[llvm] bfe5b1b - [NFC][X86][Codegen] Add missing interleaving tests after D111546
Roman Lebedev via llvm-commits
llvm-commits at lists.llvm.org
Sat Oct 16 08:20:43 PDT 2021
Author: Roman Lebedev
Date: 2021-10-16T18:20:16+03:00
New Revision: bfe5b1b3122ca3cfee977380acae01f3b409c57b
URL: https://github.com/llvm/llvm-project/commit/bfe5b1b3122ca3cfee977380acae01f3b409c57b
DIFF: https://github.com/llvm/llvm-project/commit/bfe5b1b3122ca3cfee977380acae01f3b409c57b.diff
LOG: [NFC][X86][Codegen] Add missing interleaving tests after D111546
Added:
Modified:
llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-2.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-5.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-6.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-3.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-4.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-2.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-4.ll
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
index 16142f9b2f338..689b555c40114 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
@@ -1334,3 +1334,1560 @@ define void @vf16(<80 x i16>* %in.vec, <16 x i16>* %out.vec0, <16 x i16>* %out.v
ret void
}
+
+define void @vf32(<160 x i16>* %in.vec, <32 x i16>* %out.vec0, <32 x i16>* %out.vec1, <32 x i16>* %out.vec2, <32 x i16>* %out.vec3, <32 x i16>* %out.vec4) nounwind {
+; SSE-LABEL: vf32:
+; SSE: # %bb.0:
+; SSE-NEXT: subq $344, %rsp # imm = 0x158
+; SSE-NEXT: movdqa 304(%rdi), %xmm9
+; SSE-NEXT: movdqa 240(%rdi), %xmm10
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 256(%rdi), %xmm11
+; SSE-NEXT: movdqa 288(%rdi), %xmm12
+; SSE-NEXT: movdqa 272(%rdi), %xmm7
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 144(%rdi), %xmm4
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 80(%rdi), %xmm6
+; SSE-NEXT: movdqa 96(%rdi), %xmm8
+; SSE-NEXT: movdqa 128(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 112(%rdi), %xmm15
+; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,0,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: pandn %xmm15, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,1,0,3]
+; SSE-NEXT: pand %xmm0, %xmm2
+; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[3,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm6[0,2,2,3]
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
+; SSE-NEXT: movaps {{.*#+}} xmm13 = [65535,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT: andps %xmm13, %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[0,1,0,1]
+; SSE-NEXT: movaps %xmm13, %xmm2
+; SSE-NEXT: pandn %xmm1, %xmm2
+; SSE-NEXT: por %xmm3, %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: pandn %xmm7, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm12[0,1,0,3]
+; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pand %xmm0, %xmm2
+; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[3,1,2,3]
+; SSE-NEXT: movdqa %xmm11, %xmm4
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm10[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[0,1,0,1]
+; SSE-NEXT: movaps %xmm13, %xmm2
+; SSE-NEXT: andnps %xmm1, %xmm2
+; SSE-NEXT: movdqa 32(%rdi), %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: andps %xmm13, %xmm3
+; SSE-NEXT: orps %xmm3, %xmm2
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: movdqa 48(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT: pand %xmm0, %xmm2
+; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: movdqa 16(%rdi), %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[3,1,2,3]
+; SSE-NEXT: movdqa %xmm3, %xmm14
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT: movdqa (%rdi), %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
+; SSE-NEXT: movdqa 64(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; SSE-NEXT: movaps %xmm13, %xmm2
+; SSE-NEXT: andnps %xmm1, %xmm2
+; SSE-NEXT: andps %xmm13, %xmm3
+; SSE-NEXT: orps %xmm3, %xmm2
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 208(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; SSE-NEXT: pand %xmm0, %xmm1
+; SSE-NEXT: movdqa 192(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pandn %xmm2, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa 176(%rdi), %xmm11
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[3,1,2,3]
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT: movdqa 160(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
+; SSE-NEXT: movdqa 224(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; SSE-NEXT: movaps %xmm13, %xmm1
+; SSE-NEXT: andnps %xmm0, %xmm1
+; SSE-NEXT: andps %xmm13, %xmm2
+; SSE-NEXT: orps %xmm2, %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm8, %xmm10
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm8, %xmm0
+; SSE-NEXT: psrlq $48, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: movdqa {{.*#+}} xmm0 = [0,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: pandn %xmm1, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[1,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm15[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
+; SSE-NEXT: pand %xmm0, %xmm1
+; SSE-NEXT: por %xmm2, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm7, %xmm2
+; SSE-NEXT: psllq $48, %xmm2
+; SSE-NEXT: movaps %xmm13, %xmm3
+; SSE-NEXT: andnps %xmm2, %xmm3
+; SSE-NEXT: pand %xmm13, %xmm1
+; SSE-NEXT: orps %xmm1, %xmm3
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm4, %xmm1
+; SSE-NEXT: psrlq $48, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm6[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: pandn %xmm2, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm12[1,3,2,3]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,4,7]
+; SSE-NEXT: pand %xmm0, %xmm2
+; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm9, %xmm1
+; SSE-NEXT: psllq $48, %xmm1
+; SSE-NEXT: movdqa %xmm13, %xmm3
+; SSE-NEXT: pandn %xmm1, %xmm3
+; SSE-NEXT: pand %xmm13, %xmm2
+; SSE-NEXT: por %xmm2, %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm14, %xmm1
+; SSE-NEXT: psrlq $48, %xmm1
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm5[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: pandn %xmm2, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm14[1,3,2,3]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,4,7]
+; SSE-NEXT: pand %xmm0, %xmm2
+; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: psllq $48, %xmm1
+; SSE-NEXT: movdqa %xmm13, %xmm3
+; SSE-NEXT: pandn %xmm1, %xmm3
+; SSE-NEXT: pand %xmm13, %xmm2
+; SSE-NEXT: por %xmm2, %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrlq $48, %xmm11
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm4[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[1,3,2,3]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
+; SSE-NEXT: pand %xmm0, %xmm1
+; SSE-NEXT: pandn %xmm2, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: pand %xmm13, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: psllq $48, %xmm1
+; SSE-NEXT: pandn %xmm1, %xmm13
+; SSE-NEXT: por %xmm0, %xmm13
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[0,1,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm10[2],xmm1[3],xmm10[3]
+; SSE-NEXT: movdqa {{.*#+}} xmm11 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT: pand %xmm11, %xmm1
+; SSE-NEXT: movdqa %xmm15, (%rsp) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[1,0],xmm15[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[2,0],xmm15[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[0,1,1,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm8[1,1,1,1,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT: movdqa %xmm11, %xmm0
+; SSE-NEXT: pandn %xmm2, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm2[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[0,1,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; SSE-NEXT: pand %xmm11, %xmm1
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, %xmm2
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm6[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm6[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1,1,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,1,1,1,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT: movdqa %xmm11, %xmm6
+; SSE-NEXT: pandn %xmm2, %xmm6
+; SSE-NEXT: por %xmm1, %xmm6
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm2[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm1[2,0]
+; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[0,1,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
+; 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: pand %xmm11, %xmm1
+; SSE-NEXT: movdqa %xmm14, %xmm2
+; SSE-NEXT: movdqa %xmm14, %xmm10
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm5[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm5[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1,1,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,1,1,1,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT: movdqa %xmm11, %xmm14
+; SSE-NEXT: pandn %xmm2, %xmm14
+; SSE-NEXT: por %xmm1, %xmm14
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm2[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,1],xmm1[2,0]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[0,1,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm9[2],xmm1[3],xmm9[3]
+; SSE-NEXT: pand %xmm11, %xmm1
+; SSE-NEXT: movdqa %xmm12, %xmm2
+; SSE-NEXT: movdqa %xmm12, %xmm15
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm4[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm4[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1,1,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,1,1,1,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT: movdqa %xmm11, %xmm12
+; SSE-NEXT: pandn %xmm2, %xmm12
+; SSE-NEXT: por %xmm1, %xmm12
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm2[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[0,1],xmm1[2,0]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps (%rsp), %xmm2 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[1,1,1,1]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm2[2,2,2,2,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,4,6,7]
+; SSE-NEXT: movdqa %xmm11, %xmm2
+; SSE-NEXT: pandn %xmm1, %xmm2
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm3[1,0,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm11, %xmm4
+; SSE-NEXT: por %xmm2, %xmm4
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,6]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[3,1],xmm1[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,0]
+; SSE-NEXT: movaps %xmm0, %xmm1
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
+; 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 {{.*#+}} xmm3 = xmm7[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm2[2,2,2,2,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,4,6,7]
+; SSE-NEXT: movdqa %xmm11, %xmm2
+; SSE-NEXT: pandn %xmm1, %xmm2
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,0,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm11, %xmm3
+; SSE-NEXT: por %xmm2, %xmm3
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,6]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[3,1],xmm1[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[2,0],xmm2[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm10[0,2]
+; 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 {{.*#+}} xmm0 = xmm8[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm2[2,2,2,2,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,4,6,7]
+; SSE-NEXT: movdqa %xmm11, %xmm8
+; SSE-NEXT: pandn %xmm1, %xmm8
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm0[1,0,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm11, %xmm2
+; SSE-NEXT: por %xmm8, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,5,6]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[2,0],xmm1[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm15[0,2]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm8 = xmm1[0,1,2,3,7,4,6,7]
+; SSE-NEXT: movdqa %xmm11, %xmm1
+; SSE-NEXT: pandn %xmm8, %xmm1
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm11, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm8[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
+; SSE-NEXT: pshufd $85, (%rsp), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT: # xmm15 = mem[0,2,2,3]
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[0,1,1,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm15 = xmm15[0,1,0,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm15[1,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm8[0],xmm15[1],xmm8[1]
+; SSE-NEXT: movdqa %xmm11, %xmm8
+; SSE-NEXT: pandn %xmm15, %xmm8
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[2,0],xmm5[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm7[0,2]
+; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm5[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm10 = xmm7[2,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm11, %xmm10
+; SSE-NEXT: por %xmm8, %xmm10
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,1],xmm1[2,0]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[0,2,2,3]
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[0,1,1,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[0,1,0,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm7[1,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT: movdqa %xmm11, %xmm8
+; SSE-NEXT: pandn %xmm7, %xmm8
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm7[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[0,1],xmm5[0,2]
+; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm15 = xmm7[2,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm11, %xmm15
+; SSE-NEXT: por %xmm8, %xmm15
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[0,1],xmm1[2,0]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[0,2,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[0,1,1,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[0,1,0,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm7[1,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT: movdqa %xmm11, %xmm8
+; SSE-NEXT: pandn %xmm7, %xmm8
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm9[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,1],xmm5[0,2]
+; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm9[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm7[2,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm11, %xmm9
+; SSE-NEXT: por %xmm8, %xmm9
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,1],xmm1[2,0]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[0,2,2,3]
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[0,1,1,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[0,1,0,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm7[1,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm6[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm5[0,2]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[2,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm11, %xmm6
+; SSE-NEXT: pandn %xmm7, %xmm11
+; SSE-NEXT: por %xmm6, %xmm11
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,1],xmm1[2,0]
+; 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, (%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, 16(%rsi)
+; SSE-NEXT: movdqa %xmm13, 32(%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, 48(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 16(%rdx)
+; SSE-NEXT: movaps %xmm12, 32(%rcx)
+; SSE-NEXT: movaps %xmm14, (%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, 16(%rcx)
+; SSE-NEXT: movaps %xmm0, 32(%r8)
+; SSE-NEXT: movaps %xmm2, (%r8)
+; SSE-NEXT: movaps %xmm3, 48(%r8)
+; SSE-NEXT: movaps %xmm4, 16(%r8)
+; SSE-NEXT: movaps %xmm11, (%r9)
+; SSE-NEXT: movaps %xmm9, 32(%r9)
+; SSE-NEXT: movaps %xmm15, 48(%r9)
+; SSE-NEXT: movaps %xmm10, 16(%r9)
+; SSE-NEXT: addq $344, %rsp # imm = 0x158
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: vf32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: subq $424, %rsp # imm = 0x1A8
+; AVX1-NEXT: vmovaps 224(%rdi), %xmm0
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-NEXT: vmovdqa 256(%rdi), %xmm1
+; AVX1-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; AVX1-NEXT: vmovdqa 272(%rdi), %xmm2
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX1-NEXT: vmovdqa 240(%rdi), %xmm2
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3,4,5,6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: vmovaps {{.*#+}} ymm8 = [65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,65535,65535,65535]
+; AVX1-NEXT: vandnps %ymm0, %ymm8, %ymm0
+; AVX1-NEXT: vmovdqa 208(%rdi), %xmm1
+; AVX1-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; AVX1-NEXT: vmovdqa 192(%rdi), %xmm2
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4],xmm1[5,6,7]
+; AVX1-NEXT: vmovdqa 176(%rdi), %xmm2
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[3,1,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
+; AVX1-NEXT: vmovdqa 160(%rdi), %xmm2
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm2[0,2,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
+; AVX1-NEXT: vandps %ymm1, %ymm8, %ymm1
+; AVX1-NEXT: vorps %ymm0, %ymm1, %ymm1
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT: vmovdqa 304(%rdi), %xmm3
+; AVX1-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa 288(%rdi), %xmm2
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm2[0,1],xmm3[2,3],xmm2[4,5,6,7]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm13 = [6,7,2,3,4,5,6,7,6,7,2,3,12,13,6,7]
+; AVX1-NEXT: vpshufb %xmm13, %xmm3, %xmm3
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm3[5,6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovdqa 96(%rdi), %xmm11
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm11[0,1,1,3]
+; AVX1-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
+; AVX1-NEXT: vmovdqa 112(%rdi), %xmm9
+; AVX1-NEXT: vpunpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm9[1]
+; AVX1-NEXT: vmovdqa 80(%rdi), %xmm10
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm10[0,2,2,3]
+; AVX1-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3,4,5,6,7]
+; AVX1-NEXT: vmovdqa 64(%rdi), %xmm2
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm2[0,1,0,1]
+; AVX1-NEXT: vmovdqa %xmm2, %xmm12
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm3
+; AVX1-NEXT: vmovdqa 16(%rdi), %xmm0
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
+; AVX1-NEXT: vmovdqa %xmm0, %xmm7
+; AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; AVX1-NEXT: vmovdqa (%rdi), %xmm0
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
+; AVX1-NEXT: vmovdqa %xmm0, %xmm4
+; AVX1-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-NEXT: vmovdqa 32(%rdi), %xmm6
+; AVX1-NEXT: vmovdqa 48(%rdi), %xmm0
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,0,3]
+; AVX1-NEXT: vmovdqa %xmm0, %xmm5
+; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm6[4],xmm2[5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm2[4,5,6,7]
+; AVX1-NEXT: vmovdqa 144(%rdi), %xmm2
+; AVX1-NEXT: vmovdqa 128(%rdi), %xmm1
+; AVX1-NEXT: vpblendw {{.*#+}} xmm15 = xmm1[0,1],xmm2[2,3],xmm1[4,5,6,7]
+; AVX1-NEXT: vpshufb %xmm13, %xmm15, %xmm15
+; AVX1-NEXT: vandnps %ymm3, %ymm8, %ymm3
+; AVX1-NEXT: vandps %ymm0, %ymm8, %ymm0
+; AVX1-NEXT: vorps %ymm3, %ymm0, %ymm0
+; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm3
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm15[5,6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm9[0,1],xmm11[2,3],xmm9[4,5,6,7]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,0,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm10[0,3,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3,4,5,6,7]
+; AVX1-NEXT: vpsllq $48, %xmm12, %xmm3
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm4[0,3,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
+; AVX1-NEXT: vpsrlq $48, %xmm7, %xmm7
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm7 = xmm6[0,1],xmm5[2,3],xmm6[4,5],xmm5[6,7]
+; AVX1-NEXT: vmovdqa %xmm6, %xmm15
+; AVX1-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa %xmm5, %xmm10
+; AVX1-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,0,1,0,1,0,1,10,11,4,5,14,15,6,7]
+; AVX1-NEXT: vpshufb %xmm4, %xmm7, %xmm7
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm7[3,4,5,6,7]
+; AVX1-NEXT: vandnps %ymm0, %ymm8, %ymm0
+; AVX1-NEXT: vandps %ymm3, %ymm8, %ymm3
+; AVX1-NEXT: vorps %ymm0, %ymm3, %ymm3
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT: vpblendw {{.*#+}} xmm7 = xmm1[0,1,2,3],xmm2[4,5],xmm1[6,7]
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,6,7,8,9,4,5,14,15,8,9]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm7[5,6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6,7]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,0,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm14[0,3,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3,4,5,6,7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vpsllq $48, %xmm3, %xmm3
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-NEXT: vpblendw $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm3 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm3 = mem[0,1],xmm13[2,3],mem[4,5],xmm13[6,7]
+; AVX1-NEXT: vpshufb %xmm4, %xmm3, %xmm3
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm11[0,3,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[1,2,2,3,4,5,6,7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-NEXT: vpsrlq $48, %xmm12, %xmm6
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm6[0,1,2],xmm3[3,4,5,6,7]
+; AVX1-NEXT: vandnps %ymm0, %ymm8, %ymm0
+; AVX1-NEXT: vandps %ymm3, %ymm8, %ymm3
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm4[0,1,2,3],xmm8[4,5],xmm4[6,7]
+; AVX1-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6,7,8,9,4,5,14,15,8,9]
+; AVX1-NEXT: vorps %ymm0, %ymm3, %ymm3
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm0 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm0 = xmm9[0,1,2,3],mem[4,5],xmm9[6,7]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,u,u,8,9,2,3,12,13,12,13,12,13,12,13>
+; AVX1-NEXT: vpshufb %xmm7, %xmm0, %xmm0
+; AVX1-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm5 = mem[3,1,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[2,1,2,3,4,5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm5[0,1],xmm0[2,3,4,5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm2[0,1],xmm1[2,3],xmm2[4,5,6,7]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
+; AVX1-NEXT: vpshufb %xmm6, %xmm5, %xmm5
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm15[0,1],xmm10[2,3],xmm15[4,5,6,7]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm1 = [2,3,2,3,2,3,2,3,12,13,6,7,12,13,14,15]
+; AVX1-NEXT: vpshufb %xmm1, %xmm5, %xmm5
+; AVX1-NEXT: vpshufd $212, (%rsp), %xmm3 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm3 = mem[0,1,1,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
+; AVX1-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm5[3,4,5],xmm3[6,7]
+; AVX1-NEXT: vpshufd $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm5 = mem[0,1,2,0]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6,5]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm5[6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm8[0,1],xmm4[2,3],xmm8[4,5,6,7]
+; AVX1-NEXT: vmovdqa %xmm8, %xmm10
+; AVX1-NEXT: vmovdqa %xmm4, %xmm9
+; AVX1-NEXT: vpshufb %xmm6, %xmm0, %xmm0
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm6[0,1,2,3],xmm4[4,5],xmm6[6,7]
+; AVX1-NEXT: vpshufb %xmm7, %xmm3, %xmm3
+; AVX1-NEXT: vmovdqa %xmm14, %xmm2
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm14[3,1,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[2,1,2,3,4,5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3,4,5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm14 = xmm3[0,1,2,3,4],xmm0[5,6,7]
+; AVX1-NEXT: vmovdqa %xmm13, %xmm8
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm7[0,1],xmm13[2,3],xmm7[4,5,6,7]
+; AVX1-NEXT: vpshufb %xmm1, %xmm3, %xmm3
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm11[0,1,1,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,7]
+; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm5 = xmm5[2],xmm12[2],xmm5[3],xmm12[3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm5[0,1,2],xmm3[3,4,5],xmm5[6,7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm15[0,1,2,0]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6,5]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm5[6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm14, %ymm3, %ymm3
+; AVX1-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm4[0,1],xmm6[2,3],xmm4[4,5],xmm6[6,7]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,0,1,10,11,4,5,14,15,14,15,14,15,14,15>
+; AVX1-NEXT: vpshufb %xmm3, %xmm5, %xmm4
+; AVX1-NEXT: vpsrlq $48, %xmm2, %xmm5
+; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm5[0],xmm4[1,2,3,4,5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm10[0,1,2,3],xmm9[4,5],xmm10[6,7]
+; AVX1-NEXT: vmovdqa %xmm10, %xmm13
+; AVX1-NEXT: vmovdqa %xmm9, %xmm14
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
+; AVX1-NEXT: vpshufb %xmm6, %xmm5, %xmm5
+; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm5[5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm12[0,1],xmm11[2,3],xmm12[4,5,6,7]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm9 = [6,7,0,1,10,11,10,11,8,9,10,11,12,13,14,15]
+; AVX1-NEXT: vpshufb %xmm9, %xmm5, %xmm5
+; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm7[0,1,2,3],xmm8[4,5],xmm7[6,7]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,4,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3,4,5],xmm5[6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm15[0,1,0,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,5,6]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5],xmm5[6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm0
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm2[4,5],xmm1[6,7]
+; AVX1-NEXT: vpshufb %xmm6, %xmm0, %xmm0
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0,1],xmm12[2,3],xmm6[4,5],xmm12[6,7]
+; AVX1-NEXT: vpshufb %xmm3, %xmm4, %xmm3
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-NEXT: vpsrlq $48, %xmm5, %xmm4
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1,2,3,4,5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm8 = xmm3[0,1,2,3,4],xmm0[5,6,7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovdqa (%rsp), %xmm7 # 16-byte Reload
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm0[0,1],xmm7[2,3],xmm0[4,5,6,7]
+; AVX1-NEXT: vpshufb %xmm9, %xmm3, %xmm3
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm11[0,1,2,3],xmm10[4,5],xmm11[6,7]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,4,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3,4,5],xmm3[6,7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm9[0,1,0,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,5,6]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm4[6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm8, %ymm3, %ymm8
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm6[0,3,2,3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm12[3],xmm2[4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm5[2,3,2,3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3,4,5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm0[0,1,2,3],xmm7[4,5],xmm0[6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm11[1,1,1,1]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm10[0,2,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [8,9,2,3,12,13,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT: vpshufb %xmm4, %xmm2, %xmm2
+; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3,4,5],xmm2[6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm9[0,1,1,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm3[6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm13[3,1,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,2,1,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm14[0,2,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX1-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm3 = mem[0,3,2,3]
+; AVX1-NEXT: vpblendw $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm3 = xmm3[0,1,2],mem[3],xmm3[4,5,6,7]
+; AVX1-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm5 = mem[2,3,2,3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm5[0],xmm3[1,2,3,4,5,6,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpblendw $207, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm3 = mem[0,1,2,3],xmm0[4,5],mem[6,7]
+; AVX1-NEXT: vpshufb %xmm4, %xmm3, %xmm3
+; AVX1-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm4 = mem[1,1,1,1]
+; AVX1-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm5 = mem[0,2,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,0,3,4,5,6,7]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3,4,5],xmm3[6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm15[0,1,1,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm4[6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm3, (%rsi)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm3, 32(%rsi)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 32(%rdx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, (%rdx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, (%rcx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 32(%r8)
+; AVX1-NEXT: vmovaps %ymm8, (%r8)
+; AVX1-NEXT: vmovaps %ymm2, 32(%r9)
+; AVX1-NEXT: vmovaps %ymm1, (%r9)
+; AVX1-NEXT: addq $424, %rsp # imm = 0x1A8
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-SLOW-LABEL: vf32:
+; AVX2-SLOW: # %bb.0:
+; AVX2-SLOW-NEXT: subq $280, %rsp # imm = 0x118
+; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm4
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm5
+; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm9
+; AVX2-SLOW-NEXT: vmovdqa 192(%rdi), %ymm3
+; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm10
+; AVX2-SLOW-NEXT: vmovdqa 224(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %ymm2
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm2[1,2],ymm1[3],ymm2[4],ymm1[5],ymm2[6,7],ymm1[8],ymm2[9,10],ymm1[11],ymm2[12],ymm1[13],ymm2[14,15]
+; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm8
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa %ymm1, %ymm7
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6],ymm1[7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = [0,1,10,11,4,5,14,15,8,9,10,11,4,5,6,7,16,17,26,27,20,21,30,31,24,25,26,27,20,21,22,23]
+; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm10[0],ymm3[1],ymm10[2,3],ymm3[4],ymm10[5],ymm3[6],ymm10[7,8],ymm3[9],ymm10[10,11],ymm3[12],ymm10[13],ymm3[14],ymm10[15]
+; AVX2-SLOW-NEXT: vmovdqa %ymm10, %ymm13
+; AVX2-SLOW-NEXT: vmovdqu %ymm10, {{[-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: vextracti128 $1, %ymm1, %xmm3
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm3[1,2,3],xmm1[4,5],xmm3[6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm12 = [0,1,10,11,4,5,14,15,8,9,2,3,12,13,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm1, %xmm1
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]
+; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm15[0],ymm9[1,2],ymm15[3],ymm9[4],ymm15[5],ymm9[6,7],ymm15[8],ymm9[9,10],ymm15[11],ymm9[12],ymm15[13],ymm9[14,15]
+; AVX2-SLOW-NEXT: vmovdqa %ymm9, %ymm11
+; AVX2-SLOW-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm9 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm9[5],ymm0[6],ymm9[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0],ymm5[1],ymm4[2,3],ymm5[4],ymm4[5],ymm5[6],ymm4[7,8],ymm5[9],ymm4[10,11],ymm5[12],ymm4[13],ymm5[14],ymm4[15]
+; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm6
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1,2,3],xmm2[4,5],xmm3[6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm2, %xmm2
+; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm2, %ymm0, %ymm9
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm7[1],ymm8[2,3],ymm7[4],ymm8[5],ymm7[6],ymm8[7,8],ymm7[9],ymm8[10,11],ymm7[12],ymm8[13],ymm7[14],ymm8[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = [2,3,12,13,6,7,0,1,10,11,6,7,8,9,8,9,18,19,28,29,22,23,16,17,26,27,22,23,24,25,24,25]
+; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm14[0],ymm13[1],ymm14[2],ymm13[3],ymm14[4,5],ymm13[6],ymm14[7,8],ymm13[9],ymm14[10],ymm13[11],ymm14[12,13],ymm13[14],ymm14[15]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6],xmm4[7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm4 = [2,3,12,13,6,7,0,1,10,11,4,5,14,15,10,11]
+; AVX2-SLOW-NEXT: vpshufb %xmm4, %xmm3, %xmm3
+; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm3, %ymm0, %ymm12
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm11[0],ymm15[1],ymm11[2,3],ymm15[4],ymm11[5],ymm15[6],ymm11[7,8],ymm15[9],ymm11[10,11],ymm15[12],ymm11[13],ymm15[14],ymm11[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm0, %ymm3
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0],ymm6[1],ymm5[2],ymm6[3],ymm5[4,5],ymm6[6],ymm5[7,8],ymm6[9],ymm5[10],ymm6[11],ymm5[12,13],ymm6[14],ymm5[15]
+; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm13
+; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm14
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm0[0,1],xmm2[2,3],xmm0[4,5,6],xmm2[7]
+; AVX2-SLOW-NEXT: vmovdqa 304(%rdi), %xmm7
+; AVX2-SLOW-NEXT: vpshufb %xmm4, %xmm2, %xmm4
+; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %xmm8
+; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm4, %ymm3, %ymm4
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm8[0],xmm7[1],xmm8[2,3]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm10 = [6,7,2,3,4,5,6,7,6,7,2,3,12,13,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm3, %xmm3
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm1[0,1,2,3,4],ymm3[5,6,7],ymm1[8,9,10,11,12],ymm3[13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 144(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm11 = xmm0[0],xmm1[1],xmm0[2,3]
+; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm11, %xmm5
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm9[0,1,2,3,4],ymm5[5,6,7],ymm9[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm9[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm8[0,1],xmm7[2],xmm8[3]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm9 = [0,1,2,3,4,5,6,7,8,9,4,5,14,15,8,9]
+; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm5, %xmm5
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm12[0,1,2,3,4],ymm5[5,6,7],ymm12[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm12[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm0[0,1],xmm1[2],xmm0[3]
+; AVX2-SLOW-NEXT: vmovdqa %xmm1, %xmm10
+; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm5, %xmm5
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5],ymm1[6],ymm0[7,8],ymm1[9],ymm0[10,11],ymm1[12],ymm0[13],ymm1[14],ymm0[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6],ymm4[7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm12[0,1],ymm2[2],ymm12[3],ymm2[4],ymm12[5,6],ymm2[7],ymm12[8,9],ymm2[10],ymm12[11],ymm2[12],ymm12[13,14],ymm2[15]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm6[3,4],xmm5[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,0,1,10,11,20,21,30,31,24,25,18,19,28,29,26,27,16,17,26,27>
+; AVX2-SLOW-NEXT: vpshufb %ymm6, %ymm4, %ymm4
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm9 = <4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u>
+; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm5, %xmm5
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm7[0],xmm8[1],xmm7[2,3]
+; AVX2-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm5[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: vpblendw {{.*#+}} ymm5 = ymm15[0],ymm3[1],ymm15[2,3],ymm3[4],ymm15[5],ymm3[6],ymm15[7,8],ymm3[9],ymm15[10,11],ymm3[12],ymm15[13],ymm3[14],ymm15[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm11 = ymm5[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm11[6],ymm5[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm6, %ymm5, %ymm5
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm6 = ymm14[0,1],ymm13[2],ymm14[3],ymm13[4],ymm14[5,6],ymm13[7],ymm14[8,9],ymm13[10],ymm14[11],ymm13[12],ymm14[13,14],ymm13[15]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm6, %xmm4
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0,1,2],xmm4[3,4],xmm6[5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm4, %xmm4
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm10[0],xmm9[1],xmm10[2,3]
+; AVX2-SLOW-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4,5],ymm0[6],ymm1[7,8],ymm0[9],ymm1[10],ymm0[11],ymm1[12,13],ymm0[14],ymm1[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5],ymm5[6],ymm4[7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm2[0],ymm12[1,2],ymm2[3],ymm12[4],ymm2[5],ymm12[6,7],ymm2[8],ymm12[9,10],ymm2[11],ymm12[12],ymm2[13],ymm12[14,15]
+; AVX2-SLOW-NEXT: vmovdqa %ymm12, %ymm0
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2],xmm5[3]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,2,3,12,13,22,23,16,17,26,27,20,21,30,31,30,31,18,19,28,29>
+; AVX2-SLOW-NEXT: vpshufb %ymm6, %ymm4, %ymm4
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u>
+; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm5, %xmm5
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm7[0,1],xmm8[2],xmm7[3]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm11 = [0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
+; AVX2-SLOW-NEXT: vpshufb %xmm11, %xmm5, %xmm5
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm3[0],ymm15[1],ymm3[2],ymm15[3],ymm3[4,5],ymm15[6],ymm3[7,8],ymm15[9],ymm3[10],ymm15[11],ymm3[12,13],ymm15[14],ymm3[15]
+; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm12
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5],ymm5[6],ymm4[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm6, %ymm4, %ymm4
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm13[0],ymm14[1,2],ymm13[3],ymm14[4],ymm13[5],ymm14[6,7],ymm13[8],ymm14[9,10],ymm13[11],ymm14[12],ymm13[13],ymm14[14,15]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2],xmm5[3]
+; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm5, %xmm5
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqa %xmm9, %xmm3
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm5 = xmm10[0,1],xmm9[2],xmm10[3]
+; AVX2-SLOW-NEXT: vpshufb %xmm11, %xmm5, %xmm5
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm9 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = ymm1[0,1],mem[2],ymm1[3],mem[4],ymm1[5,6],mem[7],ymm1[8,9],mem[10],ymm1[11],mem[12],ymm1[13,14],mem[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5],mem[6],ymm0[7,8],mem[9],ymm0[10,11],mem[12],ymm0[13],mem[14],ymm0[15]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3,4],xmm6[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15,24,25,18,19,28,29,22,23,u,u,u,u,u,u,u,u>
+; AVX2-SLOW-NEXT: vpshufb %ymm6, %ymm4, %ymm4
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm5
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm7[3,1,2,3]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm8[0,2,2,3]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm12[0,1],ymm15[2],ymm12[3],ymm15[4],ymm12[5,6],ymm15[7],ymm12[8,9],ymm15[10],ymm12[11],ymm15[12],ymm12[13,14],ymm15[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4],ymm2[5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm6, %ymm2, %ymm2
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm14[0],ymm13[1],ymm14[2,3],ymm13[4],ymm14[5],ymm13[6],ymm14[7,8],ymm13[9],ymm14[10,11],ymm13[12],ymm14[13],ymm13[14],ymm14[15]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3,4],xmm5[5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm4, %xmm4
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3,4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm10[3,1,2,3]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%r8)
+; AVX2-SLOW-NEXT: vmovdqa %ymm9, (%r8)
+; AVX2-SLOW-NEXT: vmovdqa %ymm0, 32(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm1, (%r9)
+; AVX2-SLOW-NEXT: addq $280, %rsp # imm = 0x118
+; AVX2-SLOW-NEXT: vzeroupper
+; AVX2-SLOW-NEXT: retq
+;
+; AVX2-FAST-ALL-LABEL: vf32:
+; AVX2-FAST-ALL: # %bb.0:
+; AVX2-FAST-ALL-NEXT: subq $200, %rsp
+; AVX2-FAST-ALL-NEXT: vmovdqa 224(%rdi), %ymm12
+; AVX2-FAST-ALL-NEXT: vmovdqa 256(%rdi), %ymm2
+; AVX2-FAST-ALL-NEXT: vmovdqa 192(%rdi), %ymm14
+; AVX2-FAST-ALL-NEXT: vmovdqa 160(%rdi), %ymm3
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rdi), %ymm0
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rdi), %ymm4
+; AVX2-FAST-ALL-NEXT: vmovdqa 64(%rdi), %ymm6
+; AVX2-FAST-ALL-NEXT: vmovdqa 96(%rdi), %ymm7
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm8 = ymm6[0],ymm7[1,2],ymm6[3],ymm7[4],ymm6[5],ymm7[6,7],ymm6[8],ymm7[9,10],ymm6[11],ymm7[12],ymm6[13],ymm7[14,15]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm10 = [1,3,0,2,4,6,1,3]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm8, %ymm10, %ymm8
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm11 = [0,1,6,7,8,9,14,15,4,5,14,15,4,5,2,3,16,17,22,23,24,25,30,31,20,21,30,31,20,21,18,19]
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm11, %ymm8, %ymm9
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm8 = ymm0[0],ymm4[1],ymm0[2,3],ymm4[4],ymm0[5],ymm4[6],ymm0[7,8],ymm4[9],ymm0[10,11],ymm4[12],ymm0[13],ymm4[14],ymm0[15]
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm0, %ymm5
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vextracti128 $1, %ymm8, %xmm0
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} xmm0 = xmm8[0],xmm0[1,2,3],xmm8[4,5],xmm0[6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} xmm13 = [0,1,10,11,4,5,14,15,8,9,2,3,12,13,6,7]
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm13, %xmm0, %xmm0
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm0, %ymm9, %ymm9
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm14[1],ymm3[2,3],ymm14[4],ymm3[5],ymm14[6],ymm3[7,8],ymm14[9],ymm3[10,11],ymm14[12],ymm3[13],ymm14[14],ymm3[15]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3],xmm0[4,5],xmm1[6,7]
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm13, %xmm0, %xmm0
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm1 = ymm12[0],ymm2[1,2],ymm12[3],ymm2[4],ymm12[5],ymm2[6,7],ymm12[8],ymm2[9,10],ymm12[11],ymm2[12],ymm12[13],ymm2[14,15]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm1, %ymm10, %ymm1
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm11, %ymm1, %ymm1
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm0, %ymm1, %ymm10
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm0 = ymm7[0],ymm6[1],ymm7[2,3],ymm6[4],ymm7[5],ymm6[6],ymm7[7,8],ymm6[9],ymm7[10,11],ymm6[12],ymm7[13],ymm6[14],ymm7[15]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm1 = <2,u,u,u,4,7,1,6>
+; AVX2-FAST-ALL-NEXT: vpermd %ymm0, %ymm1, %ymm0
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm11 = [2,3,4,5,10,11,0,1,14,15,2,3,12,13,0,1,18,19,20,21,26,27,16,17,30,31,18,19,28,29,16,17]
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm11, %ymm0, %ymm0
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm13 = ymm4[0],ymm5[1],ymm4[2],ymm5[3],ymm4[4,5],ymm5[6],ymm4[7,8],ymm5[9],ymm4[10],ymm5[11],ymm4[12,13],ymm5[14],ymm4[15]
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm4, %ymm5
+; AVX2-FAST-ALL-NEXT: vextracti128 $1, %ymm13, %xmm1
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3],xmm13[4,5,6],xmm1[7]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} xmm13 = [2,3,12,13,6,7,0,1,10,11,4,5,14,15,10,11]
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm13, %xmm1, %xmm1
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm1, %ymm0, %ymm0
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm1 = ymm14[0],ymm3[1],ymm14[2],ymm3[3],ymm14[4,5],ymm3[6],ymm14[7,8],ymm3[9],ymm14[10],ymm3[11],ymm14[12,13],ymm3[14],ymm14[15]
+; AVX2-FAST-ALL-NEXT: vextracti128 $1, %ymm1, %xmm15
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,3],xmm1[4,5,6],xmm15[7]
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm13, %xmm1, %xmm1
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm13 = ymm2[0],ymm12[1],ymm2[2,3],ymm12[4],ymm2[5],ymm12[6],ymm2[7,8],ymm12[9],ymm2[10,11],ymm12[12],ymm2[13],ymm12[14],ymm2[15]
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm12, %ymm4
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm2, %ymm12
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm2 = <2,u,u,u,4,7,1,6>
+; AVX2-FAST-ALL-NEXT: vpermd %ymm13, %ymm2, %ymm15
+; AVX2-FAST-ALL-NEXT: vmovdqa 128(%rdi), %ymm13
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm11, %ymm15, %ymm11
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm2 = [0,3,1,3,0,3,5,7]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm1, %ymm11, %ymm1
+; AVX2-FAST-ALL-NEXT: vpermd %ymm13, %ymm2, %ymm8
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27>
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm15, %ymm8, %ymm8
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm8 = ymm9[0,1,2,3,4],ymm8[5,6,7],ymm9[8,9,10,11,12],ymm8[13,14,15]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa 288(%rdi), %ymm11
+; AVX2-FAST-ALL-NEXT: vpermd %ymm11, %ymm2, %ymm2
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm15, %ymm2, %ymm2
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm2 = ymm10[0,1,2,3,4],ymm2[5,6,7],ymm10[8,9,10,11,12],ymm2[13,14,15]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm2 = [1,3,2,3,1,3,6,7]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm13, %ymm2, %ymm9
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25>
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm3, %ymm9, %ymm9
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm9 = ymm0[0,1,2,3,4],ymm9[5,6,7],ymm0[8,9,10,11,12],ymm9[13,14,15]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vpermd %ymm11, %ymm2, %ymm0
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm3, %ymm0, %ymm0
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7],ymm1[8,9,10,11,12],ymm0[13,14,15]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm1 = ymm5[0,1],ymm8[2],ymm5[3],ymm8[4],ymm5[5,6],ymm8[7],ymm5[8,9],ymm8[10],ymm5[11],ymm8[12],ymm5[13,14],ymm8[15]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3,4],xmm1[5,6,7]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm2 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5],ymm7[6],ymm6[7,8],ymm7[9],ymm6[10,11],ymm7[12],ymm6[13],ymm7[14],ymm6[15]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm9 = <0,2,u,u,5,7,2,4>
+; AVX2-FAST-ALL-NEXT: vpermd %ymm2, %ymm9, %ymm2
+; AVX2-FAST-ALL-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,20,21,22,23,16,17,22,23]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} xmm0 = <4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u>
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm0, %xmm1, %xmm1
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-ALL-NEXT: vbroadcasti128 {{.*#+}} ymm2 = [1,4,6,0,1,4,6,0]
+; AVX2-FAST-ALL-NEXT: # ymm2 = mem[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm13, %ymm2, %ymm10
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm15, %ymm10, %ymm10
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm10 = ymm1[0,1,2,3,4],ymm10[5,6,7],ymm1[8,9,10,11,12],ymm10[13,14,15]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm10 = ymm14[0,1],ymm3[2],ymm14[3],ymm3[4],ymm14[5,6],ymm3[7],ymm14[8,9],ymm3[10],ymm14[11],ymm3[12],ymm14[13,14],ymm3[15]
+; AVX2-FAST-ALL-NEXT: vextracti128 $1, %ymm10, %xmm1
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} xmm1 = xmm10[0,1,2],xmm1[3,4],xmm10[5,6,7]
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm0, %xmm1, %xmm0
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm1 = ymm4[0],ymm12[1],ymm4[2,3],ymm12[4],ymm4[5],ymm12[6],ymm4[7,8],ymm12[9],ymm4[10,11],ymm12[12],ymm4[13],ymm12[14],ymm4[15]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm1, %ymm9, %ymm1
+; AVX2-FAST-ALL-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,20,21,22,23,16,17,22,23]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm11, %ymm2, %ymm1
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm15, %ymm1, %ymm1
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0,1,2,3,4],ymm1[5,6,7],ymm0[8,9,10,11,12],ymm1[13,14,15]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm5[1,2],ymm8[3],ymm5[4],ymm8[5],ymm5[6,7],ymm8[8],ymm5[9,10],ymm8[11],ymm5[12],ymm8[13],ymm5[14,15]
+; AVX2-FAST-ALL-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2],xmm0[3]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm1 = ymm7[0],ymm6[1],ymm7[2],ymm6[3],ymm7[4,5],ymm6[6],ymm7[7,8],ymm6[9],ymm7[10],ymm6[11],ymm7[12,13],ymm6[14],ymm7[15]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm10 = <0,3,u,u,5,0,2,7>
+; AVX2-FAST-ALL-NEXT: vpermd %ymm1, %ymm10, %ymm1
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm8 = <u,u,u,u,u,u,u,u,u,u,u,u,2,3,4,5,18,19,20,21,26,27,16,17,30,31,30,31,18,19,20,21>
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm8, %ymm1, %ymm1
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} xmm2 = <6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u>
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm2, %xmm0, %xmm0
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-ALL-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [2,4,7,0,2,4,7,0]
+; AVX2-FAST-ALL-NEXT: # ymm1 = mem[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm13, %ymm1, %ymm9
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25>
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm5, %ymm9, %ymm9
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm9 = ymm0[0,1,2,3,4],ymm9[5,6,7],ymm0[8,9,10,11,12],ymm9[13,14,15]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm14[1,2],ymm3[3],ymm14[4],ymm3[5],ymm14[6,7],ymm3[8],ymm14[9,10],ymm3[11],ymm14[12],ymm3[13],ymm14[14,15]
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm3, %ymm9
+; AVX2-FAST-ALL-NEXT: vextracti128 $1, %ymm0, %xmm15
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} xmm0 = xmm15[0],xmm0[1],xmm15[2],xmm0[3]
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm2, %xmm0, %xmm0
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm4, %ymm15
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm2 = ymm12[0],ymm4[1],ymm12[2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7,8],ymm4[9],ymm12[10],ymm4[11],ymm12[12,13],ymm4[14],ymm12[15]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm2, %ymm10, %ymm2
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm8, %ymm2, %ymm2
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm11, %ymm1, %ymm1
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm5, %ymm1, %ymm1
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0,1,2,3,4],ymm1[5,6,7],ymm0[8,9,10,11,12],ymm1[13,14,15]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm10 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm0 = ymm7[0,1],ymm6[2],ymm7[3],ymm6[4],ymm7[5,6],ymm6[7],ymm7[8,9],ymm6[10],ymm7[11],ymm6[12],ymm7[13,14],ymm6[15]
+; AVX2-FAST-ALL-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-ALL-NEXT: # ymm1 = mem[0],ymm1[1],mem[2,3],ymm1[4],mem[5],ymm1[6],mem[7,8],ymm1[9],mem[10,11],ymm1[12],mem[13],ymm1[14],mem[15]
+; AVX2-FAST-ALL-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4],xmm2[5,6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm2 = <1,3,u,u,6,0,3,5>
+; AVX2-FAST-ALL-NEXT: vpermd %ymm0, %ymm2, %ymm0
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,u,u,u,u,u,u,u,u>
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm3, %ymm0, %ymm0
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} xmm4 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm4, %xmm1, %xmm1
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm1 = [0,2,1,3,0,2,5,7]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm13, %ymm1, %ymm5
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25,30,31>
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm6, %ymm5, %ymm5
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm5 = ymm14[0],ymm9[1],ymm14[2,3],ymm9[4],ymm14[5],ymm9[6],ymm14[7,8],ymm9[9],ymm14[10,11],ymm9[12],ymm14[13],ymm9[14],ymm14[15]
+; AVX2-FAST-ALL-NEXT: vextracti128 $1, %ymm5, %xmm7
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} xmm5 = xmm7[0,1,2],xmm5[3,4],xmm7[5,6,7]
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm4, %xmm5, %xmm4
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm5 = ymm12[0,1],ymm15[2],ymm12[3],ymm15[4],ymm12[5,6],ymm15[7],ymm12[8,9],ymm15[10],ymm12[11],ymm15[12],ymm12[13,14],ymm15[15]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm5, %ymm2, %ymm2
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm3, %ymm2, %ymm2
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3,4,5],ymm4[6,7]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm11, %ymm1, %ymm1
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm6, %ymm1, %ymm1
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm2, 32(%rsi)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm2, (%rsi)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm2, 32(%rdx)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm2, (%rdx)
+; AVX2-FAST-ALL-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm2, 32(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm2, (%rcx)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm10, 32(%r8)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm2, (%r8)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm1, 32(%r9)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm0, (%r9)
+; AVX2-FAST-ALL-NEXT: addq $200, %rsp
+; AVX2-FAST-ALL-NEXT: vzeroupper
+; AVX2-FAST-ALL-NEXT: retq
+;
+; AVX2-FAST-PERLANE-LABEL: vf32:
+; AVX2-FAST-PERLANE: # %bb.0:
+; AVX2-FAST-PERLANE-NEXT: subq $296, %rsp # imm = 0x128
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm9
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 192(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm2[1,2],ymm1[3],ymm2[4],ymm1[5],ymm2[6,7],ymm1[8],ymm2[9,10],ymm1[11],ymm2[12],ymm1[13],ymm2[14,15]
+; 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: vmovdqa %ymm1, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = [0,1,10,11,4,5,14,15,8,9,10,11,4,5,6,7,16,17,26,27,20,21,30,31,24,25,26,27,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5],ymm3[6],ymm4[7,8],ymm3[9],ymm4[10,11],ymm3[12],ymm4[13],ymm3[14],ymm4[15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm3[1,2,3],xmm1[4,5],xmm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,10,11,4,5,14,15,8,9,2,3,12,13,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm12
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm14 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm12, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm12 = ymm7[0],ymm9[1,2],ymm7[3],ymm9[4],ymm7[5],ymm9[6,7],ymm7[8],ymm9[9,10],ymm7[11],ymm9[12],ymm7[13],ymm9[14,15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm13 = ymm12[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm13[5],ymm12[6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm12, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm12 = ymm15[0],ymm8[1],ymm15[2,3],ymm8[4],ymm15[5],ymm8[6],ymm15[7,8],ymm8[9],ymm15[10,11],ymm8[12],ymm15[13],ymm8[14],ymm15[15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm12, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm12[0],xmm1[1,2,3],xmm12[4,5],xmm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm1, %ymm2, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm5[0],ymm10[1],ymm5[2,3],ymm10[4],ymm5[5],ymm10[6],ymm5[7,8],ymm10[9],ymm5[10,11],ymm10[12],ymm5[13],ymm10[14],ymm5[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = [2,3,12,13,6,7,0,1,10,11,6,7,8,9,8,9,18,19,28,29,22,23,16,17,26,27,22,23,24,25,24,25]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm4[0],ymm11[1],ymm4[2],ymm11[3],ymm4[4,5],ymm11[6],ymm4[7,8],ymm11[9],ymm4[10],ymm11[11],ymm4[12,13],ymm11[14],ymm4[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6],xmm4[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm4 = [2,3,12,13,6,7,0,1,10,11,4,5,14,15,10,11]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm3, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm3, %ymm1, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm9[0],ymm7[1],ymm9[2,3],ymm7[4],ymm9[5],ymm7[6],ymm9[7,8],ymm7[9],ymm9[10,11],ymm7[12],ymm9[13],ymm7[14],ymm9[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm8[0],ymm15[1],ymm8[2],ymm15[3],ymm8[4,5],ymm15[6],ymm8[7,8],ymm15[9],ymm8[10],ymm15[11],ymm8[12,13],ymm15[14],ymm8[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm2[0,1],xmm3[2,3],xmm2[4,5,6],xmm3[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 304(%rdi), %xmm10
