[4,6],ymm4[5,7] ; AVX2-SLOW-NEXT: vbroadcastss 208(%rdi), %ymm11 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,6],ymm11[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 100(%rdi), %xmm10 ; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %xmm11 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm10 = xmm11[0,1,2],xmm10[3] ; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm12 = [4,3,4,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7] ; AVX2-SLOW-NEXT: vpermd %ymm13, %ymm12, %ymm12 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm10 = xmm12[0,1],xmm10[2,3] ; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} ymm12 = [0,7,0,7,0,7,0,7] ; AVX2-SLOW-NEXT: vpermd %ymm5, %ymm12, %ymm13 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm4[6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 212(%rdi), %ymm14 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm13[4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4],ymm5[5],ymm4[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,0,3] ; AVX2-SLOW-NEXT: vpbroadcastd 216(%rdi), %ymm5 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7] ; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %xmm5 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm11 = xmm5[0,1,2],xmm11[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[0,1,3,2] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm8 = ymm8[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm8, %xmm8 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm8 = xmm8[0,1],xmm11[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3],ymm4[4,5,6,7] ; AVX2-SLOW-NEXT: vpermd 192(%rdi), %ymm12, %ymm8 ; AVX2-SLOW-NEXT: vpbroadcastd 136(%rdi), %xmm11 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm11[0],xmm3[1],xmm11[2,3] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm8[6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 80(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm8[0,1,2],xmm5[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa %ymm2, (%rsi) ; AVX2-SLOW-NEXT: vmovdqa %ymm6, (%rdx) ; AVX2-SLOW-NEXT: vmovdqa %ymm7, (%rcx) ; AVX2-SLOW-NEXT: vmovdqa %ymm9, (%r8) ; AVX2-SLOW-NEXT: vmovdqa %ymm10, (%r9) ; AVX2-SLOW-NEXT: vmovdqa %ymm4, (%r10) ; AVX2-SLOW-NEXT: vmovdqa %ymm0, (%rax) ; AVX2-SLOW-NEXT: vzeroupper ; AVX2-SLOW-NEXT: retq ; ; AVX2-FAST-LABEL: load_i32_stride7_vf8: ; AVX2-FAST: # %bb.0: ; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm9 ; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm5 ; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm0 ; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm1 ; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm10 ; AVX2-FAST-NEXT: vpbroadcastq 80(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = [0,7,6,u] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7] ; AVX2-FAST-NEXT: vpermd %ymm6, %ymm3, %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %xmm6 ; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %xmm3 ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm3[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6 ; AVX2-FAST-NEXT: vpbroadcastd 196(%rdi), %ymm7 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm6[5,6,7] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm6 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6 ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm7 = ymm4[12,13,14,15],ymm5[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm5[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm7 = ymm7[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5,6],ymm6[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm10[0,1],ymm9[2,3],ymm10[4,5],ymm9[6,7] ; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm8 = [5,6,5,6,5,6,5,6] ; AVX2-FAST-NEXT: vpermd %ymm7, %ymm8, %ymm7 ; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm11 = [1,0,7,0,1,0,7,0] ; AVX2-FAST-NEXT: # ymm11 = mem[0,1,0,1] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7] ; AVX2-FAST-NEXT: vpermd %ymm8, %ymm11, %ymm11 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2],ymm7[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7] ; AVX2-FAST-NEXT: vmovdqa 80(%rdi), %xmm7 ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm11 = ymm10[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2],ymm7[3],ymm11[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 8(%rdi), %xmm11 ; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %xmm12 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm11[0],xmm12[1],xmm11[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm11 = ymm5[0],ymm4[0],ymm5[2],ymm4[2] ; AVX2-FAST-NEXT: vpbroadcastd 204(%rdi), %ymm13 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,6],ymm13[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm11[5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm12[0,1,2],mem[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[3,2,2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm10[0],ymm9[1],ymm10[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm9 = ymm9[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm11[0,1],ymm9[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm10 = ymm5[0,2],ymm4[1,3],ymm5[4,6],ymm4[5,7] ; AVX2-FAST-NEXT: vbroadcastss 208(%rdi), %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,6],ymm11[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 100(%rdi), %xmm10 ; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %xmm11 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm10 = xmm11[0,1,2],xmm10[3] ; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm12 = [4,3,4,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7] ; AVX2-FAST-NEXT: vpermd %ymm13, %ymm12, %ymm12 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm10 = xmm12[0,1],xmm10[2,3] ; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm12 = [0,7,0,7,0,7,0,7] ; AVX2-FAST-NEXT: vpermd %ymm5, %ymm12, %ymm13 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm4[6,7] ; AVX2-FAST-NEXT: vpbroadcastd 212(%rdi), %ymm14 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm13[4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4],ymm5[5],ymm4[6,7] ; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [1,0,7,0,1,0,7,0] ; AVX2-FAST-NEXT: # ymm5 = mem[0,1,0,1] ; AVX2-FAST-NEXT: vpermd %ymm4, %ymm5, %ymm4 ; AVX2-FAST-NEXT: vpbroadcastd 216(%rdi), %ymm5 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7] ; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %xmm5 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm5[0,1,2],xmm11[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[0,1,3,2] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm8[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm8 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm8[0,1],xmm11[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-NEXT: vpermd 192(%rdi), %ymm12, %ymm8 ; AVX2-FAST-NEXT: vpbroadcastd 136(%rdi), %xmm11 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm3 = xmm11[0],xmm3[1],xmm11[2,3] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm8[6,7] ; AVX2-FAST-NEXT: vpbroadcastd 80(%rdi), %ymm8 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm5 = xmm8[0,1,2],xmm5[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm0 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm2, (%rsi) ; AVX2-FAST-NEXT: vmovdqa %ymm6, (%rdx) ; AVX2-FAST-NEXT: vmovdqa %ymm7, (%rcx) ; AVX2-FAST-NEXT: vmovdqa %ymm9, (%r8) ; AVX2-FAST-NEXT: vmovdqa %ymm10, (%r9) ; AVX2-FAST-NEXT: vmovdqa %ymm4, (%r10) ; AVX2-FAST-NEXT: vmovdqa %ymm0, (%rax) ; AVX2-FAST-NEXT: vzeroupper ; AVX2-FAST-NEXT: retq ; ; AVX2-FAST-PERLANE-LABEL: load_i32_stride7_vf8: ; AVX2-FAST-PERLANE: # %bb.0: ; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm9 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm4 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %ymm5 ; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm10 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 80(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = [0,7,6,u] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm6, %ymm3, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %xmm6 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm3[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 196(%rdi), %ymm7 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm6[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm6 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm7 = ymm4[12,13,14,15],ymm5[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm5[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm7 = ymm7[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5,6],ymm6[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm10[0,1],ymm9[2,3],ymm10[4,5],ymm9[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm7 = ymm7[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm7 = ymm7[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm8 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm11 = ymm8[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2],ymm7[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 80(%rdi), %xmm7 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm11 = ymm10[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2],ymm7[3],ymm11[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 8(%rdi), %xmm11 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %xmm12 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm11 = xmm11[0],xmm12[1],xmm11[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm11 = ymm5[0],ymm4[0],ymm5[2],ymm4[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 204(%rdi), %ymm13 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,6],ymm13[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm11[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm11 = xmm12[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm9 = ymm10[0],ymm9[1],ymm10[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm9 = ymm9[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm9 = ymm11[0,1],ymm9[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm5[0,2],ymm4[1,3],ymm5[4,6],ymm4[5,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 208(%rdi), %ymm11 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,6],ymm11[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 100(%rdi), %xmm10 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %xmm11 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm10 = xmm11[0,1,2],xmm10[3] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm12 = [4,3,4,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm13, %ymm12, %ymm12 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm10 = xmm12[0,1],xmm10[2,3] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} ymm12 = [0,7,0,7,0,7,0,7] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm5, %ymm12, %ymm13 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm4[6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 212(%rdi), %ymm14 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm13[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4],ymm5[5],ymm4[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 216(%rdi), %ymm5 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %xmm5 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm11 = xmm5[0,1,2],xmm11[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm8 = ymm8[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm8, %xmm8 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm8 = xmm8[0,1],xmm11[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermd 192(%rdi), %ymm12, %ymm8 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 136(%rdi), %xmm11 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm3 = xmm11[0],xmm3[1],xmm11[2,3] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm8[6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 80(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm8[0,1,2],xmm5[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4],ymm1[5],ymm0[6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm0 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, (%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, (%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, (%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, (%r8) ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm10, (%r9) ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, (%r10) ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, (%rax) ; AVX2-FAST-PERLANE-NEXT: vzeroupper ; AVX2-FAST-PERLANE-NEXT: retq ; ; AVX512-LABEL: load_i32_stride7_vf8: ; AVX512: # %bb.0: ; AVX512-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX512-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0 ; AVX512-NEXT: vmovdqa64 64(%rdi), %zmm1 ; AVX512-NEXT: vmovdqa64 128(%rdi), %zmm2 ; AVX512-NEXT: vmovdqa64 192(%rdi), %zmm3 ; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm4 = [0,3,10,17,0,3,10,17] ; AVX512-NEXT: # ymm4 = mem[0,1,0,1] ; AVX512-NEXT: vpermi2d %zmm3, %zmm2, %zmm4 ; AVX512-NEXT: vmovdqa {{.*#+}} ymm5 = [0,7,14,21,28,u,u,u] ; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm5 ; AVX512-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7] ; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [0,4,11,18,0,4,11,18] ; AVX512-NEXT: # ymm5 = mem[0,1,0,1] ; AVX512-NEXT: vpermi2d %zmm3, %zmm2, %zmm5 ; AVX512-NEXT: vmovdqa {{.*#+}} ymm6 = [1,8,15,22,29,u,u,u] ; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm6 ; AVX512-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6,7] ; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm6 = [0,5,12,19,0,5,12,19] ; AVX512-NEXT: # ymm6 = mem[0,1,0,1] ; AVX512-NEXT: vpermi2d %zmm3, %zmm2, %zmm6 ; AVX512-NEXT: vmovdqa {{.*#+}} ymm7 = [18,25,0,7,14,u,u,u] ; AVX512-NEXT: vpermi2d %zmm0, %zmm1, %zmm7 ; AVX512-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7] ; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm7 = [0,6,13,20,0,6,13,20] ; AVX512-NEXT: # ymm7 = mem[0,1,0,1] ; AVX512-NEXT: vpermi2d %zmm3, %zmm2, %zmm7 ; AVX512-NEXT: vmovdqa {{.*#+}} ymm8 = [19,26,1,8,15,u,u,u] ; AVX512-NEXT: vpermi2d %zmm0, %zmm1, %zmm8 ; AVX512-NEXT: vpblendd {{.*#+}} ymm7 = ymm8[0,1,2,3,4],ymm7[5,6,7] ; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm8 = [0,7,14,21,0,7,14,21] ; AVX512-NEXT: # ymm8 = mem[0,1,0,1] ; AVX512-NEXT: vpermi2d %zmm3, %zmm2, %zmm8 ; AVX512-NEXT: vmovdqa {{.*#+}} xmm9 = [4,11,18,25] ; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm9 ; AVX512-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7] ; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm9 = [1,8,15,22,1,8,15,22] ; AVX512-NEXT: # ymm9 = mem[0,1,0,1] ; AVX512-NEXT: vpermi2d %zmm3, %zmm2, %zmm9 ; AVX512-NEXT: vmovdqa {{.*#+}} xmm10 = [5,12,19,26] ; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm10 ; AVX512-NEXT: vpblendd {{.*#+}} ymm9 = ymm10[0,1,2,3],ymm9[4,5,6,7] ; AVX512-NEXT: vbroadcasti128 {{.*#+}} ymm10 = [2,9,16,23,2,9,16,23] ; AVX512-NEXT: # ymm10 = mem[0,1,0,1] ; AVX512-NEXT: vpermi2d %zmm3, %zmm2, %zmm10 ; AVX512-NEXT: vmovdqa {{.*#+}} xmm2 = [6,13,20,27] ; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm2 ; AVX512-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm10[4,5,6,7] ; AVX512-NEXT: vmovdqa %ymm4, (%rsi) ; AVX512-NEXT: vmovdqa %ymm5, (%rdx) ; AVX512-NEXT: vmovdqa %ymm6, (%rcx) ; AVX512-NEXT: vmovdqa %ymm7, (%r8) ; AVX512-NEXT: vmovdqa %ymm8, (%r9) ; AVX512-NEXT: vmovdqa %ymm9, (%r10) ; AVX512-NEXT: vmovdqa %ymm0, (%rax) ; AVX512-NEXT: vzeroupper ; AVX512-NEXT: retq %wide.vec = load <56 x i32>, ptr %in.vec, align 64 %strided.vec0 = shufflevector <56 x i32> %wide.vec, <56 x i32> poison, <8 x i32> %strided.vec1 = shufflevector <56 x i32> %wide.vec, <56 x i32> poison, <8 x i32> %strided.vec2 = shufflevector <56 x i32> %wide.vec, <56 x i32> poison, <8 x i32> %strided.vec3 = shufflevector <56 x i32> %wide.vec, <56 x i32> poison, <8 x i32> %strided.vec4 = shufflevector <56 x i32> %wide.vec, <56 x i32> poison, <8 x i32> %strided.vec5 = shufflevector <56 x i32> %wide.vec, <56 x i32> poison, <8 x i32> %strided.vec6 = shufflevector <56 x i32> %wide.vec, <56 x i32> poison, <8 x i32> store <8 x i32> %strided.vec0, ptr %out.vec0, align 64 store <8 x i32> %strided.vec1, ptr %out.vec1, align 64 store <8 x i32> %strided.vec2, ptr %out.vec2, align 64 store <8 x i32> %strided.vec3, ptr %out.vec3, align 64 store <8 x i32> %strided.vec4, ptr %out.vec4, align 64 store <8 x i32> %strided.vec5, ptr %out.vec5, align 64 store <8 x i32> %strided.vec6, ptr %out.vec6, align 64 ret void } define void @load_i32_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind { ; SSE-LABEL: load_i32_stride7_vf16: ; SSE: # %bb.0: ; SSE-NEXT: subq $440, %rsp # imm = 0x1B8 ; SSE-NEXT: movdqa 304(%rdi), %xmm3 ; SSE-NEXT: movdqa 272(%rdi), %xmm5 ; SSE-NEXT: movdqa 224(%rdi), %xmm15 ; SSE-NEXT: movdqa 240(%rdi), %xmm6 ; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 80(%rdi), %xmm7 ; SSE-NEXT: movdqa (%rdi), %xmm2 ; SSE-NEXT: movdqa 16(%rdi), %xmm8 ; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 48(%rdi), %xmm9 ; SSE-NEXT: movdqa 192(%rdi), %xmm14 ; SSE-NEXT: movdqa 160(%rdi), %xmm11 ; SSE-NEXT: movdqa 112(%rdi), %xmm4 ; SSE-NEXT: movdqa 128(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: movdqa %xmm4, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3] ; SSE-NEXT: movdqa %xmm11, %xmm12 ; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[3,3,3,3] ; SSE-NEXT: movdqa %xmm2, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3] ; SSE-NEXT: movdqa %xmm9, %xmm11 ; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3] ; SSE-NEXT: movdqa %xmm15, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3] ; SSE-NEXT: movdqa %xmm5, %xmm9 ; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 336(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 352(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: movdqa 416(%rdi), %xmm8 ; SSE-NEXT: movdqa 384(%rdi), %xmm5 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3] ; SSE-NEXT: movdqa %xmm5, %xmm13 ; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[2,2,2,2] ; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa %xmm12, %xmm1 ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1] ; SSE-NEXT: movdqa %xmm4, %xmm5 ; SSE-NEXT: movdqa 144(%rdi), %xmm4 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1] ; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,2,2,2] ; SSE-NEXT: movdqa %xmm7, %xmm12 ; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa %xmm11, %xmm1 ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1] ; SSE-NEXT: movdqa %xmm2, %xmm10 ; SSE-NEXT: movdqa 32(%rdi), %xmm7 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1] ; SSE-NEXT: movdqa %xmm7, (%rsp) # 16-byte Spill ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2] ; SSE-NEXT: movdqa %xmm9, %xmm1 ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: movdqa %xmm15, %xmm11 ; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1] ; SSE-NEXT: movdqa 256(%rdi), %xmm2 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: movdqa %xmm2, %xmm15 ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,2,2] ; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa %xmm13, %xmm1 ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1] ; SSE-NEXT: movdqa 368(%rdi), %xmm2 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: movdqa 176(%rdi), %xmm4 ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm4[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1] ; SSE-NEXT: movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3] ; SSE-NEXT: movdqa %xmm10, %xmm14 ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: movdqa 64(%rdi), %xmm9 ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm9[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm0[0],xmm6[1] ; SSE-NEXT: movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,3,2,3] ; SSE-NEXT: movdqa %xmm15, %xmm12 ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: movdqa 288(%rdi), %xmm15 ; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm3[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm13[2,3,2,3] ; SSE-NEXT: movdqa %xmm13, %xmm1 ; SSE-NEXT: movdqa %xmm2, %xmm11 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1] ; SSE-NEXT: movdqa 400(%rdi), %xmm13 ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm8[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm6[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 208(%rdi), %xmm10 ; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm10[0,0,1,1] ; SSE-NEXT: movdqa %xmm4, %xmm2 ; SSE-NEXT: movdqa %xmm4, %xmm0 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1] ; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm5[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm5[2],xmm4[3],xmm5[3] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 96(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa %xmm9, %xmm3 ; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa %xmm9, %xmm0 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1] ; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm14[2,2,3,3] ; SSE-NEXT: movdqa (%rsp), %xmm6 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm6[2],xmm4[3],xmm6[3] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 320(%rdi), %xmm0 ; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa %xmm0, %xmm7 ; SSE-NEXT: movdqa %xmm15, %xmm0 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload ; SSE-NEXT: # xmm4 = mem[2,2,3,3] ; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm12[2],xmm4[3],xmm12[3] ; SSE-NEXT: movdqa %xmm12, %xmm14 ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 432(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa %xmm13, %xmm4 ; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,2,3,3] ; SSE-NEXT: movdqa %xmm11, %xmm9 ; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm11[2],xmm0[3],xmm11[3] ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1] ; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload ; SSE-NEXT: movdqa %xmm8, %xmm5 ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3] ; SSE-NEXT: movdqa %xmm2, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm6[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; SSE-NEXT: movdqa %xmm4, %xmm0 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1] ; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm3[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm12 = xmm12[0],xmm2[0],xmm12[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm0[0],xmm12[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload ; SSE-NEXT: movdqa %xmm6, %xmm5 ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm15[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm7[0],xmm14[1],xmm7[1] ; SSE-NEXT: movdqa %xmm7, %xmm11 ; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm5[0],xmm14[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload ; SSE-NEXT: movdqa %xmm7, %xmm5 ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm13[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm3[0],xmm9[1],xmm3[1] ; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm5[0],xmm9[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,2,2,2] ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: movdqa %xmm8, %xmm5 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,2,2,2] ; SSE-NEXT: punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm0[2],xmm15[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2] ; SSE-NEXT: movdqa %xmm3, %xmm11 ; SSE-NEXT: punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm0[2],xmm13[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1] ; SSE-NEXT: movdqa %xmm7, %xmm2 ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1] ; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1] ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload ; SSE-NEXT: # xmm10 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm5[0],xmm10[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload ; SSE-NEXT: # xmm5 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm5[0],xmm7[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1] ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm6[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload ; SSE-NEXT: # xmm6 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm5[0],xmm6[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1] ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm2[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload ; SSE-NEXT: # xmm4 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm5[0],xmm4[1] ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 48(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 32(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, (%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 48(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 32(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, (%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 48(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 32(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, (%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 48(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 32(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, (%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%r8) ; SSE-NEXT: movapd %xmm9, 48(%r9) ; SSE-NEXT: movapd %xmm14, 32(%r9) ; SSE-NEXT: movapd %xmm12, (%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%r9) ; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; SSE-NEXT: movapd %xmm13, 48(%rax) ; SSE-NEXT: movapd %xmm15, 32(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, (%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%rax) ; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; SSE-NEXT: movapd %xmm4, 48(%rax) ; SSE-NEXT: movapd %xmm6, 32(%rax) ; SSE-NEXT: movapd %xmm7, (%rax) ; SSE-NEXT: movapd %xmm10, 16(%rax) ; SSE-NEXT: addq $440, %rsp # imm = 0x1B8 ; SSE-NEXT: retq ; ; AVX1-ONLY-LABEL: load_i32_stride7_vf16: ; AVX1-ONLY: # %bb.0: ; AVX1-ONLY-NEXT: subq $456, %rsp # imm = 0x1C8 ; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %ymm4 ; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm5 ; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm15 ; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm2 ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm7 ; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm0[0],ymm7[2],ymm0[2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1 ; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm13 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm12 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm12[1] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm15[0],ymm0[0],ymm15[2],ymm0[2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1 ; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm10 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm8 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm8[1] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm5 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm7[1,1],ymm5[2,2],ymm7[5,5],ymm5[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm11 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm11[0],xmm13[1],xmm11[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm6 ; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],ymm6[0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[0,0],ymm3[3,3],ymm1[4,4],ymm3[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm3[1,2],xmm12[2] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm3 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm15[1,1],ymm3[2,2],ymm15[5,5],ymm3[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm9 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm9[0],xmm10[1],xmm9[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm4 ; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm0[2,3],ymm4[0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[0,0],ymm14[3,3],ymm0[4,4],ymm14[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm14 = zero,xmm14[1,2],xmm8[2] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm14[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm13[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm11[1],xmm2[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[3,1],ymm5[0,3],ymm14[7,5],ymm5[4,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm7[2,1],ymm14[2,0],ymm7[6,5],ymm14[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm14[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm1[0],ymm6[0],ymm1[2],ymm6[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm14[0,1,2],xmm12[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm12[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm10[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm9[1],xmm2[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm12[3,1],ymm3[0,3],ymm12[7,5],ymm3[4,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm15[2,1],ymm12[2,0],ymm15[6,5],ymm12[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm12[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm0[0],ymm4[0],ymm0[2],ymm4[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm12[0,1,2],xmm8[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm8[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm5[1,0],ymm7[0,0],ymm5[5,4],ymm7[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm7[3,1],ymm2[0,2],ymm7[7,5],ymm2[4,6] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm11[0,1,2],xmm13[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm2[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm2 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm2[0,1],ymm6[1,3],ymm2[4,5],ymm6[5,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm1[0,2],ymm7[2,0],ymm1[4,6],ymm7[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm7[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[1,0],ymm15[0,0],ymm3[5,4],ymm15[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm15[3,1],ymm3[0,2],ymm15[7,5],ymm3[4,6] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm9[0,1,2],xmm10[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm5 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm5[0,1],ymm4[1,3],ymm5[4,5],ymm4[5,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm0[0,2],ymm8[2,0],ymm0[4,6],ymm8[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm8[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm3[0,0],ymm1[7,4],ymm3[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm2[1,0],ymm6[2,0],ymm2[5,4],ymm6[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm3[2,0],ymm1[6,4],ymm3[6,4] ; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm3 ; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm10 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm10[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm3[0,1,2],xmm8[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm11[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = mem[0],xmm11[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm11[0,1],xmm8[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm1[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm0[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm8[0,0],ymm0[7,4],ymm8[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm5[1,0],ymm4[2,0],ymm5[5,4],ymm4[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm8[2,0],ymm0[6,4],ymm8[6,4] ; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm8 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm8[0,1,0,1] ; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm12 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm12[0,1,2],xmm11[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm9[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = mem[0],xmm9[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm9[0,1],xmm11[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm2[2,1],ymm6[3,3],ymm2[6,5],ymm6[7,7] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm1[0],xmm0[1],xmm1[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm9[1,0],ymm6[2,0],ymm9[5,4],ymm6[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm10[0,1,2],xmm3[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm14[0,0],ymm13[1,0],ymm14[4,4],ymm13[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm9[0,1],xmm3[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm3[0,1,2,3],ymm6[4,5,6,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm5[2,1],ymm4[3,3],ymm5[6,5],ymm4[7,7] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0],xmm3[1],xmm7[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm6[1,0],ymm4[2,0],ymm6[5,4],ymm4[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm12[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm11[0,0],ymm12[1,0],ymm11[4,4],ymm12[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm9[0,1],xmm6[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4,5,6,7] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm2[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[3,0],ymm6[0,0],ymm2[7,4],ymm6[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm0[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm1[1],xmm6[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,0],ymm6[4,5],ymm2[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1,2],xmm10[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm14[1,0],ymm13[2,0],ymm14[5,4],ymm13[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1],xmm6[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm5[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[3,0],ymm6[0,0],ymm5[7,4],ymm6[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm3[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm7[1],xmm6[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,0],ymm6[4,5],ymm5[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1,2],xmm8[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm11[1,0],ymm12[2,0],ymm11[5,4],ymm12[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm8[0,1],xmm6[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm6, (%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r8) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r8) ; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r9) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r9) ; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rax) ; AVX1-ONLY-NEXT: vmovaps %ymm15, (%rax) ; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rax) ; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rax) ; AVX1-ONLY-NEXT: addq $456, %rsp # imm = 0x1C8 ; AVX1-ONLY-NEXT: vzeroupper ; AVX1-ONLY-NEXT: retq ; ; AVX2-SLOW-LABEL: load_i32_stride7_vf16: ; AVX2-SLOW: # %bb.0: ; AVX2-SLOW-NEXT: subq $264, %rsp # imm = 0x108 ; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %ymm15 ; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %ymm6 ; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %ymm5 ; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 224(%rdi), %ymm10 ; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm9 ; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm4 ; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm12 ; AVX2-SLOW-NEXT: vpbroadcastq 80(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = [0,7,6,u] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5],ymm4[6],ymm9[7] ; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm7 ; AVX2-SLOW-NEXT: vpermd %ymm3, %ymm2, %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %xmm4 ; AVX2-SLOW-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-SLOW-NEXT: vpbroadcastd 196(%rdi), %ymm4 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5],ymm0[6],ymm10[7] ; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm2, %ymm1 ; AVX2-SLOW-NEXT: vpbroadcastq 304(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpbroadcastd 420(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm2 = ymm8[12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10,11],ymm8[28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1],ymm15[2,3],ymm5[4,5],ymm15[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0],ymm10[1],ymm0[2,3,4],ymm10[5],ymm0[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm3 ; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm4 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm10 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm12[0,1],ymm10[2,3],ymm12[4,5],ymm10[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm4[2,2,2,2] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0],ymm9[1],ymm7[2,3,4],ymm9[5],ymm7[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm14 = ymm4[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm14 = ymm14[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 80(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm3 = ymm12[8,9,10,11,12,13,14,15],ymm10[0,1,2,3,4,5,6,7],ymm12[24,25,26,27,28,29,30,31],ymm10[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 8(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %xmm14 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0],xmm14[1],xmm3[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[2],ymm2[2] ; AVX2-SLOW-NEXT: vpbroadcastd 204(%rdi), %ymm13 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm13[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 304(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm3 = ymm5[8,9,10,11,12,13,14,15],ymm15[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm15[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 232(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %xmm13 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0],xmm13[1],xmm3[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm3 = ymm6[0],ymm8[0],ymm6[2],ymm8[2] ; AVX2-SLOW-NEXT: vpbroadcastd 428(%rdi), %ymm11 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm11[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0],ymm10[1],ymm12[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm14[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,2,2,3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm1[0,2],ymm2[1,3],ymm1[4,6],ymm2[5,7] ; AVX2-SLOW-NEXT: vbroadcastss 208(%rdi), %ymm10 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm10[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0],ymm15[1],ymm5[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm13[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,2,2,3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm6[0,2],ymm8[1,3],ymm6[4,6],ymm8[5,7] ; AVX2-SLOW-NEXT: vbroadcastss 432(%rdi), %ymm5 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm5[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpbroadcastd 100(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm3[0,1,2],xmm0[3] ; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm5 = [4,3,4,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm7[0,1,2,3],ymm9[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa %ymm7, %ymm12 ; AVX2-SLOW-NEXT: vpermd %ymm10, %ymm5, %ymm10 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm10 = xmm10[0,1],xmm0[2,3] ; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} ymm0 = [0,7,0,7,0,7,0,7] ; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm11 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm2[6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 212(%rdi), %ymm13 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,6],ymm13[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm15 = ymm10[0,1,2,3],ymm11[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm14[0,1,2,3],ymm7[4,5,6,7] ; AVX2-SLOW-NEXT: vpermd %ymm10, %ymm5, %ymm5 ; AVX2-SLOW-NEXT: vpbroadcastd 324(%rdi), %xmm10 ; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %xmm11 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm10 = xmm11[0,1,2],xmm10[3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm5[0,1],xmm10[2,3] ; AVX2-SLOW-NEXT: vpermd %ymm6, %ymm0, %ymm10 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm8[6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 436(%rdi), %ymm13 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,6],ymm13[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm5[0,1,2,3],ymm10[4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3,4],ymm1[5],ymm2[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,3] ; AVX2-SLOW-NEXT: vpbroadcastd 216(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %xmm5 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1,2],xmm3[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,3,2] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm4[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1,2],xmm11[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,3,2] ; AVX2-SLOW-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm4 = mem[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm4 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0],ymm6[1],ymm8[2,3,4],ymm6[5],ymm8[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,0,3] ; AVX2-SLOW-NEXT: vpbroadcastd 440(%rdi), %ymm6 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm6[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 136(%rdi), %xmm4 ; AVX2-SLOW-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4 ; AVX2-SLOW-NEXT: vpermd 192(%rdi), %ymm0, %ymm6 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 80(%rdi), %ymm6 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm6 = ymm9[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0],ymm12[1],ymm6[2,3,4],ymm12[5],ymm6[6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm6, %xmm6 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 360(%rdi), %xmm5 ; AVX2-SLOW-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3] ; AVX2-SLOW-NEXT: vpermd 416(%rdi), %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 304(%rdi), %ymm5 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm5[0,1,2],xmm3[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm5 = ymm7[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0],ymm14[1],ymm5[2,3,4],ymm14[5],ymm5[6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%r8) ; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm3, (%r8) ; AVX2-SLOW-NEXT: vmovdqa %ymm10, 32(%r9) ; AVX2-SLOW-NEXT: vmovdqa %ymm15, (%r9) ; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-SLOW-NEXT: vmovdqa %ymm2, 32(%rax) ; AVX2-SLOW-NEXT: vmovdqa %ymm1, (%rax) ; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-SLOW-NEXT: vmovdqa %ymm0, 32(%rax) ; AVX2-SLOW-NEXT: vmovdqa %ymm4, (%rax) ; AVX2-SLOW-NEXT: addq $264, %rsp # imm = 0x108 ; AVX2-SLOW-NEXT: vzeroupper ; AVX2-SLOW-NEXT: retq ; ; AVX2-FAST-LABEL: load_i32_stride7_vf16: ; AVX2-FAST: # %bb.0: ; AVX2-FAST-NEXT: subq $296, %rsp # imm = 0x128 ; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm9 ; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm10 ; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm0 ; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm5 ; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm6 ; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm11 ; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm1 ; AVX2-FAST-NEXT: vpbroadcastq 80(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = [0,7,6,u] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm5[0,1,2,3,4,5],ymm6[6],ymm5[7] ; AVX2-FAST-NEXT: vpermd %ymm8, %ymm7, %ymm8 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %xmm8 ; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %xmm12 ; AVX2-FAST-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm8 = xmm8[1],xmm12[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8 ; AVX2-FAST-NEXT: vpbroadcastd 196(%rdi), %ymm12 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm12[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm8[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm0[6],ymm4[7] ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm13 ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm2, %ymm7, %ymm2 ; AVX2-FAST-NEXT: vpbroadcastq 304(%rdi), %ymm7 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm7[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %xmm7 ; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %xmm0 ; AVX2-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm7 = xmm7[1],xmm0[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7 ; AVX2-FAST-NEXT: vpbroadcastd 420(%rdi), %ymm8 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm7[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm7 = ymm9[12,13,14,15],ymm10[0,1,2,3,4,5,6,7,8,9,10,11],ymm9[28,29,30,31],ymm10[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm7 = ymm7[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm1[0,1],ymm11[2,3],ymm1[4,5],ymm11[6,7] ; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm12 = [5,6,5,6,5,6,5,6] ; AVX2-FAST-NEXT: vpermd %ymm7, %ymm12, %ymm8 ; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm7 = [1,0,7,7,5,4,7,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0],ymm5[1],ymm6[2,3,4],ymm5[5],ymm6[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm0, %ymm7, %ymm14 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm14[0,1,2],ymm8[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %ymm14 ; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm15 ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm8 = ymm14[12,13,14,15],ymm15[0,1,2,3,4,5,6,7,8,9,10,11],ymm14[28,29,30,31],ymm15[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm8[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm8 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm8[2,3],ymm3[4,5],ymm8[6,7] ; AVX2-FAST-NEXT: vpermd %ymm2, %ymm12, %ymm12 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0],ymm4[1],ymm13[2,3,4],ymm4[5],ymm13[6,7] ; AVX2-FAST-NEXT: vpermd %ymm2, %ymm7, %ymm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm12[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 80(%rdi), %xmm0 ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm4 = ymm1[8,9,10,11,12,13,14,15],ymm11[0,1,2,3,4,5,6,7],ymm1[24,25,26,27,28,29,30,31],ymm11[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3],ymm4[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 8(%rdi), %xmm4 ; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %xmm12 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0],xmm12[1],xmm4[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm4 = ymm10[0],ymm9[0],ymm10[2],ymm9[2] ; AVX2-FAST-NEXT: vpbroadcastd 204(%rdi), %ymm7 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm7[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 304(%rdi), %xmm0 ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm4 = ymm3[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm3[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3],ymm4[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 232(%rdi), %xmm4 ; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %xmm7 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0],xmm7[1],xmm4[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm4 = ymm15[0],ymm14[0],ymm15[2],ymm14[2] ; AVX2-FAST-NEXT: vpbroadcastd 428(%rdi), %ymm13 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm13[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm11[1],ymm1[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm12[0,1,2],mem[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,2],ymm9[1,3],ymm10[4,6],ymm9[5,7] ; AVX2-FAST-NEXT: vbroadcastss 208(%rdi), %ymm4 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm4[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0],ymm8[1],ymm3[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm7[0,1,2],mem[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm15[0,2],ymm14[1,3],ymm15[4,6],ymm14[5,7] ; AVX2-FAST-NEXT: vbroadcastss 432(%rdi), %ymm3 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill ; AVX2-FAST-NEXT: vpbroadcastd 100(%rdi), %xmm0 ; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %xmm1 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] ; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm3 = [4,3,4,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm5[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm11 ; AVX2-FAST-NEXT: vmovdqa %ymm6, %ymm12 ; AVX2-FAST-NEXT: vpermd %ymm4, %ymm3, %ymm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm0[2,3] ; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm0 = [0,7,0,7,0,7,0,7] ; AVX2-FAST-NEXT: vpermd %ymm10, %ymm0, %ymm7 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm9[6,7] ; AVX2-FAST-NEXT: vpbroadcastd 212(%rdi), %ymm8 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm7[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm5[4,5,6,7] ; AVX2-FAST-NEXT: vpermd %ymm4, %ymm3, %ymm3 ; AVX2-FAST-NEXT: vpbroadcastd 324(%rdi), %xmm4 ; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %xmm13 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm13[0,1,2],xmm4[3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3] ; AVX2-FAST-NEXT: vpermd %ymm15, %ymm0, %ymm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm14[6,7] ; AVX2-FAST-NEXT: vpbroadcastd 436(%rdi), %ymm7 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm7[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm3[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0],ymm10[1],ymm9[2,3,4],ymm10[5],ymm9[6,7] ; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = [1,0,3,3,1,0,7,7] ; AVX2-FAST-NEXT: vpermd %ymm3, %ymm4, %ymm3 ; AVX2-FAST-NEXT: vpbroadcastd 216(%rdi), %ymm7 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm7[7] ; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %xmm10 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm10[0,1,2],xmm1[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,3,2] ; AVX2-FAST-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm7, %xmm7 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm7[0,1],xmm1[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %xmm9 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm9[0,1,2],xmm13[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,3,2] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0],ymm15[1],ymm14[2,3,4],ymm15[5],ymm14[6,7] ; AVX2-FAST-NEXT: vpermd %ymm2, %ymm4, %ymm2 ; AVX2-FAST-NEXT: vpbroadcastd 440(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 136(%rdi), %xmm1 ; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vpermd 192(%rdi), %ymm0, %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7] ; AVX2-FAST-NEXT: vpbroadcastd 80(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1,2],xmm10[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm11[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0],ymm12[1],ymm4[2,3,4],ymm12[5],ymm4[6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 360(%rdi), %xmm3 ; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3] ; AVX2-FAST-NEXT: vpermd 416(%rdi), %ymm0, %ymm0 ; AVX2-FAST-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7] ; AVX2-FAST-NEXT: vpbroadcastd 304(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1,2],xmm9[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm5[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0],ymm6[1],ymm4[2,3,4],ymm6[5],ymm4[6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm3, (%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm3, (%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rcx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm3, (%rcx) ; AVX2-FAST-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%r8) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm3, (%r8) ; AVX2-FAST-NEXT: vmovdqa %ymm8, 32(%r9) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm3, (%r9) ; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-NEXT: vmovdqa %ymm2, 32(%rax) ; AVX2-FAST-NEXT: vmovdqa %ymm7, (%rax) ; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-NEXT: vmovdqa %ymm0, 32(%rax) ; AVX2-FAST-NEXT: vmovdqa %ymm1, (%rax) ; AVX2-FAST-NEXT: addq $296, %rsp # imm = 0x128 ; AVX2-FAST-NEXT: vzeroupper ; AVX2-FAST-NEXT: retq ; ; AVX2-FAST-PERLANE-LABEL: load_i32_stride7_vf16: ; AVX2-FAST-PERLANE: # %bb.0: ; AVX2-FAST-PERLANE-NEXT: subq $264, %rsp # imm = 0x108 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %ymm15 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm6 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm5 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm10 ; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm9 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm4 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm12 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 80(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = [0,7,6,u] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5],ymm4[6],ymm9[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm7 ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm3, %ymm2, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %xmm4 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 196(%rdi), %ymm4 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5],ymm0[6],ymm10[7] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm2, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 304(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 420(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm2 = ymm8[12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10,11],ymm8[28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1],ymm15[2,3],ymm5[4,5],ymm15[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0],ymm10[1],ymm0[2,3,4],ymm10[5],ymm0[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm4 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm10 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm12[0,1],ymm10[2,3],ymm12[4,5],ymm10[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm4[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0],ymm9[1],ymm7[2,3,4],ymm9[5],ymm7[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm14 = ymm4[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm14 = ymm14[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 80(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm3 = ymm12[8,9,10,11,12,13,14,15],ymm10[0,1,2,3,4,5,6,7],ymm12[24,25,26,27,28,29,30,31],ymm10[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 8(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %xmm14 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0],xmm14[1],xmm3[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[2],ymm2[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 204(%rdi), %ymm13 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm13[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 304(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm3 = ymm5[8,9,10,11,12,13,14,15],ymm15[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm15[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 232(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %xmm13 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0],xmm13[1],xmm3[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm3 = ymm6[0],ymm8[0],ymm6[2],ymm8[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 428(%rdi), %ymm11 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm11[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0],ymm10[1],ymm12[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm3 = xmm14[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm1[0,2],ymm2[1,3],ymm1[4,6],ymm2[5,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 208(%rdi), %ymm10 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm10[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0],ymm15[1],ymm5[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm3 = xmm13[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm6[0,2],ymm8[1,3],ymm6[4,6],ymm8[5,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 432(%rdi), %ymm5 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm5[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 100(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm3[0,1,2],xmm0[3] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm5 = [4,3,4,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm10 = ymm7[0,1,2,3],ymm9[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, %ymm12 ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm10, %ymm5, %ymm10 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm10 = xmm10[0,1],xmm0[2,3] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} ymm0 = [0,7,0,7,0,7,0,7] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm11 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm2[6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 212(%rdi), %ymm13 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,6],ymm13[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm15 = ymm10[0,1,2,3],ymm11[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm10 = ymm14[0,1,2,3],ymm7[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm10, %ymm5, %ymm5 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 324(%rdi), %xmm10 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %xmm11 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm10 = xmm11[0,1,2],xmm10[3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm5[0,1],xmm10[2,3] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm6, %ymm0, %ymm10 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm8[6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 436(%rdi), %ymm13 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,6],ymm13[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm10 = ymm5[0,1,2,3],ymm10[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3,4],ymm1[5],ymm2[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 216(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %xmm5 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1,2],xmm3[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm3 = ymm4[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1,2],xmm11[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm4 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0],ymm6[1],ymm8[2,3,4],ymm6[5],ymm8[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 440(%rdi), %ymm6 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm6[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 136(%rdi), %xmm4 ; AVX2-FAST-PERLANE-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4 ; AVX2-FAST-PERLANE-NEXT: vpermd 192(%rdi), %ymm0, %ymm6 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 80(%rdi), %ymm6 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm6 = ymm9[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0],ymm12[1],ymm6[2,3,4],ymm12[5],ymm6[6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm6, %xmm6 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 360(%rdi), %xmm5 ; AVX2-FAST-PERLANE-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3] ; AVX2-FAST-PERLANE-NEXT: vpermd 416(%rdi), %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 304(%rdi), %ymm5 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm3 = xmm5[0,1,2],xmm3[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm5 = ymm7[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0],ymm14[1],ymm5[2,3,4],ymm14[5],ymm5[6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm5 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%r8) ; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%r8) ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm10, 32(%r9) ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm15, (%r9) ; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, 32(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, (%rax) ; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, 32(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, (%rax) ; AVX2-FAST-PERLANE-NEXT: addq $264, %rsp # imm = 0x108 ; AVX2-FAST-PERLANE-NEXT: vzeroupper ; AVX2-FAST-PERLANE-NEXT: retq ; ; AVX512F-LABEL: load_i32_stride7_vf16: ; AVX512F: # %bb.0: ; AVX512F-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX512F-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX512F-NEXT: vmovdqa64 384(%rdi), %zmm1 ; AVX512F-NEXT: vmovdqa64 320(%rdi), %zmm5 ; AVX512F-NEXT: vmovdqa64 256(%rdi), %zmm4 ; AVX512F-NEXT: vmovdqa64 (%rdi), %zmm0 ; AVX512F-NEXT: vmovdqa64 64(%rdi), %zmm2 ; AVX512F-NEXT: vmovdqa64 128(%rdi), %zmm8 ; AVX512F-NEXT: vmovdqa64 192(%rdi), %zmm7 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm6 = [8,15,0,0,0,19,26,1,8,15,0,0,0,19,26,1] ; AVX512F-NEXT: # zmm6 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm8, %zmm7, %zmm6 ; AVX512F-NEXT: vmovdqa {{.*#+}} ymm3 = [0,7,14,21,28,u,u,u] ; AVX512F-NEXT: vpermi2d %zmm2, %zmm0, %zmm3 ; AVX512F-NEXT: movw $992, %di # imm = 0x3E0 ; AVX512F-NEXT: kmovw %edi, %k1 ; AVX512F-NEXT: vmovdqa32 %zmm6, %zmm3 {%k1} ; AVX512F-NEXT: vbroadcasti32x4 {{.*#+}} zmm6 = [20,27,6,13,20,27,6,13,20,27,6,13,20,27,6,13] ; AVX512F-NEXT: # zmm6 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm5, %zmm4, %zmm6 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm9 = [0,0,10,11,12,13,18,25,0,0,10,11,12,13,18,25] ; AVX512F-NEXT: # zmm9 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm1, %zmm6, %zmm9 ; AVX512F-NEXT: movb $-32, %dil ; AVX512F-NEXT: kmovw %edi, %k1 ; AVX512F-NEXT: vmovdqa64 %zmm9, %zmm3 {%k1} ; AVX512F-NEXT: vbroadcasti32x4 {{.*#+}} zmm6 = [25,4,11,18,25,4,11,18,25,4,11,18,25,4,11,18] ; AVX512F-NEXT: # zmm6 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm8, %zmm9 ; AVX512F-NEXT: vbroadcasti32x4 {{.*#+}} zmm10 = [27,6,13,20,27,6,13,20,27,6,13,20,27,6,13,20] ; AVX512F-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm7, %zmm8, %zmm10 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm11 = [28,0,0,0,0,7,14,21,28,0,0,0,0,7,14,21] ; AVX512F-NEXT: # zmm11 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm7, %zmm8, %zmm11 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm12 = [29,0,0,0,1,8,15,22,29,0,0,0,1,8,15,22] ; AVX512F-NEXT: # zmm12 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm7, %zmm8, %zmm12 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm13 = [14,0,0,0,18,25,0,7,14,0,0,0,18,25,0,7] ; AVX512F-NEXT: # zmm13 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm8, %zmm7, %zmm13 ; AVX512F-NEXT: vpermt2d %zmm7, %zmm6, %zmm8 ; AVX512F-NEXT: vmovdqa {{.*#+}} ymm14 = [1,8,15,22,29,u,u,u] ; AVX512F-NEXT: vpermi2d %zmm2, %zmm0, %zmm14 ; AVX512F-NEXT: movw $480, %di # imm = 0x1E0 ; AVX512F-NEXT: kmovw %edi, %k2 ; AVX512F-NEXT: vmovdqa32 %zmm8, %zmm14 {%k2} ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,7,14,21,28,0,0,0,0,7,14,21,28,0,0] ; AVX512F-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm5, %zmm4, %zmm8 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm15 = [0,9,10,11,12,13,19,26,0,9,10,11,12,13,19,26] ; AVX512F-NEXT: # zmm15 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm1, %zmm8, %zmm15 ; AVX512F-NEXT: movw $-512, %di # imm = 0xFE00 ; AVX512F-NEXT: kmovw %edi, %k1 ; AVX512F-NEXT: vmovdqa32 %zmm15, %zmm14 {%k1} ; AVX512F-NEXT: vbroadcasti32x4 {{.*#+}} zmm8 = [26,5,12,19,26,5,12,19,26,5,12,19,26,5,12,19] ; AVX512F-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm7, %zmm8, %zmm9 ; AVX512F-NEXT: vmovdqa {{.*#+}} ymm7 = [18,25,0,7,14,u,u,u] ; AVX512F-NEXT: vpermi2d %zmm0, %zmm2, %zmm7 ; AVX512F-NEXT: vmovdqa32 %zmm9, %zmm7 {%k2} ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm9 = [0,1,8,15,22,29,0,0,0,1,8,15,22,29,0,0] ; AVX512F-NEXT: # zmm9 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm5, %zmm4, %zmm9 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm15 = [0,9,10,11,12,13,20,27,0,9,10,11,12,13,20,27] ; AVX512F-NEXT: # zmm15 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm1, %zmm9, %zmm15 ; AVX512F-NEXT: vmovdqa32 %zmm15, %zmm7 {%k1} ; AVX512F-NEXT: vmovdqa {{.*#+}} ymm9 = [19,26,1,8,15,u,u,u] ; AVX512F-NEXT: vpermi2d %zmm0, %zmm2, %zmm9 ; AVX512F-NEXT: vmovdqa32 %zmm10, %zmm9 {%k2} ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm10 = [0,18,25,0,7,14,0,0,0,18,25,0,7,14,0,0] ; AVX512F-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm4, %zmm5, %zmm10 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm15 = [0,9,10,11,12,13,21,28,0,9,10,11,12,13,21,28] ; AVX512F-NEXT: # zmm15 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm1, %zmm10, %zmm15 ; AVX512F-NEXT: vmovdqa32 %zmm15, %zmm9 {%k1} ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm10 = [0,19,26,1,8,15,0,0,0,19,26,1,8,15,0,0] ; AVX512F-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm4, %zmm5, %zmm10 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm15 = [0,9,10,11,12,13,22,29,0,9,10,11,12,13,22,29] ; AVX512F-NEXT: # zmm15 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm1, %zmm10, %zmm15 ; AVX512F-NEXT: vmovdqa {{.*#+}} xmm10 = [4,11,18,25] ; AVX512F-NEXT: vpermi2d %zmm2, %zmm0, %zmm10 ; AVX512F-NEXT: vinserti32x4 $0, %xmm10, %zmm11, %zmm10 ; AVX512F-NEXT: vmovdqa32 %zmm15, %zmm10 {%k1} ; AVX512F-NEXT: vpermi2d %zmm5, %zmm4, %zmm6 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm11 = [0,9,10,11,12,16,23,30,0,9,10,11,12,16,23,30] ; AVX512F-NEXT: # zmm11 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm1, %zmm6, %zmm11 ; AVX512F-NEXT: vmovdqa {{.*#+}} xmm6 = [5,12,19,26] ; AVX512F-NEXT: vpermi2d %zmm2, %zmm0, %zmm6 ; AVX512F-NEXT: vinserti32x4 $0, %xmm6, %zmm12, %zmm6 ; AVX512F-NEXT: vmovdqa32 %zmm11, %zmm6 {%k1} ; AVX512F-NEXT: vpermt2d %zmm5, %zmm8, %zmm4 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm5 = [0,9,10,11,12,17,24,31,0,9,10,11,12,17,24,31] ; AVX512F-NEXT: # zmm5 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermi2d %zmm1, %zmm4, %zmm5 ; AVX512F-NEXT: vmovdqa {{.*#+}} xmm1 = [6,13,20,27] ; AVX512F-NEXT: vpermi2d %zmm2, %zmm0, %zmm1 ; AVX512F-NEXT: vinserti32x4 $0, %xmm1, %zmm13, %zmm0 ; AVX512F-NEXT: vmovdqa32 %zmm5, %zmm0 {%k1} ; AVX512F-NEXT: vmovdqa64 %zmm3, (%rsi) ; AVX512F-NEXT: vmovdqa64 %zmm14, (%rdx) ; AVX512F-NEXT: vmovdqa64 %zmm7, (%rcx) ; AVX512F-NEXT: vmovdqa64 %zmm9, (%r8) ; AVX512F-NEXT: vmovdqa64 %zmm10, (%r9) ; AVX512F-NEXT: vmovdqa64 %zmm6, (%r10) ; AVX512F-NEXT: vmovdqa64 %zmm0, (%rax) ; AVX512F-NEXT: vzeroupper ; AVX512F-NEXT: retq ; ; AVX512BW-LABEL: load_i32_stride7_vf16: ; AVX512BW: # %bb.0: ; AVX512BW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX512BW-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX512BW-NEXT: vmovdqa64 384(%rdi), %zmm1 ; AVX512BW-NEXT: vmovdqa64 320(%rdi), %zmm5 ; AVX512BW-NEXT: vmovdqa64 256(%rdi), %zmm4 ; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0 ; AVX512BW-NEXT: vmovdqa64 64(%rdi), %zmm2 ; AVX512BW-NEXT: vmovdqa64 128(%rdi), %zmm8 ; AVX512BW-NEXT: vmovdqa64 192(%rdi), %zmm7 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm6 = [8,15,0,0,0,19,26,1,8,15,0,0,0,19,26,1] ; AVX512BW-NEXT: # zmm6 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm8, %zmm7, %zmm6 ; AVX512BW-NEXT: vmovdqa {{.*#+}} ymm3 = [0,7,14,21,28,u,u,u] ; AVX512BW-NEXT: vpermi2d %zmm2, %zmm0, %zmm3 ; AVX512BW-NEXT: movw $992, %di # imm = 0x3E0 ; AVX512BW-NEXT: kmovd %edi, %k1 ; AVX512BW-NEXT: vmovdqa32 %zmm6, %zmm3 {%k1} ; AVX512BW-NEXT: vbroadcasti32x4 {{.*#+}} zmm6 = [20,27,6,13,20,27,6,13,20,27,6,13,20,27,6,13] ; AVX512BW-NEXT: # zmm6 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm5, %zmm4, %zmm6 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm9 = [0,0,10,11,12,13,18,25,0,0,10,11,12,13,18,25] ; AVX512BW-NEXT: # zmm9 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm6, %zmm9 ; AVX512BW-NEXT: movb $-32, %dil ; AVX512BW-NEXT: kmovd %edi, %k1 ; AVX512BW-NEXT: vmovdqa64 %zmm9, %zmm3 {%k1} ; AVX512BW-NEXT: vbroadcasti32x4 {{.*#+}} zmm6 = [25,4,11,18,25,4,11,18,25,4,11,18,25,4,11,18] ; AVX512BW-NEXT: # zmm6 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm8, %zmm9 ; AVX512BW-NEXT: vbroadcasti32x4 {{.*#+}} zmm10 = [27,6,13,20,27,6,13,20,27,6,13,20,27,6,13,20] ; AVX512BW-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm7, %zmm8, %zmm10 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm11 = [28,0,0,0,0,7,14,21,28,0,0,0,0,7,14,21] ; AVX512BW-NEXT: # zmm11 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm7, %zmm8, %zmm11 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm12 = [29,0,0,0,1,8,15,22,29,0,0,0,1,8,15,22] ; AVX512BW-NEXT: # zmm12 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm7, %zmm8, %zmm12 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm13 = [14,0,0,0,18,25,0,7,14,0,0,0,18,25,0,7] ; AVX512BW-NEXT: # zmm13 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm8, %zmm7, %zmm13 ; AVX512BW-NEXT: vpermt2d %zmm7, %zmm6, %zmm8 ; AVX512BW-NEXT: vmovdqa {{.*#+}} ymm14 = [1,8,15,22,29,u,u,u] ; AVX512BW-NEXT: vpermi2d %zmm2, %zmm0, %zmm14 ; AVX512BW-NEXT: movw $480, %di # imm = 0x1E0 ; AVX512BW-NEXT: kmovd %edi, %k2 ; AVX512BW-NEXT: vmovdqa32 %zmm8, %zmm14 {%k2} ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,7,14,21,28,0,0,0,0,7,14,21,28,0,0] ; AVX512BW-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm5, %zmm4, %zmm8 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm15 = [0,9,10,11,12,13,19,26,0,9,10,11,12,13,19,26] ; AVX512BW-NEXT: # zmm15 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm8, %zmm15 ; AVX512BW-NEXT: movw $-512, %di # imm = 0xFE00 ; AVX512BW-NEXT: kmovd %edi, %k1 ; AVX512BW-NEXT: vmovdqa32 %zmm15, %zmm14 {%k1} ; AVX512BW-NEXT: vbroadcasti32x4 {{.*#+}} zmm8 = [26,5,12,19,26,5,12,19,26,5,12,19,26,5,12,19] ; AVX512BW-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm7, %zmm8, %zmm9 ; AVX512BW-NEXT: vmovdqa {{.*#+}} ymm7 = [18,25,0,7,14,u,u,u] ; AVX512BW-NEXT: vpermi2d %zmm0, %zmm2, %zmm7 ; AVX512BW-NEXT: vmovdqa32 %zmm9, %zmm7 {%k2} ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm9 = [0,1,8,15,22,29,0,0,0,1,8,15,22,29,0,0] ; AVX512BW-NEXT: # zmm9 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm5, %zmm4, %zmm9 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm15 = [0,9,10,11,12,13,20,27,0,9,10,11,12,13,20,27] ; AVX512BW-NEXT: # zmm15 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm9, %zmm15 ; AVX512BW-NEXT: vmovdqa32 %zmm15, %zmm7 {%k1} ; AVX512BW-NEXT: vmovdqa {{.*#+}} ymm9 = [19,26,1,8,15,u,u,u] ; AVX512BW-NEXT: vpermi2d %zmm0, %zmm2, %zmm9 ; AVX512BW-NEXT: vmovdqa32 %zmm10, %zmm9 {%k2} ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm10 = [0,18,25,0,7,14,0,0,0,18,25,0,7,14,0,0] ; AVX512BW-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm4, %zmm5, %zmm10 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm15 = [0,9,10,11,12,13,21,28,0,9,10,11,12,13,21,28] ; AVX512BW-NEXT: # zmm15 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm10, %zmm15 ; AVX512BW-NEXT: vmovdqa32 %zmm15, %zmm9 {%k1} ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm10 = [0,19,26,1,8,15,0,0,0,19,26,1,8,15,0,0] ; AVX512BW-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm4, %zmm5, %zmm10 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm15 = [0,9,10,11,12,13,22,29,0,9,10,11,12,13,22,29] ; AVX512BW-NEXT: # zmm15 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm10, %zmm15 ; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm10 = [4,11,18,25] ; AVX512BW-NEXT: vpermi2d %zmm2, %zmm0, %zmm10 ; AVX512BW-NEXT: vinserti32x4 $0, %xmm10, %zmm11, %zmm10 ; AVX512BW-NEXT: vmovdqa32 %zmm15, %zmm10 {%k1} ; AVX512BW-NEXT: vpermi2d %zmm5, %zmm4, %zmm6 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm11 = [0,9,10,11,12,16,23,30,0,9,10,11,12,16,23,30] ; AVX512BW-NEXT: # zmm11 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm6, %zmm11 ; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm6 = [5,12,19,26] ; AVX512BW-NEXT: vpermi2d %zmm2, %zmm0, %zmm6 ; AVX512BW-NEXT: vinserti32x4 $0, %xmm6, %zmm12, %zmm6 ; AVX512BW-NEXT: vmovdqa32 %zmm11, %zmm6 {%k1} ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm8, %zmm4 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm5 = [0,9,10,11,12,17,24,31,0,9,10,11,12,17,24,31] ; AVX512BW-NEXT: # zmm5 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm4, %zmm5 ; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm1 = [6,13,20,27] ; AVX512BW-NEXT: vpermi2d %zmm2, %zmm0, %zmm1 ; AVX512BW-NEXT: vinserti32x4 $0, %xmm1, %zmm13, %zmm0 ; AVX512BW-NEXT: vmovdqa32 %zmm5, %zmm0 {%k1} ; AVX512BW-NEXT: vmovdqa64 %zmm3, (%rsi) ; AVX512BW-NEXT: vmovdqa64 %zmm14, (%rdx) ; AVX512BW-NEXT: vmovdqa64 %zmm7, (%rcx) ; AVX512BW-NEXT: vmovdqa64 %zmm9, (%r8) ; AVX512BW-NEXT: vmovdqa64 %zmm10, (%r9) ; AVX512BW-NEXT: vmovdqa64 %zmm6, (%r10) ; AVX512BW-NEXT: vmovdqa64 %zmm0, (%rax) ; AVX512BW-NEXT: vzeroupper ; AVX512BW-NEXT: retq %wide.vec = load <112 x i32>, ptr %in.vec, align 64 %strided.vec0 = shufflevector <112 x i32> %wide.vec, <112 x i32> poison, <16 x i32> %strided.vec1 = shufflevector <112 x i32> %wide.vec, <112 x i32> poison, <16 x i32> %strided.vec2 = shufflevector <112 x i32> %wide.vec, <112 x i32> poison, <16 x i32> %strided.vec3 = shufflevector <112 x i32> %wide.vec, <112 x i32> poison, <16 x i32> %strided.vec4 = shufflevector <112 x i32> %wide.vec, <112 x i32> poison, <16 x i32> %strided.vec5 = shufflevector <112 x i32> %wide.vec, <112 x i32> poison, <16 x i32> %strided.vec6 = shufflevector <112 x i32> %wide.vec, <112 x i32> poison, <16 x i32> store <16 x i32> %strided.vec0, ptr %out.vec0, align 64 store <16 x i32> %strided.vec1, ptr %out.vec1, align 64 store <16 x i32> %strided.vec2, ptr %out.vec2, align 64 store <16 x i32> %strided.vec3, ptr %out.vec3, align 64 store <16 x i32> %strided.vec4, ptr %out.vec4, align 64 store <16 x i32> %strided.vec5, ptr %out.vec5, align 64 store <16 x i32> %strided.vec6, ptr %out.vec6, align 64 ret void } define void @load_i32_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind { ; SSE-LABEL: load_i32_stride7_vf32: ; SSE: # %bb.0: ; SSE-NEXT: subq $1160, %rsp # imm = 0x488 ; SSE-NEXT: movdqa 80(%rdi), %xmm8 ; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa (%rdi), %xmm13 ; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 16(%rdi), %xmm6 ; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 48(%rdi), %xmm5 ; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 640(%rdi), %xmm3 ; SSE-NEXT: movdqa 608(%rdi), %xmm4 ; SSE-NEXT: movdqa 560(%rdi), %xmm10 ; SSE-NEXT: movdqa 576(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 192(%rdi), %xmm7 ; SSE-NEXT: movdqa 160(%rdi), %xmm9 ; SSE-NEXT: movdqa 112(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 128(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3] ; SSE-NEXT: movdqa %xmm9, %xmm12 ; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1] ; SSE-NEXT: movdqa %xmm7, %xmm14 ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[3,3,3,3] ; SSE-NEXT: movdqa %xmm10, %xmm2 ; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3] ; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1] ; SSE-NEXT: movdqa %xmm3, %xmm7 ; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3] ; SSE-NEXT: movdqa %xmm13, %xmm2 ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 448(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 464(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1] ; SSE-NEXT: movdqa 528(%rdi), %xmm9 ; SSE-NEXT: movdqa 496(%rdi), %xmm13 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3] ; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 336(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 352(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1] ; SSE-NEXT: movdqa 416(%rdi), %xmm3 ; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 384(%rdi), %xmm11 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3] ; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 784(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 800(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1] ; SSE-NEXT: movdqa 864(%rdi), %xmm3 ; SSE-NEXT: movdqa 832(%rdi), %xmm15 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3] ; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1] ; SSE-NEXT: movdqa %xmm3, %xmm8 ; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 224(%rdi), %xmm3 ; SSE-NEXT: movdqa %xmm3, (%rsp) # 16-byte Spill ; SSE-NEXT: movdqa 240(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1] ; SSE-NEXT: movdqa 304(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 272(%rdi), %xmm6 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3] ; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 672(%rdi), %xmm3 ; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 688(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1] ; SSE-NEXT: movdqa 752(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 720(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[2,2,2,2] ; SSE-NEXT: movdqa %xmm12, %xmm3 ; SSE-NEXT: punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1] ; SSE-NEXT: movdqa 144(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1] ; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,2,2,2] ; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1] ; SSE-NEXT: movdqa 592(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1] ; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1] ; SSE-NEXT: movdqa 32(%rdi), %xmm4 ; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2] ; SSE-NEXT: punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm0[2],xmm13[3],xmm0[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: movdqa 480(%rdi), %xmm4 ; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1] ; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1] ; SSE-NEXT: movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,2,2] ; SSE-NEXT: movdqa %xmm11, %xmm4 ; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1] ; SSE-NEXT: movdqa 368(%rdi), %xmm11 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1] ; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1] ; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,2,2] ; SSE-NEXT: movdqa %xmm15, %xmm4 ; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1] ; SSE-NEXT: movdqa 816(%rdi), %xmm7 ; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1] ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1] ; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2] ; SSE-NEXT: punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm0[2],xmm6[3],xmm0[3] ; SSE-NEXT: movdqa (%rsp), %xmm9 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm9[1,1,1,1] ; SSE-NEXT: movdqa 256(%rdi), %xmm13 ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm13[0],xmm7[1],xmm13[1] ; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm7[0],xmm6[1] ; SSE-NEXT: movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm12[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm7[2],xmm6[3],xmm7[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[1,1,1,1] ; SSE-NEXT: movdqa 704(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm7[0],xmm6[1] ; SSE-NEXT: movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm2[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload ; SSE-NEXT: # xmm8 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1] ; SSE-NEXT: movdqa 176(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm14[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm8[0],xmm6[1],xmm8[1] ; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm7[0],xmm6[1] ; SSE-NEXT: movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm3[2,3,2,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm0[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1] ; SSE-NEXT: movdqa 64(%rdi), %xmm3 ; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm10[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm7[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm1[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm11[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1] ; SSE-NEXT: movdqa 400(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm5[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm7[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm9[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm13[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1] ; SSE-NEXT: movdqa 288(%rdi), %xmm2 ; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm4[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,0,1,1] ; SSE-NEXT: movdqa %xmm2, %xmm10 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm7[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload ; SSE-NEXT: # xmm8 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1] ; SSE-NEXT: movdqa 624(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload ; SSE-NEXT: # xmm8 = mem[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm7[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm9[2,3,2,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm14[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1] ; SSE-NEXT: movdqa 512(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload ; SSE-NEXT: # xmm8 = mem[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm7[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm15[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload ; SSE-NEXT: # xmm8 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1] ; SSE-NEXT: movdqa 848(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload ; SSE-NEXT: # xmm8 = mem[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm7[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm6[2,3,2,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm11[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1] ; SSE-NEXT: movdqa 736(%rdi), %xmm2 ; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm12[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,0,1,1] ; SSE-NEXT: movdqa %xmm2, %xmm12 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm7[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 96(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm1[0,0,1,1] ; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa %xmm3, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload ; SSE-NEXT: # xmm5 = mem[2,2,3,3] ; SSE-NEXT: punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 208(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; SSE-NEXT: movdqa %xmm4, %xmm2 ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm8[2],xmm1[3],xmm8[3] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 320(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa %xmm10, %xmm2 ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] ; SSE-NEXT: pshufd $250, (%rsp), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,2,3,3] ; SSE-NEXT: movdqa %xmm13, %xmm10 ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm13[2],xmm1[3],xmm13[3] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, (%rsp) # 16-byte Spill ; SSE-NEXT: movdqa 432(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload ; SSE-NEXT: movdqa %xmm13, %xmm2 ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm15[2],xmm1[3],xmm15[3] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 544(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; SSE-NEXT: movdqa %xmm5, %xmm2 ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[2,2,3,3] ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm14[2],xmm1[3],xmm14[3] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 656(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload ; SSE-NEXT: movdqa %xmm7, %xmm2 ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm14[2],xmm1[3],xmm14[3] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 768(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa %xmm12, %xmm2 ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[2,2,3,3] ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm11[2],xmm1[3],xmm11[3] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 880(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload ; SSE-NEXT: movdqa %xmm11, %xmm0 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm12[2],xmm1[3],xmm12[3] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload ; SSE-NEXT: movdqa %xmm6, %xmm2 ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload ; SSE-NEXT: movdqa %xmm8, %xmm2 ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload ; SSE-NEXT: movdqa %xmm9, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload ; SSE-NEXT: movdqa %xmm10, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload ; SSE-NEXT: movdqa %xmm13, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload ; SSE-NEXT: movdqa %xmm14, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload ; SSE-NEXT: movdqa %xmm15, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1] ; SSE-NEXT: movdqa %xmm13, %xmm5 ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, %xmm15 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,2,2,2] ; SSE-NEXT: movdqa %xmm11, %xmm1 ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload ; SSE-NEXT: # xmm8 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload ; SSE-NEXT: # xmm6 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm1[0],xmm6[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload ; SSE-NEXT: # xmm5 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload ; SSE-NEXT: # xmm4 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload ; SSE-NEXT: # xmm3 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload ; SSE-NEXT: # xmm2 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 96(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 32(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 112(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 48(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 64(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, (%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 80(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 16(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 96(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 32(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 112(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 48(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 64(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, (%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 80(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 16(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 96(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 112(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 64(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 80(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 32(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 48(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, (%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 16(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 112(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 96(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 80(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 64(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 48(%r8) ; SSE-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 32(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 16(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, (%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 112(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 96(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 80(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 64(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 48(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 32(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 16(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, (%r9) ; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 112(%rax) ; SSE-NEXT: movapd %xmm15, 96(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 80(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 64(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 48(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 32(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, 16(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: movaps %xmm1, (%rax) ; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; SSE-NEXT: movapd %xmm0, 112(%rax) ; SSE-NEXT: movapd %xmm3, 96(%rax) ; SSE-NEXT: movapd %xmm4, 80(%rax) ; SSE-NEXT: movapd %xmm5, 64(%rax) ; SSE-NEXT: movapd %xmm6, 48(%rax) ; SSE-NEXT: movapd %xmm7, 32(%rax) ; SSE-NEXT: movapd %xmm8, 16(%rax) ; SSE-NEXT: movapd %xmm9, (%rax) ; SSE-NEXT: addq $1160, %rsp # imm = 0x488 ; SSE-NEXT: retq ; ; AVX1-ONLY-LABEL: load_i32_stride7_vf32: ; AVX1-ONLY: # %bb.0: ; AVX1-ONLY-NEXT: subq $1464, %rsp # imm = 0x5B8 ; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm4 ; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm3 ; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %ymm7 ; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm2 ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm6 ; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[2],ymm0[2] ; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1 ; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm2 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3] ; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm9 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1] ; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm10 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 528(%rdi), %xmm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm0[0],ymm7[2],ymm0[2] ; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm4[6],ymm3[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1 ; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm8 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1] ; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm4 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm5 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm5[0,1],xmm0[2,3] ; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm13 ; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm13[0],ymm1[0],ymm13[2],ymm1[2] ; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm11 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm11[1] ; AVX1-ONLY-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm3 ; AVX1-ONLY-NEXT: vmovaps 752(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm1[0],ymm3[2],ymm1[2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[1,1],ymm0[2,2],ymm6[5,5],ymm0[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm9[1],xmm1[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm2 ; AVX1-ONLY-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm10[2] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm7[1,1],ymm0[2,2],ymm7[5,5],ymm0[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm10 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm10[0],xmm8[1],xmm10[2,3] ; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps %xmm8, %xmm7 ; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm14 ; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm12 ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm12[2,3],ymm14[0,1] ; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,0],ymm2[3,3],ymm12[4,4],ymm2[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm2 = zero,xmm2[1,2],xmm4[2] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm13[1,1],ymm0[2,2],ymm13[5,5],ymm0[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm0 ; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm0[0],xmm5[1],xmm0[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm2 ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm2[2,3],ymm0[0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm2[0,0],ymm5[3,3],ymm2[4,4],ymm5[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm5 = zero,xmm5[1,2],xmm11[2] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm5[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm8 