[X86] Make use of vpmovwb when possible in LowerMULH

[craig.topper]

Summary

If we have BWI, we can truncate in a much simpler way by using vpmovwb. This even works without VLX by using the wider zmm->ymm truncate with a subvector extract.

[RKSimon]

LGTM

lib/Target/X86/X86ISelLowering.cpp

if (Subtarget.hasInt256()) {
 
     if (VT == MVT::v32i8) {
       if (Subtarget.hasBWI()) {
         SDValue ExA = DAG.getNode(ExAVX, dl, MVT::v32i16, A);
         SDValue ExB = DAG.getNode(ExAVX, dl, MVT::v32i16, B);
         SDValue Mul = DAG.getNode(ISD::MUL, dl, MVT::v32i16, ExA, ExB);
         Mul = DAG.getNode(ISD::SRL, dl, MVT::v32i16, Mul,
                           DAG.getConstant(8, dl, MVT::v32i16));
-        // The ymm variant of PACKUS treats the 128-bit lanes separately, so
-        // before using PACKUS we need to permute the inputs to the correct
-        // lo/hi xmm lane.
-        const int Mask[] = { 0,  1,  2,  3,  4,  5,  6,  7,
-                            16, 17, 18, 19, 20, 21, 22, 23,
-                             8,  9, 10, 11, 12, 13, 14, 15,
-                            24, 25, 26, 27, 28, 29, 30, 31};
-        Mul = DAG.getVectorShuffle(MVT::v32i16, dl, Mul, Mul, Mask);
-        Lo = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, MVT::v16i16, Mul, Lo);
-        Hi = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, MVT::v16i16, Mul, Hi);
-        return DAG.getNode(X86ISD::PACKUS, dl, VT, Lo, Hi);
+        return DAG.getNode(ISD::TRUNCATE, dl, VT, Mul);
       }
       SDValue ALo = extract128BitVector(A, 0, DAG, dl);
       SDValue BLo = extract128BitVector(B, 0, DAG, dl);
       SDValue AHi = extract128BitVector(A, NumElems / 2, DAG, dl);
       SDValue BHi = extract128BitVector(B, NumElems / 2, DAG, dl);
       ALo = DAG.getNode(ExAVX, dl, MVT::v16i16, ALo);
       BLo = DAG.getNode(ExAVX, dl, MVT::v16i16, BLo);
       AHi = DAG.getNode(ExAVX, dl, MVT::v16i16, AHi);
if (VT == MVT::v32i8) {
       return DAG.getNode(X86ISD::PACKUS, dl, VT,
                          DAG.getVectorShuffle(MVT::v16i16, dl, Lo, Hi, LoMask),
                          DAG.getVectorShuffle(MVT::v16i16, dl, Lo, Hi, HiMask));
     }
 
     SDValue ExA = DAG.getNode(ExAVX, dl, MVT::v16i16, A);
     SDValue ExB = DAG.getNode(ExAVX, dl, MVT::v16i16, B);
     SDValue Mul = DAG.getNode(ISD::MUL, dl, MVT::v16i16, ExA, ExB);
-    SDValue MulH = DAG.getNode(ISD::SRL, dl, MVT::v16i16, Mul,
-                               DAG.getConstant(8, dl, MVT::v16i16));
-    Lo = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, MVT::v8i16, MulH, Lo);
-    Hi = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, MVT::v8i16, MulH, Hi);
+    Mul = DAG.getNode(ISD::SRL, dl, MVT::v16i16, Mul,
+    // If we have BWI we can use truncate instruction.
+    if (Subtarget.hasBWI())
+      return DAG.getNode(ISD::TRUNCATE, dl, VT, Mul);
+    Lo = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, MVT::v8i16, Mul, Lo);
+    Hi = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, MVT::v8i16, Mul, Hi);
     return DAG.getNode(X86ISD::PACKUS, dl, VT, Lo, Hi);
   }
 
   assert(VT == MVT::v16i8 &&
          "Pre-AVX2 support only supports v16i8 multiplication");
   MVT ExVT = MVT::v8i16;
   unsigned ExSSE41 = (ISD::MULHU == Opcode ? X86ISD::VZEXT : X86ISD::VSEXT);
 