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm3, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm4, %ymm1, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm12[0],xmm10[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm14 = [6,7,2,3,4,5,6,7,6,7,2,3,12,13,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0,1,2,3,4],ymm1[5,6,7],ymm0[8,9,10,11,12],ymm1[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 144(%rdi), %xmm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm11 = xmm9[0],xmm8[1],xmm9[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm11, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6,7],ymm6[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm12[0,1],xmm10[2],xmm12[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm11 = [0,1,2,3,4,5,6,7,8,9,4,5,14,15,8,9]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm13[0,1,2,3,4],ymm5[5,6,7],ymm13[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm9[0,1],xmm8[2],xmm9[3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm13[0],ymm0[1],ymm13[2,3],ymm0[4],ymm13[5],ymm0[6],ymm13[7,8],ymm0[9],ymm13[10,11],ymm0[12],ymm13[13],ymm0[14],ymm13[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6],ymm4[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), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm1[0,1],ymm14[2],ymm1[3],ymm14[4],ymm1[5,6],ymm14[7],ymm1[8,9],ymm14[10],ymm1[11],ymm14[12],ymm1[13,14],ymm14[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm6[3,4],xmm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,0,1,10,11,20,21,30,31,24,25,18,19,28,29,26,27,16,17,26,27>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm6, %ymm4, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm11 = <4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm10[0],xmm12[1],xmm10[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm5[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), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm7[0],ymm3[1],ymm7[2,3],ymm3[4],ymm7[5],ymm3[6],ymm7[7,8],ymm3[9],ymm7[10,11],ymm3[12],ymm7[13],ymm3[14],ymm7[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm6, %ymm4, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm7[0,1],ymm15[2],ymm7[3],ymm15[4],ymm7[5,6],ymm15[7],ymm7[8,9],ymm15[10],ymm7[11],ymm15[12],ymm7[13,14],ymm15[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm6[3,4],xmm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm8[0],xmm9[1],xmm8[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm0[0],ymm13[1],ymm0[2],ymm13[3],ymm0[4,5],ymm13[6],ymm0[7,8],ymm13[9],ymm0[10],ymm13[11],ymm0[12,13],ymm13[14],ymm0[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm6 = ymm5[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4],ymm5[5],ymm6[6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm6 = ymm14[0],ymm1[1,2],ymm14[3],ymm1[4],ymm14[5],ymm1[6,7],ymm14[8],ymm1[9,10],ymm14[11],ymm1[12],ymm14[13],ymm1[14,15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm6, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0],xmm6[1],xmm4[2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,2,3,12,13,22,23,16,17,26,27,20,21,30,31,30,31,18,19,28,29>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm6, %ymm5, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm0 = <6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm10[0,1],xmm12[2],xmm10[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm13 = [0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm3[0],ymm2[1],ymm3[2],ymm2[3],ymm3[4,5],ymm2[6],ymm3[7,8],ymm2[9],ymm3[10],ymm2[11],ymm3[12,13],ymm2[14],ymm3[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm14 = ymm5[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm14[4],ymm5[5],ymm14[6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm6, %ymm5, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm15, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm6 = ymm15[0],ymm7[1,2],ymm15[3],ymm7[4],ymm15[5],ymm7[6,7],ymm15[8],ymm7[9,10],ymm15[11],ymm7[12],ymm15[13],ymm7[14,15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm6, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0],xmm6[1],xmm4[2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm8[0,1],xmm9[2],xmm8[3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm15 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm13 = [12,13,14,15,4,5,14,15,8,9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm8, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm5 = [0,1,2,3,0,1,10,11,8,9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm5, %xmm9, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0,1],ymm2[2],ymm3[3],ymm2[4],ymm3[5,6],ymm2[7],ymm3[8,9],ymm2[10],ymm3[11],ymm2[12],ymm3[13,14],ymm2[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm6 = ymm1[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4],ymm1[5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm6 = ymm7[0],ymm14[1],ymm7[2,3],ymm14[4],ymm7[5],ymm14[6],ymm7[7,8],ymm14[9],ymm7[10,11],ymm14[12],ymm7[13],ymm14[14],ymm7[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm6, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm6 = xmm7[0,1,2],xmm6[3,4],xmm7[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,u,u,u,u,4,5,14,15,24,25,18,19,28,29,22,23,u,u,u,u,u,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm7, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm4 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm6, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1,2],ymm1[3,4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm1[2],mem[3],ymm1[4],mem[5,6],ymm1[7],mem[8,9],ymm1[10],mem[11],ymm1[12],mem[13,14],ymm1[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm6 = ymm1[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4],ymm1[5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm7, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0],ymm2[1],mem[2,3],ymm2[4],mem[5],ymm2[6],mem[7,8],ymm2[9],mem[10,11],ymm2[12],mem[13],ymm2[14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm6, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm6 = xmm7[0,1,2],xmm6[3,4],xmm7[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm6, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3,4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm10, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm5, %xmm12, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm15, (%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, 32(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, (%r9)
+; AVX2-FAST-PERLANE-NEXT: addq $296, %rsp # imm = 0x128
+; AVX2-FAST-PERLANE-NEXT: vzeroupper
+; AVX2-FAST-PERLANE-NEXT: retq
+;
+; AVX512-LABEL: vf32:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu64 256(%rdi), %zmm0
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm1
+; AVX512-NEXT: vmovdqu64 64(%rdi), %zmm2
+; AVX512-NEXT: vmovdqu64 128(%rdi), %zmm3
+; AVX512-NEXT: vmovdqu64 192(%rdi), %zmm4
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm5 = [16,21,26,31,36,41,46,51,56,61,0,0,0,1,6,11,16,21,26,31,36,41,46,51,56,61,0,0,0,1,6,11]
+; AVX512-NEXT: # zmm5 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm3, %zmm5
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm6 = <0,5,10,15,20,25,30,35,40,45,50,55,60,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm2, %zmm1, %zmm6
+; AVX512-NEXT: movl $67100672, %eax # imm = 0x3FFE000
+; AVX512-NEXT: kmovd %eax, %k1
+; AVX512-NEXT: vmovdqu16 %zmm5, %zmm6 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm5 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,34,39,44,49,54,59]
+; AVX512-NEXT: vpermi2w %zmm0, %zmm6, %zmm5
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm6 = [49,54,59,0,5,10,15,20,25,30,0,0,0,34,39,44,49,54,59,0,5,10,15,20,25,30,0,0,0,34,39,44]
+; AVX512-NEXT: # zmm6 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm3, %zmm4, %zmm6
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm7 = <1,6,11,16,21,26,31,36,41,46,51,56,61,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm2, %zmm1, %zmm7
+; AVX512-NEXT: vmovdqu16 %zmm6, %zmm7 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm6 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,35,40,45,50,55,60]
+; AVX512-NEXT: vpermi2w %zmm0, %zmm7, %zmm6
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm7 = [50,55,60,1,6,11,16,21,26,31,0,0,0,35,40,45,50,55,60,1,6,11,16,21,26,31,0,0,0,35,40,45]
+; AVX512-NEXT: # zmm7 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm3, %zmm4, %zmm7
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm8 = <34,39,44,49,54,59,0,5,10,15,20,25,30,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm1, %zmm2, %zmm8
+; AVX512-NEXT: vmovdqu16 %zmm7, %zmm8 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm7 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,36,41,46,51,56,61]
+; AVX512-NEXT: vpermi2w %zmm0, %zmm8, %zmm7
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [19,24,29,34,39,44,49,54,59,0,0,0,0,4,9,14,19,24,29,34,39,44,49,54,59,0,0,0,0,4,9,14]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm3, %zmm8
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm9 = <35,40,45,50,55,60,1,6,11,16,21,26,31,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm1, %zmm2, %zmm9
+; AVX512-NEXT: movl $33546240, %eax # imm = 0x1FFE000
+; AVX512-NEXT: kmovd %eax, %k1
+; AVX512-NEXT: vmovdqu16 %zmm8, %zmm9 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm8 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,32,37,42,47,52,57,62]
+; AVX512-NEXT: vpermi2w %zmm0, %zmm9, %zmm8
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm9 = <4,9,14,19,24,29,34,39,44,49,54,59,u,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm2, %zmm1, %zmm9
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm1 = [20,25,30,35,40,45,50,55,60,0,0,0,0,5,10,15,20,25,30,35,40,45,50,55,60,0,0,0,0,5,10,15]
+; AVX512-NEXT: # zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm3, %zmm1
+; AVX512-NEXT: movb $7, %al
+; AVX512-NEXT: kmovd %eax, %k1
+; AVX512-NEXT: vmovdqa64 %zmm9, %zmm1 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm2 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,33,38,43,48,53,58,63]
+; AVX512-NEXT: vpermi2w %zmm0, %zmm1, %zmm2
+; AVX512-NEXT: vmovdqu64 %zmm5, (%rsi)
+; AVX512-NEXT: vmovdqu64 %zmm6, (%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm7, (%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm8, (%r8)
+; AVX512-NEXT: vmovdqu64 %zmm2, (%r9)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %wide.vec = load <160 x i16>, <160 x i16>* %in.vec, align 32
+
+ %strided.vec0 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 0, i32 5, i32 10, i32 15, i32 20, i32 25, i32 30, i32 35, i32 40, i32 45, i32 50, i32 55, i32 60, i32 65, i32 70, i32 75, i32 80, i32 85, i32 90, i32 95, i32 100, i32 105, i32 110, i32 115, i32 120, i32 125, i32 130, i32 135, i32 140, i32 145, i32 150, i32 155>
+ %strided.vec1 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 1, i32 6, i32 11, i32 16, i32 21, i32 26, i32 31, i32 36, i32 41, i32 46, i32 51, i32 56, i32 61, i32 66, i32 71, i32 76, i32 81, i32 86, i32 91, i32 96, i32 101, i32 106, i32 111, i32 116, i32 121, i32 126, i32 131, i32 136, i32 141, i32 146, i32 151, i32 156>
+ %strided.vec2 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 2, i32 7, i32 12, i32 17, i32 22, i32 27, i32 32, i32 37, i32 42, i32 47, i32 52, i32 57, i32 62, i32 67, i32 72, i32 77, i32 82, i32 87, i32 92, i32 97, i32 102, i32 107, i32 112, i32 117, i32 122, i32 127, i32 132, i32 137, i32 142, i32 147, i32 152, i32 157>
+ %strided.vec3 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 3, i32 8, i32 13, i32 18, i32 23, i32 28, i32 33, i32 38, i32 43, i32 48, i32 53, i32 58, i32 63, i32 68, i32 73, i32 78, i32 83, i32 88, i32 93, i32 98, i32 103, i32 108, i32 113, i32 118, i32 123, i32 128, i32 133, i32 138, i32 143, i32 148, i32 153, i32 158>
+ %strided.vec4 = shufflevector <160 x i16> %wide.vec, <160 x i16> poison, <32 x i32> <i32 4, i32 9, i32 14, i32 19, i32 24, i32 29, i32 34, i32 39, i32 44, i32 49, i32 54, i32 59, i32 64, i32 69, i32 74, i32 79, i32 84, i32 89, i32 94, i32 99, i32 104, i32 109, i32 114, i32 119, i32 124, i32 129, i32 134, i32 139, i32 144, i32 149, i32 154, i32 159>
+
+ store <32 x i16> %strided.vec0, <32 x i16>* %out.vec0, align 32
+ store <32 x i16> %strided.vec1, <32 x i16>* %out.vec1, align 32
+ store <32 x i16> %strided.vec2, <32 x i16>* %out.vec2, align 32
+ store <32 x i16> %strided.vec3, <32 x i16>* %out.vec3, align 32
+ store <32 x i16> %strided.vec4, <32 x i16>* %out.vec4, align 32
+
+ ret void
+}
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
index f0052e24e59fc..144b1b12f0eab 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
@@ -1650,3 +1650,1836 @@ define void @vf16(<96 x i16>* %in.vec, <16 x i16>* %out.vec0, <16 x i16>* %out.v
ret void
}
+
+define void @vf32(<192 x i16>* %in.vec, <32 x i16>* %out.vec0, <32 x i16>* %out.vec1, <32 x i16>* %out.vec2, <32 x i16>* %out.vec3, <32 x i16>* %out.vec4, <32 x i16>* %out.vec5) nounwind {
+; SSE-LABEL: vf32:
+; SSE: # %bb.0:
+; SSE-NEXT: subq $456, %rsp # imm = 0x1C8
+; SSE-NEXT: movdqa 256(%rdi), %xmm3
+; SSE-NEXT: movdqa 272(%rdi), %xmm0
+; SSE-NEXT: movdqa 64(%rdi), %xmm2
+; SSE-NEXT: movdqa 80(%rdi), %xmm6
+; SSE-NEXT: movaps (%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 16(%rdi), %xmm14
+; SSE-NEXT: movdqa 32(%rdi), %xmm9
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 48(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrlq $16, %xmm9
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm10 = xmm1[0,1,0,2,4,5,6,7]
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm2[2,2,3,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1],xmm4[2],xmm6[2],xmm4[3],xmm6[3]
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm6, %xmm4
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm2[3,0]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm6[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm6[2,3]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pslld $16, %xmm6
+; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm10[1,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
+; SSE-NEXT: movdqa {{.*#+}} xmm12 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT: movdqa %xmm12, %xmm2
+; SSE-NEXT: pandn %xmm10, %xmm2
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm6[0,1,2,3,4,6,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm14[2],xmm4[3],xmm14[3]
+; SSE-NEXT: pand %xmm12, %xmm4
+; SSE-NEXT: por %xmm2, %xmm4
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm1[2,0]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm3, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm3[2,2,3,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
+; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm3[3,0]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm3, %xmm2
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,3]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pslld $16, %xmm0
+; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT: movdqa 224(%rdi), %xmm8
+; SSE-NEXT: movdqa %xmm8, %xmm0
+; SSE-NEXT: psrlq $16, %xmm0
+; SSE-NEXT: movdqa 240(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm1[0,1,0,2,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[3,1],xmm3[1,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: pandn %xmm3, %xmm0
+; SSE-NEXT: movdqa 208(%rdi), %xmm11
+; SSE-NEXT: movdqa 192(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm11[2],xmm1[3],xmm11[3]
+; SSE-NEXT: pand %xmm12, %xmm1
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 160(%rdi), %xmm1
+; SSE-NEXT: movdqa 176(%rdi), %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,2,3,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm1[3,0]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm1, %xmm3
+; SSE-NEXT: movaps %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,0],xmm2[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,3]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pslld $16, %xmm2
+; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE-NEXT: movdqa 128(%rdi), %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrlq $16, %xmm5
+; SSE-NEXT: movdqa 144(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm1[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm2[0,1,0,2,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm4[1,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
+; SSE-NEXT: movdqa %xmm12, %xmm5
+; SSE-NEXT: pandn %xmm4, %xmm5
+; SSE-NEXT: movdqa 112(%rdi), %xmm4
+; SSE-NEXT: movdqa %xmm4, {{[-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 {{.*#+}} xmm13 = xmm1[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm13[0,1,2,3,4,6,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT: pand %xmm12, %xmm1
+; SSE-NEXT: por %xmm5, %xmm1
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 352(%rdi), %xmm1
+; SSE-NEXT: movdqa 368(%rdi), %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[2,2,3,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; SSE-NEXT: movdqa %xmm4, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm1[3,0]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm1, %xmm7
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,0],xmm4[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,3]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pslld $16, %xmm4
+; SSE-NEXT: psrldq {{.*#+}} xmm7 = xmm7[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3]
+; SSE-NEXT: movdqa 320(%rdi), %xmm4
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrlq $16, %xmm4
+; SSE-NEXT: movdqa 336(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm1[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm15 = xmm10[0,1,0,2,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[3,1],xmm15[1,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm4[0],xmm15[1],xmm4[1]
+; SSE-NEXT: movdqa %xmm12, %xmm4
+; SSE-NEXT: pandn %xmm15, %xmm4
+; SSE-NEXT: movdqa 304(%rdi), %xmm15
+; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 288(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,6,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm15[2],xmm3[3],xmm15[3]
+; SSE-NEXT: pand %xmm12, %xmm3
+; SSE-NEXT: por %xmm4, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm7[2,0]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,1,1,3,4,5,6,7]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[3,1],xmm4[1,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm3[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
+; SSE-NEXT: movdqa %xmm12, %xmm7
+; SSE-NEXT: pandn %xmm4, %xmm7
+; SSE-NEXT: movdqa %xmm14, %xmm4
+; SSE-NEXT: psrld $16, %xmm4
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,7,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm4[2],xmm6[3],xmm4[3]
+; SSE-NEXT: pand %xmm12, %xmm6
+; SSE-NEXT: por %xmm7, %xmm6
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm15[2,0]
+; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,1,1,3,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[3,1],xmm4[1,3]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm8[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
+; SSE-NEXT: movdqa %xmm12, %xmm7
+; SSE-NEXT: pandn %xmm4, %xmm7
+; SSE-NEXT: movdqa %xmm11, %xmm4
+; SSE-NEXT: psrld $16, %xmm4
+; SSE-NEXT: pshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,2,3,5,7,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm4[2],xmm6[3],xmm4[3]
+; SSE-NEXT: pand %xmm12, %xmm6
+; SSE-NEXT: por %xmm7, %xmm6
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm9[2,0]
+; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1],xmm2[1,3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
+; SSE-NEXT: movdqa %xmm12, %xmm4
+; SSE-NEXT: pandn %xmm2, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm2
+; SSE-NEXT: pshufhw {{.*#+}} xmm15 = xmm13[0,1,2,3,5,7,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm2[2],xmm15[3],xmm2[3]
+; SSE-NEXT: pand %xmm12, %xmm15
+; SSE-NEXT: por %xmm4, %xmm15
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[0,1],xmm0[2,0]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm10[0,1,1,3,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[3,1],xmm0[1,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm6[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm10, %xmm2
+; SSE-NEXT: psrld $16, %xmm2
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,7,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE-NEXT: pand %xmm12, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm12
+; SSE-NEXT: por %xmm1, %xmm12
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[0,1],xmm5[2,0]
+; SSE-NEXT: movdqa %xmm3, %xmm0
+; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[1,1,1,1]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT: movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm1, %xmm2
+; SSE-NEXT: pandn %xmm0, %xmm2
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm14[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm14[2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm4[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm0[1,0,2,3,4,5,6,7]
+; SSE-NEXT: pand %xmm1, %xmm5
+; SSE-NEXT: por %xmm2, %xmm5
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,2]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,6,5,4]
+; SSE-NEXT: movdqa {{.*#+}} xmm14 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: pandn %xmm2, %xmm3
+; SSE-NEXT: pand %xmm14, %xmm5
+; SSE-NEXT: por %xmm5, %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrldq {{.*#+}} xmm8 = xmm8[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm9[1,1,1,1]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm5[0]
+; SSE-NEXT: movdqa %xmm1, %xmm7
+; SSE-NEXT: pandn %xmm8, %xmm7
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[1,0],xmm11[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm11[2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm5[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,0,2,3,4,5,6,7]
+; SSE-NEXT: pand %xmm1, %xmm2
+; SSE-NEXT: por %xmm7, %xmm2
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,2]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm0[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,6,5,4]
+; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: pandn %xmm7, %xmm3
+; SSE-NEXT: pand %xmm14, %xmm2
+; SSE-NEXT: por %xmm2, %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm6, %xmm2
+; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[1,1,1,1]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm7[0]
+; SSE-NEXT: movdqa %xmm1, %xmm7
+; SSE-NEXT: pandn %xmm2, %xmm7
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[1,0],xmm10[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[2,0],xmm10[2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm8[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,0,2,3,4,5,6,7]
+; SSE-NEXT: pand %xmm1, %xmm3
+; SSE-NEXT: por %xmm7, %xmm3
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,2]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm0[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,6,5,4]
+; SSE-NEXT: movdqa %xmm14, %xmm2
+; SSE-NEXT: pandn %xmm7, %xmm2
+; SSE-NEXT: pand %xmm14, %xmm3
+; SSE-NEXT: por %xmm3, %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm11, %xmm3
+; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm10[1,1,1,1]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm7[0]
+; SSE-NEXT: movdqa %xmm1, %xmm7
+; SSE-NEXT: pandn %xmm3, %xmm7
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm0[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm0[2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm3[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[1,0,2,3,4,5,6,7]
+; SSE-NEXT: pand %xmm1, %xmm6
+; SSE-NEXT: por %xmm7, %xmm6
+; SSE-NEXT: movaps (%rsp), %xmm2 # 16-byte Reload
+; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = xmm2[0,1],mem[0,2]
+; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm2[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,6,5,4]
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm7, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm6
+; SSE-NEXT: por %xmm6, %xmm0
+; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT: psrlq $48, %xmm13
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[2,2,3,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm7 = xmm7[0],xmm13[0]
+; SSE-NEXT: movdqa %xmm1, %xmm6
+; SSE-NEXT: pandn %xmm7, %xmm6
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm1, %xmm4
+; SSE-NEXT: por %xmm6, %xmm4
+; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,7,4,5]
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm6, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm4
+; SSE-NEXT: por %xmm4, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrlq $48, %xmm9
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[2,2,3,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm9[0]
+; SSE-NEXT: movdqa %xmm1, %xmm4
+; SSE-NEXT: pandn %xmm6, %xmm4
+; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[0,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm1, %xmm5
+; SSE-NEXT: por %xmm4, %xmm5
+; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,4,5]
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm4, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm5
+; SSE-NEXT: por %xmm5, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrlq $48, %xmm10
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm11[2,2,3,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm10[0]
+; SSE-NEXT: movdqa %xmm1, %xmm4
+; SSE-NEXT: pandn %xmm5, %xmm4
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm1, %xmm3
+; SSE-NEXT: por %xmm4, %xmm3
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,4,5]
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm4, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm3
+; SSE-NEXT: por %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm5, %xmm3
+; SSE-NEXT: psrlq $48, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm13[2,2,3,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm3[0]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm8[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm1, %xmm2
+; SSE-NEXT: pandn %xmm4, %xmm1
+; SSE-NEXT: por %xmm2, %xmm1
+; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,4,5]
+; SSE-NEXT: movdqa %xmm14, %xmm10
+; SSE-NEXT: pandn %xmm2, %xmm10
+; SSE-NEXT: pand %xmm14, %xmm1
+; SSE-NEXT: por %xmm1, %xmm10
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm6[2,3,2,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT: # xmm11 = mem[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm11[0,1,2,3,4,5,4,6]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm7[1]
+; SSE-NEXT: movss {{.*#+}} xmm3 = xmm2[0],xmm3[1,2,3]
+; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,6]
+; SSE-NEXT: movdqa %xmm14, %xmm9
+; SSE-NEXT: pandn %xmm2, %xmm9
+; SSE-NEXT: andps %xmm14, %xmm3
+; SSE-NEXT: por %xmm3, %xmm9
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,3,2,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[0,1,0,3]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,5,4,6]
+; SSE-NEXT: punpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = xmm2[1],mem[1]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3]
+; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
+; SSE-NEXT: movdqa %xmm14, %xmm8
+; SSE-NEXT: pandn %xmm3, %xmm8
+; SSE-NEXT: andps %xmm14, %xmm2
+; SSE-NEXT: por %xmm2, %xmm8
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,3,2,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[0,1,0,3]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,5,4,6]
+; SSE-NEXT: punpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = xmm2[1],mem[1]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3]
+; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
+; SSE-NEXT: movdqa %xmm14, %xmm4
+; SSE-NEXT: pandn %xmm3, %xmm4
+; SSE-NEXT: andps %xmm14, %xmm2
+; SSE-NEXT: por %xmm2, %xmm4
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,3,2,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[0,1,0,3]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,5,4,6]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm5[1]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3]
+; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,4,5,4,6]
+; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: pandn %xmm1, %xmm3
+; SSE-NEXT: andps %xmm14, %xmm2
+; SSE-NEXT: por %xmm2, %xmm3
+; SSE-NEXT: psrlq $48, %xmm0
+; SSE-NEXT: psrldq {{.*#+}} xmm6 = xmm6[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1],xmm6[2],xmm0[2],xmm6[3],xmm0[3]
+; SSE-NEXT: psrld $16, %xmm7
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm11[0,1,2,3,4,5,5,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm7[1]
+; SSE-NEXT: movss {{.*#+}} xmm1 = xmm6[0],xmm1[1,2,3]
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm2[0,1,2,3,4,4,5,7]
+; SSE-NEXT: movdqa %xmm14, %xmm2
+; SSE-NEXT: pandn %xmm6, %xmm2
+; SSE-NEXT: andps %xmm14, %xmm1
+; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm5
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT: movdqa %xmm1, %xmm5
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm1
+; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,2,3,4,5,5,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm1[1]
+; SSE-NEXT: movss {{.*#+}} xmm6 = xmm5[0],xmm6[1,2,3]
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm1[0,1,2,3,4,4,5,7]
+; SSE-NEXT: movdqa %xmm14, %xmm1
+; SSE-NEXT: pandn %xmm13, %xmm1
+; SSE-NEXT: andps %xmm14, %xmm6
+; SSE-NEXT: por %xmm6, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm5 = xmm5[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE-NEXT: movdqa %xmm5, %xmm7
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm5
+; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,2,3,4,5,5,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm5[1]
+; SSE-NEXT: movss {{.*#+}} xmm6 = xmm7[0],xmm6[1,2,3]
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm7[0,1,2,3,4,4,5,7]
+; SSE-NEXT: movdqa %xmm14, %xmm7
+; SSE-NEXT: pandn %xmm13, %xmm7
+; SSE-NEXT: andps %xmm14, %xmm6
+; SSE-NEXT: por %xmm6, %xmm7
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm6
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm5 = xmm5[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; SSE-NEXT: movdqa %xmm5, %xmm11
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm5
+; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,2,3,4,5,5,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm5[1]
+; SSE-NEXT: movss {{.*#+}} xmm6 = xmm11[0],xmm6[1,2,3]
+; SSE-NEXT: andps %xmm14, %xmm6
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,5,7]
+; SSE-NEXT: pandn %xmm5, %xmm14
+; SSE-NEXT: por %xmm6, %xmm14
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movaps %xmm5, 48(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movaps %xmm5, 16(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movaps %xmm5, 32(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movaps %xmm5, (%rsi)
+; SSE-NEXT: movaps %xmm12, 48(%rdx)
+; SSE-NEXT: movaps %xmm15, 16(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movaps %xmm5, 32(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movaps %xmm5, (%rdx)
+; SSE-NEXT: movaps (%rsp), %xmm5 # 16-byte Reload
+; SSE-NEXT: movaps %xmm5, 16(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movaps %xmm5, 48(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movaps %xmm5, 32(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movaps %xmm5, (%rcx)
+; SSE-NEXT: movdqa %xmm10, 48(%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, 32(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, (%r8)
+; SSE-NEXT: movdqa %xmm3, 48(%r9)
+; SSE-NEXT: movdqa %xmm4, 16(%r9)
+; SSE-NEXT: movdqa %xmm8, 32(%r9)
+; SSE-NEXT: movdqa %xmm9, (%r9)
+; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT: movdqa %xmm14, 48(%rax)
+; SSE-NEXT: movdqa %xmm7, 16(%rax)
+; SSE-NEXT: movdqa %xmm1, 32(%rax)
+; SSE-NEXT: movdqa %xmm2, (%rax)
+; SSE-NEXT: addq $456, %rsp # imm = 0x1C8
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: vf32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: subq $584, %rsp # imm = 0x248
+; AVX1-NEXT: vmovdqa 288(%rdi), %xmm0
+; AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
+; AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; AVX1-NEXT: vmovdqa 304(%rdi), %xmm1
+; AVX1-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT: vmovdqa 272(%rdi), %xmm1
+; AVX1-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpslld $16, %xmm1, %xmm1
+; AVX1-NEXT: vmovdqa 256(%rdi), %xmm2
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpsrldq {{.*#+}} xmm2 = xmm2[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT: vmovdqa 224(%rdi), %xmm1
+; AVX1-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpsrlq $16, %xmm1, %xmm1
+; AVX1-NEXT: vmovdqa 240(%rdi), %xmm2
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,0,2,4,5,6,7]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-NEXT: vmovdqa 192(%rdi), %xmm2
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
+; AVX1-NEXT: vmovdqa 208(%rdi), %xmm3
+; AVX1-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4,5],xmm2[6,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-NEXT: vmovdqa 368(%rdi), %xmm1
+; AVX1-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpslld $16, %xmm1, %xmm1
+; AVX1-NEXT: vmovdqa 352(%rdi), %xmm2
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpsrldq {{.*#+}} xmm2 = xmm2[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX1-NEXT: vmovdqa 320(%rdi), %xmm2
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpsrlq $16, %xmm2, %xmm2
+; AVX1-NEXT: vmovdqa 336(%rdi), %xmm3
+; AVX1-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; AVX1-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm4 = xmm3[0,1,0,2,4,5,6,7]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5],xmm1[6,7]
+; AVX1-NEXT: vmovaps {{.*#+}} ymm2 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0]
+; AVX1-NEXT: vandps %ymm2, %ymm0, %ymm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT: vandnps %ymm1, %ymm2, %ymm1
+; AVX1-NEXT: vmovaps %ymm2, %ymm9
+; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovdqa 80(%rdi), %xmm12
+; AVX1-NEXT: vpslld $16, %xmm12, %xmm0
+; AVX1-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa 64(%rdi), %xmm11
+; AVX1-NEXT: vpsrldq {{.*#+}} xmm1 = xmm11[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX1-NEXT: vmovdqa 96(%rdi), %xmm0
+; AVX1-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[0,1,0,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,6,6,7]
+; AVX1-NEXT: vmovdqa 112(%rdi), %xmm14
+; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm14[2],xmm2[3],xmm14[3]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm15
+; AVX1-NEXT: vmovdqa 32(%rdi), %xmm13
+; AVX1-NEXT: vmovdqa 48(%rdi), %xmm0
+; AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpsrlq $16, %xmm13, %xmm2
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[0,3,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm5 = xmm1[0,1,0,2,4,5,6,7]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; AVX1-NEXT: vmovdqa (%rdi), %xmm0
+; AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa 16(%rdi), %xmm8
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,0,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm7 = xmm2[0,1,2,3,4,6,6,7]
+; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm7 = xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm7[0,1,2],xmm5[3,4,5],xmm7[6,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm15[3,4,5],ymm5[6,7]
+; AVX1-NEXT: vmovdqa 176(%rdi), %xmm0
+; AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpslld $16, %xmm0, %xmm5
+; AVX1-NEXT: vmovdqa 160(%rdi), %xmm10
+; AVX1-NEXT: vpsrldq {{.*#+}} xmm7 = xmm10[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm7[0],xmm5[0],xmm7[1],xmm5[1],xmm7[2],xmm5[2],xmm7[3],xmm5[3]
+; AVX1-NEXT: vmovdqa 128(%rdi), %xmm6
+; AVX1-NEXT: vpsrlq $16, %xmm6, %xmm7
+; AVX1-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa 144(%rdi), %xmm0
+; AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm15 = xmm0[0,1,0,2,4,5,6,7]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm15[0],xmm7[0],xmm15[1],xmm7[1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm7[0,1,2,3,4,5],xmm5[6,7]
+; AVX1-NEXT: vandps %ymm4, %ymm9, %ymm4
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-NEXT: vandnps %ymm5, %ymm9, %ymm5
+; AVX1-NEXT: vmovaps %ymm9, %ymm7
+; AVX1-NEXT: vorps %ymm5, %ymm4, %ymm4
+; AVX1-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,7,6,7]
+; AVX1-NEXT: vpsrld $16, %xmm14, %xmm4
+; AVX1-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm11[2,2,3,3]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm12[0],xmm4[1],xmm12[1],xmm4[2],xmm12[2],xmm4[3],xmm12[3]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm3
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
+; AVX1-NEXT: vmovdqa %xmm13, %xmm11
+; AVX1-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm13[1,1,1,1]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,7,6,7]
+; AVX1-NEXT: vpsrld $16, %xmm8, %xmm4
+; AVX1-NEXT: vmovdqa %xmm8, %xmm9
+; AVX1-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4,5],xmm2[6,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5],ymm1[6,7]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm6[1,1,1,1]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm10[2,2,3,3]
+; AVX1-NEXT: vmovdqa %xmm10, %xmm13
+; AVX1-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm2[6,7]
+; AVX1-NEXT: vandps %ymm7, %ymm1, %ymm1
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vandnps %ymm0, %ymm7, %ymm0
+; AVX1-NEXT: vmovaps %ymm7, %ymm4
+; AVX1-NEXT: vorps %ymm0, %ymm1, %ymm0
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vpshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm0 = mem[0,1,2,3,5,7,6,7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-NEXT: vpsrld $16, %xmm7, %xmm1
+; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm1 = mem[2,2,3,3]
+; AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT: vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm1 = mem[0,1,1,3,4,5,6,7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm15[1,1,1,1]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX1-NEXT: vpshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm2 = mem[0,1,2,3,5,7,6,7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-NEXT: vpsrld $16, %xmm5, %xmm3
+; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4,5],xmm2[6,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
+; AVX1-NEXT: vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm1 = mem[0,1,1,3,4,5,6,7]
+; AVX1-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm2 = mem[1,1,1,1]
+; AVX1-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm6[2,2,3,3]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm2[6,7]
+; AVX1-NEXT: vandps %ymm4, %ymm0, %ymm0
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT: vandnps %ymm1, %ymm4, %ymm1
+; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm0 = mem[1,1,1,1]
+; AVX1-NEXT: vpsrldq {{.*#+}} xmm1 = xmm11[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm0 = [4,5,0,1,12,13,14,15,8,9,10,11,12,13,14,15]
+; AVX1-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm2 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm2 = xmm9[0,1],mem[2,3],xmm9[4,5,6,7]
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufb %xmm0, %xmm2, %xmm2
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4],xmm2[5,6,7]
+; AVX1-NEXT: vpblendw $12, (%rsp), %xmm14, %xmm2 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm2 = xmm14[0,1],mem[2,3],xmm14[4,5,6,7]
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufb %xmm0, %xmm2, %xmm2
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm11 = [0,1,2,3,4,5,6,7,8,9,0,1,12,13,8,9]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vpblendw $207, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm3 = mem[0,1,2,3],xmm3[4,5],mem[6,7]
+; AVX1-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufb %xmm11, %xmm3, %xmm3
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2
+; AVX1-NEXT: vmovaps {{.*#+}} ymm12 = [0,0,0,0,0,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX1-NEXT: vandnps %ymm1, %ymm12, %ymm1
+; AVX1-NEXT: vandps %ymm2, %ymm12, %ymm2
+; AVX1-NEXT: vorps %ymm1, %ymm2, %ymm1
+; AVX1-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm2 = mem[1,1,1,1]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vpsrldq {{.*#+}} xmm4 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm4[0],xmm2[0]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm13[0,1,2,3],xmm10[4,5],xmm13[6,7]
+; AVX1-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufb %xmm11, %xmm3, %xmm4
+; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3,4],xmm4[5,6,7]
+; AVX1-NEXT: vmovaps {{.*#+}} ymm3 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0]
+; AVX1-NEXT: vandps %ymm3, %ymm1, %ymm1
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-NEXT: vandnps %ymm2, %ymm3, %ymm2
+; AVX1-NEXT: vmovaps %ymm3, %ymm4
+; AVX1-NEXT: vorps %ymm2, %ymm1, %ymm1
+; AVX1-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm1 = mem[1,1,1,1]
+; AVX1-NEXT: vpsrldq {{.*#+}} xmm2 = xmm15[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm14 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm14 = xmm5[0,1],mem[2,3],xmm5[4,5,6,7]
+; AVX1-NEXT: vpshufb %xmm0, %xmm14, %xmm2
+; AVX1-NEXT: vpblendw {{.*#+}} xmm15 = xmm2[0,1,2],xmm1[3,4],xmm2[5,6,7]
+; AVX1-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm2 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm2 = xmm7[0,1],mem[2,3],xmm7[4,5,6,7]
+; AVX1-NEXT: vpshufb %xmm0, %xmm2, %xmm0
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vpblendw $207, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm1 = mem[0,1,2,3],xmm1[4,5],mem[6,7]
+; AVX1-NEXT: vpshufb %xmm11, %xmm1, %xmm13
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm13, %ymm0
+; AVX1-NEXT: vandnps %ymm15, %ymm12, %ymm13
+; AVX1-NEXT: vandps %ymm0, %ymm12, %ymm0
+; AVX1-NEXT: vorps %ymm0, %ymm13, %ymm13
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-NEXT: vpsrldq {{.*#+}} xmm7 = xmm9[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm7 = xmm7[0],xmm0[0]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm15 = xmm6[0,1,2,3],xmm8[4,5],xmm6[6,7]
+; AVX1-NEXT: vpshufb %xmm11, %xmm15, %xmm5
+; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm7[3,4],xmm5[5,6,7]
+; AVX1-NEXT: vandps %ymm4, %ymm13, %ymm7
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-NEXT: vandnps %ymm5, %ymm4, %ymm5
+; AVX1-NEXT: vmovaps %ymm4, %ymm13
+; AVX1-NEXT: vorps %ymm5, %ymm7, %ymm0
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-NEXT: vpsrlq $48, %xmm11, %xmm5
+; AVX1-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm7 = mem[2,2,3,3]
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm5 = xmm7[0],xmm5[0]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [6,7,2,3,14,15,14,15,8,9,10,11,12,13,14,15]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpshufb %xmm7, %xmm0, %xmm6
+; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3,4],xmm6[5,6,7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpshufb %xmm7, %xmm0, %xmm6
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm10 = [0,1,2,3,4,5,6,7,14,15,2,3,14,15,10,11]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpshufb %xmm10, %xmm0, %xmm8
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm8, %ymm6
+; AVX1-NEXT: vandnps %ymm5, %ymm12, %ymm5
+; AVX1-NEXT: vandps %ymm6, %ymm12, %ymm6
+; AVX1-NEXT: vorps %ymm5, %ymm6, %ymm5
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-NEXT: vpsrlq $48, %xmm8, %xmm6
+; AVX1-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm4 = mem[2,2,3,3]
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm6[0]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpshufb %xmm10, %xmm0, %xmm6
+; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0,1,2],xmm4[3,4],xmm6[5,6,7]
+; AVX1-NEXT: vmovaps %ymm13, %ymm6
+; AVX1-NEXT: vandps %ymm5, %ymm13, %ymm5
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-NEXT: vandnps %ymm4, %ymm13, %ymm4
+; AVX1-NEXT: vorps %ymm4, %ymm5, %ymm4
+; AVX1-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-NEXT: vpsrlq $48, %xmm13, %xmm4
+; AVX1-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm5 = mem[2,2,3,3]
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm4 = xmm5[0],xmm4[0]
+; AVX1-NEXT: vpshufb %xmm7, %xmm14, %xmm5
+; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3,4],xmm5[5,6,7]
+; AVX1-NEXT: vpshufb %xmm7, %xmm2, %xmm2
+; AVX1-NEXT: vpshufb %xmm10, %xmm1, %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT: vandnps %ymm4, %ymm12, %ymm2
+; AVX1-NEXT: vandps %ymm1, %ymm12, %ymm1
+; AVX1-NEXT: vorps %ymm2, %ymm1, %ymm1
+; AVX1-NEXT: vpsrlq $48, %xmm3, %xmm2
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm9[2,2,3,3]
+; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
+; AVX1-NEXT: vpshufb %xmm10, %xmm15, %xmm0
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm2[3,4],xmm0[5,6,7]
+; AVX1-NEXT: vandps %ymm6, %ymm1, %ymm1
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vandnps %ymm0, %ymm6, %ymm0
+; AVX1-NEXT: vorps %ymm0, %ymm1, %ymm0
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm0 = mem[1,1,1,1]
+; AVX1-NEXT: vpshufd $238, (%rsp), %xmm1 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm1 = mem[2,3,2,3]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm1 = xmm1[0,1],mem[2,3],xmm1[4,5,6,7]
+; AVX1-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [0,1,4,5,4,5,6,7,0,1,4,5,0,1,12,13]
+; AVX1-NEXT: vpshufb %xmm5, %xmm1, %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm1 = mem[1,1,1,1]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm15[2,3,2,3]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX1-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm2 = mem[0,1,0,3]
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,6]
+; AVX1-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm11[1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3,4,5,6,7]
+; AVX1-NEXT: vmovaps {{.*#+}} ymm2 = [65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
+; AVX1-NEXT: vandnps %ymm0, %ymm2, %ymm0
+; AVX1-NEXT: vandps %ymm2, %ymm1, %ymm1
+; AVX1-NEXT: vorps %ymm0, %ymm1, %ymm0
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm3 = xmm1[0,1],mem[2,3],xmm1[4,5,6,7]
+; AVX1-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm14 = mem[0,1,0,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm1 = xmm14[0,1,2,3,4,5,4,6]
+; AVX1-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm8[1]
+; AVX1-NEXT: vpshufb %xmm5, %xmm3, %xmm3
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm3[5,6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm0 = mem[1,1,1,1]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm11[2,3,2,3]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm12 = xmm0[0,1],mem[2,3],xmm0[4,5,6,7]
+; AVX1-NEXT: vpshufb %xmm5, %xmm12, %xmm3
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm3, %ymm3
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm9[1,1,1,1]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm8[2,3,2,3]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1],xmm6[2],xmm1[2],xmm6[3],xmm1[3]
+; AVX1-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm1 = mem[0,1,0,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm4 = xmm1[0,1,2,3,4,5,4,6]
+; AVX1-NEXT: vpunpckhqdq {{.*#+}} xmm4 = xmm4[1],xmm13[1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3,4,5,6,7]
+; AVX1-NEXT: vandnps %ymm3, %ymm2, %ymm3
+; AVX1-NEXT: vandps %ymm2, %ymm4, %ymm4
+; AVX1-NEXT: vorps %ymm3, %ymm4, %ymm6
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm10 = xmm0[0,1],mem[2,3],xmm0[4,5,6,7]
+; AVX1-NEXT: vpshufb %xmm5, %xmm10, %xmm7
+; AVX1-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm4 = mem[0,1,0,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm5 = xmm4[0,1,2,3,4,5,4,6]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vpunpckhqdq {{.*#+}} xmm5 = xmm5[1],xmm3[1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm7[5,6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm5[5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpsrlq $48, %xmm0, %xmm5
+; AVX1-NEXT: vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpsrldq {{.*#+}} xmm6 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm13 = [6,7,2,3,4,5,6,7,6,7,6,7,2,3,14,15]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpshufb %xmm13, %xmm0, %xmm7
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm7, %ymm5
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpsrlq $48, %xmm0, %xmm7
+; AVX1-NEXT: vpsrldq {{.*#+}} xmm6 = xmm15[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpsrld $16, %xmm0, %xmm7
+; AVX1-NEXT: vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm0 = mem[0,1,2,3,4,5,5,7]
+; AVX1-NEXT: vpunpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm7[1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm6[0,1],xmm0[2,3,4,5,6,7]
+; AVX1-NEXT: vandnps %ymm5, %ymm2, %ymm5
+; AVX1-NEXT: vandps %ymm2, %ymm0, %ymm0
+; AVX1-NEXT: vorps %ymm5, %ymm0, %ymm0
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-NEXT: vpsrld $16, %xmm5, %xmm5
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm6 = xmm14[0,1,2,3,4,5,5,7]
+; AVX1-NEXT: vpunpckhqdq {{.*#+}} xmm5 = xmm6[1],xmm5[1]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-NEXT: vpshufb %xmm13, %xmm6, %xmm6
+; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm6[5,6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3,4],ymm5[5,6,7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpsrlq $48, %xmm0, %xmm0
+; AVX1-NEXT: vpsrldq {{.*#+}} xmm6 = xmm11[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm6[0],xmm0[0],xmm6[1],xmm0[1],xmm6[2],xmm0[2],xmm6[3],xmm0[3]
+; AVX1-NEXT: vpshufb %xmm13, %xmm12, %xmm6
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm6, %ymm0
+; AVX1-NEXT: vpsrlq $48, %xmm9, %xmm6
+; AVX1-NEXT: vpsrldq {{.*#+}} xmm7 = xmm8[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-NEXT: vpsrld $16, %xmm7, %xmm7
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,7]
+; AVX1-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm7[1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3,4,5,6,7]
+; AVX1-NEXT: vandnps %ymm0, %ymm2, %ymm0
+; AVX1-NEXT: vandps %ymm2, %ymm1, %ymm1
+; AVX1-NEXT: vorps %ymm0, %ymm1, %ymm0
+; AVX1-NEXT: vpshufb %xmm13, %xmm10, %xmm1
+; AVX1-NEXT: vpsrld $16, %xmm3, %xmm2
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,4,5,5,7]
+; AVX1-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4],xmm1[5,6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, (%rsi)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, 32(%rsi)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, 32(%rdx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, (%rdx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, 32(%rcx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, (%rcx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, 32(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, (%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, 32(%r9)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, (%r9)
+; AVX1-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX1-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX1-NEXT: vmovaps %ymm5, (%rax)
+; AVX1-NEXT: addq $584, %rsp # imm = 0x248
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-SLOW-LABEL: vf32:
+; AVX2-SLOW: # %bb.0:
+; AVX2-SLOW-NEXT: subq $520, %rsp # imm = 0x208
+; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm5
+; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm6
+; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %ymm2
+; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %ymm3
+; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm12 = ymm3[2,3],ymm2[2,3]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm12[0,2,2,1,4,6,6,5]
+; AVX2-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm4 = ymm4[0,1,2,3,6,6,6,6,8,9,10,11,14,14,14,14]
+; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm2 = ymm3[0,1],ymm2[0,1]
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm2[0,3,2,3,4,7,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm2 = ymm3[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[0,1,2,2,4,5,6,6]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4,5,6],ymm4[7],ymm2[8,9],ymm4[10],ymm2[11,12,13,14],ymm4[15]
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm9 = ymm0[2,3],ymm1[2,3]
+; AVX2-SLOW-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm7 = ymm0[0,1],ymm1[0,1]
+; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1],ymm5[2],ymm6[3,4],ymm5[5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm4 = <8,9,u,u,0,1,12,13,u,u,12,13,12,13,14,15>
+; AVX2-SLOW-NEXT: vpshufb %xmm4, %xmm1, %xmm0
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm2[2,2,2,2,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm0[0],xmm3[1],xmm0[2,3],xmm3[4],xmm0[5,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm10 = ymm7[2,1,2,1,6,5,6,5]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm0 = ymm10[0,1,2,3,6,6,6,6,8,9,10,11,14,14,14,14]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm13 = ymm9[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm11 = ymm13[2,2,2,2,4,5,6,7,10,10,10,10,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm11 = ymm0[0],ymm11[1,2,3,4],ymm0[5],ymm11[6,7],ymm0[8],ymm11[9,10,11,12],ymm0[13],ymm11[14,15]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
+; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm3, %ymm11, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 224(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovdqa 192(%rdi), %ymm6
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0,1],ymm6[2],ymm0[3,4],ymm6[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm9
+; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa %ymm0, %ymm8
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufb %xmm4, %xmm3, %xmm7
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm6 = xmm4[2,2,2,2,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm6 = xmm7[0],xmm6[1],xmm7[2,3],xmm6[4],xmm7[5,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm14 = ymm12[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm7 = ymm14[2,2,2,2,4,5,6,7,10,10,10,10,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,4,6,8,9,10,11,12,13,12,14]
+; AVX2-SLOW-NEXT: vpshufd $102, (%rsp), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm0 = mem[2,1,2,1,6,5,6,5]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm15 = ymm0[0,1,2,3,6,6,6,6,8,9,10,11,14,14,14,14]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm15[0],ymm7[1,2,3,4],ymm15[5],ymm7[6,7],ymm15[8],ymm7[9,10,11,12],ymm15[13],ymm7[14,15]
+; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm6, %ymm7, %ymm6
+; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm6 = ymm10[1,1,1,1,4,5,6,7,9,9,9,9,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm6 = ymm6[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm7 = ymm13[3,3,3,3,4,5,6,7,11,11,11,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3,4],ymm6[5],ymm7[6,7],ymm6[8],ymm7[9,10,11,12],ymm6[13],ymm7[14,15]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm7 = <10,11,u,u,2,3,14,15,u,u,10,11,12,13,14,15>
+; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm1, %xmm1
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,2,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3],xmm2[4],xmm1[5,6,7]
+; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm1, %ymm6, %ymm1
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[1,1,1,1,4,5,6,7,9,9,9,9,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm1 = ymm14[3,3,3,3,4,5,6,7,11,11,11,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,5,7,8,9,10,11,12,13,13,15]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6,7],ymm0[8],ymm1[9,10,11,12],ymm0[13],ymm1[14,15]
+; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm3, %xmm1
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm4[1,1,2,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3],xmm2[4],xmm1[5,6,7]
+; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0],ymm8[1],ymm9[2,3],ymm8[4],ymm9[5,6],ymm8[7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm9
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm9[0,2,0,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm10 = <0,1,12,13,u,u,4,5,u,u,u,u,12,13,14,15>
+; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm2, %xmm1
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3],xmm0[4,5],xmm1[6,7]
+; AVX2-SLOW-NEXT: vpblendd $56, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm0 = ymm0[0,1,2],mem[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %ymm12
+; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %ymm3
+; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm3[0,1],ymm12[2],ymm3[3,4],ymm12[5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm5[2,2,2,2,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,2]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm6 = <0,1,4,5,4,5,u,u,0,1,12,13,u,u,4,5>
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm4
+; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm4, %xmm7
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm7[0,1,2],xmm1[3],xmm7[4,5],xmm1[6],xmm7[7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5,6,7],ymm0[8,9,10],ymm1[11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm3
+; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm3, %xmm6
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm13 = xmm7[2,2,2,2,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm13[0,1,2,2]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm8 = xmm6[0,1,2],xmm1[3],xmm6[4,5],xmm1[6],xmm6[7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm6 = ymm15[0,2,2,1,4,6,6,5]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm6 = ymm6[0,1,2,3,6,6,6,6,8,9,10,11,14,14,14,14]
+; AVX2-SLOW-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm13 = mem[0,3,2,3,4,7,6,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm11 = ymm13[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm11 = ymm11[0,1,2,2,4,5,6,6]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm6 = ymm11[0,1],ymm6[2],ymm11[3,4,5,6],ymm6[7],ymm11[8,9],ymm6[10],ymm11[11,12,13,14],ymm6[15]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $146, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm11 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5,6],mem[7]
+; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm11, %xmm10
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm11, %xmm0
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm14 = xmm0[0,2,0,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm14[0,1,2,3,4,6,6,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm10[0,1],xmm1[2],xmm10[3],xmm1[4,5],xmm10[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm6[3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm6
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm6 = ymm1[0,1,2],ymm6[3,4,5,6,7],ymm1[8,9,10],ymm6[11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm10[2,1,2,1,6,5,6,5]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm1 = ymm1[1,1,1,1,4,5,6,7,9,9,9,9,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshuflw $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm8 = ymm8[0,1,2,3,5,5,5,5,8,9,10,11,13,13,13,13]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm8[0,1],ymm1[2],ymm8[3,4,5,6],ymm1[7],ymm8[8,9],ymm1[10],ymm8[11,12,13,14],ymm1[15]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm8 = <u,u,u,u,10,11,u,u,2,3,14,15,u,u,u,u>
+; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm9, %xmm6
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm9 = <2,3,14,15,u,u,6,7,u,u,u,u,12,13,14,15>
+; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm2, %xmm2
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm6[2],xmm2[3],xmm6[4,5],xmm2[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3,4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
+; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm4, %xmm4
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,5,5]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2],xmm5[3],xmm4[4,5],xmm5[6],xmm4[7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm1[0,1,2],ymm4[3,4,5,6,7],ymm1[8,9,10],ymm4[11,12,13,14,15]
+; 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: vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm3 = xmm7[0,1,2,3,5,5,5,5]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3],xmm2[4,5],xmm3[6],xmm2[7]
+; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm0, %xmm0
+; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm11, %xmm3
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0,1],xmm0[2],xmm3[3],xmm0[4,5],xmm3[6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm15[2,1,2,1,6,5,6,5]
+; AVX2-SLOW-NEXT: vmovdqa %ymm15, %ymm1
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm3 = ymm3[1,1,1,1,4,5,6,7,9,9,9,9,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm4 = ymm13[1,3,2,3,4,5,6,7,9,11,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm4 = ymm4[0,1,2,3,5,5,5,5,8,9,10,11,13,13,13,13]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm4[0,1],ymm3[2],ymm4[3,4,5,6],ymm3[7],ymm4[8,9],ymm3[10],ymm4[11,12,13,14],ymm3[15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm0[0,1,2],ymm2[3,4,5,6,7],ymm0[8,9,10],ymm2[11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufd $197, (%rsp), %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm0 = mem[1,1,0,3,5,5,4,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,6,7,8,9,10,11,12,12,14,15]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm15 = ymm10[0,3,2,3,4,7,6,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm4 = ymm15[0,0,2,3,4,5,6,7,8,8,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm4 = ymm4[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm4[1,2,3,4],ymm0[5,6],ymm4[7],ymm0[8],ymm4[9,10,11,12],ymm0[13,14],ymm4[15]
+; AVX2-SLOW-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = ymm12[0,1],mem[2],ymm12[3,4],mem[5],ymm12[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm12 = xmm5[0,1,2,1]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm14 = xmm4[2,1,0,3]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm5 = xmm14[0,0,0,0,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm7 = xmm12[0,1,2,3,6,5,6,4]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3],xmm7[4],xmm5[5,6],xmm7[7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = ymm2[0,1],mem[2],ymm2[3,4],mem[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm11 = xmm7[2,1,2,3]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm7