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[1,1],ymm8[2,2],ymm3[5,5],ymm8[6,6] ; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps %ymm3, %ymm9 ; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm6 ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm6[0],xmm2[1],xmm6[2,3] ; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm11[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2],ymm1[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm5 ; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %ymm15 ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm15[2,3],ymm5[0,1] ; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm15[0,0],ymm13[3,3],ymm15[4,4],ymm13[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13 ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm13 = zero,xmm13[1,2],xmm4[2] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm13[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm0 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm13 = ymm13[3,1],mem[0,3],ymm13[7,5],mem[4,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm11[2,1],ymm13[2,0],ymm11[6,5],ymm13[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm13[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload ; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm13 = ymm3[0],mem[0],ymm3[2],mem[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13 ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm13 = xmm13[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm13[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm7[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm10[1],xmm0[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm13[3,1],ymm10[0,3],ymm13[7,5],ymm10[4,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm7[2,1],ymm13[2,0],ymm7[6,5],ymm13[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm13[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm12[0],ymm14[0],ymm12[2],ymm14[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13 ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm13 = xmm13[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm13[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm2[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm6[1],xmm0[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm13[3,1],ymm8[0,3],ymm13[7,5],ymm8[4,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm9[2,1],ymm13[2,0],ymm9[6,5],ymm13[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm13[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm15[0],ymm5[0],ymm15[2],ymm5[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm13[0,1,2],xmm4[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm5[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[3,1],ymm14[0,3],ymm8[7,5],ymm14[4,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm4[2,1],ymm8[2,0],ymm4[6,5],ymm8[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm8[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm6[0],ymm13[0],ymm6[2],ymm13[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8 ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm8 = xmm8[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm11[0,0],ymm0[5,4],ymm11[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm11[3,1],ymm0[0,2],ymm11[7,5],ymm0[4,6] ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm8 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm8 = xmm1[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm8[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm1[0,1],ymm9[1,3],ymm1[4,5],ymm9[5,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm3[0,2],ymm8[2,0],ymm3[4,6],ymm8[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm10[1,0],ymm7[0,0],ymm10[5,4],ymm7[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm7[3,1],ymm0[0,2],ymm7[7,5],ymm0[4,6] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm8 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm8 = xmm11[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm8[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm1[0,1],ymm8[1,3],ymm1[4,5],ymm8[5,7] ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm12[0,2],ymm10[2,0],ymm12[4,6],ymm10[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm10[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm14[1,0],ymm4[0,0],ymm14[5,4],ymm4[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm4[3,1],ymm0[0,2],ymm4[7,5],ymm0[4,6] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm2[0,1,2],xmm5[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm7[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm0[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm7 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm7[0,1],ymm13[1,3],ymm7[4,5],ymm13[5,7] ; AVX1-ONLY-NEXT: vmovaps %ymm7, %ymm14 ; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm6[0,2],ymm10[2,0],ymm6[4,6],ymm10[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm10[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm2[0,0],ymm0[5,4],ymm2[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[3,1],ymm0[0,2],ymm2[7,5],ymm0[4,6] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm5 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2],xmm2[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm0[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm5 ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm5[0,1],ymm7[1,3],ymm5[4,5],ymm7[5,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm15[0,2],ymm6[2,0],ymm15[4,6],ymm6[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm6[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm12[3,0],ymm0[0,0],ymm12[7,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[1,0],ymm8[2,0],ymm1[5,4],ymm8[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm3[2,0],ymm0[6,4],ymm3[6,4] ; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm1[0,1,0,1] ; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm6 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm6[0,1,2],xmm10[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm12 = xmm11[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = mem[0],xmm12[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm12[0,1],xmm10[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm2[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[3,0],ymm0[0,0],ymm2[7,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm3[1,0],ymm9[2,0],ymm3[5,4],ymm9[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm0[2,0],ymm4[2,0],ymm0[6,4],ymm4[6,4] ; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm0 ; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm4 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm12 = xmm4[0,1,0,1] ; AVX1-ONLY-NEXT: vmovaps %xmm4, (%rsp) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm0[0,1,2],xmm12[3] ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm13 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = mem[0],xmm13[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm13[0,1],xmm12[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1,2,3],ymm10[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm1[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[3,0],ymm10[0,0],ymm1[7,4],ymm10[4,4] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm14[1,0],ymm11[2,0],ymm14[5,4],ymm11[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm2[2,0],ymm10[2,0],ymm2[6,4],ymm10[6,4] ; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm1[0,1,0,1] ; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm2[0,1,2],xmm13[3] ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm14 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = mem[0],xmm14[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm14[0,1],xmm13[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3],ymm12[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm15[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm15[3,0],ymm12[0,0],ymm15[7,4],ymm12[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm5[1,0],ymm7[2,0],ymm5[5,4],ymm7[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm1[2,0],ymm12[2,0],ymm1[6,4],ymm12[6,4] ; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm1[0,1,0,1] ; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm13 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm13[0,1,2],xmm14[3] ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm15 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm15 = mem[0],xmm15[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3],ymm12[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm3[2,1],ymm9[3,3],ymm3[6,5],ymm9[7,7] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm15 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm15 = xmm1[0],mem[1],xmm1[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm15[1,0],ymm12[2,0],ymm15[5,4],ymm12[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm4[0,1,2],xmm0[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm15 = ymm1[0,0],mem[1,0],ymm1[4,4],mem[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm15[0,1],xmm0[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm12[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm3[2,1],ymm8[3,3],ymm3[6,5],ymm8[7,7] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm15 = xmm8[0],xmm7[1],xmm8[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm15[1,0],ymm0[2,0],ymm15[5,4],ymm0[6,4] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm10[0,1,2],xmm6[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm9[0,0],ymm12[1,0],ymm9[4,4],ymm12[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm15[0,1],xmm6[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm11[3,3],ymm0[6,5],ymm11[7,7] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm15 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm15 = mem[0],xmm1[1],mem[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm15[1,0],ymm0[2,0],ymm15[5,4],ymm0[6,4] ; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2],xmm2[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm15 = ymm1[0,0],mem[1,0],ymm1[4,4],mem[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm15[0,1],xmm2[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm2[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm0 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm0 = ymm5[2,1],mem[3,3],ymm5[6,5],mem[7,7] ; AVX1-ONLY-NEXT: vmovaps %ymm5, %ymm14 ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0],xmm1[1],xmm4[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[1,0],ymm0[2,0],ymm2[5,4],ymm0[6,4] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm6[0,1,2],xmm13[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm5[0,0],ymm13[1,0],ymm5[4,4],ymm13[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm11[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm11 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm11[0,1],xmm2[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm3[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm3[3,0],ymm0[0,0],ymm3[7,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm7[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm8[1],xmm2[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,0],ymm2[4,5],ymm0[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm10[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm9[1,0],ymm12[2,0],ymm9[5,4],ymm12[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm14[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm14[3,0],ymm0[0,0],ymm14[7,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm1[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm4[1],xmm2[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,0],ymm2[4,5],ymm0[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1,2],xmm6[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm5[1,0],ymm13[2,0],ymm5[5,4],ymm13[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[3,0],ymm1[0,0],ymm2[7,4],ymm1[4,4] ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm2 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm2 = xmm2[0],mem[1],xmm2[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,0],ymm2[4,5],ymm1[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps $8, (%rsp), %xmm2, %xmm2 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm4 = ymm4[1,0],mem[2,0],ymm4[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm4[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[3,0],ymm2[0,0],ymm4[7,4],ymm2[4,4] ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,0],ymm4[4,5],ymm2[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm4 = xmm4[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm5 = ymm5[1,0],mem[2,0],ymm5[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%r8) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%r8) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%r8) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, (%r8) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%r9) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%r9) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, (%r9) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%r9) ; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX1-ONLY-NEXT: vmovaps %ymm11, 96(%rax) ; AVX1-ONLY-NEXT: vmovaps %ymm15, 32(%rax) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rax) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rax) ; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rax) ; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rax) ; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rax) ; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rax) ; AVX1-ONLY-NEXT: addq $1464, %rsp # imm = 0x5B8 ; AVX1-ONLY-NEXT: vzeroupper ; AVX1-ONLY-NEXT: retq ; ; AVX2-SLOW-LABEL: load_i32_stride7_vf32: ; AVX2-SLOW: # %bb.0: ; AVX2-SLOW-NEXT: subq $1224, %rsp # imm = 0x4C8 ; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %ymm9 ; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %ymm4 ; AVX2-SLOW-NEXT: vmovdqa 224(%rdi), %ymm5 ; AVX2-SLOW-NEXT: vmovdqa 544(%rdi), %ymm10 ; AVX2-SLOW-NEXT: vmovdqa 480(%rdi), %ymm7 ; AVX2-SLOW-NEXT: vmovdqa 448(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm3 ; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm6 ; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpbroadcastq 80(%rdi), %ymm0 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = [0,7,6,u] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm6[6],ymm3[7] ; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm12 ; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm14 ; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpermd %ymm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpbroadcastd 196(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm7[6],ymm8[7] ; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-SLOW-NEXT: vpbroadcastq 528(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 576(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 608(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpbroadcastd 644(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7] ; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-SLOW-NEXT: vpbroadcastq 304(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm9[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa %ymm9, %ymm13 ; AVX2-SLOW-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpbroadcastd 420(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 704(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqa 672(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7] ; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm6 ; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm3 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 768(%rdi), %ymm11 ; AVX2-SLOW-NEXT: vpbroadcastq 752(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm11[4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 800(%rdi), %xmm1 ; AVX2-SLOW-NEXT: vmovdqa 832(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-SLOW-NEXT: vpbroadcastd 868(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 608(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 576(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-SLOW-NEXT: vmovdqa 512(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1],ymm2[2,3],ymm10[4,5],ymm2[6,7] ; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm9 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0],ymm8[1],ymm7[2,3,4],ymm8[5],ymm7[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1],ymm1[2,3],ymm13[4,5],ymm1[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0],ymm5[1],ymm4[2,3,4],ymm5[5],ymm4[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 832(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqa 800(%rdi), %ymm4 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm13 ; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm15 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-SLOW-NEXT: vmovdqa 736(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm2[2,3],ymm11[4,5],ymm2[6,7] ; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm10 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, (%rsp) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0],ymm6[1],ymm3[2,3,4],ymm6[5],ymm3[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm6 ; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm5 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1],ymm1[2,3],ymm4[4,5],ymm1[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm12[0],ymm14[1],ymm12[2,3,4],ymm14[5],ymm12[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 80(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm2 = ymm4[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3],ymm2[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 8(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm2 = ymm6[0],ymm5[0],ymm6[2],ymm5[2] ; AVX2-SLOW-NEXT: vpbroadcastd 204(%rdi), %ymm4 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm4[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 528(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm2 = ymm8[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3],ymm2[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 456(%rdi), %xmm4 ; AVX2-SLOW-NEXT: vmovdqa 480(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0],xmm2[1],xmm4[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm4 = ymm7[0],ymm9[0],ymm7[2],ymm9[2] ; AVX2-SLOW-NEXT: vpbroadcastd 652(%rdi), %ymm12 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm12[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 752(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm4 = ymm11[8,9,10,11,12,13,14,15],ymm10[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm10[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3],ymm4[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 680(%rdi), %xmm12 ; AVX2-SLOW-NEXT: vmovdqa 704(%rdi), %xmm4 ; AVX2-SLOW-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm12 = xmm12[0],xmm4[1],xmm12[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm12 = ymm13[0],ymm15[0],ymm13[2],ymm15[2] ; AVX2-SLOW-NEXT: vmovdqa %ymm13, %ymm10 ; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpbroadcastd 876(%rdi), %ymm15 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5,6],ymm15[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm12[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 304(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm12 = ymm5[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2],ymm0[3],ymm12[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 232(%rdi), %xmm15 ; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm0[1],xmm15[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm13[0],ymm4[0],ymm13[2],ymm4[2] ; AVX2-SLOW-NEXT: vpbroadcastd 428(%rdi), %ymm14 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5,6],ymm14[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm14[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd $253, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm1 = mem[0],ymm1[1],mem[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,2,2,3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm14[0,2],ymm15[1,3],ymm14[4,6],ymm15[5,7] ; AVX2-SLOW-NEXT: vbroadcastss 208(%rdi), %ymm11 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm11[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm1 = ymm8[0],mem[1],ymm8[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa %ymm9, %ymm12 ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm7[0,2],ymm9[1,3],ymm7[4,6],ymm9[5,7] ; AVX2-SLOW-NEXT: vbroadcastss 656(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0],ymm6[1],ymm5[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,2],ymm4[1,3],ymm13[4,6],ymm4[5,7] ; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm8 ; AVX2-SLOW-NEXT: vbroadcastss 432(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $2, (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,2],ymm9[1,3],ymm10[4,6],ymm9[5,7] ; AVX2-SLOW-NEXT: vbroadcastss 880(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm1 = [4,3,4,3] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,2,3],ymm0[4,5,6,7] ; AVX2-SLOW-NEXT: vpermd %ymm0, %ymm1, %ymm0 ; AVX2-SLOW-NEXT: vpbroadcastd 548(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 512(%rdi), %xmm6 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm6[0,1,2],xmm2[3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3] ; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} ymm11 = [0,7,0,7,0,7,0,7] ; AVX2-SLOW-NEXT: vpermd %ymm7, %ymm11, %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm12[6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 660(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill ; AVX2-SLOW-NEXT: vpbroadcastd 100(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm0[0,1,2],xmm2[3] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vpermd %ymm3, %ymm1, %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3] ; AVX2-SLOW-NEXT: vmovaps %ymm14, %ymm10 ; AVX2-SLOW-NEXT: vpermd %ymm14, %ymm11, %ymm3 ; AVX2-SLOW-NEXT: vmovaps %ymm15, %ymm7 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm15[6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 212(%rdi), %ymm4 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7] ; AVX2-SLOW-NEXT: vpermd %ymm2, %ymm1, %ymm3 ; AVX2-SLOW-NEXT: vpbroadcastd 324(%rdi), %xmm4 ; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm2[0,1,2],xmm4[3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3] ; AVX2-SLOW-NEXT: vpermd %ymm13, %ymm11, %ymm4 ; AVX2-SLOW-NEXT: vmovdqa %ymm13, %ymm14 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm8[6,7] ; AVX2-SLOW-NEXT: vmovdqa %ymm8, %ymm13 ; AVX2-SLOW-NEXT: vpbroadcastd 436(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm8[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vpermd %ymm3, %ymm1, %ymm1 ; AVX2-SLOW-NEXT: vpbroadcastd 772(%rdi), %xmm4 ; AVX2-SLOW-NEXT: vmovdqa 736(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm3[0,1,2],xmm4[3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload ; AVX2-SLOW-NEXT: vpermd %ymm15, %ymm11, %ymm4 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm9[6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 884(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm8[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0],ymm10[1],ymm7[2,3,4],ymm10[5],ymm7[6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3] ; AVX2-SLOW-NEXT: vbroadcastss 216(%rdi), %ymm4 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm4[7] ; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %xmm10 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1,2],xmm0[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2] ; AVX2-SLOW-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm4 = mem[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm4 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm4[0,1],xmm0[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 544(%rdi), %xmm4 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm4[0,1,2],xmm6[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,3,2] ; AVX2-SLOW-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3] ; AVX2-SLOW-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm5 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm5[0,1,0,3] ; AVX2-SLOW-NEXT: vpbroadcastd 664(%rdi), %ymm7 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm7[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %xmm12 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm12[0,1,2],xmm2[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,3,2] ; AVX2-SLOW-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm5[0,1,0,3] ; AVX2-SLOW-NEXT: vpbroadcastd 440(%rdi), %ymm7 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm7[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm5[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 768(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm2[0,1,2],xmm3[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,3,2] ; AVX2-SLOW-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm7 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm7[0,1],xmm3[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm9[0],ymm15[1],ymm9[2,3,4],ymm15[5],ymm9[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-SLOW-NEXT: vpbroadcastd 888(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm3[0,1,2,3],ymm7[4,5,6,7] ; AVX2-SLOW-NEXT: vbroadcastss 584(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3] ; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-SLOW-NEXT: vpermd 640(%rdi), %ymm11, %ymm8 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm8[6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 528(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm8[0,1,2],xmm4[3] ; AVX2-SLOW-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm8 = mem[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm8 = ymm8[0],mem[1],ymm8[2,3,4],mem[5],ymm8[6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm8, %xmm8 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm8[0,1],xmm4[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vbroadcastss 808(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3] ; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-SLOW-NEXT: vpermd 864(%rdi), %ymm11, %ymm4 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 752(%rdi), %ymm4 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3] ; AVX2-SLOW-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm4 = mem[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm4 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vbroadcastss 136(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3] ; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-SLOW-NEXT: vpermd 192(%rdi), %ymm11, %ymm4 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 80(%rdi), %ymm4 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1,2],xmm10[3] ; AVX2-SLOW-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm6 = mem[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm6 = ymm6[0],mem[1],ymm6[2,3,4],mem[5],ymm6[6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm6, %xmm6 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vbroadcastss 360(%rdi), %xmm4 ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3] ; AVX2-SLOW-NEXT: vpermd 416(%rdi), %ymm11, %ymm6 ; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 304(%rdi), %ymm6 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm6[0,1,2],xmm12[3] ; AVX2-SLOW-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7] ; AVX2-SLOW-NEXT: vextracti128 $1, %ymm9, %xmm9 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm6 = xmm9[0,1],xmm6[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 96(%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 64(%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, (%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 96(%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 64(%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, (%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 96(%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 64(%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, (%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 96(%r8) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%r8) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 64(%r8) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, (%r8) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 96(%r9) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%r9) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, (%r9) ; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 64(%r9) ; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-SLOW-NEXT: vmovdqa %ymm7, 96(%rax) ; AVX2-SLOW-NEXT: vmovdqa %ymm5, 32(%rax) ; AVX2-SLOW-NEXT: vmovdqa %ymm1, 64(%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm0, (%rax) ; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm2, 96(%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm8, 64(%rax) ; AVX2-SLOW-NEXT: addq $1224, %rsp # imm = 0x4C8 ; AVX2-SLOW-NEXT: vzeroupper ; AVX2-SLOW-NEXT: retq ; ; AVX2-FAST-LABEL: load_i32_stride7_vf32: ; AVX2-FAST: # %bb.0: ; AVX2-FAST-NEXT: subq $1192, %rsp # imm = 0x4A8 ; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm10 ; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm5 ; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 544(%rdi), %ymm11 ; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 480(%rdi), %ymm7 ; AVX2-FAST-NEXT: vmovdqa 448(%rdi), %ymm13 ; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm9 ; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpbroadcastq 80(%rdi), %ymm0 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm12 ; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = [0,7,6,u] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm9[6],ymm3[7] ; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm6 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %xmm3 ; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpbroadcastd 196(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],ymm7[6],ymm13[7] ; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa %ymm7, %ymm8 ; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vpbroadcastq 528(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 576(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa 608(%rdi), %xmm3 ; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpbroadcastd 644(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7] ; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm11 ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vpbroadcastq 304(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %xmm3 ; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpbroadcastd 420(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 704(%rdi), %ymm14 ; AVX2-FAST-NEXT: vmovdqa 672(%rdi), %ymm1 ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm14[6],ymm1[7] ; AVX2-FAST-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm0 ; AVX2-FAST-NEXT: vmovdqa 768(%rdi), %ymm2 ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpbroadcastq 752(%rdi), %ymm1 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 800(%rdi), %xmm1 ; AVX2-FAST-NEXT: vmovdqa 832(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vpbroadcastd 868(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm2 ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm1 ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm0 ; AVX2-FAST-NEXT: vmovdqa %ymm12, %ymm7 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1],ymm0[2,3],ymm12[4,5],ymm0[6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm12 ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm0 = [5,6,5,6,5,6,5,6] ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm3 ; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [1,0,7,7,5,4,7,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0],ymm6[1],ymm9[2,3,4],ymm6[5],ymm9[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vmovdqa 608(%rdi), %ymm15 ; AVX2-FAST-NEXT: vmovdqa 576(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm15[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm15[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %ymm10 ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1],ymm10[2,3],ymm5[4,5],ymm10[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0],ymm13[1],ymm8[2,3,4],ymm13[5],ymm8[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm4[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, (%rsp) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm3[2,3],ymm9[4,5],ymm3[6,7] ; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm3 ; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm4 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vmovdqa 832(%rdi), %ymm6 ; AVX2-FAST-NEXT: vmovdqa 800(%rdi), %ymm11 ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm6[12,13,14,15],ymm11[0,1,2,3,4,5,6,7,8,9,10,11],ymm6[28,29,30,31],ymm11[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vmovdqa %ymm11, %ymm13 ; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vmovdqa 736(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1],ymm4[2,3],ymm8[4,5],ymm4[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm0 ; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm3 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm1 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 80(%rdi), %xmm0 ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm7[8,9,10,11,12,13,14,15],ymm12[0,1,2,3,4,5,6,7],ymm7[24,25,26,27,28,29,30,31],ymm12[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 8(%rdi), %xmm1 ; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %xmm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm14[0],ymm1[2],ymm14[2] ; AVX2-FAST-NEXT: vpbroadcastd 204(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 528(%rdi), %xmm0 ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm5[8,9,10,11,12,13,14,15],ymm10[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm10[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 456(%rdi), %xmm3 ; AVX2-FAST-NEXT: vmovdqa 480(%rdi), %xmm1 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0],xmm1[1],xmm3[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm3 = ymm12[0],ymm15[0],ymm12[2],ymm15[2] ; AVX2-FAST-NEXT: vpbroadcastd 652(%rdi), %ymm11 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm11[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 752(%rdi), %xmm0 ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm8[8,9,10,11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3],ymm3[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 680(%rdi), %xmm11 ; AVX2-FAST-NEXT: vmovdqa 704(%rdi), %xmm10 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm11[0],xmm10[1],xmm11[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm11 = ymm13[0],ymm6[0],ymm13[2],ymm6[2] ; AVX2-FAST-NEXT: vpbroadcastd 876(%rdi), %ymm15 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,6],ymm15[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm11[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 304(%rdi), %xmm0 ; AVX2-FAST-NEXT: vmovdqu (%rsp), %ymm4 # 32-byte Reload ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm11 = ymm9[8,9,10,11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2],ymm0[3],ymm11[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 232(%rdi), %xmm15 ; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %xmm0 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm0[1],xmm15[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm15[0,1],ymm11[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm3[0],ymm8[0],ymm3[2],ymm8[2] ; AVX2-FAST-NEXT: vpbroadcastd 428(%rdi), %ymm13 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5,6],ymm13[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm11[0,1,2,3,4],ymm13[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],mem[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm7 = ymm7[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm7[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm14, %ymm5 ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm7 = ymm14[0,2],ymm5[1,3],ymm14[4,6],ymm5[5,7] ; AVX2-FAST-NEXT: vbroadcastss 208(%rdi), %ymm9 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm9[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm7[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],mem[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm12, %ymm9 ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,2],ymm13[1,3],ymm12[4,6],ymm13[5,7] ; AVX2-FAST-NEXT: vbroadcastss 656(%rdi), %ymm6 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm6[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd $253, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = mem[0],ymm4[1],mem[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],mem[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm7 ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm8[1,3],ymm3[4,6],ymm8[5,7] ; AVX2-FAST-NEXT: vbroadcastss 432(%rdi), %ymm2 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm10[0,1,2],mem[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,2],ymm12[1,3],ymm10[4,6],ymm12[5,7] ; AVX2-FAST-NEXT: vbroadcastss 880(%rdi), %ymm2 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm0 = [4,3,4,3] ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vpbroadcastd 548(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %xmm15 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm15[0,1,2],xmm2[3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3] ; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm11 = [0,7,0,7,0,7,0,7] ; AVX2-FAST-NEXT: vpermd %ymm9, %ymm11, %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm13[6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm13, %ymm9 ; AVX2-FAST-NEXT: vpbroadcastd 660(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpbroadcastd 100(%rdi), %xmm1 ; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %xmm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3] ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-NEXT: vpermd %ymm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3] ; AVX2-FAST-NEXT: vpermd %ymm14, %ymm11, %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm5[6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm4 ; AVX2-FAST-NEXT: vpbroadcastd 212(%rdi), %ymm5 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm5[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vpbroadcastd 324(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %xmm6 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm6[0,1,2],xmm2[3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3] ; AVX2-FAST-NEXT: vpermd %ymm7, %ymm11, %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm8[6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm8, %ymm13 ; AVX2-FAST-NEXT: vpbroadcastd 436(%rdi), %ymm5 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm5[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7] ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm0 ; AVX2-FAST-NEXT: vpbroadcastd 772(%rdi), %xmm1 ; AVX2-FAST-NEXT: vmovdqa 736(%rdi), %xmm8 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3] ; AVX2-FAST-NEXT: vpermd %ymm10, %ymm11, %ymm1 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm12[6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm12, %ymm2 ; AVX2-FAST-NEXT: vpbroadcastd 884(%rdi), %ymm5 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm5[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0],ymm14[1],ymm4[2,3,4],ymm14[5],ymm4[6,7] ; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm12 = [1,0,3,3,1,0,7,7] ; AVX2-FAST-NEXT: vpermd %ymm0, %ymm12, %ymm0 ; AVX2-FAST-NEXT: vpbroadcastd 216(%rdi), %ymm5 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm5[7] ; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %xmm7 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm3 = xmm7[0,1,2],xmm3[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,3,2] ; AVX2-FAST-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm5, %xmm5 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm0[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 544(%rdi), %xmm0 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm0[0,1,2],xmm15[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,3,2] ; AVX2-FAST-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm5, %xmm5 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3] ; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm5 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm5 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7] ; AVX2-FAST-NEXT: vpermd %ymm5, %ymm12, %ymm5 ; AVX2-FAST-NEXT: vpbroadcastd 664(%rdi), %ymm9 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm9[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm4[0,1,2,3],ymm5[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %xmm9 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm9[0,1,2],xmm6[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,3,2] ; AVX2-FAST-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm6 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm6 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3] ; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm6 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm6 = ymm13[0],mem[1],ymm13[2,3,4],mem[5],ymm13[6,7] ; AVX2-FAST-NEXT: vpermd %ymm6, %ymm12, %ymm6 ; AVX2-FAST-NEXT: vpbroadcastd 440(%rdi), %ymm13 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm13[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 768(%rdi), %xmm14 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm14[0,1,2],xmm8[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,1,3,2] ; AVX2-FAST-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm8 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm8 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1],xmm6[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm2[0],ymm10[1],ymm2[2,3,4],ymm10[5],ymm2[6,7] ; AVX2-FAST-NEXT: vpermd %ymm8, %ymm12, %ymm8 ; AVX2-FAST-NEXT: vpbroadcastd 888(%rdi), %ymm10 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm10[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 584(%rdi), %xmm8 ; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm8 = xmm8[0],mem[1],xmm8[2,3] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8 ; AVX2-FAST-NEXT: vpermd 640(%rdi), %ymm11, %ymm10 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm10[6,7] ; AVX2-FAST-NEXT: vpbroadcastd 528(%rdi), %ymm10 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1,2],xmm0[3] ; AVX2-FAST-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm10 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm10 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm10, %xmm10 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1],xmm0[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm8[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 808(%rdi), %xmm0 ; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm0 = xmm0[0],mem[1],xmm0[2,3] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-NEXT: vpermd 864(%rdi), %ymm11, %ymm10 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm10[6,7] ; AVX2-FAST-NEXT: vpbroadcastd 752(%rdi), %ymm10 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm10 = xmm10[0,1,2],xmm14[3] ; AVX2-FAST-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm12, %xmm12 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm10 = xmm12[0,1],xmm10[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm0[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 136(%rdi), %xmm0 ; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm0 = xmm0[0],mem[1],xmm0[2,3] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-NEXT: vpermd 192(%rdi), %ymm11, %ymm12 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7] ; AVX2-FAST-NEXT: vpbroadcastd 80(%rdi), %ymm12 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm12[0,1,2],xmm7[3] ; AVX2-FAST-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm12, %xmm12 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm12[0,1],xmm7[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 360(%rdi), %xmm7 ; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm7 = xmm7[0],mem[1],xmm7[2,3] ; AVX2-FAST-NEXT: vpermd 416(%rdi), %ymm11, %ymm11 ; AVX2-FAST-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm11[6,7] ; AVX2-FAST-NEXT: vpbroadcastd 304(%rdi), %ymm11 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm9 = xmm11[0,1,2],xmm9[3] ; AVX2-FAST-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7] ; AVX2-FAST-NEXT: vextracti128 $1, %ymm11, %xmm11 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm9 = xmm11[0,1],xmm9[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm9, 96(%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm9, 32(%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm9, 64(%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm9, (%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm2, 96(%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm2, 32(%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm2, 64(%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm9, (%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm2, 32(%rcx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm2, 96(%rcx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm2, 64(%rcx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm2, (%rcx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm2, 96(%r8) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm2, 32(%r8) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm2, 