[x86] match concat of 128-bit high half vectors before lowering to VPERM2X128

[spatel]

Summary

shuffle (ins ?, X, C1), (ins ?, Y, C2), Mask --> concat X, Y

This is another shuffle problem seen with PR42024:
https://bugs.llvm.org/show_bug.cgi?id=42024

We have this small crack in legalization/lowering/combining/demanded that allows forming a vperm2f128 of high halves with AVX1 when we could do better by peeking through the insert_subvector nodes. AFAICT, it requires IR as shown in the diffs - much larger than legal vectors - to avoid all of the usual folds.

Another option might be to allow forming the 256-bit vperm here and then doing custom shuffle combining on that opcode.

[RKSimon]

This looks OK, but I'd be curious if we'd hit more cases by adding a vperm2f128(insert_subvector,insert_subvector) combine in combineTargetShuffle instead.

[spatel]

In D73197#1834183, @RKSimon wrote:

This looks OK, but I'd be curious if we'd hit more cases by adding a vperm2f128(insert_subvector,insert_subvector) combine in combineTargetShuffle instead.

I drafted a combineTargetShuffle() alternative, and it doesn't cause any difference on existing regression tests. I'll post it here for comparison.

[spatel]

Patch updated:
Move the transform to combineTargetShuffle(); no test diffs.

[RKSimon]

LGTM - cheers

llvm/lib/Target/X86/X86ISelLowering.cpp

if (N0.getOpcode() == ISD::BITCAST &&
         N0.getOperand(0).getScalarValueSizeInBits() == EltSizeInBits) {
       SDValue Src = N0.getOperand(0);
       EVT SrcVT = Src.getValueType();
       SDValue Res = DAG.getNode(X86ISD::VPERMI, DL, SrcVT, Src, N1);
       return DAG.getBitcast(VT, Res);
     }
     return SDValue();
   }
+  case X86ISD::VPERM2X128: {
+    // If both 128-bit values were inserted into high halves of 256-bit values,
+    // the shuffle can be reduced to a concatenation of subvectors:
+    // vperm2x128 (ins ?, X, C1), (ins ?, Y, C2), 0x31 --> concat X, Y
+    // Note: We are only looking for the exact high/high shuffle mask because we
+    //       expect to fold other similar patterns before creating this opcode.
+    SDValue Ins0 = peekThroughBitcasts(N.getOperand(0));
+    SDValue Ins1 = peekThroughBitcasts(N.getOperand(1));
+    unsigned Imm = N.getConstantOperandVal(2);
+    if (!(Imm == 0x31 &&
+          Ins0.getOpcode() == ISD::INSERT_SUBVECTOR &&
+          Ins1.getOpcode() == ISD::INSERT_SUBVECTOR &&
+          Ins0.getValueType() == Ins1.getValueType() &&
+          isa<ConstantSDNode>(Ins0.getOperand(2)) &&
+          isa<ConstantSDNode>(Ins1.getOperand(2))))
+      return SDValue();
+
+    SDValue X = Ins0.getOperand(1);
+    SDValue Y = Ins1.getOperand(1);
+    unsigned C1 = Ins0.getConstantOperandVal(2);
+    unsigned C2 = Ins1.getConstantOperandVal(2);
+    MVT SrcVT = X.getSimpleValueType();
+    unsigned SrcElts = SrcVT.getVectorNumElements();
+    if (SrcVT != Y.getSimpleValueType() || SrcVT.getSizeInBits() != 128 ||
+        C1 != SrcElts || C2 != SrcElts)
+      return SDValue();
+
+    return DAG.getBitcast(VT, DAG.getNode(ISD::CONCAT_VECTORS, DL,
+                                          Ins1.getValueType(), X, Y));
+  }
   case X86ISD::PSHUFD:
   case X86ISD::PSHUFLW:
   case X86ISD::PSHUFHW:
     Mask = getPSHUFShuffleMask(N);
     assert(Mask.size() == 4);
     break;
   case X86ISD::MOVSD:
   case X86ISD::MOVSS: {