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[0,3,2,1]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm7[0,0,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,6,6,6]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm11[2,1,2,0,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1,2],xmm3[3],xmm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm3
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm0[0,1,2],ymm3[3,4,5,6,7],ymm0[8,9,10],ymm3[11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1,2,3,4],xmm0[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; 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 $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1],ymm0[2],mem[3,4],ymm0[5],mem[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,1,0,3]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm5 = xmm2[0,0,0,0,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,6,5,6,4]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm9 = xmm5[0,1,2,3],xmm0[4],xmm5[5,6],xmm0[7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm5 = ymm8[1,1,0,3,5,5,4,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm5 = ymm5[0,1,2,3,4,4,6,7,8,9,10,11,12,12,14,15]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm13 = ymm1[0,3,2,3,4,7,6,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm10 = ymm13[0,0,2,3,4,5,6,7,8,8,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm10 = ymm10[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0],ymm10[1,2,3,4],ymm5[5,6],ymm10[7],ymm5[8],ymm10[9,10,11,12],ymm5[13,14],ymm10[15]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm10 = ymm0[0,1],mem[2],ymm0[3,4],mem[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm10[2,1,2,3]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm10, %xmm6
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,3,2,1]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm6[0,0,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,6,6,6]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm0[2,1,2,0,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm4[1,2],xmm1[3],xmm4[4,5,6,7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm4
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7],ymm5[8,9,10],ymm4[11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm5[5,6,7]
+; 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: vmovdqa {{.*#+}} ymm10 = <6,7,u,u,u,u,u,u,u,u,2,3,14,15,u,u,22,23,u,u,u,u,u,u,u,u,18,19,30,31,u,u>
+; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpshufb %ymm10, %ymm1, %ymm4
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm9 = ymm15[0,1,3,3,4,5,6,7,8,9,11,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm9 = ymm9[0,1,2,3,5,5,5,5,8,9,10,11,13,13,13,13]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm9[1,2,3,4],ymm4[5,6],ymm9[7],ymm4[8],ymm9[9,10,11,12],ymm4[13,14],ymm9[15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm12[0,1,2,3,7,5,6,5]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm14[1,1,1,1,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm5[4],xmm1[5,6],xmm5[7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm5 = xmm11[3,1,2,1,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,1,3,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0],xmm7[1,2],xmm5[3],xmm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3,4,5,6,7],ymm4[8,9,10],ymm1[11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2,3,4],xmm4[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm9 = ymm4[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,5]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,1,1,1,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4],xmm2[5,6],xmm3[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm10, %ymm8, %ymm3
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm4 = ymm13[0,1,3,3,4,5,6,7,8,9,11,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm4 = ymm4[0,1,2,3,5,5,5,5,8,9,10,11,13,13,13,13]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3,4],ymm3[5,6],ymm4[7],ymm3[8],ymm4[9,10,11,12],ymm3[13,14],ymm4[15]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,1,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm6[0,1,3,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm4[1,2],xmm0[3],xmm4[4,5,6,7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7],ymm3[8,9,10],ymm2[11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm3[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $146, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm1[0],mem[1],ymm1[2,3],mem[4],ymm1[5,6],mem[7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm2[0,1,0,2,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,6,6,6]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,u,u,u,u,u,u,8,9,u,u,0,1,12,13>
+; AVX2-SLOW-NEXT: vpshufb %xmm5, %xmm3, %xmm6
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm6[4],xmm4[5],xmm6[6,7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $109, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[0],ymm1[1],mem[2,3],ymm1[4],mem[5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm6, %xmm7
+; AVX2-SLOW-NEXT: vpshufb %xmm5, %xmm7, %xmm5
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,3,2,1]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm6[0,1,0,2,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,6,6,6]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm5[4],xmm1[5],xmm5[6,7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,u,u,u,u,u,u,10,11,u,u,2,3,14,15>
+; AVX2-SLOW-NEXT: vpshufb %xmm5, %xmm3, %xmm3
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,3,3]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4],xmm2[5],xmm3[6,7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm5, %xmm7, %xmm3
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm5 = xmm6[0,1,1,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,3,3]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm5[0,1,2,3],xmm3[4],xmm5[5],xmm3[6,7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm5, 32(%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm5, (%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm5, 32(%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm5, (%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm5, 32(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm5, (%rcx)
+; AVX2-SLOW-NEXT: vmovdqa %ymm9, 32(%r8)
+; AVX2-SLOW-NEXT: vmovdqa %ymm0, (%r8)
+; AVX2-SLOW-NEXT: vmovdqa %ymm1, 32(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm4, (%r9)
+; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX2-SLOW-NEXT: vmovdqa %ymm3, 32(%rax)
+; AVX2-SLOW-NEXT: vmovdqa %ymm2, (%rax)
+; AVX2-SLOW-NEXT: addq $520, %rsp # imm = 0x208
+; AVX2-SLOW-NEXT: vzeroupper
+; AVX2-SLOW-NEXT: retq
+;
+; AVX2-FAST-LABEL: vf32:
+; AVX2-FAST: # %bb.0:
+; AVX2-FAST-NEXT: subq $568, %rsp # imm = 0x238
+; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm9
+; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm11
+; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm3
+; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm5 = ymm2[2,3],ymm3[2,3]
+; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm2 = ymm2[0,1],ymm3[0,1]
+; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm12 = ymm1[2,3],ymm0[2,3]
+; AVX2-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm7 = ymm1[0,1],ymm0[0,1]
+; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <u,u,4,5,4,5,4,5,8,9,u,u,8,9,12,13,u,u,20,21,20,21,20,21,24,25,u,u,24,25,28,29>
+; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm5[0,1,0,3,4,5,4,7]
+; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm8, %ymm0
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm3 = ymm2[2,1,2,1,6,5,6,5]
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} ymm2 = ymm3[0,1,2,3,6,6,6,6,8,9,10,11,14,14,14,14]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm2[0],ymm0[1,2,3,4],ymm2[5],ymm0[6,7],ymm2[8],ymm0[9,10,11,12],ymm2[13],ymm0[14,15]
+; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0,1],ymm9[2],ymm11[3,4],ymm9[5],ymm11[6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm15 = <8,9,u,u,0,1,12,13,u,u,12,13,12,13,14,15>
+; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm2, %xmm1
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm0
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm6 = xmm0[2,2,2,2,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm6[1],xmm1[2,3],xmm6[4],xmm1[5,6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
+; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm1, %ymm4, %ymm1
+; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm12 = ymm12[0,1,0,3,4,5,4,7]
+; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm12, %ymm1
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm14 = ymm7[2,1,2,1,6,5,6,5]
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} ymm4 = ymm14[0,1,2,3,6,6,6,6,8,9,10,11,14,14,14,14]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm1[1,2,3,4],ymm4[5],ymm1[6,7],ymm4[8],ymm1[9,10,11,12],ymm4[13],ymm1[14,15]
+; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 192(%rdi), %ymm6
+; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1],ymm6[2],ymm1[3,4],ymm6[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm13, %xmm6
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm13, %xmm1
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm7 = xmm1[2,2,2,2,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0],xmm7[1],xmm6[2,3],xmm7[4],xmm6[5,6,7]
+; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm6, %ymm4, %ymm4
+; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <2,3,u,u,u,u,u,u,u,u,14,15,u,u,u,u,18,19,u,u,u,u,u,u,u,u,30,31,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,6,7,6,7,6,7,8,9,u,u,10,11,14,15,u,u,22,23,22,23,22,23,24,25,u,u,26,27,30,31>
+; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm8, %ymm7
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm7[1,2,3,4],ymm3[5],ymm7[6,7],ymm3[8],ymm7[9,10,11,12],ymm3[13],ymm7[14,15]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,6,7,u,u,u,u,10,11,u,u,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm0, %xmm0
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = <10,11,u,u,2,3,14,15,u,u,10,11,12,13,14,15>
+; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm2, %xmm2
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3],xmm0[4],xmm2[5,6,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm10, %ymm8
+; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm0, %ymm3, %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm14, %ymm0
+; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm12, %ymm2
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3,4],ymm0[5],ymm2[6,7],ymm0[8],ymm2[9,10,11,12],ymm0[13],ymm2[14,15]
+; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm1, %xmm1
+; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm13, %xmm2
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,12,13,u,u,u,u,16,17,u,u,u,u,u,u,u,u,u,u>
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2,3],xmm1[4],xmm2[5,6,7]
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm5[2,1,2,1,6,5,6,5]
+; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendvb %ymm8, %ymm1, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpshufb %ymm10, %ymm2, %ymm1
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm8 = <u,u,u,u,u,u,u,u,u,u,u,u,8,9,u,u,16,17,20,21,u,u,22,23,u,u,u,u,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = mem[0,3,2,3,4,7,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpshufb %ymm8, %ymm0, %ymm6
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm12 = ymm6[0,1],ymm1[2],ymm6[3,4,5,6],ymm1[7],ymm6[8,9],ymm1[10],ymm6[11,12,13,14],ymm1[15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0],ymm11[1],ymm9[2,3],ymm11[4],ymm9[5,6],ymm11[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm0 = xmm7[2,1,0,3]
+; AVX2-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm9 = <0,1,12,13,u,u,4,5,u,u,u,u,12,13,14,15>
+; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm3, %xmm1
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm14 = <u,u,u,u,0,1,u,u,8,9,12,13,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm0, %xmm5
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm5[2],xmm1[3],xmm5[4,5],xmm1[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm12[3,4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm2[0,1],ymm0[2],ymm2[3,4],ymm0[5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm12 = xmm5[u,u,u,u,u,u,4,5,u,u,u,u,8,9,u,u]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm5, %xmm7
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm15 = xmm7[0,1,4,5,4,5,u,u,0,1,12,13,u,u,4,5]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm15[0,1,2],xmm12[3],xmm15[4,5],xmm12[6],xmm15[7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm1[0,1,2],ymm4[3,4,5,6,7],ymm1[8,9,10],ymm4[11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm13[2,1,2,1,6,5,6,5]
+; AVX2-FAST-NEXT: vpshufb %ymm10, %ymm4, %ymm1
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm15 = ymm6[0,3,2,3,4,7,6,7]
+; AVX2-FAST-NEXT: vpshufb %ymm8, %ymm15, %ymm12
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm12[0,1],ymm1[2],ymm12[3,4,5,6],ymm1[7],ymm12[8,9],ymm1[10],ymm12[11,12,13,14],ymm1[15]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $146, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm12 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5,6],mem[7]
+; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm12, %xmm9
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm12, %xmm11
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[2,1,0,3]
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm11, %xmm8
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm8 = xmm9[0,1],xmm8[2],xmm9[3],xmm8[4,5],xmm9[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2],ymm1[3,4,5],ymm8[6,7]
+; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm10
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1],ymm1[2],ymm10[3,4],ymm1[5],ymm10[6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm9
+; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm0[u,u,u,u,u,u,4,5,u,u,u,u,8,9,u,u]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm14 = xmm1[0,1,4,5,4,5,u,u,0,1,12,13,u,u,4,5]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm14[0,1,2],xmm2[3],xmm14[4,5],xmm2[6],xmm14[7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm8[0,1,2],ymm2[3,4,5,6,7],ymm8[8,9,10],ymm2[11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm8 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,u,u,18,19,22,23,u,u,22,23,u,u,u,u,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufb %ymm8, %ymm15, %ymm14
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} ymm4 = ymm4[1,1,1,1,4,5,6,7,9,9,9,9,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm15 = ymm14[0,1],ymm4[2],ymm14[3,4,5,6],ymm4[7],ymm14[8,9],ymm4[10],ymm14[11,12,13,14],ymm4[15]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm14 = <2,3,14,15,u,u,6,7,u,u,u,u,12,13,14,15>
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm12, %xmm4
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm12 = <u,u,u,u,2,3,u,u,10,11,14,15,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm11, %xmm2
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1],xmm2[2],xmm4[3],xmm2[4,5],xmm4[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm15[3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
+; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm1, %xmm1
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3],xmm1[4,5],xmm0[6],xmm1[7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7],ymm2[8,9,10],ymm0[11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm7, %xmm0
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm2 = xmm5[0,1,2,3,5,5,5,5]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm2[3],xmm0[4,5],xmm2[6],xmm0[7]
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm3, %xmm2
+; AVX2-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm1, %xmm3
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2],xmm2[3],xmm3[4,5],xmm2[6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpshufb %ymm8, %ymm1, %ymm3
+; AVX2-FAST-NEXT: vpshuflw $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[1,1,1,1,4,5,6,7,9,9,9,9,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4,5,6],ymm4[7],ymm3[8,9],ymm4[10],ymm3[11,12,13,14],ymm4[15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7],ymm2[8,9,10],ymm0[11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm15 = ymm13[0,3,2,3,4,7,6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm7 = <4,5,u,u,u,u,u,u,u,u,0,1,12,13,u,u,20,21,u,u,u,u,u,u,u,u,16,17,28,29,u,u>
+; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm6, %ymm4
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm12 = <u,u,0,1,4,5,6,7,8,9,u,u,u,u,8,9,u,u,16,17,20,21,22,23,24,25,u,u,u,u,24,25>
+; AVX2-FAST-NEXT: vpshufb %ymm12, %ymm15, %ymm6
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm6[1,2,3,4],ymm4[5,6],ymm6[7],ymm4[8],ymm6[9,10,11,12],ymm4[13,14],ymm6[15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm9[0,1],ymm10[2],ymm9[3,4],ymm10[5],ymm9[6,7]
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm8 = xmm6[2,1,0,3]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm6
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm13 = xmm6[0,1,2,1]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm14 = <0,1,0,1,0,1,0,1,u,u,8,9,12,13,u,u>
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm8, %xmm6
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,6,5,6,4]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm6[0,1,2,3],xmm0[4],xmm6[5,6],xmm0[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = ymm1[0,1],mem[2],ymm1[3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm5
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,3,2,1]
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[2,1,2,3]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,0,1,4,5,u,u,12,13,12,13,12,13,12,13>
+; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm5, %xmm2
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm3 = xmm6[2,1,2,0,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1,2],xmm3[3],xmm2[4,5,6,7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7],ymm4[8,9,10],ymm0[11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm4[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm9, %ymm0
+; AVX2-FAST-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0,3,2,3,4,7,6,7]
+; AVX2-FAST-NEXT: vpshufb %ymm12, %ymm10, %ymm3
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm3[1,2,3,4],ymm0[5,6],ymm3[7],ymm0[8],ymm3[9,10,11,12],ymm0[13,14],ymm3[15]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm11[0,1],mem[2],ymm11[3,4],mem[5],ymm11[6,7]
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm12 = xmm3[2,1,0,3]
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm12, %xmm2
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm4 = xmm3[0,1,2,3,6,5,6,4]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm14 = xmm2[0,1,2,3],xmm4[4],xmm2[5,6],xmm4[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1],ymm2[2],mem[3,4],ymm2[5],mem[6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm7
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[0,3,2,1]
+; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm7, %xmm2
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm1 = xmm4[2,1,2,0,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1,2],xmm1[3],xmm2[4,5,6,7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm0[0,1,2],ymm2[3,4,5,6,7],ymm0[8,9,10],ymm2[11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4],xmm0[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <6,7,u,u,u,u,u,u,u,u,2,3,14,15,u,u,22,23,u,u,u,u,u,u,u,u,18,19,30,31,u,u>
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <u,u,2,3,6,7,6,7,10,11,u,u,u,u,10,11,u,u,18,19,22,23,22,23,26,27,u,u,u,u,26,27>
+; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm15, %ymm15
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm15[1,2,3,4],ymm2[5,6],ymm15[7],ymm2[8],ymm15[9,10,11,12],ymm2[13,14],ymm15[15]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm15 = <2,3,2,3,2,3,2,3,u,u,10,11,14,15,u,u>
+; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm8, %xmm8
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,7,5,6,5]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm8[0,1,2,3],xmm0[4],xmm8[5,6],xmm0[7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm8 = <u,u,2,3,6,7,u,u,14,15,14,15,14,15,14,15>
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm5, %xmm5
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[3,1,2,1,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0],xmm5[1,2],xmm6[3],xmm5[4,5,6,7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7],ymm2[8,9,10],ymm0[11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm5[0,1,2,3,4],xmm2[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm9, %ymm1
+; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm10, %ymm2
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3,4],ymm1[5,6],ymm2[7],ymm1[8],ymm2[9,10,11,12],ymm1[13,14],ymm2[15]
+; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm12, %xmm2
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,5]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4],xmm2[5,6],xmm3[7]
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm7, %xmm3
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[3,1,2,1,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1,2],xmm4[3],xmm3[4,5,6,7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm1[0,1,2],ymm2[3,4,5,6,7],ymm1[8,9,10],ymm2[11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3,4],xmm1[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd $109, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0],ymm11[1],mem[2,3],ymm11[4],mem[5,6],ymm11[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = <u,u,u,u,u,u,u,u,8,9,u,u,0,1,12,13>
+; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm3, %xmm5
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm6 = <0,1,2,3,0,1,4,5,u,u,12,13,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm2, %xmm7
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm5 = xmm7[0,1,2,3],xmm5[4],xmm7[5],xmm5[6,7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $109, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = mem[0],ymm0[1],mem[2,3],ymm0[4],mem[5,6],ymm0[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm7, %xmm0
+; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm0, %xmm4
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[0,3,2,1]
+; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm7, %xmm6
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0,1,2,3],xmm4[4],xmm6[5],xmm4[6,7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm6 = <u,u,u,u,u,u,u,u,10,11,u,u,2,3,14,15>
+; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm3, %xmm3
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = <0,1,2,3,2,3,6,7,u,u,14,15,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm2, %xmm2
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4],xmm2[5],xmm3[6,7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm0, %xmm0
+; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm7, %xmm1
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4],xmm1[5],xmm0[6,7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = mem[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, (%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, (%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, (%rcx)
+; AVX2-FAST-NEXT: vmovdqa %ymm13, 32(%r8)
+; AVX2-FAST-NEXT: vmovdqa %ymm8, (%r8)
+; AVX2-FAST-NEXT: vmovdqa %ymm4, 32(%r9)
+; AVX2-FAST-NEXT: vmovdqa %ymm5, (%r9)
+; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX2-FAST-NEXT: vmovdqa %ymm0, 32(%rax)
+; AVX2-FAST-NEXT: vmovdqa %ymm2, (%rax)
+; AVX2-FAST-NEXT: addq $568, %rsp # imm = 0x238
+; AVX2-FAST-NEXT: vzeroupper
+; AVX2-FAST-NEXT: retq
+;
+; AVX512-LABEL: vf32:
+; AVX512: # %bb.0:
+; AVX512-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm3
+; AVX512-NEXT: vmovdqu64 64(%rdi), %zmm5
+; AVX512-NEXT: vmovdqu64 128(%rdi), %zmm2
+; AVX512-NEXT: vmovdqu64 192(%rdi), %zmm4
+; AVX512-NEXT: vmovdqu64 320(%rdi), %zmm0
+; AVX512-NEXT: vmovdqu64 256(%rdi), %zmm1
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,0,0,0,4,10,16,22,28,34,40,46,52,58,0,0,0,0,0,0,4,10,16,22,28,34,40,46,52,58]
+; AVX512-NEXT: # zmm7 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm0, %zmm1, %zmm7
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,6,12,18,24,30,0,0,0,0,0,34,40,46,52,58,0,6,12,18,24,30,0,0,0,0,0,34,40,46,52,58]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm2, %zmm4, %zmm8
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm6 = <0,6,12,18,24,30,36,42,48,54,60,u,u,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm5, %zmm3, %zmm6
+; AVX512-NEXT: movl $4192256, %edi # imm = 0x3FF800
+; AVX512-NEXT: kmovd %edi, %k1
+; AVX512-NEXT: vmovdqu16 %zmm8, %zmm6 {%k1}
+; AVX512-NEXT: movw $-2048, %di # imm = 0xF800
+; AVX512-NEXT: kmovd %edi, %k2
+; AVX512-NEXT: vmovdqa32 %zmm7, %zmm6 {%k2}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,0,0,0,5,11,17,23,29,35,41,47,53,59,0,0,0,0,0,0,5,11,17,23,29,35,41,47,53,59]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm0, %zmm1, %zmm8
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm9 = [1,7,13,19,25,31,0,0,0,0,0,35,41,47,53,59,1,7,13,19,25,31,0,0,0,0,0,35,41,47,53,59]
+; AVX512-NEXT: # zmm9 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm2, %zmm4, %zmm9
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm7 = <1,7,13,19,25,31,37,43,49,55,61,u,u,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm5, %zmm3, %zmm7
+; AVX512-NEXT: vmovdqu16 %zmm9, %zmm7 {%k1}
+; AVX512-NEXT: vmovdqa32 %zmm8, %zmm7 {%k2}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,0,0,0,6,12,18,24,30,36,42,48,54,60,0,0,0,0,0,0,6,12,18,24,30,36,42,48,54,60]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm0, %zmm1, %zmm8
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm9 = [34,40,46,52,58,0,0,0,0,0,0,4,10,16,22,28,34,40,46,52,58,0,0,0,0,0,0,4,10,16,22,28]
+; AVX512-NEXT: # zmm9 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm2, %zmm9
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm10 = <34,40,46,52,58,0,6,12,18,24,30,u,u,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm3, %zmm5, %zmm10
+; AVX512-NEXT: movl $2095104, %edi # imm = 0x1FF800
+; AVX512-NEXT: kmovd %edi, %k2
+; AVX512-NEXT: vmovdqu16 %zmm9, %zmm10 {%k2}
+; AVX512-NEXT: movl $-2097152, %edi # imm = 0xFFE00000
+; AVX512-NEXT: kmovd %edi, %k1
+; AVX512-NEXT: vmovdqu16 %zmm8, %zmm10 {%k1}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,0,0,1,7,13,19,25,31,37,43,49,55,61,0,0,0,0,0,1,7,13,19,25,31,37,43,49,55,61]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm0, %zmm1, %zmm8
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm9 = [35,41,47,53,59,0,0,0,0,0,0,5,11,17,23,29,35,41,47,53,59,0,0,0,0,0,0,5,11,17,23,29]
+; AVX512-NEXT: # zmm9 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm2, %zmm9
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm11 = <35,41,47,53,59,1,7,13,19,25,31,u,u,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm3, %zmm5, %zmm11
+; AVX512-NEXT: vmovdqu16 %zmm9, %zmm11 {%k2}
+; AVX512-NEXT: vmovdqu16 %zmm8, %zmm11 {%k1}
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm8 = <4,10,16,22,28,34,40,46,52,58,u,u,u,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm5, %zmm3, %zmm8
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm9 = [36,42,48,54,60,0,0,0,0,0,0,6,12,18,24,30,36,42,48,54,60,0,0,0,0,0,0,6,12,18,24,30]
+; AVX512-NEXT: # zmm9 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm2, %zmm9
+; AVX512-NEXT: movw $31, %di
+; AVX512-NEXT: kmovd %edi, %k2
+; AVX512-NEXT: vmovdqa32 %zmm8, %zmm9 {%k2}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,0,0,0,34,40,46,52,58,0,6,12,18,24,30,0,0,0,0,0,34,40,46,52,58,0,6,12,18,24,30]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm1, %zmm0, %zmm8
+; AVX512-NEXT: vmovdqu16 %zmm8, %zmm9 {%k1}
+; AVX512-NEXT: vmovdqa {{.*#+}} ymm8 = <5,11,17,23,29,35,41,47,53,59,u,u,u,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm5, %zmm3, %zmm8
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm3 = [37,43,49,55,61,0,0,0,0,0,1,7,13,19,25,31,37,43,49,55,61,0,0,0,0,0,1,7,13,19,25,31]
+; AVX512-NEXT: # zmm3 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm2, %zmm3
+; AVX512-NEXT: vmovdqa32 %zmm8, %zmm3 {%k2}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm2 = [0,0,0,0,0,35,41,47,53,59,1,7,13,19,25,31,0,0,0,0,0,35,41,47,53,59,1,7,13,19,25,31]
+; AVX512-NEXT: # zmm2 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm1, %zmm0, %zmm2
+; AVX512-NEXT: vmovdqu16 %zmm2, %zmm3 {%k1}
+; AVX512-NEXT: vmovdqu64 %zmm6, (%rsi)
+; AVX512-NEXT: vmovdqu64 %zmm7, (%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm10, (%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm11, (%r8)
+; AVX512-NEXT: vmovdqu64 %zmm9, (%r9)
+; AVX512-NEXT: vmovdqu64 %zmm3, (%rax)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %wide.vec = load <192 x i16>, <192 x i16>* %in.vec, align 32
+
+ %strided.vec0 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <32 x i32> <i32 0, i32 6, i32 12, i32 18, i32 24, i32 30, i32 36, i32 42, i32 48, i32 54, i32 60, i32 66, i32 72, i32 78, i32 84, i32 90, i32 96, i32 102, i32 108, i32 114, i32 120, i32 126, i32 132, i32 138, i32 144, i32 150, i32 156, i32 162, i32 168, i32 174, i32 180, i32 186>
+ %strided.vec1 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <32 x i32> <i32 1, i32 7, i32 13, i32 19, i32 25, i32 31, i32 37, i32 43, i32 49, i32 55, i32 61, i32 67, i32 73, i32 79, i32 85, i32 91, i32 97, i32 103, i32 109, i32 115, i32 121, i32 127, i32 133, i32 139, i32 145, i32 151, i32 157, i32 163, i32 169, i32 175, i32 181, i32 187>
+ %strided.vec2 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <32 x i32> <i32 2, i32 8, i32 14, i32 20, i32 26, i32 32, i32 38, i32 44, i32 50, i32 56, i32 62, i32 68, i32 74, i32 80, i32 86, i32 92, i32 98, i32 104, i32 110, i32 116, i32 122, i32 128, i32 134, i32 140, i32 146, i32 152, i32 158, i32 164, i32 170, i32 176, i32 182, i32 188>
+ %strided.vec3 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <32 x i32> <i32 3, i32 9, i32 15, i32 21, i32 27, i32 33, i32 39, i32 45, i32 51, i32 57, i32 63, i32 69, i32 75, i32 81, i32 87, i32 93, i32 99, i32 105, i32 111, i32 117, i32 123, i32 129, i32 135, i32 141, i32 147, i32 153, i32 159, i32 165, i32 171, i32 177, i32 183, i32 189>
+ %strided.vec4 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <32 x i32> <i32 4, i32 10, i32 16, i32 22, i32 28, i32 34, i32 40, i32 46, i32 52, i32 58, i32 64, i32 70, i32 76, i32 82, i32 88, i32 94, i32 100, i32 106, i32 112, i32 118, i32 124, i32 130, i32 136, i32 142, i32 148, i32 154, i32 160, i32 166, i32 172, i32 178, i32 184, i32 190>
+ %strided.vec5 = shufflevector <192 x i16> %wide.vec, <192 x i16> poison, <32 x i32> <i32 5, i32 11, i32 17, i32 23, i32 29, i32 35, i32 41, i32 47, i32 53, i32 59, i32 65, i32 71, i32 77, i32 83, i32 89, i32 95, i32 101, i32 107, i32 113, i32 119, i32 125, i32 131, i32 137, i32 143, i32 149, i32 155, i32 161, i32 167, i32 173, i32 179, i32 185, i32 191>
+
+ store <32 x i16> %strided.vec0, <32 x i16>* %out.vec0, align 32
+ store <32 x i16> %strided.vec1, <32 x i16>* %out.vec1, align 32
+ store <32 x i16> %strided.vec2, <32 x i16>* %out.vec2, align 32
+ store <32 x i16> %strided.vec3, <32 x i16>* %out.vec3, align 32
+ store <32 x i16> %strided.vec4, <32 x i16>* %out.vec4, align 32
+ store <32 x i16> %strided.vec5, <32 x i16>* %out.vec5, align 32
+
+ ret void
+}
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
index a52dce62834c9..e48a1e3e88541 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
@@ -713,3 +713,780 @@ define void @load_i32_stride3_vf16(<48 x i32>* %in.vec, <16 x i32>* %out.vec0, <
ret void
}
+
+define void @load_i32_stride3_vf32(<96 x i32>* %in.vec, <32 x i32>* %out.vec0, <32 x i32>* %out.vec1, <32 x i32>* %out.vec2) nounwind {
+; SSE-LABEL: load_i32_stride3_vf32:
+; SSE: # %bb.0:
+; SSE-NEXT: subq $360, %rsp # imm = 0x168
+; SSE-NEXT: movaps 336(%rdi), %xmm11
+; SSE-NEXT: movaps 368(%rdi), %xmm14
+; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 352(%rdi), %xmm13
+; SSE-NEXT: movaps 240(%rdi), %xmm12
+; SSE-NEXT: movaps 272(%rdi), %xmm6
+; SSE-NEXT: movaps 256(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 144(%rdi), %xmm10
+; SSE-NEXT: movaps 176(%rdi), %xmm3
+; SSE-NEXT: movaps 160(%rdi), %xmm7
+; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 48(%rdi), %xmm5
+; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 80(%rdi), %xmm8
+; SSE-NEXT: movaps 64(%rdi), %xmm9
+; SSE-NEXT: movaps %xmm9, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm8[1,1]
+; SSE-NEXT: movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm5, %xmm2
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm7, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm3[1,1]
+; SSE-NEXT: movaps %xmm3, %xmm5
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm10, %xmm2
+; SSE-NEXT: movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm1, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm6[1,1]
+; SSE-NEXT: movaps %xmm6, %xmm15
+; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm12, %xmm2
+; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm13, %xmm2
+; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm13, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm14[1,1]
+; SSE-NEXT: movaps %xmm11, %xmm1
+; SSE-NEXT: movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm11, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 16(%rdi), %xmm4
+; SSE-NEXT: movaps 32(%rdi), %xmm3
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm4, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm3[1,1]
+; SSE-NEXT: movaps (%rdi), %xmm11
+; SSE-NEXT: movaps %xmm11, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 128(%rdi), %xmm3
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 112(%rdi), %xmm7
+; SSE-NEXT: movaps %xmm7, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm3[1,1]
+; SSE-NEXT: movaps 96(%rdi), %xmm6
+; SSE-NEXT: movaps %xmm6, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 224(%rdi), %xmm3
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 208(%rdi), %xmm13
+; SSE-NEXT: movaps %xmm13, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm3[1,1]
+; SSE-NEXT: movaps 192(%rdi), %xmm3
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 320(%rdi), %xmm3
+; SSE-NEXT: movaps 304(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,3],xmm3[1,1]
+; SSE-NEXT: movaps %xmm3, %xmm14
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 288(%rdi), %xmm3
+; SSE-NEXT: movaps %xmm3, (%rsp) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[2,1],xmm9[3,3]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm9[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,2],xmm8[2,0]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm5, %xmm0
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,1],xmm5[3,3]
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[1,0],xmm5[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,2],xmm0[2,0]
+; SSE-NEXT: movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm15, %xmm0
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,1],xmm3[3,3]
+; SSE-NEXT: movaps %xmm12, %xmm15
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[1,0],xmm3[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[0,2],xmm0[2,0]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,1],xmm2[3,3]
+; SSE-NEXT: movaps %xmm1, %xmm12
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[1,0],xmm2[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[0,2],xmm0[2,0]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: movaps %xmm8, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,1],xmm7[3,3]
+; SSE-NEXT: movaps %xmm6, %xmm1
+; SSE-NEXT: movaps %xmm6, %xmm10
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[1,0],xmm7[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,2],xmm0[2,0]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,1],xmm13[3,3]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[1,0],xmm13[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,2],xmm0[2,0]
+; SSE-NEXT: movaps %xmm14, %xmm2
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,1],xmm5[3,3]
+; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm5[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,2],xmm2[2,0]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, %xmm14
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[2,1],xmm4[3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm11[2,3,2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[1,0],xmm4[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,2],xmm14[2,0]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT: # xmm14 = mem[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],xmm14[0],xmm5[1],xmm14[1]
+; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = xmm5[0,1],mem[0,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1]
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,1],xmm2[0,3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT: # xmm13 = mem[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm4[0],xmm13[1],xmm4[1]
+; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT: # xmm13 = xmm13[0,1],mem[0,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm7[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm1[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm4[0],xmm14[1],xmm4[1]
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,1],xmm8[0,3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
+; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = xmm4[0,1],mem[0,3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = xmm7[0,1],mem[0,3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[1,1,1,1]
+; SSE-NEXT: 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],xmm8[0],xmm1[1],xmm8[1]
+; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[0,1],mem[0,3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
+; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,3]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 96(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 64(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 32(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, (%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 112(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 80(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 48(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 16(%rsi)
+; SSE-NEXT: movaps %xmm3, 96(%rdx)
+; SSE-NEXT: movaps %xmm6, 64(%rdx)
+; SSE-NEXT: movaps %xmm10, 32(%rdx)
+; SSE-NEXT: movaps %xmm11, (%rdx)
+; SSE-NEXT: movaps %xmm12, 112(%rdx)
+; SSE-NEXT: movaps %xmm15, 80(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 48(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 16(%rdx)
+; SSE-NEXT: movaps %xmm0, 96(%rcx)
+; SSE-NEXT: movaps %xmm1, 112(%rcx)
+; SSE-NEXT: movaps %xmm7, 64(%rcx)
+; SSE-NEXT: movaps %xmm4, 80(%rcx)
+; SSE-NEXT: movaps %xmm14, 32(%rcx)
+; SSE-NEXT: movaps %xmm13, 48(%rcx)
+; SSE-NEXT: movaps %xmm9, (%rcx)
+; SSE-NEXT: movaps %xmm5, 16(%rcx)
+; SSE-NEXT: addq $360, %rsp # imm = 0x168
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: load_i32_stride3_vf32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: subq $392, %rsp # imm = 0x188
+; AVX1-NEXT: vmovaps 320(%rdi), %ymm3
+; AVX1-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 288(%rdi), %ymm4
+; AVX1-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps (%rdi), %ymm2
+; AVX1-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX1-NEXT: vmovaps 32(%rdi), %ymm6
+; AVX1-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 128(%rdi), %ymm1
+; AVX1-NEXT: vmovaps 96(%rdi), %ymm5
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0],ymm1[1],ymm5[2,3],ymm1[4],ymm5[5,6],ymm1[7]
+; AVX1-NEXT: vmovaps %ymm5, %ymm9
+; AVX1-NEXT: vmovaps %ymm1, %ymm10
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,1]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: vmovaps 176(%rdi), %xmm5
+; AVX1-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 160(%rdi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vinsertps {{.*#+}} xmm1 = zero,zero,xmm1[2],xmm5[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 80(%rdi), %xmm1
+; AVX1-NEXT: vmovaps 64(%rdi), %xmm0
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = zero,zero,xmm0[2],xmm1[1]
+; AVX1-NEXT: vmovaps %xmm1, %xmm8
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0],ymm6[1],ymm2[2,3],ymm6[4],ymm2[5,6],ymm6[7]
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,1]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7]
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,1]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: vmovaps 368(%rdi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 352(%rdi), %xmm11
+; AVX1-NEXT: vinsertps {{.*#+}} xmm1 = zero,zero,xmm11[2],xmm1[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 224(%rdi), %ymm1
+; AVX1-NEXT: vmovaps 192(%rdi), %ymm2
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6],ymm1[7]
+; AVX1-NEXT: vmovaps %ymm2, %ymm3
+; AVX1-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps %ymm1, %ymm2
+; AVX1-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,3,2,1]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm12
+; AVX1-NEXT: vmovaps 272(%rdi), %xmm15
+; AVX1-NEXT: vmovaps 256(%rdi), %xmm14
+; AVX1-NEXT: vinsertps {{.*#+}} xmm13 = zero,zero,xmm14[2],xmm15[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3,4,5],ymm13[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vblendps {{.*#+}} ymm12 = ymm3[0,1],ymm2[2],ymm3[3,4],ymm2[5],ymm3[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm12, %xmm13
+; AVX1-NEXT: vblendps {{.*#+}} xmm7 = xmm13[0],xmm12[1,2],xmm13[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[1,0,3,2]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm13[1,1,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm7, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} xmm7 = xmm14[0,1],xmm15[2],xmm14[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[0,0,3,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm7[5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps %ymm10, %ymm12
+; AVX1-NEXT: vmovaps %ymm9, %ymm5
+; AVX1-NEXT: vblendps {{.*#+}} ymm6 = ymm9[0,1],ymm10[2],ymm9[3,4],ymm10[5],ymm9[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm6, %xmm7
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0],xmm6[1,2],xmm7[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[1,0,3,2]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[1,1,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-NEXT: vblendps {{.*#+}} xmm7 = xmm9[0,1],xmm10[2],xmm9[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[0,0,3,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm7[5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
+; AVX1-NEXT: vblendps {{.*#+}} ymm6 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm6, %xmm7
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0],xmm6[1,2],xmm7[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[1,0,3,2]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[1,1,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm8, %xmm0
+; AVX1-NEXT: vblendps {{.*#+}} xmm7 = xmm1[0,1],xmm8[2],xmm1[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[0,0,3,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4],ymm7[5,6,7]
+; AVX1-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-NEXT: vblendps {{.*#+}} ymm6 = ymm13[0,1],ymm2[2],ymm13[3,4],ymm2[5],ymm13[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm6, %xmm7
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0],xmm6[1,2],xmm7[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[1,0,3,2]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[1,1,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-NEXT: vblendps {{.*#+}} xmm7 = xmm11[0,1],xmm8[2],xmm11[3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[0,0,3,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5,6,7]
+; AVX1-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vblendps {{.*#+}} ymm6 = ymm12[0,1],ymm5[2],ymm12[3,4],ymm5[5],ymm12[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm6, %xmm7
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm7[1],xmm6[2,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[2,1,0,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[2,3,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
+; AVX1-NEXT: vshufps {{.*#+}} xmm7 = xmm9[0,1],xmm10[0,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-NEXT: vblendps {{.*#+}} ymm12 = ymm6[0,1,2,3,4],ymm7[5,6,7]
+; AVX1-NEXT: vshufps {{.*#+}} xmm5 = xmm1[0,1],xmm0[0,3]
+; AVX1-NEXT: vblendps {{.*#+}} ymm6 = ymm4[0,1],ymm3[2],ymm4[3,4],ymm3[5],ymm4[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm6, %xmm7
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm7[1],xmm6[2,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[2,1,0,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[2,3,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm6 = ymm2[0,1],ymm13[2],ymm2[3,4],ymm13[5],ymm2[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm6, %xmm7
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm7[1],xmm6[2,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[2,1,0,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[2,3,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
+; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm11[0,1],xmm8[0,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vblendps $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX1-NEXT: # ymm2 = ymm0[0,1],mem[2],ymm0[3,4],mem[5],ymm0[6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm4
+; AVX1-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm4[1],xmm2[2,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[2,1,0,3]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm4 = xmm4[2,3,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm14[0,1],xmm15[0,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, 64(%rsi)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, 96(%rsi)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, (%rsi)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, 32(%rsi)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, 96(%rdx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, (%rdx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, 32(%rdx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, 64(%rdx)
+; AVX1-NEXT: vmovaps %ymm0, 64(%rcx)
+; AVX1-NEXT: vmovaps %ymm3, 96(%rcx)
+; AVX1-NEXT: vmovaps %ymm5, (%rcx)
+; AVX1-NEXT: vmovaps %ymm12, 32(%rcx)
+; AVX1-NEXT: addq $392, %rsp # imm = 0x188
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-SLOW-LABEL: load_i32_stride3_vf32:
+; AVX2-SLOW: # %bb.0:
+; AVX2-SLOW-NEXT: subq $136, %rsp
+; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vmovaps 224(%rdi), %ymm3
+; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %ymm2
+; AVX2-SLOW-NEXT: vmovaps 256(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovaps 128(%rdi), %ymm6
+; AVX2-SLOW-NEXT: vmovaps 160(%rdi), %ymm10
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm4
+; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %ymm14
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm12 = [21474836482,21474836482,21474836482,21474836482]
+; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm12, %ymm0
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm4[0],ymm7[1],ymm4[2,3],ymm7[4],ymm4[5,6],ymm7[7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm13 = <0,3,6,1,4,7,u,u>
+; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm13, %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm12, %ymm0
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0],ymm6[1],ymm14[2,3],ymm6[4],ymm14[5,6],ymm6[7]
+; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm13, %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm12, %ymm0
+; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm2[0],ymm3[1],ymm2[2,3],ymm3[4],ymm2[5,6],ymm3[7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm13, %ymm9
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %ymm11
+; AVX2-SLOW-NEXT: vmovaps %ymm15, %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm11[0],ymm15[1],ymm11[2,3],ymm15[4],ymm11[5,6],ymm15[7]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm13, %ymm13
+; AVX2-SLOW-NEXT: vmovaps 352(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm12, %ymm12
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm3 = <1,4,7,2,5,u,u,u>
+; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm3, %ymm12
+; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,0,3,6,0,0,3,6]
+; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm0, %ymm15
+; AVX2-SLOW-NEXT: vmovaps %ymm1, %ymm13
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm14[0,1],ymm6[2],ymm14[3,4],ymm6[5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm3, %ymm15
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1],ymm7[2],ymm1[3,4],ymm7[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm3, %ymm5
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm9
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1],ymm4[2],ymm11[3,4],ymm4[5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm3, %ymm3
+; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm14[2],ymm6[3,4],ymm14[5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm9 = <2,5,0,3,6,u,u,u>
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm9, %ymm3
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm10 = ymm10[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm1[2],ymm7[3,4],ymm1[5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm9, %ymm6
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = ymm2[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm11[2],ymm4[3,4],ymm11[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm9, %ymm2
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm8[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm2[0,1],mem[2],ymm2[3,4],mem[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm9, %ymm2
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm4 = ymm13[0,1,0,3,4,5,4,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 96(%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 64(%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, (%rsi)
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 96(%rdx)
+; AVX2-SLOW-NEXT: vmovaps %ymm5, (%rdx)
+; AVX2-SLOW-NEXT: vmovaps %ymm15, 32(%rdx)
+; AVX2-SLOW-NEXT: vmovaps %ymm12, 64(%rdx)
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 64(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 96(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm6, (%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%rcx)
+; AVX2-SLOW-NEXT: addq $136, %rsp
+; AVX2-SLOW-NEXT: vzeroupper
+; AVX2-SLOW-NEXT: retq
+;
+; AVX2-FAST-ALL-LABEL: load_i32_stride3_vf32:
+; AVX2-FAST-ALL: # %bb.0:
+; AVX2-FAST-ALL-NEXT: subq $72, %rsp
+; AVX2-FAST-ALL-NEXT: vmovaps 320(%rdi), %ymm1
+; AVX2-FAST-ALL-NEXT: vmovaps 224(%rdi), %ymm0
+; AVX2-FAST-ALL-NEXT: vmovaps 192(%rdi), %ymm2
+; AVX2-FAST-ALL-NEXT: vmovaps 256(%rdi), %ymm3
+; AVX2-FAST-ALL-NEXT: vmovaps 128(%rdi), %ymm9
+; AVX2-FAST-ALL-NEXT: vmovaps 160(%rdi), %ymm10
+; AVX2-FAST-ALL-NEXT: vmovaps (%rdi), %ymm5
+; AVX2-FAST-ALL-NEXT: vmovaps 32(%rdi), %ymm7
+; AVX2-FAST-ALL-NEXT: vmovaps 64(%rdi), %ymm8
+; AVX2-FAST-ALL-NEXT: vmovaps 96(%rdi), %ymm12
+; AVX2-FAST-ALL-NEXT: vbroadcastsd {{.*#+}} ymm4 = [21474836482,21474836482,21474836482,21474836482]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm8, %ymm4, %ymm6
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm11 = ymm5[0],ymm7[1],ymm5[2,3],ymm7[4],ymm5[5,6],ymm7[7]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm13 = <0,3,6,1,4,7,u,u>
+; AVX2-FAST-ALL-NEXT: vpermps %ymm11, %ymm13, %ymm11
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm6 = ymm11[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-FAST-ALL-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vpermps %ymm10, %ymm4, %ymm6
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0],ymm9[1],ymm12[2,3],ymm9[4],ymm12[5,6],ymm9[7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm11, %ymm13, %ymm11
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm6 = ymm11[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-FAST-ALL-NEXT: vmovups %ymm6, (%rsp) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm3, %ymm6
+; AVX2-FAST-ALL-NEXT: vpermps %ymm3, %ymm4, %ymm11
+; AVX2-FAST-ALL-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm14 = ymm2[0],ymm0[1],ymm2[2,3],ymm0[4],ymm2[5,6],ymm0[7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm14, %ymm13, %ymm14
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-FAST-ALL-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovaps 288(%rdi), %ymm15
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0],ymm1[1],ymm15[2,3],ymm1[4],ymm15[5,6],ymm1[7]
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm1, %ymm3
+; AVX2-FAST-ALL-NEXT: vpermps %ymm14, %ymm13, %ymm13
+; AVX2-FAST-ALL-NEXT: vmovaps 352(%rdi), %ymm1
+; AVX2-FAST-ALL-NEXT: vpermps %ymm1, %ymm4, %ymm4
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm4 = ymm13[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-ALL-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm4 = ymm2[0,1],ymm0[2],ymm2[3,4],ymm0[5],ymm2[6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm2 = <1,4,7,2,5,u,u,u>
+; AVX2-FAST-ALL-NEXT: vpermps %ymm4, %ymm2, %ymm4
+; AVX2-FAST-ALL-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,0,3,6,0,0,3,6]
+; AVX2-FAST-ALL-NEXT: # ymm0 = mem[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm6, %ymm0, %ymm13
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm6, %ymm14
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm13 = ymm4[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm4 = ymm12[0,1],ymm9[2],ymm12[3,4],ymm9[5],ymm12[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm4, %ymm2, %ymm4
+; AVX2-FAST-ALL-NEXT: vpermps %ymm10, %ymm0, %ymm6
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm6 = ymm5[0,1],ymm7[2],ymm5[3,4],ymm7[5],ymm5[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm6, %ymm2, %ymm6
+; AVX2-FAST-ALL-NEXT: vpermps %ymm8, %ymm0, %ymm11
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm11[5,6,7]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm11 = ymm15[0,1],ymm3[2],ymm15[3,4],ymm3[5],ymm15[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm11, %ymm2, %ymm2
+; AVX2-FAST-ALL-NEXT: vpermps %ymm1, %ymm0, %ymm0
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm9[0,1],ymm12[2],ymm9[3,4],ymm12[5],ymm9[6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm9 = [0,1,0,3,0,1,4,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm10, %ymm9, %ymm10
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm11 = <2,5,0,3,6,u,u,u>
+; AVX2-FAST-ALL-NEXT: vpermps %ymm2, %ymm11, %ymm2
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1],ymm5[2],ymm7[3,4],ymm5[5],ymm7[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm8, %ymm9, %ymm7
+; AVX2-FAST-ALL-NEXT: vpermps %ymm5, %ymm11, %ymm5
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm7 = ymm3[0,1],ymm15[2],ymm3[3,4],ymm15[5],ymm3[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm1, %ymm9, %ymm1
+; AVX2-FAST-ALL-NEXT: vpermps %ymm7, %ymm11, %ymm7
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm14, %ymm9, %ymm3
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vblendps $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-ALL-NEXT: # ymm7 = ymm7[0,1],mem[2],ymm7[3,4],mem[5],ymm7[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm7, %ymm11, %ymm7
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm7, 96(%rsi)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm7, 64(%rsi)
+; AVX2-FAST-ALL-NEXT: vmovups (%rsp), %ymm7 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm7, 32(%rsi)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm7, (%rsi)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, 96(%rdx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm6, (%rdx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm13, 64(%rdx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm3, 64(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm1, 96(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm5, (%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm2, 32(%rcx)
+; AVX2-FAST-ALL-NEXT: addq $72, %rsp
+; AVX2-FAST-ALL-NEXT: vzeroupper
+; AVX2-FAST-ALL-NEXT: retq
+;
+; AVX2-FAST-PERLANE-LABEL: load_i32_stride3_vf32:
+; AVX2-FAST-PERLANE: # %bb.0:
+; AVX2-FAST-PERLANE-NEXT: subq $136, %rsp
+; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovaps 256(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps 128(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovaps 160(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %ymm14
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm12 = [21474836482,21474836482,21474836482,21474836482]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm12, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm4[0],ymm7[1],ymm4[2,3],ymm7[4],ymm4[5,6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm13 = <0,3,6,1,4,7,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm13, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm12, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm14[0],ymm6[1],ymm14[2,3],ymm6[4],ymm14[5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm13, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm12, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm2[0],ymm3[1],ymm2[2,3],ymm3[4],ymm2[5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm13, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %ymm11
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm15, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm11[0],ymm15[1],ymm11[2,3],ymm15[4],ymm11[5,6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm13, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vmovaps 352(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm12, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm3 = <1,4,7,2,5,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm3, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,0,3,6,0,0,3,6]
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm0, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm14[0,1],ymm6[2],ymm14[3,4],ymm6[5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm3, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1],ymm7[2],ymm1[3,4],ymm7[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm3, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1],ymm4[2],ymm11[3,4],ymm4[5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm3, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm14[2],ymm6[3,4],ymm14[5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm9 = <2,5,0,3,6,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm9, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm10 = ymm10[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm1[2],ymm7[3,4],ymm1[5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm9, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = ymm2[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm11[2],ymm4[3,4],ymm11[5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm9, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm8[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm2[0,1],mem[2],ymm2[3,4],mem[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm9, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm4 = ymm13[0,1,0,3,4,5,4,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 96(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 64(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, (%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 96(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm15, 32(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm12, 64(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 64(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 96(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: addq $136, %rsp