64(%r8) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm2, (%r8) ; AVX2-FAST-NEXT: vmovdqa %ymm1, 96(%r9) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%r9) ; AVX2-FAST-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm1, (%r9) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm1, 64(%r9) ; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-NEXT: vmovdqa %ymm6, 96(%rax) ; AVX2-FAST-NEXT: vmovdqa %ymm4, 32(%rax) ; AVX2-FAST-NEXT: vmovdqa %ymm5, 64(%rax) ; AVX2-FAST-NEXT: vmovdqa %ymm3, (%rax) ; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-NEXT: vmovdqa %ymm7, 32(%rax) ; AVX2-FAST-NEXT: vmovdqa %ymm0, (%rax) ; AVX2-FAST-NEXT: vmovdqa %ymm10, 96(%rax) ; AVX2-FAST-NEXT: vmovdqa %ymm8, 64(%rax) ; AVX2-FAST-NEXT: addq $1192, %rsp # imm = 0x4A8 ; AVX2-FAST-NEXT: vzeroupper ; AVX2-FAST-NEXT: retq ; ; AVX2-FAST-PERLANE-LABEL: load_i32_stride7_vf32: ; AVX2-FAST-PERLANE: # %bb.0: ; AVX2-FAST-PERLANE-NEXT: subq $1224, %rsp # imm = 0x4C8 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm9 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm4 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm5 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 544(%rdi), %ymm10 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 480(%rdi), %ymm7 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 448(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm6 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 80(%rdi), %ymm0 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm0 = [0,7,6,u] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm6[6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, %ymm12 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm14 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 196(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm7[6],ymm8[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 528(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 576(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 608(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 644(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 304(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm9[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm13 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 420(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 704(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 672(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm6 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 768(%rdi), %ymm11 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 752(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm11[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 800(%rdi), %xmm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 832(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 868(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 608(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 576(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 512(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1],ymm2[2,3],ymm10[4,5],ymm2[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm9 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0],ymm8[1],ymm7[2,3,4],ymm8[5],ymm7[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1],ymm1[2,3],ymm13[4,5],ymm1[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0],ymm5[1],ymm4[2,3,4],ymm5[5],ymm4[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 832(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 800(%rdi), %ymm4 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm13 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm15 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 736(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm2[2,3],ymm11[4,5],ymm2[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm10 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, (%rsp) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0],ymm6[1],ymm3[2,3,4],ymm6[5],ymm3[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm6 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm5 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1],ymm1[2,3],ymm4[4,5],ymm1[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm12[0],ymm14[1],ymm12[2,3,4],ymm14[5],ymm12[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 80(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm2 = ymm4[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3],ymm2[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 8(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm2 = ymm6[0],ymm5[0],ymm6[2],ymm5[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 204(%rdi), %ymm4 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm4[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 528(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm2 = ymm8[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3],ymm2[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 456(%rdi), %xmm4 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 480(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0],xmm2[1],xmm4[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm4 = ymm7[0],ymm9[0],ymm7[2],ymm9[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 652(%rdi), %ymm12 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm12[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 752(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm4 = ymm11[8,9,10,11,12,13,14,15],ymm10[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm10[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3],ymm4[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 680(%rdi), %xmm12 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 704(%rdi), %xmm4 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm12 = xmm12[0],xmm4[1],xmm12[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm12 = ymm13[0],ymm15[0],ymm13[2],ymm15[2] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm13, %ymm10 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 876(%rdi), %ymm15 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5,6],ymm15[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm12[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 304(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm12 = ymm5[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2],ymm0[3],ymm12[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 232(%rdi), %xmm15 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm0[1],xmm15[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm13[0],ymm4[0],ymm13[2],ymm4[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 428(%rdi), %ymm14 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5,6],ymm14[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm14[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd $253, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0],ymm1[1],mem[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm14[0,2],ymm15[1,3],ymm14[4,6],ymm15[5,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 208(%rdi), %ymm11 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm11[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm8[0],mem[1],ymm8[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm12 ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm7[0,2],ymm9[1,3],ymm7[4,6],ymm9[5,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 656(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0],ymm6[1],ymm5[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,2],ymm4[1,3],ymm13[4,6],ymm4[5,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm8 ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 432(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $2, (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,2],ymm9[1,3],ymm10[4,6],ymm9[5,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 880(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm1 = [4,3,4,3] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,2,3],ymm0[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm0, %ymm1, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 548(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 512(%rdi), %xmm6 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm6[0,1,2],xmm2[3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} ymm11 = [0,7,0,7,0,7,0,7] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm7, %ymm11, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm12[6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 660(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 100(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm0[0,1,2],xmm2[3] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm3, %ymm1, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3] ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm14, %ymm10 ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm14, %ymm11, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm15, %ymm7 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm15[6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 212(%rdi), %ymm4 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm2, %ymm1, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 324(%rdi), %xmm4 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm2[0,1,2],xmm4[3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm13, %ymm11, %ymm4 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm13, %ymm14 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm8[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm8, %ymm13 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 436(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm8[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm3, %ymm1, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 772(%rdi), %xmm4 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 736(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm3[0,1,2],xmm4[3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm15, %ymm11, %ymm4 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm9[6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 884(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm8[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0],ymm10[1],ymm7[2,3,4],ymm10[5],ymm7[6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 216(%rdi), %ymm4 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm4[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %xmm10 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1,2],xmm0[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm4 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm4[0,1],xmm0[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 544(%rdi), %xmm4 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm4[0,1,2],xmm6[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm5 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm5[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 664(%rdi), %ymm7 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm7[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %xmm12 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm12[0,1,2],xmm2[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm5 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm5[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 440(%rdi), %ymm7 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm7[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm5[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 768(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm3 = xmm2[0,1,2],xmm3[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm7, %xmm7 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm3 = xmm7[0,1],xmm3[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm9[0],ymm15[1],ymm9[2,3,4],ymm15[5],ymm9[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 888(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm3[0,1,2,3],ymm7[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 584(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3] ; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpermd 640(%rdi), %ymm11, %ymm8 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm8[6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 528(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm8[0,1,2],xmm4[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm8 = ymm8[0],mem[1],ymm8[2,3,4],mem[5],ymm8[6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm8, %xmm8 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm8[0,1],xmm4[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 808(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3] ; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpermd 864(%rdi), %ymm11, %ymm4 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 752(%rdi), %ymm4 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm4 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 136(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3] ; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpermd 192(%rdi), %ymm11, %ymm4 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 80(%rdi), %ymm4 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1,2],xmm10[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm6[0],mem[1],ymm6[2,3,4],mem[5],ymm6[6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm6, %xmm6 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 360(%rdi), %xmm4 ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3] ; AVX2-FAST-PERLANE-NEXT: vpermd 416(%rdi), %ymm11, %ymm6 ; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 304(%rdi), %ymm6 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm6 = xmm6[0,1,2],xmm12[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7] ; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm9, %xmm9 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm6 = xmm9[0,1],xmm6[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 96(%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 64(%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 96(%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 64(%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 96(%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 64(%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 96(%r8) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%r8) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 64(%r8) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%r8) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 96(%r9) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%r9) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%r9) ; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 64(%r9) ; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, 96(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, 32(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, 64(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, (%rax) ; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 32(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 96(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, 64(%rax) ; AVX2-FAST-PERLANE-NEXT: addq $1224, %rsp # imm = 0x4C8 ; AVX2-FAST-PERLANE-NEXT: vzeroupper ; AVX2-FAST-PERLANE-NEXT: retq ; ; AVX512F-LABEL: load_i32_stride7_vf32: ; AVX512F: # %bb.0: ; AVX512F-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX512F-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX512F-NEXT: vmovdqa64 512(%rdi), %zmm1 ; AVX512F-NEXT: vmovdqa64 448(%rdi), %zmm0 ; AVX512F-NEXT: vmovdqa64 576(%rdi), %zmm4 ; AVX512F-NEXT: vmovdqa64 640(%rdi), %zmm2 ; AVX512F-NEXT: vmovdqa64 832(%rdi), %zmm5 ; AVX512F-NEXT: vmovdqa64 768(%rdi), %zmm6 ; AVX512F-NEXT: vmovdqa64 704(%rdi), %zmm3 ; AVX512F-NEXT: vmovdqa64 384(%rdi), %zmm13 ; AVX512F-NEXT: vmovdqa64 320(%rdi), %zmm15 ; AVX512F-NEXT: vmovdqa64 256(%rdi), %zmm9 ; AVX512F-NEXT: vmovdqa64 (%rdi), %zmm10 ; AVX512F-NEXT: vmovdqa64 64(%rdi), %zmm11 ; AVX512F-NEXT: vmovdqa64 128(%rdi), %zmm14 ; AVX512F-NEXT: vmovdqa64 192(%rdi), %zmm12 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm16 = [8,15,0,0,0,19,26,1,8,15,0,0,0,19,26,1] ; AVX512F-NEXT: # zmm16 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm12, %zmm17 ; AVX512F-NEXT: vpermt2d %zmm14, %zmm16, %zmm17 ; AVX512F-NEXT: vmovdqa {{.*#+}} ymm7 = [0,7,14,21,28,u,u,u] ; AVX512F-NEXT: vmovdqa64 %zmm10, %zmm8 ; AVX512F-NEXT: vpermt2d %zmm11, %zmm7, %zmm8 ; AVX512F-NEXT: movw $992, %di # imm = 0x3E0 ; AVX512F-NEXT: kmovw %edi, %k1 ; AVX512F-NEXT: vmovdqa32 %zmm17, %zmm8 {%k1} ; AVX512F-NEXT: vbroadcasti32x4 {{.*#+}} zmm17 = [20,27,6,13,20,27,6,13,20,27,6,13,20,27,6,13] ; AVX512F-NEXT: # zmm17 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm9, %zmm18 ; AVX512F-NEXT: vpermt2d %zmm15, %zmm17, %zmm18 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm19 = [0,0,10,11,12,13,18,25,0,0,10,11,12,13,18,25] ; AVX512F-NEXT: # zmm19 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm13, %zmm19, %zmm18 ; AVX512F-NEXT: movb $-32, %dil ; AVX512F-NEXT: kmovw %edi, %k2 ; AVX512F-NEXT: vmovdqa64 %zmm18, %zmm8 {%k2} ; AVX512F-NEXT: vpermi2d %zmm6, %zmm3, %zmm17 ; AVX512F-NEXT: vpermt2d %zmm5, %zmm19, %zmm17 ; AVX512F-NEXT: vpermi2d %zmm4, %zmm2, %zmm16 ; AVX512F-NEXT: vpermi2d %zmm1, %zmm0, %zmm7 ; AVX512F-NEXT: vmovdqa32 %zmm16, %zmm7 {%k1} ; AVX512F-NEXT: vmovdqa64 %zmm17, %zmm7 {%k2} ; AVX512F-NEXT: vbroadcasti32x4 {{.*#+}} zmm18 = [25,4,11,18,25,4,11,18,25,4,11,18,25,4,11,18] ; AVX512F-NEXT: # zmm18 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm14, %zmm19 ; AVX512F-NEXT: vpermt2d %zmm12, %zmm18, %zmm19 ; AVX512F-NEXT: vmovdqa64 {{.*#+}} ymm16 = [1,8,15,22,29,u,u,u] ; AVX512F-NEXT: vmovdqa64 %zmm10, %zmm17 ; AVX512F-NEXT: vpermt2d %zmm11, %zmm16, %zmm17 ; AVX512F-NEXT: movw $480, %di # imm = 0x1E0 ; AVX512F-NEXT: kmovw %edi, %k2 ; AVX512F-NEXT: vmovdqa32 %zmm19, %zmm17 {%k2} ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm19 = [0,0,7,14,21,28,0,0,0,0,7,14,21,28,0,0] ; AVX512F-NEXT: # zmm19 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm9, %zmm20 ; AVX512F-NEXT: vpermt2d %zmm15, %zmm19, %zmm20 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm21 = [0,9,10,11,12,13,19,26,0,9,10,11,12,13,19,26] ; AVX512F-NEXT: # zmm21 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm13, %zmm21, %zmm20 ; AVX512F-NEXT: movw $-512, %di # imm = 0xFE00 ; AVX512F-NEXT: kmovw %edi, %k1 ; AVX512F-NEXT: vmovdqa32 %zmm20, %zmm17 {%k1} ; AVX512F-NEXT: vpermi2d %zmm6, %zmm3, %zmm19 ; AVX512F-NEXT: vpermt2d %zmm5, %zmm21, %zmm19 ; AVX512F-NEXT: vmovdqa64 %zmm4, %zmm20 ; AVX512F-NEXT: vpermt2d %zmm2, %zmm18, %zmm20 ; AVX512F-NEXT: vpermi2d %zmm1, %zmm0, %zmm16 ; AVX512F-NEXT: vmovdqa32 %zmm20, %zmm16 {%k2} ; AVX512F-NEXT: vmovdqa32 %zmm19, %zmm16 {%k1} ; AVX512F-NEXT: vbroadcasti32x4 {{.*#+}} zmm21 = [26,5,12,19,26,5,12,19,26,5,12,19,26,5,12,19] ; AVX512F-NEXT: # zmm21 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm14, %zmm22 ; AVX512F-NEXT: vpermt2d %zmm12, %zmm21, %zmm22 ; AVX512F-NEXT: vmovdqa64 {{.*#+}} ymm19 = [18,25,0,7,14,u,u,u] ; AVX512F-NEXT: vmovdqa64 %zmm11, %zmm20 ; AVX512F-NEXT: vpermt2d %zmm10, %zmm19, %zmm20 ; AVX512F-NEXT: vmovdqa32 %zmm22, %zmm20 {%k2} ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm22 = [0,1,8,15,22,29,0,0,0,1,8,15,22,29,0,0] ; AVX512F-NEXT: # zmm22 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm9, %zmm23 ; AVX512F-NEXT: vpermt2d %zmm15, %zmm22, %zmm23 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm24 = [0,9,10,11,12,13,20,27,0,9,10,11,12,13,20,27] ; AVX512F-NEXT: # zmm24 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm13, %zmm24, %zmm23 ; AVX512F-NEXT: vmovdqa32 %zmm23, %zmm20 {%k1} ; AVX512F-NEXT: vpermi2d %zmm6, %zmm3, %zmm22 ; AVX512F-NEXT: vpermt2d %zmm5, %zmm24, %zmm22 ; AVX512F-NEXT: vmovdqa64 %zmm4, %zmm23 ; AVX512F-NEXT: vpermt2d %zmm2, %zmm21, %zmm23 ; AVX512F-NEXT: vpermi2d %zmm0, %zmm1, %zmm19 ; AVX512F-NEXT: vmovdqa32 %zmm23, %zmm19 {%k2} ; AVX512F-NEXT: vmovdqa32 %zmm22, %zmm19 {%k1} ; AVX512F-NEXT: vbroadcasti32x4 {{.*#+}} zmm24 = [27,6,13,20,27,6,13,20,27,6,13,20,27,6,13,20] ; AVX512F-NEXT: # zmm24 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm14, %zmm25 ; AVX512F-NEXT: vpermt2d %zmm12, %zmm24, %zmm25 ; AVX512F-NEXT: vmovdqa64 {{.*#+}} ymm22 = [19,26,1,8,15,u,u,u] ; AVX512F-NEXT: vmovdqa64 %zmm11, %zmm23 ; AVX512F-NEXT: vpermt2d %zmm10, %zmm22, %zmm23 ; AVX512F-NEXT: vmovdqa32 %zmm25, %zmm23 {%k2} ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm25 = [0,18,25,0,7,14,0,0,0,18,25,0,7,14,0,0] ; AVX512F-NEXT: # zmm25 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm15, %zmm26 ; AVX512F-NEXT: vpermt2d %zmm9, %zmm25, %zmm26 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm27 = [0,9,10,11,12,13,21,28,0,9,10,11,12,13,21,28] ; AVX512F-NEXT: # zmm27 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm13, %zmm27, %zmm26 ; AVX512F-NEXT: vmovdqa32 %zmm26, %zmm23 {%k1} ; AVX512F-NEXT: vpermi2d %zmm3, %zmm6, %zmm25 ; AVX512F-NEXT: vpermt2d %zmm5, %zmm27, %zmm25 ; AVX512F-NEXT: vpermi2d %zmm2, %zmm4, %zmm24 ; AVX512F-NEXT: vpermi2d %zmm0, %zmm1, %zmm22 ; AVX512F-NEXT: vmovdqa32 %zmm24, %zmm22 {%k2} ; AVX512F-NEXT: vmovdqa32 %zmm25, %zmm22 {%k1} ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm26 = [0,19,26,1,8,15,0,0,0,19,26,1,8,15,0,0] ; AVX512F-NEXT: # zmm26 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm15, %zmm25 ; AVX512F-NEXT: vpermt2d %zmm9, %zmm26, %zmm25 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm27 = [0,9,10,11,12,13,22,29,0,9,10,11,12,13,22,29] ; AVX512F-NEXT: # zmm27 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm13, %zmm27, %zmm25 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm28 = [28,0,0,0,0,7,14,21,28,0,0,0,0,7,14,21] ; AVX512F-NEXT: # zmm28 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm14, %zmm24 ; AVX512F-NEXT: vpermt2d %zmm12, %zmm28, %zmm24 ; AVX512F-NEXT: vmovdqa64 {{.*#+}} xmm29 = [4,11,18,25] ; AVX512F-NEXT: vmovdqa64 %zmm10, %zmm30 ; AVX512F-NEXT: vpermt2d %zmm11, %zmm29, %zmm30 ; AVX512F-NEXT: vinserti32x4 $0, %xmm30, %zmm24, %zmm24 ; AVX512F-NEXT: vmovdqa32 %zmm25, %zmm24 {%k1} ; AVX512F-NEXT: vpermi2d %zmm3, %zmm6, %zmm26 ; AVX512F-NEXT: vpermt2d %zmm5, %zmm27, %zmm26 ; AVX512F-NEXT: vpermi2d %zmm2, %zmm4, %zmm28 ; AVX512F-NEXT: vpermi2d %zmm1, %zmm0, %zmm29 ; AVX512F-NEXT: vinserti32x4 $0, %xmm29, %zmm28, %zmm25 ; AVX512F-NEXT: vmovdqa32 %zmm26, %zmm25 {%k1} ; AVX512F-NEXT: vmovdqa64 %zmm9, %zmm26 ; AVX512F-NEXT: vpermt2d %zmm15, %zmm18, %zmm26 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm27 = [0,9,10,11,12,16,23,30,0,9,10,11,12,16,23,30] ; AVX512F-NEXT: # zmm27 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm13, %zmm27, %zmm26 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm28 = [29,0,0,0,1,8,15,22,29,0,0,0,1,8,15,22] ; AVX512F-NEXT: # zmm28 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm14, %zmm29 ; AVX512F-NEXT: vpermt2d %zmm12, %zmm28, %zmm29 ; AVX512F-NEXT: vmovdqa64 {{.*#+}} xmm30 = [5,12,19,26] ; AVX512F-NEXT: vmovdqa64 %zmm10, %zmm31 ; AVX512F-NEXT: vpermt2d %zmm11, %zmm30, %zmm31 ; AVX512F-NEXT: vinserti32x4 $0, %xmm31, %zmm29, %zmm29 ; AVX512F-NEXT: vmovdqa32 %zmm26, %zmm29 {%k1} ; AVX512F-NEXT: vpermi2d %zmm6, %zmm3, %zmm18 ; AVX512F-NEXT: vpermt2d %zmm5, %zmm27, %zmm18 ; AVX512F-NEXT: vpermi2d %zmm2, %zmm4, %zmm28 ; AVX512F-NEXT: vpermi2d %zmm1, %zmm0, %zmm30 ; AVX512F-NEXT: vinserti32x4 $0, %xmm30, %zmm28, %zmm26 ; AVX512F-NEXT: vmovdqa32 %zmm18, %zmm26 {%k1} ; AVX512F-NEXT: vpermt2d %zmm15, %zmm21, %zmm9 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm15 = [0,9,10,11,12,17,24,31,0,9,10,11,12,17,24,31] ; AVX512F-NEXT: # zmm15 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm13, %zmm15, %zmm9 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm13 = [14,0,0,0,18,25,0,7,14,0,0,0,18,25,0,7] ; AVX512F-NEXT: # zmm13 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm14, %zmm13, %zmm12 ; AVX512F-NEXT: vmovdqa {{.*#+}} xmm14 = [6,13,20,27] ; AVX512F-NEXT: vpermt2d %zmm11, %zmm14, %zmm10 ; AVX512F-NEXT: vinserti32x4 $0, %xmm10, %zmm12, %zmm10 ; AVX512F-NEXT: vmovdqa32 %zmm9, %zmm10 {%k1} ; AVX512F-NEXT: vpermt2d %zmm6, %zmm21, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm5, %zmm15, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm4, %zmm13, %zmm2 ; AVX512F-NEXT: vpermt2d %zmm1, %zmm14, %zmm0 ; AVX512F-NEXT: vinserti32x4 $0, %xmm0, %zmm2, %zmm0 ; AVX512F-NEXT: vmovdqa32 %zmm3, %zmm0 {%k1} ; AVX512F-NEXT: vmovdqa64 %zmm7, 64(%rsi) ; AVX512F-NEXT: vmovdqa64 %zmm8, (%rsi) ; AVX512F-NEXT: vmovdqa64 %zmm16, 64(%rdx) ; AVX512F-NEXT: vmovdqa64 %zmm17, (%rdx) ; AVX512F-NEXT: vmovdqa64 %zmm19, 64(%rcx) ; AVX512F-NEXT: vmovdqa64 %zmm20, (%rcx) ; AVX512F-NEXT: vmovdqa64 %zmm22, 64(%r8) ; AVX512F-NEXT: vmovdqa64 %zmm23, (%r8) ; AVX512F-NEXT: vmovdqa64 %zmm25, 64(%r9) ; AVX512F-NEXT: vmovdqa64 %zmm24, (%r9) ; AVX512F-NEXT: vmovdqa64 %zmm26, 64(%r10) ; AVX512F-NEXT: vmovdqa64 %zmm29, (%r10) ; AVX512F-NEXT: vmovdqa64 %zmm0, 64(%rax) ; AVX512F-NEXT: vmovdqa64 %zmm10, (%rax) ; AVX512F-NEXT: vzeroupper ; AVX512F-NEXT: retq ; ; AVX512BW-LABEL: load_i32_stride7_vf32: ; AVX512BW: # %bb.0: ; AVX512BW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX512BW-NEXT: movq {{[0-9]+}}(%rsp), %r10 ; AVX512BW-NEXT: vmovdqa64 512(%rdi), %zmm1 ; AVX512BW-NEXT: vmovdqa64 448(%rdi), %zmm0 ; AVX512BW-NEXT: vmovdqa64 576(%rdi), %zmm4 ; AVX512BW-NEXT: vmovdqa64 640(%rdi), %zmm2 ; AVX512BW-NEXT: vmovdqa64 832(%rdi), %zmm5 ; AVX512BW-NEXT: vmovdqa64 768(%rdi), %zmm6 ; AVX512BW-NEXT: vmovdqa64 704(%rdi), %zmm3 ; AVX512BW-NEXT: vmovdqa64 384(%rdi), %zmm13 ; AVX512BW-NEXT: vmovdqa64 320(%rdi), %zmm15 ; AVX512BW-NEXT: vmovdqa64 256(%rdi), %zmm9 ; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm10 ; AVX512BW-NEXT: vmovdqa64 64(%rdi), %zmm11 ; AVX512BW-NEXT: vmovdqa64 128(%rdi), %zmm14 ; AVX512BW-NEXT: vmovdqa64 192(%rdi), %zmm12 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm16 = [8,15,0,0,0,19,26,1,8,15,0,0,0,19,26,1] ; AVX512BW-NEXT: # zmm16 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm12, %zmm17 ; AVX512BW-NEXT: vpermt2d %zmm14, %zmm16, %zmm17 ; AVX512BW-NEXT: vmovdqa {{.*#+}} ymm7 = [0,7,14,21,28,u,u,u] ; AVX512BW-NEXT: vmovdqa64 %zmm10, %zmm8 ; AVX512BW-NEXT: vpermt2d %zmm11, %zmm7, %zmm8 ; AVX512BW-NEXT: movw $992, %di # imm = 0x3E0 ; AVX512BW-NEXT: kmovd %edi, %k1 ; AVX512BW-NEXT: vmovdqa32 %zmm17, %zmm8 {%k1} ; AVX512BW-NEXT: vbroadcasti32x4 {{.*#+}} zmm17 = [20,27,6,13,20,27,6,13,20,27,6,13,20,27,6,13] ; AVX512BW-NEXT: # zmm17 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm9, %zmm18 ; AVX512BW-NEXT: vpermt2d %zmm15, %zmm17, %zmm18 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm19 = [0,0,10,11,12,13,18,25,0,0,10,11,12,13,18,25] ; AVX512BW-NEXT: # zmm19 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm13, %zmm19, %zmm18 ; AVX512BW-NEXT: movb $-32, %dil ; AVX512BW-NEXT: kmovd %edi, %k2 ; AVX512BW-NEXT: vmovdqa64 %zmm18, %zmm8 {%k2} ; AVX512BW-NEXT: vpermi2d %zmm6, %zmm3, %zmm17 ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm19, %zmm17 ; AVX512BW-NEXT: vpermi2d %zmm4, %zmm2, %zmm16 ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm0, %zmm7 ; AVX512BW-NEXT: vmovdqa32 %zmm16, %zmm7 {%k1} ; AVX512BW-NEXT: vmovdqa64 %zmm17, %zmm7 {%k2} ; AVX512BW-NEXT: vbroadcasti32x4 {{.*#+}} zmm18 = [25,4,11,18,25,4,11,18,25,4,11,18,25,4,11,18] ; AVX512BW-NEXT: # zmm18 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm14, %zmm19 ; AVX512BW-NEXT: vpermt2d %zmm12, %zmm18, %zmm19 ; AVX512BW-NEXT: vmovdqa64 {{.*#+}} ymm16 = [1,8,15,22,29,u,u,u] ; AVX512BW-NEXT: vmovdqa64 %zmm10, %zmm17 ; AVX512BW-NEXT: vpermt2d %zmm11, %zmm16, %zmm17 ; AVX512BW-NEXT: movw $480, %di # imm = 0x1E0 ; AVX512BW-NEXT: kmovd %edi, %k2 ; AVX512BW-NEXT: vmovdqa32 %zmm19, %zmm17 {%k2} ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm19 = [0,0,7,14,21,28,0,0,0,0,7,14,21,28,0,0] ; AVX512BW-NEXT: # zmm19 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm9, %zmm20 ; AVX512BW-NEXT: vpermt2d %zmm15, %zmm19, %zmm20 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm21 = [0,9,10,11,12,13,19,26,0,9,10,11,12,13,19,26] ; AVX512BW-NEXT: # zmm21 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm13, %zmm21, %zmm20 ; AVX512BW-NEXT: movw $-512, %di # imm = 0xFE00 ; AVX512BW-NEXT: kmovd %edi, %k1 ; AVX512BW-NEXT: vmovdqa32 %zmm20, %zmm17 {%k1} ; AVX512BW-NEXT: vpermi2d %zmm6, %zmm3, %zmm19 ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm21, %zmm19 ; AVX512BW-NEXT: vmovdqa64 %zmm4, %zmm20 ; AVX512BW-NEXT: vpermt2d %zmm2, %zmm18, %zmm20 ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm0, %zmm16 ; AVX512BW-NEXT: vmovdqa32 %zmm20, %zmm16 {%k2} ; AVX512BW-NEXT: vmovdqa32 %zmm19, %zmm16 {%k1} ; AVX512BW-NEXT: vbroadcasti32x4 {{.*#+}} zmm21 = [26,5,12,19,26,5,12,19,26,5,12,19,26,5,12,19] ; AVX512BW-NEXT: # zmm21 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm14, %zmm22 ; AVX512BW-NEXT: vpermt2d %zmm12, %zmm21, %zmm22 ; AVX512BW-NEXT: vmovdqa64 {{.*#+}} ymm19 = [18,25,0,7,14,u,u,u] ; AVX512BW-NEXT: vmovdqa64 %zmm11, %zmm20 ; AVX512BW-NEXT: vpermt2d %zmm10, %zmm19, %zmm20 ; AVX512BW-NEXT: vmovdqa32 %zmm22, %zmm20 {%k2} ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm22 = [0,1,8,15,22,29,0,0,0,1,8,15,22,29,0,0] ; AVX512BW-NEXT: # zmm22 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm9, %zmm23 ; AVX512BW-NEXT: vpermt2d %zmm15, %zmm22, %zmm23 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm24 = [0,9,10,11,12,13,20,27,0,9,10,11,12,13,20,27] ; AVX512BW-NEXT: # zmm24 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm13, %zmm24, %zmm23 ; AVX512BW-NEXT: vmovdqa32 %zmm23, %zmm20 {%k1} ; AVX512BW-NEXT: vpermi2d %zmm6, %zmm3, %zmm22 ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm24, %zmm22 ; AVX512BW-NEXT: vmovdqa64 %zmm4, %zmm23 ; AVX512BW-NEXT: vpermt2d %zmm2, %zmm21, %zmm23 ; AVX512BW-NEXT: vpermi2d %zmm0, %zmm1, %zmm19 ; AVX512BW-NEXT: vmovdqa32 %zmm23, %zmm19 {%k2} ; AVX512BW-NEXT: vmovdqa32 %zmm22, %zmm19 {%k1} ; AVX512BW-NEXT: vbroadcasti32x4 {{.*#+}} zmm24 = [27,6,13,20,27,6,13,20,27,6,13,20,27,6,13,20] ; AVX512BW-NEXT: # zmm24 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm14, %zmm25 ; AVX512BW-NEXT: vpermt2d %zmm12, %zmm24, %zmm25 ; AVX512BW-NEXT: vmovdqa64 {{.*#+}} ymm22 = [19,26,1,8,15,u,u,u] ; AVX512BW-NEXT: vmovdqa64 %zmm11, %zmm23 ; AVX512BW-NEXT: vpermt2d %zmm10, %zmm22, %zmm23 ; AVX512BW-NEXT: vmovdqa32 %zmm25, %zmm23 {%k2} ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm25 = [0,18,25,0,7,14,0,0,0,18,25,0,7,14,0,0] ; AVX512BW-NEXT: # zmm25 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm15, %zmm26 ; AVX512BW-NEXT: vpermt2d %zmm9, %zmm25, %zmm26 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm27 = [0,9,10,11,12,13,21,28,0,9,10,11,12,13,21,28] ; AVX512BW-NEXT: # zmm27 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm13, %zmm27, %zmm26 ; AVX512BW-NEXT: vmovdqa32 %zmm26, %zmm23 {%k1} ; AVX512BW-NEXT: vpermi2d %zmm3, %zmm6, %zmm25 ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm27, %zmm25 ; AVX512BW-NEXT: vpermi2d %zmm2, %zmm4, %zmm24 ; AVX512BW-NEXT: vpermi2d %zmm0, %zmm1, %zmm22 ; AVX512BW-NEXT: vmovdqa32 %zmm24, %zmm22 {%k2} ; AVX512BW-NEXT: vmovdqa32 %zmm25, %zmm22 {%k1} ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm26 = [0,19,26,1,8,15,0,0,0,19,26,1,8,15,0,0] ; AVX512BW-NEXT: # zmm26 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm15, %zmm25 ; AVX512BW-NEXT: vpermt2d %zmm9, %zmm26, %zmm25 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm27 = [0,9,10,11,12,13,22,29,0,9,10,11,12,13,22,29] ; AVX512BW-NEXT: # zmm27 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm13, %zmm27, %zmm25 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm28 = [28,0,0,0,0,7,14,21,28,0,0,0,0,7,14,21] ; AVX512BW-NEXT: # zmm28 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm14, %zmm24 ; AVX512BW-NEXT: vpermt2d %zmm12, %zmm28, %zmm24 ; AVX512BW-NEXT: vmovdqa64 {{.*#+}} xmm29 = [4,11,18,25] ; AVX512BW-NEXT: vmovdqa64 %zmm10, %zmm30 ; AVX512BW-NEXT: vpermt2d %zmm11, %zmm29, %zmm30 ; AVX512BW-NEXT: vinserti32x4 $0, %xmm30, %zmm24, %zmm24 ; AVX512BW-NEXT: vmovdqa32 %zmm25, %zmm24 {%k1} ; AVX512BW-NEXT: vpermi2d %zmm3, %zmm6, %zmm26 ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm27, %zmm26 ; AVX512BW-NEXT: vpermi2d %zmm2, %zmm4, %zmm28 ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm0, %zmm29 ; AVX512BW-NEXT: vinserti32x4 $0, %xmm29, %zmm28, %zmm25 ; AVX512BW-NEXT: vmovdqa32 %zmm26, %zmm25 {%k1} ; AVX512BW-NEXT: vmovdqa64 %zmm9, %zmm26 ; AVX512BW-NEXT: vpermt2d %zmm15, %zmm18, %zmm26 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm27 = [0,9,10,11,12,16,23,30,0,9,10,11,12,16,23,30] ; AVX512BW-NEXT: # zmm27 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm13, %zmm27, %zmm26 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm28 = [29,0,0,0,1,8,15,22,29,0,0,0,1,8,15,22] ; AVX512BW-NEXT: # zmm28 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm14, %zmm29 ; AVX512BW-NEXT: vpermt2d %zmm12, %zmm28, %zmm29 ; AVX512BW-NEXT: vmovdqa64 {{.*#+}} xmm30 = [5,12,19,26] ; AVX512BW-NEXT: vmovdqa64 %zmm10, %zmm31 ; AVX512BW-NEXT: vpermt2d %zmm11, %zmm30, %zmm31 ; AVX512BW-NEXT: vinserti32x4 $0, %xmm31, %zmm29, %zmm29 ; AVX512BW-NEXT: vmovdqa32 %zmm26, %zmm29 {%k1} ; AVX512BW-NEXT: vpermi2d %zmm6, %zmm3, %zmm18 ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm27, %zmm18 ; AVX512BW-NEXT: vpermi2d %zmm2, %zmm4, %zmm28 ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm0, %zmm30 ; AVX512BW-NEXT: vinserti32x4 $0, %xmm30, %zmm28, %zmm26 ; AVX512BW-NEXT: vmovdqa32 %zmm18, %zmm26 {%k1} ; AVX512BW-NEXT: vpermt2d %zmm15, %zmm21, %zmm9 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm15 = [0,9,10,11,12,17,24,31,0,9,10,11,12,17,24,31] ; AVX512BW-NEXT: # zmm15 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm13, %zmm15, %zmm9 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm13 = [14,0,0,0,18,25,0,7,14,0,0,0,18,25,0,7] ; AVX512BW-NEXT: # zmm13 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm14, %zmm13, %zmm12 ; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm14 = [6,13,20,27] ; AVX512BW-NEXT: vpermt2d %zmm11, %zmm14, %zmm10 ; AVX512BW-NEXT: vinserti32x4 $0, %xmm10, %zmm12, %zmm10 ; AVX512BW-NEXT: vmovdqa32 %zmm9, %zmm10 {%k1} ; AVX512BW-NEXT: vpermt2d %zmm6, %zmm21, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm15, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm4, %zmm13, %zmm2 ; AVX512BW-NEXT: vpermt2d %zmm1, %zmm14, %zmm0 ; AVX512BW-NEXT: vinserti32x4 $0, %xmm0, %zmm2, %zmm0 ; AVX512BW-NEXT: vmovdqa32 %zmm3, %zmm0 {%k1} ; AVX512BW-NEXT: vmovdqa64 %zmm7, 64(%rsi) ; AVX512BW-NEXT: vmovdqa64 %zmm8, (%rsi) ; AVX512BW-NEXT: vmovdqa64 %zmm16, 64(%rdx) ; AVX512BW-NEXT: vmovdqa64 %zmm17, (%rdx) ; AVX512BW-NEXT: vmovdqa64 %zmm19, 64(%rcx) ; AVX512BW-NEXT: vmovdqa64 %zmm20, (%rcx) ; AVX512BW-NEXT: vmovdqa64 %zmm22, 64(%r8) ; AVX512BW-NEXT: vmovdqa64 %zmm23, (%r8) ; AVX512BW-NEXT: vmovdqa64 %zmm25, 64(%r9) ; AVX512BW-NEXT: vmovdqa64 %zmm24, (%r9) ; AVX512BW-NEXT: vmovdqa64 %zmm26, 64(%r10) ; AVX512BW-NEXT: vmovdqa64 %zmm29, (%r10) ; AVX512BW-NEXT: vmovdqa64 %zmm0, 64(%rax) ; AVX512BW-NEXT: vmovdqa64 %zmm10, (%rax) ; AVX512BW-NEXT: vzeroupper ; AVX512BW-NEXT: retq %wide.vec = load <224 x i32>, ptr %in.vec, align 64 %strided.vec0 = shufflevector <224 x i32> %wide.vec, <224 x i32> poison, <32 x i32> %strided.vec1 = shufflevector <224 x i32> %wide.vec, <224 x i32> poison, <32 x i32> %strided.vec2 = shufflevector <224 x i32> %wide.vec, <224 x i32> poison, <32 x i32> %strided.vec3 = shufflevector <224 x i32> %wide.vec, <224 x i32> poison, <32 x i32> %strided.vec4 = shufflevector <224 x i32> %wide.vec, <224 x i32> poison, <32 x i32> %strided.vec5 = shufflevector <224 x i32> %wide.vec, <224 x i32> poison, <32 x i32> %strided.vec6 = shufflevector <224 x i32> %wide.vec, <224 x i32> poison, <32 x i32> store <32 x i32> %strided.vec0, ptr %out.vec0, align 64 store <32 x i32> %strided.vec1, ptr %out.vec1, align 64 store <32 x i32> %strided.vec2, ptr %out.vec2, align 64 store <32 x i32> %strided.vec3, ptr %out.vec3, align 64 store <32 x i32> %strided.vec4, ptr %out.vec4, align 64 store <32 x i32> %strided.vec5, ptr %out.vec5, align 64 store <32 x i32> %strided.vec6, ptr %out.vec6, align 64 ret void } define void @load_i32_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind { ; SSE-LABEL: load_i32_stride7_vf64: ; SSE: # %bb.0: ; SSE-NEXT: subq $2456, %rsp # imm = 0x998 ; SSE-NEXT: movdqa 1088(%rdi), %xmm3 ; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1056(%rdi), %xmm4 ; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1008(%rdi), %xmm9 ; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1024(%rdi), %xmm5 ; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 640(%rdi), %xmm13 ; SSE-NEXT: movdqa 608(%rdi), %xmm6 ; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 560(%rdi), %xmm10 ; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 576(%rdi), %xmm7 ; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 192(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 160(%rdi), %xmm15 ; SSE-NEXT: movdqa 112(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 128(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3] ; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3] ; SSE-NEXT: movdqa %xmm10, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1] ; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3] ; SSE-NEXT: movdqa %xmm9, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1456(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1472(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: movdqa 1536(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1504(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa (%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 16(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: movdqa 80(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 48(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 448(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 464(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: movdqa 528(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 496(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 896(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 912(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: movdqa 976(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 944(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1344(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1360(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: movdqa 1424(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1392(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 336(%rdi), %xmm12 ; SSE-NEXT: movdqa 352(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3] ; SSE-NEXT: movdqa %xmm12, %xmm5 ; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1] ; SSE-NEXT: movdqa 416(%rdi), %xmm4 ; SSE-NEXT: movdqa 384(%rdi), %xmm10 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3] ; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1] ; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 784(%rdi), %xmm6 ; SSE-NEXT: movdqa %xmm6, (%rsp) # 16-byte Spill ; SSE-NEXT: movdqa 800(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1] ; SSE-NEXT: movdqa 864(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 832(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm6[0],xmm5[1] ; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1232(%rdi), %xmm6 ; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1248(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1] ; SSE-NEXT: movdqa 1312(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1280(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm6[0],xmm5[1] ; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1680(%rdi), %xmm6 ; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1696(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1] ; SSE-NEXT: movdqa 1760(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1728(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm6[0],xmm5[1] ; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 224(%rdi), %xmm8 ; SSE-NEXT: movdqa 240(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[3,3,3,3] ; SSE-NEXT: movdqa %xmm8, %xmm6 ; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1] ; SSE-NEXT: movdqa 304(%rdi), %xmm2 ; SSE-NEXT: movdqa 272(%rdi), %xmm3 ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm3[2,2,3,3] ; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1] ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm6[0],xmm5[1] ; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 672(%rdi), %xmm11 ; SSE-NEXT: movdqa 688(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[3,3,3,3] ; SSE-NEXT: movdqa %xmm11, %xmm7 ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1] ; SSE-NEXT: movdqa 752(%rdi), %xmm14 ; SSE-NEXT: movdqa 720(%rdi), %xmm1 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,2,3,3] ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1120(%rdi), %xmm9 ; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1136(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm7[0],xmm9[1],xmm7[1] ; SSE-NEXT: movdqa 1200(%rdi), %xmm5 ; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1168(%rdi), %xmm6 ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3] ; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm9[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1568(%rdi), %xmm9 ; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1584(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm0[3,3,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm7[0],xmm9[1],xmm7[1] ; SSE-NEXT: movdqa 1648(%rdi), %xmm5 ; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1616(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm9[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm0[2,2,2,2] ; SSE-NEXT: punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm7[2],xmm15[3],xmm7[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[1,1,1,1] ; SSE-NEXT: movdqa 144(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm7[0],xmm15[1] ; SSE-NEXT: movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm15[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm7[2],xmm9[3],xmm7[3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm0[1,1,1,1] ; SSE-NEXT: movdqa 32(%rdi), %xmm5 ; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1] ; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm7[0],xmm9[1] ; SSE-NEXT: movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm4[2,2,2,2] ; SSE-NEXT: punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm7[2],xmm10[3],xmm7[3] ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm12[1,1,1,1] ; SSE-NEXT: movdqa 368(%rdi), %xmm4 ; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1] ; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm7[0],xmm10[1] ; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm2[2,2,2,2] ; SSE-NEXT: punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm7[2],xmm3[3],xmm7[3] ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm8[1,1,1,1] ; SSE-NEXT: movdqa 256(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm7[0],xmm3[1] ; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm13[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[1,1,1,1] ; SSE-NEXT: movdqa 592(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm7[0],xmm4[1] ; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm4[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm7[2],xmm2[3],xmm7[3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm3[1,1,1,1] ; SSE-NEXT: movdqa 480(%rdi), %xmm5 ; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm7[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm7[2],xmm2[3],xmm7[3] ; SSE-NEXT: movdqa (%rsp), %xmm13 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm13[1,1,1,1] ; SSE-NEXT: movdqa 816(%rdi), %xmm5 ; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm7[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm14[2,2,2,2] ; SSE-NEXT: movdqa %xmm1, %xmm2 ; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm7[2],xmm2[3],xmm7[3] ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm11[1,1,1,1] ; SSE-NEXT: movdqa %xmm11, %xmm12 ; SSE-NEXT: movdqa 704(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm7[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[1,1,1,1] ; SSE-NEXT: movdqa 1040(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm7[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm5[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[1,1,1,1] ; SSE-NEXT: movdqa 928(%rdi), %xmm2 ; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm7[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm1[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm7[2],xmm2[3],xmm7[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[1,1,1,1] ; SSE-NEXT: movdqa 1264(%rdi), %xmm1 ; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm7[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm8[2,2,2,2] ; SSE-NEXT: movdqa %xmm6, %xmm2 ; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm7[2],xmm2[3],xmm7[3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm1[1,1,1,1] ; SSE-NEXT: movdqa 1152(%rdi), %xmm6 ; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm7[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm7[2],xmm2[3],xmm7[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[1,1,1,1] ; SSE-NEXT: movdqa 1488(%rdi), %xmm6 ; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm7[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm10[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm7[2],xmm2[3],xmm7[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[1,1,1,1] ; SSE-NEXT: movdqa 1376(%rdi), %xmm6 ; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm7[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm11[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm7[2],xmm2[3],xmm7[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[1,1,1,1] ; SSE-NEXT: movdqa 1712(%rdi), %xmm6 ; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm7[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm6[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm7[2],xmm2[3],xmm7[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[1,1,1,1] ; SSE-NEXT: movdqa 1600(%rdi), %xmm9 ; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm7[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm0[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 64(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm15[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm7[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm2[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 176(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 288(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 400(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm3[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 512(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm4[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 624(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa %xmm12, %xmm15 ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm12[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 736(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm14[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm13[2,3,2,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm12[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 848(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 960(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm5[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[2,3,2,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm14[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 1072(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm1[2,3,2,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm3[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 1184(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm8[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm8[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 1296(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 1408(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm10[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 1520(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 1632(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm6[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[2,3,2,3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1] ; SSE-NEXT: movdqa 1744(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm11[2,3,2,3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 96(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; SSE-NEXT: movdqa %xmm5, %xmm0 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload ; SSE-NEXT: # xmm4 = mem[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 208(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1] ; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,2,3,3] ; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload ; SSE-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 320(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload ; SSE-NEXT: movdqa %xmm13, %xmm0 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload ; SSE-NEXT: # xmm2 = mem[2,2,3,3] ; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload ; SSE-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 432(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload ; SSE-NEXT: movdqa %xmm11, %xmm0 ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm10[2],xmm1[3],xmm10[3] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 544(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,2,3,3] ; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 656(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,2,3,3] ; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 768(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[2,2,3,3] ; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 880(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload ; SSE-NEXT: movdqa %xmm15, %xmm2 ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] ; SSE-NEXT: pshufd $250, (%rsp), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 992(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1104(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload ; SSE-NEXT: movdqa %xmm12, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm14[2],xmm0[3],xmm14[3] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1216(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1328(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload ; SSE-NEXT: movdqa %xmm14, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3] ; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1440(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload ; SSE-NEXT: movdqa %xmm9, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1552(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload ; SSE-NEXT: movdqa %xmm6, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1664(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa 1776(%rdi), %xmm0 ; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; SSE-NEXT: movdqa %xmm4, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload ; SSE-NEXT: movdqa %xmm7, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; SSE-NEXT: movdqa %xmm5, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload ; SSE-NEXT: movdqa %xmm8, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload ; SSE-NEXT: movdqa %xmm10, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload ; SSE-NEXT: movdqa %xmm11, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload ; SSE-NEXT: movdqa %xmm13, %xmm1 ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload ; SSE-NEXT: # xmm15 = mem[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload ; SSE-NEXT: # xmm15 = xmm15[0],mem[0],xmm15[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1] ; SSE-NEXT: movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3] ; SSE-NEXT: movdqa (%rsp), %xmm15 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload ; SSE-NEXT: # xmm2 = mem[3,3,3,3] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1] ; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1] ; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,2,2] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1] ; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,2,2,2] ; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1] ; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload ; SSE-NEXT: # xmm4 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1] ; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload ; SSE-NEXT: # xmm4 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1] ; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload ; SSE-NEXT: # xmm15 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload ; SSE-NEXT: # xmm14 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm1[0],xmm14[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload ; SSE-NEXT: # xmm13 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm1[0],xmm13[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload ; SSE-NEXT: # xmm12 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm1[0],xmm12[1] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload ; SSE-NEXT: # xmm11 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm11 = xmm1[0],xmm11[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload ; SSE-NEXT: # xmm10 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm1[0],xmm10[1] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; SSE-NEXT: # xmm9 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, (%rsp), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload ; SSE-NEXT: # xmm8 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload ; SSE-NEXT: # xmm7 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload ; SSE-NEXT: # xmm6 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm1[0],xmm6[1] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload ; SSE-NEXT: # xmm5 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload ; SSE-NEXT: # xmm4 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1] ; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload ; SSE-NEXT: # xmm3 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1] ; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1] ; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; SSE-NEXT: # xmm1 = mem[2,3,2,3] ; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] ; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload ; SSE-NEXT: # xmm0 = mem[2,3,2,3] ; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload ; SSE-NEXT: # xmm2 = mem[0,0,1,1] ; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1] ; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1] ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 224(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 160(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 96(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 32(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 240(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 176(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 112(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 48(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 192(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 128(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 64(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, (%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 208(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 144(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 80(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%rsi) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 224(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 240(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 192(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 208(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 160(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 176(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 128(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 144(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 96(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 112(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 64(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 80(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 32(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 48(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, (%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%rdx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 240(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 224(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 208(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 192(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 176(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 160(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 144(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 128(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 112(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 96(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 80(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 64(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 48(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 32(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, (%rcx) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 240(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 224(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 208(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 192(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 176(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 160(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 144(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 128(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 112(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 96(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 80(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 64(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 48(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 32(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, (%r8) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 240(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 224(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 208(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 192(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 176(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 160(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 144(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 128(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 112(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 96(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 80(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 64(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 48(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 32(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%r9) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, (%r9) ; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 240(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 224(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 208(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 192(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 176(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 160(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 144(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 128(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 112(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 96(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 80(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 64(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 48(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 32(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, (%rax) ; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; SSE-NEXT: movapd %xmm2, 240(%rax) ; SSE-NEXT: movapd %xmm3, 224(%rax) ; SSE-NEXT: movapd %xmm4, 208(%rax) ; SSE-NEXT: movapd %xmm5, 192(%rax) ; SSE-NEXT: movapd %xmm6, 176(%rax) ; SSE-NEXT: movapd %xmm7, 160(%rax) ; SSE-NEXT: movapd %xmm8, 144(%rax) ; SSE-NEXT: movapd %xmm9, 128(%rax) ; SSE-NEXT: movapd %xmm10, 112(%rax) ; SSE-NEXT: movapd %xmm11, 96(%rax) ; SSE-NEXT: movapd %xmm12, 80(%rax) ; SSE-NEXT: movapd %xmm13, 64(%rax) ; SSE-NEXT: movapd %xmm14, 48(%rax) ; SSE-NEXT: movapd %xmm15, 32(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, 16(%rax) ; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload ; SSE-NEXT: movaps %xmm0, (%rax) ; SSE-NEXT: addq $2456, %rsp # imm = 0x998 ; SSE-NEXT: retq ; ; AVX1-ONLY-LABEL: load_i32_stride7_vf64: ; AVX1-ONLY: # %bb.0: ; AVX1-ONLY-NEXT: subq $3176, %rsp # imm = 0xC68 ; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm2 ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %ymm3 ; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm5 ; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %ymm4 ; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm6 ; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[2],ymm0[2] ; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6],ymm1[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1 ; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm13 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm4 ; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm4[1] ; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm4 ; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm4[1] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 752(%rdi), %xmm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm0[0],ymm5[2],ymm0[2] ; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1 ; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3] ; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm3 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1120(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX1-ONLY-NEXT: vmovaps 1120(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %ymm9 ; AVX1-ONLY-NEXT: vmovaps 1200(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm9[0],ymm1[0],ymm9[2],ymm1[2] ; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 1280(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1] ; AVX1-ONLY-NEXT: vmovaps 1312(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1600(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1568(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX1-ONLY-NEXT: vmovaps 1568(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %ymm2 ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1648(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1696(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1] ; AVX1-ONLY-NEXT: vmovaps 1760(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm15 ; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm15[0],ymm1[0],ymm15[2],ymm1[2] ; AVX1-ONLY-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %ymm2 ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 528(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1] ; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %xmm12 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm12[0,1],xmm0[2,3] ; AVX1-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %ymm4 ; AVX1-ONLY-NEXT: vmovaps 976(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm4[0],ymm1[0],ymm4[2],ymm1[2] ; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1] ; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %xmm10 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm10[1] ; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1376(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1344(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0 ; AVX1-ONLY-NEXT: vmovaps 1344(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 1440(%rdi), %ymm8 ; AVX1-ONLY-NEXT: vmovaps 1424(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm8[0],ymm1[0],ymm8[2],ymm1[2] ; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 1504(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX1-ONLY-NEXT: vmovaps 1536(%rdi), %xmm14 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm14[1] ; AVX1-ONLY-NEXT: vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[1,1],ymm0[2,2],ymm6[5,5],ymm0[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm13[1],xmm1[2,3] ; AVX1-ONLY-NEXT: vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm2 ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1 ; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm1 = zero,xmm1[1,2],mem[0] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm5[1,1],ymm0[2,2],ymm5[5,5],ymm0[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3] ; AVX1-ONLY-NEXT: vmovaps %xmm3, %xmm5 ; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %ymm3 ; AVX1-ONLY-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm3[2,3],ymm1[0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,0],ymm1[3,3],ymm3[4,4],ymm1[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1 ; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm1 = zero,xmm1[1,2],mem[0] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1184(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm9[1,1],ymm0[2,2],ymm9[5,5],ymm0[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %xmm11 ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm11[0],xmm9[1],xmm11[2,3] ; AVX1-ONLY-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 1280(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %ymm2 ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1 ; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm1 = zero,xmm1[1,2],mem[0] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1632(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,1],ymm0[2,2],ymm1[5,5],ymm0[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 1600(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm7[1],xmm1[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1696(%rdi), %ymm2 ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1 ; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm1 = zero,xmm1[1,2],mem[0] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm15[1,1],ymm0[2,2],ymm15[5,5],ymm0[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm2 ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3],ymm2[0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[0,0],ymm2[3,3],ymm0[4,4],ymm2[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2 ; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm2 = zero,xmm2[1,2],mem[0] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[1,1],ymm0[2,2],ymm6[5,5],ymm0[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm0 ; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm0[0],xmm15[1],xmm0[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],ymm0[0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[0,0],ymm3[3,3],ymm1[4,4],ymm3[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3 ; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm3 = zero,xmm3[1,2],mem[0] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[1,1],ymm0[2,2],ymm4[5,5],ymm0[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %xmm0 ; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0],xmm12[1],xmm0[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm0[0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm1[0,0],ymm4[3,3],ymm1[4,4],ymm4[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm4 = zero,xmm4[1,2],xmm10[2] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm8[1,1],ymm0[2,2],ymm8[5,5],ymm0[6,6] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm4[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps 1376(%rdi), %xmm0 ; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm0[0],xmm3[1],xmm0[2,3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm12 = xmm12[1,0],mem[3,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm4[3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 1504(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm1[2,3],ymm0[0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm1[0,0],ymm10[3,3],ymm1[4,4],ymm10[7,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10 ; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm10 = zero,xmm10[1,2],xmm14[2] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3,4],ymm10[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm13[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],xmm14[1],xmm10[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm12 = ymm12[3,1],mem[0,3],ymm12[7,5],mem[4,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm0[2,1],ymm12[2,0],ymm0[6,5],ymm12[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm12[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm0[0],ymm8[0],ymm0[2],ymm8[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12 ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm12 = xmm12[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm12[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm5[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],xmm5[1],xmm10[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm12 = ymm12[3,1],mem[0,3],ymm12[7,5],mem[4,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm0[2,1],ymm12[2,0],ymm0[6,5],ymm12[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm12[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12 ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm12 = xmm12[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm12[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm9[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],xmm11[1],xmm10[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm12 = ymm12[3,1],mem[0,3],ymm12[7,5],mem[4,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm0[2,1],ymm12[2,0],ymm0[6,5],ymm12[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm12[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm13[0],ymm0[0],ymm13[2],ymm0[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12 ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm12 = xmm12[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm12[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm7[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],xmm9[1],xmm10[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm12 = ymm12[3,1],mem[0,3],ymm12[7,5],mem[4,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm2[2,1],ymm12[2,0],ymm2[6,5],ymm12[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm12[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm12 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm12 = ymm11[0],mem[0],ymm11[2],mem[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12 ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm12 = xmm12[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm12[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm10 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm10 = xmm10[0],mem[1],xmm10[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm12 = ymm12[3,1],mem[0,3],ymm12[7,5],mem[4,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm2[2,1],ymm12[2,0],ymm2[6,5],ymm12[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm12[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm12 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm12 = ymm2[0],mem[0],ymm2[2],mem[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12 ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm12 = xmm12[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm12[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm15[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm10 = xmm10[0],mem[1],xmm10[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm12 = ymm12[3,1],mem[0,3],ymm12[7,5],mem[4,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm6[2,1],ymm12[2,0],ymm6[6,5],ymm12[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm12[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm12 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm12 = ymm2[0],mem[0],ymm2[2],mem[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12 ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm12 = xmm12[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm12[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm10 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm10 = xmm10[0],mem[1],xmm10[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm12 = ymm12[3,1],mem[0,3],ymm12[7,5],mem[4,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm2[2,1],ymm12[2,0],ymm2[6,5],ymm12[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm12[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm12 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm12 = ymm2[0],mem[0],ymm2[2],mem[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12 ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm12 = xmm12[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm12[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm3[2,3,2,3] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],xmm4[1],xmm10[2,3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm12[3,1],ymm6[0,3],ymm12[7,5],ymm6[4,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm2[2,1],ymm12[2,0],ymm2[6,5],ymm12[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm12[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm7[0],ymm3[0],ymm7[2],ymm3[2] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12 ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm12 = xmm12[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm12[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm10[1,0],ymm12[0,0],ymm10[5,4],ymm12[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm12[3,1],ymm10[0,2],ymm12[7,5],ymm10[4,6] ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm12 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm12 = xmm14[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm12 = xmm12[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm12 ; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm12[0,1],ymm8[1,3],ymm12[4,5],ymm8[5,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[0,2],ymm15[2,0],ymm8[4,6],ymm15[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm15[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm8[1,0],ymm12[0,0],ymm8[5,4],ymm12[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm12[3,1],ymm10[0,2],ymm12[7,5],ymm10[4,6] ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm15 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm15 = xmm5[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm15[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm15[0,1],ymm10[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm5 ; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm5[0,1],ymm1[1,3],ymm5[4,5],ymm1[5,7] ; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,2],ymm15[2,0],ymm1[4,6],ymm15[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3,4],ymm15[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm5[1,0],ymm1[0,0],ymm5[5,4],ymm1[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm1[3,1],ymm10[0,2],ymm1[7,5],ymm10[4,6] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm15 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm15 = mem[0,1,2],xmm1[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm15[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm15[0,1],ymm10[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 1312(%rdi), %ymm1 ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,1],ymm0[1,3],ymm1[4,5],ymm0[5,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm13[0,2],ymm15[2,0],ymm13[4,6],ymm15[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3,4],ymm15[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm1[1,0],ymm0[0,0],ymm1[5,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm0[3,1],ymm10[0,2],ymm0[7,5],ymm10[4,6] ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm15 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm15 = xmm9[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm15[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm15[0,1],ymm10[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 1760(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm15 = ymm0[0,1],mem[1,3],ymm0[4,5],mem[5,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm11[0,2],ymm15[2,0],ymm11[4,6],ymm15[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3,4],ymm15[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm0 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm6[1,0],ymm2[0,0],ymm6[5,4],ymm2[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,1],ymm2[0,2],ymm0[7,5],ymm2[4,6] ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm9 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm9 = xmm4[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm9[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm2[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 1536(%rdi), %ymm0 ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm0[0,1],ymm3[1,3],ymm0[4,5],ymm3[5,7] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm7[0,2],ymm9[2,0],ymm7[4,6],ymm9[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm9[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[1,0],ymm0[0,0],ymm1[5,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,1],ymm2[0,2],ymm0[7,5],ymm2[4,6] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm7 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm7 = xmm15[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm7[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm2[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %ymm12 ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm12[0,1],ymm8[1,3],ymm12[4,5],ymm8[5,7] ; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm11[0,2],ymm7[2,0],ymm11[4,6],ymm7[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm7[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[1,0],ymm0[0,0],ymm1[5,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,1],ymm2[0,2],ymm0[7,5],ymm2[4,6] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm7 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm7 = xmm14[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm7[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm2[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm4 ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm4[0,1],ymm10[1,3],ymm4[4,5],ymm10[5,7] ; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm3[0,2],ymm7[2,0],ymm3[4,6],ymm7[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm7[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[1,0],ymm0[0,0],ymm1[5,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,1],ymm2[0,2],ymm0[7,5],ymm2[4,6] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm7 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm7 = xmm5[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm7[3,2,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm2[2,3,4,5,6,7] ; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm13 ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm13[0,1],ymm1[1,3],ymm13[4,5],ymm1[5,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm2[0,2],ymm9[2,0],ymm2[4,6],ymm9[6,4] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3,4],ymm9[5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm2[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[3,0],ymm7[0,0],ymm2[7,4],ymm7[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm13[1,0],ymm1[2,0],ymm13[5,4],ymm1[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm7[2,0],ymm0[6,4],ymm7[6,4] ; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm9 ; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm1[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm9[0,1,2],xmm7[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm5[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = mem[0],xmm6[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm5 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm5 = ymm7[1,0],mem[2,0],ymm7[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm5[2,0],ymm0[6,4],ymm5[6,4] ; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm1[0,1,0,1] ; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm1[0,1,2],xmm5[3] ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm6 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = mem[0],xmm6[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm3[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm3[3,0],ymm0[0,0],ymm3[7,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm4[1,0],ymm10[2,0],ymm4[5,4],ymm10[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4] ; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1] ; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm6 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm6[0,1,2],xmm1[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm14[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = mem[0],xmm5[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm1 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm1 = ymm5[1,0],mem[2,0],ymm5[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4] ; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, (%rsp) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1] ; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3] ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm14 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = mem[0],xmm14[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm14[0,1],xmm1[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm11[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm11[3,0],ymm0[0,0],ymm11[7,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm12[1,0],ymm8[2,0],ymm12[5,4],ymm8[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4] ; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1] ; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %xmm3 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm15[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = mem[0],xmm14[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm14[0,1],xmm1[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm1 = ymm8[1,0],mem[2,0],ymm8[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4] ; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1] ; AVX1-ONLY-NEXT: vmovaps 1184(%rdi), %xmm2 ; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3] ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm14 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = mem[0],xmm14[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm14[0,1],xmm1[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm1 = ymm1[1,0],mem[2,0],ymm1[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4] ; AVX1-ONLY-NEXT: vmovaps 1440(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm1[0,1,0,1] ; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %xmm1 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm1[0,1,2],xmm14[3] ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm14[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm2[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[3,0],ymm0[0,0],ymm2[7,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm4 = ymm2[1,0],mem[2,0],ymm2[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm0[2,0],ymm4[2,0],ymm0[6,4],ymm4[6,4] ; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %xmm0 ; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm0[0,1,0,1] ; AVX1-ONLY-NEXT: vmovaps 1632(%rdi), %xmm0 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm0[0,1,2],xmm14[3] ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm11 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = mem[0],xmm11[1],mem[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm11[0,1],xmm14[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm11[0,1,2,3],ymm4[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm4 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm4 = ymm13[2,1],mem[3,3],ymm13[6,5],mem[7,7] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm12[0],xmm10[1],xmm12[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm11 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm11[1,0],ymm4[2,0],ymm11[5,4],ymm4[6,4] ; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm11 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm11 = mem[0,1,2],xmm9[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm9[0,0],ymm15[1,0],ymm9[4,4],ymm15[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm14[0,1],xmm11[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm11[0,1,2,3],ymm4[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm4 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm4 = ymm7[2,1],mem[3,3],ymm7[6,5],mem[7,7] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm14[0],xmm9[1],xmm14[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm11 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm11[1,0],ymm4[2,0],ymm11[5,4],ymm4[6,4] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,2],xmm7[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm11 = ymm11[0,0],mem[1,0],ymm11[4,4],mem[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm11[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm11 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm11[0,1],xmm7[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2,3],ymm4[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm4 = ymm4[2,1],mem[3,3],ymm4[6,5],mem[7,7] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm7 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm7 = mem[0],xmm11[1],mem[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm7[1,0],ymm4[2,0],ymm7[5,4],ymm4[6,4] ; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2],xmm6[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm7 = ymm7[0,0],mem[1,0],ymm7[4,4],mem[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm7[0,1],xmm6[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm4 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm4 = ymm5[2,1],mem[3,3],ymm5[6,5],mem[7,7] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm6 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm6 = mem[0],xmm7[1],mem[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm6[1,0],ymm4[2,0],ymm6[5,4],ymm4[6,4] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $7, (%rsp), %xmm5, %xmm5 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2],xmm5[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm6 = ymm6[0,0],mem[1,0],ymm6[4,4],mem[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm6[0,1],xmm5[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm4 = ymm4[2,1],mem[3,3],ymm4[6,5],mem[7,7] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm5 = mem[0],xmm5[1],mem[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm5[1,0],ymm4[2,0],ymm5[5,4],ymm4[6,4] ; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,2],xmm3[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm5 = ymm5[0,0],mem[1,0],ymm5[4,4],mem[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm5[0,1],xmm3[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm3 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm3 = ymm8[2,1],mem[3,3],ymm8[6,5],mem[7,7] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm4 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm4 = mem[0],xmm6[1],mem[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[1,0],ymm3[2,0],ymm4[5,4],ymm3[6,4] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,2],xmm4[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm5 = ymm5[0,0],mem[1,0],ymm5[4,4],mem[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm5[0,1],xmm4[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm3 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm3 = ymm8[2,1],mem[3,3],ymm8[6,5],mem[7,7] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm4 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm4 = xmm5[0],mem[1],xmm5[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[1,0],ymm3[2,0],ymm4[5,4],ymm3[6,4] ; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2],xmm1[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm4 = ymm4[0,0],mem[1,0],ymm4[4,4],mem[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm4[0,1],xmm1[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm1 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm1 = ymm2[2,1],mem[3,3],ymm2[6,5],mem[7,7] ; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm3 = mem[0],xmm3[1],mem[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[1,0],ymm1[2,0],ymm3[5,4],ymm1[6,4] ; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm0 = mem[0,1,2],xmm0[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm3 = ymm3[0,0],mem[1,0],ymm3[4,4],mem[5,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm3[0,1],xmm0[3,2] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm13[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[3,0],ymm0[0,0],ymm13[7,4],ymm0[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm10[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm12[1],xmm1[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm1 = xmm1[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[1,0],ymm15[2,0],ymm2[5,4],ymm15[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm9[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm14[1],xmm2[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,0],ymm2[4,5],ymm1[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm3 = ymm0[1,0],mem[2,0],ymm0[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm11[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,0],ymm3[4,5],ymm1[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm3 = xmm3[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm4 = ymm0[1,0],mem[2,0],ymm0[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm1[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm3[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[3,0],ymm1[0,0],ymm3[7,4],ymm1[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm7[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,0],ymm3[4,5],ymm1[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps $8, (%rsp), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm3 = xmm3[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm4 = ymm4[1,0],mem[2,0],ymm4[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm4[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[3,0],ymm3[0,0],ymm4[7,4],ymm3[4,4] ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,0],ymm4[4,5],ymm3[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm4 = xmm4[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm9 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm9 = ymm7[1,0],mem[2,0],ymm7[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm9[0,1],xmm4[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm7[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm7[3,0],ymm4[0,0],ymm7[7,4],ymm4[4,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm6[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm9 = xmm9[0],mem[1],xmm9[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm9[0,1],ymm4[2,0],ymm9[4,5],ymm4[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm9 