llvm/test/CodeGen/X86/x86-interleaved-access.ll

; AVX512-NEXT:    retq
   store <16 x i64> %interleaved.vec, <16 x i64>* %ptr, align 16
   ret void
 }
 
 
 define void @interleaved_store_vf32_i8_stride4(<32 x i8> %x1, <32 x i8> %x2, <32 x i8> %x3, <32 x i8> %x4, <128 x i8>* %p) {
 ; AVX1-LABEL: interleaved_store_vf32_i8_stride4:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm4 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm9 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm5
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm6
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm7 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3],xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm8 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm9 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm5 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
 ; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm6
 ; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm0
-; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm1 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm4 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm7[0],xmm1[0],xmm7[1],xmm1[1],xmm7[2],xmm1[2],xmm7[3],xmm1[3]
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm6 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
-; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm7[4],xmm1[4],xmm7[5],xmm1[5],xmm7[6],xmm1[6],xmm7[7],xmm1[7]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm3[2,3],ymm1[2,3]
-; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm3 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
-; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm8[0],xmm2[0],xmm8[1],xmm2[1],xmm8[2],xmm2[2],xmm8[3],xmm2[3]
-; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm9[4],xmm0[4],xmm9[5],xmm0[5],xmm9[6],xmm0[6],xmm9[7],xmm0[7]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm4[2,3],ymm0[2,3]
-; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm8[4],xmm2[4],xmm8[5],xmm2[5],xmm8[6],xmm2[6],xmm8[7],xmm2[7]
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm6, %ymm3
-; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm5, %ymm2
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm6 = xmm9[0],xmm5[0],xmm9[1],xmm5[1],xmm9[2],xmm5[2],xmm9[3],xmm5[3]
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm7[4],xmm4[4],xmm7[5],xmm4[5],xmm7[6],xmm4[6],xmm7[7],xmm4[7]
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm5 = xmm9[4],xmm5[4],xmm9[5],xmm5[5],xmm9[6],xmm5[6],xmm9[7],xmm5[7]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm9 = xmm8[0],xmm2[0],xmm8[1],xmm2[1],xmm8[2],xmm2[2],xmm8[3],xmm2[3]
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm8[4],xmm2[4],xmm8[5],xmm2[5],xmm8[6],xmm2[6],xmm8[7],xmm2[7]
+; AVX1-NEXT:    vinsertf128 $1, %xmm5, %ymm6, %ymm2
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm9, %ymm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm7, %ymm0
 ; AVX1-NEXT:    vmovaps %ymm0, 96(%rdi)
-; AVX1-NEXT:    vmovaps %ymm1, 64(%rdi)
-; AVX1-NEXT:    vmovaps %ymm2, 32(%rdi)
-; AVX1-NEXT:    vmovaps %ymm3, (%rdi)
+; AVX1-NEXT:    vmovaps %ymm3, 64(%rdi)
+; AVX1-NEXT:    vmovaps %ymm1, 32(%rdi)
+; AVX1-NEXT:    vmovaps %ymm2, (%rdi)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: interleaved_store_vf32_i8_stride4:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm4 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23]
 ; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm0 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm1 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[4],ymm3[4],ymm2[5],ymm3[5],ymm2[6],ymm3[6],ymm2[7],ymm3[7],ymm2[16],ymm3[16],ymm2[17],ymm3[17],ymm2[18],ymm3[18],ymm2[19],ymm3[19],ymm2[20],ymm3[20],ymm2[21],ymm3[21],ymm2[22],ymm3[22],ymm2[23],ymm3[23]