+; AVX2-FAST-PERLANE-NEXT: vzeroupper
+; AVX2-FAST-PERLANE-NEXT: retq
+;
+; AVX512-LABEL: load_i32_stride3_vf32:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu64 320(%rdi), %zmm0
+; AVX512-NEXT: vmovdqu64 256(%rdi), %zmm1
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm2
+; AVX512-NEXT: vmovdqu64 64(%rdi), %zmm3
+; AVX512-NEXT: vmovdqu64 128(%rdi), %zmm4
+; AVX512-NEXT: vmovdqu64 192(%rdi), %zmm5
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm6 = <0,3,6,9,12,15,18,21,24,27,30,u,u,u,u,u>
+; AVX512-NEXT: vmovdqa64 %zmm5, %zmm7
+; AVX512-NEXT: vpermt2d %zmm1, %zmm6, %zmm7
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm8 = [0,1,2,3,4,5,6,7,8,9,10,17,20,23,26,29]
+; AVX512-NEXT: vpermt2d %zmm0, %zmm8, %zmm7
+; AVX512-NEXT: vpermi2d %zmm3, %zmm2, %zmm6
+; AVX512-NEXT: vpermt2d %zmm4, %zmm8, %zmm6
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm8 = <17,20,23,26,29,0,3,6,9,12,15,u,u,u,u,u>
+; AVX512-NEXT: vmovdqa64 %zmm1, %zmm9
+; AVX512-NEXT: vpermt2d %zmm5, %zmm8, %zmm9
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm10 = [0,1,2,3,4,5,6,7,8,9,10,18,21,24,27,30]
+; AVX512-NEXT: vpermt2d %zmm0, %zmm10, %zmm9
+; AVX512-NEXT: vpermi2d %zmm2, %zmm3, %zmm8
+; AVX512-NEXT: vpermt2d %zmm4, %zmm10, %zmm8
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm10 = <2,5,8,11,14,17,20,23,26,29,u,u,u,u,u,u>
+; AVX512-NEXT: vpermt2d %zmm1, %zmm10, %zmm5
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,1,2,3,4,5,6,7,8,9,16,19,22,25,28,31]
+; AVX512-NEXT: vpermt2d %zmm0, %zmm1, %zmm5
+; AVX512-NEXT: vpermt2d %zmm3, %zmm10, %zmm2
+; AVX512-NEXT: vpermt2d %zmm4, %zmm1, %zmm2
+; AVX512-NEXT: vmovdqu64 %zmm7, 64(%rsi)
+; AVX512-NEXT: vmovdqu64 %zmm6, (%rsi)
+; AVX512-NEXT: vmovdqu64 %zmm9, 64(%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm8, (%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm5, 64(%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm2, (%rcx)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %wide.vec = load <96 x i32>, <96 x i32>* %in.vec, align 32
+
+ %strided.vec0 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <32 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45, i32 48, i32 51, i32 54, i32 57, i32 60, i32 63, i32 66, i32 69, i32 72, i32 75, i32 78, i32 81, i32 84, i32 87, i32 90, i32 93>
+ %strided.vec1 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <32 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46, i32 49, i32 52, i32 55, i32 58, i32 61, i32 64, i32 67, i32 70, i32 73, i32 76, i32 79, i32 82, i32 85, i32 88, i32 91, i32 94>
+ %strided.vec2 = shufflevector <96 x i32> %wide.vec, <96 x i32> poison, <32 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47, i32 50, i32 53, i32 56, i32 59, i32 62, i32 65, i32 68, i32 71, i32 74, i32 77, i32 80, i32 83, i32 86, i32 89, i32 92, i32 95>
+
+ store <32 x i32> %strided.vec0, <32 x i32>* %out.vec0, align 32
+ store <32 x i32> %strided.vec1, <32 x i32>* %out.vec1, align 32
+ store <32 x i32> %strided.vec2, <32 x i32>* %out.vec2, align 32
+
+ ret void
+}
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
index ab4fb54967cd5..dc73c043b2a3d 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
@@ -796,3 +796,908 @@ define void @load_i32_stride4_vf16(<64 x i32>* %in.vec, <16 x i32>* %out.vec0, <
ret void
}
+
+define void @load_i32_stride4_vf32(<128 x i32>* %in.vec, <32 x i32>* %out.vec0, <32 x i32>* %out.vec1, <32 x i32>* %out.vec2, <32 x i32>* %out.vec3) nounwind {
+; SSE-LABEL: load_i32_stride4_vf32:
+; SSE: # %bb.0:
+; SSE-NEXT: subq $456, %rsp # imm = 0x1C8
+; SSE-NEXT: movaps 336(%rdi), %xmm8
+; SSE-NEXT: movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 368(%rdi), %xmm9
+; SSE-NEXT: movaps 352(%rdi), %xmm10
+; SSE-NEXT: movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 208(%rdi), %xmm7
+; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 192(%rdi), %xmm5
+; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 240(%rdi), %xmm6
+; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 224(%rdi), %xmm14
+; SSE-NEXT: movaps 80(%rdi), %xmm4
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 64(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 112(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 96(%rdi), %xmm11
+; SSE-NEXT: movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpcklps {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT: movaps %xmm1, %xmm2
+; SSE-NEXT: unpcklps {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
+; SSE-NEXT: movaps %xmm2, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm11[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm14, %xmm1
+; SSE-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
+; SSE-NEXT: movaps %xmm5, %xmm3
+; SSE-NEXT: unpcklps {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm11[1]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm3, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm1[1]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm10, %xmm0
+; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
+; SSE-NEXT: movaps %xmm9, %xmm7
+; SSE-NEXT: movaps 320(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
+; SSE-NEXT: movaps %xmm1, %xmm2
+; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 496(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 480(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: movaps 464(%rdi), %xmm3
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 448(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm1, %xmm2
+; SSE-NEXT: unpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT: movaps %xmm2, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 176(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 160(%rdi), %xmm9
+; SSE-NEXT: movaps %xmm9, %xmm0
+; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: movaps 144(%rdi), %xmm3
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 128(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm1, %xmm2
+; SSE-NEXT: unpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT: movaps %xmm2, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 304(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 288(%rdi), %xmm13
+; SSE-NEXT: movaps %xmm13, %xmm0
+; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: movaps 272(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 256(%rdi), %xmm15
+; SSE-NEXT: movaps %xmm15, %xmm12
+; SSE-NEXT: unpcklps {{.*#+}} xmm12 = xmm12[0],xmm1[0],xmm12[1],xmm1[1]
+; SSE-NEXT: movaps %xmm12, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm0[1]
+; SSE-NEXT: movaps 432(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 416(%rdi), %xmm10
+; SSE-NEXT: movaps %xmm10, %xmm0
+; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: movaps 400(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 384(%rdi), %xmm8
+; SSE-NEXT: movaps %xmm8, %xmm3
+; SSE-NEXT: unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT: movaps %xmm3, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm0[1]
+; SSE-NEXT: movaps 32(%rdi), %xmm2
+; SSE-NEXT: movaps 48(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm2, %xmm1
+; SSE-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: movaps (%rdi), %xmm5
+; SSE-NEXT: movaps 16(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT: movaps %xmm5, %xmm4
+; SSE-NEXT: unpcklps {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT: movaps %xmm4, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm1[1]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT: # xmm14 = xmm14[2],mem[2],xmm14[3],mem[3]
+; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT: # xmm14 = xmm14[2],mem[2],xmm14[3],mem[3]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: unpckhps {{.*#+}} xmm11 = xmm11[2],xmm7[2],xmm11[3],xmm7[3]
+; SSE-NEXT: movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT: # xmm11 = xmm11[2],mem[2],xmm11[3],mem[3]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT: # xmm9 = xmm9[2],mem[2],xmm9[3],mem[3]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT: # xmm13 = xmm13[2],mem[2],xmm13[3],mem[3]
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT: # xmm15 = xmm15[2],mem[2],xmm15[3],mem[3]
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT: # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; SSE-NEXT: unpckhps (%rsp), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; SSE-NEXT: movaps %xmm0, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm6[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm6[1]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm14, %xmm0
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm1[1]
+; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, %xmm1
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm14[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm14[1]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: movaps %xmm6, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm11[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm11[1]
+; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm7, %xmm14
+; SSE-NEXT: movaps %xmm7, %xmm11
+; SSE-NEXT: movlhps {{.*#+}} xmm11 = xmm11[0],xmm9[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm9[1]
+; SSE-NEXT: movaps %xmm15, %xmm9
+; SSE-NEXT: movlhps {{.*#+}} xmm9 = xmm9[0],xmm13[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm13[1]
+; SSE-NEXT: movaps %xmm8, %xmm7
+; SSE-NEXT: movlhps {{.*#+}} xmm7 = xmm7[0],xmm10[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm10[1]
+; SSE-NEXT: movaps %xmm5, %xmm6
+; SSE-NEXT: movlhps {{.*#+}} xmm6 = xmm6[0],xmm2[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm2[1]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 96(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 64(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 32(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, (%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 112(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 80(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 48(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 16(%rsi)
+; SSE-NEXT: movaps %xmm3, 96(%rdx)
+; SSE-NEXT: movaps %xmm12, 64(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 32(%rdx)
+; SSE-NEXT: movaps %xmm4, (%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 112(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 80(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 48(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 16(%rdx)
+; SSE-NEXT: movaps %xmm7, 96(%rcx)
+; SSE-NEXT: movaps %xmm9, 64(%rcx)
+; SSE-NEXT: movaps %xmm11, 32(%rcx)
+; SSE-NEXT: movaps %xmm6, (%rcx)
+; SSE-NEXT: movaps %xmm0, 112(%rcx)
+; SSE-NEXT: movaps %xmm1, 80(%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, 16(%rcx)
+; SSE-NEXT: movaps %xmm8, 96(%r8)
+; SSE-NEXT: movaps %xmm15, 64(%r8)
+; SSE-NEXT: movaps %xmm14, 32(%r8)
+; SSE-NEXT: movaps %xmm5, (%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, 80(%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, 16(%r8)
+; SSE-NEXT: addq $456, %rsp # imm = 0x1C8
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: load_i32_stride4_vf32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: subq $744, %rsp # imm = 0x2E8
+; AVX1-NEXT: vmovaps 416(%rdi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 432(%rdi), %xmm0
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT: vmovaps 400(%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 384(%rdi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,0]
+; AVX1-NEXT: vmovaps 464(%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 448(%rdi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT: vmovaps 480(%rdi), %xmm3
+; AVX1-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 496(%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,0],xmm3[0,0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 288(%rdi), %xmm0
+; AVX1-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX1-NEXT: vmovaps 304(%rdi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm1[0],xmm0[0]
+; AVX1-NEXT: vmovaps 272(%rdi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 256(%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,0]
+; AVX1-NEXT: vmovaps 336(%rdi), %xmm14
+; AVX1-NEXT: vmovaps 320(%rdi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
+; AVX1-NEXT: vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT: vmovaps 352(%rdi), %xmm11
+; AVX1-NEXT: vmovaps 368(%rdi), %xmm12
+; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm12[0,0],xmm11[0,0]
+; AVX1-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 160(%rdi), %xmm0
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 176(%rdi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm1[0],xmm0[0]
+; AVX1-NEXT: vmovaps 144(%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 128(%rdi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX1-NEXT: vshufps {{.*#+}} xmm8 = xmm1[0,1],xmm0[2,0]
+; AVX1-NEXT: vmovaps 208(%rdi), %xmm5
+; AVX1-NEXT: vmovaps 192(%rdi), %xmm10
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm0 = xmm10[0],xmm5[0],xmm10[1],xmm5[1]
+; AVX1-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm9
+; AVX1-NEXT: vmovaps 224(%rdi), %xmm3
+; AVX1-NEXT: vmovaps 240(%rdi), %xmm4
+; AVX1-NEXT: vshufps {{.*#+}} xmm7 = xmm4[0,0],xmm3[0,0]
+; AVX1-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[0,1,2,0]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 80(%rdi), %xmm0
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 64(%rdi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-NEXT: vmovaps 96(%rdi), %xmm13
+; AVX1-NEXT: vmovaps 112(%rdi), %xmm15
+; AVX1-NEXT: vshufps {{.*#+}} xmm6 = xmm15[0,0],xmm13[0,0]
+; AVX1-NEXT: vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[0,1,2,0]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm6[6,7]
+; AVX1-NEXT: vmovaps 32(%rdi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 48(%rdi), %xmm0
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT: vmovaps (%rdi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 16(%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm6 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm6[0,1],xmm0[2,0]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm0 = xmm11[0],xmm12[0],xmm11[1],xmm12[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm6 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm6 = mem[0],xmm14[1],zero,zero
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-NEXT: vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm6 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm6 = mem[0],xmm11[1],zero,zero
+; AVX1-NEXT: vmovaps (%rsp), %xmm14 # 16-byte Reload
+; AVX1-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm7 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm7 = xmm14[0],mem[0],xmm14[1],mem[1]
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm0 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vinsertps {{.*#+}} xmm6 = xmm10[1],xmm5[1],zero,zero
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm6 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm6 = mem[0],xmm1[1],zero,zero
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm7 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm7 = xmm1[0],mem[0],xmm1[1],mem[1]
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm0 = xmm3[0],xmm10[0],xmm3[1],xmm10[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-NEXT: vinsertps {{.*#+}} xmm6 = xmm12[1],xmm2[1],zero,zero
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-NEXT: vinsertps {{.*#+}} xmm6 = xmm9[1],xmm5[1],zero,zero
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm0 = xmm13[0],xmm15[0],xmm13[1],xmm15[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-NEXT: vinsertps {{.*#+}} xmm6 = xmm13[1],xmm4[1],zero,zero
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm6 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm6 = xmm15[0],mem[0],xmm15[1],mem[1]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-NEXT: vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm7 = mem[0],xmm7[1],zero,zero
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm0 = zero,zero,xmm0[2],mem[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm6 = xmm6[2],xmm11[2],xmm6[3],xmm11[3]
+; AVX1-NEXT: vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm7 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm7 = zero,zero,xmm14[2],mem[0]
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm0 = zero,zero,xmm0[2],mem[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm6 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm6 = xmm14[2],mem[2],xmm14[3],mem[3]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-NEXT: vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm7 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm7 = zero,zero,xmm11[2],mem[0]
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = zero,zero,xmm3[2],xmm10[2]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm6 = xmm12[2],xmm2[2],xmm12[3],xmm2[3]
+; AVX1-NEXT: vmovaps %xmm12, %xmm3
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm6 = xmm9[2],xmm5[2],xmm9[3],xmm5[3]
+; AVX1-NEXT: vinsertps {{.*#+}} xmm7 = zero,zero,xmm1[2],xmm8[2]
+; AVX1-NEXT: vmovaps %xmm1, %xmm12
+; AVX1-NEXT: vmovaps %xmm8, %xmm10
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = zero,zero,xmm8[2],xmm5[2]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm4[2],xmm13[3],xmm4[3]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-NEXT: vinsertps {{.*#+}} xmm6 = zero,zero,xmm15[2],xmm13[2]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm7 = xmm9[2],xmm15[2],xmm9[3],xmm15[3]
+; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm1 = xmm1[3,0],mem[3,0]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,0,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm1 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm1 = xmm11[2],mem[2],xmm11[3],mem[3]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,0],xmm14[3,0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[2,0,2,3]
+; AVX1-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,0],xmm3[3,0]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[2,0,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm2 = xmm12[2],xmm10[2],xmm12[3],xmm10[3]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm3 = xmm3[3,0],mem[3,0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[2,0,2,3]
+; AVX1-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm3 = xmm3[3,0],mem[3,0]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[2,0,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-NEXT: vmovaps (%rsp), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm4 = xmm4[3,0],mem[3,0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm4 = xmm4[2,0,2,3]
+; AVX1-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm3 = xmm8[2],xmm5[2],xmm8[3],xmm5[3]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm4 = xmm4[3,0],mem[3,0]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-NEXT: vpermilps {{.*#+}} xmm4 = xmm4[2,0,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm4 = xmm4[2],xmm13[2],xmm4[3],xmm13[3]
+; AVX1-NEXT: vshufps {{.*#+}} xmm5 = xmm15[3,0],xmm9[3,0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[2,0,2,3]
+; AVX1-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX1-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm4, (%rsi)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm4, 64(%rsi)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm4, 96(%rsi)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm4, 96(%rdx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm4, (%rdx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm4, 64(%rdx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm4, 96(%rcx)
+; AVX1-NEXT: vmovaps %ymm6, (%rcx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm4, 32(%rcx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm4, 64(%rcx)
+; AVX1-NEXT: vmovaps %ymm2, 64(%r8)
+; AVX1-NEXT: vmovaps %ymm1, 96(%r8)
+; AVX1-NEXT: vmovaps %ymm3, (%r8)
+; AVX1-NEXT: vmovaps %ymm0, 32(%r8)
+; AVX1-NEXT: addq $744, %rsp # imm = 0x2E8
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: load_i32_stride4_vf32:
+; AVX2: # %bb.0:
+; AVX2-NEXT: subq $696, %rsp # imm = 0x2B8
+; AVX2-NEXT: vmovaps 288(%rdi), %ymm12
+; AVX2-NEXT: vmovaps 320(%rdi), %ymm5
+; AVX2-NEXT: vmovaps 352(%rdi), %ymm6
+; AVX2-NEXT: vmovaps 416(%rdi), %ymm7
+; AVX2-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 448(%rdi), %ymm2
+; AVX2-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 480(%rdi), %ymm1
+; AVX2-NEXT: vbroadcastsd {{.*#+}} ymm3 = [17179869184,17179869184,17179869184,17179869184]
+; AVX2-NEXT: vpermps %ymm1, %ymm3, %ymm0
+; AVX2-NEXT: vmovaps %ymm1, %ymm14
+; AVX2-NEXT: vpermps %ymm2, %ymm3, %ymm1
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-NEXT: vmovaps 400(%rdi), %xmm2
+; AVX2-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 384(%rdi), %xmm1
+; AVX2-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX2-NEXT: vmovaps {{.*#+}} xmm4 = <u,u,0,4>
+; AVX2-NEXT: vpermps %ymm7, %ymm4, %ymm2
+; AVX2-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vpermps %ymm6, %ymm3, %ymm0
+; AVX2-NEXT: vpermps %ymm5, %ymm3, %ymm1
+; AVX2-NEXT: vmovaps %ymm5, %ymm9
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-NEXT: vmovaps 272(%rdi), %xmm2
+; AVX2-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 256(%rdi), %xmm1
+; AVX2-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX2-NEXT: vpermps %ymm12, %ymm4, %ymm2
+; AVX2-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-NEXT: vmovaps 192(%rdi), %ymm5
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 224(%rdi), %ymm10
+; AVX2-NEXT: vpermps %ymm10, %ymm3, %ymm0
+; AVX2-NEXT: vpermps %ymm5, %ymm3, %ymm2
+; AVX2-NEXT: vmovaps %ymm5, %ymm8
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-NEXT: vmovaps 160(%rdi), %ymm1
+; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 144(%rdi), %xmm5
+; AVX2-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 128(%rdi), %xmm2
+; AVX2-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1]
+; AVX2-NEXT: vpermps %ymm1, %ymm4, %ymm5
+; AVX2-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm5[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 64(%rdi), %ymm15
+; AVX2-NEXT: vmovaps 96(%rdi), %ymm11
+; AVX2-NEXT: vpermps %ymm11, %ymm3, %ymm5
+; AVX2-NEXT: vpermps %ymm15, %ymm3, %ymm3
+; AVX2-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-NEXT: vmovaps 32(%rdi), %ymm0
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vpermps %ymm0, %ymm4, %ymm4
+; AVX2-NEXT: vmovaps (%rdi), %xmm0
+; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 16(%rdi), %xmm1
+; AVX2-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm5 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; AVX2-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vbroadcastsd {{.*#+}} ymm3 = [21474836481,21474836481,21474836481,21474836481]
+; AVX2-NEXT: vpermps %ymm6, %ymm3, %ymm4
+; AVX2-NEXT: vmovaps %ymm6, %ymm2
+; AVX2-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps %ymm9, %ymm1
+; AVX2-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vpermps %ymm9, %ymm3, %ymm5
+; AVX2-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-NEXT: vmovaps 256(%rdi), %ymm0
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 304(%rdi), %xmm4
+; AVX2-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 288(%rdi), %xmm6
+; AVX2-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
+; AVX2-NEXT: vmovaps {{.*#+}} xmm4 = <1,5,u,u>
+; AVX2-NEXT: vpermps %ymm0, %ymm4, %ymm7
+; AVX2-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vpermps %ymm10, %ymm3, %ymm5
+; AVX2-NEXT: vpermps %ymm8, %ymm3, %ymm6
+; AVX2-NEXT: vmovaps %ymm8, %ymm7
+; AVX2-NEXT: vblendps {{.*#+}} ymm8 = ymm6[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-NEXT: vmovaps 128(%rdi), %ymm0
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 176(%rdi), %xmm6
+; AVX2-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 160(%rdi), %xmm5
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm13 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; AVX2-NEXT: vpermps %ymm0, %ymm4, %ymm9
+; AVX2-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1],xmm13[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vpermps %ymm11, %ymm3, %ymm6
+; AVX2-NEXT: vpermps %ymm15, %ymm3, %ymm8
+; AVX2-NEXT: vblendps {{.*#+}} ymm9 = ymm8[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-NEXT: vmovaps (%rdi), %ymm13
+; AVX2-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 32(%rdi), %xmm6
+; AVX2-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 48(%rdi), %xmm0
+; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm8 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
+; AVX2-NEXT: vpermps %ymm13, %ymm4, %ymm13
+; AVX2-NEXT: vblendps {{.*#+}} xmm0 = xmm13[0,1],xmm8[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps %ymm14, %ymm9
+; AVX2-NEXT: vpermps %ymm14, %ymm3, %ymm0
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-NEXT: vpermps %ymm8, %ymm3, %ymm3
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-NEXT: vmovaps 384(%rdi), %ymm3
+; AVX2-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vpermps %ymm3, %ymm4, %ymm13
+; AVX2-NEXT: vmovaps 432(%rdi), %xmm4
+; AVX2-NEXT: vmovaps 416(%rdi), %xmm3
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm14 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; AVX2-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm14[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vbroadcastsd {{.*#+}} ymm0 = [25769803778,25769803778,25769803778,25769803778]
+; AVX2-NEXT: vpermps %ymm2, %ymm0, %ymm13
+; AVX2-NEXT: vpermps %ymm1, %ymm0, %ymm14
+; AVX2-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm14 # 16-byte Folded Reload
+; AVX2-NEXT: # xmm14 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX2-NEXT: vmovaps {{.*#+}} xmm1 = <u,u,2,6>
+; AVX2-NEXT: vpermps %ymm12, %ymm1, %ymm12
+; AVX2-NEXT: vblendps {{.*#+}} xmm12 = xmm14[0,1],xmm12[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vpermps %ymm10, %ymm0, %ymm12
+; AVX2-NEXT: vpermps %ymm7, %ymm0, %ymm13
+; AVX2-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm13 # 16-byte Folded Reload
+; AVX2-NEXT: # xmm13 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX2-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX2-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm14[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm2 = ymm13[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vpermps %ymm9, %ymm0, %ymm12
+; AVX2-NEXT: vpermps %ymm8, %ymm0, %ymm13
+; AVX2-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm13 # 16-byte Folded Reload
+; AVX2-NEXT: # xmm13 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX2-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX2-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm14[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm14 = ymm13[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-NEXT: vpermps %ymm11, %ymm0, %ymm12
+; AVX2-NEXT: vpermps %ymm15, %ymm0, %ymm0
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm6 # 16-byte Folded Reload
+; AVX2-NEXT: # xmm6 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX2-NEXT: vblendps {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-NEXT: vbroadcastsd {{.*#+}} ymm6 = [30064771075,30064771075,30064771075,30064771075]
+; AVX2-NEXT: vpermps %ymm10, %ymm6, %ymm0
+; AVX2-NEXT: vpermps %ymm7, %ymm6, %ymm1
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm1 # 16-byte Folded Reload
+; AVX2-NEXT: # xmm1 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX2-NEXT: vmovaps {{.*#+}} xmm5 = <3,7,u,u>
+; AVX2-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm7 # 32-byte Folded Reload
+; AVX2-NEXT: vblendps {{.*#+}} xmm1 = xmm7[0,1],xmm1[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-NEXT: vpermps %ymm11, %ymm6, %ymm0
+; AVX2-NEXT: vpermps %ymm15, %ymm6, %ymm2
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX2-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm7 # 32-byte Folded Reload
+; AVX2-NEXT: vblendps {{.*#+}} xmm2 = xmm7[0,1],xmm2[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-NEXT: vpermps %ymm9, %ymm6, %ymm2
+; AVX2-NEXT: vpermps %ymm8, %ymm6, %ymm7
+; AVX2-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-NEXT: vunpckhps {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; AVX2-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm4 # 32-byte Folded Reload
+; AVX2-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm3 # 32-byte Folded Reload
+; AVX2-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm4 # 32-byte Folded Reload
+; AVX2-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX2-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm4, (%rsi)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm4, 64(%rsi)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm4, 96(%rsi)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm4, 96(%rdx)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm4, (%rdx)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm4, 64(%rdx)
+; AVX2-NEXT: vmovaps %ymm14, 96(%rcx)
+; AVX2-NEXT: vmovaps %ymm13, (%rcx)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm4, 32(%rcx)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm4, 64(%rcx)
+; AVX2-NEXT: vmovaps %ymm3, 64(%r8)
+; AVX2-NEXT: vmovaps %ymm2, 96(%r8)
+; AVX2-NEXT: vmovaps %ymm0, (%r8)
+; AVX2-NEXT: vmovaps %ymm1, 32(%r8)
+; AVX2-NEXT: addq $696, %rsp # imm = 0x2B8
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+;
+; AVX512-LABEL: load_i32_stride4_vf32:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm0
+; AVX512-NEXT: vmovdqu64 64(%rdi), %zmm1
+; AVX512-NEXT: vmovdqu64 128(%rdi), %zmm2
+; AVX512-NEXT: vmovdqu64 192(%rdi), %zmm3
+; AVX512-NEXT: vmovdqu64 320(%rdi), %zmm4
+; AVX512-NEXT: vmovdqu64 256(%rdi), %zmm5
+; AVX512-NEXT: vmovdqu64 448(%rdi), %zmm6
+; AVX512-NEXT: vmovdqu64 384(%rdi), %zmm7
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,16,20,24,28,0,4,8,12,16,20,24,28]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vmovdqa64 %zmm7, %zmm9
+; AVX512-NEXT: vpermt2d %zmm6, %zmm8, %zmm9
+; AVX512-NEXT: vmovdqa64 %zmm5, %zmm10
+; AVX512-NEXT: vpermt2d %zmm4, %zmm8, %zmm10
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
+; AVX512-NEXT: vmovdqa64 %zmm2, %zmm10
+; AVX512-NEXT: vpermt2d %zmm3, %zmm8, %zmm10
+; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm8
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,17,21,25,29,1,5,9,13,17,21,25,29]
+; AVX512-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vmovdqa64 %zmm7, %zmm11
+; AVX512-NEXT: vpermt2d %zmm6, %zmm10, %zmm11
+; AVX512-NEXT: vmovdqa64 %zmm5, %zmm12
+; AVX512-NEXT: vpermt2d %zmm4, %zmm10, %zmm12
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
+; AVX512-NEXT: vmovdqa64 %zmm2, %zmm12
+; AVX512-NEXT: vpermt2d %zmm3, %zmm10, %zmm12
+; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm10
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,18,22,26,30,2,6,10,14,18,22,26,30]
+; AVX512-NEXT: # zmm12 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vmovdqa64 %zmm7, %zmm13
+; AVX512-NEXT: vpermt2d %zmm6, %zmm12, %zmm13
+; AVX512-NEXT: vmovdqa64 %zmm5, %zmm14
+; AVX512-NEXT: vpermt2d %zmm4, %zmm12, %zmm14
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
+; AVX512-NEXT: vmovdqa64 %zmm2, %zmm14
+; AVX512-NEXT: vpermt2d %zmm3, %zmm12, %zmm14
+; AVX512-NEXT: vpermi2d %zmm1, %zmm0, %zmm12
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,19,23,27,31,3,7,11,15,19,23,27,31]
+; AVX512-NEXT: # zmm14 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermt2d %zmm6, %zmm14, %zmm7
+; AVX512-NEXT: vpermt2d %zmm4, %zmm14, %zmm5
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
+; AVX512-NEXT: vpermt2d %zmm3, %zmm14, %zmm2
+; AVX512-NEXT: vpermt2d %zmm1, %zmm14, %zmm0
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
+; AVX512-NEXT: vmovdqu64 %zmm9, 64(%rsi)
+; AVX512-NEXT: vmovdqu64 %zmm8, (%rsi)
+; AVX512-NEXT: vmovdqu64 %zmm11, 64(%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm10, (%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm13, 64(%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm12, (%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm4, 64(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm0, (%r8)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %wide.vec = load <128 x i32>, <128 x i32>* %in.vec, align 32
+
+ %strided.vec0 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <32 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60, i32 64, i32 68, i32 72, i32 76, i32 80, i32 84, i32 88, i32 92, i32 96, i32 100, i32 104, i32 108, i32 112, i32 116, i32 120, i32 124>
+ %strided.vec1 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <32 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61, i32 65, i32 69, i32 73, i32 77, i32 81, i32 85, i32 89, i32 93, i32 97, i32 101, i32 105, i32 109, i32 113, i32 117, i32 121, i32 125>
+ %strided.vec2 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <32 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62, i32 66, i32 70, i32 74, i32 78, i32 82, i32 86, i32 90, i32 94, i32 98, i32 102, i32 106, i32 110, i32 114, i32 118, i32 122, i32 126>
+ %strided.vec3 = shufflevector <128 x i32> %wide.vec, <128 x i32> poison, <32 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63, i32 67, i32 71, i32 75, i32 79, i32 83, i32 87, i32 91, i32 95, i32 99, i32 103, i32 107, i32 111, i32 115, i32 119, i32 123, i32 127>
+
+ store <32 x i32> %strided.vec0, <32 x i32>* %out.vec0, align 32
+ store <32 x i32> %strided.vec1, <32 x i32>* %out.vec1, align 32
+ store <32 x i32> %strided.vec2, <32 x i32>* %out.vec2, align 32
+ store <32 x i32> %strided.vec3, <32 x i32>* %out.vec3, align 32
+
+ ret void
+}
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-2.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-2.ll
index ede9b77af3abf..e6fafb606c521 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-2.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-2.ll
@@ -382,3 +382,346 @@ define void @load_i64_stride2_vf16(<32 x i64>* %in.vec, <16 x i64>* %out.vec0, <
ret void
}
+
+define void @load_i64_stride2_vf32(<64 x i64>* %in.vec, <32 x i64>* %out.vec0, <32 x i64>* %out.vec1) nounwind {
+; SSE-LABEL: load_i64_stride2_vf32:
+; SSE: # %bb.0:
+; SSE-NEXT: subq $152, %rsp
+; SSE-NEXT: movaps 272(%rdi), %xmm15
+; SSE-NEXT: movaps 208(%rdi), %xmm9
+; SSE-NEXT: movaps 192(%rdi), %xmm5
+; SSE-NEXT: movaps 144(%rdi), %xmm10
+; SSE-NEXT: movaps 128(%rdi), %xmm4
+; SSE-NEXT: movaps 80(%rdi), %xmm13
+; SSE-NEXT: movaps 64(%rdi), %xmm3
+; SSE-NEXT: movaps 304(%rdi), %xmm14
+; SSE-NEXT: movaps 288(%rdi), %xmm12
+; SSE-NEXT: movaps 240(%rdi), %xmm2
+; SSE-NEXT: movaps 224(%rdi), %xmm11
+; SSE-NEXT: movaps 176(%rdi), %xmm7
+; SSE-NEXT: movaps 160(%rdi), %xmm8
+; SSE-NEXT: movaps 112(%rdi), %xmm0
+; SSE-NEXT: movaps 96(%rdi), %xmm6
+; SSE-NEXT: movaps %xmm6, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm0[1]
+; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm3, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm13[0]
+; SSE-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm13[1]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm8, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm7[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm7[1]
+; SSE-NEXT: movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm4, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm10[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm10[1]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm11, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm2[1]
+; SSE-NEXT: movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm5, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm9[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm9[1]
+; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm12, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm14[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm14[1]
+; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 256(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm0, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm15[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm15[1]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 368(%rdi), %xmm0
+; SSE-NEXT: movaps 352(%rdi), %xmm15
+; SSE-NEXT: movaps %xmm15, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm0[1]
+; SSE-NEXT: movaps 336(%rdi), %xmm0
+; SSE-NEXT: movaps 320(%rdi), %xmm11
+; SSE-NEXT: movaps %xmm11, %xmm13
+; SSE-NEXT: movlhps {{.*#+}} xmm13 = xmm13[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm0[1]
+; SSE-NEXT: movaps 432(%rdi), %xmm0
+; SSE-NEXT: movaps 416(%rdi), %xmm12
+; SSE-NEXT: movaps %xmm12, %xmm14
+; SSE-NEXT: movlhps {{.*#+}} xmm14 = xmm14[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm0[1]
+; SSE-NEXT: movaps 400(%rdi), %xmm0
+; SSE-NEXT: movaps 384(%rdi), %xmm5
+; SSE-NEXT: movaps %xmm5, %xmm3
+; SSE-NEXT: movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm0[1]
+; SSE-NEXT: movaps 496(%rdi), %xmm1
+; SSE-NEXT: movaps 480(%rdi), %xmm4
+; SSE-NEXT: movaps %xmm4, %xmm9
+; SSE-NEXT: movlhps {{.*#+}} xmm9 = xmm9[0],xmm1[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm1[1]
+; SSE-NEXT: movaps 464(%rdi), %xmm2
+; SSE-NEXT: movaps 448(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; SSE-NEXT: movaps 32(%rdi), %xmm10
+; SSE-NEXT: movaps 48(%rdi), %xmm2
+; SSE-NEXT: movaps %xmm10, %xmm8
+; SSE-NEXT: movlhps {{.*#+}} xmm8 = xmm8[0],xmm2[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm2[1]
+; SSE-NEXT: movaps (%rdi), %xmm7
+; SSE-NEXT: movaps 16(%rdi), %xmm2
+; SSE-NEXT: movaps %xmm7, %xmm6
+; SSE-NEXT: movlhps {{.*#+}} xmm6 = xmm6[0],xmm2[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm2[1]
+; SSE-NEXT: movaps %xmm0, 224(%rsi)
+; SSE-NEXT: movaps %xmm3, 192(%rsi)
+; SSE-NEXT: movaps %xmm13, 160(%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, 96(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 64(%rsi)
+; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 32(%rsi)
+; SSE-NEXT: movaps %xmm6, (%rsi)
+; SSE-NEXT: movaps %xmm9, 240(%rsi)
+; SSE-NEXT: movaps %xmm14, 208(%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, 144(%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, 80(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 48(%rsi)
+; SSE-NEXT: movaps %xmm8, 16(%rsi)
+; SSE-NEXT: movaps %xmm1, 224(%rdx)
+; SSE-NEXT: movaps %xmm4, 240(%rdx)
+; SSE-NEXT: movaps %xmm5, 192(%rdx)
+; SSE-NEXT: movaps %xmm12, 208(%rdx)
+; SSE-NEXT: movaps %xmm11, 160(%rdx)
+; SSE-NEXT: movaps %xmm15, 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 %xmm7, (%rdx)
+; SSE-NEXT: movaps %xmm10, 16(%rdx)
+; SSE-NEXT: addq $152, %rsp
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: load_i64_stride2_vf32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovaps (%rdi), %ymm0
+; AVX1-NEXT: vmovaps 64(%rdi), %ymm1
+; AVX1-NEXT: vmovaps 96(%rdi), %ymm2
+; AVX1-NEXT: vmovaps 480(%rdi), %ymm3
+; AVX1-NEXT: vmovaps 448(%rdi), %ymm4
+; AVX1-NEXT: vmovaps 288(%rdi), %ymm5
+; AVX1-NEXT: vmovaps 256(%rdi), %ymm6
+; AVX1-NEXT: vmovaps 352(%rdi), %ymm7
+; AVX1-NEXT: vmovaps 320(%rdi), %ymm8
+; AVX1-NEXT: vmovaps 160(%rdi), %ymm9
+; AVX1-NEXT: vmovaps 128(%rdi), %ymm10
+; AVX1-NEXT: vmovaps 224(%rdi), %ymm11
+; AVX1-NEXT: vmovaps 192(%rdi), %ymm12
+; AVX1-NEXT: vmovaps 416(%rdi), %ymm13
+; AVX1-NEXT: vmovaps 384(%rdi), %ymm14
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm14[2,3],ymm13[2,3]
+; AVX1-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm14[0,1],ymm13[0,1]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm12[2,3],ymm11[2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm12[0,1],ymm11[0,1]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm10[2,3],ymm9[2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm10[0,1],ymm9[0,1]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm8[2,3],ymm7[2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm8[0,1],ymm7[0,1]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm6[2,3],ymm5[2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm6[0,1],ymm5[0,1]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm4[2,3],ymm3[2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm4[0,1],ymm3[0,1]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm2[2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[0,1],ymm2[0,1]
+; AVX1-NEXT: vmovaps 32(%rdi), %ymm2
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm2[2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[0,1],ymm2[0,1]
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm13[0],ymm14[0],ymm13[2],ymm14[2]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm13[1],ymm14[1],ymm13[3],ymm14[3]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm11 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm9[0],ymm10[0],ymm9[2],ymm10[2]
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm9 = ymm9[1],ymm10[1],ymm9[3],ymm10[3]
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm7 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm5[0],ymm6[0],ymm5[2],ymm6[2]
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm5 = ymm5[1],ymm6[1],ymm5[3],ymm6[3]
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm15[0],ymm0[0],ymm15[2],ymm0[2]
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm13 = ymm15[1],ymm0[1],ymm15[3],ymm0[3]
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
+; AVX1-NEXT: vmovaps %ymm8, 224(%rsi)
+; AVX1-NEXT: vmovaps %ymm10, 128(%rsi)
+; AVX1-NEXT: vmovaps %ymm12, 160(%rsi)
+; AVX1-NEXT: vmovaps %ymm14, 64(%rsi)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, 96(%rsi)
+; AVX1-NEXT: vmovaps %ymm4, (%rsi)
+; AVX1-NEXT: vmovaps %ymm15, 32(%rsi)
+; AVX1-NEXT: vmovaps %ymm6, 192(%rsi)
+; AVX1-NEXT: vmovaps %ymm13, 192(%rdx)
+; AVX1-NEXT: vmovaps %ymm5, 224(%rdx)
+; AVX1-NEXT: vmovaps %ymm7, 128(%rdx)
+; AVX1-NEXT: vmovaps %ymm9, 160(%rdx)
+; AVX1-NEXT: vmovaps %ymm11, 64(%rdx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm1, 96(%rdx)
+; AVX1-NEXT: vmovaps %ymm0, (%rdx)
+; AVX1-NEXT: vmovaps %ymm3, 32(%rdx)
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: load_i64_stride2_vf32:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vmovaps (%rdi), %ymm5
+; AVX2-NEXT: vmovaps 64(%rdi), %ymm1
+; AVX2-NEXT: vmovaps 96(%rdi), %ymm2
+; AVX2-NEXT: vmovaps 480(%rdi), %ymm3
+; AVX2-NEXT: vmovaps 448(%rdi), %ymm4
+; AVX2-NEXT: vmovaps 288(%rdi), %ymm6
+; AVX2-NEXT: vmovaps 256(%rdi), %ymm7
+; AVX2-NEXT: vmovaps 352(%rdi), %ymm8
+; AVX2-NEXT: vmovaps 320(%rdi), %ymm9
+; AVX2-NEXT: vmovaps 160(%rdi), %ymm10
+; AVX2-NEXT: vmovaps 128(%rdi), %ymm11
+; AVX2-NEXT: vmovaps 224(%rdi), %ymm12
+; AVX2-NEXT: vmovaps 192(%rdi), %ymm13
+; AVX2-NEXT: vmovaps 416(%rdi), %ymm14
+; AVX2-NEXT: vmovaps 384(%rdi), %ymm15
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm13[0],ymm12[0],ymm13[2],ymm12[2]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm12 = ymm13[1],ymm12[1],ymm13[3],ymm12[3]
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm10 = ymm11[1],ymm10[1],ymm11[3],ymm10[3]
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm4[1],ymm3[1],ymm4[3],ymm3[3]
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm15[0],ymm14[0],ymm15[2],ymm14[2]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm14 = ymm15[1],ymm14[1],ymm15[3],ymm14[3]
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX2-NEXT: vmovaps 32(%rdi), %ymm2
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm2[0],ymm5[2],ymm2[2]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm5[1],ymm2[1],ymm5[3],ymm2[3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm5 = ymm7[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm5, 224(%rsi)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm5 = ymm9[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm5, 128(%rsi)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm5 = ymm11[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm5, 160(%rsi)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm5 = ymm13[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm5, 64(%rsi)
+; AVX2-NEXT: vpermpd $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-NEXT: # ymm5 = mem[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm5, 96(%rsi)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm0, (%rsi)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm15[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm0, 32(%rsi)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm4[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm0, 192(%rsi)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm14[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm0, 192(%rdx)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm3[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm0, 224(%rdx)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm6[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm0, 128(%rdx)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm8[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm0, 160(%rdx)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm10[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm0, 64(%rdx)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm12[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm0, 96(%rdx)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm2[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm0, (%rdx)
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm1[0,2,1,3]
+; AVX2-NEXT: vmovaps %ymm0, 32(%rdx)
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+;
+; AVX512-LABEL: load_i64_stride2_vf32:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm0
+; AVX512-NEXT: vmovdqu64 64(%rdi), %zmm1
+; AVX512-NEXT: vmovdqu64 128(%rdi), %zmm2
+; AVX512-NEXT: vmovdqu64 192(%rdi), %zmm3
+; AVX512-NEXT: vmovdqu64 448(%rdi), %zmm4
+; AVX512-NEXT: vmovdqu64 384(%rdi), %zmm5
+; AVX512-NEXT: vmovdqu64 320(%rdi), %zmm6
+; AVX512-NEXT: vmovdqu64 256(%rdi), %zmm7
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm8 = [0,2,4,6,8,10,12,14]
+; AVX512-NEXT: vmovdqa64 %zmm7, %zmm9
+; AVX512-NEXT: vpermt2q %zmm6, %zmm8, %zmm9
+; AVX512-NEXT: vmovdqa64 %zmm5, %zmm10
+; AVX512-NEXT: vpermt2q %zmm4, %zmm8, %zmm10
+; AVX512-NEXT: vmovdqa64 %zmm2, %zmm11
+; AVX512-NEXT: vpermt2q %zmm3, %zmm8, %zmm11
+; AVX512-NEXT: vpermi2q %zmm1, %zmm0, %zmm8
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm12 = [1,3,5,7,9,11,13,15]
+; AVX512-NEXT: vpermt2q %zmm4, %zmm12, %zmm5
+; AVX512-NEXT: vpermt2q %zmm6, %zmm12, %zmm7
+; AVX512-NEXT: vpermt2q %zmm3, %zmm12, %zmm2
+; AVX512-NEXT: vpermt2q %zmm1, %zmm12, %zmm0
+; AVX512-NEXT: vmovdqu64 %zmm10, 192(%rsi)
+; AVX512-NEXT: vmovdqu64 %zmm8, (%rsi)
+; AVX512-NEXT: vmovdqu64 %zmm11, 64(%rsi)
+; AVX512-NEXT: vmovdqu64 %zmm9, 128(%rsi)
+; AVX512-NEXT: vmovdqu64 %zmm7, 128(%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm5, 192(%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm0, (%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm2, 64(%rdx)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %wide.vec = load <64 x i64>, <64 x i64>* %in.vec, align 32
+
+ %strided.vec0 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46, i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62>
+ %strided.vec1 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <32 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31, i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47, i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63>
+
+ store <32 x i64> %strided.vec0, <32 x i64>* %out.vec0, align 32
+ store <32 x i64> %strided.vec1, <32 x i64>* %out.vec1, align 32
+
+ ret void
+}
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.ll
index 854efcbeef223..238f602986e87 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.ll
@@ -410,3 +410,522 @@ define void @load_i64_stride4_vf8(<32 x i64>* %in.vec, <8 x i64>* %out.vec0, <8
ret void
}
+
+define void @load_i64_stride4_vf16(<64 x i64>* %in.vec, <16 x i64>* %out.vec0, <16 x i64>* %out.vec1, <16 x i64>* %out.vec2, <16 x i64>* %out.vec3) nounwind {
+; SSE-LABEL: load_i64_stride4_vf16:
+; SSE: # %bb.0:
+; SSE-NEXT: subq $152, %rsp
+; SSE-NEXT: movaps (%rdi), %xmm8
+; SSE-NEXT: movaps 416(%rdi), %xmm13
+; SSE-NEXT: movaps 384(%rdi), %xmm7
+; SSE-NEXT: movaps 288(%rdi), %xmm15
+; SSE-NEXT: movaps 256(%rdi), %xmm9
+; SSE-NEXT: movaps 160(%rdi), %xmm2
+; SSE-NEXT: movaps 128(%rdi), %xmm10
+; SSE-NEXT: movaps 480(%rdi), %xmm3
+; SSE-NEXT: movaps 448(%rdi), %xmm11
+; SSE-NEXT: movaps 352(%rdi), %xmm4
+; SSE-NEXT: movaps 320(%rdi), %xmm12
+; SSE-NEXT: movaps 224(%rdi), %xmm5
+; SSE-NEXT: movaps 192(%rdi), %xmm14
+; SSE-NEXT: movaps 96(%rdi), %xmm6
+; SSE-NEXT: movaps 64(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm0, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm6[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm6[1]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm14, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm5[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm5[1]
+; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm12, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm4[1]
+; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm11, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; SSE-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm3[1]
+; SSE-NEXT: movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm10, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm2[1]
+; SSE-NEXT: movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm9, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm15[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm15[1]
+; SSE-NEXT: movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm7, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm13[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm13[1]
+; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 32(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm8, %xmm15
+; SSE-NEXT: movlhps {{.*#+}} xmm15 = xmm15[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
+; SSE-NEXT: movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 112(%rdi), %xmm0
+; SSE-NEXT: movaps 80(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, %xmm2
+; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 240(%rdi), %xmm0
+; SSE-NEXT: movaps 208(%rdi), %xmm12
+; SSE-NEXT: movaps %xmm12, %xmm14
+; SSE-NEXT: movlhps {{.*#+}} xmm14 = xmm14[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm0[1]
+; SSE-NEXT: movaps 368(%rdi), %xmm0
+; SSE-NEXT: movaps 336(%rdi), %xmm10
+; SSE-NEXT: movaps %xmm10, %xmm13
+; SSE-NEXT: movlhps {{.*#+}} xmm13 = xmm13[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm0[1]
+; SSE-NEXT: movaps 496(%rdi), %xmm0
+; SSE-NEXT: movaps 464(%rdi), %xmm9
+; SSE-NEXT: movaps %xmm9, %xmm11
+; SSE-NEXT: movlhps {{.*#+}} xmm11 = xmm11[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm0[1]
+; SSE-NEXT: movaps 176(%rdi), %xmm0
+; SSE-NEXT: movaps 144(%rdi), %xmm7
+; SSE-NEXT: movaps %xmm7, %xmm8
+; SSE-NEXT: movlhps {{.*#+}} xmm8 = xmm8[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm0[1]
+; SSE-NEXT: movaps 304(%rdi), %xmm0
+; SSE-NEXT: movaps 272(%rdi), %xmm5
+; SSE-NEXT: movaps %xmm5, %xmm6
+; SSE-NEXT: movlhps {{.*#+}} xmm6 = xmm6[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm0[1]
+; SSE-NEXT: movaps 432(%rdi), %xmm0
+; SSE-NEXT: movaps 400(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, %xmm3
+; SSE-NEXT: movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT: movaps 16(%rdi), %xmm2
+; SSE-NEXT: movaps 48(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm2, %xmm4
+; SSE-NEXT: movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 96(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 64(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 32(%rsi)
+; SSE-NEXT: movaps %xmm15, (%rsi)
+; SSE-NEXT: movaps (%rsp), %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, 80(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 48(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 16(%rsi)
+; SSE-NEXT: movaps {{[-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, 64(%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, 112(%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, 48(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 16(%rdx)
+; SSE-NEXT: movaps %xmm3, 96(%rcx)
+; SSE-NEXT: movaps %xmm6, 64(%rcx)
+; SSE-NEXT: movaps %xmm8, 32(%rcx)
+; SSE-NEXT: movaps %xmm4, (%rcx)
+; SSE-NEXT: movaps %xmm11, 112(%rcx)
+; SSE-NEXT: movaps %xmm13, 80(%rcx)
+; SSE-NEXT: movaps %xmm14, 48(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 16(%rcx)
+; SSE-NEXT: movaps %xmm1, 96(%r8)
+; SSE-NEXT: movaps %xmm5, 64(%r8)
+; SSE-NEXT: movaps %xmm7, 32(%r8)
+; SSE-NEXT: movaps %xmm2, (%r8)
+; SSE-NEXT: movaps %xmm9, 112(%r8)
+; SSE-NEXT: movaps %xmm10, 80(%r8)
+; SSE-NEXT: movaps %xmm12, 48(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 16(%r8)
+; SSE-NEXT: addq $152, %rsp
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: load_i64_stride4_vf16:
+; AVX1: # %bb.0:
+; AVX1-NEXT: subq $296, %rsp # imm = 0x128
+; AVX1-NEXT: vmovaps 224(%rdi), %xmm8
+; AVX1-NEXT: vmovaps 192(%rdi), %xmm9
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm9[0],xmm8[0]
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 96(%rdi), %xmm10
+; AVX1-NEXT: vmovaps 64(%rdi), %xmm11
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm11[0],xmm10[0]
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 352(%rdi), %xmm4
+; AVX1-NEXT: vmovaps 320(%rdi), %xmm5
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm5[0],xmm4[0]
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 160(%rdi), %xmm12
+; AVX1-NEXT: vmovaps 128(%rdi), %xmm1
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm1[0],xmm12[0]
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 32(%rdi), %xmm2
+; AVX1-NEXT: vmovaps (%rdi), %xmm3
+; AVX1-NEXT: vmovaps 288(%rdi), %xmm6
+; AVX1-NEXT: vmovaps 256(%rdi), %xmm0
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
+; AVX1-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm4 = xmm0[0],xmm6[0]
+; AVX1-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm6[1]
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm3[0],xmm2[0]
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm9[1],xmm8[1]
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 416(%rdi), %xmm0
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm12[1]
+; AVX1-NEXT: vmovaps %xmm1, (%rsp) # 16-byte Spill
+; AVX1-NEXT: vmovaps 480(%rdi), %xmm1
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm11[1],xmm10[1]
+; AVX1-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 448(%rdi), %xmm4
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm1[0]
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm1[1]
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 384(%rdi), %xmm1
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 224(%rdi), %ymm13
+; AVX1-NEXT: vmovaps 192(%rdi), %ymm12
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm12[0],ymm13[0],ymm12[2],ymm13[2]
+; AVX1-NEXT: vmovaps 176(%rdi), %xmm11
+; AVX1-NEXT: vmovaps 144(%rdi), %xmm10
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm5 = xmm10[0],xmm11[0]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 352(%rdi), %ymm0
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 320(%rdi), %ymm1
+; AVX1-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-NEXT: vmovaps 304(%rdi), %xmm15
+; AVX1-NEXT: vmovaps 272(%rdi), %xmm14
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm9 = xmm14[0],xmm15[0]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 480(%rdi), %ymm8
+; AVX1-NEXT: vmovaps 448(%rdi), %ymm7
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
+; AVX1-NEXT: vmovaps 432(%rdi), %xmm5
+; AVX1-NEXT: vmovaps 400(%rdi), %xmm4
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm9 = xmm4[0],xmm5[0]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 96(%rdi), %ymm3
+; AVX1-NEXT: vmovaps 64(%rdi), %ymm2
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
+; AVX1-NEXT: vmovaps 48(%rdi), %xmm1
+; AVX1-NEXT: vmovaps 16(%rdi), %xmm0
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm9 = xmm0[0],xmm1[0]
+; AVX1-NEXT: vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm9 = ymm12[1],ymm13[1],ymm12[3],ymm13[3]
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm10[1],xmm11[1]
+; AVX1-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm4[1],xmm5[1]
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-NEXT: # ymm2 = ymm2[1],mem[1],ymm2[3],mem[3]
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm14[1],xmm15[1]
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 112(%rsi)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 96(%rsi)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 64(%rsi)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, (%rsi)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 32(%rsi)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 80(%rsi)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 16(%rsi)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 48(%rsi)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 96(%rdx)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 112(%rdx)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, (%rdx)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 16(%rdx)
+; AVX1-NEXT: vmovaps (%rsp), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 32(%rdx)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 48(%rdx)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 64(%rdx)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm3, 80(%rdx)
+; AVX1-NEXT: vmovaps %ymm6, (%rcx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm3, 96(%rcx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm3, 64(%rcx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm3, 32(%rcx)
+; AVX1-NEXT: vmovaps %ymm2, 64(%r8)
+; AVX1-NEXT: vmovaps %ymm1, 96(%r8)
+; AVX1-NEXT: vmovaps %ymm0, (%r8)
+; AVX1-NEXT: vmovaps %ymm10, 32(%r8)
+; AVX1-NEXT: addq $296, %rsp # imm = 0x128
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: load_i64_stride4_vf16:
+; AVX2: # %bb.0:
+; AVX2-NEXT: subq $296, %rsp # imm = 0x128
+; AVX2-NEXT: vmovaps 224(%rdi), %xmm11
+; AVX2-NEXT: vmovaps 192(%rdi), %xmm15
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm0 = xmm15[0],xmm11[0]
+; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 96(%rdi), %xmm7
+; AVX2-NEXT: vmovaps (%rdi), %xmm9
+; AVX2-NEXT: vmovaps 64(%rdi), %xmm4
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm0 = xmm4[0],xmm7[0]
+; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 352(%rdi), %xmm8
+; AVX2-NEXT: vmovaps 320(%rdi), %xmm6
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm0 = xmm6[0],xmm8[0]
+; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 160(%rdi), %xmm3
+; AVX2-NEXT: vmovaps 128(%rdi), %xmm2
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm0 = xmm2[0],xmm3[0]
+; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 288(%rdi), %xmm1
+; AVX2-NEXT: vmovaps 256(%rdi), %xmm5
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm0 = xmm5[0],xmm1[0]
+; AVX2-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm6[1],xmm8[1]
+; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 416(%rdi), %xmm0
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm1[1]
+; AVX2-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 384(%rdi), %xmm1
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm15[1],xmm11[1]
+; AVX2-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 480(%rdi), %xmm11
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; AVX2-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 448(%rdi), %xmm2
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm4[1],xmm7[1]
+; AVX2-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm11[0]
+; AVX2-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm2[1],xmm11[1]
+; AVX2-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 32(%rdi), %xmm0
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm0[0]
+; AVX2-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm9[1],xmm0[1]
+; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 160(%rdi), %ymm9
+; AVX2-NEXT: vmovaps 128(%rdi), %ymm8
+; AVX2-NEXT: vmovaps 224(%rdi), %ymm5
+; AVX2-NEXT: vmovaps 192(%rdi), %ymm4
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm8[0],ymm9[0],ymm8[2],ymm9[2]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm0[2,3]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 288(%rdi), %ymm0
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 256(%rdi), %ymm1
+; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 352(%rdi), %ymm14
+; AVX2-NEXT: vmovaps 320(%rdi), %ymm13
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm13[0],ymm14[0],ymm13[2],ymm14[2]
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm11[2,3]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 416(%rdi), %ymm10
+; AVX2-NEXT: vmovaps 384(%rdi), %ymm7
+; AVX2-NEXT: vmovaps 480(%rdi), %ymm6
+; AVX2-NEXT: vmovaps 448(%rdi), %ymm3
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm3[0],ymm6[0],ymm3[2],ymm6[2]
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm7[0],ymm10[0],ymm7[2],ymm10[2]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm15[2,3]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 32(%rdi), %ymm11
+; AVX2-NEXT: vmovaps (%rdi), %ymm2
+; AVX2-NEXT: vmovaps 96(%rdi), %ymm1
+; AVX2-NEXT: vmovaps 64(%rdi), %ymm0
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm2[0],ymm11[0],ymm2[2],ymm11[2]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm15[2,3],ymm12[2,3]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm5 = ymm8[1],ymm9[1],ymm8[3],ymm9[3]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm5[2,3],ymm4[2,3]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm11[1],ymm2[3],ymm11[3]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm6[1],ymm3[3],ymm6[3]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm7[1],ymm10[1],ymm7[3],ymm10[3]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm13[1],ymm14[1],ymm13[3],ymm14[3]
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-NEXT: # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm3[2,3],ymm2[2,3]
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 112(%rsi)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 96(%rsi)
+; AVX2-NEXT: vmovaps (%rsp), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 64(%rsi)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, (%rsi)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 32(%rsi)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 80(%rsi)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 16(%rsi)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 48(%rsi)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 96(%rdx)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 112(%rdx)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, (%rdx)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 16(%rdx)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 32(%rdx)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 48(%rdx)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 64(%rdx)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm3, 80(%rdx)
+; AVX2-NEXT: vmovaps %ymm12, (%rcx)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm3, 96(%rcx)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm3, 64(%rcx)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm3, 32(%rcx)
+; AVX2-NEXT: vmovaps %ymm2, 64(%r8)
+; AVX2-NEXT: vmovaps %ymm1, 96(%r8)
+; AVX2-NEXT: vmovaps %ymm0, (%r8)
+; AVX2-NEXT: vmovaps %ymm5, 32(%r8)
+; AVX2-NEXT: addq $296, %rsp # imm = 0x128
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+;
+; AVX512-LABEL: load_i64_stride4_vf16:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm0
+; AVX512-NEXT: vmovdqu64 64(%rdi), %zmm1
+; AVX512-NEXT: vmovdqu64 128(%rdi), %zmm2
+; AVX512-NEXT: vmovdqu64 192(%rdi), %zmm3
+; AVX512-NEXT: vmovdqu64 320(%rdi), %zmm4
+; AVX512-NEXT: vmovdqu64 256(%rdi), %zmm5
+; AVX512-NEXT: vmovdqu64 448(%rdi), %zmm6
+; AVX512-NEXT: vmovdqu64 384(%rdi), %zmm7
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,4,8,12,0,4,8,12]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vmovdqa64 %zmm7, %zmm9
+; AVX512-NEXT: vpermt2q %zmm6, %zmm8, %zmm9
+; AVX512-NEXT: vmovdqa64 %zmm5, %zmm10
+; AVX512-NEXT: vpermt2q %zmm4, %zmm8, %zmm10
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm9 = zmm10[0,1,2,3],zmm9[4,5,6,7]
+; AVX512-NEXT: vmovdqa64 %zmm2, %zmm10
+; AVX512-NEXT: vpermt2q %zmm3, %zmm8, %zmm10
+; AVX512-NEXT: vpermi2q %zmm1, %zmm0, %zmm8
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm8 = zmm8[0,1,2,3],zmm10[4,5,6,7]
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm10 = [1,5,9,13,1,5,9,13]
+; AVX512-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vmovdqa64 %zmm7, %zmm11
+; AVX512-NEXT: vpermt2q %zmm6, %zmm10, %zmm11
+; AVX512-NEXT: vmovdqa64 %zmm5, %zmm12
+; AVX512-NEXT: vpermt2q %zmm4, %zmm10, %zmm12
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm11 = zmm12[0,1,2,3],zmm11[4,5,6,7]
+; AVX512-NEXT: vmovdqa64 %zmm2, %zmm12
+; AVX512-NEXT: vpermt2q %zmm3, %zmm10, %zmm12
+; AVX512-NEXT: vpermi2q %zmm1, %zmm0, %zmm10
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm10 = zmm10[0,1,2,3],zmm12[4,5,6,7]
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm12 = [2,6,10,14,2,6,10,14]
+; AVX512-NEXT: # zmm12 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vmovdqa64 %zmm7, %zmm13
+; AVX512-NEXT: vpermt2q %zmm6, %zmm12, %zmm13
+; AVX512-NEXT: vmovdqa64 %zmm5, %zmm14
+; AVX512-NEXT: vpermt2q %zmm4, %zmm12, %zmm14
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm13 = zmm14[0,1,2,3],zmm13[4,5,6,7]
+; AVX512-NEXT: vmovdqa64 %zmm2, %zmm14
+; AVX512-NEXT: vpermt2q %zmm3, %zmm12, %zmm14
+; AVX512-NEXT: vpermi2q %zmm1, %zmm0, %zmm12
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm12 = zmm12[0,1,2,3],zmm14[4,5,6,7]
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm14 = [3,7,11,15,3,7,11,15]
+; AVX512-NEXT: # zmm14 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermt2q %zmm6, %zmm14, %zmm7
+; AVX512-NEXT: vpermt2q %zmm4, %zmm14, %zmm5
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm5[0,1,2,3],zmm7[4,5,6,7]
+; AVX512-NEXT: vpermt2q %zmm3, %zmm14, %zmm2
+; AVX512-NEXT: vpermt2q %zmm1, %zmm14, %zmm0
+; AVX512-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm2[4,5,6,7]
+; AVX512-NEXT: vmovdqu64 %zmm9, 64(%rsi)
+; AVX512-NEXT: vmovdqu64 %zmm8, (%rsi)
+; AVX512-NEXT: vmovdqu64 %zmm11, 64(%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm10, (%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm13, 64(%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm12, (%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm4, 64(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm0, (%r8)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %wide.vec = load <64 x i64>, <64 x i64>* %in.vec, align 32
+
+ %strided.vec0 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
+ %strided.vec1 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
+ %strided.vec2 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
+ %strided.vec3 = shufflevector <64 x i64> %wide.vec, <64 x i64> poison, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
+
+ store <16 x i64> %strided.vec0, <16 x i64>* %out.vec0, align 32
+ store <16 x i64> %strided.vec1, <16 x i64>* %out.vec1, align 32
+ store <16 x i64> %strided.vec2, <16 x i64>* %out.vec2, align 32
+ store <16 x i64> %strided.vec3, <16 x i64>* %out.vec3, align 32
+
+ ret void
+}
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-5.ll
index a1913f66c521e..e06b1faeae676 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-5.ll
@@ -1225,3 +1225,1292 @@ define void @vf16(<16 x i16>* %in.vecptr0, <16 x i16>* %in.vecptr1, <16 x i16>*
ret void
}
+
+define void @vf32(<32 x i16>* %in.vecptr0, <32 x i16>* %in.vecptr1, <32 x i16>* %in.vecptr2, <32 x i16>* %in.vecptr3, <32 x i16>* %in.vecptr4, <160 x i16>* %out.vec) nounwind {
+; SSE-LABEL: vf32:
+; SSE: # %bb.0:
+; SSE-NEXT: subq $328, %rsp # imm = 0x148
+; SSE-NEXT: movdqa (%rdi), %xmm10
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa (%rsi), %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa (%rdx), %xmm14
+; SSE-NEXT: movdqa 16(%rdx), %xmm9
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa (%rcx), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 16(%rcx), %xmm8
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa (%r8), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 16(%r8), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{.*#+}} xmm4 = [65535,65535,65535,0,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm1, %xmm0
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm2[3,3,3,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pand %xmm4, %xmm1
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,65535,0,0,65535,65535,65535,0]
+; SSE-NEXT: movdqa %xmm2, %xmm0
+; SSE-NEXT: pandn %xmm1, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm14[1,1,2,2]
+; SSE-NEXT: movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,65535,65535,0,65535]
+; SSE-NEXT: movdqa %xmm1, %xmm7
+; SSE-NEXT: pandn %xmm6, %xmm7
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm5[3,3,3,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pand %xmm1, %xmm6
+; SSE-NEXT: por %xmm7, %xmm6
+; SSE-NEXT: pand %xmm2, %xmm6
+; SSE-NEXT: por %xmm0, %xmm6
+; SSE-NEXT: movdqa {{.*#+}} xmm12 = [65535,65535,65535,65535,0,65535,65535,65535]
+; SSE-NEXT: pand %xmm12, %xmm6
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: pandn %xmm10, %xmm0
+; SSE-NEXT: por %xmm6, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm3, %xmm0
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm8[3,3,3,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pand %xmm4, %xmm6
+; SSE-NEXT: por %xmm0, %xmm6
+; SSE-NEXT: movdqa %xmm2, %xmm11
+; SSE-NEXT: pandn %xmm6, %xmm11
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm9[1,1,2,2]
+; SSE-NEXT: movdqa %xmm1, %xmm7
+; SSE-NEXT: pandn %xmm6, %xmm7
+; SSE-NEXT: movdqa 16(%rsi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pand %xmm1, %xmm0
+; SSE-NEXT: por %xmm7, %xmm0
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm11, %xmm0
+; SSE-NEXT: movdqa 16(%rdi), %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: movdqa %xmm12, %xmm3
+; SSE-NEXT: pandn %xmm5, %xmm3
+; SSE-NEXT: por %xmm0, %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 32(%r8), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm3, %xmm0
+; SSE-NEXT: movdqa 32(%rcx), %xmm3
+; SSE-NEXT: movdqa %xmm3, (%rsp) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm3[3,3,3,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pand %xmm4, %xmm7
+; SSE-NEXT: por %xmm0, %xmm7
+; SSE-NEXT: movdqa %xmm2, %xmm13
+; SSE-NEXT: pandn %xmm7, %xmm13
+; SSE-NEXT: movdqa 32(%rdx), %xmm15
+; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm15[1,1,2,2]
+; SSE-NEXT: movdqa %xmm1, %xmm0
+; SSE-NEXT: pandn %xmm11, %xmm0
+; SSE-NEXT: movdqa 32(%rsi), %xmm11
+; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm11[3,3,3,3,4,5,6,7]
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pand %xmm1, %xmm5
+; SSE-NEXT: por %xmm0, %xmm5
+; SSE-NEXT: pand %xmm2, %xmm5
+; SSE-NEXT: por %xmm13, %xmm5
+; SSE-NEXT: pand %xmm12, %xmm5
+; SSE-NEXT: movdqa 32(%rdi), %xmm9
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: pandn %xmm9, %xmm0
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: por %xmm5, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 48(%rcx), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: movdqa 48(%r8), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pandn %xmm3, %xmm4
+; SSE-NEXT: por %xmm0, %xmm4
+; SSE-NEXT: movdqa %xmm2, %xmm0
+; SSE-NEXT: pandn %xmm4, %xmm0
+; SSE-NEXT: movdqa 48(%rsi), %xmm10
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm10[3,3,3,3,4,5,6,7]
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pand %xmm1, %xmm4
+; SSE-NEXT: movdqa 48(%rdx), %xmm8
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm8[1,1,2,2]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: por %xmm4, %xmm1
+; SSE-NEXT: pand %xmm2, %xmm1
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: pand %xmm12, %xmm1
+; SSE-NEXT: movdqa 48(%rdi), %xmm7
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: pandn %xmm7, %xmm0
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm14[4],xmm0[5],xmm14[5],xmm0[6],xmm14[6],xmm0[7],xmm14[7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
+; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: pandn %xmm0, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm13[4],xmm6[5],xmm13[5],xmm6[6],xmm13[6],xmm6[7],xmm13[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,7]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm4, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm3, %xmm4
+; SSE-NEXT: psrld $16, %xmm4
+; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: pandn %xmm4, %xmm1
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
+; SSE-NEXT: movdqa %xmm2, %xmm5
+; SSE-NEXT: pandn %xmm0, %xmm5
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,7]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm5, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm5
+; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm15[4],xmm11[5],xmm15[5],xmm11[6],xmm15[6],xmm11[7],xmm15[7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,4,4,4,4]
+; SSE-NEXT: movdqa %xmm2, %xmm5
+; SSE-NEXT: pandn %xmm0, %xmm5
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,7]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm5, %xmm0
+; SSE-NEXT: movdqa (%rsp), %xmm5 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm5
+; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm8[4],xmm10[5],xmm8[5],xmm10[6],xmm8[6],xmm10[7],xmm8[7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm10[0,1,2,3,4,4,4,4]
+; SSE-NEXT: movdqa %xmm2, %xmm5
+; SSE-NEXT: pandn %xmm0, %xmm5
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,7]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm5, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm5
+; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
+; SSE-NEXT: movdqa %xmm2, %xmm5
+; SSE-NEXT: pandn %xmm0, %xmm5
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm8[0],xmm13[1],xmm8[1],xmm13[2],xmm8[2],xmm13[3],xmm8[3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm13[0,1,3,2,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,1,1]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm5, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm14[1,1,1,1]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm3[4],xmm14[5],xmm3[5],xmm14[6],xmm3[6],xmm14[7],xmm3[7]
+; SSE-NEXT: movdqa %xmm3, %xmm5
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; SSE-NEXT: movdqa %xmm12, %xmm3
+; SSE-NEXT: pandn %xmm1, %xmm3
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,4,5,6,6]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,2,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,7]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: pandn %xmm9, %xmm1
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,4,7,6,6]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm14[0,1,2,3,4,5,7,6]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: psrlq $48, %xmm1
+; SSE-NEXT: movdqa %xmm12, %xmm3
+; SSE-NEXT: pandn %xmm1, %xmm3
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm1, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: pandn %xmm0, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[0,1,3,2,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,1,1]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm3, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm1[1,1,1,1]
+; SSE-NEXT: movdqa %xmm5, %xmm3
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm5[0,1,0,1]
+; SSE-NEXT: movdqa %xmm12, %xmm11
+; SSE-NEXT: pandn %xmm6, %xmm11
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm11
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,4,5,6,6]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: movdqa %xmm2, %xmm6
+; SSE-NEXT: pandn %xmm0, %xmm6
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,2,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,7]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm6, %xmm0
+; SSE-NEXT: movdqa %xmm12, %xmm9
+; SSE-NEXT: pandn %xmm10, %xmm9
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm9
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,4,7,6,6]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: pandn %xmm0, %xmm4
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5,7,6]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm4, %xmm0
+; SSE-NEXT: movdqa %xmm7, %xmm1
+; SSE-NEXT: psrlq $48, %xmm1
+; SSE-NEXT: movdqa %xmm12, %xmm10
+; SSE-NEXT: pandn %xmm1, %xmm10
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm10
+; SSE-NEXT: movdqa %xmm15, %xmm5
+; SSE-NEXT: movdqa %xmm15, %xmm0
+; SSE-NEXT: movdqa (%rsp), %xmm3 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
+; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: pandn %xmm0, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1],xmm7[2],xmm1[2],xmm7[3],xmm1[3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm7[0,1,3,2,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,1,1]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm4, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm15[1,1,1,1]
+; SSE-NEXT: movdqa %xmm3, %xmm4
+; SSE-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,1]
+; SSE-NEXT: movdqa %xmm12, %xmm6
+; SSE-NEXT: pandn %xmm3, %xmm6
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm6
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,4,5,6,6]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: pandn %xmm0, %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,2,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,7]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm12, %xmm4
+; SSE-NEXT: pandn %xmm15, %xmm4
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm4
+; SSE-NEXT: pshufhw $172, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,2,3,4,7,6,6]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: pandn %xmm0, %xmm3
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,4,5,7,6]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm3, %xmm0
+; SSE-NEXT: psrlq $48, %xmm1
+; SSE-NEXT: movdqa %xmm12, %xmm15
+; SSE-NEXT: pandn %xmm1, %xmm15
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm15
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm1, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1],xmm0[2],xmm14[2],xmm0[3],xmm14[3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: pandn %xmm0, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1],xmm5[2],xmm8[2],xmm5[3],xmm8[3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm5[0,1,3,2,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,1,1]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm3, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm1[1,1,1,1]
+; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm14[4],xmm1[5],xmm14[5],xmm1[6],xmm14[6],xmm1[7],xmm14[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1],xmm3[2],xmm7[2],xmm3[3],xmm7[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,1,0,1]
+; SSE-NEXT: movdqa %xmm12, %xmm14
+; SSE-NEXT: pandn %xmm7, %xmm14
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm14
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,4,5,6,6]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: movdqa %xmm2, %xmm7
+; SSE-NEXT: pandn %xmm0, %xmm7
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,2,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,7]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm7, %xmm0
+; SSE-NEXT: movdqa %xmm12, %xmm3
+; SSE-NEXT: pandn %xmm13, %xmm3
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm3
+; SSE-NEXT: pshufhw $172, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,2,3,4,7,6,6]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm1[0,1,2,3,4,5,7,6]
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[2,1,3,3]
+; SSE-NEXT: pand %xmm2, %xmm7
+; SSE-NEXT: pandn %xmm0, %xmm2
+; SSE-NEXT: por %xmm7, %xmm2
+; SSE-NEXT: pand %xmm12, %xmm2
+; SSE-NEXT: psrlq $48, %xmm8
+; SSE-NEXT: pandn %xmm8, %xmm12
+; SSE-NEXT: por %xmm2, %xmm12
+; SSE-NEXT: movdqa %xmm12, 304(%r9)
+; SSE-NEXT: movdqa %xmm3, 256(%r9)
+; SSE-NEXT: movdqa %xmm14, 240(%r9)
+; SSE-NEXT: movdqa %xmm15, 224(%r9)
+; SSE-NEXT: movdqa %xmm4, 176(%r9)
+; SSE-NEXT: movdqa %xmm6, 160(%r9)
+; SSE-NEXT: movdqa %xmm10, 144(%r9)
+; SSE-NEXT: movdqa %xmm9, 96(%r9)
+; SSE-NEXT: movdqa %xmm11, 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, 16(%r9)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, (%r9)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 288(%r9)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 272(%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, 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, 48(%r9)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 32(%r9)
+; SSE-NEXT: addq $328, %rsp # imm = 0x148
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: vf32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: subq $152, %rsp
+; AVX1-NEXT: vmovdqa 32(%rdi), %xmm14
+; AVX1-NEXT: vmovdqa 48(%rdi), %xmm9
+; AVX1-NEXT: vmovdqa 32(%r8), %xmm8
+; AVX1-NEXT: vmovdqa 48(%r8), %xmm10
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm8[4],xmm14[4],xmm8[5],xmm14[5],xmm8[6],xmm14[6],xmm8[7],xmm14[7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,6]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm9[0],xmm10[0],xmm9[1],xmm10[1],xmm9[2],xmm10[2],xmm9[3],xmm10[3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,1,2,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,1,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT: vmovaps {{.*#+}} ymm12 = [65535,65535,0,0,65535,65535,65535,0,0,65535,65535,65535,0,0,65535,65535]
+; AVX1-NEXT: vandnps %ymm0, %ymm12, %ymm2
+; AVX1-NEXT: vmovdqa 32(%rcx), %xmm0
+; AVX1-NEXT: vmovdqa 48(%rcx), %xmm11
+; AVX1-NEXT: vmovdqa 32(%rdx), %xmm3
+; AVX1-NEXT: vmovdqa 48(%rdx), %xmm13
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,7,6]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[2,1,3,3]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm13[0],xmm11[0],xmm13[1],xmm11[1],xmm13[2],xmm11[2],xmm13[3],xmm11[3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,2,2,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,0,2,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm5, %ymm5
+; AVX1-NEXT: vandps %ymm5, %ymm12, %ymm5
+; AVX1-NEXT: vorps %ymm2, %ymm5, %ymm5
+; AVX1-NEXT: vmovdqa 32(%rsi), %xmm2
+; AVX1-NEXT: vmovdqa 48(%rsi), %xmm1
+; AVX1-NEXT: vpsrlq $48, %xmm2, %xmm6
+; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2,3],xmm6[4],xmm5[5,6,7]
+; AVX1-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm6 = xmm1[1,0,2,3,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,0,0,0]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm6[1],xmm5[2,3,4,5],xmm6[6],xmm5[7]
+; AVX1-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm5 = xmm0[3,3,3,3,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,4,4]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm8[3],xmm5[4,5,6,7]
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm8[4],xmm0[4],xmm8[5],xmm0[5],xmm8[6],xmm0[6],xmm8[7],xmm0[7]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,3,2,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,1,1,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm5, %ymm5
+; AVX1-NEXT: vandnps %ymm5, %ymm12, %ymm5
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm3[1,1,2,2]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm7 = xmm2[3,3,3,3,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,4,4]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm6 = xmm7[0],xmm6[1],xmm7[2,3,4,5],xmm6[6],xmm7[7]
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm7 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,4,4]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
+; AVX1-NEXT: vandps %ymm6, %ymm12, %ymm6
+; AVX1-NEXT: vorps %ymm5, %ymm6, %ymm5
+; AVX1-NEXT: vextractf128 $1, %ymm5, %xmm6
+; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm14[2,3,2,3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0],xmm7[1],xmm6[2,3,4,5],xmm7[6],xmm6[7]
+; AVX1-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm6 = xmm1[3,3,3,3,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,4,4]
+; AVX1-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3],xmm9[4],xmm6[5,6,7]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,6,6]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm7, %ymm6
+; AVX1-NEXT: vmovdqa %xmm11, (%rsp) # 16-byte Spill
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm7 = xmm11[3,3,3,3,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,4,4]
+; AVX1-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1,2],xmm10[3],xmm7[4,5,6,7]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[1,1,2,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,5,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm4, %ymm4
+; AVX1-NEXT: vandnps %ymm6, %ymm12, %ymm6
+; AVX1-NEXT: vandps %ymm4, %ymm12, %ymm4
+; AVX1-NEXT: vorps %ymm6, %ymm4, %ymm4
+; AVX1-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm6 = xmm13[0],zero,xmm13[1],zero
+; AVX1-NEXT: vpblendw {{.*#+}} xmm6 = xmm4[0,1,2,3],xmm6[4],xmm4[5,6,7]
+; AVX1-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vextractf128 $1, %ymm4, %xmm4
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm13[1,1,2,2]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0],xmm6[1],xmm4[2,3,4,5],xmm6[6],xmm4[7]
+; AVX1-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm3 = xmm4[0,1,2,2,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,0,2,1]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm5[0,1,2,3],xmm14[4],xmm5[5,6,7]
+; AVX1-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm14[0],xmm2[0],xmm14[1],xmm2[1],xmm14[2],xmm2[2],xmm14[3],xmm2[3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm1[0,1,3,2,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,1,1]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,6]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-NEXT: vmovdqa (%rcx), %xmm4
+; AVX1-NEXT: vandnps %ymm0, %ymm12, %ymm0
+; AVX1-NEXT: vandps %ymm1, %ymm12, %ymm1
+; AVX1-NEXT: vorps %ymm0, %ymm1, %ymm0
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm8[0,1,0,1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm0[0,1,2,3],xmm1[4],xmm0[5,6,7]
+; AVX1-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5],xmm1[6],xmm0[7]
+; AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm4[3,3,3,3,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
+; AVX1-NEXT: vmovdqa (%r8), %xmm10
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm10[3],xmm0[4,5,6,7]
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm10[4],xmm4[4],xmm10[5],xmm4[5],xmm10[6],xmm4[6],xmm10[7],xmm4[7]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,3,2,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT: vmovdqa (%rsi), %xmm8
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm8[3,3,3,3,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,4,4,4]
+; AVX1-NEXT: vmovdqa (%rdx), %xmm3
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm3[1,1,2,2]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm5[1],xmm2[2,3,4,5],xmm5[6],xmm2[7]
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm8[4],xmm3[4],xmm8[5],xmm3[5],xmm8[6],xmm3[6],xmm8[7],xmm3[7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,4,4]
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm2, %ymm2
+; AVX1-NEXT: vandnps %ymm1, %ymm12, %ymm1
+; AVX1-NEXT: vandps %ymm2, %ymm12, %ymm2
+; AVX1-NEXT: vorps %ymm1, %ymm2, %ymm14
+; AVX1-NEXT: vmovdqa (%rdi), %xmm2
+; AVX1-NEXT: vextractf128 $1, %ymm14, %xmm1
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm2[2,3,2,3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm5[1],xmm1[2,3,4,5],xmm5[6],xmm1[7]
+; AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa 16(%rsi), %xmm5
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm1 = xmm5[3,3,3,3,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
+; AVX1-NEXT: vmovdqa 16(%rdi), %xmm13
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm13[4],xmm1[5,6,7]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm13[0],xmm5[0],xmm13[1],xmm5[1],xmm13[2],xmm5[2],xmm13[3],xmm5[3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,6,6]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm6, %ymm1
+; AVX1-NEXT: vmovdqa 16(%rcx), %xmm9
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm7 = xmm9[3,3,3,3,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,4,4,4,4]
+; AVX1-NEXT: vmovdqa 16(%r8), %xmm11
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm11[3],xmm0[4,5,6,7]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm15 = xmm9[0],xmm11[0],xmm9[1],xmm11[1],xmm9[2],xmm11[2],xmm9[3],xmm11[3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm15[1,1,2,3]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,5,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm7, %ymm0
+; AVX1-NEXT: vandnps %ymm1, %ymm12, %ymm1
+; AVX1-NEXT: vandps %ymm0, %ymm12, %ymm0
+; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
+; AVX1-NEXT: vmovdqa 16(%rdx), %xmm1
+; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm7 = xmm1[0],zero,xmm1[1],zero
+; AVX1-NEXT: vpblendw {{.*#+}} xmm15 = xmm0[0,1,2,3],xmm7[4],xmm0[5,6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm1[1,1,2,2]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm7[1],xmm0[2,3,4,5],xmm7[6],xmm0[7]
+; AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm10[4],xmm2[4],xmm10[5],xmm2[5],xmm10[6],xmm2[6],xmm10[7],xmm2[7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,6]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm13[0],xmm11[0],xmm13[1],xmm11[1],xmm13[2],xmm11[2],xmm13[3],xmm11[3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,1,1,2,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[0,1,1,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm0
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm7 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,7,6]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[2,1,3,3]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,2,2,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,0,2,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm7, %ymm6
+; AVX1-NEXT: vandnps %ymm0, %ymm12, %ymm0
+; AVX1-NEXT: vandps %ymm6, %ymm12, %ymm6
+; AVX1-NEXT: vorps %ymm0, %ymm6, %ymm0
+; AVX1-NEXT: vpsrlq $48, %xmm8, %xmm6
+; AVX1-NEXT: vpblendw {{.*#+}} xmm6 = xmm0[0,1,2,3],xmm6[4],xmm0[5,6,7]
+; AVX1-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm6 = xmm5[1,0,2,3,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,0,0,0]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm6[1],xmm0[2,3,4,5],xmm6[6],xmm0[7]
+; AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[2,2,2,2,4,5,6,7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,4,6,7]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-NEXT: vpblendw {{.*#+}} xmm14 = xmm14[0,1,2,3],xmm2[4],xmm14[5,6,7]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; AVX1-NEXT: vpshuflw {{.*#+}} xmm4 = xmm2[0,1,3,2,4,5,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,1,1]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,6]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm2
+; AVX1-NEXT: vandnps %ymm3, %ymm12, %ymm3
+; AVX1-NEXT: vandps %ymm2, %ymm12, %ymm2
+; AVX1-NEXT: vorps %ymm3, %ymm2, %ymm3
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm10[0,1,0,1]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm4[4],xmm3[5,6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm3
+; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3,4,5],xmm4[6],xmm3[7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm6[4],xmm0[4],xmm6[5],xmm0[5],xmm6[6],xmm0[6],xmm6[7],xmm0[7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,7,6]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[2,1,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm4, %ymm4
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm6 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [0,1,6,7,4,5,6,7,8,9,4,5,10,11,6,7]
+; AVX1-NEXT: vpshufb %xmm7, %xmm6, %xmm8
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,7,6,6]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm8, %ymm6
+; AVX1-NEXT: vandnps %ymm4, %ymm12, %ymm4
+; AVX1-NEXT: vandps %ymm6, %ymm12, %ymm6
+; AVX1-NEXT: vorps %ymm4, %ymm6, %ymm4
+; AVX1-NEXT: vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vpsrld $16, %xmm0, %xmm6
+; AVX1-NEXT: vpblendw {{.*#+}} xmm6 = xmm4[0,1,2,3],xmm6[4],xmm4[5,6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm4, %xmm4
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm4[0],xmm0[1],xmm4[2,3,4,5],xmm0[6],xmm4[7]
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm5[4],xmm1[4],xmm5[5],xmm1[5],xmm5[6],xmm1[6],xmm5[7],xmm1[7]
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,7,6]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,1,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm4, %ymm1
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm11[4],xmm13[4],xmm11[5],xmm13[5],xmm11[6],xmm13[6],xmm11[7],xmm13[7]
+; AVX1-NEXT: vpshufb %xmm7, %xmm4, %xmm5
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,6]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm5, %ymm4
+; AVX1-NEXT: vandnps %ymm1, %ymm12, %ymm1
+; AVX1-NEXT: vandps %ymm4, %ymm12, %ymm4
+; AVX1-NEXT: vorps %ymm1, %ymm4, %ymm1
+; AVX1-NEXT: vpsrld $16, %xmm9, %xmm4
+; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm1[0,1,2,3],xmm4[4],xmm1[5,6,7]
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX1-NEXT: vpshufhw {{.*#+}} xmm5 = xmm9[0,1,2,3,7,6,6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[2,2,2,2]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3,4,5],xmm5[6],xmm1[7]
+; AVX1-NEXT: vmovdqa %xmm1, 144(%r9)
+; AVX1-NEXT: vmovdqa %xmm4, 128(%r9)
+; AVX1-NEXT: vmovdqa %xmm0, 304(%r9)
+; AVX1-NEXT: vmovdqa %xmm6, 288(%r9)
+; AVX1-NEXT: vmovdqa %xmm3, 16(%r9)
+; AVX1-NEXT: vmovdqa %xmm2, (%r9)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm0, 80(%r9)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm0, 64(%r9)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm0, 112(%r9)
+; AVX1-NEXT: vmovdqa %xmm15, 96(%r9)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm0, 48(%r9)
+; AVX1-NEXT: vmovdqa %xmm14, 32(%r9)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm0, 176(%r9)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm0, 160(%r9)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm0, 272(%r9)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm0, 256(%r9)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm0, 208(%r9)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm0, 192(%r9)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm0, 240(%r9)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm0, 224(%r9)
+; AVX1-NEXT: addq $152, %rsp
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-SLOW-LABEL: vf32:
+; AVX2-SLOW: # %bb.0:
+; AVX2-SLOW-NEXT: subq $72, %rsp
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm12
+; AVX2-SLOW-NEXT: vmovdqa 32(%rsi), %ymm2
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdx), %ymm14
+; AVX2-SLOW-NEXT: vmovdqa (%rdx), %xmm11
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdx), %xmm9
+; AVX2-SLOW-NEXT: vmovdqa (%rcx), %xmm10
+; AVX2-SLOW-NEXT: vmovdqa 32(%rcx), %xmm4
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm10[0],xmm11[0],xmm10[1],xmm11[1],xmm10[2],xmm11[2],xmm10[3],xmm11[3]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm8 = [4,5,2,3,2,3,0,1,10,11,8,9,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm0, %xmm0
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-SLOW-NEXT: vmovdqa (%rsi), %xmm13
+; AVX2-SLOW-NEXT: vmovdqa 32(%rsi), %xmm7
+; AVX2-SLOW-NEXT: vmovdqa (%rdi), %xmm3
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm13[0],xmm3[1],xmm13[1],xmm3[2],xmm13[2],xmm3[3],xmm13[3]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,2,1,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,5,6]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm3[0,1,0,1]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = <255,255,255,255,0,0,0,0,u,u,255,255,255,255,0,0,0,0,u,u,255,255,255,255,0,0,0,0,u,u,255,255>
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm5, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm15 = <6,7,u,u,10,11,u,u,8,9,8,9,u,u,12,13>
+; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm7, %xmm5
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm9[1,2,2,2]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm6 = xmm5[0],xmm6[1],xmm5[2],xmm6[3],xmm5[4,5],xmm6[6],xmm5[7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = mem[2,1,2,3]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm4[3,3,3,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,6,4]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm5[0,1],xmm1[2],xmm5[3],xmm1[4],xmm5[5,6],xmm1[7]
+; AVX2-SLOW-NEXT: vmovdqa (%rcx), %ymm5
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm6[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm6, %ymm1, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 32(%rcx), %ymm6
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3]
+; AVX2-SLOW-NEXT: vmovdqa 32(%r8), %ymm9
+; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm4, %xmm4
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,0,1]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,6]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm0, %ymm4, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm13, %xmm0
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm11[1,2,2,2]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm4[1],xmm0[2],xmm4[3],xmm0[4,5],xmm4[6],xmm0[7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = mem[2,1,2,3]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm7 = xmm10[3,3,3,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,5,5,6,4]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm7[2],xmm4[3],xmm7[4],xmm4[5,6],xmm7[7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,0,1]
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm0, %ymm4, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm12[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm9[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1],ymm4[2],ymm0[3],ymm4[4,5],ymm0[6],ymm4[7,8],ymm0[9],ymm4[10],ymm0[11],ymm4[12,13],ymm0[14],ymm4[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm14[3,2,3,3,7,6,7,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm7 = ymm2[0,1,2,3,5,6,7,7,8,9,10,11,13,14,15,15]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm7 = ymm7[0,2,3,2,4,6,7,6]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1],ymm7[2],ymm4[3],ymm7[4],ymm4[5,6],ymm7[7],ymm4[8,9],ymm7[10],ymm4[11],ymm7[12],ymm4[13,14],ymm7[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,3,2,3]
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm0, %ymm4, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm0 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,u,u,22,23,u,u,24,25,20,21,u,u,24,25>
+; AVX2-SLOW-NEXT: vpshufb %ymm0, %ymm6, %ymm4
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm7 = ymm9[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm7[1],ymm4[2],ymm7[3],ymm4[4,5],ymm7[6],ymm4[7,8],ymm7[9],ymm4[10],ymm7[11],ymm4[12,13],ymm7[14],ymm4[15]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm7 = ymm12[0,1,2,1,4,5,6,5]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm13 = ymm2[3,1,2,2,4,5,6,7,11,9,10,10,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm13 = ymm13[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm13 = ymm13[0,1],ymm7[2],ymm13[3],ymm7[4],ymm13[5,6],ymm7[7],ymm13[8,9],ymm7[10],ymm13[11],ymm7[12],ymm13[13,14],ymm7[15]
+; AVX2-SLOW-NEXT: vmovdqa (%r8), %ymm7
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm13 = ymm13[2,3,2,3]
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm4, %ymm13, %ymm4
+; AVX2-SLOW-NEXT: vpshufb %ymm0, %ymm5, %ymm0
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm13 = ymm7[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm15 = ymm0[0],ymm13[1],ymm0[2],ymm13[3],ymm0[4,5],ymm13[6],ymm0[7,8],ymm13[9],ymm0[10],ymm13[11],ymm0[12,13],ymm13[14],ymm0[15]
+; AVX2-SLOW-NEXT: vmovdqa (%rsi), %ymm13
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm0 = ymm13[3,1,2,2,4,5,6,7,11,9,10,10,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
+; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm1
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm8 = ymm1[0,1,2,1,4,5,6,5]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm8[2],ymm0[3],ymm8[4],ymm0[5,6],ymm8[7],ymm0[8,9],ymm8[10],ymm0[11],ymm8[12],ymm0[13,14],ymm8[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm8 = ymm15[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm8, %ymm0, %ymm15
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm1[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm8 = ymm7[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm8[0],ymm0[1],ymm8[2],ymm0[3],ymm8[4,5],ymm0[6],ymm8[7,8],ymm0[9],ymm8[10],ymm0[11],ymm8[12,13],ymm0[14],ymm8[15]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm8 = ymm13[0,1,2,3,5,6,7,7,8,9,10,11,13,14,15,15]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm8 = ymm8[0,2,3,2,4,6,7,6]
+; AVX2-SLOW-NEXT: vmovdqa (%rdx), %ymm0
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm10 = ymm0[3,2,3,3,7,6,7,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm8 = ymm10[0,1],ymm8[2],ymm10[3],ymm8[4],ymm10[5,6],ymm8[7],ymm10[8,9],ymm8[10],ymm10[11],ymm8[12],ymm10[13,14],ymm8[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm8 = ymm8[2,3,2,3]
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm2, %ymm8, %ymm2
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm8 = <u,u,12,13,u,u,0,1,u,u,u,u,14,15,u,u,u,u,28,29,u,u,16,17,u,u,u,u,30,31,u,u>
+; AVX2-SLOW-NEXT: vpshufb %ymm8, %ymm6, %ymm10
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm11 = ymm14[3,0,3,0,7,4,7,4]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm10 = ymm11[0],ymm10[1],ymm11[2],ymm10[3],ymm11[4,5],ymm10[6],ymm11[7,8],ymm10[9],ymm11[10],ymm10[11],ymm11[12,13],ymm10[14],ymm11[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm11 = ymm12[1,1,2,2]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm9 = ymm9[1,1,2,2]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm9 = ymm11[0],ymm9[1,2],ymm11[3],ymm9[4],ymm11[5],ymm9[6,7],ymm11[8],ymm9[9,10],ymm11[11],ymm9[12],ymm11[13],ymm9[14,15]
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm10, %ymm9, %ymm9
+; AVX2-SLOW-NEXT: vpshufb %ymm8, %ymm5, %ymm8
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm10 = ymm0[3,0,3,0,7,4,7,4]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm8 = ymm10[0],ymm8[1],ymm10[2],ymm8[3],ymm10[4,5],ymm8[6],ymm10[7,8],ymm8[9],ymm10[10],ymm8[11],ymm10[12,13],ymm8[14],ymm10[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,1,2,2]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm7 = ymm7[1,1,2,2]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm7[1,2],ymm1[3],ymm7[4],ymm1[5],ymm7[6,7],ymm1[8],ymm7[9,10],ymm1[11],ymm7[12],ymm1[13],ymm7[14,15]
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm8, %ymm1, %ymm1
+; AVX2-SLOW-NEXT: vpbroadcastq (%r8), %ymm3
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,0,0,255,255]
+; AVX2-SLOW-NEXT: vpblendvb %ymm7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpbroadcastq 40(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vpblendvb %ymm7, (%rsp), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpbroadcastq 32(%r8), %ymm10
+; AVX2-SLOW-NEXT: vpblendvb %ymm7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpbroadcastq 8(%rdi), %ymm11
+; AVX2-SLOW-NEXT: vpblendvb %ymm7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpsrldq {{.*#+}} ymm6 = ymm6[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm6[2,2,2,2]
+; AVX2-SLOW-NEXT: vpblendvb %ymm7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm12 = ymm14[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm12 = ymm12[2,2,2,2]
+; AVX2-SLOW-NEXT: vpblendvb %ymm7, %ymm4, %ymm12, %ymm4
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,2,2]
+; AVX2-SLOW-NEXT: vpblendvb %ymm7, %ymm15, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vpsrldq {{.*#+}} ymm5 = ymm5[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm5[26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm5[2,2,2,2]
+; AVX2-SLOW-NEXT: vpblendvb %ymm7, %ymm2, %ymm5, %ymm2
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm5 = [0,1,0,1,0,1,0,1,14,15,2,3,2,3,14,15,16,17,16,17,16,17,16,17,30,31,18,19,18,19,30,31]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpshufb %ymm5, %ymm12, %ymm12
+; AVX2-SLOW-NEXT: vpblendvb %ymm7, %ymm9, %ymm12, %ymm9
+; AVX2-SLOW-NEXT: vpshufb %ymm5, %ymm13, %ymm5
+; AVX2-SLOW-NEXT: vpblendvb %ymm7, %ymm1, %ymm5, %ymm1
+; AVX2-SLOW-NEXT: vmovdqa %ymm1, 64(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm9, 224(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm2, 128(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm0, 96(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm4, 256(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm6, 288(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm11, 32(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm10, 160(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm8, 192(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm3, (%r9)
+; AVX2-SLOW-NEXT: addq $72, %rsp
+; AVX2-SLOW-NEXT: vzeroupper
+; AVX2-SLOW-NEXT: retq
+;
+; AVX2-FAST-ALL-LABEL: vf32:
+; AVX2-FAST-ALL: # %bb.0:
+; AVX2-FAST-ALL-NEXT: subq $72, %rsp
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rdi), %ymm15
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rsi), %xmm9
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rsi), %xmm0
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rdi), %xmm1
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3]
+; AVX2-FAST-ALL-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,1,2,3,8,9,10,11,4,5,4,5,6,7,12,13]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm3 = ymm1[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rdx), %xmm11
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rdx), %xmm5
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rcx), %xmm10
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rcx), %xmm1
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm10[0],xmm11[0],xmm10[1],xmm11[1],xmm10[2],xmm11[2],xmm10[3],xmm11[3]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} xmm12 = [4,5,2,3,2,3,0,1,10,11,8,9,4,5,6,7]
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm12, %xmm2, %xmm2
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm4 = ymm2[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm8 = <255,255,255,255,0,0,0,0,u,u,255,255,255,255,0,0,0,0,u,u,255,255,255,255,0,0,0,0,u,u,255,255>
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm3, %ymm4, %ymm2
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} xmm13 = <6,7,u,u,10,11,u,u,8,9,8,9,u,u,12,13>
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm13, %xmm0, %xmm3
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} xmm4 = xmm5[1,2,2,2]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} xmm4 = xmm3[0],xmm4[1],xmm3[2],xmm4[3],xmm3[4,5],xmm4[6],xmm3[7]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,u,u,6,7,u,u,10,11,u,u,u,u,8,9>
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm7, %xmm1, %xmm3
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} xmm6 = mem[2,1,2,3]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1],xmm3[2],xmm6[3],xmm3[4],xmm6[5,6],xmm3[7]
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rsi), %ymm14
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm6 = ymm6[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm4, %ymm6, %ymm2
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm2, (%rsp) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rdi), %xmm4
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rdx), %ymm4
+; AVX2-FAST-ALL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,8,9,10,11,4,5,4,5,6,7,12,13]
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rcx), %ymm6
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%r8), %ymm3
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm12, %xmm1, %xmm1
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm13, %xmm9, %xmm0
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} xmm1 = xmm11[1,2,2,2]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4,5],xmm1[6],xmm0[7]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm7, %xmm10, %xmm1
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} xmm5 = mem[2,1,2,3]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} xmm1 = xmm5[0,1],xmm1[2],xmm5[3],xmm1[4],xmm5[5,6],xmm1[7]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm0 = ymm15[2,3,2,3,6,7,6,7]
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm1 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4,5],ymm0[6],ymm1[7,8],ymm0[9],ymm1[10],ymm0[11],ymm1[12,13],ymm0[14],ymm1[15]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,28,29,30,31,30,31,26,27,28,29>
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm5, %ymm14, %ymm1
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm7 = ymm4[3,2,3,3,7,6,7,7]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm1 = ymm7[0,1],ymm1[2],ymm7[3],ymm1[4],ymm7[5,6],ymm1[7],ymm7[8,9],ymm1[10],ymm7[11],ymm1[12],ymm7[13,14],ymm1[15]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm0 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,u,u,22,23,u,u,24,25,20,21,u,u,24,25>
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm0, %ymm6, %ymm1
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm7 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm7[1],ymm1[2],ymm7[3],ymm1[4,5],ymm7[6],ymm1[7,8],ymm7[9],ymm1[10],ymm7[11],ymm1[12,13],ymm7[14],ymm1[15]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,22,23,18,19,u,u,20,21,u,u,24,25,24,25,u,u>
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm7, %ymm14, %ymm11
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm12 = ymm15[0,1,2,1,4,5,6,5]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm11 = ymm11[0,1],ymm12[2],ymm11[3],ymm12[4],ymm11[5,6],ymm12[7],ymm11[8,9],ymm12[10],ymm11[11],ymm12[12],ymm11[13,14],ymm12[15]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm11 = ymm11[2,3,2,3]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm1, %ymm11, %ymm1
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rcx), %ymm11
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm0, %ymm11, %ymm1
+; AVX2-FAST-ALL-NEXT: vmovdqa (%r8), %ymm2
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm13 = ymm2[0,1,0,1,4,5,4,5]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm13[1],ymm1[2],ymm13[3],ymm1[4,5],ymm13[6],ymm1[7,8],ymm13[9],ymm1[10],ymm13[11],ymm1[12,13],ymm13[14],ymm1[15]
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rsi), %ymm13
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm7, %ymm13, %ymm7
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rdi), %ymm0
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm9 = ymm0[0,1,2,1,4,5,6,5]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0,1],ymm9[2],ymm7[3],ymm9[4],ymm7[5,6],ymm9[7],ymm7[8,9],ymm9[10],ymm7[11],ymm9[12],ymm7[13,14],ymm9[15]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm7 = ymm7[2,3,2,3]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm1, %ymm7, %ymm1
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm7 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm9 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm7 = ymm9[0],ymm7[1],ymm9[2],ymm7[3],ymm9[4,5],ymm7[6],ymm9[7,8],ymm7[9],ymm9[10],ymm7[11],ymm9[12,13],ymm7[14],ymm9[15]
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm5, %ymm13, %ymm5
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rdx), %ymm9
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm10 = ymm9[3,2,3,3,7,6,7,7]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm5 = ymm10[0,1],ymm5[2],ymm10[3],ymm5[4],ymm10[5,6],ymm5[7],ymm10[8,9],ymm5[10],ymm10[11],ymm5[12],ymm10[13,14],ymm5[15]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm7 = ymm7[2,3,2,3]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm5 = ymm5[2,3,2,3]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm7, %ymm5, %ymm5
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,12,13,u,u,0,1,u,u,u,u,14,15,u,u,u,u,28,29,u,u,16,17,u,u,u,u,30,31,u,u>
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm7, %ymm6, %ymm10
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm12 = ymm4[3,0,3,0,7,4,7,4]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm10 = ymm12[0],ymm10[1],ymm12[2],ymm10[3],ymm12[4,5],ymm10[6],ymm12[7,8],ymm10[9],ymm12[10],ymm10[11],ymm12[12,13],ymm10[14],ymm12[15]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm12 = ymm15[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm3 = ymm3[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm3 = ymm12[0],ymm3[1,2],ymm12[3],ymm3[4],ymm12[5],ymm3[6,7],ymm12[8],ymm3[9,10],ymm12[11],ymm3[12],ymm12[13],ymm3[14,15]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm10, %ymm3, %ymm3
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm7, %ymm11, %ymm7
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm10 = ymm9[3,0,3,0,7,4,7,4]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm7 = ymm10[0],ymm7[1],ymm10[2],ymm7[3],ymm10[4,5],ymm7[6],ymm10[7,8],ymm7[9],ymm10[10],ymm7[11],ymm10[12,13],ymm7[14],ymm10[15]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm2 = ymm2[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1,2],ymm0[3],ymm2[4],ymm0[5],ymm2[6,7],ymm0[8],ymm2[9,10],ymm0[11],ymm2[12],ymm0[13],ymm2[14,15]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm8, %ymm7, %ymm0, %ymm0
+; AVX2-FAST-ALL-NEXT: vpbroadcastq (%r8), %ymm2
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,0,0,255,255]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-ALL-NEXT: vpbroadcastq 40(%rdi), %ymm8
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm7, (%rsp), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-ALL-NEXT: vpbroadcastq 32(%r8), %ymm10
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-ALL-NEXT: vpbroadcastq 8(%rdi), %ymm12
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-ALL-NEXT: vpsrldq {{.*#+}} ymm6 = ymm6[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm6 = ymm6[2,2,2,2]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-ALL-NEXT: vpbroadcastq {{.*#+}} ymm15 = [25769803781,25769803781,25769803781,25769803781]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm4, %ymm15, %ymm4
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-ALL-NEXT: vpermd %ymm9, %ymm15, %ymm9
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm7, %ymm1, %ymm9, %ymm1
+; AVX2-FAST-ALL-NEXT: vpsrldq {{.*#+}} ymm9 = ymm11[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm11[26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm9 = ymm9[2,2,2,2]
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm7, %ymm5, %ymm9, %ymm5
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm9 = [0,1,0,1,0,1,0,1,14,15,2,3,2,3,14,15,16,17,16,17,16,17,16,17,30,31,18,19,18,19,30,31]
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm9, %ymm14, %ymm11
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm7, %ymm3, %ymm11, %ymm3
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm9, %ymm13, %ymm9
+; AVX2-FAST-ALL-NEXT: vpblendvb %ymm7, %ymm0, %ymm9, %ymm0
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm0, 64(%r9)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm3, 224(%r9)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm5, 128(%r9)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm1, 96(%r9)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm4, 256(%r9)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm6, 288(%r9)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm12, 32(%r9)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm10, 160(%r9)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm8, 192(%r9)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm2, (%r9)
+; AVX2-FAST-ALL-NEXT: addq $72, %rsp
+; AVX2-FAST-ALL-NEXT: vzeroupper
+; AVX2-FAST-ALL-NEXT: retq
+;
+; AVX2-FAST-PERLANE-LABEL: vf32:
+; AVX2-FAST-PERLANE: # %bb.0:
+; AVX2-FAST-PERLANE-NEXT: subq $72, %rsp
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rsi), %xmm9
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rsi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm8 = [0,1,2,3,8,9,10,11,4,5,4,5,6,7,12,13]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdx), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdx), %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rcx), %xmm11
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rcx), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm11[0],xmm12[0],xmm11[1],xmm12[1],xmm11[2],xmm12[2],xmm11[3],xmm12[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm13 = [4,5,2,3,2,3,0,1,10,11,8,9,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm3, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm3[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm3 = <255,255,255,255,0,0,0,0,u,u,255,255,255,255,0,0,0,0,u,u,255,255,255,255,0,0,0,0,u,u,255,255>