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm9 = xmm9[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm10 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm10 = ymm6[1,0],mem[2,0],ymm6[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm4[4,5,6,7] ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm8[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm8[3,0],ymm4[0,0],ymm8[7,4],ymm4[4,4] ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm10 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],xmm5[1],xmm10[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm10[0,1],ymm4[2,0],ymm10[4,5],ymm4[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm10 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm10 = xmm10[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm11 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm11 = ymm6[1,0],mem[2,0],ymm6[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm11[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm11 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm11[0,1],xmm10[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm10[0,1,2,3],ymm4[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm5[2,3,0,1] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm5[3,0],ymm10[0,0],ymm5[7,4],ymm10[4,4] ; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm11 = mem[2,3,2,3] ; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm11 = xmm11[0],mem[1],xmm11[2,3] ; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm11 ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,0],ymm11[4,5],ymm10[6,4] ; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm11 = mem[0,1,0,1] ; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload ; AVX1-ONLY-NEXT: # xmm11 = xmm11[0,1,2],mem[3] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm12 # 32-byte Folded Reload ; AVX1-ONLY-NEXT: # ymm12 = ymm5[1,0],mem[2,0],ymm5[5,4],mem[6,4] ; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm12[2,0,2,3,6,4,6,7] ; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12 ; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm12[0,1],xmm11[2,3] ; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7] ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm11, 224(%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm11, 160(%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm11, 96(%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm11, 32(%rsi) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rdx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rcx) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, (%r8) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%r8) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%r8) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%r8) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%r8) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%r8) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%r8) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%r8) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%r9) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%r9) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%r9) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%r9) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%r9) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%r9) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%r9) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, (%r9) ; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%rax) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rax) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%rax) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rax) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%rax) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rax) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rax) ; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rax) ; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX1-ONLY-NEXT: vmovaps %ymm10, 224(%rax) ; AVX1-ONLY-NEXT: vmovaps %ymm4, 192(%rax) ; AVX1-ONLY-NEXT: vmovaps %ymm9, 160(%rax) ; AVX1-ONLY-NEXT: vmovaps %ymm3, 128(%rax) ; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rax) ; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%rax) ; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rax) ; AVX1-ONLY-NEXT: vmovaps %ymm13, (%rax) ; AVX1-ONLY-NEXT: addq $3176, %rsp # imm = 0xC68 ; AVX1-ONLY-NEXT: vzeroupper ; AVX1-ONLY-NEXT: retq ; ; AVX2-SLOW-LABEL: load_i32_stride7_vf64: ; AVX2-SLOW: # %bb.0: ; AVX2-SLOW-NEXT: subq $2680, %rsp # imm = 0xA78 ; AVX2-SLOW-NEXT: vmovdqa 1216(%rdi), %ymm12 ; AVX2-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 1152(%rdi), %ymm4 ; AVX2-SLOW-NEXT: vmovdqa 1120(%rdi), %ymm5 ; AVX2-SLOW-NEXT: vmovdqa 768(%rdi), %ymm13 ; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 704(%rdi), %ymm6 ; AVX2-SLOW-NEXT: vmovdqa 672(%rdi), %ymm7 ; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqa 224(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = [0,7,6,u] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7] ; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm11 ; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm10 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-SLOW-NEXT: vpbroadcastq 304(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpbroadcastd 420(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm6[6],ymm7[7] ; AVX2-SLOW-NEXT: vmovdqa %ymm7, %ymm9 ; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm8 ; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-SLOW-NEXT: vpbroadcastq 752(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 800(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 832(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpbroadcastd 868(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7] ; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm6 ; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm5 ; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-SLOW-NEXT: vpbroadcastq 1200(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 1248(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 1280(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpbroadcastd 1316(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 1600(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqa 1568(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7] ; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm14 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-SLOW-NEXT: vmovdqa 1664(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpbroadcastq 1648(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 1696(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 1728(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpbroadcastd 1764(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpbroadcastq 80(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm3 ; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7] ; AVX2-SLOW-NEXT: vpermd %ymm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpbroadcastd 196(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 480(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqa 448(%rdi), %ymm13 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],ymm2[6],ymm13[7] ; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm12 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-SLOW-NEXT: vmovdqa 544(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vmovdqu %ymm3, (%rsp) # 32-byte Spill ; AVX2-SLOW-NEXT: vpbroadcastq 528(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 576(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 608(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpbroadcastd 644(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 928(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 896(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6],ymm3[7] ; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm15 ; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-SLOW-NEXT: vmovdqa 992(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpbroadcastq 976(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 1024(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa 1056(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpbroadcastd 1092(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 1376(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 1344(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7] ; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 1440(%rdi), %ymm7 ; AVX2-SLOW-NEXT: vpbroadcastq 1424(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqa 1472(%rdi), %xmm1 ; AVX2-SLOW-NEXT: vmovdqa 1504(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-SLOW-NEXT: vpbroadcastd 1540(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm3 ; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %ymm0 ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10,11],ymm1[28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3],ymm4[4,5],ymm1[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0],ymm11[1],ymm10[2,3,4],ymm11[5],ymm10[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 832(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 800(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-SLOW-NEXT: vmovdqa 736(%rdi), %ymm10 ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0],ymm9[1],ymm8[2,3,4],ymm9[5],ymm8[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 1280(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 1248(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-SLOW-NEXT: vmovdqa 1184(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm1[2,3],ymm9[4,5],ymm1[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0],ymm6[1],ymm5[2,3,4],ymm6[5],ymm5[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 1728(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 1696(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-SLOW-NEXT: vmovdqa 1632(%rdi), %ymm6 ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm6[2,3],ymm5[4,5],ymm6[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-SLOW-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm2 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm2 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 608(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 576(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-SLOW-NEXT: vmovdqa 512(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm14 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1],ymm1[2,3],ymm14[4,5],ymm1[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 1056(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 1024(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-SLOW-NEXT: vmovdqa 960(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm1[2,3],ymm8[4,5],ymm1[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-SLOW-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm2 = mem[0],ymm15[1],mem[2,3,4],ymm15[5],mem[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 1504(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 1472(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-SLOW-NEXT: vmovdqa 1408(%rdi), %ymm1 ; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3],ymm7[4,5],ymm1[6,7] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm13 ; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm12 ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm3[2,3],ymm7[4,5],ymm3[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm2 = mem[0],ymm2[1],mem[2,3,4],ymm2[5],mem[6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 304(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm4[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 232(%rdi), %xmm1 ; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-SLOW-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm1 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm1 = ymm4[0],mem[0],ymm4[2],mem[2] ; AVX2-SLOW-NEXT: vpbroadcastd 428(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 752(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm11[8,9,10,11,12,13,14,15],ymm10[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm10[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 680(%rdi), %xmm1 ; AVX2-SLOW-NEXT: vmovdqa 704(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX2-SLOW-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm1 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm1 = ymm11[0],mem[0],ymm11[2],mem[2] ; AVX2-SLOW-NEXT: vpbroadcastd 876(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 1200(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm9[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 1128(%rdi), %xmm1 ; AVX2-SLOW-NEXT: vmovdqa 1152(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-SLOW-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] ; AVX2-SLOW-NEXT: vpbroadcastd 1324(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 1648(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm5[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 1576(%rdi), %xmm1 ; AVX2-SLOW-NEXT: vmovdqa 1600(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-SLOW-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] ; AVX2-SLOW-NEXT: vpbroadcastd 1772(%rdi), %ymm6 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm6[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 80(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm7[8,9,10,11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7],ymm7[24,25,26,27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 8(%rdi), %xmm1 ; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm13[0],ymm12[0],ymm13[2],ymm12[2] ; AVX2-SLOW-NEXT: vpbroadcastd 204(%rdi), %ymm6 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm6[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 528(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm14[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm14[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 456(%rdi), %xmm1 ; AVX2-SLOW-NEXT: vmovdqa 480(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-SLOW-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] ; AVX2-SLOW-NEXT: vpbroadcastd 652(%rdi), %ymm15 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm15[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 976(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm8[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 904(%rdi), %xmm15 ; AVX2-SLOW-NEXT: vmovdqa 928(%rdi), %xmm12 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm12[1],xmm15[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm8[0],ymm7[0],ymm8[2],ymm7[2] ; AVX2-SLOW-NEXT: vpbroadcastd 1100(%rdi), %ymm14 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5,6],ymm14[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqa 1424(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm14 = ymm10[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm14[0,1,2],ymm0[3],ymm14[4,5,6,7] ; AVX2-SLOW-NEXT: vpbroadcastd 1352(%rdi), %xmm15 ; AVX2-SLOW-NEXT: vmovdqa 1376(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm0[1],xmm15[2,3] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm6[0],ymm1[0],ymm6[2],ymm1[2] ; AVX2-SLOW-NEXT: vpbroadcastd 1548(%rdi), %ymm13 ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5,6],ymm13[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm13 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm3[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm13[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm4[0,2],ymm15[1,3],ymm4[4,6],ymm15[5,7] ; AVX2-SLOW-NEXT: vbroadcastss 432(%rdi), %ymm14 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm13[5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm5 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm3[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm4 = xmm4[3,2,2,3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm5 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm5 = ymm11[0,2],mem[1,3],ymm11[4,6],mem[5,7] ; AVX2-SLOW-NEXT: vbroadcastss 880(%rdi), %ymm13 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm13[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm4 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm13[0,2],ymm14[1,3],ymm13[4,6],ymm14[5,7] ; AVX2-SLOW-NEXT: vbroadcastss 1328(%rdi), %ymm5 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm3 = ymm3[0,2],mem[1,3],ymm3[4,6],mem[5,7] ; AVX2-SLOW-NEXT: vbroadcastss 1776(%rdi), %ymm4 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0],ymm9[1],ymm10[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,2],ymm1[1,3],ymm6[4,6],ymm1[5,7] ; AVX2-SLOW-NEXT: vbroadcastss 1552(%rdi), %ymm3 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX2-SLOW-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm12[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,2],ymm7[1,3],ymm8[4,6],ymm7[5,7] ; AVX2-SLOW-NEXT: vmovaps %ymm7, %ymm9 ; AVX2-SLOW-NEXT: vmovaps %ymm8, %ymm11 ; AVX2-SLOW-NEXT: vbroadcastss 1104(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,2],ymm8[1,3],ymm6[4,6],ymm8[5,7] ; AVX2-SLOW-NEXT: vbroadcastss 656(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm4[1,3],ymm3[4,6],ymm4[5,7] ; AVX2-SLOW-NEXT: vbroadcastss 208(%rdi), %ymm2 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vbroadcastss 100(%rdi), %xmm0 ; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %xmm1 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] ; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm7 = [4,3,4,3] ; AVX2-SLOW-NEXT: # xmm7 = mem[0,0] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7] ; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm7, %ymm2 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3] ; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm0 = [0,7,0,7,0,7,0,7] ; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm0, %ymm3 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 212(%rdi), %ymm4 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7] ; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm7, %ymm3 ; AVX2-SLOW-NEXT: vbroadcastss 324(%rdi), %xmm4 ; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm2[0,1,2],xmm4[3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3] ; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm15[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 436(%rdi), %ymm5 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm7, %ymm4 ; AVX2-SLOW-NEXT: vbroadcastss 548(%rdi), %xmm5 ; AVX2-SLOW-NEXT: vmovaps 512(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm3[0,1,2],xmm5[3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3] ; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm0, %ymm5 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm8[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 660(%rdi), %ymm6 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,2,3],ymm4[4,5,6,7] ; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm7, %ymm5 ; AVX2-SLOW-NEXT: vbroadcastss 772(%rdi), %xmm6 ; AVX2-SLOW-NEXT: vmovaps 736(%rdi), %xmm4 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm4[0,1,2],xmm6[3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3] ; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm10[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 884(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm8[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,2,3],ymm5[4,5,6,7] ; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm7, %ymm6 ; AVX2-SLOW-NEXT: vbroadcastss 996(%rdi), %xmm8 ; AVX2-SLOW-NEXT: vmovaps 960(%rdi), %xmm5 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm8 = xmm5[0,1,2],xmm8[3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3] ; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm0, %ymm8 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 1108(%rdi), %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5,6,7] ; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm7, %ymm8 ; AVX2-SLOW-NEXT: vbroadcastss 1220(%rdi), %xmm9 ; AVX2-SLOW-NEXT: vmovaps 1184(%rdi), %xmm6 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm9 = xmm6[0,1,2],xmm9[3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm9[2,3] ; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm0, %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm14[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 1332(%rdi), %ymm12 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5,6],ymm12[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm8 = mem[0,1,2,3],ymm8[4,5,6,7] ; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm7, %ymm8 ; AVX2-SLOW-NEXT: vbroadcastss 1444(%rdi), %xmm9 ; AVX2-SLOW-NEXT: vmovaps 1408(%rdi), %xmm13 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm9 = xmm13[0,1,2],xmm9[3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm9[2,3] ; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm9 = ymm9[0,1,2,3,4,5],mem[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 1556(%rdi), %ymm12 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5,6],ymm12[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm8 = mem[0,1,2,3],ymm8[4,5,6,7] ; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm7, %ymm7 ; AVX2-SLOW-NEXT: vbroadcastss 1668(%rdi), %xmm8 ; AVX2-SLOW-NEXT: vmovaps 1632(%rdi), %xmm14 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm8 = xmm14[0,1,2],xmm8[3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2,3] ; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm11[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 1780(%rdi), %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-SLOW-NEXT: vbroadcastss 216(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7] ; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %xmm12 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm12[0,1,2],xmm1[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2] ; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm8 = mem[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm8, %xmm8 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %xmm1 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm2[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,1,3,2] ; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm7, %xmm7 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm7[0,1],xmm2[2,3] ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm7 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm7 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-SLOW-NEXT: vbroadcastss 440(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovaps 544(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm3[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm3 = xmm3[0,1,3,2] ; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm7, %xmm7 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm7[0,1],xmm3[2,3] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-SLOW-NEXT: vbroadcastss 664(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm7[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovaps 768(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm3[0,1,2],xmm4[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2] ; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm7, %xmm7 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3] ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm7 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm7 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-SLOW-NEXT: vbroadcastss 888(%rdi), %ymm8 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm7[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-SLOW-NEXT: vmovaps 992(%rdi), %xmm4 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm4[0,1,2],xmm5[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2] ; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm7, %xmm7 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-SLOW-NEXT: vbroadcastss 1112(%rdi), %ymm10 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm10[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm5[0,1,2,3],ymm7[4,5,6,7] ; AVX2-SLOW-NEXT: vmovaps 1216(%rdi), %xmm15 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm15[0,1,2],xmm6[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2] ; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm6 = mem[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm6, %xmm6 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm6 = ymm6[0],mem[1],ymm6[2,3,4],mem[5],ymm6[6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm6[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3] ; AVX2-SLOW-NEXT: vbroadcastss 1336(%rdi), %ymm7 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm5[0,1,2,3],ymm6[4,5,6,7] ; AVX2-SLOW-NEXT: vmovaps 1440(%rdi), %xmm5 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm5[0,1,2],xmm13[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,3,2] ; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm7, %xmm7 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-SLOW-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm7 = mem[0],ymm7[1],mem[2,3,4],ymm7[5],mem[6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-SLOW-NEXT: vbroadcastss 1560(%rdi), %ymm13 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm13[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm6[0,1,2,3],ymm7[4,5,6,7] ; AVX2-SLOW-NEXT: vmovaps 1664(%rdi), %xmm6 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm13 = xmm6[0,1,2],xmm14[3] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm13 = xmm13[0,1,3,2] ; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm14 = mem[1,0,2,3,5,4,6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm14, %xmm14 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm13 = xmm14[0,1],xmm13[2,3] ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm14 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm14 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7] ; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm14 = ymm14[1,0,3,3,5,4,7,7] ; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm14 = ymm14[0,1,0,3] ; AVX2-SLOW-NEXT: vbroadcastss 1784(%rdi), %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm14[0,1,2,3,4,5,6],ymm9[7] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm9[4,5,6,7] ; AVX2-SLOW-NEXT: vbroadcastss 136(%rdi), %xmm9 ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm9 = xmm9[0],mem[1],xmm9[2,3] ; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9 ; AVX2-SLOW-NEXT: vpermps 192(%rdi), %ymm0, %ymm14 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm14[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 80(%rdi), %ymm14 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm12 = xmm14[0,1,2],xmm12[3] ; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm14 = mem[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm14 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm14, %xmm14 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm12 = xmm14[0,1],xmm12[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm9[4,5,6,7] ; AVX2-SLOW-NEXT: vbroadcastss 360(%rdi), %xmm9 ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm9 = xmm9[0],mem[1],xmm9[2,3] ; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9 ; AVX2-SLOW-NEXT: vpermps 416(%rdi), %ymm0, %ymm14 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm14[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 304(%rdi), %ymm14 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm14[0,1,2],xmm1[3] ; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm14 = mem[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm14 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm14, %xmm14 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm14[0,1],xmm1[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm9[4,5,6,7] ; AVX2-SLOW-NEXT: vbroadcastss 584(%rdi), %xmm1 ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3] ; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-SLOW-NEXT: vpermps 640(%rdi), %ymm0, %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm9[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 528(%rdi), %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm9[0,1,2],xmm2[3] ; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm9, %xmm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm9[0,1],xmm2[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7] ; AVX2-SLOW-NEXT: vbroadcastss 808(%rdi), %xmm2 ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm2 = xmm2[0],mem[1],xmm2[2,3] ; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-SLOW-NEXT: vpermps 864(%rdi), %ymm0, %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm9[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 752(%rdi), %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm9[0,1,2],xmm3[3] ; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm9, %xmm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm9[0,1],xmm3[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7] ; AVX2-SLOW-NEXT: vbroadcastss 1032(%rdi), %xmm3 ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3] ; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-SLOW-NEXT: vpermps 1088(%rdi), %ymm0, %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm9[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 976(%rdi), %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm9[0,1,2],xmm4[3] ; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm9, %xmm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm9[0,1],xmm4[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7] ; AVX2-SLOW-NEXT: vbroadcastss 1256(%rdi), %xmm4 ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3] ; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4 ; AVX2-SLOW-NEXT: vpermps 1312(%rdi), %ymm0, %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm9[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 1200(%rdi), %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm9 = xmm9[0,1,2],xmm15[3] ; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm15, %xmm15 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm9 = xmm15[0,1],xmm9[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm9[0,1,2,3],ymm4[4,5,6,7] ; AVX2-SLOW-NEXT: vbroadcastss 1480(%rdi), %xmm9 ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm9 = xmm9[0],mem[1],xmm9[2,3] ; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9 ; AVX2-SLOW-NEXT: vpermps 1536(%rdi), %ymm0, %ymm15 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm15[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 1424(%rdi), %ymm15 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm15[0,1,2],xmm5[3] ; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm15, %xmm15 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm15[0,1],xmm5[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm9[4,5,6,7] ; AVX2-SLOW-NEXT: vbroadcastss 1704(%rdi), %xmm9 ; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload ; AVX2-SLOW-NEXT: # xmm9 = xmm9[0],mem[1],xmm9[2,3] ; AVX2-SLOW-NEXT: vpermps 1760(%rdi), %ymm0, %ymm0 ; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3,4,5],ymm0[6,7] ; AVX2-SLOW-NEXT: vbroadcastss 1648(%rdi), %ymm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1,2],xmm6[3] ; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7] ; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload ; AVX2-SLOW-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7] ; AVX2-SLOW-NEXT: vextractf128 $1, %ymm9, %xmm9 ; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1],xmm6[2,3] ; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7] ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 192(%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 128(%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 64(%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, (%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 224(%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 160(%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 96(%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%rsi) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 192(%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 128(%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 64(%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, (%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 224(%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 160(%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 96(%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%rdx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 192(%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 128(%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 64(%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, (%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 224(%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 160(%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 96(%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%rcx) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, (%r8) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 64(%r8) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 128(%r8) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 192(%r8) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 224(%r8) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 160(%r8) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 96(%r8) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%r8) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 224(%r9) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 192(%r9) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 160(%r9) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 128(%r9) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 96(%r9) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 64(%r9) ; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%r9) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, (%r9) ; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-SLOW-NEXT: vmovaps %ymm13, 224(%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm7, 192(%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm8, 160(%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm10, 128(%rax) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 96(%rax) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 64(%rax) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%rax) ; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-SLOW-NEXT: vmovaps %ymm6, (%rax) ; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-SLOW-NEXT: vmovaps %ymm0, 224(%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm5, 192(%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm4, 160(%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm3, 128(%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm2, 96(%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm1, 64(%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm14, 32(%rax) ; AVX2-SLOW-NEXT: vmovaps %ymm12, (%rax) ; AVX2-SLOW-NEXT: addq $2680, %rsp # imm = 0xA78 ; AVX2-SLOW-NEXT: vzeroupper ; AVX2-SLOW-NEXT: retq ; ; AVX2-FAST-LABEL: load_i32_stride7_vf64: ; AVX2-FAST: # %bb.0: ; AVX2-FAST-NEXT: subq $2680, %rsp # imm = 0xA78 ; AVX2-FAST-NEXT: vmovdqa 1216(%rdi), %ymm13 ; AVX2-FAST-NEXT: vmovdqa 1152(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqa 1120(%rdi), %ymm5 ; AVX2-FAST-NEXT: vmovdqa 768(%rdi), %ymm8 ; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 704(%rdi), %ymm6 ; AVX2-FAST-NEXT: vmovdqa 672(%rdi), %ymm7 ; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm12 ; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm2 ; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm10 ; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = [0,7,6,u] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5],ymm2[6],ymm10[7] ; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm11 ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vpbroadcastq 304(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %xmm3 ; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpbroadcastd 420(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm6[6],ymm7[7] ; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa %ymm6, %ymm9 ; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vpbroadcastq 752(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 800(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa 832(%rdi), %xmm3 ; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpbroadcastd 868(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7] ; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm6 ; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm8 ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vpbroadcastq 1200(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm13, %ymm15 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 1248(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa 1280(%rdi), %xmm3 ; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpbroadcastd 1316(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 1600(%rdi), %ymm2 ; AVX2-FAST-NEXT: vmovdqa 1568(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7] ; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm13 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm5 ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vmovdqa 1664(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpbroadcastq 1648(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 1696(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa 1728(%rdi), %xmm3 ; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpbroadcastd 1764(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpbroadcastq 80(%rdi), %ymm1 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm14 ; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm2 ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7] ; AVX2-FAST-NEXT: vpermd %ymm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %xmm3 ; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpbroadcastd 196(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 480(%rdi), %ymm1 ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 448(%rdi), %ymm2 ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6],ymm2[7] ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vmovdqa 544(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpbroadcastq 528(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 576(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa 608(%rdi), %xmm3 ; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpbroadcastd 644(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 928(%rdi), %ymm2 ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 896(%rdi), %ymm1 ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7] ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vmovdqa 992(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpbroadcastq 976(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 1024(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa 1056(%rdi), %xmm3 ; AVX2-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpbroadcastd 1092(%rdi), %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 1376(%rdi), %ymm2 ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 1344(%rdi), %ymm1 ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7] ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm0 ; AVX2-FAST-NEXT: vmovdqa 1440(%rdi), %ymm2 ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpbroadcastq 1424(%rdi), %ymm1 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa 1472(%rdi), %xmm1 ; AVX2-FAST-NEXT: vmovdqa 1504(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vpbroadcastd 1540(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %ymm2 ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm1 ; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm0 ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1],ymm0[2,3],ymm12[4,5],ymm0[6,7] ; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm0 = [5,6,5,6,5,6,5,6] ; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm3 ; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [1,0,7,7,5,4,7,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm11[0],ymm10[1],ymm11[2,3,4],ymm10[5],ymm11[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vmovdqa 832(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 800(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm4[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vmovdqa 736(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1],ymm3[2,3],ymm11[4,5],ymm3[6,7] ; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0],ymm7[1],ymm9[2,3,4],ymm7[5],ymm9[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vmovdqa 1280(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 1248(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm4[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vmovdqa 1184(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa %ymm15, %ymm10 ; AVX2-FAST-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1],ymm3[2,3],ymm15[4,5],ymm3[6,7] ; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0],ymm6[1],ymm8[2,3,4],ymm6[5],ymm8[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vmovdqa 1728(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 1696(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm4[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vmovdqa 1632(%rdi), %ymm15 ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1],ymm15[2,3],ymm6[4,5],ymm15[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm3 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0],ymm13[1],ymm5[2,3,4],ymm13[5],ymm5[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm4[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqa %ymm14, %ymm8 ; AVX2-FAST-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0,1],ymm4[2,3],ymm14[4,5],ymm4[6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm14 ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm3 ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm4 = mem[0],ymm4[1],mem[2,3,4],ymm4[5],mem[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vmovdqa 608(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 576(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, (%rsp) # 32-byte Spill ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm4[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %ymm13 ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm13[2,3],ymm9[4,5],ymm13[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm3 ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm4 = mem[0],ymm4[1],mem[2,3,4],ymm4[5],mem[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vmovdqa 1056(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 1024(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm4[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vmovdqa 960(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm3[2,3],ymm7[4,5],ymm3[6,7] ; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm3 ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2] ; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vmovdqa 1504(%rdi), %ymm4 ; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 1472(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm4[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vmovdqa 1408(%rdi), %ymm3 ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3],ymm5[4,5],ymm3[6,7] ; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm0 ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm1 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 304(%rdi), %xmm0 ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm12[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm12[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 232(%rdi), %xmm1 ; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm12[0],ymm4[0],ymm12[2],ymm4[2] ; AVX2-FAST-NEXT: vpbroadcastd 428(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 752(%rdi), %xmm0 ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm11[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 680(%rdi), %xmm1 ; AVX2-FAST-NEXT: vmovdqa 704(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[2],ymm3[2] ; AVX2-FAST-NEXT: vpbroadcastd 876(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 1200(%rdi), %xmm0 ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm10[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 1128(%rdi), %xmm1 ; AVX2-FAST-NEXT: vmovdqa 1152(%rdi), %xmm2 ; AVX2-FAST-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] ; AVX2-FAST-NEXT: vpbroadcastd 1324(%rdi), %ymm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 1648(%rdi), %xmm0 ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm6[8,9,10,11,12,13,14,15],ymm15[0,1,2,3,4,5,6,7],ymm6[24,25,26,27,28,29,30,31],ymm15[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 1576(%rdi), %xmm1 ; AVX2-FAST-NEXT: vmovdqa 1600(%rdi), %xmm2 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] ; AVX2-FAST-NEXT: vpbroadcastd 1772(%rdi), %ymm6 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm6[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 80(%rdi), %xmm0 ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm8[8,9,10,11,12,13,14,15],ymm14[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm14[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 8(%rdi), %xmm1 ; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %xmm6 ; AVX2-FAST-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm6[1],xmm1[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] ; AVX2-FAST-NEXT: vpbroadcastd 204(%rdi), %ymm6 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm6[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 