 %add2 = add <64 x i8> %v3, %add1
 ret <64 x i8> %add2
 }
 
 define void @interleaved_store_vf64_i8_stride4(<64 x i8> %a, <64 x i8> %b, <64 x i8> %c,<64 x i8> %d, <256 x i8>* %p) {
 ; AVX1-LABEL: interleaved_store_vf64_i8_stride4:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm8 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
-; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm11
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm12
-; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm9 = xmm12[0],xmm11[0],xmm12[1],xmm11[1],xmm12[2],xmm11[2],xmm12[3],xmm11[3],xmm12[4],xmm11[4],xmm12[5],xmm11[5],xmm12[6],xmm11[6],xmm12[7],xmm11[7]
-; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm10 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm10
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm11
+; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm9 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3],xmm11[4],xmm10[4],xmm11[5],xmm10[5],xmm11[6],xmm10[6],xmm11[7],xmm10[7]
+; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm12 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
 ; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm13
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm14
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm15 = xmm14[0],xmm13[0],xmm14[1],xmm13[1],xmm14[2],xmm13[2],xmm14[3],xmm13[3],xmm14[4],xmm13[4],xmm14[5],xmm13[5],xmm14[6],xmm13[6],xmm14[7],xmm13[7]
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
 ; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm11 = xmm12[8],xmm11[8],xmm12[9],xmm11[9],xmm12[10],xmm11[10],xmm12[11],xmm11[11],xmm12[12],xmm11[12],xmm12[13],xmm11[13],xmm12[14],xmm11[14],xmm12[15],xmm11[15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm12 = xmm1[8],xmm3[8],xmm1[9],xmm3[9],xmm1[10],xmm3[10],xmm1[11],xmm3[11],xmm1[12],xmm3[12],xmm1[13],xmm3[13],xmm1[14],xmm3[14],xmm1[15],xmm3[15]
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm10 = xmm11[8],xmm10[8],xmm11[9],xmm10[9],xmm11[10],xmm10[10],xmm11[11],xmm10[11],xmm11[12],xmm10[12],xmm11[13],xmm10[13],xmm11[14],xmm10[14],xmm11[15],xmm10[15]
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm11 = xmm1[8],xmm3[8],xmm1[9],xmm3[9],xmm1[10],xmm3[10],xmm1[11],xmm3[11],xmm1[12],xmm3[12],xmm1[13],xmm3[13],xmm1[14],xmm3[14],xmm1[15],xmm3[15]
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm13 = xmm14[8],xmm13[8],xmm14[9],xmm13[9],xmm14[10],xmm13[10],xmm14[11],xmm13[11],xmm14[12],xmm13[12],xmm14[13],xmm13[13],xmm14[14],xmm13[14],xmm14[15],xmm13[15]
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm14 = xmm4[0],xmm6[0],xmm4[1],xmm6[1],xmm4[2],xmm6[2],xmm4[3],xmm6[3],xmm4[4],xmm6[4],xmm4[5],xmm6[5],xmm4[6],xmm6[6],xmm4[7],xmm6[7]
 ; AVX1-NEXT:    vextractf128 $1, %ymm6, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm4, %xmm3
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm2 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3],xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm6[8],xmm4[9],xmm6[9],xmm4[10],xmm6[10],xmm4[11],xmm6[11],xmm4[12],xmm6[12],xmm4[13],xmm6[13],xmm4[14],xmm6[14],xmm4[15],xmm6[15]
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm6 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3],xmm5[4],xmm7[4],xmm5[5],xmm7[5],xmm5[6],xmm7[6],xmm5[7],xmm7[7]