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm1, %ymm4, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm14 = <6,7,u,u,10,11,u,u,8,9,8,9,u,u,12,13>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm0, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm7 = xmm5[1,2,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0],xmm7[1],xmm4[2],xmm7[3],xmm4[4,5],xmm7[6],xmm4[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,u,u,6,7,u,u,10,11,u,u,u,u,8,9>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm2, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm6 = mem[2,1,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1],xmm7[2],xmm6[3],xmm7[4],xmm6[5,6],xmm7[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rsi), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm6 = ymm6[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm4, %ymm6, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdx), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rcx), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%r8), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm2, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm0, %ymm2, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm9, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm2 = xmm12[1,2,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4,5],xmm2[6],xmm0[7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm11, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,1,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2],xmm2[3],xmm1[4],xmm2[5,6],xmm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm0 = ymm15[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm8[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4,5],ymm0[6],ymm1[7,8],ymm0[9],ymm1[10],ymm0[11],ymm1[12,13],ymm0[14],ymm1[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,28,29,30,31,30,31,26,27,28,29>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm10, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm7[3,2,3,3,7,6,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3],ymm1[4],ymm2[5,6],ymm1[7],ymm2[8,9],ymm1[10],ymm2[11],ymm1[12],ymm2[13,14],ymm1[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm0 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,u,u,22,23,u,u,24,25,20,21,u,u,24,25>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm0, %ymm6, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2],ymm2[3],ymm1[4,5],ymm2[6],ymm1[7,8],ymm2[9],ymm1[10],ymm2[11],ymm1[12,13],ymm2[14],ymm1[15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm4 = <u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,22,23,18,19,u,u,20,21,u,u,24,25,24,25,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm4, %ymm10, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm12 = ymm15[0,1,2,1,4,5,6,5]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm11 = ymm11[0,1],ymm12[2],ymm11[3],ymm12[4],ymm11[5,6],ymm12[7],ymm11[8,9],ymm12[10],ymm11[11],ymm12[12],ymm11[13,14],ymm12[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm11 = ymm11[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm1, %ymm11, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rcx), %ymm11
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm0, %ymm11, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%r8), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm13 = ymm2[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0],ymm13[1],ymm0[2],ymm13[3],ymm0[4,5],ymm13[6],ymm0[7,8],ymm13[9],ymm0[10],ymm13[11],ymm0[12,13],ymm13[14],ymm0[15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rsi), %ymm13
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm4, %ymm13, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm14 = ymm0[0,1,2,1,4,5,6,5]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1],ymm14[2],ymm4[3],ymm14[4],ymm4[5,6],ymm14[7],ymm4[8,9],ymm14[10],ymm4[11],ymm14[12],ymm4[13,14],ymm14[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm1, %ymm4, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm4 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm14 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm14[0],ymm4[1],ymm14[2],ymm4[3],ymm14[4,5],ymm4[6],ymm14[7,8],ymm4[9],ymm14[10],ymm4[11],ymm14[12,13],ymm4[14],ymm14[15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm13, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdx), %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm9 = ymm14[3,2,3,3,7,6,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm9[0,1],ymm5[2],ymm9[3],ymm5[4],ymm9[5,6],ymm5[7],ymm9[8,9],ymm5[10],ymm9[11],ymm5[12],ymm9[13,14],ymm5[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm5[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm4, %ymm5, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,12,13,u,u,0,1,u,u,u,u,14,15,u,u,u,u,28,29,u,u,16,17,u,u,u,u,30,31,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm6, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm12 = ymm7[3,0,3,0,7,4,7,4]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm9 = ymm12[0],ymm9[1],ymm12[2],ymm9[3],ymm12[4,5],ymm9[6],ymm12[7,8],ymm9[9],ymm12[10],ymm9[11],ymm12[12,13],ymm9[14],ymm12[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm12 = ymm15[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm8 = ymm8[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm8 = ymm12[0],ymm8[1,2],ymm12[3],ymm8[4],ymm12[5],ymm8[6,7],ymm12[8],ymm8[9,10],ymm12[11],ymm8[12],ymm12[13],ymm8[14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm9, %ymm8, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm11, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm9 = ymm14[3,0,3,0,7,4,7,4]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm9[0],ymm5[1],ymm9[2],ymm5[3],ymm9[4,5],ymm5[6],ymm9[7,8],ymm5[9],ymm9[10],ymm5[11],ymm9[12,13],ymm5[14],ymm9[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1,2],ymm0[3],ymm2[4],ymm0[5],ymm2[6,7],ymm0[8],ymm2[9,10],ymm0[11],ymm2[12],ymm0[13],ymm2[14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm5, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastq (%r8), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,0,0,255,255]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 40(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, (%rsp), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 32(%r8), %ymm9
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 8(%rdi), %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpsrldq {{.*#+}} ymm6 = ymm6[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm6[26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm6 = ymm6[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; 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: vpblendvb %ymm3, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm14 = ymm14[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm14 = ymm14[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm1, %ymm14, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpsrldq {{.*#+}} ymm11 = ymm11[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm11[26,27,28,29,30,31],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm11 = ymm11[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm4, %ymm11, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm11 = [0,1,0,1,0,1,0,1,14,15,2,3,2,3,14,15,16,17,16,17,16,17,16,17,30,31,18,19,18,19,30,31]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm11, %ymm10, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm8, %ymm10, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm11, %ymm13, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm0, %ymm10, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, 64(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm8, 224(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, 128(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, 96(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, 256(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, 288(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm12, 32(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, 160(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, 192(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, (%r9)
+; AVX2-FAST-PERLANE-NEXT: addq $72, %rsp
+; AVX2-FAST-PERLANE-NEXT: vzeroupper
+; AVX2-FAST-PERLANE-NEXT: retq
+;
+; AVX512-LABEL: vf32:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm0
+; AVX512-NEXT: vmovdqu64 (%rsi), %zmm1
+; AVX512-NEXT: vmovdqu64 (%rdx), %zmm2
+; AVX512-NEXT: vmovdqu64 (%rcx), %zmm3
+; AVX512-NEXT: vmovdqu64 (%r8), %zmm4
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm5 = <u,u,0,32,u,u,u,1,33,u,u,u,2,34,u,u,u,3,35,u,u,u,4,36,u,u,u,5,37,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm3, %zmm2, %zmm5
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm6 = <0,32,u,u,u,1,33,u,u,u,2,34,u,u,u,3,35,u,u,u,4,36,u,u,u,5,37,u,u,u,6,38>
+; AVX512-NEXT: vpermi2w %zmm1, %zmm0, %zmm6
+; AVX512-NEXT: movl $415641996, %eax # imm = 0x18C6318C
+; AVX512-NEXT: kmovd %eax, %k1
+; AVX512-NEXT: vmovdqu16 %zmm5, %zmm6 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm5 = [0,1,2,3,32,5,6,7,8,33,10,11,12,13,34,15,16,17,18,35,20,21,22,23,36,25,26,27,28,37,30,31]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm6, %zmm5
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm6 = <u,u,6,39,u,u,u,7,40,u,u,u,8,41,u,u,u,9,42,u,u,u,10,43,u,u,u,11,44,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm0, %zmm4, %zmm6
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm7 = <6,38,u,u,u,7,39,u,u,u,8,40,u,u,u,9,41,u,u,u,10,42,u,u,u,11,43,u,u,u,12,44>
+; AVX512-NEXT: vpermi2w %zmm3, %zmm2, %zmm7
+; AVX512-NEXT: vmovdqu16 %zmm6, %zmm7 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm6 = [0,1,2,3,39,5,6,7,8,40,10,11,12,13,41,15,16,17,18,42,20,21,22,23,43,25,26,27,28,44,30,31]
+; AVX512-NEXT: vpermi2w %zmm1, %zmm7, %zmm6
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm7 = <u,u,13,45,u,u,u,14,46,u,u,u,15,47,u,u,u,16,48,u,u,u,17,49,u,u,u,18,50,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm2, %zmm1, %zmm7
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm8 = <12,45,u,u,u,13,46,u,u,u,14,47,u,u,u,15,48,u,u,u,16,49,u,u,u,17,50,u,u,u,18,51>
+; AVX512-NEXT: vpermi2w %zmm0, %zmm4, %zmm8
+; AVX512-NEXT: vmovdqu16 %zmm7, %zmm8 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm7 = [0,1,2,3,45,5,6,7,8,46,10,11,12,13,47,15,16,17,18,48,20,21,22,23,49,25,26,27,28,50,30,31]
+; AVX512-NEXT: vpermi2w %zmm3, %zmm8, %zmm7
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm8 = <u,u,19,51,u,u,u,20,52,u,u,u,21,53,u,u,u,22,54,u,u,u,23,55,u,u,u,24,56,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm4, %zmm3, %zmm8
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm9 = <19,51,u,u,u,20,52,u,u,u,21,53,u,u,u,22,54,u,u,u,23,55,u,u,u,24,56,u,u,u,25,57>
+; AVX512-NEXT: vpermi2w %zmm2, %zmm1, %zmm9
+; AVX512-NEXT: vmovdqu16 %zmm8, %zmm9 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm8 = [0,1,2,3,52,5,6,7,8,53,10,11,12,13,54,15,16,17,18,55,20,21,22,23,56,25,26,27,28,57,30,31]
+; AVX512-NEXT: vpermi2w %zmm0, %zmm9, %zmm8
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm9 = <u,u,26,58,u,u,u,27,59,u,u,u,28,60,u,u,u,29,61,u,u,u,30,62,u,u,u,31,63,u,u,u>
+; AVX512-NEXT: vpermi2w %zmm1, %zmm0, %zmm9
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm0 = <25,57,u,u,u,26,58,u,u,u,27,59,u,u,u,28,60,u,u,u,29,61,u,u,u,30,62,u,u,u,31,63>
+; AVX512-NEXT: vpermi2w %zmm4, %zmm3, %zmm0
+; AVX512-NEXT: vmovdqu16 %zmm9, %zmm0 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,1,2,3,58,5,6,7,8,59,10,11,12,13,60,15,16,17,18,61,20,21,22,23,62,25,26,27,28,63,30,31]
+; AVX512-NEXT: vpermi2w %zmm2, %zmm0, %zmm1
+; AVX512-NEXT: vmovdqu64 %zmm1, 256(%r9)
+; AVX512-NEXT: vmovdqu64 %zmm8, 192(%r9)
+; AVX512-NEXT: vmovdqu64 %zmm7, 128(%r9)
+; AVX512-NEXT: vmovdqu64 %zmm6, 64(%r9)
+; AVX512-NEXT: vmovdqu64 %zmm5, (%r9)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %in.vec0 = load <32 x i16>, <32 x i16>* %in.vecptr0, align 32
+ %in.vec1 = load <32 x i16>, <32 x i16>* %in.vecptr1, align 32
+ %in.vec2 = load <32 x i16>, <32 x i16>* %in.vecptr2, align 32
+ %in.vec3 = load <32 x i16>, <32 x i16>* %in.vecptr3, align 32
+ %in.vec4 = load <32 x i16>, <32 x i16>* %in.vecptr4, align 32
+
+ %concat01 = shufflevector <32 x i16> %in.vec0, <32 x i16> %in.vec1, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+ %concat23 = shufflevector <32 x i16> %in.vec2, <32 x i16> %in.vec3, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+ %concat0123 = shufflevector <64 x i16> %concat01, <64 x i16> %concat23, <128 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64, i32 65, i32 66, i32 67, i32 68, i32 69, i32 70, i32 71, i32 72, i32 73, i32 74, i32 75, i32 76, i32 77, i32 78, i32 79, i32 80, i32 81, i32 82, i32 83, i32 84, i32 85, i32 86, i32 87, i32 88, i32 89, i32 90, i32 91, i32 92, i32 93, i32 94, i32 95, i32 96, i32 97, i32 98, i32 99, i32 100, i32 101, i32 102, i32 103, i32 104, i32 105, i32 106, i32 107, i32 108, i32 109, i32 110, i32 111, i32 112, i32 113, i32 114, i32 115, i32 116, i32 117, i32 118, i32 119, i32 120, i32 121, i32 122, i32 123, i32 124, i32 125, i32 126, i32 127>
+ %concat4uuu = shufflevector <32 x i16> %in.vec4, <32 x i16> poison, <128 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+ %concat01234 = shufflevector <128 x i16> %concat0123, <128 x i16> %concat4uuu, <160 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64, i32 65, i32 66, i32 67, i32 68, i32 69, i32 70, i32 71, i32 72, i32 73, i32 74, i32 75, i32 76, i32 77, i32 78, i32 79, i32 80, i32 81, i32 82, i32 83, i32 84, i32 85, i32 86, i32 87, i32 88, i32 89, i32 90, i32 91, i32 92, i32 93, i32 94, i32 95, i32 96, i32 97, i32 98, i32 99, i32 100, i32 101, i32 102, i32 103, i32 104, i32 105, i32 106, i32 107, i32 108, i32 109, i32 110, i32 111, i32 112, i32 113, i32 114, i32 115, i32 116, i32 117, i32 118, i32 119, i32 120, i32 121, i32 122, i32 123, i32 124, i32 125, i32 126, i32 127, i32 128, i32 129, i32 130, i32 131, i32 132, i32 133, i32 134, i32 135, i32 136, i32 137, i32 138, i32 139, i32 140, i32 141, i32 142, i32 143, i32 144, i32 145, i32 146, i32 147, i32 148, i32 149, i32 150, i32 151, i32 152, i32 153, i32 154, i32 155, i32 156, i32 157, i32 158, i32 159>
+ %interleaved.vec = shufflevector <160 x i16> %concat01234, <160 x i16> poison, <160 x i32> <i32 0, i32 32, i32 64, i32 96, i32 128, i32 1, i32 33, i32 65, i32 97, i32 129, i32 2, i32 34, i32 66, i32 98, i32 130, i32 3, i32 35, i32 67, i32 99, i32 131, i32 4, i32 36, i32 68, i32 100, i32 132, i32 5, i32 37, i32 69, i32 101, i32 133, i32 6, i32 38, i32 70, i32 102, i32 134, i32 7, i32 39, i32 71, i32 103, i32 135, i32 8, i32 40, i32 72, i32 104, i32 136, i32 9, i32 41, i32 73, i32 105, i32 137, i32 10, i32 42, i32 74, i32 106, i32 138, i32 11, i32 43, i32 75, i32 107, i32 139, i32 12, i32 44, i32 76, i32 108, i32 140, i32 13, i32 45, i32 77, i32 109, i32 141, i32 14, i32 46, i32 78, i32 110, i32 142, i32 15, i32 47, i32 79, i32 111, i32 143, i32 16, i32 48, i32 80, i32 112, i32 144, i32 17, i32 49, i32 81, i32 113, i32 145, i32 18, i32 50, i32 82, i32 114, i32 146, i32 19, i32 51, i32 83, i32 115, i32 147, i32 20, i32 52, i32 84, i32 116, i32 148, i32 21, i32 53, i32 85, i32 117, i32 149, i32 22, i32 54, i32 86, i32 118, i32 150, i32 23, i32 55, i32 87, i32 119, i32 151, i32 24, i32 56, i32 88, i32 120, i32 152, i32 25, i32 57, i32 89, i32 121, i32 153, i32 26, i32 58, i32 90, i32 122, i32 154, i32 27, i32 59, i32 91, i32 123, i32 155, i32 28, i32 60, i32 92, i32 124, i32 156, i32 29, i32 61, i32 93, i32 125, i32 157, i32 30, i32 62, i32 94, i32 126, i32 158, i32 31, i32 63, i32 95, i32 127, i32 159>
+
+ store <160 x i16> %interleaved.vec, <160 x i16>* %out.vec, align 32
+
+ ret void
+}
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-6.ll
index 8b64ab36c8b63..30efc4dee308e 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-6.ll
@@ -1143,3 +1143,1165 @@ define void @vf16(<16 x i16>* %in.vecptr0, <16 x i16>* %in.vecptr1, <16 x i16>*
ret void
}
+
+define void @vf32(<32 x i16>* %in.vecptr0, <32 x i16>* %in.vecptr1, <32 x i16>* %in.vecptr2, <32 x i16>* %in.vecptr3, <32 x i16>* %in.vecptr4, <32 x i16>* %in.vecptr5, <192 x i16>* %out.vec) nounwind {
+; SSE-LABEL: vf32:
+; SSE: # %bb.0:
+; SSE-NEXT: subq $104, %rsp
+; SSE-NEXT: movdqa (%rdi), %xmm9
+; SSE-NEXT: movdqa 16(%rdi), %xmm12
+; SSE-NEXT: movdqa (%rsi), %xmm10
+; SSE-NEXT: movdqa 16(%rsi), %xmm8
+; SSE-NEXT: movdqa (%rdx), %xmm1
+; SSE-NEXT: movdqa 16(%rdx), %xmm6
+; SSE-NEXT: movdqa (%rcx), %xmm13
+; SSE-NEXT: movdqa 16(%rcx), %xmm11
+; SSE-NEXT: movdqa (%r8), %xmm5
+; SSE-NEXT: movdqa (%r9), %xmm14
+; SSE-NEXT: movdqa %xmm9, %xmm4
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm10[0],xmm4[1],xmm10[1],xmm4[2],xmm10[2],xmm4[3],xmm10[3]
+; SSE-NEXT: movdqa %xmm5, %xmm3
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm14[0],xmm3[1],xmm14[1],xmm3[2],xmm14[2],xmm3[3],xmm14[3]
+; SSE-NEXT: movdqa %xmm3, %xmm7
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[0,1],xmm4[1,3]
+; SSE-NEXT: movdqa %xmm1, %xmm2
+; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm13[0],xmm2[1],xmm13[1],xmm2[2],xmm13[2],xmm2[3],xmm13[3]
+; SSE-NEXT: movdqa %xmm2, %xmm0
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[0,2]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm4, %xmm7
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm7 = xmm7[1],xmm2[1]
+; SSE-NEXT: movdqa %xmm3, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1],xmm2[1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[0,2]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 16(%r8), %xmm15
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[3,1],xmm3[2,3]
+; SSE-NEXT: movdqa 16(%r9), %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm2[0,2]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm14[4],xmm5[5],xmm14[5],xmm5[6],xmm14[6],xmm5[7],xmm14[7]
+; SSE-NEXT: movdqa %xmm5, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm9[1,3]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
+; SSE-NEXT: movdqa %xmm1, %xmm2
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm9[0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm9, %xmm0
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT: movdqa %xmm5, %xmm2
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,1],xmm1[1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,3],xmm5[3,3]
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[3,1],xmm5[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[2,0],xmm1[0,2]
+; SSE-NEXT: movaps %xmm9, (%rsp) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm12, %xmm13
+; SSE-NEXT: punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm8[0],xmm13[1],xmm8[1],xmm13[2],xmm8[2],xmm13[3],xmm8[3]
+; SSE-NEXT: movdqa %xmm15, %xmm0
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm13[1,3]
+; SSE-NEXT: movdqa %xmm6, %xmm2
+; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1],xmm2[2],xmm11[2],xmm2[3],xmm11[3]
+; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm13[0]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm1[0,2]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm13, %xmm1
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; SSE-NEXT: movdqa %xmm0, %xmm4
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,1],xmm2[1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm1[0,2]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[3,3],xmm0[3,3]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[2,0],xmm2[0,2]
+; SSE-NEXT: movdqa 32(%rdi), %xmm10
+; SSE-NEXT: punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm8[4],xmm12[5],xmm8[5],xmm12[6],xmm8[6],xmm12[7],xmm8[7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm3[4],xmm15[5],xmm3[5],xmm15[6],xmm3[6],xmm15[7],xmm3[7]
+; SSE-NEXT: movdqa %xmm15, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm12[1,3]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm11[4],xmm6[5],xmm11[5],xmm6[6],xmm11[6],xmm6[7],xmm11[7]
+; SSE-NEXT: movdqa %xmm6, %xmm1
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm12[0]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[0,2]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm6[1]
+; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,1],xmm6[1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[0,2]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 32(%rsi), %xmm8
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[3,3],xmm15[3,3]
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[3,1],xmm15[2,3]
+; SSE-NEXT: movdqa 32(%r8), %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[2,0],xmm6[0,2]
+; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm10, %xmm9
+; SSE-NEXT: punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
+; SSE-NEXT: movdqa 32(%r9), %xmm3
+; SSE-NEXT: movdqa %xmm0, %xmm4
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; SSE-NEXT: movdqa %xmm4, %xmm5
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm9[1,3]
+; SSE-NEXT: movdqa 32(%rdx), %xmm7
+; SSE-NEXT: movdqa 32(%rcx), %xmm6
+; SSE-NEXT: movdqa %xmm7, %xmm1
+; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; SSE-NEXT: movdqa %xmm1, %xmm2
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm9[0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm5[0,2]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm9, %xmm5
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm5 = xmm5[1],xmm1[1]
+; SSE-NEXT: movdqa %xmm4, %xmm2
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,1],xmm1[1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm5[0,2]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,3],xmm4[3,3]
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[3,1],xmm4[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[2,0],xmm1[0,2]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm8[4],xmm10[5],xmm8[5],xmm10[6],xmm8[6],xmm10[7],xmm8[7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]
+; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm10[1,3]
+; SSE-NEXT: movdqa %xmm7, %xmm12
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm12 = xmm12[0],xmm10[0]
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[2,0],xmm2[0,2]
+; SSE-NEXT: movdqa %xmm10, %xmm2
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm7[1]
+; SSE-NEXT: movdqa %xmm0, %xmm11
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[1,1],xmm7[1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[2,0],xmm2[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[3,3],xmm0[3,3]
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[2,0],xmm7[0,2]
+; SSE-NEXT: movdqa 48(%rdi), %xmm0
+; SSE-NEXT: movdqa 48(%rsi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE-NEXT: movdqa 48(%r8), %xmm7
+; SSE-NEXT: movdqa 48(%r9), %xmm15
+; SSE-NEXT: movdqa %xmm7, %xmm4
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm15[0],xmm4[1],xmm15[1],xmm4[2],xmm15[2],xmm4[3],xmm15[3]
+; SSE-NEXT: movdqa %xmm4, %xmm5
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm2[1,3]
+; SSE-NEXT: movdqa 48(%rdx), %xmm1
+; SSE-NEXT: movdqa 48(%rcx), %xmm14
+; SSE-NEXT: movdqa %xmm1, %xmm6
+; SSE-NEXT: punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm14[0],xmm6[1],xmm14[1],xmm6[2],xmm14[2],xmm6[3],xmm14[3]
+; SSE-NEXT: movdqa %xmm6, %xmm8
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm2[0]
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[2,0],xmm5[0,2]
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm6[1]
+; SSE-NEXT: movdqa %xmm4, %xmm5
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[1,1],xmm6[1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm3[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[3,3],xmm4[3,3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[3,1],xmm4[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm6[0,2]
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm15[4],xmm7[5],xmm15[5],xmm7[6],xmm15[6],xmm7[7],xmm15[7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm14[4],xmm1[5],xmm14[5],xmm1[6],xmm14[6],xmm1[7],xmm14[7]
+; SSE-NEXT: movdqa %xmm7, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[1,3]
+; SSE-NEXT: movdqa %xmm1, %xmm4
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm0[0]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm3[0,2]
+; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm1[1]
+; SSE-NEXT: movdqa %xmm7, %xmm6
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[1,1],xmm1[1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[2,0],xmm3[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,3],xmm7[3,3]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1],xmm7[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[0,2]
+; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT: movaps %xmm0, 368(%rax)
+; SSE-NEXT: movaps %xmm6, 352(%rax)
+; SSE-NEXT: movaps %xmm4, 336(%rax)
+; SSE-NEXT: movaps %xmm2, 320(%rax)
+; SSE-NEXT: movaps %xmm5, 304(%rax)
+; SSE-NEXT: movaps %xmm8, 288(%rax)
+; SSE-NEXT: movaps %xmm10, 272(%rax)
+; SSE-NEXT: movaps %xmm11, 256(%rax)
+; SSE-NEXT: movaps %xmm12, 240(%rax)
+; SSE-NEXT: movaps %xmm9, 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 %xmm13, 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 (%rsp), %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: addq $104, %rsp
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: vf32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: subq $88, %rsp
+; AVX1-NEXT: vmovdqa 16(%rcx), %xmm0
+; AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa 16(%rdx), %xmm1
+; AVX1-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX1-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm1[0,0,1,1]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,1,2,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: vmovdqa 16(%rsi), %xmm3
+; AVX1-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa 32(%rsi), %xmm1
+; AVX1-NEXT: vmovdqa 16(%rdi), %xmm4
+; AVX1-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa 32(%rdi), %xmm2
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm10[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm10, %ymm3, %ymm3
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3],ymm0[4],ymm3[5,6],ymm0[7]
+; AVX1-NEXT: vmovdqa 16(%r9), %xmm5
+; AVX1-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa 32(%r9), %xmm3
+; AVX1-NEXT: vmovdqa 16(%r8), %xmm6
+; AVX1-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovdqa 32(%r8), %xmm4
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm14 = xmm9[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm9, %ymm14, %ymm14
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm14[2],ymm0[3,4],ymm14[5],ymm0[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm14 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm14[2,3,2,3]
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm15 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm15[2,3,2,3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm2[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm0[1],ymm3[2,3],ymm0[4],ymm3[5,6],ymm0[7]
+; AVX1-NEXT: vmovdqa 32(%rcx), %xmm4
+; AVX1-NEXT: vmovdqa 32(%rdx), %xmm0
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm8[2,2,3,3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm13 = xmm0[0,0,1,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm13, %ymm4, %ymm4
+; AVX1-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7]
+; AVX1-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[2,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[1,1,2,2]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3],ymm2[4],ymm0[5,6],ymm2[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovdqa 48(%rcx), %xmm13
+; AVX1-NEXT: vmovdqa 48(%rdx), %xmm7
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm7[4],xmm13[4],xmm7[5],xmm13[5],xmm7[6],xmm13[6],xmm7[7],xmm13[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm6[1,1,2,2]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm6[2,2,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm3
+; AVX1-NEXT: vmovdqa 48(%r9), %xmm5
+; AVX1-NEXT: vmovdqa 48(%r8), %xmm4
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm12 = xmm2[2,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm12, %ymm2, %ymm12
+; AVX1-NEXT: vblendps {{.*#+}} ymm12 = ymm3[0],ymm12[1],ymm3[2,3],ymm12[4],ymm3[5,6],ymm12[7]
+; AVX1-NEXT: vmovdqa 48(%rsi), %xmm3
+; AVX1-NEXT: vmovdqa 48(%rdi), %xmm1
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm11 = xmm0[2,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm11
+; AVX1-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm11[2],ymm12[3,4],ymm11[5],ymm12[6,7]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm3[2,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm1
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm2
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6],ymm1[7]
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm7[0],xmm13[0],xmm7[1],xmm13[1],xmm7[2],xmm13[2],xmm7[3],xmm13[3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm6[0,0,1,1]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm2[2,2,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm5, %ymm4
+; AVX1-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm8[0,0,1,1]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm8[1,1,2,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm14[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm14, %ymm4, %ymm4
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0],ymm1[1],ymm4[2,3],ymm1[4],ymm4[5,6],ymm1[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm15[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm15, %ymm4, %ymm4
+; AVX1-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm2[0,0,1,1]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,2,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm3[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6],ymm1[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
+; AVX1-NEXT: vblendps {{.*#+}} ymm15 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
+; AVX1-NEXT: vmovdqa (%rsi), %xmm0
+; AVX1-NEXT: vmovdqa (%rdi), %xmm1
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm8[2,3,2,3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm3[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: vmovdqa (%r9), %xmm4
+; AVX1-NEXT: vmovdqa (%r8), %xmm5
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm5[4],xmm4[4],xmm5[5],xmm4[5],xmm5[6],xmm4[6],xmm5[7],xmm4[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm1[2,3,2,3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm5[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm4, %ymm4
+; AVX1-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm0[1],ymm4[2,3],ymm0[4],ymm4[5,6],ymm0[7]
+; AVX1-NEXT: vmovdqa (%rcx), %xmm6
+; AVX1-NEXT: vmovdqa (%rdx), %xmm7
+; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm0[2,2,3,3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm6[0,0,1,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm7, %ymm2
+; AVX1-NEXT: vblendps {{.*#+}} ymm11 = ymm4[0,1],ymm2[2],ymm4[3,4],ymm2[5],ymm4[6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm5[2,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm5, %ymm2
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm6[1,1,2,2]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[2,2,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm5, %ymm5
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1],ymm5[2,3],ymm2[4],ymm5[5,6],ymm2[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm3[2,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm3
+; AVX1-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm5 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm5 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm5[1,1,2,2]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm5[2,2,3,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
+; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm2[2,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm2, %ymm7
+; AVX1-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2,3],ymm7[4],ymm6[5,6],ymm7[7]
+; AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm7 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm7 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm7[2,3,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm7, %ymm4
+; AVX1-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1],ymm4[2],ymm6[3,4],ymm4[5],ymm6[6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm10[2,3,2,3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
+; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm9[2,3,2,3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm7, %ymm2
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm6[1],ymm2[2,3],ymm6[4],ymm2[5,6],ymm6[7]
+; AVX1-NEXT: vpermilps $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-NEXT: # xmm6 = mem[2,2,3,3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,0,1,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm6, %ymm5
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm5[2],ymm2[3,4],ymm5[5],ymm2[6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm0[0,0,1,1]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm5, %ymm0
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm8[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm8, %ymm5, %ymm5
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0],ymm0[1],ymm5[2,3],ymm0[4],ymm5[5,6],ymm0[7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm1[0,1,0,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm5, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX1-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX1-NEXT: vmovaps %ymm0, (%rax)
+; AVX1-NEXT: vmovaps %ymm2, 128(%rax)
+; AVX1-NEXT: vmovaps %ymm4, 160(%rax)
+; AVX1-NEXT: vmovaps %ymm3, 64(%rax)
+; AVX1-NEXT: vmovaps %ymm11, 32(%rax)
+; AVX1-NEXT: vmovaps %ymm15, 288(%rax)
+; AVX1-NEXT: vmovaps %ymm14, 192(%rax)
+; AVX1-NEXT: vmovaps %ymm13, 320(%rax)
+; AVX1-NEXT: vmovaps %ymm12, 352(%rax)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 256(%rax)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 224(%rax)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 96(%rax)
+; AVX1-NEXT: addq $88, %rsp
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-SLOW-LABEL: vf32:
+; AVX2-SLOW: # %bb.0:
+; AVX2-SLOW-NEXT: subq $296, %rsp # imm = 0x128
+; AVX2-SLOW-NEXT: vmovdqa (%r9), %xmm10
+; AVX2-SLOW-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 32(%r9), %xmm9
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm9[1,2,2,3]
+; AVX2-SLOW-NEXT: vmovdqa (%r8), %xmm11
+; AVX2-SLOW-NEXT: vmovdqa 32(%r8), %xmm8
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm8[1,2,2,3]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,0,2,1]
+; AVX2-SLOW-NEXT: vmovdqa (%rsi), %xmm15
+; AVX2-SLOW-NEXT: vmovdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 32(%rsi), %xmm5
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm5[0,1,2,1]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
+; AVX2-SLOW-NEXT: vmovdqa (%rdi), %xmm12
+; AVX2-SLOW-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %xmm3
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm3[0,1,2,1]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,5]
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vmovdqa (%rcx), %xmm13
+; AVX2-SLOW-NEXT: vmovdqa 32(%rcx), %xmm2
+; AVX2-SLOW-NEXT: vpsrldq {{.*#+}} xmm7 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX2-SLOW-NEXT: vmovdqa (%rdx), %xmm14
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdx), %xmm1
+; AVX2-SLOW-NEXT: vpsrldq {{.*#+}} xmm6 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; AVX2-SLOW-NEXT: vpbroadcastq %xmm6, %ymm6
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1],ymm6[2],ymm4[3,4],ymm6[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm10[1,2,2,3]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm11[1,2,2,3]
+; AVX2-SLOW-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,0,2,1]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm15[0,1,2,1]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,7,6,5]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm7 = xmm12[0,1,2,1]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,7,6,5]
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm6[0,1,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0],ymm6[1],ymm4[2,3],ymm6[4],ymm4[5,6],ymm6[7]
+; AVX2-SLOW-NEXT: vpsrldq {{.*#+}} xmm6 = xmm13[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX2-SLOW-NEXT: vpsrldq {{.*#+}} xmm7 = xmm14[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; AVX2-SLOW-NEXT: vpbroadcastq %xmm6, %ymm6
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1],ymm6[2],ymm4[3,4],ymm6[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm8[4],xmm9[4],xmm8[5],xmm9[5],xmm8[6],xmm9[6],xmm8[7],xmm9[7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,0,1]
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[1,2,3,3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm6[0,0,2,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0],ymm4[1],ymm6[2,3],ymm4[4],ymm6[5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm6[1,1,1,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1],ymm6[2],ymm4[3,4],ymm6[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 32(%r8), %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 32(%r9), %ymm4
+; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm6 = ymm0[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} ymm4 = ymm6[0],ymm4[0],ymm6[1],ymm4[1],ymm6[2],ymm4[2],ymm6[3],ymm4[3],ymm6[8],ymm4[8],ymm6[9],ymm4[9],ymm6[10],ymm4[10],ymm6[11],ymm4[11]
+; AVX2-SLOW-NEXT: vmovdqa 32(%rsi), %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm6 = ymm0[2,1,2,3,6,5,6,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm6 = ymm6[0,3,2,1,4,5,6,7,8,11,10,9,12,13,14,15]
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm7 = ymm0[2,1,2,3,6,5,6,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm7 = ymm7[0,3,2,1,4,5,6,7,8,11,10,9,12,13,14,15]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} ymm6 = ymm7[0],ymm6[0],ymm7[1],ymm6[1],ymm7[2],ymm6[2],ymm7[3],ymm6[3],ymm7[8],ymm6[8],ymm7[9],ymm6[9],ymm7[10],ymm6[10],ymm7[11],ymm6[11]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,2,2,3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm6[2,1,2,3]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm4[0],ymm6[1],ymm4[2,3],ymm6[4],ymm4[5,6],ymm6[7]
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdx), %ymm10
+; AVX2-SLOW-NEXT: vmovdqa 32(%rcx), %ymm4
+; AVX2-SLOW-NEXT: vpsrldq {{.*#+}} ymm0 = ymm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm4[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX2-SLOW-NEXT: vpsrldq {{.*#+}} ymm15 = ymm10[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm10[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm15[0],ymm0[0],ymm15[1],ymm0[1],ymm15[2],ymm0[2],ymm15[3],ymm0[3],ymm15[8],ymm0[8],ymm15[9],ymm0[9],ymm15[10],ymm0[10],ymm15[11],ymm0[11]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,2,2]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1],ymm0[2],ymm6[3,4],ymm0[5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,0,2,1]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,0,2,2]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm8[0],xmm9[0],xmm8[1],xmm9[1],xmm8[2],xmm9[2],xmm8[3],xmm9[3]
+; AVX2-SLOW-NEXT: vpbroadcastq %xmm1, %ymm1
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm11[4],xmm3[4],xmm11[5],xmm3[5],xmm11[6],xmm3[6],xmm11[7],xmm3[7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm14[4],xmm13[4],xmm14[5],xmm13[5],xmm14[6],xmm13[6],xmm14[7],xmm13[7]
+; AVX2-SLOW-NEXT: vmovdqa %xmm13, %xmm12
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,2,3,3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5,6],ymm0[7]
+; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm2[4],xmm7[4],xmm2[5],xmm7[5],xmm2[6],xmm7[6],xmm2[7],xmm7[7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,1,1,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa (%r8), %ymm6
+; AVX2-SLOW-NEXT: vmovdqa (%r9), %ymm5
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm5[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm6[1,2,2,3,5,6,6,7]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
+; AVX2-SLOW-NEXT: vmovdqa (%rsi), %ymm13
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm13[2,1,2,3,6,5,6,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm1 = ymm1[0,3,2,1,4,5,6,7,8,11,10,9,12,13,14,15]
+; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm11
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm8 = ymm11[2,1,2,3,6,5,6,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} ymm8 = ymm8[0,3,2,1,4,5,6,7,8,11,10,9,12,13,14,15]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm8[0],ymm1[0],ymm8[1],ymm1[1],ymm8[2],ymm1[2],ymm8[3],ymm1[3],ymm8[8],ymm1[8],ymm8[9],ymm1[9],ymm8[10],ymm1[10],ymm8[11],ymm1[11]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,2,3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,1,2,3]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vmovdqa (%rdx), %ymm1
+; AVX2-SLOW-NEXT: vmovdqa (%rcx), %ymm0
+; AVX2-SLOW-NEXT: vpsrldq {{.*#+}} ymm9 = ymm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm0[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX2-SLOW-NEXT: vpsrldq {{.*#+}} ymm15 = ymm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm1[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} ymm9 = ymm15[0],ymm9[0],ymm15[1],ymm9[1],ymm15[2],ymm9[2],ymm15[3],ymm9[3],ymm15[8],ymm9[8],ymm15[9],ymm9[9],ymm15[10],ymm9[10],ymm15[11],ymm9[11]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm9 = ymm9[2,2,2,2]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4],ymm9[5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm2[0],xmm7[0],xmm2[1],xmm7[1],xmm2[2],xmm7[2],xmm2[3],xmm7[3]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm14[0],xmm12[0],xmm14[1],xmm12[1],xmm14[2],xmm12[2],xmm14[3],xmm12[3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm8 = ymm8[0,0,2,1]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[1,0,2,2]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm9 = ymm9[0,1,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3],ymm9[4],ymm8[5,6],ymm9[7]
+; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX2-SLOW-NEXT: vpbroadcastq %xmm7, %ymm7
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1],ymm7[2],ymm8[3,4],ymm7[5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} ymm7 = ymm11[0],ymm13[0],ymm11[1],ymm13[1],ymm11[2],ymm13[2],ymm11[3],ymm13[3],ymm11[8],ymm13[8],ymm11[9],ymm13[9],ymm11[10],ymm13[10],ymm11[11],ymm13[11]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm7 = ymm7[2,2,2,3]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} ymm9 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm9 = ymm9[1,0,2,2,5,4,6,6]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm9 = ymm9[2,1,2,3]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0],ymm9[1],ymm7[2,3],ymm9[4],ymm7[5,6],ymm9[7]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} ymm9 = ymm6[0],ymm5[0],ymm6[1],ymm5[1],ymm6[2],ymm5[2],ymm6[3],ymm5[3],ymm6[8],ymm5[8],ymm6[9],ymm5[9],ymm6[10],ymm5[10],ymm6[11],ymm5[11]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm9 = ymm9[2,2,2,2]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm9 = ymm7[0,1],ymm9[2],ymm7[3,4],ymm9[5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} ymm7 = ymm3[4],ymm2[4],ymm3[5],ymm2[5],ymm3[6],ymm2[6],ymm3[7],ymm2[7],ymm3[12],ymm2[12],ymm3[13],ymm2[13],ymm3[14],ymm2[14],ymm3[15],ymm2[15]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm7 = ymm7[2,1,2,3,6,5,6,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm7 = ymm7[2,1,2,3]
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} ymm15 = ymm10[4],ymm4[4],ymm10[5],ymm4[5],ymm10[6],ymm4[6],ymm10[7],ymm4[7],ymm10[12],ymm4[12],ymm10[13],ymm4[13],ymm10[14],ymm4[14],ymm10[15],ymm4[15]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm15 = ymm15[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm15 = ymm15[2,2,2,3]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm15[0],ymm7[1],ymm15[2,3],ymm7[4],ymm15[5,6],ymm7[7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} ymm15 = ymm14[4],ymm12[4],ymm14[5],ymm12[5],ymm14[6],ymm12[6],ymm14[7],ymm12[7],ymm14[12],ymm12[12],ymm14[13],ymm12[13],ymm14[14],ymm12[14],ymm14[15],ymm12[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm15 = ymm15[3,3,3,3]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1],ymm15[2],ymm7[3,4],ymm15[5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} ymm12 = ymm14[0],ymm12[0],ymm14[1],ymm12[1],ymm14[2],ymm12[2],ymm14[3],ymm12[3],ymm14[8],ymm12[8],ymm14[9],ymm12[9],ymm14[10],ymm12[10],ymm14[11],ymm12[11]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} ymm4 = ymm10[0],ymm4[0],ymm10[1],ymm4[1],ymm10[2],ymm4[2],ymm10[3],ymm4[3],ymm10[8],ymm4[8],ymm10[9],ymm4[9],ymm10[10],ymm4[10],ymm10[11],ymm4[11]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm12[2,2,2,3]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[1,0,2,2,5,4,6,6]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,1,2,3]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm10[0],ymm4[1],ymm10[2,3],ymm4[4],ymm10[5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} ymm10 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[8],ymm2[8],ymm3[9],ymm2[9],ymm3[10],ymm2[10],ymm3[11],ymm2[11]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm10[2,2,2,2]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm10[2],ymm4[3,4],ymm10[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm6[4],ymm5[4],ymm6[5],ymm5[5],ymm6[6],ymm5[6],ymm6[7],ymm5[7],ymm6[12],ymm5[12],ymm6[13],ymm5[13],ymm6[14],ymm5[14],ymm6[15],ymm5[15]
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm2[2,1,2,3,6,5,6,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,1,2,3]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,2,3]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} ymm1 = ymm11[4],ymm13[4],ymm11[5],ymm13[5],ymm11[6],ymm13[6],ymm11[7],ymm13[7],ymm11[12],ymm13[12],ymm11[13],ymm13[13],ymm11[14],ymm13[14],ymm11[15],ymm13[15]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[3,3,3,3]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX2-SLOW-NEXT: vmovdqa %ymm0, 160(%rax)
+; AVX2-SLOW-NEXT: vmovdqa %ymm4, 288(%rax)
+; AVX2-SLOW-NEXT: vmovdqa %ymm7, 352(%rax)
+; AVX2-SLOW-NEXT: vmovdqa %ymm9, 96(%rax)
+; AVX2-SLOW-NEXT: vmovdqa %ymm8, (%rax)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 128(%rax)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 64(%rax)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 192(%rax)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 320(%rax)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 256(%rax)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 224(%rax)
+; AVX2-SLOW-NEXT: addq $296, %rsp # imm = 0x128
+; AVX2-SLOW-NEXT: vzeroupper
+; AVX2-SLOW-NEXT: retq
+;
+; AVX2-FAST-ALL-LABEL: vf32:
+; AVX2-FAST-ALL: # %bb.0:
+; AVX2-FAST-ALL-NEXT: subq $392, %rsp # imm = 0x188
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rsi), %xmm15
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rsi), %xmm14
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} xmm12 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm12, %xmm14, %xmm0
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rdi), %xmm8
+; AVX2-FAST-ALL-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rdi), %xmm9
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm12, %xmm9, %xmm1
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vmovdqa (%r9), %xmm10
+; AVX2-FAST-ALL-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%r9), %xmm3
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} xmm1 = xmm3[1,2,2,3]
+; AVX2-FAST-ALL-NEXT: vmovdqa (%r8), %xmm11
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%r8), %xmm2
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} xmm6 = xmm2[1,2,2,3]
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm6[0],xmm1[0],xmm6[1],xmm1[1],xmm6[2],xmm1[2],xmm6[3],xmm1[3]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm6 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5,6],ymm0[7]
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rcx), %xmm13
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rcx), %xmm1
+; AVX2-FAST-ALL-NEXT: vpsrldq {{.*#+}} xmm4 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rdx), %xmm7
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rdx), %xmm0
+; AVX2-FAST-ALL-NEXT: vpsrldq {{.*#+}} xmm5 = xmm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; AVX2-FAST-ALL-NEXT: vpbroadcastq %xmm5, %ymm5
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1],ymm5[2],ymm6[3,4],ymm5[5],ymm6[6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm12, %xmm15, %xmm5
+; AVX2-FAST-ALL-NEXT: vpshufb %xmm12, %xmm8, %xmm4
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} xmm5 = xmm10[1,2,2,3]
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} xmm6 = xmm11[1,2,2,3]
+; AVX2-FAST-ALL-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm5 = ymm5[0,0,2,1]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5,6],ymm4[7]
+; AVX2-FAST-ALL-NEXT: vpsrldq {{.*#+}} xmm5 = xmm13[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-ALL-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-ALL-NEXT: vpsrldq {{.*#+}} xmm6 = xmm7[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-ALL-NEXT: vmovdqa %xmm7, %xmm12
+; AVX2-FAST-ALL-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; AVX2-FAST-ALL-NEXT: vpbroadcastq %xmm5, %ymm5
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm5 = <1,2,1,2,u,u,3,3>
+; AVX2-FAST-ALL-NEXT: vpermd %ymm4, %ymm5, %ymm4
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm5 = ymm5[0,1,1,1]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3],ymm5[4],ymm4[5,6],ymm5[7]
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm9[4],xmm14[4],xmm9[5],xmm14[5],xmm9[6],xmm14[6],xmm9[7],xmm14[7]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm5 = ymm5[1,1,1,1]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rdi), %ymm5
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rsi), %ymm4
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm10 = <u,u,u,u,4,5,10,11,u,u,u,u,u,u,u,u,24,25,22,23,20,21,26,27,u,u,u,u,u,u,u,u>
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm10, %ymm4, %ymm4
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm10, %ymm5, %ymm5
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} ymm4 = ymm5[0],ymm4[0],ymm5[1],ymm4[1],ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[8],ymm4[8],ymm5[9],ymm4[9],ymm5[10],ymm4[10],ymm5[11],ymm4[11]
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%r8), %ymm6
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%r9), %ymm5
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm5 = ymm5[1,2,2,3,5,6,6,7]
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm7 = ymm6[1,2,2,3,5,6,6,7]
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} ymm5 = ymm7[0],ymm5[0],ymm7[1],ymm5[1],ymm7[2],ymm5[2],ymm7[3],ymm5[3],ymm7[8],ymm5[8],ymm7[9],ymm5[9],ymm7[10],ymm5[10],ymm7[11],ymm5[11]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,1,2,3]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm5 = ymm5[2,2,2,3]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5,6],ymm4[7]
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rdx), %ymm6
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa 32(%rcx), %ymm8
+; AVX2-FAST-ALL-NEXT: vpsrldq {{.*#+}} ymm5 = ymm8[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm8[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-ALL-NEXT: vpsrldq {{.*#+}} ymm6 = ymm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm6[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} ymm5 = ymm6[0],ymm5[0],ymm6[1],ymm5[1],ymm6[2],ymm5[2],ymm6[3],ymm5[3],ymm6[8],ymm5[8],ymm6[9],ymm5[9],ymm6[10],ymm5[10],ymm6[11],ymm5[11]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm5 = ymm5[2,2,2,2]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm9[0],xmm14[0],xmm9[1],xmm14[1],xmm9[2],xmm14[2],xmm9[3],xmm14[3]
+; AVX2-FAST-ALL-NEXT: vbroadcasti128 {{.*#+}} ymm4 = [1,0,2,2,1,0,2,2]
+; AVX2-FAST-ALL-NEXT: # ymm4 = mem[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm0, %ymm4, %ymm0
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5,6],ymm0[7]
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX2-FAST-ALL-NEXT: vpbroadcastq %xmm1, %ymm1
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm12[4],xmm13[4],xmm12[5],xmm13[5],xmm12[6],xmm13[6],xmm12[7],xmm13[7]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm1 = <1,2,1,2,u,u,3,3>
+; AVX2-FAST-ALL-NEXT: vpermd %ymm0, %ymm1, %ymm0
+; AVX2-FAST-ALL-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm11[4],xmm5[4],xmm11[5],xmm5[5],xmm11[6],xmm5[6],xmm11[7],xmm5[7]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,1,1]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX2-FAST-ALL-NEXT: vmovdqa %xmm15, %xmm7
+; AVX2-FAST-ALL-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,1,1,1]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rdi), %ymm1
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rsi), %ymm15
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm10, %ymm15, %ymm0
+; AVX2-FAST-ALL-NEXT: vpshufb %ymm10, %ymm1, %ymm1
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
+; AVX2-FAST-ALL-NEXT: vmovdqa (%r8), %ymm13
+; AVX2-FAST-ALL-NEXT: vmovdqa (%r9), %ymm12
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm3 = ymm12[1,2,2,3,5,6,6,7]
+; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} ymm9 = ymm13[1,2,2,3,5,6,6,7]
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm9[0],ymm3[0],ymm9[1],ymm3[1],ymm9[2],ymm3[2],ymm9[3],ymm3[3],ymm9[8],ymm3[8],ymm9[9],ymm3[9],ymm9[10],ymm3[10],ymm9[11],ymm3[11]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,1,2,3]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,2,2,3]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm9 = ymm3[0],ymm0[1],ymm3[2,3],ymm0[4],ymm3[5,6],ymm0[7]
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rdx), %ymm11
+; AVX2-FAST-ALL-NEXT: vmovdqa (%rcx), %ymm6
+; AVX2-FAST-ALL-NEXT: vpsrldq {{.*#+}} ymm10 = ymm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm6[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-ALL-NEXT: vpsrldq {{.*#+}} ymm14 = ymm11[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm11[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} ymm10 = ymm14[0],ymm10[0],ymm14[1],ymm10[1],ymm14[2],ymm10[2],ymm14[3],ymm10[3],ymm14[8],ymm10[8],ymm14[9],ymm10[9],ymm14[10],ymm10[10],ymm14[11],ymm10[11]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm10 = ymm10[2,2,2,2]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1],ymm10[2],ymm9[3,4],ymm10[5],ymm9[6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm9 # 16-byte Folded Reload
+; AVX2-FAST-ALL-NEXT: # xmm9 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm9, %ymm4, %ymm4
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm2[0],xmm7[0],xmm2[1],xmm7[1],xmm2[2],xmm7[2],xmm2[3],xmm7[3]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm9 = ymm9[0,0,2,1]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0],ymm4[1],ymm9[2,3],ymm4[4],ymm9[5,6],ymm4[7]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
+; AVX2-FAST-ALL-NEXT: vpbroadcastq %xmm7, %ymm7
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1],ymm7[2],ymm4[3,4],ymm7[5],ymm4[6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} ymm7 = ymm11[0],ymm6[0],ymm11[1],ymm6[1],ymm11[2],ymm6[2],ymm11[3],ymm6[3],ymm11[8],ymm6[8],ymm11[9],ymm6[9],ymm11[10],ymm6[10],ymm11[11],ymm6[11]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm9 = [5,4,2,2,5,4,6,6]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm7, %ymm9, %ymm7
+; AVX2-FAST-ALL-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} ymm10 = ymm4[0],ymm15[0],ymm4[1],ymm15[1],ymm4[2],ymm15[2],ymm4[3],ymm15[3],ymm4[8],ymm15[8],ymm4[9],ymm15[9],ymm4[10],ymm15[10],ymm4[11],ymm15[11]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm10 = ymm10[2,2,2,3]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm7 = ymm10[0],ymm7[1],ymm10[2,3],ymm7[4],ymm10[5,6],ymm7[7]
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} ymm10 = ymm13[0],ymm12[0],ymm13[1],ymm12[1],ymm13[2],ymm12[2],ymm13[3],ymm12[3],ymm13[8],ymm12[8],ymm13[9],ymm12[9],ymm13[10],ymm12[10],ymm13[11],ymm12[11]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm10 = ymm10[2,2,2,2]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1],ymm10[2],ymm7[3,4],ymm10[5],ymm7[6,7]
+; AVX2-FAST-ALL-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} ymm10 = ymm0[4],ymm8[4],ymm0[5],ymm8[5],ymm0[6],ymm8[6],ymm0[7],ymm8[7],ymm0[12],ymm8[12],ymm0[13],ymm8[13],ymm0[14],ymm8[14],ymm0[15],ymm8[15]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm1 = [5,6,5,6,5,6,7,7]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm10, %ymm1, %ymm10
+; AVX2-FAST-ALL-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} ymm14 = ymm3[4],ymm1[4],ymm3[5],ymm1[5],ymm3[6],ymm1[6],ymm3[7],ymm1[7],ymm3[12],ymm1[12],ymm3[13],ymm1[13],ymm3[14],ymm1[14],ymm3[15],ymm1[15]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm14 = ymm14[2,1,3,3]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0],ymm14[1],ymm10[2,3],ymm14[4],ymm10[5,6],ymm14[7]
+; AVX2-FAST-ALL-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} ymm14 = ymm5[4],ymm2[4],ymm5[5],ymm2[5],ymm5[6],ymm2[6],ymm5[7],ymm2[7],ymm5[12],ymm2[12],ymm5[13],ymm2[13],ymm5[14],ymm2[14],ymm5[15],ymm2[15]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm14 = ymm14[3,3,3,3]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1],ymm14[2],ymm10[3,4],ymm14[5],ymm10[6,7]
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} ymm8 = ymm0[0],ymm8[0],ymm0[1],ymm8[1],ymm0[2],ymm8[2],ymm0[3],ymm8[3],ymm0[8],ymm8[8],ymm0[9],ymm8[9],ymm0[10],ymm8[10],ymm0[11],ymm8[11]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm8, %ymm9, %ymm8
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} ymm9 = ymm5[0],ymm2[0],ymm5[1],ymm2[1],ymm5[2],ymm2[2],ymm5[3],ymm2[3],ymm5[8],ymm2[8],ymm5[9],ymm2[9],ymm5[10],ymm2[10],ymm5[11],ymm2[11]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm9 = ymm9[2,2,2,3]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0],ymm8[1],ymm9[2,3],ymm8[4],ymm9[5,6],ymm8[7]
+; AVX2-FAST-ALL-NEXT: vpunpcklwd {{.*#+}} ymm9 = ymm3[0],ymm1[0],ymm3[1],ymm1[1],ymm3[2],ymm1[2],ymm3[3],ymm1[3],ymm3[8],ymm1[8],ymm3[9],ymm1[9],ymm3[10],ymm1[10],ymm3[11],ymm1[11]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm9 = ymm9[2,2,2,2]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4],ymm9[5],ymm8[6,7]
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm11[4],ymm6[4],ymm11[5],ymm6[5],ymm11[6],ymm6[6],ymm11[7],ymm6[7],ymm11[12],ymm6[12],ymm11[13],ymm6[13],ymm11[14],ymm6[14],ymm11[15],ymm6[15]
+; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm1 = [5,6,5,6,5,6,7,7]
+; AVX2-FAST-ALL-NEXT: vpermd %ymm0, %ymm1, %ymm0
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} ymm1 = ymm13[4],ymm12[4],ymm13[5],ymm12[5],ymm13[6],ymm12[6],ymm13[7],ymm12[7],ymm13[12],ymm12[12],ymm13[13],ymm12[13],ymm13[14],ymm12[14],ymm13[15],ymm12[15]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,1,3,3]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} ymm1 = ymm4[4],ymm15[4],ymm4[5],ymm15[5],ymm4[6],ymm15[6],ymm4[7],ymm15[7],ymm4[12],ymm15[12],ymm4[13],ymm15[13],ymm4[14],ymm15[14],ymm4[15],ymm15[15]
+; AVX2-FAST-ALL-NEXT: vpermq {{.*#+}} ymm1 = ymm1[3,3,3,3]
+; AVX2-FAST-ALL-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-ALL-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm0, 160(%rax)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm8, 288(%rax)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm10, 352(%rax)
+; AVX2-FAST-ALL-NEXT: vmovdqa %ymm7, 96(%rax)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, (%rax)
+; AVX2-FAST-ALL-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, 128(%rax)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, 64(%rax)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, 192(%rax)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, 320(%rax)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, 256(%rax)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, 224(%rax)
+; AVX2-FAST-ALL-NEXT: addq $392, %rsp # imm = 0x188
+; AVX2-FAST-ALL-NEXT: vzeroupper
+; AVX2-FAST-ALL-NEXT: retq
+;
+; AVX2-FAST-PERLANE-LABEL: vf32:
+; AVX2-FAST-PERLANE: # %bb.0:
+; AVX2-FAST-PERLANE-NEXT: subq $168, %rsp
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rsi), %ymm9
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdx), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rcx), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rsi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rsi), %xmm13