528(%rdi), %xmm0 ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm9[8,9,10,11,12,13,14,15],ymm13[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm13[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 456(%rdi), %xmm1 ; AVX2-FAST-NEXT: vmovdqa 480(%rdi), %xmm6 ; AVX2-FAST-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm6[1],xmm1[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu (%rsp), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] ; AVX2-FAST-NEXT: vpbroadcastd 652(%rdi), %ymm15 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm15[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 976(%rdi), %xmm0 ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm7[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm7[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 904(%rdi), %xmm15 ; AVX2-FAST-NEXT: vmovdqa 928(%rdi), %xmm11 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm11[1],xmm15[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload ; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm10[0],ymm7[0],ymm10[2],ymm7[2] ; AVX2-FAST-NEXT: vpbroadcastd 1100(%rdi), %ymm13 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5,6],ymm13[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm13[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqa 1424(%rdi), %xmm0 ; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm9 ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm13 = ymm5[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2],ymm0[3],ymm13[4,5,6,7] ; AVX2-FAST-NEXT: vpbroadcastd 1352(%rdi), %xmm15 ; AVX2-FAST-NEXT: vmovdqa 1376(%rdi), %xmm0 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm0[1],xmm15[2,3] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1],ymm13[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm6[0],ymm1[0],ymm6[2],ymm1[2] ; AVX2-FAST-NEXT: vpbroadcastd 1548(%rdi), %ymm14 ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5,6],ymm14[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm13 = ymm5[0],mem[1],ymm5[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1,2],mem[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm13[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm13 = ymm12[0,2],ymm4[1,3],ymm12[4,6],ymm4[5,7] ; AVX2-FAST-NEXT: vbroadcastss 432(%rdi), %ymm14 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm13[5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm5 = ymm4[0],mem[1],ymm4[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1,2],mem[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm4 = xmm4[3,2,2,3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm15[0,2],ymm3[1,3],ymm15[4,6],ymm3[5,7] ; AVX2-FAST-NEXT: vmovaps %ymm3, %ymm13 ; AVX2-FAST-NEXT: vbroadcastss 880(%rdi), %ymm12 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm12[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm4 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm14[0,2],ymm12[1,3],ymm14[4,6],ymm12[5,7] ; AVX2-FAST-NEXT: vbroadcastss 1328(%rdi), %ymm5 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],mem[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm3 = ymm3[0,2],mem[1,3],ymm3[4,6],mem[5,7] ; AVX2-FAST-NEXT: vbroadcastss 1776(%rdi), %ymm4 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0],ymm8[1],ymm9[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],mem[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,2],ymm1[1,3],ymm6[4,6],ymm1[5,7] ; AVX2-FAST-NEXT: vbroadcastss 1552(%rdi), %ymm3 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm11[0,1,2],mem[3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovdqa %ymm10, %ymm8 ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,2],ymm7[1,3],ymm10[4,6],ymm7[5,7] ; AVX2-FAST-NEXT: vmovaps %ymm7, %ymm10 ; AVX2-FAST-NEXT: vbroadcastss 1104(%rdi), %ymm2 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovups (%rsp), %ymm6 # 32-byte Reload ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,2],ymm7[1,3],ymm6[4,6],ymm7[5,7] ; AVX2-FAST-NEXT: vbroadcastss 656(%rdi), %ymm2 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm5[1,3],ymm3[4,6],ymm5[5,7] ; AVX2-FAST-NEXT: vbroadcastss 208(%rdi), %ymm2 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vbroadcastss 100(%rdi), %xmm0 ; AVX2-FAST-NEXT: vmovaps 64(%rdi), %xmm9 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm0 = xmm9[0,1,2],xmm0[3] ; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm4 = [4,3,4,3] ; AVX2-FAST-NEXT: # xmm4 = mem[0,0] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm1 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3] ; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm0 = [0,7,0,7,0,7,0,7] ; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm5[6,7] ; AVX2-FAST-NEXT: vbroadcastss 212(%rdi), %ymm3 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm1 ; AVX2-FAST-NEXT: vbroadcastss 324(%rdi), %xmm3 ; AVX2-FAST-NEXT: vmovaps 288(%rdi), %xmm2 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm3[3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3] ; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload ; AVX2-FAST-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm3 = ymm3[0,1,2,3,4,5],mem[6,7] ; AVX2-FAST-NEXT: vbroadcastss 436(%rdi), %ymm5 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm5[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm1 ; AVX2-FAST-NEXT: vbroadcastss 548(%rdi), %xmm5 ; AVX2-FAST-NEXT: vmovaps 512(%rdi), %xmm3 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm3[0,1,2],xmm5[3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3] ; AVX2-FAST-NEXT: vpermps %ymm6, %ymm0, %ymm5 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm7[6,7] ; AVX2-FAST-NEXT: vbroadcastss 660(%rdi), %ymm6 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm1 ; AVX2-FAST-NEXT: vbroadcastss 772(%rdi), %xmm6 ; AVX2-FAST-NEXT: vmovaps 736(%rdi), %xmm5 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm6 = xmm5[0,1,2],xmm6[3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm6[2,3] ; AVX2-FAST-NEXT: vpermps %ymm15, %ymm0, %ymm6 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm13[6,7] ; AVX2-FAST-NEXT: vbroadcastss 884(%rdi), %ymm7 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7] ; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm1 ; AVX2-FAST-NEXT: vbroadcastss 996(%rdi), %xmm7 ; AVX2-FAST-NEXT: vmovaps 960(%rdi), %xmm6 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm7 = xmm6[0,1,2],xmm7[3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2,3] ; AVX2-FAST-NEXT: vpermps %ymm8, %ymm0, %ymm7 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm10[6,7] ; AVX2-FAST-NEXT: vbroadcastss 1108(%rdi), %ymm8 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4,5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm1 ; AVX2-FAST-NEXT: vbroadcastss 1220(%rdi), %xmm8 ; AVX2-FAST-NEXT: vmovaps 1184(%rdi), %xmm7 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm8 = xmm7[0,1,2],xmm8[3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm8[2,3] ; AVX2-FAST-NEXT: vpermps %ymm14, %ymm0, %ymm8 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm12[6,7] ; AVX2-FAST-NEXT: vmovaps %ymm12, %ymm15 ; AVX2-FAST-NEXT: vbroadcastss 1332(%rdi), %ymm10 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm10[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm1 ; AVX2-FAST-NEXT: vbroadcastss 1444(%rdi), %xmm10 ; AVX2-FAST-NEXT: vmovaps 1408(%rdi), %xmm8 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm10 = xmm8[0,1,2],xmm10[3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm10[2,3] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload ; AVX2-FAST-NEXT: vpermps %ymm14, %ymm0, %ymm10 ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm13[6,7] ; AVX2-FAST-NEXT: vbroadcastss 1556(%rdi), %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,6],ymm11[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vpermps %ymm1, %ymm4, %ymm1 ; AVX2-FAST-NEXT: vbroadcastss 1668(%rdi), %xmm4 ; AVX2-FAST-NEXT: vmovaps 1632(%rdi), %xmm12 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm4 = xmm12[0,1,2],xmm4[3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3] ; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm4 = ymm4[0,1,2,3,4,5],mem[6,7] ; AVX2-FAST-NEXT: vbroadcastss 1780(%rdi), %ymm10 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm10[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm4 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7] ; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm1 = [1,0,3,3,1,0,7,7] ; AVX2-FAST-NEXT: vpermps %ymm4, %ymm1, %ymm4 ; AVX2-FAST-NEXT: vbroadcastss 216(%rdi), %ymm10 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm10[7] ; AVX2-FAST-NEXT: vmovaps 96(%rdi), %xmm10 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1,2],xmm9[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm9 = xmm9[0,1,3,2] ; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm11, %xmm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm9 = xmm11[0,1],xmm9[2,3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm9[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovaps 320(%rdi), %xmm4 ; AVX2-FAST-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,1,3,2] ; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm4 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm4, %xmm4 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm4 = mem[0],ymm4[1],mem[2,3,4],ymm4[5],mem[6,7] ; AVX2-FAST-NEXT: vpermps %ymm4, %ymm1, %ymm4 ; AVX2-FAST-NEXT: vbroadcastss 440(%rdi), %ymm9 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm9[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovaps 544(%rdi), %xmm2 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm3[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm3 = xmm3[0,1,3,2] ; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm4 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm4, %xmm4 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $34, (%rsp), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7] ; AVX2-FAST-NEXT: vpermps %ymm4, %ymm1, %ymm4 ; AVX2-FAST-NEXT: vbroadcastss 664(%rdi), %ymm9 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm9[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill ; AVX2-FAST-NEXT: vmovaps 768(%rdi), %xmm4 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1,2],xmm5[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm3 = xmm3[0,1,3,2] ; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm5, %xmm5 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7] ; AVX2-FAST-NEXT: vpermps %ymm5, %ymm1, %ymm5 ; AVX2-FAST-NEXT: vbroadcastss 888(%rdi), %ymm9 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm9[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7] ; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-NEXT: vmovaps 992(%rdi), %xmm9 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm9[0,1,2],xmm6[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm3 = xmm3[0,1,3,2] ; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm5, %xmm5 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7] ; AVX2-FAST-NEXT: vpermps %ymm5, %ymm1, %ymm5 ; AVX2-FAST-NEXT: vbroadcastss 1112(%rdi), %ymm6 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm3[0,1,2,3],ymm5[4,5,6,7] ; AVX2-FAST-NEXT: vmovaps 1216(%rdi), %xmm3 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm3[0,1,2],xmm7[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2] ; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm7, %xmm7 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3] ; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm7 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm7 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7] ; AVX2-FAST-NEXT: vpermps %ymm7, %ymm1, %ymm7 ; AVX2-FAST-NEXT: vbroadcastss 1336(%rdi), %ymm15 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm15[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm5[0,1,2,3],ymm7[4,5,6,7] ; AVX2-FAST-NEXT: vmovaps 1440(%rdi), %xmm5 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm8 = xmm5[0,1,2],xmm8[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm8 = xmm8[0,1,3,2] ; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm15 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm15, %xmm15 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm8 = xmm15[0,1],xmm8[2,3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm15 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7] ; AVX2-FAST-NEXT: vpermps %ymm15, %ymm1, %ymm15 ; AVX2-FAST-NEXT: vbroadcastss 1560(%rdi), %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm15[0,1,2,3,4,5,6],ymm11[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm11[4,5,6,7] ; AVX2-FAST-NEXT: vmovaps 1664(%rdi), %xmm15 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm11 = xmm15[0,1,2],xmm12[3] ; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm11 = xmm11[0,1,3,2] ; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm12 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm12, %xmm12 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm11 = xmm12[0,1],xmm11[2,3] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload ; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7] ; AVX2-FAST-NEXT: vpermps %ymm12, %ymm1, %ymm1 ; AVX2-FAST-NEXT: vbroadcastss 1784(%rdi), %ymm12 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm12[7] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm11[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vbroadcastss 136(%rdi), %xmm1 ; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3] ; AVX2-FAST-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vpermps 192(%rdi), %ymm0, %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm11[6,7] ; AVX2-FAST-NEXT: vbroadcastss 80(%rdi), %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm10 = xmm11[0,1,2],xmm10[3] ; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm11, %xmm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm10 = xmm11[0,1],xmm10[2,3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vbroadcastss 360(%rdi), %xmm1 ; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3] ; AVX2-FAST-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vpermps 416(%rdi), %ymm0, %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm11[6,7] ; AVX2-FAST-NEXT: vbroadcastss 304(%rdi), %ymm11 ; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm11 = xmm11[0,1,2],mem[3] ; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm13 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm13 = ymm13[0],mem[1],ymm13[2,3,4],mem[5],ymm13[6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm13, %xmm13 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm11 = xmm13[0,1],xmm11[2,3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm11[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vbroadcastss 584(%rdi), %xmm1 ; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3] ; AVX2-FAST-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-FAST-NEXT: vpermps 640(%rdi), %ymm0, %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm11[6,7] ; AVX2-FAST-NEXT: vbroadcastss 528(%rdi), %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm11[0,1,2],xmm2[3] ; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm11, %xmm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm11[0,1],xmm2[2,3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-NEXT: vbroadcastss 808(%rdi), %xmm2 ; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm2 = xmm2[0],mem[1],xmm2[2,3] ; AVX2-FAST-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpermps 864(%rdi), %ymm0, %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm11[6,7] ; AVX2-FAST-NEXT: vbroadcastss 752(%rdi), %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm4 = xmm11[0,1,2],xmm4[3] ; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm11, %xmm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm4 = xmm11[0,1],xmm4[2,3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-NEXT: vbroadcastss 1032(%rdi), %xmm2 ; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm2 = xmm2[0],mem[1],xmm2[2,3] ; AVX2-FAST-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpermps 1088(%rdi), %ymm0, %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm11[6,7] ; AVX2-FAST-NEXT: vbroadcastss 976(%rdi), %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm11 = xmm11[0,1,2],xmm9[3] ; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm14 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm14 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm14, %xmm14 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm11 = xmm14[0,1],xmm11[2,3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm11[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-NEXT: vbroadcastss 1256(%rdi), %xmm2 ; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm2 = xmm2[0],mem[1],xmm2[2,3] ; AVX2-FAST-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-NEXT: vpermps 1312(%rdi), %ymm0, %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm11[6,7] ; AVX2-FAST-NEXT: vbroadcastss 1200(%rdi), %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm11[0,1,2],xmm3[3] ; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm11, %xmm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm11[0,1],xmm3[2,3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-NEXT: vbroadcastss 1480(%rdi), %xmm3 ; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3] ; AVX2-FAST-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-FAST-NEXT: vpermps 1536(%rdi), %ymm0, %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7] ; AVX2-FAST-NEXT: vbroadcastss 1424(%rdi), %ymm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm11[0,1,2],xmm5[3] ; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm11, %xmm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm11[0,1],xmm5[2,3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-NEXT: vbroadcastss 1704(%rdi), %xmm5 ; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload ; AVX2-FAST-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3] ; AVX2-FAST-NEXT: vpermps 1760(%rdi), %ymm0, %ymm0 ; AVX2-FAST-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5 ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7] ; AVX2-FAST-NEXT: vbroadcastss 1648(%rdi), %ymm5 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1,2],xmm15[3] ; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload ; AVX2-FAST-NEXT: # ymm11 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7] ; AVX2-FAST-NEXT: vextractf128 $1, %ymm11, %xmm11 ; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm11[0,1],xmm5[2,3] ; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7] ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 192(%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 128(%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, (%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 224(%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 160(%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%rsi) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 192(%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 128(%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, (%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 224(%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 160(%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%rdx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 192(%rcx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 128(%rcx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%rcx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, (%rcx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 224(%rcx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 160(%rcx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%rcx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%rcx) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, (%r8) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%r8) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 128(%r8) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 192(%r8) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 224(%r8) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 160(%r8) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%r8) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%r8) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 224(%r9) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 192(%r9) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 160(%r9) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 128(%r9) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%r9) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%r9) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%r9) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, (%r9) ; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-NEXT: vmovaps %ymm12, 224(%rax) ; AVX2-FAST-NEXT: vmovaps %ymm8, 192(%rax) ; AVX2-FAST-NEXT: vmovaps %ymm7, 160(%rax) ; AVX2-FAST-NEXT: vmovaps %ymm6, 128(%rax) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%rax) ; AVX2-FAST-NEXT: vmovups (%rsp), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%rax) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%rax) ; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-NEXT: vmovaps %ymm5, (%rax) ; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-NEXT: vmovaps %ymm0, 224(%rax) ; AVX2-FAST-NEXT: vmovaps %ymm3, 192(%rax) ; AVX2-FAST-NEXT: vmovaps %ymm2, 160(%rax) ; AVX2-FAST-NEXT: vmovaps %ymm14, 128(%rax) ; AVX2-FAST-NEXT: vmovaps %ymm4, 96(%rax) ; AVX2-FAST-NEXT: vmovaps %ymm1, 64(%rax) ; AVX2-FAST-NEXT: vmovaps %ymm13, 32(%rax) ; AVX2-FAST-NEXT: vmovaps %ymm10, (%rax) ; AVX2-FAST-NEXT: addq $2680, %rsp # imm = 0xA78 ; AVX2-FAST-NEXT: vzeroupper ; AVX2-FAST-NEXT: retq ; ; AVX2-FAST-PERLANE-LABEL: load_i32_stride7_vf64: ; AVX2-FAST-PERLANE: # %bb.0: ; AVX2-FAST-PERLANE-NEXT: subq $2680, %rsp # imm = 0xA78 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1216(%rdi), %ymm12 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1152(%rdi), %ymm4 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1120(%rdi), %ymm5 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 768(%rdi), %ymm13 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 704(%rdi), %ymm6 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 672(%rdi), %ymm7 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm0 = [0,7,6,u] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm11 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm10 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 304(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 420(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm6[6],ymm7[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, %ymm9 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, %ymm8 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 752(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 800(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 832(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 868(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, %ymm6 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm5 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 1200(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1248(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1280(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1316(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1600(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1568(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm14 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1664(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 1648(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1696(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1728(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1764(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 80(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 196(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 480(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 448(%rdi), %ymm13 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],ymm2[6],ymm13[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm12 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 544(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, (%rsp) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 528(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 576(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 608(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 644(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 928(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 896(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm15 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 992(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 976(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1024(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1056(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1092(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1376(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1344(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7] ; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1440(%rdi), %ymm7 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 1424(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1472(%rdi), %xmm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1504(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1540(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[12,13,14,15],ymm0[0,1,2,3,4,5,6,7,8,9,10,11],ymm1[28,29,30,31],ymm0[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3],ymm4[4,5],ymm1[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0],ymm11[1],ymm10[2,3,4],ymm11[5],ymm10[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 832(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 800(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 736(%rdi), %ymm10 ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0],ymm9[1],ymm8[2,3,4],ymm9[5],ymm8[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1280(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1248(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1184(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm1[2,3],ymm9[4,5],ymm1[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0],ymm6[1],ymm5[2,3,4],ymm6[5],ymm5[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1728(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1696(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1632(%rdi), %ymm6 ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm6[2,3],ymm5[4,5],ymm6[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm2 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 608(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 576(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 512(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm14 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1],ymm1[2,3],ymm14[4,5],ymm1[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1056(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1024(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 960(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm1[2,3],ymm8[4,5],ymm1[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0],ymm15[1],mem[2,3,4],ymm15[5],mem[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1504(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1472(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1408(%rdi), %ymm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3],ymm7[4,5],ymm1[6,7] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27] ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm13 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm12 ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7] ; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm3[2,3],ymm7[4,5],ymm3[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0],ymm2[1],mem[2,3,4],ymm2[5],mem[6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 304(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm4[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 232(%rdi), %xmm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm4[0],mem[0],ymm4[2],mem[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 428(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 752(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm11[8,9,10,11,12,13,14,15],ymm10[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm10[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 680(%rdi), %xmm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 704(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm11[0],mem[0],ymm11[2],mem[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 876(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1200(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm9[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1128(%rdi), %xmm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1152(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1324(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1648(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm5[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1576(%rdi), %xmm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1600(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1772(%rdi), %ymm6 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm6[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 80(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm7[8,9,10,11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7],ymm7[24,25,26,27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 8(%rdi), %xmm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm13[0],ymm12[0],ymm13[2],ymm12[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 204(%rdi), %ymm6 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm6[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 528(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm14[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm14[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 456(%rdi), %xmm1 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 480(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 652(%rdi), %ymm15 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm15[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 976(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm8[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 904(%rdi), %xmm15 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 928(%rdi), %xmm12 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm12[1],xmm15[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm8[0],ymm7[0],ymm8[2],ymm7[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1100(%rdi), %ymm14 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5,6],ymm14[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1424(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm14 = ymm10[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm14 = ymm14[0,1,2],ymm0[3],ymm14[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1352(%rdi), %xmm15 ; AVX2-FAST-PERLANE-NEXT: vmovdqa 1376(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm0[1],xmm15[2,3] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm6[0],ymm1[0],ymm6[2],ymm1[2] ; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1548(%rdi), %ymm13 ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5,6],ymm13[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm13 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm3[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm13[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm4[0,2],ymm15[1,3],ymm4[4,6],ymm15[5,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 432(%rdi), %ymm14 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm13[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm3[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm4 = xmm4[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm5 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm11[0,2],mem[1,3],ymm11[4,6],mem[5,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 880(%rdi), %ymm13 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm13[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm13[0,2],ymm14[1,3],ymm13[4,6],ymm14[5,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1328(%rdi), %ymm5 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm3[0,2],mem[1,3],ymm3[4,6],mem[5,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1776(%rdi), %ymm4 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0],ymm9[1],ymm10[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,2],ymm1[1,3],ymm6[4,6],ymm1[5,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1552(%rdi), %ymm3 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm12[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,2],ymm7[1,3],ymm8[4,6],ymm7[5,7] ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm7, %ymm9 ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, %ymm11 ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1104(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,2],ymm8[1,3],ymm6[4,6],ymm8[5,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 656(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm4[1,3],ymm3[4,6],ymm4[5,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 208(%rdi), %ymm2 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 100(%rdi), %xmm0 ; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %xmm1 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] ; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm7 = [4,3,4,3] ; AVX2-FAST-PERLANE-NEXT: # xmm7 = mem[0,0] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm7, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3] ; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm0 = [0,7,0,7,0,7,0,7] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm0, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 212(%rdi), %ymm4 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm7, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 324(%rdi), %xmm4 ; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm2[0,1,2],xmm4[3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3] ; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm15[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 436(%rdi), %ymm5 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm7, %ymm4 ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 548(%rdi), %xmm5 ; AVX2-FAST-PERLANE-NEXT: vmovaps 512(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm3[0,1,2],xmm5[3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm0, %ymm5 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm8[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 660(%rdi), %ymm6 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm7, %ymm5 ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 772(%rdi), %xmm6 ; AVX2-FAST-PERLANE-NEXT: vmovaps 736(%rdi), %xmm4 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm4[0,1,2],xmm6[3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3] ; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm10[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 884(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm8[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1,2,3],ymm5[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm7, %ymm6 ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 996(%rdi), %xmm8 ; AVX2-FAST-PERLANE-NEXT: vmovaps 960(%rdi), %xmm5 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm8 = xmm5[0,1,2],xmm8[3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm0, %ymm8 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1108(%rdi), %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm7, %ymm8 ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1220(%rdi), %xmm9 ; AVX2-FAST-PERLANE-NEXT: vmovaps 1184(%rdi), %xmm6 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm9 = xmm6[0,1,2],xmm9[3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm9[2,3] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm0, %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm14[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1332(%rdi), %ymm12 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5,6],ymm12[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,1,2,3],ymm8[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm7, %ymm8 ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1444(%rdi), %xmm9 ; AVX2-FAST-PERLANE-NEXT: vmovaps 1408(%rdi), %xmm13 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm9 = xmm13[0,1,2],xmm9[3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm9[2,3] ; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm9[0,1,2,3,4,5],mem[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1556(%rdi), %ymm12 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5,6],ymm12[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,1,2,3],ymm8[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm7, %ymm7 ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1668(%rdi), %xmm8 ; AVX2-FAST-PERLANE-NEXT: vmovaps 1632(%rdi), %xmm14 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm8 = xmm14[0,1,2],xmm8[3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2,3] ; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm11[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1780(%rdi), %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 216(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7] ; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %xmm12 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm12[0,1,2],xmm1[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm8, %xmm8 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %xmm1 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm2[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm7, %xmm7 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm7[0,1],xmm2[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm7 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 440(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovaps 544(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm3[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm3 = xmm3[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm7, %xmm7 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm7[0,1],xmm3[2,3] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 664(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm7[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovaps 768(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm3[0,1,2],xmm4[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm7, %xmm7 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm7 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 888(%rdi), %ymm8 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm7[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill ; AVX2-FAST-PERLANE-NEXT: vmovaps 992(%rdi), %xmm4 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm4[0,1,2],xmm5[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm7, %xmm7 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1112(%rdi), %ymm10 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm10[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm5[0,1,2,3],ymm7[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovaps 1216(%rdi), %xmm15 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm15[0,1,2],xmm6[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm6, %xmm6 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm6[0],mem[1],ymm6[2,3,4],mem[5],ymm6[6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm6[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1336(%rdi), %ymm7 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm5[0,1,2,3],ymm6[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovaps 1440(%rdi), %xmm5 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm5[0,1,2],xmm13[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm7, %xmm7 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0],ymm7[1],mem[2,3,4],ymm7[5],mem[6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1560(%rdi), %ymm13 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm13[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm6[0,1,2,3],ymm7[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovaps 1664(%rdi), %xmm6 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm13 = xmm6[0,1,2],xmm14[3] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm13 = xmm13[0,1,3,2] ; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,0,2,3,5,4,6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm14, %xmm14 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm13 = xmm14[0,1],xmm13[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm14 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7] ; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm14 = ymm14[1,0,3,3,5,4,7,7] ; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm14 = ymm14[0,1,0,3] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1784(%rdi), %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm14[0,1,2,3,4,5,6],ymm9[7] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm9[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 136(%rdi), %xmm9 ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm9 = xmm9[0],mem[1],xmm9[2,3] ; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9 ; AVX2-FAST-PERLANE-NEXT: vpermps 192(%rdi), %ymm0, %ymm14 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm14[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 80(%rdi), %ymm14 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm12 = xmm14[0,1,2],xmm12[3] ; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm14, %xmm14 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm12 = xmm14[0,1],xmm12[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm9[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 360(%rdi), %xmm9 ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm9 = xmm9[0],mem[1],xmm9[2,3] ; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9 ; AVX2-FAST-PERLANE-NEXT: vpermps 416(%rdi), %ymm0, %ymm14 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm14[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 304(%rdi), %ymm14 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm14[0,1,2],xmm1[3] ; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm14, %xmm14 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm14[0,1],xmm1[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm9[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 584(%rdi), %xmm1 ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3] ; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 ; AVX2-FAST-PERLANE-NEXT: vpermps 640(%rdi), %ymm0, %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm9[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 528(%rdi), %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm9[0,1,2],xmm2[3] ; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm9, %xmm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm9[0,1],xmm2[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 808(%rdi), %xmm2 ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm2 = xmm2[0],mem[1],xmm2[2,3] ; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2 ; AVX2-FAST-PERLANE-NEXT: vpermps 864(%rdi), %ymm0, %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm9[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 752(%rdi), %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm9[0,1,2],xmm3[3] ; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm9, %xmm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm9[0,1],xmm3[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1032(%rdi), %xmm3 ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3] ; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3 ; AVX2-FAST-PERLANE-NEXT: vpermps 1088(%rdi), %ymm0, %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm9[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 976(%rdi), %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm9[0,1,2],xmm4[3] ; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm9, %xmm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm9[0,1],xmm4[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1256(%rdi), %xmm4 ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3] ; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4 ; AVX2-FAST-PERLANE-NEXT: vpermps 1312(%rdi), %ymm0, %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm9[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1200(%rdi), %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm9 = xmm9[0,1,2],xmm15[3] ; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm15, %xmm15 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm9 = xmm15[0,1],xmm9[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm9[0,1,2,3],ymm4[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1480(%rdi), %xmm9 ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm9 = xmm9[0],mem[1],xmm9[2,3] ; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9 ; AVX2-FAST-PERLANE-NEXT: vpermps 1536(%rdi), %ymm0, %ymm15 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm15[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1424(%rdi), %ymm15 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm15[0,1,2],xmm5[3] ; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm15, %xmm15 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm15[0,1],xmm5[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm9[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1704(%rdi), %xmm9 ; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # xmm9 = xmm9[0],mem[1],xmm9[2,3] ; AVX2-FAST-PERLANE-NEXT: vpermps 1760(%rdi), %ymm0, %ymm0 ; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3,4,5],ymm0[6,7] ; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1648(%rdi), %ymm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1,2],xmm6[3] ; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7] ; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload ; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7] ; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm9, %xmm9 ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1],xmm6[2,3] ; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7] ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 192(%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 128(%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 64(%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 224(%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 160(%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 96(%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%rsi) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 192(%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 128(%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 64(%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 224(%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 160(%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 96(%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%rdx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 192(%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 128(%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 64(%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 224(%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 160(%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 96(%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%rcx) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%r8) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 64(%r8) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 128(%r8) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 192(%r8) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 224(%r8) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 160(%r8) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 96(%r8) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%r8) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 224(%r9) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 192(%r9) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 160(%r9) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 128(%r9) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 96(%r9) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 64(%r9) ; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%r9) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%r9) ; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm13, 224(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm7, 192(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, 160(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm10, 128(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 96(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 64(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%rax) ; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 224(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 192(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 160(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 128(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 96(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm14, 32(%rax) ; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm12, (%rax) ; AVX2-FAST-PERLANE-NEXT: addq $2680, %rsp # imm = 0xA78 ; AVX2-FAST-PERLANE-NEXT: vzeroupper ; AVX2-FAST-PERLANE-NEXT: retq ; ; AVX512F-LABEL: load_i32_stride7_vf64: ; AVX512F: # %bb.0: ; AVX512F-NEXT: subq $3400, %rsp # imm = 0xD48 ; AVX512F-NEXT: vmovdqa64 1728(%rdi), %zmm2 ; AVX512F-NEXT: vmovdqa64 1664(%rdi), %zmm17 ; AVX512F-NEXT: vmovdqa64 1600(%rdi), %zmm11 ; AVX512F-NEXT: vmovdqa64 1280(%rdi), %zmm7 ; AVX512F-NEXT: vmovdqa64 1216(%rdi), %zmm5 ; AVX512F-NEXT: vmovdqa64 1152(%rdi), %zmm12 ; AVX512F-NEXT: vmovdqa64 832(%rdi), %zmm6 ; AVX512F-NEXT: vmovdqa64 768(%rdi), %zmm8 ; AVX512F-NEXT: vmovdqa64 704(%rdi), %zmm13 ; AVX512F-NEXT: vmovdqa64 384(%rdi), %zmm20 ; AVX512F-NEXT: vmovdqa64 320(%rdi), %zmm4 ; AVX512F-NEXT: vmovdqa64 256(%rdi), %zmm14 ; AVX512F-NEXT: vbroadcasti32x4 {{.