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
 ; AVX1-NEXT:    vextractf128 $1, %ymm7, %xmm3
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm7 = xmm5[8],xmm7[8],xmm5[9],xmm7[9],xmm5[10],xmm7[10],xmm5[11],xmm7[11],xmm5[12],xmm7[12],xmm5[13],xmm7[13],xmm5[14],xmm7[14],xmm5[15],xmm7[15]
 ; AVX1-NEXT:    vextractf128 $1, %ymm5, %xmm5
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3],xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm5[8],xmm3[8],xmm5[9],xmm3[9],xmm5[10],xmm3[10],xmm5[11],xmm3[11],xmm5[12],xmm3[12],xmm5[13],xmm3[13],xmm5[14],xmm3[14],xmm5[15],xmm3[15]
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm15[0],xmm0[0],xmm15[1],xmm0[1],xmm15[2],xmm0[2],xmm15[3],xmm0[3]
-; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm15[4],xmm0[4],xmm15[5],xmm0[5],xmm15[6],xmm0[6],xmm15[7],xmm0[7]
-; AVX1-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm5[2,3],ymm0[2,3]
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm5 = xmm10[0],xmm6[0],xmm10[1],xmm6[1],xmm10[2],xmm6[2],xmm10[3],xmm6[3]
-; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm10 = xmm10[4],xmm6[4],xmm10[5],xmm6[5],xmm10[6],xmm6[6],xmm10[7],xmm6[7]
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm9[0],xmm2[0],xmm9[1],xmm2[1],xmm9[2],xmm2[2],xmm9[3],xmm2[3]
+; AVX1-NEXT:    vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm5 = xmm15[4],xmm0[4],xmm15[5],xmm0[5],xmm15[6],xmm0[6],xmm15[7],xmm0[7]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm15 = xmm12[0],xmm6[0],xmm12[1],xmm6[1],xmm12[2],xmm6[2],xmm12[3],xmm6[3]
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm6 = xmm12[4],xmm6[4],xmm12[5],xmm6[5],xmm12[6],xmm6[6],xmm12[7],xmm6[7]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm12 = xmm9[0],xmm2[0],xmm9[1],xmm2[1],xmm9[2],xmm2[2],xmm9[3],xmm2[3]
 ; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm2 = xmm9[4],xmm2[4],xmm9[5],xmm2[5],xmm9[6],xmm2[6],xmm9[7],xmm2[7]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm9 = ymm0[2,3],ymm2[2,3]
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm8[0],xmm14[0],xmm8[1],xmm14[1],xmm8[2],xmm14[2],xmm8[3],xmm14[3]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm9 = xmm8[0],xmm14[0],xmm8[1],xmm14[1],xmm8[2],xmm14[2],xmm8[3],xmm14[3]
 ; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm8 = xmm8[4],xmm14[4],xmm8[5],xmm14[5],xmm8[6],xmm14[6],xmm8[7],xmm14[7]
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm6 = xmm13[0],xmm3[0],xmm13[1],xmm3[1],xmm13[2],xmm3[2],xmm13[3],xmm3[3]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm14 = xmm13[0],xmm3[0],xmm13[1],xmm3[1],xmm13[2],xmm3[2],xmm13[3],xmm3[3]
 ; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm3 = xmm13[4],xmm3[4],xmm13[5],xmm3[5],xmm13[6],xmm3[6],xmm13[7],xmm3[7]
-; AVX1-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm6[2,3],ymm3[2,3]
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm6 = xmm12[0],xmm7[0],xmm12[1],xmm7[1],xmm12[2],xmm7[2],xmm12[3],xmm7[3]
-; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm7 = xmm12[4],xmm7[4],xmm12[5],xmm7[5],xmm12[6],xmm7[6],xmm12[7],xmm7[7]
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm11[0],xmm1[0],xmm11[1],xmm1[1],xmm11[2],xmm1[2],xmm11[3],xmm1[3]
-; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm11[4],xmm1[4],xmm11[5],xmm1[5],xmm11[6],xmm1[6],xmm11[7],xmm1[7]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm11 = ymm0[2,3],ymm1[2,3]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm13 = xmm11[0],xmm7[0],xmm11[1],xmm7[1],xmm11[2],xmm7[2],xmm11[3],xmm7[3]
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm7 = xmm11[4],xmm7[4],xmm11[5],xmm7[5],xmm11[6],xmm7[6],xmm11[7],xmm7[7]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm11 = xmm10[0],xmm1[0],xmm10[1],xmm1[1],xmm10[2],xmm1[2],xmm10[3],xmm1[3]
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm10[4],xmm1[4],xmm10[5],xmm1[5],xmm10[6],xmm1[6],xmm10[7],xmm1[7]
 ; AVX1-NEXT:    vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm10 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
 ; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm4 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
-; AVX1-NEXT:    vinsertf128 $1, %xmm8, %ymm2, %ymm2
-; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm1, %ymm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm10, %ymm5, %ymm4
-; AVX1-NEXT:    vinsertf128 $1, %xmm7, %ymm6, %ymm5
+; AVX1-NEXT:    vinsertf128 $1, %xmm8, %ymm9, %ymm8
+; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm10, %ymm4
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm12, %ymm2
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm11, %ymm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm6, %ymm15, %ymm6
+; AVX1-NEXT:    vinsertf128 $1, %xmm7, %ymm13, %ymm7
+; AVX1-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm14, %ymm3
 ; AVX1-NEXT:    vmovaps %ymm3, 224(%rdi)
-; AVX1-NEXT:    vmovaps %ymm15, 192(%rdi)
-; AVX1-NEXT:    vmovaps %ymm5, 160(%rdi)
-; AVX1-NEXT:    vmovaps %ymm4, 128(%rdi)
-; AVX1-NEXT:    vmovaps %ymm11, 96(%rdi)
-; AVX1-NEXT:    vmovaps %ymm9, 64(%rdi)
-; AVX1-NEXT:    vmovaps %ymm1, 32(%rdi)
-; AVX1-NEXT:    vmovaps %ymm2, (%rdi)
+; AVX1-NEXT:    vmovaps %ymm0, 192(%rdi)
+; AVX1-NEXT:    vmovaps %ymm7, 160(%rdi)
+; AVX1-NEXT:    vmovaps %ymm6, 128(%rdi)
+; AVX1-NEXT:    vmovaps %ymm1, 96(%rdi)
+; AVX1-NEXT:    vmovaps %ymm2, 64(%rdi)
+; AVX1-NEXT:    vmovaps %ymm4, 32(%rdi)
+; AVX1-NEXT:    vmovaps %ymm8, (%rdi)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: interleaved_store_vf64_i8_stride4:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm8 = ymm0[0],ymm2[0],ymm0[1],ymm2[1],ymm0[2],ymm2[2],ymm0[3],ymm2[3],ymm0[4],ymm2[4],ymm0[5],ymm2[5],ymm0[6],ymm2[6],ymm0[7],ymm2[7],ymm0[16],ymm2[16],ymm0[17],ymm2[17],ymm0[18],ymm2[18],ymm0[19],ymm2[19],ymm0[20],ymm2[20],ymm0[21],ymm2[21],ymm0[22],ymm2[22],ymm0[23],ymm2[23]
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm9 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[4],ymm3[4],ymm1[5],ymm3[5],ymm1[6],ymm3[6],ymm1[7],ymm3[7],ymm1[16],ymm3[16],ymm1[17],ymm3[17],ymm1[18],ymm3[18],ymm1[19],ymm3[19],ymm1[20],ymm3[20],ymm1[21],ymm3[21],ymm1[22],ymm3[22],ymm1[23],ymm3[23]
 ; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm0 = ymm0[8],ymm2[8],ymm0[9],ymm2[9],ymm0[10],ymm2[10],ymm0[11],ymm2[11],ymm0[12],ymm2[12],ymm0[13],ymm2[13],ymm0[14],ymm2[14],ymm0[15],ymm2[15],ymm0[24],ymm2[24],ymm0[25],ymm2[25],ymm0[26],ymm2[26],ymm0[27],ymm2[27],ymm0[28],ymm2[28],ymm0[29],ymm2[29],ymm0[30],ymm2[30],ymm0[31],ymm2[31]