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm8 = [0,1,2,3,4,5,6,7,8,9,6,7,4,5,10,11]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm13, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm4, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm11 = ymm0[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%r9), %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = xmm7[1,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%r8), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[1,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm11 = ymm1[0],ymm11[1],ymm1[2,3],ymm11[4],ymm1[5,6],ymm11[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rcx), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpsrldq {{.*#+}} xmm1 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdx), %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpsrldq {{.*#+}} xmm3 = xmm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastq %xmm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm1[2],ymm11[3,4],ymm1[5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm6[4],xmm2[4],xmm6[5],xmm2[5],xmm6[6],xmm2[6],xmm6[7],xmm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[1,2,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0],ymm1[1],ymm3[2,3],ymm1[4],ymm3[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm4[4],xmm13[4],xmm4[5],xmm13[5],xmm4[6],xmm13[6],xmm4[7],xmm13[7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%r9), %xmm11
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm4[0],xmm13[0],xmm4[1],xmm13[1],xmm4[2],xmm13[2],xmm4[3],xmm13[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%r8), %xmm13
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm6[0],xmm2[0],xmm6[1],xmm2[1],xmm6[2],xmm2[2],xmm6[3],xmm2[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rcx), %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,0,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3],ymm2[4],ymm1[5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdx), %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastq %xmm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm12, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm5, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = xmm11[1,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm3 = xmm13[1,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT: vpsrldq {{.*#+}} xmm1 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpsrldq {{.*#+}} xmm3 = xmm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastq %xmm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm13[4],xmm11[4],xmm13[5],xmm11[5],xmm13[6],xmm11[6],xmm13[7],xmm11[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm6[4],xmm4[4],xmm6[5],xmm4[5],xmm6[6],xmm4[6],xmm6[7],xmm4[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,2,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm5[4],xmm12[4],xmm5[5],xmm12[5],xmm5[6],xmm12[6],xmm5[7],xmm12[7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%r8), %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%r8), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%r9), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,0,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%r9), %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm13[0],xmm11[0],xmm13[1],xmm11[1],xmm13[2],xmm11[2],xmm13[3],xmm11[3]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastq %xmm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[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: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[1],ymm9[1],ymm0[2],ymm9[2],ymm0[3],ymm9[3],ymm0[8],ymm9[8],ymm0[9],ymm9[9],ymm0[10],ymm9[10],ymm0[11],ymm9[11]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm10[0],ymm7[0],ymm10[1],ymm7[1],ymm10[2],ymm7[2],ymm10[3],ymm7[3],ymm10[8],ymm7[8],ymm10[9],ymm7[9],ymm10[10],ymm7[10],ymm10[11],ymm7[11]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm10, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,0,2,2,5,4,6,6]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,1,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm15[0],ymm12[0],ymm15[1],ymm12[1],ymm15[2],ymm12[2],ymm15[3],ymm12[3],ymm15[8],ymm12[8],ymm15[9],ymm12[9],ymm15[10],ymm12[10],ymm15[11],ymm12[11]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm14[4],ymm5[4],ymm14[5],ymm5[5],ymm14[6],ymm5[6],ymm14[7],ymm5[7],ymm14[12],ymm5[12],ymm14[13],ymm5[13],ymm14[14],ymm5[14],ymm14[15],ymm5[15]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,1,2,3,6,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,1,2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdx), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rcx), %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} ymm2 = ymm4[4],ymm3[4],ymm4[5],ymm3[5],ymm4[6],ymm3[6],ymm4[7],ymm3[7],ymm4[12],ymm3[12],ymm4[13],ymm3[13],ymm4[14],ymm3[14],ymm4[15],ymm3[15]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm0[1],ymm2[2,3],ymm0[4],ymm2[5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rsi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} ymm6 = ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm6 = ymm6[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm11 = ymm2[0,1],ymm6[2],ymm2[3,4],ymm6[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,4,5,10,11,u,u,u,u,u,u,u,u,24,25,22,23,20,21,26,27,u,u,u,u,u,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm0, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} ymm6 = ymm8[0],ymm6[0],ymm8[1],ymm6[1],ymm8[2],ymm6[2],ymm8[3],ymm6[3],ymm8[8],ymm6[8],ymm8[9],ymm6[9],ymm8[10],ymm6[10],ymm8[11],ymm6[11]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm8 = ymm5[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm10 = ymm14[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} ymm8 = ymm10[0],ymm8[0],ymm10[1],ymm8[1],ymm10[2],ymm8[2],ymm10[3],ymm8[3],ymm10[8],ymm8[8],ymm10[9],ymm8[9],ymm10[10],ymm8[10],ymm10[11],ymm8[11]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm6 = ymm6[2,1,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm8 = ymm8[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0],ymm6[1],ymm8[2,3],ymm6[4],ymm8[5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT: vpsrldq {{.*#+}} ymm8 = ymm3[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm3[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpsrldq {{.*#+}} ymm10 = ymm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm4[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} ymm8 = ymm10[0],ymm8[0],ymm10[1],ymm8[1],ymm10[2],ymm8[2],ymm10[3],ymm8[3],ymm10[8],ymm8[8],ymm10[9],ymm8[9],ymm10[10],ymm8[10],ymm10[11],ymm8[11]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm8 = ymm8[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1],ymm8[2],ymm6[3,4],ymm8[5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[8],ymm3[8],ymm4[9],ymm3[9],ymm4[10],ymm3[10],ymm4[11],ymm3[11]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,0,2,2,5,4,6,6]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,1,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm14[0],ymm5[0],ymm14[1],ymm5[1],ymm14[2],ymm5[2],ymm14[3],ymm5[3],ymm14[8],ymm5[8],ymm14[9],ymm5[9],ymm14[10],ymm5[10],ymm14[11],ymm5[11]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} ymm1 = ymm15[4],ymm12[4],ymm15[5],ymm12[5],ymm15[6],ymm12[6],ymm15[7],ymm12[7],ymm15[12],ymm12[12],ymm15[13],ymm12[13],ymm15[14],ymm12[14],ymm15[15],ymm12[15]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[2,1,2,3,6,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,1,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm9[4],ymm7[4],ymm9[5],ymm7[5],ymm9[6],ymm7[6],ymm9[7],ymm7[7],ymm9[12],ymm7[12],ymm9[13],ymm7[13],ymm9[14],ymm7[14],ymm9[15],ymm7[15]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0],ymm1[1],ymm3[2,3],ymm1[4],ymm3[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm4[4],ymm13[4],ymm4[5],ymm13[5],ymm4[6],ymm13[6],ymm4[7],ymm13[7],ymm4[12],ymm13[12],ymm4[13],ymm13[13],ymm4[14],ymm13[14],ymm4[15],ymm13[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm13, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm4, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[8],ymm3[8],ymm2[9],ymm3[9],ymm2[10],ymm3[10],ymm2[11],ymm3[11]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm3 = ymm12[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm4 = ymm15[1,2,2,3,5,6,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[8],ymm3[8],ymm4[9],ymm3[9],ymm4[10],ymm3[10],ymm4[11],ymm3[11]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,1,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpsrldq {{.*#+}} ymm3 = ymm7[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm7[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpsrldq {{.*#+}} ymm4 = ymm9[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,ymm9[22,23,24,25,26,27,28,29,30,31],zero,zero,zero,zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[8],ymm3[8],ymm4[9],ymm3[9],ymm4[10],ymm3[10],ymm4[11],ymm3[11]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, 128(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, 160(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, 288(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, 320(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm11, 352(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 96(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, (%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 64(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 192(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 256(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 224(%rax)
+; AVX2-FAST-PERLANE-NEXT: addq $168, %rsp
+; AVX2-FAST-PERLANE-NEXT: vzeroupper
+; AVX2-FAST-PERLANE-NEXT: retq
+;
+; AVX512-LABEL: vf32:
+; AVX512: # %bb.0:
+; AVX512-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm3
+; AVX512-NEXT: vmovdqu64 (%rsi), %zmm4
+; AVX512-NEXT: vmovdqu64 (%rdx), %zmm1
+; AVX512-NEXT: vmovdqu64 (%rcx), %zmm2
+; AVX512-NEXT: vmovdqu64 (%r8), %zmm5
+; AVX512-NEXT: vmovdqu64 (%r9), %zmm6
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm7 = [0,0,0,32,3,35,0,0,1,33,4,36,0,0,2,34,0,0,0,32,3,35,0,0,1,33,4,36,0,0,2,34]
+; AVX512-NEXT: # zmm7 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm2, %zmm1, %zmm7
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = [0,32,3,35,0,0,1,33,4,36,0,0,2,34,5,37,0,32,3,35,0,0,1,33,4,36,0,0,2,34,5,37]
+; AVX512-NEXT: # zmm0 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm3, %zmm0
+; AVX512-NEXT: movw $9362, %cx # imm = 0x2492
+; AVX512-NEXT: kmovd %ecx, %k1
+; AVX512-NEXT: vmovdqa32 %zmm7, %zmm0 {%k1}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm7 = [2,34,0,0,0,32,3,35,0,0,1,33,4,36,0,0,2,34,0,0,0,32,3,35,0,0,1,33,4,36,0,0]
+; AVX512-NEXT: # zmm7 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm6, %zmm5, %zmm7
+; AVX512-NEXT: movw $18724, %cx # imm = 0x4924
+; AVX512-NEXT: kmovd %ecx, %k2
+; AVX512-NEXT: vmovdqa32 %zmm7, %zmm0 {%k2}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,5,37,8,40,0,0,6,38,9,41,0,0,7,39,0,0,5,37,8,40,0,0,6,38,9,41,0,0,7,39]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm6, %zmm5, %zmm8
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm7 = [5,37,8,40,0,0,6,38,9,41,0,0,7,39,10,42,5,37,8,40,0,0,6,38,9,41,0,0,7,39,10,42]
+; AVX512-NEXT: # zmm7 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm2, %zmm1, %zmm7
+; AVX512-NEXT: vmovdqa32 %zmm8, %zmm7 {%k1}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [8,40,0,0,6,38,9,41,0,0,7,39,10,42,0,0,8,40,0,0,6,38,9,41,0,0,7,39,10,42,0,0]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm3, %zmm8
+; AVX512-NEXT: vmovdqa32 %zmm8, %zmm7 {%k2}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,11,43,14,46,0,0,12,44,15,47,0,0,13,45,0,0,11,43,14,46,0,0,12,44,15,47,0,0,13,45]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm3, %zmm8
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm9 = [10,42,13,45,0,0,11,43,14,46,0,0,12,44,15,47,10,42,13,45,0,0,11,43,14,46,0,0,12,44,15,47]
+; AVX512-NEXT: # zmm9 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm6, %zmm5, %zmm9
+; AVX512-NEXT: vmovdqa32 %zmm8, %zmm9 {%k1}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [13,45,0,0,11,43,14,46,0,0,12,44,15,47,0,0,13,45,0,0,11,43,14,46,0,0,12,44,15,47,0,0]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm2, %zmm1, %zmm8
+; AVX512-NEXT: vmovdqa32 %zmm8, %zmm9 {%k2}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,16,48,19,51,0,0,17,49,20,52,0,0,18,50,0,0,16,48,19,51,0,0,17,49,20,52,0,0,18,50]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm2, %zmm1, %zmm8
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm10 = [16,48,19,51,0,0,17,49,20,52,0,0,18,50,21,53,16,48,19,51,0,0,17,49,20,52,0,0,18,50,21,53]
+; AVX512-NEXT: # zmm10 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm3, %zmm10
+; AVX512-NEXT: vmovdqa32 %zmm8, %zmm10 {%k1}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [18,50,0,0,16,48,19,51,0,0,17,49,20,52,0,0,18,50,0,0,16,48,19,51,0,0,17,49,20,52,0,0]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm6, %zmm5, %zmm8
+; AVX512-NEXT: vmovdqa32 %zmm8, %zmm10 {%k2}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,21,53,24,56,0,0,22,54,25,57,0,0,23,55,0,0,21,53,24,56,0,0,22,54,25,57,0,0,23,55]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm6, %zmm5, %zmm8
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm11 = [21,53,24,56,0,0,22,54,25,57,0,0,23,55,26,58,21,53,24,56,0,0,22,54,25,57,0,0,23,55,26,58]
+; AVX512-NEXT: # zmm11 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm2, %zmm1, %zmm11
+; AVX512-NEXT: vmovdqa32 %zmm8, %zmm11 {%k1}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [24,56,0,0,22,54,25,57,0,0,23,55,26,58,0,0,24,56,0,0,22,54,25,57,0,0,23,55,26,58,0,0]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm3, %zmm8
+; AVX512-NEXT: vmovdqa32 %zmm8, %zmm11 {%k2}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,0,27,59,30,62,0,0,28,60,31,63,0,0,29,61,0,0,27,59,30,62,0,0,28,60,31,63,0,0,29,61]
+; AVX512-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm4, %zmm3, %zmm8
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm3 = [26,58,29,61,0,0,27,59,30,62,0,0,28,60,31,63,26,58,29,61,0,0,27,59,30,62,0,0,28,60,31,63]
+; AVX512-NEXT: # zmm3 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm6, %zmm5, %zmm3
+; AVX512-NEXT: vmovdqa32 %zmm8, %zmm3 {%k1}
+; AVX512-NEXT: vbroadcasti64x4 {{.*#+}} zmm4 = [29,61,0,0,27,59,30,62,0,0,28,60,31,63,0,0,29,61,0,0,27,59,30,62,0,0,28,60,31,63,0,0]
+; AVX512-NEXT: # zmm4 = mem[0,1,2,3,0,1,2,3]
+; AVX512-NEXT: vpermi2w %zmm2, %zmm1, %zmm4
+; AVX512-NEXT: vmovdqa32 %zmm4, %zmm3 {%k2}
+; AVX512-NEXT: vmovdqu64 %zmm3, 320(%rax)
+; AVX512-NEXT: vmovdqu64 %zmm11, 256(%rax)
+; AVX512-NEXT: vmovdqu64 %zmm10, 192(%rax)
+; AVX512-NEXT: vmovdqu64 %zmm9, 128(%rax)
+; AVX512-NEXT: vmovdqu64 %zmm7, 64(%rax)
+; AVX512-NEXT: vmovdqu64 %zmm0, (%rax)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %in.vec0 = load <32 x i16>, <32 x i16>* %in.vecptr0, align 32
+ %in.vec1 = load <32 x i16>, <32 x i16>* %in.vecptr1, align 32
+ %in.vec2 = load <32 x i16>, <32 x i16>* %in.vecptr2, align 32
+ %in.vec3 = load <32 x i16>, <32 x i16>* %in.vecptr3, align 32
+ %in.vec4 = load <32 x i16>, <32 x i16>* %in.vecptr4, align 32
+ %in.vec5 = load <32 x i16>, <32 x i16>* %in.vecptr5, align 32
+
+ %concat01 = shufflevector <32 x i16> %in.vec0, <32 x i16> %in.vec1, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+ %concat23 = shufflevector <32 x i16> %in.vec2, <32 x i16> %in.vec3, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+ %concat45 = shufflevector <32 x i16> %in.vec4, <32 x i16> %in.vec5, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+ %concat0123 = shufflevector <64 x i16> %concat01, <64 x i16> %concat23, <128 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64, i32 65, i32 66, i32 67, i32 68, i32 69, i32 70, i32 71, i32 72, i32 73, i32 74, i32 75, i32 76, i32 77, i32 78, i32 79, i32 80, i32 81, i32 82, i32 83, i32 84, i32 85, i32 86, i32 87, i32 88, i32 89, i32 90, i32 91, i32 92, i32 93, i32 94, i32 95, i32 96, i32 97, i32 98, i32 99, i32 100, i32 101, i32 102, i32 103, i32 104, i32 105, i32 106, i32 107, i32 108, i32 109, i32 110, i32 111, i32 112, i32 113, i32 114, i32 115, i32 116, i32 117, i32 118, i32 119, i32 120, i32 121, i32 122, i32 123, i32 124, i32 125, i32 126, i32 127>
+ %concat45uu = shufflevector <64 x i16> %concat45, <64 x i16> poison, <128 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+ %concat012345 = shufflevector <128 x i16> %concat0123, <128 x i16> %concat45uu, <192 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64, i32 65, i32 66, i32 67, i32 68, i32 69, i32 70, i32 71, i32 72, i32 73, i32 74, i32 75, i32 76, i32 77, i32 78, i32 79, i32 80, i32 81, i32 82, i32 83, i32 84, i32 85, i32 86, i32 87, i32 88, i32 89, i32 90, i32 91, i32 92, i32 93, i32 94, i32 95, i32 96, i32 97, i32 98, i32 99, i32 100, i32 101, i32 102, i32 103, i32 104, i32 105, i32 106, i32 107, i32 108, i32 109, i32 110, i32 111, i32 112, i32 113, i32 114, i32 115, i32 116, i32 117, i32 118, i32 119, i32 120, i32 121, i32 122, i32 123, i32 124, i32 125, i32 126, i32 127, i32 128, i32 129, i32 130, i32 131, i32 132, i32 133, i32 134, i32 135, i32 136, i32 137, i32 138, i32 139, i32 140, i32 141, i32 142, i32 143, i32 144, i32 145, i32 146, i32 147, i32 148, i32 149, i32 150, i32 151, i32 152, i32 153, i32 154, i32 155, i32 156, i32 157, i32 158, i32 159, i32 160, i32 161, i32 162, i32 163, i32 164, i32 165, i32 166, i32 167, i32 168, i32 169, i32 170, i32 171, i32 172, i32 173, i32 174, i32 175, i32 176, i32 177, i32 178, i32 179, i32 180, i32 181, i32 182, i32 183, i32 184, i32 185, i32 186, i32 187, i32 188, i32 189, i32 190, i32 191>
+ %interleaved.vec = shufflevector <192 x i16> %concat012345, <192 x i16> poison, <192 x i32> <i32 0, i32 32, i32 64, i32 96, i32 128, i32 160, i32 1, i32 33, i32 65, i32 97, i32 129, i32 161, i32 2, i32 34, i32 66, i32 98, i32 130, i32 162, i32 3, i32 35, i32 67, i32 99, i32 131, i32 163, i32 4, i32 36, i32 68, i32 100, i32 132, i32 164, i32 5, i32 37, i32 69, i32 101, i32 133, i32 165, i32 6, i32 38, i32 70, i32 102, i32 134, i32 166, i32 7, i32 39, i32 71, i32 103, i32 135, i32 167, i32 8, i32 40, i32 72, i32 104, i32 136, i32 168, i32 9, i32 41, i32 73, i32 105, i32 137, i32 169, i32 10, i32 42, i32 74, i32 106, i32 138, i32 170, i32 11, i32 43, i32 75, i32 107, i32 139, i32 171, i32 12, i32 44, i32 76, i32 108, i32 140, i32 172, i32 13, i32 45, i32 77, i32 109, i32 141, i32 173, i32 14, i32 46, i32 78, i32 110, i32 142, i32 174, i32 15, i32 47, i32 79, i32 111, i32 143, i32 175, i32 16, i32 48, i32 80, i32 112, i32 144, i32 176, i32 17, i32 49, i32 81, i32 113, i32 145, i32 177, i32 18, i32 50, i32 82, i32 114, i32 146, i32 178, i32 19, i32 51, i32 83, i32 115, i32 147, i32 179, i32 20, i32 52, i32 84, i32 116, i32 148, i32 180, i32 21, i32 53, i32 85, i32 117, i32 149, i32 181, i32 22, i32 54, i32 86, i32 118, i32 150, i32 182, i32 23, i32 55, i32 87, i32 119, i32 151, i32 183, i32 24, i32 56, i32 88, i32 120, i32 152, i32 184, i32 25, i32 57, i32 89, i32 121, i32 153, i32 185, i32 26, i32 58, i32 90, i32 122, i32 154, i32 186, i32 27, i32 59, i32 91, i32 123, i32 155, i32 187, i32 28, i32 60, i32 92, i32 124, i32 156, i32 188, i32 29, i32 61, i32 93, i32 125, i32 157, i32 189, i32 30, i32 62, i32 94, i32 126, i32 158, i32 190, i32 31, i32 63, i32 95, i32 127, i32 159, i32 191>
+
+ store <192 x i16> %interleaved.vec, <192 x i16>* %out.vec, align 32
+
+ ret void
+}
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-3.ll
index 133346878cced..0c05b3e104be2 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-3.ll
@@ -715,3 +715,658 @@ define void @store_i32_stride3_vf16(<16 x i32>* %in.vecptr0, <16 x i32>* %in.vec
ret void
}
+
+define void @store_i32_stride3_vf32(<32 x i32>* %in.vecptr0, <32 x i32>* %in.vecptr1, <32 x i32>* %in.vecptr2, <96 x i32>* %out.vec) nounwind {
+; SSE-LABEL: store_i32_stride3_vf32:
+; SSE: # %bb.0:
+; SSE-NEXT: subq $56, %rsp
+; SSE-NEXT: movaps 64(%rdi), %xmm12
+; SSE-NEXT: movaps (%rdi), %xmm2
+; SSE-NEXT: movaps 16(%rdi), %xmm10
+; SSE-NEXT: movaps 32(%rdi), %xmm9
+; SSE-NEXT: movaps 48(%rdi), %xmm8
+; SSE-NEXT: movaps (%rsi), %xmm7
+; SSE-NEXT: movaps 16(%rsi), %xmm3
+; SSE-NEXT: movaps 32(%rsi), %xmm5
+; SSE-NEXT: movaps 48(%rsi), %xmm6
+; SSE-NEXT: movaps (%rdx), %xmm1
+; SSE-NEXT: movaps 16(%rdx), %xmm11
+; SSE-NEXT: movaps 32(%rdx), %xmm13
+; SSE-NEXT: movaps 48(%rdx), %xmm14
+; SSE-NEXT: movaps %xmm1, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[1,3]
+; SSE-NEXT: movaps %xmm2, %xmm4
+; SSE-NEXT: unpcklps {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[0,2]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm2, %xmm0
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm7[1]
+; SSE-NEXT: movaps %xmm1, %xmm4
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm7[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm0[0,2]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[3,3],xmm1[3,3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm7[0,2]
+; SSE-NEXT: movaps %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT: movaps %xmm11, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm10[1,3]
+; SSE-NEXT: movaps %xmm10, %xmm2
+; SSE-NEXT: unpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm10, %xmm0
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
+; SSE-NEXT: movaps %xmm11, %xmm2
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm3[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,3],xmm11[3,3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm11 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,3],xmm3[0,2]
+; SSE-NEXT: movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm13, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm9[1,3]
+; SSE-NEXT: movaps %xmm9, %xmm2
+; SSE-NEXT: unpcklps {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm9, %xmm0
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm5[1]
+; SSE-NEXT: movaps %xmm13, %xmm2
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm5[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[3,3],xmm13[3,3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm13 = xmm13[2],xmm9[2],xmm13[3],xmm9[3]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[0,3],xmm5[0,2]
+; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm14, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm8[1,3]
+; SSE-NEXT: movaps %xmm8, %xmm2
+; SSE-NEXT: unpcklps {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm8, %xmm0
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm6[1]
+; SSE-NEXT: movaps %xmm14, %xmm15
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[1,0],xmm6[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[2,0],xmm0[0,2]
+; SSE-NEXT: movaps 64(%rsi), %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[3,3],xmm14[3,3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm14 = xmm14[2],xmm8[2],xmm14[3],xmm8[3]
+; SSE-NEXT: movaps 64(%rdx), %xmm10
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,3],xmm6[0,2]
+; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm10, %xmm1
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm12[1,3]
+; SSE-NEXT: movaps %xmm12, %xmm14
+; SSE-NEXT: unpcklps {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,1],xmm1[0,2]
+; SSE-NEXT: movaps %xmm12, %xmm1
+; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT: movaps %xmm10, %xmm13
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[1,0],xmm0[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[2,0],xmm1[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3],xmm10[3,3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm10 = xmm10[2],xmm12[2],xmm10[3],xmm12[3]
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,3],xmm0[0,2]
+; SSE-NEXT: movaps 80(%rdi), %xmm3
+; SSE-NEXT: movaps 80(%rdx), %xmm12
+; SSE-NEXT: movaps %xmm12, %xmm1
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm3[1,3]
+; SSE-NEXT: movaps 80(%rsi), %xmm0
+; SSE-NEXT: movaps %xmm3, %xmm11
+; SSE-NEXT: unpcklps {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,1],xmm1[0,2]
+; SSE-NEXT: movaps %xmm3, %xmm1
+; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT: movaps %xmm12, %xmm9
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[1,0],xmm0[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[2,0],xmm1[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3],xmm12[3,3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm12 = xmm12[2],xmm3[2],xmm12[3],xmm3[3]
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[0,3],xmm0[0,2]
+; SSE-NEXT: movaps 96(%rdi), %xmm3
+; SSE-NEXT: movaps 96(%rdx), %xmm6
+; SSE-NEXT: movaps %xmm6, %xmm2
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm3[1,3]
+; SSE-NEXT: movaps 96(%rsi), %xmm0
+; SSE-NEXT: movaps %xmm3, %xmm8
+; SSE-NEXT: unpcklps {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[0,1],xmm2[0,2]
+; SSE-NEXT: movaps %xmm3, %xmm2
+; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT: movaps %xmm6, %xmm7
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[1,0],xmm0[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[2,0],xmm2[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3],xmm6[3,3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,3],xmm0[0,2]
+; SSE-NEXT: movaps 112(%rdi), %xmm5
+; SSE-NEXT: movaps 112(%rdx), %xmm2
+; SSE-NEXT: movaps %xmm2, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm5[1,3]
+; SSE-NEXT: movaps 112(%rsi), %xmm0
+; SSE-NEXT: movaps %xmm5, %xmm4
+; SSE-NEXT: unpcklps {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[0,2]
+; SSE-NEXT: movaps %xmm5, %xmm3
+; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm0[1]
+; SSE-NEXT: movaps %xmm2, %xmm1
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[1,0]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm2 = xmm2[2],xmm5[2],xmm2[3],xmm5[3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, 368(%rcx)
+; SSE-NEXT: movaps %xmm1, 352(%rcx)
+; SSE-NEXT: movaps %xmm4, 336(%rcx)
+; SSE-NEXT: movaps %xmm6, 320(%rcx)
+; SSE-NEXT: movaps %xmm7, 304(%rcx)
+; SSE-NEXT: movaps %xmm8, 288(%rcx)
+; SSE-NEXT: movaps %xmm12, 272(%rcx)
+; SSE-NEXT: movaps %xmm9, 256(%rcx)
+; SSE-NEXT: movaps %xmm11, 240(%rcx)
+; SSE-NEXT: movaps %xmm10, 224(%rcx)
+; SSE-NEXT: movaps %xmm13, 208(%rcx)
+; SSE-NEXT: movaps %xmm14, 192(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 176(%rcx)
+; SSE-NEXT: movaps %xmm15, 160(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 144(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 128(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 112(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 96(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 80(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 64(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 48(%rcx)
+; SSE-NEXT: movaps (%rsp), %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: addq $56, %rsp
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: store_i32_stride3_vf32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovaps (%rsi), %xmm0
+; AVX1-NEXT: vmovaps (%rdi), %xmm1
+; AVX1-NEXT: vmovaps 32(%rdi), %xmm5
+; AVX1-NEXT: vmovaps 64(%rdi), %xmm4
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm1[1],xmm0[1]
+; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm0[1,1],xmm2[0,2]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT: vbroadcastsd (%rdx), %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm8 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX1-NEXT: vmovaps 80(%rdx), %xmm1
+; AVX1-NEXT: vmovaps 80(%rsi), %xmm2
+; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm2[3,0],xmm1[3,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm1[2,1],xmm3[0,2]
+; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],xmm2[1,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-NEXT: vbroadcastsd 88(%rdi), %ymm2
+; AVX1-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
+; AVX1-NEXT: vmovaps 48(%rdx), %xmm2
+; AVX1-NEXT: vmovaps 48(%rsi), %xmm3
+; AVX1-NEXT: vshufps {{.*#+}} xmm6 = xmm3[3,0],xmm2[3,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm6 = xmm2[2,1],xmm6[0,2]
+; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,0],xmm3[1,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm2[2,0],xmm3[2,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm2, %ymm2
+; AVX1-NEXT: vbroadcastsd 56(%rdi), %ymm3
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX1-NEXT: vmovaps 112(%rdx), %xmm3
+; AVX1-NEXT: vmovaps 112(%rsi), %xmm6
+; AVX1-NEXT: vshufps {{.*#+}} xmm7 = xmm6[3,0],xmm3[3,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm7 = xmm3[2,1],xmm7[0,2]
+; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,0],xmm6[1,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm3[2,0],xmm6[2,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm3, %ymm3
+; AVX1-NEXT: vbroadcastsd 120(%rdi), %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7]
+; AVX1-NEXT: vmovaps 64(%rsi), %xmm6
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm4[1],xmm6[1]
+; AVX1-NEXT: vshufps {{.*#+}} xmm7 = xmm6[1,1],xmm7[0,2]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm6 = xmm6[0],xmm4[0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm4 = xmm6[2,0],xmm4[2,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm4, %ymm4
+; AVX1-NEXT: vbroadcastsd 64(%rdx), %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm6[2],ymm4[3,4],ymm6[5],ymm4[6,7]
+; AVX1-NEXT: vmovaps 32(%rsi), %xmm6
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm5[1],xmm6[1]
+; AVX1-NEXT: vshufps {{.*#+}} xmm7 = xmm6[1,1],xmm7[0,2]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm6 = xmm6[0],xmm5[0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm5 = xmm6[2,0],xmm5[2,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm5, %ymm5
+; AVX1-NEXT: vbroadcastsd 32(%rdx), %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2],ymm5[3,4],ymm6[5],ymm5[6,7]
+; AVX1-NEXT: vmovaps 96(%rsi), %xmm6
+; AVX1-NEXT: vmovaps 96(%rdi), %xmm7
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm7[1],xmm6[1]
+; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm6[1,1],xmm0[0,2]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm6 = xmm6[0],xmm7[0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm6 = xmm6[2,0],xmm7[2,1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm6, %ymm0
+; AVX1-NEXT: vbroadcastsd 96(%rdx), %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1],ymm6[2],ymm0[3,4],ymm6[5],ymm0[6,7]
+; AVX1-NEXT: vmovaps 16(%rdx), %xmm0
+; AVX1-NEXT: vmovaps 16(%rsi), %xmm7
+; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm7[3,0],xmm0[3,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[2,1],xmm1[0,2]
+; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[1,0],xmm7[1,0]
+; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[2,2]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT: vbroadcastsd 24(%rdi), %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm1 = mem[1,0,2,2]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm7 = mem[1,1,2,2]
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0],ymm1[1],ymm7[2,3],ymm1[4],ymm7[5,6],ymm1[7]
+; AVX1-NEXT: vpermilps {{.*#+}} ymm7 = mem[0,0,3,3,4,4,7,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm7[2],ymm1[3,4],ymm7[5],ymm1[6,7]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm7 = mem[1,0,2,2]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm10 = mem[1,1,2,2]
+; AVX1-NEXT: vblendps {{.*#+}} ymm7 = ymm10[0],ymm7[1],ymm10[2,3],ymm7[4],ymm10[5,6],ymm7[7]
+; AVX1-NEXT: vpermilps {{.*#+}} ymm10 = mem[0,0,3,3,4,4,7,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm10[2],ymm7[3,4],ymm10[5],ymm7[6,7]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm10 = mem[1,0,2,2]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm11 = mem[1,1,2,2]
+; AVX1-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0],ymm10[1],ymm11[2,3],ymm10[4],ymm11[5,6],ymm10[7]
+; AVX1-NEXT: vpermilps {{.*#+}} ymm11 = mem[0,0,3,3,4,4,7,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm11[2],ymm10[3,4],ymm11[5],ymm10[6,7]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm11 = mem[1,0,2,2]
+; AVX1-NEXT: vpermilpd {{.*#+}} ymm12 = mem[1,1,2,2]
+; AVX1-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0],ymm11[1],ymm12[2,3],ymm11[4],ymm12[5,6],ymm11[7]
+; AVX1-NEXT: vpermilps {{.*#+}} ymm12 = mem[0,0,3,3,4,4,7,7]
+; AVX1-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm12[2],ymm11[3,4],ymm12[5],ymm11[6,7]
+; AVX1-NEXT: vmovaps %ymm11, 32(%rcx)
+; AVX1-NEXT: vmovaps %ymm10, 320(%rcx)
+; AVX1-NEXT: vmovaps %ymm7, 128(%rcx)
+; AVX1-NEXT: vmovaps %ymm1, 224(%rcx)
+; AVX1-NEXT: vmovaps %ymm0, 64(%rcx)
+; AVX1-NEXT: vmovaps %ymm6, 288(%rcx)
+; AVX1-NEXT: vmovaps %ymm5, 96(%rcx)
+; AVX1-NEXT: vmovaps %ymm4, 192(%rcx)
+; AVX1-NEXT: vmovaps %ymm3, 352(%rcx)
+; AVX1-NEXT: vmovaps %ymm2, 160(%rcx)
+; AVX1-NEXT: vmovaps %ymm9, 256(%rcx)
+; AVX1-NEXT: vmovaps %ymm8, (%rcx)
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-SLOW-LABEL: store_i32_stride3_vf32:
+; AVX2-SLOW: # %bb.0:
+; AVX2-SLOW-NEXT: subq $40, %rsp
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vmovaps 32(%rsi), %ymm5
+; AVX2-SLOW-NEXT: vmovaps 64(%rsi), %ymm9
+; AVX2-SLOW-NEXT: vmovaps 96(%rsi), %ymm2
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 32(%rdx), %ymm10
+; AVX2-SLOW-NEXT: vmovaps 64(%rdx), %ymm12
+; AVX2-SLOW-NEXT: vmovaps 96(%rdx), %ymm6
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm1 = mem[1,0,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm0[0,0,2,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0],ymm1[1],ymm3[2,3],ymm1[4],ymm3[5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vbroadcastsd (%rdx), %ymm3
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm12[2,1,3,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm7 = ymm9[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[2,2,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0],ymm3[1],ymm7[2,3],ymm3[4],ymm7[5,6],ymm3[7]
+; AVX2-SLOW-NEXT: vbroadcastsd 88(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm7[2],ymm3[3,4],ymm7[5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm10[2,1,3,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm11 = ymm5[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[2,2,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm11[0],ymm7[1],ymm11[2,3],ymm7[4],ymm11[5,6],ymm7[7]
+; AVX2-SLOW-NEXT: vbroadcastsd 56(%rdi), %ymm11
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm11[2],ymm7[3,4],ymm11[5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm11 = ymm6[2,1,3,3]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm13 = ymm2[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[2,2,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm13[0],ymm11[1],ymm13[2,3],ymm11[4],ymm13[5,6],ymm11[7]
+; AVX2-SLOW-NEXT: vbroadcastsd 120(%rdi), %ymm13
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm13[2],ymm11[3,4],ymm13[5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm13 = mem[1,0,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm14 = ymm8[0,0,2,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0],ymm13[1],ymm14[2,3],ymm13[4],ymm14[5,6],ymm13[7]
+; AVX2-SLOW-NEXT: vbroadcastsd 64(%rdx), %ymm14
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2],ymm13[3,4],ymm14[5],ymm13[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm14 = mem[1,0,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm14 = ymm14[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm15 = ymm4[0,0,2,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0],ymm14[1],ymm15[2,3],ymm14[4],ymm15[5,6],ymm14[7]
+; AVX2-SLOW-NEXT: vbroadcastsd 32(%rdx), %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3,4],ymm15[5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm15 = mem[1,0,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm15 = ymm15[0,1,0,1]
+; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %ymm2
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm2[0,0,2,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3],ymm15[4],ymm1[5,6],ymm15[7]
+; AVX2-SLOW-NEXT: vbroadcastsd 96(%rdx), %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm15[2],ymm1[3,4],ymm15[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovaps (%rsi), %ymm15
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm3 = ymm15[1,2,3,3,5,6,7,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[2,2,2,3]
+; AVX2-SLOW-NEXT: vmovaps (%rdx), %ymm0
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm0[2,1,3,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm7[1],ymm3[2,3],ymm7[4],ymm3[5,6],ymm7[7]
+; AVX2-SLOW-NEXT: vbroadcastsd 24(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm7[2],ymm3[3,4],ymm7[5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm8[1,1,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm12[1,1,2,2]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1],ymm8[2,3],ymm7[4],ymm8[5,6],ymm7[7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm8 = ymm9[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm8[2],ymm7[3,4],ymm8[5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[1,1,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm10[1,1,2,2]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0],ymm4[1],ymm8[2,3],ymm4[4],ymm8[5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[1,1,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm6[1,1,2,2]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1],ymm5[2,3],ymm2[4],ymm5[5,6],ymm2[7]
+; AVX2-SLOW-NEXT: vpermilps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm5[2],ymm2[3,4],ymm5[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpermpd $165, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[1,1,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3],ymm5[4],ymm0[5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm15[0,0,3,3,4,4,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm5[2],ymm0[3,4],ymm5[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 320(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 128(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm7, 224(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm3, 64(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm1, 288(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm14, 96(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm13, 192(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm11, 352(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 160(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 256(%rcx)
+; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, (%rcx)
+; AVX2-SLOW-NEXT: addq $40, %rsp
+; AVX2-SLOW-NEXT: vzeroupper
+; AVX2-SLOW-NEXT: retq
+;
+; AVX2-FAST-ALL-LABEL: store_i32_stride3_vf32:
+; AVX2-FAST-ALL: # %bb.0:
+; AVX2-FAST-ALL-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-FAST-ALL-NEXT: vmovaps 32(%rdi), %ymm5
+; AVX2-FAST-ALL-NEXT: vmovaps 64(%rdi), %ymm9
+; AVX2-FAST-ALL-NEXT: vmovaps (%rsi), %ymm1
+; AVX2-FAST-ALL-NEXT: vmovaps 32(%rsi), %ymm7
+; AVX2-FAST-ALL-NEXT: vmovaps 64(%rsi), %ymm11
+; AVX2-FAST-ALL-NEXT: vmovaps 96(%rsi), %ymm6
+; AVX2-FAST-ALL-NEXT: vmovaps 32(%rdx), %ymm12
+; AVX2-FAST-ALL-NEXT: vmovaps 64(%rdx), %ymm4
+; AVX2-FAST-ALL-NEXT: vmovaps 96(%rdx), %ymm15
+; AVX2-FAST-ALL-NEXT: vbroadcastf128 {{.*#+}} ymm10 = [1,0,2,2,1,0,2,2]
+; AVX2-FAST-ALL-NEXT: # ymm10 = mem[0,1,0,1]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm1, %ymm10, %ymm2
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm3 = ymm0[0,0,2,1]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5,6],ymm2[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd (%rdx), %ymm3
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX2-FAST-ALL-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm3 = ymm9[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm8 = ymm4[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0],ymm3[1],ymm8[2,3],ymm3[4],ymm8[5,6],ymm3[7]
+; AVX2-FAST-ALL-NEXT: vpermilps {{.*#+}} ymm8 = ymm11[0,0,3,3,4,4,7,7]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm8[2],ymm3[3,4],ymm8[5],ymm3[6,7]
+; AVX2-FAST-ALL-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm13 = [5,6,5,6,5,6,7,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm11, %ymm13, %ymm8
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[2,1,3,3]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0],ymm4[1],ymm8[2,3],ymm4[4],ymm8[5,6],ymm4[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd 88(%rdi), %ymm8
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm8[2],ymm4[3,4],ymm8[5],ymm4[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm7, %ymm13, %ymm8
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm14 = ymm12[2,1,3,3]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0],ymm14[1],ymm8[2,3],ymm14[4],ymm8[5,6],ymm14[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd 56(%rdi), %ymm14
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm14[2],ymm8[3,4],ymm14[5],ymm8[6,7]
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm14 = ymm5[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0],ymm14[1],ymm12[2,3],ymm14[4],ymm12[5,6],ymm14[7]
+; AVX2-FAST-ALL-NEXT: vpermilps {{.*#+}} ymm14 = ymm7[0,0,3,3,4,4,7,7]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm14[2],ymm12[3,4],ymm14[5],ymm12[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm6, %ymm13, %ymm14
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm2 = ymm15[2,1,3,3]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm14[0],ymm2[1],ymm14[2,3],ymm2[4],ymm14[5,6],ymm2[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd 120(%rdi), %ymm14
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm14 = ymm2[0,1],ymm14[2],ymm2[3,4],ymm14[5],ymm2[6,7]
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm2 = ymm15[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vmovaps 96(%rdi), %ymm15
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm3 = ymm15[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3],ymm3[4],ymm2[5,6],ymm3[7]
+; AVX2-FAST-ALL-NEXT: vpermilps {{.*#+}} ymm3 = ymm6[0,0,3,3,4,4,7,7]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm11, %ymm10, %ymm3
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,0,2,1]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm3 = ymm9[0],ymm3[1],ymm9[2,3],ymm3[4],ymm9[5,6],ymm3[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd 64(%rdx), %ymm9
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm9[2],ymm3[3,4],ymm9[5],ymm3[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm7, %ymm10, %ymm7
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,0,2,1]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2,3],ymm7[4],ymm5[5,6],ymm7[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd 32(%rdx), %ymm7
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm7[2],ymm5[3,4],ymm7[5],ymm5[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm6, %ymm10, %ymm6
+; AVX2-FAST-ALL-NEXT: vmovaps (%rdx), %ymm7
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm9 = ymm15[0,0,2,1]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm6 = ymm9[0],ymm6[1],ymm9[2,3],ymm6[4],ymm9[5,6],ymm6[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd 96(%rdx), %ymm9
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm9[2],ymm6[3,4],ymm9[5],ymm6[6,7]
+; AVX2-FAST-ALL-NEXT: vpermps %ymm1, %ymm13, %ymm9
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm10 = ymm7[2,1,3,3]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1],ymm9[2,3],ymm10[4],ymm9[5,6],ymm10[7]
+; AVX2-FAST-ALL-NEXT: vbroadcastsd 24(%rdi), %ymm10
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm10[2],ymm9[3,4],ymm10[5],ymm9[6,7]
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[1,1,2,2]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0],ymm0[1],ymm7[2,3],ymm0[4],ymm7[5,6],ymm0[7]
+; AVX2-FAST-ALL-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7]
+; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm9, 64(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm6, 288(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm5, 96(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm3, 192(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm2, 320(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm14, 352(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm12, 128(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm8, 160(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm4, 256(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, 224(%rcx)
+; AVX2-FAST-ALL-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-ALL-NEXT: vmovaps %ymm0, (%rcx)
+; AVX2-FAST-ALL-NEXT: vzeroupper
+; AVX2-FAST-ALL-NEXT: retq
+;
+; AVX2-FAST-PERLANE-LABEL: store_i32_stride3_vf32:
+; AVX2-FAST-PERLANE: # %bb.0:
+; AVX2-FAST-PERLANE-NEXT: subq $40, %rsp
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rsi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rsi), %ymm9
+; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rsi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdx), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdx), %ymm12
+; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdx), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm1 = mem[1,0,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm0[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0],ymm1[1],ymm3[2,3],ymm1[4],ymm3[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd (%rdx), %ymm3
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm12[2,1,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm7 = ymm9[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0],ymm3[1],ymm7[2,3],ymm3[4],ymm7[5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 88(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm7[2],ymm3[3,4],ymm7[5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm10[2,1,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm11 = ymm5[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm11[0],ymm7[1],ymm11[2,3],ymm7[4],ymm11[5,6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 56(%rdi), %ymm11
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm11[2],ymm7[3,4],ymm11[5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm11 = ymm6[2,1,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm13 = ymm2[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm13[0],ymm11[1],ymm13[2,3],ymm11[4],ymm13[5,6],ymm11[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 120(%rdi), %ymm13
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm13[2],ymm11[3,4],ymm13[5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm13 = mem[1,0,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm14 = ymm8[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0],ymm13[1],ymm14[2,3],ymm13[4],ymm14[5,6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 64(%rdx), %ymm14
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2],ymm13[3,4],ymm14[5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm14 = mem[1,0,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm14 = ymm14[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm15 = ymm4[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0],ymm14[1],ymm15[2,3],ymm14[4],ymm15[5,6],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 32(%rdx), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3,4],ymm15[5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm15 = mem[1,0,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm15 = ymm15[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm2[0,0,2,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3],ymm15[4],ymm1[5,6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 96(%rdx), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm15[2],ymm1[3,4],ymm15[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rsi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm3 = ymm15[1,2,3,3,5,6,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[2,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdx), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm0[2,1,3,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm7[1],ymm3[2,3],ymm7[4],ymm3[5,6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 24(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm7[2],ymm3[3,4],ymm7[5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm8[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm12[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0],ymm7[1],ymm8[2,3],ymm7[4],ymm8[5,6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm8 = ymm9[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm8[2],ymm7[3,4],ymm8[5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm10[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0],ymm4[1],ymm8[2,3],ymm4[4],ymm8[5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm6[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0],ymm2[1],ymm5[2,3],ymm2[4],ymm5[5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm5[2],ymm2[3,4],ymm5[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd $165, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3],ymm5[4],ymm0[5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm15[0,0,3,3,4,4,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm5[2],ymm0[3,4],ymm5[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 320(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 128(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm7, 224(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 64(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 288(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm14, 96(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm13, 192(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm11, 352(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 160(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 256(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, (%rcx)
+; AVX2-FAST-PERLANE-NEXT: addq $40, %rsp
+; AVX2-FAST-PERLANE-NEXT: vzeroupper
+; AVX2-FAST-PERLANE-NEXT: retq
+;
+; AVX512-LABEL: store_i32_stride3_vf32:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm0
+; AVX512-NEXT: vmovdqu64 64(%rdi), %zmm1
+; AVX512-NEXT: vmovdqu64 (%rsi), %zmm2
+; AVX512-NEXT: vmovdqu64 64(%rsi), %zmm3
+; AVX512-NEXT: vmovdqu64 (%rdx), %zmm4
+; AVX512-NEXT: vmovdqu64 64(%rdx), %zmm5
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm6 = <0,16,u,1,17,u,2,18,u,3,19,u,4,20,u,5>
+; AVX512-NEXT: vmovdqa64 %zmm0, %zmm7
+; AVX512-NEXT: vpermt2d %zmm2, %zmm6, %zmm7
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm8 = [0,1,16,3,4,17,6,7,18,9,10,19,12,13,20,15]
+; AVX512-NEXT: vpermt2d %zmm4, %zmm8, %zmm7
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm9 = <10,27,u,11,28,u,12,29,u,13,30,u,14,31,u,15>
+; AVX512-NEXT: vmovdqa64 %zmm5, %zmm10
+; AVX512-NEXT: vpermt2d %zmm1, %zmm9, %zmm10
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm11 = [0,1,27,3,4,28,6,7,29,9,10,30,12,13,31,15]
+; AVX512-NEXT: vpermt2d %zmm3, %zmm11, %zmm10
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm12 = <5,21,u,6,22,u,7,23,u,8,24,u,9,25,u,10>
+; AVX512-NEXT: vmovdqa64 %zmm3, %zmm13
+; AVX512-NEXT: vpermt2d %zmm5, %zmm12, %zmm13
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm14 = [0,1,22,3,4,23,6,7,24,9,10,25,12,13,26,15]
+; AVX512-NEXT: vpermt2d %zmm1, %zmm14, %zmm13
+; AVX512-NEXT: vpermt2d %zmm3, %zmm6, %zmm1
+; AVX512-NEXT: vpermt2d %zmm5, %zmm8, %zmm1
+; AVX512-NEXT: vpermi2d %zmm0, %zmm4, %zmm9
+; AVX512-NEXT: vpermt2d %zmm2, %zmm11, %zmm9
+; AVX512-NEXT: vpermt2d %zmm4, %zmm12, %zmm2
+; AVX512-NEXT: vpermt2d %zmm0, %zmm14, %zmm2
+; AVX512-NEXT: vmovdqu64 %zmm2, 64(%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm9, 128(%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm1, 192(%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm13, 256(%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm10, 320(%rcx)
+; AVX512-NEXT: vmovdqu64 %zmm7, (%rcx)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %in.vec0 = load <32 x i32>, <32 x i32>* %in.vecptr0, align 32
+ %in.vec1 = load <32 x i32>, <32 x i32>* %in.vecptr1, align 32
+ %in.vec2 = load <32 x i32>, <32 x i32>* %in.vecptr2, align 32
+
+ %concat01 = shufflevector <32 x i32> %in.vec0, <32 x i32> %in.vec1, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+ %concat2u = shufflevector <32 x i32> %in.vec2, <32 x i32> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+ %concat012 = shufflevector <64 x i32> %concat01, <64 x i32> %concat2u, <96 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64, i32 65, i32 66, i32 67, i32 68, i32 69, i32 70, i32 71, i32 72, i32 73, i32 74, i32 75, i32 76, i32 77, i32 78, i32 79, i32 80, i32 81, i32 82, i32 83, i32 84, i32 85, i32 86, i32 87, i32 88, i32 89, i32 90, i32 91, i32 92, i32 93, i32 94, i32 95>
+ %interleaved.vec = shufflevector <96 x i32> %concat012, <96 x i32> poison, <96 x i32> <i32 0, i32 32, i32 64, i32 1, i32 33, i32 65, i32 2, i32 34, i32 66, i32 3, i32 35, i32 67, i32 4, i32 36, i32 68, i32 5, i32 37, i32 69, i32 6, i32 38, i32 70, i32 7, i32 39, i32 71, i32 8, i32 40, i32 72, i32 9, i32 41, i32 73, i32 10, i32 42, i32 74, i32 11, i32 43, i32 75, i32 12, i32 44, i32 76, i32 13, i32 45, i32 77, i32 14, i32 46, i32 78, i32 15, i32 47, i32 79, i32 16, i32 48, i32 80, i32 17, i32 49, i32 81, i32 18, i32 50, i32 82, i32 19, i32 51, i32 83, i32 20, i32 52, i32 84, i32 21, i32 53, i32 85, i32 22, i32 54, i32 86, i32 23, i32 55, i32 87, i32 24, i32 56, i32 88, i32 25, i32 57, i32 89, i32 26, i32 58, i32 90, i32 27, i32 59, i32 91, i32 28, i32 60, i32 92, i32 29, i32 61, i32 93, i32 30, i32 62, i32 94, i32 31, i32 63, i32 95>
+
+ store <96 x i32> %interleaved.vec, <96 x i32>* %out.vec, align 32
+
+ ret void
+}
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-4.ll
index 71ec3e1d81bc6..019068553ca47 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-4.ll
@@ -690,3 +690,678 @@ define void @store_i32_stride4_vf16(<16 x i32>* %in.vecptr0, <16 x i32>* %in.vec
ret void
}
+
+define void @store_i32_stride4_vf32(<32 x i32>* %in.vecptr0, <32 x i32>* %in.vecptr1, <32 x i32>* %in.vecptr2, <32 x i32>* %in.vecptr3, <128 x i32>* %out.vec) nounwind {
+; SSE-LABEL: store_i32_stride4_vf32:
+; SSE: # %bb.0:
+; SSE-NEXT: subq $184, %rsp
+; SSE-NEXT: movaps (%rdi), %xmm14
+; SSE-NEXT: movaps 16(%rdi), %xmm11
+; SSE-NEXT: movaps 32(%rdi), %xmm12
+; SSE-NEXT: movaps 48(%rdi), %xmm13
+; SSE-NEXT: movaps (%rsi), %xmm15
+; SSE-NEXT: movaps 16(%rsi), %xmm9
+; SSE-NEXT: movaps 32(%rsi), %xmm8
+; SSE-NEXT: movaps (%rdx), %xmm2
+; SSE-NEXT: movaps 16(%rdx), %xmm4
+; SSE-NEXT: movaps 32(%rdx), %xmm1
+; SSE-NEXT: movaps (%rcx), %xmm7
+; SSE-NEXT: movaps 16(%rcx), %xmm0
+; SSE-NEXT: movaps 32(%rcx), %xmm10
+; SSE-NEXT: movaps %xmm2, %xmm6
+; SSE-NEXT: unpcklps {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
+; SSE-NEXT: movaps %xmm14, %xmm5
+; SSE-NEXT: unpcklps {{.*#+}} xmm5 = xmm5[0],xmm15[0],xmm5[1],xmm15[1]
+; SSE-NEXT: movaps %xmm5, %xmm3
+; SSE-NEXT: movlhps {{.*#+}} xmm3 = xmm3[0],xmm6[0]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm6[1]
+; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhps {{.*#+}} xmm2 = xmm2[2],xmm7[2],xmm2[3],xmm7[3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm14 = xmm14[2],xmm15[2],xmm14[3],xmm15[3]
+; SSE-NEXT: movaps %xmm14, %xmm5
+; SSE-NEXT: movlhps {{.*#+}} xmm5 = xmm5[0],xmm2[0]
+; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm2[1]
+; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm4, %xmm2
+; SSE-NEXT: unpcklps {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: movaps %xmm11, %xmm3
+; SSE-NEXT: unpcklps {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1]
+; SSE-NEXT: movaps %xmm3, %xmm5
+; SSE-NEXT: movlhps {{.*#+}} xmm5 = xmm5[0],xmm2[0]
+; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm2[1]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhps {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm11 = xmm11[2],xmm9[2],xmm11[3],xmm9[3]
+; SSE-NEXT: movaps %xmm11, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm4[1]
+; SSE-NEXT: movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm1, %xmm0
+; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT: movaps %xmm12, %xmm2
+; SSE-NEXT: unpcklps {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1]
+; SSE-NEXT: movaps %xmm2, %xmm3
+; SSE-NEXT: movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 48(%rdx), %xmm0
+; SSE-NEXT: unpckhps {{.*#+}} xmm1 = xmm1[2],xmm10[2],xmm1[3],xmm10[3]
+; SSE-NEXT: movaps 48(%rcx), %xmm2
+; SSE-NEXT: unpckhps {{.*#+}} xmm12 = xmm12[2],xmm8[2],xmm12[3],xmm8[3]
+; SSE-NEXT: movaps %xmm12, %xmm3
+; SSE-NEXT: movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
+; SSE-NEXT: movaps %xmm3, (%rsp) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm1[1]
+; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm0, %xmm1
+; SSE-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT: movaps 48(%rsi), %xmm3
+; SSE-NEXT: movaps %xmm13, %xmm15
+; SSE-NEXT: unpcklps {{.*#+}} xmm15 = xmm15[0],xmm3[0],xmm15[1],xmm3[1]
+; SSE-NEXT: movaps %xmm15, %xmm4
+; SSE-NEXT: movlhps {{.*#+}} xmm4 = xmm4[0],xmm1[0]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm1[1]
+; SSE-NEXT: unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm13 = xmm13[2],xmm3[2],xmm13[3],xmm3[3]
+; SSE-NEXT: movaps %xmm13, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm0[1]
+; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 64(%rdx), %xmm0
+; SSE-NEXT: movaps 64(%rcx), %xmm2
+; SSE-NEXT: movaps %xmm0, %xmm3
+; SSE-NEXT: unpcklps {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSE-NEXT: movaps 64(%rdi), %xmm10
+; SSE-NEXT: movaps 64(%rsi), %xmm4
+; SSE-NEXT: movaps %xmm10, %xmm12
+; SSE-NEXT: unpcklps {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1]
+; SSE-NEXT: movaps %xmm12, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm3[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm3[1]
+; SSE-NEXT: unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm10 = xmm10[2],xmm4[2],xmm10[3],xmm4[3]
+; SSE-NEXT: movaps %xmm10, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm0[1]
+; SSE-NEXT: movaps 80(%rdx), %xmm0
+; SSE-NEXT: movaps 80(%rcx), %xmm3
+; SSE-NEXT: movaps %xmm0, %xmm4
+; SSE-NEXT: unpcklps {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT: movaps 80(%rdi), %xmm5
+; SSE-NEXT: movaps 80(%rsi), %xmm6
+; SSE-NEXT: movaps %xmm5, %xmm14
+; SSE-NEXT: unpcklps {{.*#+}} xmm14 = xmm14[0],xmm6[0],xmm14[1],xmm6[1]
+; SSE-NEXT: movaps %xmm14, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm4[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm4[1]
+; SSE-NEXT: unpckhps {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm5 = xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; SSE-NEXT: movaps %xmm5, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm0[1]
+; SSE-NEXT: movaps 96(%rdx), %xmm1
+; SSE-NEXT: movaps 96(%rcx), %xmm6
+; SSE-NEXT: movaps %xmm1, %xmm0
+; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; SSE-NEXT: movaps 96(%rdi), %xmm3
+; SSE-NEXT: movaps 96(%rsi), %xmm7
+; SSE-NEXT: movaps %xmm3, %xmm4
+; SSE-NEXT: unpcklps {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
+; SSE-NEXT: movaps %xmm4, %xmm13
+; SSE-NEXT: movlhps {{.*#+}} xmm13 = xmm13[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm0[1]
+; SSE-NEXT: unpckhps {{.*#+}} xmm1 = xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm3 = xmm3[2],xmm7[2],xmm3[3],xmm7[3]
+; SSE-NEXT: movaps %xmm3, %xmm11
+; SSE-NEXT: movlhps {{.*#+}} xmm11 = xmm11[0],xmm1[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm1[1]