*#+}} zmm1 = [20,27,6,13,20,27,6,13,20,27,6,13,20,27,6,13] ; AVX512F-NEXT: # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm14, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm4, %zmm1, %zmm3 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,10,11,12,13,18,25,0,0,10,11,12,13,18,25] ; AVX512F-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm20, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm13, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm8, %zmm1, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm6, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm12, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm5, %zmm1, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm7, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermi2d %zmm17, %zmm11, %zmm1 ; AVX512F-NEXT: vpermt2d %zmm2, %zmm0, %zmm1 ; AVX512F-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,7,14,21,28,0,0,0,0,7,14,21,28,0,0] ; AVX512F-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm12, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm5, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm1 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,19,26,0,9,10,11,12,13,19,26] ; AVX512F-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm7, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm13, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm8, %zmm1, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm6, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm14, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm4, %zmm1, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm20, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermi2d %zmm17, %zmm11, %zmm1 ; AVX512F-NEXT: vpermt2d %zmm2, %zmm0, %zmm1 ; AVX512F-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,1,8,15,22,29,0,0,0,1,8,15,22,29,0,0] ; AVX512F-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm12, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm5, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm1 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,20,27,0,9,10,11,12,13,20,27] ; AVX512F-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm7, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm13, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm8, %zmm1, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm6, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm14, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm4, %zmm1, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm20, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermi2d %zmm17, %zmm11, %zmm1 ; AVX512F-NEXT: vpermt2d %zmm2, %zmm0, %zmm1 ; AVX512F-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,18,25,0,7,14,0,0,0,18,25,0,7,14,0,0] ; AVX512F-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm5, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm12, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm1 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,21,28,0,9,10,11,12,13,21,28] ; AVX512F-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm7, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm8, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm13, %zmm1, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm6, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm4, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm14, %zmm1, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm20, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermi2d %zmm11, %zmm17, %zmm1 ; AVX512F-NEXT: vpermt2d %zmm2, %zmm0, %zmm1 ; AVX512F-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,19,26,1,8,15,0,0,0,19,26,1,8,15,0,0] ; AVX512F-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm5, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm12, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm1 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,22,29,0,9,10,11,12,13,22,29] ; AVX512F-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm7, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm8, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm13, %zmm1, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm6, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm4, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm14, %zmm1, %zmm3 ; AVX512F-NEXT: vpermt2d %zmm20, %zmm0, %zmm3 ; AVX512F-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermi2d %zmm11, %zmm17, %zmm1 ; AVX512F-NEXT: vpermt2d %zmm2, %zmm0, %zmm1 ; AVX512F-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 1024(%rdi), %zmm3 ; AVX512F-NEXT: vmovdqa64 1088(%rdi), %zmm15 ; AVX512F-NEXT: vbroadcasti32x4 {{.*#+}} zmm30 = [25,4,11,18,25,4,11,18,25,4,11,18,25,4,11,18] ; AVX512F-NEXT: # zmm30 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm3, %zmm0 ; AVX512F-NEXT: vpermt2d %zmm15, %zmm30, %zmm0 ; AVX512F-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 576(%rdi), %zmm9 ; AVX512F-NEXT: vmovdqa64 640(%rdi), %zmm16 ; AVX512F-NEXT: vmovdqa64 %zmm9, %zmm0 ; AVX512F-NEXT: vpermt2d %zmm16, %zmm30, %zmm0 ; AVX512F-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 128(%rdi), %zmm0 ; AVX512F-NEXT: vmovdqa64 192(%rdi), %zmm18 ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm1 ; AVX512F-NEXT: vpermt2d %zmm18, %zmm30, %zmm1 ; AVX512F-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 1472(%rdi), %zmm1 ; AVX512F-NEXT: vmovdqa64 1536(%rdi), %zmm19 ; AVX512F-NEXT: vmovdqa64 %zmm1, %zmm10 ; AVX512F-NEXT: vpermt2d %zmm19, %zmm30, %zmm10 ; AVX512F-NEXT: vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm12, %zmm21 ; AVX512F-NEXT: vpermt2d %zmm5, %zmm30, %zmm21 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm10 = [0,9,10,11,12,16,23,30,0,9,10,11,12,16,23,30] ; AVX512F-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm7, %zmm10, %zmm21 ; AVX512F-NEXT: vmovdqu64 %zmm21, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm13, %zmm21 ; AVX512F-NEXT: vpermt2d %zmm8, %zmm30, %zmm21 ; AVX512F-NEXT: vpermt2d %zmm6, %zmm10, %zmm21 ; AVX512F-NEXT: vmovdqu64 %zmm21, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm14, %zmm21 ; AVX512F-NEXT: vpermt2d %zmm4, %zmm30, %zmm21 ; AVX512F-NEXT: vpermt2d %zmm20, %zmm10, %zmm21 ; AVX512F-NEXT: vmovdqu64 %zmm21, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermi2d %zmm17, %zmm11, %zmm30 ; AVX512F-NEXT: vpermt2d %zmm2, %zmm10, %zmm30 ; AVX512F-NEXT: vbroadcasti32x4 {{.*#+}} zmm10 = [26,5,12,19,26,5,12,19,26,5,12,19,26,5,12,19] ; AVX512F-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm8, %zmm10, %zmm13 ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,9,10,11,12,17,24,31,0,9,10,11,12,17,24,31] ; AVX512F-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm6, %zmm8, %zmm13 ; AVX512F-NEXT: vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermt2d %zmm4, %zmm10, %zmm14 ; AVX512F-NEXT: vpermt2d %zmm20, %zmm8, %zmm14 ; AVX512F-NEXT: vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermt2d %zmm17, %zmm10, %zmm11 ; AVX512F-NEXT: vpermt2d %zmm2, %zmm8, %zmm11 ; AVX512F-NEXT: vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermt2d %zmm5, %zmm10, %zmm12 ; AVX512F-NEXT: vpermt2d %zmm7, %zmm8, %zmm12 ; AVX512F-NEXT: vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm3, %zmm2 ; AVX512F-NEXT: vpermt2d %zmm15, %zmm10, %zmm2 ; AVX512F-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm9, %zmm2 ; AVX512F-NEXT: vpermt2d %zmm16, %zmm10, %zmm2 ; AVX512F-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm2 ; AVX512F-NEXT: vpermt2d %zmm18, %zmm10, %zmm2 ; AVX512F-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm1, %zmm2 ; AVX512F-NEXT: vpermt2d %zmm19, %zmm10, %zmm2 ; AVX512F-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm25 = [8,15,0,0,0,19,26,1,8,15,0,0,0,19,26,1] ; AVX512F-NEXT: # zmm25 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm16, %zmm2 ; AVX512F-NEXT: vpermt2d %zmm9, %zmm25, %zmm2 ; AVX512F-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vbroadcasti32x4 {{.*#+}} zmm27 = [27,6,13,20,27,6,13,20,27,6,13,20,27,6,13,20] ; AVX512F-NEXT: # zmm27 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm9, %zmm2 ; AVX512F-NEXT: vpermt2d %zmm16, %zmm27, %zmm2 ; AVX512F-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm28 = [28,0,0,0,0,7,14,21,28,0,0,0,0,7,14,21] ; AVX512F-NEXT: # zmm28 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm9, %zmm2 ; AVX512F-NEXT: vpermt2d %zmm16, %zmm28, %zmm2 ; AVX512F-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm31 = [29,0,0,0,1,8,15,22,29,0,0,0,1,8,15,22] ; AVX512F-NEXT: # zmm31 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vmovdqa64 %zmm9, %zmm2 ; AVX512F-NEXT: vpermt2d %zmm16, %zmm31, %zmm2 ; AVX512F-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm2 = [14,0,0,0,18,25,0,7,14,0,0,0,18,25,0,7] ; AVX512F-NEXT: # zmm2 = mem[0,1,2,3,0,1,2,3] ; AVX512F-NEXT: vpermt2d %zmm9, %zmm2, %zmm16 ; AVX512F-NEXT: vmovdqu64 %zmm16, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm18, %zmm4 ; AVX512F-NEXT: vpermt2d %zmm0, %zmm25, %zmm4 ; AVX512F-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm4 ; AVX512F-NEXT: vpermt2d %zmm18, %zmm27, %zmm4 ; AVX512F-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm4 ; AVX512F-NEXT: vpermt2d %zmm18, %zmm28, %zmm4 ; AVX512F-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm4 ; AVX512F-NEXT: vpermt2d %zmm18, %zmm31, %zmm4 ; AVX512F-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermt2d %zmm0, %zmm2, %zmm18 ; AVX512F-NEXT: vmovdqu64 %zmm18, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm15, %zmm0 ; AVX512F-NEXT: vpermt2d %zmm3, %zmm25, %zmm0 ; AVX512F-NEXT: vmovdqu64 %zmm0, (%rsp) # 64-byte Spill ; AVX512F-NEXT: vpermi2d %zmm1, %zmm19, %zmm25 ; AVX512F-NEXT: vmovdqa64 %zmm3, %zmm0 ; AVX512F-NEXT: vpermt2d %zmm15, %zmm27, %zmm0 ; AVX512F-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermi2d %zmm19, %zmm1, %zmm27 ; AVX512F-NEXT: vmovdqa64 %zmm3, %zmm0 ; AVX512F-NEXT: vpermt2d %zmm15, %zmm28, %zmm0 ; AVX512F-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermi2d %zmm19, %zmm1, %zmm28 ; AVX512F-NEXT: vmovdqa64 %zmm3, %zmm0 ; AVX512F-NEXT: vpermt2d %zmm15, %zmm31, %zmm0 ; AVX512F-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermi2d %zmm19, %zmm1, %zmm31 ; AVX512F-NEXT: vpermt2d %zmm1, %zmm2, %zmm19 ; AVX512F-NEXT: vmovdqu64 %zmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermt2d %zmm3, %zmm2, %zmm15 ; AVX512F-NEXT: vmovdqu64 %zmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 512(%rdi), %zmm0 ; AVX512F-NEXT: vmovdqa64 448(%rdi), %zmm17 ; AVX512F-NEXT: vmovdqa {{.*#+}} ymm2 = [0,7,14,21,28,u,u,u] ; AVX512F-NEXT: vmovdqa64 %zmm17, %zmm22 ; AVX512F-NEXT: vpermt2d %zmm0, %zmm2, %zmm22 ; AVX512F-NEXT: vmovdqa {{.*#+}} ymm3 = [1,8,15,22,29,u,u,u] ; AVX512F-NEXT: vmovdqa64 %zmm17, %zmm23 ; AVX512F-NEXT: vpermt2d %zmm0, %zmm3, %zmm23 ; AVX512F-NEXT: vmovdqa {{.*#+}} ymm4 = [18,25,0,7,14,u,u,u] ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm24 ; AVX512F-NEXT: vpermt2d %zmm17, %zmm4, %zmm24 ; AVX512F-NEXT: vmovdqa {{.*#+}} ymm7 = [19,26,1,8,15,u,u,u] ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm29 ; AVX512F-NEXT: vpermt2d %zmm17, %zmm7, %zmm29 ; AVX512F-NEXT: vmovdqa64 {{.*#+}} xmm18 = [4,11,18,25] ; AVX512F-NEXT: vmovdqa64 %zmm17, %zmm1 ; AVX512F-NEXT: vpermt2d %zmm0, %zmm18, %zmm1 ; AVX512F-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 {{.*#+}} xmm19 = [5,12,19,26] ; AVX512F-NEXT: vmovdqa64 %zmm17, %zmm1 ; AVX512F-NEXT: vpermt2d %zmm0, %zmm19, %zmm1 ; AVX512F-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 {{.*#+}} xmm20 = [6,13,20,27] ; AVX512F-NEXT: vpermt2d %zmm0, %zmm20, %zmm17 ; AVX512F-NEXT: vmovdqa64 (%rdi), %zmm5 ; AVX512F-NEXT: vmovdqa64 64(%rdi), %zmm0 ; AVX512F-NEXT: vmovdqa64 %zmm5, %zmm13 ; AVX512F-NEXT: vpermt2d %zmm0, %zmm2, %zmm13 ; AVX512F-NEXT: vmovdqa64 %zmm5, %zmm14 ; AVX512F-NEXT: vpermt2d %zmm0, %zmm3, %zmm14 ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm15 ; AVX512F-NEXT: vpermt2d %zmm5, %zmm4, %zmm15 ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm16 ; AVX512F-NEXT: vpermt2d %zmm5, %zmm7, %zmm16 ; AVX512F-NEXT: vmovdqa64 %zmm5, %zmm1 ; AVX512F-NEXT: vpermt2d %zmm0, %zmm18, %zmm1 ; AVX512F-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vmovdqa64 %zmm5, %zmm1 ; AVX512F-NEXT: vpermt2d %zmm0, %zmm19, %zmm1 ; AVX512F-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512F-NEXT: vpermt2d %zmm0, %zmm20, %zmm5 ; AVX512F-NEXT: vmovdqa64 960(%rdi), %zmm9 ; AVX512F-NEXT: vmovdqa64 896(%rdi), %zmm6 ; AVX512F-NEXT: vmovdqa64 %zmm6, %zmm8 ; AVX512F-NEXT: vpermt2d %zmm9, %zmm2, %zmm8 ; AVX512F-NEXT: vmovdqa64 1408(%rdi), %zmm0 ; AVX512F-NEXT: vmovdqa64 1344(%rdi), %zmm1 ; AVX512F-NEXT: vpermi2d %zmm0, %zmm1, %zmm2 ; AVX512F-NEXT: vmovdqa64 %zmm6, %zmm10 ; AVX512F-NEXT: vpermt2d %zmm9, %zmm3, %zmm10 ; AVX512F-NEXT: vpermi2d %zmm0, %zmm1, %zmm3 ; AVX512F-NEXT: vmovdqa64 %zmm9, %zmm11 ; AVX512F-NEXT: vpermt2d %zmm6, %zmm4, %zmm11 ; AVX512F-NEXT: vpermi2d %zmm1, %zmm0, %zmm4 ; AVX512F-NEXT: vmovdqa64 %zmm9, %zmm12 ; AVX512F-NEXT: vpermt2d %zmm6, %zmm7, %zmm12 ; AVX512F-NEXT: vpermi2d %zmm1, %zmm0, %zmm7 ; AVX512F-NEXT: vmovdqa64 %zmm6, %zmm21 ; AVX512F-NEXT: vpermt2d %zmm9, %zmm18, %zmm21 ; AVX512F-NEXT: vpermi2d %zmm0, %zmm1, %zmm18 ; AVX512F-NEXT: vmovdqa64 %zmm6, %zmm26 ; AVX512F-NEXT: vpermt2d %zmm9, %zmm19, %zmm26 ; AVX512F-NEXT: vpermi2d %zmm0, %zmm1, %zmm19 ; AVX512F-NEXT: vpermt2d %zmm0, %zmm20, %zmm1 ; AVX512F-NEXT: vpermt2d %zmm9, %zmm20, %zmm6 ; AVX512F-NEXT: movw $992, %ax # imm = 0x3E0 ; AVX512F-NEXT: kmovw %eax, %k1 ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm13 {%k1} ; AVX512F-NEXT: movb $-32, %al ; AVX512F-NEXT: kmovw %eax, %k2 ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm13 {%k2} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm22 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm22 {%k2} ; AVX512F-NEXT: vmovdqu64 (%rsp), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm8 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm8 {%k2} ; AVX512F-NEXT: vmovdqa32 %zmm25, %zmm2 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm2 {%k2} ; AVX512F-NEXT: movw $480, %ax # imm = 0x1E0 ; AVX512F-NEXT: kmovw %eax, %k2 ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm10 {%k2} ; AVX512F-NEXT: movw $-512, %ax # imm = 0xFE00 ; AVX512F-NEXT: kmovw %eax, %k1 ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm10 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm23 {%k2} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm23 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm14 {%k2} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm14 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm3 {%k2} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm3 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm11 {%k2} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm11 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm24 {%k2} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm24 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm15 {%k2} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm15 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm4 {%k2} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm4 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm12 {%k2} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm12 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm29 {%k2} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm29 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm16 {%k2} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm16 {%k1} ; AVX512F-NEXT: vmovdqa32 %zmm27, %zmm7 {%k2} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm7 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512F-NEXT: vinserti32x4 $0, %xmm21, %zmm0, %zmm0 ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm9, %zmm0 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload ; AVX512F-NEXT: vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm9 # 16-byte Folded Reload ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm20, %zmm9 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload ; AVX512F-NEXT: vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm20, %zmm20 # 16-byte Folded Reload ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm25, %zmm20 {%k1} ; AVX512F-NEXT: vinserti32x4 $0, %xmm18, %zmm28, %zmm18 ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm25, %zmm18 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload ; AVX512F-NEXT: vinserti32x4 $0, %xmm26, %zmm25, %zmm25 ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm26, %zmm25 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload ; AVX512F-NEXT: vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm26, %zmm26 # 16-byte Folded Reload ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm27, %zmm26 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload ; AVX512F-NEXT: vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm27, %zmm27 # 16-byte Folded Reload ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm28, %zmm27 {%k1} ; AVX512F-NEXT: vinserti32x4 $0, %xmm19, %zmm31, %zmm19 ; AVX512F-NEXT: vmovdqa32 %zmm30, %zmm19 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512F-NEXT: vinserti32x4 $0, %xmm17, %zmm28, %zmm17 ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm28, %zmm17 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512F-NEXT: vinserti32x4 $0, %xmm5, %zmm28, %zmm5 ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm28, %zmm5 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512F-NEXT: vinserti32x4 $0, %xmm1, %zmm28, %zmm1 ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm28, %zmm1 {%k1} ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512F-NEXT: vinserti32x4 $0, %xmm6, %zmm28, %zmm6 ; AVX512F-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512F-NEXT: vmovdqa32 %zmm28, %zmm6 {%k1} ; AVX512F-NEXT: vmovdqa64 %zmm2, 192(%rsi) ; AVX512F-NEXT: vmovdqa64 %zmm8, 128(%rsi) ; AVX512F-NEXT: vmovdqa64 %zmm22, 64(%rsi) ; AVX512F-NEXT: vmovdqa64 %zmm13, (%rsi) ; AVX512F-NEXT: vmovdqa64 %zmm3, 192(%rdx) ; AVX512F-NEXT: vmovdqa64 %zmm14, (%rdx) ; AVX512F-NEXT: vmovdqa64 %zmm23, 64(%rdx) ; AVX512F-NEXT: vmovdqa64 %zmm10, 128(%rdx) ; AVX512F-NEXT: vmovdqa64 %zmm4, 192(%rcx) ; AVX512F-NEXT: vmovdqa64 %zmm15, (%rcx) ; AVX512F-NEXT: vmovdqa64 %zmm24, 64(%rcx) ; AVX512F-NEXT: vmovdqa64 %zmm11, 128(%rcx) ; AVX512F-NEXT: vmovdqa64 %zmm7, 192(%r8) ; AVX512F-NEXT: vmovdqa64 %zmm16, (%r8) ; AVX512F-NEXT: vmovdqa64 %zmm29, 64(%r8) ; AVX512F-NEXT: vmovdqa64 %zmm12, 128(%r8) ; AVX512F-NEXT: vmovdqa64 %zmm18, 192(%r9) ; AVX512F-NEXT: vmovdqa64 %zmm20, (%r9) ; AVX512F-NEXT: vmovdqa64 %zmm9, 64(%r9) ; AVX512F-NEXT: vmovdqa64 %zmm0, 128(%r9) ; AVX512F-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX512F-NEXT: vmovdqa64 %zmm19, 192(%rax) ; AVX512F-NEXT: vmovdqa64 %zmm27, (%rax) ; AVX512F-NEXT: vmovdqa64 %zmm26, 64(%rax) ; AVX512F-NEXT: vmovdqa64 %zmm25, 128(%rax) ; AVX512F-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX512F-NEXT: vmovdqa64 %zmm6, 128(%rax) ; AVX512F-NEXT: vmovdqa64 %zmm1, 192(%rax) ; AVX512F-NEXT: vmovdqa64 %zmm5, (%rax) ; AVX512F-NEXT: vmovdqa64 %zmm17, 64(%rax) ; AVX512F-NEXT: addq $3400, %rsp # imm = 0xD48 ; AVX512F-NEXT: vzeroupper ; AVX512F-NEXT: retq ; ; AVX512BW-LABEL: load_i32_stride7_vf64: ; AVX512BW: # %bb.0: ; AVX512BW-NEXT: subq $3400, %rsp # imm = 0xD48 ; AVX512BW-NEXT: vmovdqa64 1728(%rdi), %zmm2 ; AVX512BW-NEXT: vmovdqa64 1664(%rdi), %zmm17 ; AVX512BW-NEXT: vmovdqa64 1600(%rdi), %zmm11 ; AVX512BW-NEXT: vmovdqa64 1280(%rdi), %zmm7 ; AVX512BW-NEXT: vmovdqa64 1216(%rdi), %zmm5 ; AVX512BW-NEXT: vmovdqa64 1152(%rdi), %zmm12 ; AVX512BW-NEXT: vmovdqa64 832(%rdi), %zmm6 ; AVX512BW-NEXT: vmovdqa64 768(%rdi), %zmm8 ; AVX512BW-NEXT: vmovdqa64 704(%rdi), %zmm13 ; AVX512BW-NEXT: vmovdqa64 384(%rdi), %zmm20 ; AVX512BW-NEXT: vmovdqa64 320(%rdi), %zmm4 ; AVX512BW-NEXT: vmovdqa64 256(%rdi), %zmm14 ; AVX512BW-NEXT: vbroadcasti32x4 {{.*#+}} zmm1 = [20,27,6,13,20,27,6,13,20,27,6,13,20,27,6,13] ; AVX512BW-NEXT: # zmm1 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm14, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm4, %zmm1, %zmm3 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,10,11,12,13,18,25,0,0,10,11,12,13,18,25] ; AVX512BW-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm20, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm13, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm8, %zmm1, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm6, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm12, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm1, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm7, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermi2d %zmm17, %zmm11, %zmm1 ; AVX512BW-NEXT: vpermt2d %zmm2, %zmm0, %zmm1 ; AVX512BW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,0,7,14,21,28,0,0,0,0,7,14,21,28,0,0] ; AVX512BW-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm12, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm1 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,19,26,0,9,10,11,12,13,19,26] ; AVX512BW-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm7, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm13, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm8, %zmm1, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm6, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm14, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm4, %zmm1, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm20, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermi2d %zmm17, %zmm11, %zmm1 ; AVX512BW-NEXT: vpermt2d %zmm2, %zmm0, %zmm1 ; AVX512BW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,1,8,15,22,29,0,0,0,1,8,15,22,29,0,0] ; AVX512BW-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm12, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm1 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,20,27,0,9,10,11,12,13,20,27] ; AVX512BW-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm7, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm13, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm8, %zmm1, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm6, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm14, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm4, %zmm1, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm20, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermi2d %zmm17, %zmm11, %zmm1 ; AVX512BW-NEXT: vpermt2d %zmm2, %zmm0, %zmm1 ; AVX512BW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,18,25,0,7,14,0,0,0,18,25,0,7,14,0,0] ; AVX512BW-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm5, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm12, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm1 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,21,28,0,9,10,11,12,13,21,28] ; AVX512BW-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm7, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm8, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm13, %zmm1, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm6, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm4, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm14, %zmm1, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm20, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermi2d %zmm11, %zmm17, %zmm1 ; AVX512BW-NEXT: vpermt2d %zmm2, %zmm0, %zmm1 ; AVX512BW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,19,26,1,8,15,0,0,0,19,26,1,8,15,0,0] ; AVX512BW-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm5, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm12, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm1 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,9,10,11,12,13,22,29,0,9,10,11,12,13,22,29] ; AVX512BW-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm7, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm8, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm13, %zmm1, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm6, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm4, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm14, %zmm1, %zmm3 ; AVX512BW-NEXT: vpermt2d %zmm20, %zmm0, %zmm3 ; AVX512BW-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermi2d %zmm11, %zmm17, %zmm1 ; AVX512BW-NEXT: vpermt2d %zmm2, %zmm0, %zmm1 ; AVX512BW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 1024(%rdi), %zmm3 ; AVX512BW-NEXT: vmovdqa64 1088(%rdi), %zmm15 ; AVX512BW-NEXT: vbroadcasti32x4 {{.*#+}} zmm30 = [25,4,11,18,25,4,11,18,25,4,11,18,25,4,11,18] ; AVX512BW-NEXT: # zmm30 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm3, %zmm0 ; AVX512BW-NEXT: vpermt2d %zmm15, %zmm30, %zmm0 ; AVX512BW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 576(%rdi), %zmm9 ; AVX512BW-NEXT: vmovdqa64 640(%rdi), %zmm16 ; AVX512BW-NEXT: vmovdqa64 %zmm9, %zmm0 ; AVX512BW-NEXT: vpermt2d %zmm16, %zmm30, %zmm0 ; AVX512BW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 128(%rdi), %zmm0 ; AVX512BW-NEXT: vmovdqa64 192(%rdi), %zmm18 ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm1 ; AVX512BW-NEXT: vpermt2d %zmm18, %zmm30, %zmm1 ; AVX512BW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 1472(%rdi), %zmm1 ; AVX512BW-NEXT: vmovdqa64 1536(%rdi), %zmm19 ; AVX512BW-NEXT: vmovdqa64 %zmm1, %zmm10 ; AVX512BW-NEXT: vpermt2d %zmm19, %zmm30, %zmm10 ; AVX512BW-NEXT: vmovdqu64 %zmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm12, %zmm21 ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm30, %zmm21 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm10 = [0,9,10,11,12,16,23,30,0,9,10,11,12,16,23,30] ; AVX512BW-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm7, %zmm10, %zmm21 ; AVX512BW-NEXT: vmovdqu64 %zmm21, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm13, %zmm21 ; AVX512BW-NEXT: vpermt2d %zmm8, %zmm30, %zmm21 ; AVX512BW-NEXT: vpermt2d %zmm6, %zmm10, %zmm21 ; AVX512BW-NEXT: vmovdqu64 %zmm21, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm14, %zmm21 ; AVX512BW-NEXT: vpermt2d %zmm4, %zmm30, %zmm21 ; AVX512BW-NEXT: vpermt2d %zmm20, %zmm10, %zmm21 ; AVX512BW-NEXT: vmovdqu64 %zmm21, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermi2d %zmm17, %zmm11, %zmm30 ; AVX512BW-NEXT: vpermt2d %zmm2, %zmm10, %zmm30 ; AVX512BW-NEXT: vbroadcasti32x4 {{.*#+}} zmm10 = [26,5,12,19,26,5,12,19,26,5,12,19,26,5,12,19] ; AVX512BW-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm8, %zmm10, %zmm13 ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,9,10,11,12,17,24,31,0,9,10,11,12,17,24,31] ; AVX512BW-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm6, %zmm8, %zmm13 ; AVX512BW-NEXT: vmovdqu64 %zmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermt2d %zmm4, %zmm10, %zmm14 ; AVX512BW-NEXT: vpermt2d %zmm20, %zmm8, %zmm14 ; AVX512BW-NEXT: vmovdqu64 %zmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermt2d %zmm17, %zmm10, %zmm11 ; AVX512BW-NEXT: vpermt2d %zmm2, %zmm8, %zmm11 ; AVX512BW-NEXT: vmovdqu64 %zmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm10, %zmm12 ; AVX512BW-NEXT: vpermt2d %zmm7, %zmm8, %zmm12 ; AVX512BW-NEXT: vmovdqu64 %zmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm3, %zmm2 ; AVX512BW-NEXT: vpermt2d %zmm15, %zmm10, %zmm2 ; AVX512BW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm9, %zmm2 ; AVX512BW-NEXT: vpermt2d %zmm16, %zmm10, %zmm2 ; AVX512BW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm2 ; AVX512BW-NEXT: vpermt2d %zmm18, %zmm10, %zmm2 ; AVX512BW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm1, %zmm2 ; AVX512BW-NEXT: vpermt2d %zmm19, %zmm10, %zmm2 ; AVX512BW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm25 = [8,15,0,0,0,19,26,1,8,15,0,0,0,19,26,1] ; AVX512BW-NEXT: # zmm25 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm16, %zmm2 ; AVX512BW-NEXT: vpermt2d %zmm9, %zmm25, %zmm2 ; AVX512BW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vbroadcasti32x4 {{.*#+}} zmm27 = [27,6,13,20,27,6,13,20,27,6,13,20,27,6,13,20] ; AVX512BW-NEXT: # zmm27 = mem[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm9, %zmm2 ; AVX512BW-NEXT: vpermt2d %zmm16, %zmm27, %zmm2 ; AVX512BW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm28 = [28,0,0,0,0,7,14,21,28,0,0,0,0,7,14,21] ; AVX512BW-NEXT: # zmm28 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm9, %zmm2 ; AVX512BW-NEXT: vpermt2d %zmm16, %zmm28, %zmm2 ; AVX512BW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm31 = [29,0,0,0,1,8,15,22,29,0,0,0,1,8,15,22] ; AVX512BW-NEXT: # zmm31 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vmovdqa64 %zmm9, %zmm2 ; AVX512BW-NEXT: vpermt2d %zmm16, %zmm31, %zmm2 ; AVX512BW-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm2 = [14,0,0,0,18,25,0,7,14,0,0,0,18,25,0,7] ; AVX512BW-NEXT: # zmm2 = mem[0,1,2,3,0,1,2,3] ; AVX512BW-NEXT: vpermt2d %zmm9, %zmm2, %zmm16 ; AVX512BW-NEXT: vmovdqu64 %zmm16, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm18, %zmm4 ; AVX512BW-NEXT: vpermt2d %zmm0, %zmm25, %zmm4 ; AVX512BW-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm4 ; AVX512BW-NEXT: vpermt2d %zmm18, %zmm27, %zmm4 ; AVX512BW-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm4 ; AVX512BW-NEXT: vpermt2d %zmm18, %zmm28, %zmm4 ; AVX512BW-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm4 ; AVX512BW-NEXT: vpermt2d %zmm18, %zmm31, %zmm4 ; AVX512BW-NEXT: vmovdqu64 %zmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermt2d %zmm0, %zmm2, %zmm18 ; AVX512BW-NEXT: vmovdqu64 %zmm18, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm15, %zmm0 ; AVX512BW-NEXT: vpermt2d %zmm3, %zmm25, %zmm0 ; AVX512BW-NEXT: vmovdqu64 %zmm0, (%rsp) # 64-byte Spill ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm19, %zmm25 ; AVX512BW-NEXT: vmovdqa64 %zmm3, %zmm0 ; AVX512BW-NEXT: vpermt2d %zmm15, %zmm27, %zmm0 ; AVX512BW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermi2d %zmm19, %zmm1, %zmm27 ; AVX512BW-NEXT: vmovdqa64 %zmm3, %zmm0 ; AVX512BW-NEXT: vpermt2d %zmm15, %zmm28, %zmm0 ; AVX512BW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermi2d %zmm19, %zmm1, %zmm28 ; AVX512BW-NEXT: vmovdqa64 %zmm3, %zmm0 ; AVX512BW-NEXT: vpermt2d %zmm15, %zmm31, %zmm0 ; AVX512BW-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermi2d %zmm19, %zmm1, %zmm31 ; AVX512BW-NEXT: vpermt2d %zmm1, %zmm2, %zmm19 ; AVX512BW-NEXT: vmovdqu64 %zmm19, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermt2d %zmm3, %zmm2, %zmm15 ; AVX512BW-NEXT: vmovdqu64 %zmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 512(%rdi), %zmm0 ; AVX512BW-NEXT: vmovdqa64 448(%rdi), %zmm17 ; AVX512BW-NEXT: vmovdqa {{.*#+}} ymm2 = [0,7,14,21,28,u,u,u] ; AVX512BW-NEXT: vmovdqa64 %zmm17, %zmm22 ; AVX512BW-NEXT: vpermt2d %zmm0, %zmm2, %zmm22 ; AVX512BW-NEXT: vmovdqa {{.*#+}} ymm3 = [1,8,15,22,29,u,u,u] ; AVX512BW-NEXT: vmovdqa64 %zmm17, %zmm23 ; AVX512BW-NEXT: vpermt2d %zmm0, %zmm3, %zmm23 ; AVX512BW-NEXT: vmovdqa {{.*#+}} ymm4 = [18,25,0,7,14,u,u,u] ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm24 ; AVX512BW-NEXT: vpermt2d %zmm17, %zmm4, %zmm24 ; AVX512BW-NEXT: vmovdqa {{.*#+}} ymm7 = [19,26,1,8,15,u,u,u] ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm29 ; AVX512BW-NEXT: vpermt2d %zmm17, %zmm7, %zmm29 ; AVX512BW-NEXT: vmovdqa64 {{.*#+}} xmm18 = [4,11,18,25] ; AVX512BW-NEXT: vmovdqa64 %zmm17, %zmm1 ; AVX512BW-NEXT: vpermt2d %zmm0, %zmm18, %zmm1 ; AVX512BW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 {{.*#+}} xmm19 = [5,12,19,26] ; AVX512BW-NEXT: vmovdqa64 %zmm17, %zmm1 ; AVX512BW-NEXT: vpermt2d %zmm0, %zmm19, %zmm1 ; AVX512BW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 {{.*#+}} xmm20 = [6,13,20,27] ; AVX512BW-NEXT: vpermt2d %zmm0, %zmm20, %zmm17 ; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm5 ; AVX512BW-NEXT: vmovdqa64 64(%rdi), %zmm0 ; AVX512BW-NEXT: vmovdqa64 %zmm5, %zmm13 ; AVX512BW-NEXT: vpermt2d %zmm0, %zmm2, %zmm13 ; AVX512BW-NEXT: vmovdqa64 %zmm5, %zmm14 ; AVX512BW-NEXT: vpermt2d %zmm0, %zmm3, %zmm14 ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm15 ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm4, %zmm15 ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm16 ; AVX512BW-NEXT: vpermt2d %zmm5, %zmm7, %zmm16 ; AVX512BW-NEXT: vmovdqa64 %zmm5, %zmm1 ; AVX512BW-NEXT: vpermt2d %zmm0, %zmm18, %zmm1 ; AVX512BW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vmovdqa64 %zmm5, %zmm1 ; AVX512BW-NEXT: vpermt2d %zmm0, %zmm19, %zmm1 ; AVX512BW-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill ; AVX512BW-NEXT: vpermt2d %zmm0, %zmm20, %zmm5 ; AVX512BW-NEXT: vmovdqa64 960(%rdi), %zmm9 ; AVX512BW-NEXT: vmovdqa64 896(%rdi), %zmm6 ; AVX512BW-NEXT: vmovdqa64 %zmm6, %zmm8 ; AVX512BW-NEXT: vpermt2d %zmm9, %zmm2, %zmm8 ; AVX512BW-NEXT: vmovdqa64 1408(%rdi), %zmm0 ; AVX512BW-NEXT: vmovdqa64 1344(%rdi), %zmm1 ; AVX512BW-NEXT: vpermi2d %zmm0, %zmm1, %zmm2 ; AVX512BW-NEXT: vmovdqa64 %zmm6, %zmm10 ; AVX512BW-NEXT: vpermt2d %zmm9, %zmm3, %zmm10 ; AVX512BW-NEXT: vpermi2d %zmm0, %zmm1, %zmm3 ; AVX512BW-NEXT: vmovdqa64 %zmm9, %zmm11 ; AVX512BW-NEXT: vpermt2d %zmm6, %zmm4, %zmm11 ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm0, %zmm4 ; AVX512BW-NEXT: vmovdqa64 %zmm9, %zmm12 ; AVX512BW-NEXT: vpermt2d %zmm6, %zmm7, %zmm12 ; AVX512BW-NEXT: vpermi2d %zmm1, %zmm0, %zmm7 ; AVX512BW-NEXT: vmovdqa64 %zmm6, %zmm21 ; AVX512BW-NEXT: vpermt2d %zmm9, %zmm18, %zmm21 ; AVX512BW-NEXT: vpermi2d %zmm0, %zmm1, %zmm18 ; AVX512BW-NEXT: vmovdqa64 %zmm6, %zmm26 ; AVX512BW-NEXT: vpermt2d %zmm9, %zmm19, %zmm26 ; AVX512BW-NEXT: vpermi2d %zmm0, %zmm1, %zmm19 ; AVX512BW-NEXT: vpermt2d %zmm0, %zmm20, %zmm1 ; AVX512BW-NEXT: vpermt2d %zmm9, %zmm20, %zmm6 ; AVX512BW-NEXT: movw $992, %ax # imm = 0x3E0 ; AVX512BW-NEXT: kmovd %eax, %k1 ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm13 {%k1} ; AVX512BW-NEXT: movb $-32, %al ; AVX512BW-NEXT: kmovd %eax, %k2 ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm13 {%k2} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm22 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm22 {%k2} ; AVX512BW-NEXT: vmovdqu64 (%rsp), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm8 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm8 {%k2} ; AVX512BW-NEXT: vmovdqa32 %zmm25, %zmm2 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm2 {%k2} ; AVX512BW-NEXT: movw $480, %ax # imm = 0x1E0 ; AVX512BW-NEXT: kmovd %eax, %k2 ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm10 {%k2} ; AVX512BW-NEXT: movw $-512, %ax # imm = 0xFE00 ; AVX512BW-NEXT: kmovd %eax, %k1 ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm10 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm23 {%k2} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm23 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm14 {%k2} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm14 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm3 {%k2} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm3 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm11 {%k2} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm11 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm24 {%k2} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm24 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm15 {%k2} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm15 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm4 {%k2} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm4 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm12 {%k2} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm12 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm29 {%k2} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm29 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm16 {%k2} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm16 {%k1} ; AVX512BW-NEXT: vmovdqa32 %zmm27, %zmm7 {%k2} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm0, %zmm7 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload ; AVX512BW-NEXT: vinserti32x4 $0, %xmm21, %zmm0, %zmm0 ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm9, %zmm0 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload ; AVX512BW-NEXT: vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm9 # 16-byte Folded Reload ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm20, %zmm9 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload ; AVX512BW-NEXT: vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm20, %zmm20 # 16-byte Folded Reload ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm25, %zmm20 {%k1} ; AVX512BW-NEXT: vinserti32x4 $0, %xmm18, %zmm28, %zmm18 ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm25, %zmm18 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload ; AVX512BW-NEXT: vinserti32x4 $0, %xmm26, %zmm25, %zmm25 ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm26, %zmm25 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload ; AVX512BW-NEXT: vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm26, %zmm26 # 16-byte Folded Reload ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm27, %zmm26 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm27 # 64-byte Reload ; AVX512BW-NEXT: vinserti32x4 $0, {{[-0-9]+}}(%r{{[sb]}}p), %zmm27, %zmm27 # 16-byte Folded Reload ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm28, %zmm27 {%k1} ; AVX512BW-NEXT: vinserti32x4 $0, %xmm19, %zmm31, %zmm19 ; AVX512BW-NEXT: vmovdqa32 %zmm30, %zmm19 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512BW-NEXT: vinserti32x4 $0, %xmm17, %zmm28, %zmm17 ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm28, %zmm17 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512BW-NEXT: vinserti32x4 $0, %xmm5, %zmm28, %zmm5 ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm28, %zmm5 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512BW-NEXT: vinserti32x4 $0, %xmm1, %zmm28, %zmm1 ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm28, %zmm1 {%k1} ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512BW-NEXT: vinserti32x4 $0, %xmm6, %zmm28, %zmm6 ; AVX512BW-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload ; AVX512BW-NEXT: vmovdqa32 %zmm28, %zmm6 {%k1} ; AVX512BW-NEXT: vmovdqa64 %zmm2, 192(%rsi) ; AVX512BW-NEXT: vmovdqa64 %zmm8, 128(%rsi) ; AVX512BW-NEXT: vmovdqa64 %zmm22, 64(%rsi) ; AVX512BW-NEXT: vmovdqa64 %zmm13, (%rsi) ; AVX512BW-NEXT: vmovdqa64 %zmm3, 192(%rdx) ; AVX512BW-NEXT: vmovdqa64 %zmm14, (%rdx) ; AVX512BW-NEXT: vmovdqa64 %zmm23, 64(%rdx) ; AVX512BW-NEXT: vmovdqa64 %zmm10, 128(%rdx) ; AVX512BW-NEXT: vmovdqa64 %zmm4, 192(%rcx) ; AVX512BW-NEXT: vmovdqa64 %zmm15, (%rcx) ; AVX512BW-NEXT: vmovdqa64 %zmm24, 64(%rcx) ; AVX512BW-NEXT: vmovdqa64 %zmm11, 128(%rcx) ; AVX512BW-NEXT: vmovdqa64 %zmm7, 192(%r8) ; AVX512BW-NEXT: vmovdqa64 %zmm16, (%r8) ; AVX512BW-NEXT: vmovdqa64 %zmm29, 64(%r8) ; AVX512BW-NEXT: vmovdqa64 %zmm12, 128(%r8) ; AVX512BW-NEXT: vmovdqa64 %zmm18, 192(%r9) ; AVX512BW-NEXT: vmovdqa64 %zmm20, (%r9) ; AVX512BW-NEXT: vmovdqa64 %zmm9, 64(%r9) ; AVX512BW-NEXT: vmovdqa64 %zmm0, 128(%r9) ; AVX512BW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX512BW-NEXT: vmovdqa64 %zmm19, 192(%rax) ; AVX512BW-NEXT: vmovdqa64 %zmm27, (%rax) ; AVX512BW-NEXT: vmovdqa64 %zmm26, 64(%rax) ; AVX512BW-NEXT: vmovdqa64 %zmm25, 128(%rax) ; AVX512BW-NEXT: movq {{[0-9]+}}(%rsp), %rax ; AVX512BW-NEXT: vmovdqa64 %zmm6, 128(%rax) ; AVX512BW-NEXT: vmovdqa64 %zmm1, 192(%rax) ; AVX512BW-NEXT: vmovdqa64 %zmm5, (%rax) ; AVX512BW-NEXT: vmovdqa64 %zmm17, 64(%rax) ; AVX512BW-NEXT: addq $3400, %rsp # imm = 0xD48 ; AVX512BW-NEXT: vzeroupper ; AVX512BW-NEXT: retq %wide.vec = load <448 x i32>, ptr %in.vec, align 64 %strided.vec0 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> %strided.vec1 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> %strided.vec2 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> %strided.vec3 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> %strided.vec4 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> %strided.vec5 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> %strided.vec6 = shufflevector <448 x i32> %wide.vec, <448 x i32> poison, <64 x i32> store <64 x i32> %strided.vec0, ptr %out.vec0, align 64 store <64 x i32> %strided.vec1, ptr %out.vec1, align 64 store <64 x i32> %strided.vec2, ptr %out.vec2, align 64 store <64 x i32> %strided.vec3, ptr %out.vec3, align 64 store <64 x i32> %strided.vec4, ptr %out.vec4, align 64 store <64 x i32> %strided.vec5, ptr %out.vec5, align 64 store <64 x i32> %strided.vec6, ptr %out.vec6, align 64 ret void } ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: ; AVX: {{.*}} ; AVX1: {{.*}} ; AVX2: {{.*}} ; AVX512BW-FAST: {{.*}} ; AVX512BW-ONLY: {{.*}} ; AVX512BW-ONLY-FAST: {{.*}} ; AVX512BW-ONLY-SLOW: {{.*}} ; AVX512BW-SLOW: {{.*}} ; AVX512DQ-FAST: {{.*}} ; AVX512DQ-ONLY: {{.*}} ; AVX512DQ-SLOW: {{.*}} ; AVX512DQBW-FAST: {{.*}} ; AVX512DQBW-ONLY: {{.*}} ; AVX512DQBW-SLOW: {{.*}} ; AVX512F-FAST: {{.*}} ; AVX512F-ONLY: {{.*}} ; AVX512F-ONLY-FAST: {{.*}} ; AVX512F-ONLY-SLOW: {{.*}} ; AVX512F-SLOW: {{.*}} ; FALLBACK0: {{.*}} ; FALLBACK1: {{.*}} ; FALLBACK10: {{.*}} ; FALLBACK11: {{.*}} ; FALLBACK12: {{.*}} ; FALLBACK2: {{.*}} ; FALLBACK3: {{.*}} ; FALLBACK4: {{.*}} ; FALLBACK5: {{.*}} ; FALLBACK6: {{.*}} ; FALLBACK7: {{.*}} ; FALLBACK8: {{.*}} ; FALLBACK9: {{.*}}