+; SSE-NEXT: movaps 112(%rdx), %xmm2
+; SSE-NEXT: movaps 112(%rcx), %xmm9
+; SSE-NEXT: movaps %xmm2, %xmm7
+; SSE-NEXT: unpcklps {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1]
+; SSE-NEXT: movaps 112(%rdi), %xmm0
+; SSE-NEXT: movaps 112(%rsi), %xmm8
+; SSE-NEXT: movaps %xmm0, %xmm1
+; SSE-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
+; SSE-NEXT: movaps %xmm1, %xmm6
+; SSE-NEXT: movlhps {{.*#+}} xmm6 = xmm6[0],xmm7[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm7[1]
+; SSE-NEXT: unpckhps {{.*#+}} xmm2 = xmm2[2],xmm9[2],xmm2[3],xmm9[3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm0 = xmm0[2],xmm8[2],xmm0[3],xmm8[3]
+; SSE-NEXT: movaps %xmm0, %xmm7
+; SSE-NEXT: movlhps {{.*#+}} xmm7 = xmm7[0],xmm2[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSE-NEXT: movaps %xmm0, 496(%r8)
+; SSE-NEXT: movaps %xmm7, 480(%r8)
+; SSE-NEXT: movaps %xmm1, 464(%r8)
+; SSE-NEXT: movaps %xmm6, 448(%r8)
+; SSE-NEXT: movaps %xmm3, 432(%r8)
+; SSE-NEXT: movaps %xmm11, 416(%r8)
+; SSE-NEXT: movaps %xmm4, 400(%r8)
+; SSE-NEXT: movaps %xmm13, 384(%r8)
+; SSE-NEXT: movaps %xmm5, 368(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 352(%r8)
+; SSE-NEXT: movaps %xmm14, 336(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 320(%r8)
+; SSE-NEXT: movaps %xmm10, 304(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 288(%r8)
+; SSE-NEXT: movaps %xmm12, 272(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 256(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 240(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 224(%r8)
+; SSE-NEXT: movaps %xmm15, 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 (%rsp), %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: addq $184, %rsp
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: store_i32_stride4_vf32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: subq $456, %rsp # imm = 0x1C8
+; AVX1-NEXT: vmovaps 16(%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 16(%rsi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = xmm2[1],xmm1[1],zero,zero
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT: vmovaps 16(%rcx), %xmm3
+; AVX1-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 16(%rdx), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm2 = xmm3[0],xmm2[0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 64(%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 64(%rsi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = xmm2[1],xmm1[1],zero,zero
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT: vmovaps 64(%rcx), %xmm3
+; AVX1-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 64(%rdx), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm2 = xmm3[0],xmm2[0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 80(%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 80(%rsi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = xmm2[1],xmm1[1],zero,zero
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT: vmovaps 80(%rcx), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 80(%rdx), %xmm7
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm7[0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm2 = xmm7[0],xmm2[0],xmm7[1],xmm2[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 32(%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 32(%rsi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = xmm2[1],xmm1[1],zero,zero
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT: vmovaps 32(%rcx), %xmm13
+; AVX1-NEXT: vmovaps 32(%rdx), %xmm10
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm1 = xmm13[0],xmm10[0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm2 = xmm10[0],xmm13[0],xmm10[1],xmm13[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 48(%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 48(%rsi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = xmm2[1],xmm1[1],zero,zero
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT: vmovaps 48(%rcx), %xmm11
+; AVX1-NEXT: vmovaps 48(%rdx), %xmm9
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm1 = xmm11[0],xmm9[0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm2 = xmm9[0],xmm11[0],xmm9[1],xmm11[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 96(%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 96(%rsi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, (%rsp) # 16-byte Spill
+; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = xmm2[1],xmm1[1],zero,zero
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT: vmovaps 96(%rcx), %xmm12
+; AVX1-NEXT: vmovaps 96(%rdx), %xmm5
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm1 = xmm12[0],xmm5[0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm2 = xmm5[0],xmm12[0],xmm5[1],xmm12[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 112(%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 112(%rsi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = xmm2[1],xmm1[1],zero,zero
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT: vmovaps 112(%rcx), %xmm4
+; AVX1-NEXT: vmovaps 112(%rdx), %xmm3
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm1 = xmm4[0],xmm3[0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm6 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm1, %ymm1
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps (%rdi), %xmm2
+; AVX1-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps (%rsi), %xmm1
+; AVX1-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = xmm2[1],xmm1[1],zero,zero
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT: vmovaps (%rcx), %xmm8
+; AVX1-NEXT: vmovaps (%rdx), %xmm2
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm15 = xmm8[0],xmm2[0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm15 = xmm15[0,1,2,0]
+; AVX1-NEXT: vunpcklps {{.*#+}} xmm14 = xmm2[0],xmm8[0],xmm2[1],xmm8[1]
+; AVX1-NEXT: vinsertf128 $1, %xmm14, %ymm15, %ymm14
+; AVX1-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0,1],ymm14[2,3],ymm0[4,5],ymm14[6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; AVX1-NEXT: vinsertps {{.*#+}} xmm14 = zero,zero,xmm1[2],xmm6[2]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm14, %ymm1
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm14 = xmm0[2],xmm6[2],xmm0[3],xmm6[3]
+; AVX1-NEXT: vshufps {{.*#+}} xmm6 = xmm6[3,0],xmm0[3,0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[2,0,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm14, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1],ymm1[2,3],ymm6[4,5],ymm1[6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm0 = xmm7[2],xmm1[2],xmm7[3],xmm1[3]
+; AVX1-NEXT: vinsertps {{.*#+}} xmm6 = zero,zero,xmm7[2],xmm1[2]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm6, %ymm0
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm6 = xmm7[2],xmm1[2],xmm7[3],xmm1[3]
+; AVX1-NEXT: vshufps {{.*#+}} xmm7 = xmm1[3,0],xmm7[3,0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[2,0,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm14 = ymm6[0,1],ymm0[2,3],ymm6[4,5],ymm0[6,7]
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm6 = xmm10[2],xmm13[2],xmm10[3],xmm13[3]
+; AVX1-NEXT: vinsertps {{.*#+}} xmm7 = zero,zero,xmm10[2],xmm13[2]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm7, %ymm6
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm7 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm1[3,0],xmm0[3,0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,0,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm7, %ymm0
+; AVX1-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm0 = xmm9[2],xmm11[2],xmm9[3],xmm11[3]
+; AVX1-NEXT: vinsertps {{.*#+}} xmm6 = zero,zero,xmm9[2],xmm11[2]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm6, %ymm0
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm6 = xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; AVX1-NEXT: vshufps {{.*#+}} xmm7 = xmm7[3,0],xmm1[3,0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm7 = xmm7[2,0,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
+; AVX1-NEXT: vblendps {{.*#+}} ymm9 = ymm6[0,1],ymm0[2,3],ymm6[4,5],ymm0[6,7]
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm0 = xmm5[2],xmm12[2],xmm5[3],xmm12[3]
+; AVX1-NEXT: vinsertps {{.*#+}} xmm5 = zero,zero,xmm5[2],xmm12[2]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm5, %ymm0
+; AVX1-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm5 = xmm6[2],xmm1[2],xmm6[3],xmm1[3]
+; AVX1-NEXT: vshufps {{.*#+}} xmm6 = xmm1[3,0],xmm6[3,0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[2,0,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm5, %ymm5
+; AVX1-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm0[2,3],ymm5[4,5],ymm0[6,7]
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm0 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; AVX1-NEXT: vinsertps {{.*#+}} xmm3 = zero,zero,xmm3[2],xmm4[2]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm3 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; AVX1-NEXT: vshufps {{.*#+}} xmm4 = xmm1[3,0],xmm4[3,0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm4 = xmm4[2,0,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5],ymm0[6,7]
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm3 = xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; AVX1-NEXT: vinsertps {{.*#+}} xmm1 = zero,zero,xmm2[2],xmm8[2]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm2 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm4[3,0],xmm3[3,0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm3 = xmm3[2,0,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
+; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm2 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; AVX1-NEXT: vinsertps {{.*#+}} xmm3 = zero,zero,xmm3[2],xmm4[2]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-NEXT: vunpckhps {{.*#+}} xmm3 = xmm6[2],xmm4[2],xmm6[3],xmm4[3]
+; AVX1-NEXT: vshufps {{.*#+}} xmm4 = xmm4[3,0],xmm6[3,0]
+; AVX1-NEXT: vpermilps {{.*#+}} xmm4 = xmm4[2,0,2,3]
+; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-NEXT: vmovaps %ymm2, 96(%r8)
+; AVX1-NEXT: vmovaps %ymm1, 32(%r8)
+; AVX1-NEXT: vmovaps %ymm0, 480(%r8)
+; AVX1-NEXT: vmovaps %ymm5, 416(%r8)
+; AVX1-NEXT: vmovaps %ymm9, 224(%r8)
+; AVX1-NEXT: vmovaps %ymm10, 160(%r8)
+; AVX1-NEXT: vmovaps %ymm14, 352(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 288(%r8)
+; AVX1-NEXT: vmovaps %ymm15, (%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 448(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 384(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 192(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 128(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 320(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 256(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 64(%r8)
+; AVX1-NEXT: addq $456, %rsp # imm = 0x1C8
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: store_i32_stride4_vf32:
+; AVX2: # %bb.0:
+; AVX2-NEXT: subq $168, %rsp
+; AVX2-NEXT: vmovaps (%rdi), %ymm0
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps (%rsi), %ymm0
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 64(%rdx), %ymm6
+; AVX2-NEXT: vmovaps (%rdx), %ymm2
+; AVX2-NEXT: vmovaps (%rcx), %xmm13
+; AVX2-NEXT: vmovaps 32(%rcx), %xmm10
+; AVX2-NEXT: vmovaps 64(%rcx), %xmm11
+; AVX2-NEXT: vmovaps (%rdx), %xmm14
+; AVX2-NEXT: vmovaps 32(%rdx), %xmm12
+; AVX2-NEXT: vmovaps 64(%rdx), %xmm3
+; AVX2-NEXT: vunpckhps {{.*#+}} xmm8 = xmm3[2],xmm11[2],xmm3[3],xmm11[3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,0,2,1]
+; AVX2-NEXT: vmovaps 32(%rsi), %xmm4
+; AVX2-NEXT: vmovaps 64(%rsi), %xmm7
+; AVX2-NEXT: vmovaps 32(%rdi), %xmm0
+; AVX2-NEXT: vmovaps 64(%rdi), %xmm5
+; AVX2-NEXT: vunpckhps {{.*#+}} xmm9 = xmm5[2],xmm7[2],xmm5[3],xmm7[3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm8[2,3],ymm9[4,5],ymm8[6,7]
+; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm3 = xmm3[0],xmm11[0],xmm3[1],xmm11[1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm3[2,3],ymm5[4,5],ymm3[6,7]
+; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vunpckhps {{.*#+}} xmm3 = xmm12[2],xmm10[2],xmm12[3],xmm10[3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
+; AVX2-NEXT: vunpckhps {{.*#+}} xmm5 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm3[2,3],ymm5[4,5],ymm3[6,7]
+; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm3 = xmm12[0],xmm10[0],xmm12[1],xmm10[1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3],ymm0[4,5],ymm3[6,7]
+; AVX2-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-NEXT: vmovaps 96(%rcx), %xmm10
+; AVX2-NEXT: vmovaps 96(%rdx), %xmm3
+; AVX2-NEXT: vunpckhps {{.*#+}} xmm4 = xmm3[2],xmm10[2],xmm3[3],xmm10[3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm8 = ymm4[0,0,2,1]
+; AVX2-NEXT: vmovaps 96(%rsi), %xmm4
+; AVX2-NEXT: vmovaps 96(%rdi), %xmm0
+; AVX2-NEXT: vunpckhps {{.*#+}} xmm12 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1],ymm8[2,3],ymm12[4,5],ymm8[6,7]
+; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps (%rsi), %xmm1
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1]
+; AVX2-NEXT: vmovaps (%rdi), %xmm10
+; AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3],ymm0[4,5],ymm3[6,7]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vunpckhps {{.*#+}} xmm3 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
+; AVX2-NEXT: vunpckhps {{.*#+}} xmm4 = xmm10[2],xmm1[2],xmm10[3],xmm1[3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm3[2,3],ymm4[4,5],ymm3[6,7]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 64(%rcx), %ymm0
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm3 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
+; AVX2-NEXT: vmovaps (%rcx), %ymm13
+; AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
+; AVX2-NEXT: vunpcklps {{.*#+}} xmm1 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm15 = ymm1[0,1],ymm3[2,3],ymm1[4,5],ymm3[6,7]
+; AVX2-NEXT: vmovaps %ymm2, %ymm5
+; AVX2-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm13[0],ymm2[1],ymm13[1],ymm2[4],ymm13[4],ymm2[5],ymm13[5]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-NEXT: vunpcklps {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm12 = ymm4[0,1],ymm1[2,3],ymm4[4,5],ymm1[6,7]
+; AVX2-NEXT: vunpckhps {{.*#+}} ymm1 = ymm6[2],ymm0[2],ymm6[3],ymm0[3],ymm6[6],ymm0[6],ymm6[7],ymm0[7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-NEXT: vmovaps 64(%rdi), %ymm10
+; AVX2-NEXT: vmovaps 64(%rsi), %ymm14
+; AVX2-NEXT: vunpckhps {{.*#+}} ymm9 = ymm10[2],ymm14[2],ymm10[3],ymm14[3],ymm10[6],ymm14[6],ymm10[7],ymm14[7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm1[2,3],ymm9[4,5],ymm1[6,7]
+; AVX2-NEXT: vunpcklps {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[1],ymm0[1],ymm6[4],ymm0[4],ymm6[5],ymm0[5]
+; AVX2-NEXT: vunpcklps {{.*#+}} ymm6 = ymm10[0],ymm14[0],ymm10[1],ymm14[1],ymm10[4],ymm14[4],ymm10[5],ymm14[5]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1],ymm0[2,3],ymm6[4,5],ymm0[6,7]
+; AVX2-NEXT: vmovaps 32(%rdx), %ymm6
+; AVX2-NEXT: vmovaps 32(%rcx), %ymm9
+; AVX2-NEXT: vunpckhps {{.*#+}} ymm10 = ymm6[2],ymm9[2],ymm6[3],ymm9[3],ymm6[6],ymm9[6],ymm6[7],ymm9[7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,2,2,3]
+; AVX2-NEXT: vmovaps 32(%rdi), %ymm14
+; AVX2-NEXT: vmovaps 32(%rsi), %ymm0
+; AVX2-NEXT: vunpckhps {{.*#+}} ymm11 = ymm14[2],ymm0[2],ymm14[3],ymm0[3],ymm14[6],ymm0[6],ymm14[7],ymm0[7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7]
+; AVX2-NEXT: vunpcklps {{.*#+}} ymm6 = ymm6[0],ymm9[0],ymm6[1],ymm9[1],ymm6[4],ymm9[4],ymm6[5],ymm9[5]
+; AVX2-NEXT: vunpcklps {{.*#+}} ymm0 = ymm14[0],ymm0[0],ymm14[1],ymm0[1],ymm14[4],ymm0[4],ymm14[5],ymm0[5]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,2,3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
+; AVX2-NEXT: vmovaps 96(%rdx), %ymm6
+; AVX2-NEXT: vmovaps 96(%rcx), %ymm9
+; AVX2-NEXT: vunpckhps {{.*#+}} ymm11 = ymm6[2],ymm9[2],ymm6[3],ymm9[3],ymm6[6],ymm9[6],ymm6[7],ymm9[7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[0,2,2,3]
+; AVX2-NEXT: vmovaps 96(%rdi), %ymm14
+; AVX2-NEXT: vmovaps 96(%rsi), %ymm0
+; AVX2-NEXT: vunpckhps {{.*#+}} ymm8 = ymm14[2],ymm0[2],ymm14[3],ymm0[3],ymm14[6],ymm0[6],ymm14[7],ymm0[7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm11[2,3],ymm8[4,5],ymm11[6,7]
+; AVX2-NEXT: vunpcklps {{.*#+}} ymm6 = ymm6[0],ymm9[0],ymm6[1],ymm9[1],ymm6[4],ymm9[4],ymm6[5],ymm9[5]
+; AVX2-NEXT: vunpcklps {{.*#+}} ymm0 = ymm14[0],ymm0[0],ymm14[1],ymm0[1],ymm14[4],ymm0[4],ymm14[5],ymm0[5]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,2,3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
+; AVX2-NEXT: vunpckhps {{.*#+}} ymm6 = ymm5[2],ymm13[2],ymm5[3],ymm13[3],ymm5[6],ymm13[6],ymm5[7],ymm13[7]
+; AVX2-NEXT: vunpckhps {{.*#+}} ymm9 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,2,3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm6 = ymm9[0,1],ymm6[2,3],ymm9[4,5],ymm6[6,7]
+; AVX2-NEXT: vmovaps %ymm6, 96(%r8)
+; AVX2-NEXT: vmovaps %ymm0, 448(%r8)
+; AVX2-NEXT: vmovaps %ymm8, 480(%r8)
+; AVX2-NEXT: vmovaps %ymm1, 192(%r8)
+; AVX2-NEXT: vmovaps %ymm10, 224(%r8)
+; AVX2-NEXT: vmovaps %ymm4, 320(%r8)
+; AVX2-NEXT: vmovaps %ymm7, 352(%r8)
+; AVX2-NEXT: vmovaps %ymm12, 64(%r8)
+; AVX2-NEXT: vmovaps %ymm15, (%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 32(%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 384(%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 416(%r8)
+; AVX2-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 128(%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 160(%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 256(%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 288(%r8)
+; AVX2-NEXT: addq $168, %rsp
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+;
+; AVX512-LABEL: store_i32_stride4_vf32:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm0
+; AVX512-NEXT: vmovdqu64 64(%rdi), %zmm1
+; AVX512-NEXT: vmovdqu64 (%rsi), %zmm2
+; AVX512-NEXT: vmovdqu64 64(%rsi), %zmm3
+; AVX512-NEXT: vmovdqu64 (%rdx), %zmm4
+; AVX512-NEXT: vmovdqu64 64(%rdx), %zmm5
+; AVX512-NEXT: vmovdqu64 (%rcx), %zmm6
+; AVX512-NEXT: vmovdqu64 64(%rcx), %zmm7
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm8 = <u,u,4,20,u,u,5,21,u,u,6,22,u,u,7,23>
+; AVX512-NEXT: vmovdqa64 %zmm4, %zmm9
+; AVX512-NEXT: vpermt2d %zmm6, %zmm8, %zmm9
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm10 = <4,20,u,u,5,21,u,u,6,22,u,u,7,23,u,u>
+; AVX512-NEXT: vmovdqa64 %zmm0, %zmm11
+; AVX512-NEXT: vpermt2d %zmm2, %zmm10, %zmm11
+; AVX512-NEXT: movb $-86, %al
+; AVX512-NEXT: kmovd %eax, %k1
+; AVX512-NEXT: vmovdqa64 %zmm9, %zmm11 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm9 = <u,u,0,16,u,u,1,17,u,u,2,18,u,u,3,19>
+; AVX512-NEXT: vmovdqa64 %zmm4, %zmm12
+; AVX512-NEXT: vpermt2d %zmm6, %zmm9, %zmm12
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm13 = <0,16,u,u,1,17,u,u,2,18,u,u,3,19,u,u>
+; AVX512-NEXT: vmovdqa64 %zmm0, %zmm14
+; AVX512-NEXT: vpermt2d %zmm2, %zmm13, %zmm14
+; AVX512-NEXT: vmovdqa64 %zmm12, %zmm14 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm12 = <u,u,12,28,u,u,13,29,u,u,14,30,u,u,15,31>
+; AVX512-NEXT: vmovdqa64 %zmm4, %zmm15
+; AVX512-NEXT: vpermt2d %zmm6, %zmm12, %zmm15
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm16 = <12,28,u,u,13,29,u,u,14,30,u,u,15,31,u,u>
+; AVX512-NEXT: vmovdqa64 %zmm0, %zmm17
+; AVX512-NEXT: vpermt2d %zmm2, %zmm16, %zmm17
+; AVX512-NEXT: vmovdqa64 %zmm15, %zmm17 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm15 = <u,u,8,24,u,u,9,25,u,u,10,26,u,u,11,27>
+; AVX512-NEXT: vpermt2d %zmm6, %zmm15, %zmm4
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm6 = <8,24,u,u,9,25,u,u,10,26,u,u,11,27,u,u>
+; AVX512-NEXT: vpermt2d %zmm2, %zmm6, %zmm0
+; AVX512-NEXT: vmovdqa64 %zmm4, %zmm0 {%k1}
+; AVX512-NEXT: vpermi2d %zmm7, %zmm5, %zmm8
+; AVX512-NEXT: vpermi2d %zmm3, %zmm1, %zmm10
+; AVX512-NEXT: vmovdqa64 %zmm8, %zmm10 {%k1}
+; AVX512-NEXT: vpermi2d %zmm7, %zmm5, %zmm9
+; AVX512-NEXT: vpermi2d %zmm3, %zmm1, %zmm13
+; AVX512-NEXT: vmovdqa64 %zmm9, %zmm13 {%k1}
+; AVX512-NEXT: vpermi2d %zmm7, %zmm5, %zmm12
+; AVX512-NEXT: vpermi2d %zmm3, %zmm1, %zmm16
+; AVX512-NEXT: vmovdqa64 %zmm12, %zmm16 {%k1}
+; AVX512-NEXT: vpermt2d %zmm7, %zmm15, %zmm5
+; AVX512-NEXT: vpermt2d %zmm3, %zmm6, %zmm1
+; AVX512-NEXT: vmovdqa64 %zmm5, %zmm1 {%k1}
+; AVX512-NEXT: vmovdqu64 %zmm1, 384(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm16, 448(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm13, 256(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm10, 320(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm0, 128(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm17, 192(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm14, (%r8)
+; AVX512-NEXT: vmovdqu64 %zmm11, 64(%r8)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %in.vec0 = load <32 x i32>, <32 x i32>* %in.vecptr0, align 32
+ %in.vec1 = load <32 x i32>, <32 x i32>* %in.vecptr1, align 32
+ %in.vec2 = load <32 x i32>, <32 x i32>* %in.vecptr2, align 32
+ %in.vec3 = load <32 x i32>, <32 x i32>* %in.vecptr3, align 32
+
+ %concat01 = shufflevector <32 x i32> %in.vec0, <32 x i32> %in.vec1, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+ %concat23 = shufflevector <32 x i32> %in.vec2, <32 x i32> %in.vec3, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+ %concat0123 = shufflevector <64 x i32> %concat01, <64 x i32> %concat23, <128 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64, i32 65, i32 66, i32 67, i32 68, i32 69, i32 70, i32 71, i32 72, i32 73, i32 74, i32 75, i32 76, i32 77, i32 78, i32 79, i32 80, i32 81, i32 82, i32 83, i32 84, i32 85, i32 86, i32 87, i32 88, i32 89, i32 90, i32 91, i32 92, i32 93, i32 94, i32 95, i32 96, i32 97, i32 98, i32 99, i32 100, i32 101, i32 102, i32 103, i32 104, i32 105, i32 106, i32 107, i32 108, i32 109, i32 110, i32 111, i32 112, i32 113, i32 114, i32 115, i32 116, i32 117, i32 118, i32 119, i32 120, i32 121, i32 122, i32 123, i32 124, i32 125, i32 126, i32 127>
+ %interleaved.vec = shufflevector <128 x i32> %concat0123, <128 x i32> poison, <128 x i32> <i32 0, i32 32, i32 64, i32 96, i32 1, i32 33, i32 65, i32 97, i32 2, i32 34, i32 66, i32 98, i32 3, i32 35, i32 67, i32 99, i32 4, i32 36, i32 68, i32 100, i32 5, i32 37, i32 69, i32 101, i32 6, i32 38, i32 70, i32 102, i32 7, i32 39, i32 71, i32 103, i32 8, i32 40, i32 72, i32 104, i32 9, i32 41, i32 73, i32 105, i32 10, i32 42, i32 74, i32 106, i32 11, i32 43, i32 75, i32 107, i32 12, i32 44, i32 76, i32 108, i32 13, i32 45, i32 77, i32 109, i32 14, i32 46, i32 78, i32 110, i32 15, i32 47, i32 79, i32 111, i32 16, i32 48, i32 80, i32 112, i32 17, i32 49, i32 81, i32 113, i32 18, i32 50, i32 82, i32 114, i32 19, i32 51, i32 83, i32 115, i32 20, i32 52, i32 84, i32 116, i32 21, i32 53, i32 85, i32 117, i32 22, i32 54, i32 86, i32 118, i32 23, i32 55, i32 87, i32 119, i32 24, i32 56, i32 88, i32 120, i32 25, i32 57, i32 89, i32 121, i32 26, i32 58, i32 90, i32 122, i32 27, i32 59, i32 91, i32 123, i32 28, i32 60, i32 92, i32 124, i32 29, i32 61, i32 93, i32 125, i32 30, i32 62, i32 94, i32 126, i32 31, i32 63, i32 95, i32 127>
+
+ store <128 x i32> %interleaved.vec, <128 x i32>* %out.vec, align 32
+
+ ret void
+}
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-2.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-2.ll
index b476e57653562..5b808dbb4c1f6 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-2.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-2.ll
@@ -416,3 +416,375 @@ define void @store_i64_stride2_vf16(<16 x i64>* %in.vecptr0, <16 x i64>* %in.vec
ret void
}
+
+define void @store_i64_stride2_vf32(<32 x i64>* %in.vecptr0, <32 x i64>* %in.vecptr1, <64 x i64>* %out.vec) nounwind {
+; SSE-LABEL: store_i64_stride2_vf32:
+; SSE: # %bb.0:
+; SSE-NEXT: subq $152, %rsp
+; SSE-NEXT: movaps 112(%rdi), %xmm14
+; SSE-NEXT: movaps 96(%rdi), %xmm13
+; SSE-NEXT: movaps 80(%rdi), %xmm11
+; SSE-NEXT: movaps 64(%rdi), %xmm10
+; SSE-NEXT: movaps (%rdi), %xmm0
+; SSE-NEXT: movaps 16(%rdi), %xmm8
+; SSE-NEXT: movaps 32(%rdi), %xmm9
+; SSE-NEXT: movaps 48(%rdi), %xmm12
+; SSE-NEXT: movaps 96(%rsi), %xmm15
+; SSE-NEXT: movaps 80(%rsi), %xmm1
+; SSE-NEXT: movaps 64(%rsi), %xmm2
+; SSE-NEXT: movaps (%rsi), %xmm3
+; SSE-NEXT: movaps 16(%rsi), %xmm4
+; SSE-NEXT: movaps 32(%rsi), %xmm5
+; SSE-NEXT: movaps 48(%rsi), %xmm6
+; SSE-NEXT: movaps %xmm0, %xmm7
+; SSE-NEXT: movlhps {{.*#+}} xmm7 = xmm7[0],xmm3[0]
+; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm8, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm4[1]
+; SSE-NEXT: movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm9, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm5[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm5[1]
+; SSE-NEXT: movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm12, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm6[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm6[1]
+; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm10, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSE-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm2[1]
+; SSE-NEXT: movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm11, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm1[1]
+; SSE-NEXT: movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm13, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm15[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm15[1]
+; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 112(%rsi), %xmm0
+; SSE-NEXT: movaps %xmm14, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm0[1]
+; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 128(%rdi), %xmm14
+; SSE-NEXT: movaps 128(%rsi), %xmm1
+; SSE-NEXT: movaps %xmm14, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm1[1]
+; SSE-NEXT: movaps 144(%rdi), %xmm10
+; SSE-NEXT: movaps 144(%rsi), %xmm2
+; SSE-NEXT: movaps %xmm10, %xmm15
+; SSE-NEXT: movlhps {{.*#+}} xmm15 = xmm15[0],xmm2[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm2[1]
+; SSE-NEXT: movaps 160(%rdi), %xmm12
+; SSE-NEXT: movaps 160(%rsi), %xmm0
+; SSE-NEXT: movaps %xmm12, %xmm13
+; SSE-NEXT: movlhps {{.*#+}} xmm13 = xmm13[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm0[1]
+; SSE-NEXT: movaps 176(%rdi), %xmm9
+; SSE-NEXT: movaps 176(%rsi), %xmm0
+; SSE-NEXT: movaps %xmm9, %xmm11
+; SSE-NEXT: movlhps {{.*#+}} xmm11 = xmm11[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm0[1]
+; SSE-NEXT: movaps 192(%rdi), %xmm7
+; SSE-NEXT: movaps 192(%rsi), %xmm1
+; SSE-NEXT: movaps %xmm7, %xmm8
+; SSE-NEXT: movlhps {{.*#+}} xmm8 = xmm8[0],xmm1[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm1[1]
+; SSE-NEXT: movaps 208(%rdi), %xmm1
+; SSE-NEXT: movaps 208(%rsi), %xmm2
+; SSE-NEXT: movaps %xmm1, %xmm6
+; SSE-NEXT: movlhps {{.*#+}} xmm6 = xmm6[0],xmm2[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; SSE-NEXT: movaps 224(%rdi), %xmm2
+; SSE-NEXT: movaps 224(%rsi), %xmm3
+; SSE-NEXT: movaps %xmm2, %xmm4
+; SSE-NEXT: movlhps {{.*#+}} xmm4 = xmm4[0],xmm3[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; SSE-NEXT: movaps 240(%rdi), %xmm3
+; SSE-NEXT: movaps 240(%rsi), %xmm5
+; SSE-NEXT: movaps %xmm3, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm5[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm5[1]
+; SSE-NEXT: movaps %xmm3, 496(%rdx)
+; SSE-NEXT: movaps %xmm0, 480(%rdx)
+; SSE-NEXT: movaps %xmm2, 464(%rdx)
+; SSE-NEXT: movaps %xmm4, 448(%rdx)
+; SSE-NEXT: movaps %xmm1, 432(%rdx)
+; SSE-NEXT: movaps %xmm6, 416(%rdx)
+; SSE-NEXT: movaps %xmm7, 400(%rdx)
+; SSE-NEXT: movaps %xmm8, 384(%rdx)
+; SSE-NEXT: movaps %xmm9, 368(%rdx)
+; SSE-NEXT: movaps %xmm11, 352(%rdx)
+; SSE-NEXT: movaps %xmm12, 336(%rdx)
+; SSE-NEXT: movaps %xmm13, 320(%rdx)
+; SSE-NEXT: movaps %xmm10, 304(%rdx)
+; SSE-NEXT: movaps %xmm15, 288(%rdx)
+; SSE-NEXT: movaps %xmm14, 272(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 256(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 240(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 224(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 208(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 192(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 176(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 160(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 144(%rdx)
+; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 128(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 112(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 96(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 80(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 64(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 48(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 32(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 16(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, (%rdx)
+; SSE-NEXT: addq $152, %rsp
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: store_i64_stride2_vf32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovaps 224(%rsi), %xmm0
+; AVX1-NEXT: vmovaps 224(%rdi), %xmm1
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm1[1],xmm0[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm1[0],xmm0[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 128(%rsi), %xmm1
+; AVX1-NEXT: vmovaps 128(%rdi), %xmm2
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm2[1],xmm1[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm9
+; AVX1-NEXT: vmovaps (%rsi), %xmm2
+; AVX1-NEXT: vmovaps 32(%rsi), %xmm3
+; AVX1-NEXT: vmovaps 64(%rsi), %xmm4
+; AVX1-NEXT: vmovaps 96(%rsi), %xmm10
+; AVX1-NEXT: vmovaps (%rdi), %xmm6
+; AVX1-NEXT: vmovaps 32(%rdi), %xmm7
+; AVX1-NEXT: vmovaps 64(%rdi), %xmm0
+; AVX1-NEXT: vmovaps 96(%rdi), %xmm1
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm6[1],xmm2[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm2 = xmm6[0],xmm2[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm2, %ymm2
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm7[1],xmm3[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm3 = xmm7[0],xmm3[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm3
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm0[1],xmm4[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm4
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm10[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm10[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm5
+; AVX1-NEXT: vmovaps 160(%rsi), %xmm0
+; AVX1-NEXT: vmovaps 160(%rdi), %xmm1
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm1[1],xmm0[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm1[0],xmm0[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-NEXT: vmovapd 192(%rsi), %xmm0
+; AVX1-NEXT: vmovapd 192(%rdi), %xmm1
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm1[1],xmm0[1]
+; AVX1-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0]
+; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = mem[2,3,2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = mem[2,3,2,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[3],ymm0[3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = mem[2,3,2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm10 = mem[2,3,2,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm1 = ymm10[0],ymm1[0],ymm10[3],ymm1[3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm10 = mem[2,3,2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm11 = mem[2,3,2,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm10 = ymm11[0],ymm10[0],ymm11[3],ymm10[3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm11 = mem[2,3,2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm12 = mem[2,3,2,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm11 = ymm12[0],ymm11[0],ymm12[3],ymm11[3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm12 = mem[2,3,2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm13 = mem[2,3,2,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm12 = ymm13[0],ymm12[0],ymm13[3],ymm12[3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm13 = mem[2,3,2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm14 = mem[2,3,2,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm13 = ymm14[0],ymm13[0],ymm14[3],ymm13[3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm14 = mem[2,3,2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm15 = mem[2,3,2,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm14 = ymm15[0],ymm14[0],ymm15[3],ymm14[3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm15 = mem[2,3,2,3]
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm8 = mem[2,3,2,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm8 = ymm8[0],ymm15[0],ymm8[3],ymm15[3]
+; AVX1-NEXT: vmovapd %ymm8, 480(%rdx)
+; AVX1-NEXT: vmovapd %ymm14, 416(%rdx)
+; AVX1-NEXT: vmovapd %ymm13, 352(%rdx)
+; AVX1-NEXT: vmovapd %ymm12, 224(%rdx)
+; AVX1-NEXT: vmovapd %ymm11, 160(%rdx)
+; AVX1-NEXT: vmovapd %ymm10, 96(%rdx)
+; AVX1-NEXT: vmovapd %ymm1, 32(%rdx)
+; AVX1-NEXT: vmovapd %ymm0, 288(%rdx)
+; AVX1-NEXT: vmovaps %ymm7, 384(%rdx)
+; AVX1-NEXT: vmovaps %ymm6, 320(%rdx)
+; AVX1-NEXT: vmovaps %ymm5, 192(%rdx)
+; AVX1-NEXT: vmovaps %ymm4, 128(%rdx)
+; AVX1-NEXT: vmovaps %ymm3, 64(%rdx)
+; AVX1-NEXT: vmovaps %ymm2, (%rdx)
+; AVX1-NEXT: vmovaps %ymm9, 256(%rdx)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 448(%rdx)
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: store_i64_stride2_vf32:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vmovaps 192(%rdi), %ymm2
+; AVX2-NEXT: vmovaps 160(%rdi), %ymm5
+; AVX2-NEXT: vmovaps 128(%rdi), %ymm8
+; AVX2-NEXT: vmovaps (%rdi), %ymm1
+; AVX2-NEXT: vmovaps 32(%rdi), %ymm4
+; AVX2-NEXT: vmovaps 64(%rdi), %ymm9
+; AVX2-NEXT: vmovaps 96(%rdi), %ymm12
+; AVX2-NEXT: vmovaps 192(%rsi), %ymm6
+; AVX2-NEXT: vmovaps 160(%rsi), %ymm10
+; AVX2-NEXT: vmovaps 128(%rsi), %ymm13
+; AVX2-NEXT: vmovaps (%rsi), %ymm3
+; AVX2-NEXT: vmovaps 32(%rsi), %ymm7
+; AVX2-NEXT: vmovaps 64(%rsi), %ymm11
+; AVX2-NEXT: vmovaps 96(%rsi), %ymm14
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm3[0,2,2,3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm15 = ymm1[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3],ymm15[4,5],ymm0[6,7]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,0,2,1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm3[2,3],ymm1[4,5],ymm3[6,7]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm7[0,2,2,3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm15 = ymm4[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm3 = ymm15[0,1],ymm3[2,3],ymm15[4,5],ymm3[6,7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,0,2,1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3],ymm4[4,5],ymm7[6,7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm7 = ymm11[0,2,2,3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm15 = ymm9[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3],ymm15[4,5],ymm7[6,7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[0,0,2,1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm11[2,3],ymm9[4,5],ymm11[6,7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm11 = ymm14[0,2,2,3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm15 = ymm12[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm11 = ymm15[0,1],ymm11[2,3],ymm15[4,5],ymm11[6,7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm14 = ymm14[0,0,2,1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm14[2,3],ymm12[4,5],ymm14[6,7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm14 = ymm13[0,2,2,3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm15 = ymm8[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3],ymm15[4,5],ymm14[6,7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[0,0,2,1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm13[2,3],ymm8[4,5],ymm13[6,7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm13 = ymm10[0,2,2,3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm15 = ymm5[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1],ymm13[2,3],ymm15[4,5],ymm13[6,7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,0,2,1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm10[2,3],ymm5[4,5],ymm10[6,7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm10 = ymm6[0,2,2,3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm15 = ymm2[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm10 = ymm15[0,1],ymm10[2,3],ymm15[4,5],ymm10[6,7]
+; AVX2-NEXT: vmovaps 224(%rdi), %ymm15
+; AVX2-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,0,2,1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm6[2,3],ymm2[4,5],ymm6[6,7]
+; AVX2-NEXT: vmovaps 224(%rsi), %ymm6
+; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm6[0,2,2,3]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm1 = ymm15[2,1,3,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm1 = ymm6[0,0,2,1]
+; AVX2-NEXT: vpermpd {{.*#+}} ymm6 = ymm15[0,1,1,3]
+; AVX2-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3],ymm6[4,5],ymm1[6,7]
+; AVX2-NEXT: vmovaps %ymm1, 448(%rdx)
+; AVX2-NEXT: vmovaps %ymm0, 480(%rdx)
+; AVX2-NEXT: vmovaps %ymm2, 384(%rdx)
+; AVX2-NEXT: vmovaps %ymm10, 416(%rdx)
+; AVX2-NEXT: vmovaps %ymm5, 320(%rdx)
+; AVX2-NEXT: vmovaps %ymm13, 352(%rdx)
+; AVX2-NEXT: vmovaps %ymm8, 256(%rdx)
+; AVX2-NEXT: vmovaps %ymm14, 288(%rdx)
+; AVX2-NEXT: vmovaps %ymm12, 192(%rdx)
+; AVX2-NEXT: vmovaps %ymm11, 224(%rdx)
+; AVX2-NEXT: vmovaps %ymm9, 128(%rdx)
+; AVX2-NEXT: vmovaps %ymm7, 160(%rdx)
+; AVX2-NEXT: vmovaps %ymm4, 64(%rdx)
+; AVX2-NEXT: vmovaps %ymm3, 96(%rdx)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, (%rdx)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 32(%rdx)
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+;
+; AVX512-LABEL: store_i64_stride2_vf32:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm0
+; AVX512-NEXT: vmovdqu64 64(%rdi), %zmm1
+; AVX512-NEXT: vmovdqu64 128(%rdi), %zmm2
+; AVX512-NEXT: vmovdqu64 192(%rdi), %zmm3
+; AVX512-NEXT: vmovdqu64 (%rsi), %zmm4
+; AVX512-NEXT: vmovdqu64 64(%rsi), %zmm5
+; AVX512-NEXT: vmovdqu64 128(%rsi), %zmm6
+; AVX512-NEXT: vmovdqu64 192(%rsi), %zmm7
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm8 = [4,12,5,13,6,14,7,15]
+; AVX512-NEXT: vmovdqa64 %zmm0, %zmm9
+; AVX512-NEXT: vpermt2q %zmm4, %zmm8, %zmm9
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm10 = [0,8,1,9,2,10,3,11]
+; AVX512-NEXT: vpermt2q %zmm4, %zmm10, %zmm0
+; AVX512-NEXT: vmovdqa64 %zmm1, %zmm4
+; AVX512-NEXT: vpermt2q %zmm5, %zmm8, %zmm4
+; AVX512-NEXT: vpermt2q %zmm5, %zmm10, %zmm1
+; AVX512-NEXT: vmovdqa64 %zmm2, %zmm5
+; AVX512-NEXT: vpermt2q %zmm6, %zmm8, %zmm5
+; AVX512-NEXT: vpermt2q %zmm6, %zmm10, %zmm2
+; AVX512-NEXT: vpermi2q %zmm7, %zmm3, %zmm8
+; AVX512-NEXT: vpermt2q %zmm7, %zmm10, %zmm3
+; AVX512-NEXT: vmovdqu64 %zmm3, 384(%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm8, 448(%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm2, 256(%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm5, 320(%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm1, 128(%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm4, 192(%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm0, (%rdx)
+; AVX512-NEXT: vmovdqu64 %zmm9, 64(%rdx)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %in.vec0 = load <32 x i64>, <32 x i64>* %in.vecptr0, align 32
+ %in.vec1 = load <32 x i64>, <32 x i64>* %in.vecptr1, align 32
+
+ %concat01 = shufflevector <32 x i64> %in.vec0, <32 x i64> %in.vec1, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+ %interleaved.vec = shufflevector <64 x i64> %concat01, <64 x i64> poison, <64 x i32> <i32 0, i32 32, i32 1, i32 33, i32 2, i32 34, i32 3, i32 35, i32 4, i32 36, i32 5, i32 37, i32 6, i32 38, i32 7, i32 39, i32 8, i32 40, i32 9, i32 41, i32 10, i32 42, i32 11, i32 43, i32 12, i32 44, i32 13, i32 45, i32 14, i32 46, i32 15, i32 47, i32 16, i32 48, i32 17, i32 49, i32 18, i32 50, i32 19, i32 51, i32 20, i32 52, i32 21, i32 53, i32 22, i32 54, i32 23, i32 55, i32 24, i32 56, i32 25, i32 57, i32 26, i32 58, i32 27, i32 59, i32 28, i32 60, i32 29, i32 61, i32 30, i32 62, i32 31, i32 63>
+
+ store <64 x i64> %interleaved.vec, <64 x i64>* %out.vec, align 32
+ ret void
+}
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-4.ll
index 12bb5aa4c5efc..15eaad9b0527f 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-4.ll
@@ -391,3 +391,473 @@ define void @store_i64_stride4_vf8(<8 x i64>* %in.vecptr0, <8 x i64>* %in.vecptr
ret void
}
+
+define void @store_i64_stride4_vf16(<16 x i64>* %in.vecptr0, <16 x i64>* %in.vecptr1, <16 x i64>* %in.vecptr2, <16 x i64>* %in.vecptr3, <64 x i64>* %out.vec) nounwind {
+; SSE-LABEL: store_i64_stride4_vf16:
+; SSE: # %bb.0:
+; SSE-NEXT: subq $152, %rsp
+; SSE-NEXT: movaps (%rdi), %xmm0
+; SSE-NEXT: movaps 16(%rdi), %xmm8
+; SSE-NEXT: movaps 32(%rdi), %xmm9
+; SSE-NEXT: movaps 48(%rdi), %xmm10
+; SSE-NEXT: movaps (%rsi), %xmm1
+; SSE-NEXT: movaps 16(%rsi), %xmm2
+; SSE-NEXT: movaps 32(%rsi), %xmm13
+; SSE-NEXT: movaps 48(%rsi), %xmm11
+; SSE-NEXT: movaps (%rdx), %xmm6
+; SSE-NEXT: movaps 16(%rdx), %xmm12
+; SSE-NEXT: movaps 32(%rdx), %xmm14
+; SSE-NEXT: movaps 48(%rdx), %xmm15
+; SSE-NEXT: movaps (%rcx), %xmm3
+; SSE-NEXT: movaps 16(%rcx), %xmm4
+; SSE-NEXT: movaps 32(%rcx), %xmm5
+; SSE-NEXT: movaps %xmm0, %xmm7
+; SSE-NEXT: movlhps {{.*#+}} xmm7 = xmm7[0],xmm1[0]
+; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm6, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm3[1]
+; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm8, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm2[1]
+; SSE-NEXT: movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm12, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm4[1]
+; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm9, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm13[0]
+; SSE-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm13[1]
+; SSE-NEXT: movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm14, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm5[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm5[1]
+; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm10, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm11[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm11[1]
+; SSE-NEXT: movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 48(%rcx), %xmm0
+; SSE-NEXT: movaps %xmm15, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm0[1]
+; SSE-NEXT: movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 64(%rdi), %xmm15
+; SSE-NEXT: movaps 64(%rsi), %xmm1
+; SSE-NEXT: movaps %xmm15, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm1[1]
+; SSE-NEXT: movaps 64(%rdx), %xmm11
+; SSE-NEXT: movaps 64(%rcx), %xmm2
+; SSE-NEXT: movaps %xmm11, %xmm14
+; SSE-NEXT: movlhps {{.*#+}} xmm14 = xmm14[0],xmm2[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm2[1]
+; SSE-NEXT: movaps 80(%rdi), %xmm9
+; SSE-NEXT: movaps 80(%rsi), %xmm0
+; SSE-NEXT: movaps %xmm9, %xmm13
+; SSE-NEXT: movlhps {{.*#+}} xmm13 = xmm13[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm0[1]
+; SSE-NEXT: movaps 80(%rdx), %xmm10
+; SSE-NEXT: movaps 80(%rcx), %xmm0
+; SSE-NEXT: movaps %xmm10, %xmm12
+; SSE-NEXT: movlhps {{.*#+}} xmm12 = xmm12[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm0[1]
+; SSE-NEXT: movaps 96(%rdi), %xmm7
+; SSE-NEXT: movaps 96(%rsi), %xmm1
+; SSE-NEXT: movaps %xmm7, %xmm8
+; SSE-NEXT: movlhps {{.*#+}} xmm8 = xmm8[0],xmm1[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm1[1]
+; SSE-NEXT: movaps 96(%rdx), %xmm5
+; SSE-NEXT: movaps 96(%rcx), %xmm0
+; SSE-NEXT: movaps %xmm5, %xmm6
+; SSE-NEXT: movlhps {{.*#+}} xmm6 = xmm6[0],xmm0[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm0[1]
+; SSE-NEXT: movaps 112(%rdi), %xmm0
+; SSE-NEXT: movaps 112(%rsi), %xmm3
+; SSE-NEXT: movaps %xmm0, %xmm2
+; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
+; SSE-NEXT: movaps 112(%rdx), %xmm3
+; SSE-NEXT: movaps 112(%rcx), %xmm4
+; SSE-NEXT: movaps %xmm3, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm4[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm4[1]
+; SSE-NEXT: movaps %xmm3, 496(%r8)
+; SSE-NEXT: movaps %xmm0, 480(%r8)
+; SSE-NEXT: movaps %xmm1, 464(%r8)
+; SSE-NEXT: movaps %xmm2, 448(%r8)
+; SSE-NEXT: movaps %xmm5, 432(%r8)
+; SSE-NEXT: movaps %xmm7, 416(%r8)
+; SSE-NEXT: movaps %xmm6, 400(%r8)
+; SSE-NEXT: movaps %xmm8, 384(%r8)
+; SSE-NEXT: movaps %xmm10, 368(%r8)
+; SSE-NEXT: movaps %xmm9, 352(%r8)
+; SSE-NEXT: movaps %xmm12, 336(%r8)
+; SSE-NEXT: movaps %xmm13, 320(%r8)
+; SSE-NEXT: movaps %xmm11, 304(%r8)
+; SSE-NEXT: movaps %xmm15, 288(%r8)
+; SSE-NEXT: movaps %xmm14, 272(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 256(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 240(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 224(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 208(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 192(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 176(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 160(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 144(%r8)
+; SSE-NEXT: movaps (%rsp), %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: addq $152, %rsp
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: store_i64_stride4_vf16:
+; AVX1: # %bb.0:
+; AVX1-NEXT: subq $152, %rsp
+; AVX1-NEXT: vmovaps 96(%rdx), %ymm12
+; AVX1-NEXT: vmovaps 64(%rdx), %ymm1
+; AVX1-NEXT: vmovaps 32(%rdx), %ymm2
+; AVX1-NEXT: vmovaps (%rdx), %ymm3
+; AVX1-NEXT: vmovaps 96(%rcx), %ymm13
+; AVX1-NEXT: vmovaps 64(%rcx), %ymm5
+; AVX1-NEXT: vmovaps 32(%rcx), %ymm4
+; AVX1-NEXT: vmovaps (%rcx), %ymm6
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm3[1],ymm6[1],ymm3[3],ymm6[3]
+; AVX1-NEXT: vmovaps 16(%rsi), %xmm0
+; AVX1-NEXT: vmovaps 16(%rdi), %xmm7
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm9 = xmm7[1],xmm0[1]
+; AVX1-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm6[0],ymm3[2],ymm6[2]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm7[0],xmm0[0]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm4[1],ymm2[3],ymm4[3]
+; AVX1-NEXT: vmovaps 48(%rsi), %xmm3
+; AVX1-NEXT: vmovaps 48(%rdi), %xmm6
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm6[1],xmm3[1]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm2[0],ymm4[0],ymm2[2],ymm4[2]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm2 = xmm6[0],xmm3[0]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm5[1],ymm1[3],ymm5[3]
+; AVX1-NEXT: vmovaps 80(%rsi), %xmm2
+; AVX1-NEXT: vmovaps 80(%rdi), %xmm3
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm3[1],xmm2[1]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm5[0],ymm1[2],ymm5[2]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm2[0]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm12[1],ymm13[1],ymm12[3],ymm13[3]
+; AVX1-NEXT: vmovaps 112(%rsi), %xmm1
+; AVX1-NEXT: vmovaps 112(%rdi), %xmm2
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm2[1],xmm1[1]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm12[0],ymm13[0],ymm12[2],ymm13[2]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT: vmovaps 64(%rsi), %xmm2
+; AVX1-NEXT: vmovaps 64(%rdi), %xmm3
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
+; AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-NEXT: vmovaps 32(%rcx), %xmm4
+; AVX1-NEXT: vmovaps 64(%rcx), %xmm5
+; AVX1-NEXT: vmovaps 64(%rdx), %xmm7
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm15 = xmm7[1],xmm5[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm14 = xmm3[0],xmm2[0]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm13 = xmm7[0],xmm5[0]
+; AVX1-NEXT: vmovaps 32(%rsi), %xmm5
+; AVX1-NEXT: vmovaps 32(%rdi), %xmm7
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm12 = xmm7[1],xmm5[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm10 = xmm7[0],xmm5[0]
+; AVX1-NEXT: vmovaps 32(%rdx), %xmm7
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm11 = xmm7[1],xmm4[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm9 = xmm7[0],xmm4[0]
+; AVX1-NEXT: vmovaps 96(%rsi), %xmm7
+; AVX1-NEXT: vmovaps 96(%rdi), %xmm0
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm0[1],xmm7[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm6 = xmm0[0],xmm7[0]
+; AVX1-NEXT: vmovaps 96(%rcx), %xmm7
+; AVX1-NEXT: vmovaps 96(%rdx), %xmm0
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm0[1],xmm7[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm4 = xmm0[0],xmm7[0]
+; AVX1-NEXT: vmovaps (%rsi), %xmm7
+; AVX1-NEXT: vmovaps (%rdi), %xmm0
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm0[1],xmm7[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm7[0]
+; AVX1-NEXT: vmovaps (%rcx), %xmm7
+; AVX1-NEXT: vmovaps (%rdx), %xmm0
+; AVX1-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm0[1],xmm7[1]
+; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm7[0]
+; AVX1-NEXT: vmovaps %xmm0, 16(%r8)
+; AVX1-NEXT: vmovaps %xmm2, (%r8)
+; AVX1-NEXT: vmovaps %xmm1, 48(%r8)
+; AVX1-NEXT: vmovaps %xmm3, 32(%r8)
+; AVX1-NEXT: vmovaps %xmm4, 400(%r8)
+; AVX1-NEXT: vmovaps %xmm6, 384(%r8)
+; AVX1-NEXT: vmovaps %xmm5, 432(%r8)
+; AVX1-NEXT: vmovaps %xmm8, 416(%r8)
+; AVX1-NEXT: vmovaps %xmm9, 144(%r8)
+; AVX1-NEXT: vmovaps %xmm10, 128(%r8)
+; AVX1-NEXT: vmovaps %xmm11, 176(%r8)
+; AVX1-NEXT: vmovaps %xmm12, 160(%r8)
+; AVX1-NEXT: vmovaps %xmm13, 272(%r8)
+; AVX1-NEXT: vmovaps %xmm14, 256(%r8)
+; AVX1-NEXT: vmovaps %xmm15, 304(%r8)
+; AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-NEXT: vmovaps %xmm0, 288(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 448(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 480(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 320(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 352(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 192(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 224(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 64(%r8)
+; AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT: vmovaps %ymm0, 96(%r8)
+; AVX1-NEXT: addq $152, %rsp
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: store_i64_stride4_vf16:
+; AVX2: # %bb.0:
+; AVX2-NEXT: subq $152, %rsp
+; AVX2-NEXT: vmovaps 96(%rdi), %ymm0
+; AVX2-NEXT: vmovaps 64(%rdi), %ymm5
+; AVX2-NEXT: vmovaps 32(%rdi), %ymm3
+; AVX2-NEXT: vmovaps (%rdi), %ymm1
+; AVX2-NEXT: vmovaps 96(%rsi), %ymm7
+; AVX2-NEXT: vmovaps 64(%rsi), %ymm6
+; AVX2-NEXT: vmovaps 32(%rsi), %ymm4
+; AVX2-NEXT: vmovaps (%rsi), %ymm2
+; AVX2-NEXT: vmovaps 96(%rdx), %ymm9
+; AVX2-NEXT: vmovaps 64(%rdx), %ymm10
+; AVX2-NEXT: vmovaps 32(%rdx), %ymm11
+; AVX2-NEXT: vmovaps (%rdx), %ymm12
+; AVX2-NEXT: vmovaps 96(%rcx), %ymm13
+; AVX2-NEXT: vmovaps (%rcx), %ymm14
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm12[1],ymm14[1],ymm12[3],ymm14[3]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm15 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm15[2,3],ymm8[2,3]
+; AVX2-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 64(%rcx), %ymm15
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm12[0],ymm14[0],ymm12[2],ymm14[2]
+; AVX2-NEXT: vmovaps 32(%rcx), %ymm14
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm12[2,3]
+; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm11[1],ymm14[1],ymm11[3],ymm14[3]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm12 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm12[2,3],ymm2[2,3]
+; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm11[0],ymm14[0],ymm11[2],ymm14[2]
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm3[2,3],ymm11[2,3]
+; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm10[1],ymm15[1],ymm10[3],ymm15[3]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm11 = ymm5[1],ymm6[1],ymm5[3],ymm6[3]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm11[2,3],ymm4[2,3]
+; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm10[0],ymm15[0],ymm10[2],ymm15[2]
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm5[0],ymm6[0],ymm5[2],ymm6[2]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm10[2,3]
+; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm9[1],ymm13[1],ymm9[3],ymm13[3]
+; AVX2-NEXT: vunpckhpd {{.*#+}} ymm10 = ymm0[1],ymm7[1],ymm0[3],ymm7[3]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm10[2,3],ymm6[2,3]
+; AVX2-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm9[0],ymm13[0],ymm9[2],ymm13[2]
+; AVX2-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm7[0],ymm0[2],ymm7[2]
+; AVX2-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm9[2,3]
+; AVX2-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-NEXT: vmovaps 32(%rsi), %xmm2
+; AVX2-NEXT: vmovaps 64(%rsi), %xmm3
+; AVX2-NEXT: vmovaps 32(%rdi), %xmm4
+; AVX2-NEXT: vmovaps 64(%rdi), %xmm5
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm5[1],xmm3[1]
+; AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-NEXT: vmovaps 32(%rcx), %xmm6
+; AVX2-NEXT: vmovaps 64(%rcx), %xmm7
+; AVX2-NEXT: vmovaps 64(%rdx), %xmm0
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm14 = xmm5[0],xmm3[0]
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm15 = xmm0[1],xmm7[1]
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm13 = xmm0[0],xmm7[0]
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm12 = xmm4[1],xmm2[1]
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm10 = xmm4[0],xmm2[0]
+; AVX2-NEXT: vmovaps 32(%rdx), %xmm4
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm11 = xmm4[1],xmm6[1]
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm9 = xmm4[0],xmm6[0]
+; AVX2-NEXT: vmovaps 96(%rsi), %xmm6
+; AVX2-NEXT: vmovaps 96(%rdi), %xmm0
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm0[1],xmm6[1]
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm7 = xmm0[0],xmm6[0]
+; AVX2-NEXT: vmovaps 96(%rcx), %xmm6
+; AVX2-NEXT: vmovaps 96(%rdx), %xmm0
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm0[1],xmm6[1]
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm4 = xmm0[0],xmm6[0]
+; AVX2-NEXT: vmovaps (%rsi), %xmm6
+; AVX2-NEXT: vmovaps (%rdi), %xmm0
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm0[1],xmm6[1]
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm6[0]
+; AVX2-NEXT: vmovaps (%rcx), %xmm6
+; AVX2-NEXT: vmovaps (%rdx), %xmm0
+; AVX2-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm0[1],xmm6[1]
+; AVX2-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm6[0]
+; AVX2-NEXT: vmovaps %xmm0, 16(%r8)
+; AVX2-NEXT: vmovaps %xmm2, (%r8)
+; AVX2-NEXT: vmovaps %xmm1, 48(%r8)
+; AVX2-NEXT: vmovaps %xmm3, 32(%r8)
+; AVX2-NEXT: vmovaps %xmm4, 400(%r8)
+; AVX2-NEXT: vmovaps %xmm7, 384(%r8)
+; AVX2-NEXT: vmovaps %xmm5, 432(%r8)
+; AVX2-NEXT: vmovaps %xmm8, 416(%r8)
+; AVX2-NEXT: vmovaps %xmm9, 144(%r8)
+; AVX2-NEXT: vmovaps %xmm10, 128(%r8)
+; AVX2-NEXT: vmovaps %xmm11, 176(%r8)
+; AVX2-NEXT: vmovaps %xmm12, 160(%r8)
+; AVX2-NEXT: vmovaps %xmm13, 272(%r8)
+; AVX2-NEXT: vmovaps %xmm14, 256(%r8)
+; AVX2-NEXT: vmovaps %xmm15, 304(%r8)
+; AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-NEXT: vmovaps %xmm0, 288(%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 448(%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 480(%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 320(%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 352(%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 192(%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 224(%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 64(%r8)
+; AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-NEXT: vmovaps %ymm0, 96(%r8)
+; AVX2-NEXT: addq $152, %rsp
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+;
+; AVX512-LABEL: store_i64_stride4_vf16:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu64 (%rdi), %zmm0
+; AVX512-NEXT: vmovdqu64 64(%rdi), %zmm1
+; AVX512-NEXT: vmovdqu64 (%rsi), %zmm2
+; AVX512-NEXT: vmovdqu64 64(%rsi), %zmm3
+; AVX512-NEXT: vmovdqu64 (%rdx), %zmm4
+; AVX512-NEXT: vmovdqu64 64(%rdx), %zmm5
+; AVX512-NEXT: vmovdqu64 (%rcx), %zmm6
+; AVX512-NEXT: vmovdqu64 64(%rcx), %zmm7
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm8 = <u,u,2,10,u,u,3,11>
+; AVX512-NEXT: vmovdqa64 %zmm4, %zmm9
+; AVX512-NEXT: vpermt2q %zmm6, %zmm8, %zmm9
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm10 = <2,10,u,u,3,11,u,u>
+; AVX512-NEXT: vmovdqa64 %zmm0, %zmm11
+; AVX512-NEXT: vpermt2q %zmm2, %zmm10, %zmm11
+; AVX512-NEXT: movb $-52, %al
+; AVX512-NEXT: kmovd %eax, %k1
+; AVX512-NEXT: vmovdqa64 %zmm9, %zmm11 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm9 = <u,u,0,8,u,u,1,9>
+; AVX512-NEXT: vmovdqa64 %zmm4, %zmm12
+; AVX512-NEXT: vpermt2q %zmm6, %zmm9, %zmm12
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm13 = <0,8,u,u,1,9,u,u>
+; AVX512-NEXT: vmovdqa64 %zmm0, %zmm14
+; AVX512-NEXT: vpermt2q %zmm2, %zmm13, %zmm14
+; AVX512-NEXT: vmovdqa64 %zmm12, %zmm14 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm12 = <u,u,6,14,u,u,7,15>
+; AVX512-NEXT: vmovdqa64 %zmm4, %zmm15
+; AVX512-NEXT: vpermt2q %zmm6, %zmm12, %zmm15
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm16 = <6,14,u,u,7,15,u,u>
+; AVX512-NEXT: vmovdqa64 %zmm0, %zmm17
+; AVX512-NEXT: vpermt2q %zmm2, %zmm16, %zmm17
+; AVX512-NEXT: vmovdqa64 %zmm15, %zmm17 {%k1}
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm15 = <u,u,4,12,u,u,5,13>
+; AVX512-NEXT: vpermt2q %zmm6, %zmm15, %zmm4
+; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm6 = <4,12,u,u,5,13,u,u>
+; AVX512-NEXT: vpermt2q %zmm2, %zmm6, %zmm0
+; AVX512-NEXT: vmovdqa64 %zmm4, %zmm0 {%k1}
+; AVX512-NEXT: vpermi2q %zmm7, %zmm5, %zmm8
+; AVX512-NEXT: vpermi2q %zmm3, %zmm1, %zmm10
+; AVX512-NEXT: vmovdqa64 %zmm8, %zmm10 {%k1}
+; AVX512-NEXT: vpermi2q %zmm7, %zmm5, %zmm9
+; AVX512-NEXT: vpermi2q %zmm3, %zmm1, %zmm13
+; AVX512-NEXT: vmovdqa64 %zmm9, %zmm13 {%k1}
+; AVX512-NEXT: vpermi2q %zmm7, %zmm5, %zmm12
+; AVX512-NEXT: vpermi2q %zmm3, %zmm1, %zmm16
+; AVX512-NEXT: vmovdqa64 %zmm12, %zmm16 {%k1}
+; AVX512-NEXT: vpermt2q %zmm7, %zmm15, %zmm5
+; AVX512-NEXT: vpermt2q %zmm3, %zmm6, %zmm1
+; AVX512-NEXT: vmovdqa64 %zmm5, %zmm1 {%k1}
+; AVX512-NEXT: vmovdqu64 %zmm1, 384(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm16, 448(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm13, 256(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm10, 320(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm0, 128(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm17, 192(%r8)
+; AVX512-NEXT: vmovdqu64 %zmm14, (%r8)
+; AVX512-NEXT: vmovdqu64 %zmm11, 64(%r8)
+; AVX512-NEXT: vzeroupper
+; AVX512-NEXT: retq
+ %in.vec0 = load <16 x i64>, <16 x i64>* %in.vecptr0, align 32
+ %in.vec1 = load <16 x i64>, <16 x i64>* %in.vecptr1, align 32
+ %in.vec2 = load <16 x i64>, <16 x i64>* %in.vecptr2, align 32
+ %in.vec3 = load <16 x i64>, <16 x i64>* %in.vecptr3, align 32
+
+ %concat01 = shufflevector <16 x i64> %in.vec0, <16 x i64> %in.vec1, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+ %concat23 = shufflevector <16 x i64> %in.vec2, <16 x i64> %in.vec3, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+ %concat0123 = shufflevector <32 x i64> %concat01, <32 x i64> %concat23, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+ %interleaved.vec = shufflevector <64 x i64> %concat0123, <64 x i64> poison, <64 x i32> <i32 0, i32 16, i32 32, i32 48, i32 1, i32 17, i32 33, i32 49, i32 2, i32 18, i32 34, i32 50, i32 3, i32 19, i32 35, i32 51, i32 4, i32 20, i32 36, i32 52, i32 5, i32 21, i32 37, i32 53, i32 6, i32 22, i32 38, i32 54, i32 7, i32 23, i32 39, i32 55, i32 8, i32 24, i32 40, i32 56, i32 9, i32 25, i32 41, i32 57, i32 10, i32 26, i32 42, i32 58, i32 11, i32 27, i32 43, i32 59, i32 12, i32 28, i32 44, i32 60, i32 13, i32 29, i32 45, i32 61, i32 14, i32 30, i32 46, i32 62, i32 15, i32 31, i32 47, i32 63>
+
+ store <64 x i64> %interleaved.vec, <64 x i64>* %out.vec, align 32
+
+ ret void
+}
More information about the llvm-commits
mailing list