test/CodeGen/X86/vector-idiv-sdiv-128.ll

 ; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32]
 ; AVX1-NEXT:    vpxor %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpsubb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpsrlw $7, %xmm0, %xmm0
 ; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: test_div7_16i8:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vpmovsxbw %xmm0, %ymm1
-; AVX2-NEXT:    vpmullw {{.*}}(%rip), %ymm1, %ymm1
-; AVX2-NEXT:    vpsrlw $8, %ymm1, %ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX2-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpsrlw $2, %xmm0, %xmm1
-; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
-; AVX2-NEXT:    vmovdqa {{.*#+}} xmm2 = [32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32]
-; AVX2-NEXT:    vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpsubb %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpsrlw $7, %xmm0, %xmm0
-; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
-; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
+; AVX2NOBW-LABEL: test_div7_16i8:
+; AVX2NOBW:       # BB#0:
+; AVX2NOBW-NEXT:    vpmovsxbw %xmm0, %ymm1
+; AVX2NOBW-NEXT:    vpmullw {{.*}}(%rip), %ymm1, %ymm1
+; AVX2NOBW-NEXT:    vpsrlw $8, %ymm1, %ymm1
+; AVX2NOBW-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2NOBW-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
+; AVX2NOBW-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2NOBW-NEXT:    vpsrlw $2, %xmm0, %xmm1
+; AVX2NOBW-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX2NOBW-NEXT:    vmovdqa {{.*#+}} xmm2 = [32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32]
+; AVX2NOBW-NEXT:    vpxor %xmm2, %xmm1, %xmm1
+; AVX2NOBW-NEXT:    vpsubb %xmm2, %xmm1, %xmm1
+; AVX2NOBW-NEXT:    vpsrlw $7, %xmm0, %xmm0
+; AVX2NOBW-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
+; AVX2NOBW-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2NOBW-NEXT:    vzeroupper
+; AVX2NOBW-NEXT:    retq
+;
+; AVX512BW-LABEL: test_div7_16i8:
+; AVX512BW:       # BB#0:
+; AVX512BW-NEXT:    vpmovsxbw %xmm0, %ymm1
+; AVX512BW-NEXT:    vpmullw {{.*}}(%rip), %ymm1, %ymm1
+; AVX512BW-NEXT:    vpsrlw $8, %ymm1, %ymm1
+; AVX512BW-NEXT:    vpmovwb %zmm1, %ymm1
+; AVX512BW-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX512BW-NEXT:    vpsrlw $2, %xmm0, %xmm1
+; AVX512BW-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm2 = [32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32]
+; AVX512BW-NEXT:    vpxor %xmm2, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsubb %xmm2, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsrlw $7, %xmm0, %xmm0
+; AVX512BW-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
+; AVX512BW-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
   %res = sdiv <16 x i8> %a, <i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7>
   ret <16 x i8> %res
 }
 
 ;
 ; srem by 7
 ;
 

 ; AVX2NOBW-NEXT:    vzeroupper
 ; AVX2NOBW-NEXT:    retq
 ;
 ; AVX512BW-LABEL: test_rem7_16i8:
 ; AVX512BW:       # BB#0:
 ; AVX512BW-NEXT:    vpmovsxbw %xmm0, %ymm1
 ; AVX512BW-NEXT:    vpmullw {{.*}}(%rip), %ymm1, %ymm1
 ; AVX512BW-NEXT:    vpsrlw $8, %ymm1, %ymm1
-; AVX512BW-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX512BW-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpmovwb %zmm1, %ymm1
 ; AVX512BW-NEXT:    vpaddb %xmm0, %xmm1, %xmm1
 ; AVX512BW-NEXT:    vpsrlw $2, %xmm1, %xmm2
 ; AVX512BW-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
 ; AVX512BW-NEXT:    vmovdqa {{.*#+}} xmm3 = [32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32]
 ; AVX512BW-NEXT:    vpxor %xmm3, %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpsubb %xmm3, %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpsrlw $7, %xmm1, %xmm1
 ; AVX512BW-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1

test/CodeGen/X86/vector-idiv-sdiv-256.ll

 ; AVX2NOBW-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
 ; AVX2NOBW-NEXT:    retq
 ;
 ; AVX512BW-LABEL: test_div7_32i8:
 ; AVX512BW:       # BB#0:
 ; AVX512BW-NEXT:    vpmovsxbw %ymm0, %zmm1
 ; AVX512BW-NEXT:    vpmullw {{.*}}(%rip), %zmm1, %zmm1
 ; AVX512BW-NEXT:    vpsrlw $8, %zmm1, %zmm1
-; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,4,5,2,3,6,7]
-; AVX512BW-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
-; AVX512BW-NEXT:    vpackuswb %ymm2, %ymm1, %ymm1
+; AVX512BW-NEXT:    vpmovwb %zmm1, %ymm1
 ; AVX512BW-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
 ; AVX512BW-NEXT:    vpsrlw $2, %ymm0, %ymm1
 ; AVX512BW-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1
 ; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm2 = [32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32]
 ; AVX512BW-NEXT:    vpxor %ymm2, %ymm1, %ymm1
 ; AVX512BW-NEXT:    vpsubb %ymm2, %ymm1, %ymm1
 ; AVX512BW-NEXT:    vpsrlw $7, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpand {{.*}}(%rip), %ymm0, %ymm0

 ; AVX2NOBW-NEXT:    vpsubb %ymm1, %ymm0, %ymm0
 ; AVX2NOBW-NEXT:    retq
 ;
 ; AVX512BW-LABEL: test_rem7_32i8:
 ; AVX512BW:       # BB#0:
 ; AVX512BW-NEXT:    vpmovsxbw %ymm0, %zmm1
 ; AVX512BW-NEXT:    vpmullw {{.*}}(%rip), %zmm1, %zmm1
 ; AVX512BW-NEXT:    vpsrlw $8, %zmm1, %zmm1
-; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,4,5,2,3,6,7]
-; AVX512BW-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
-; AVX512BW-NEXT:    vpackuswb %ymm2, %ymm1, %ymm1
+; AVX512BW-NEXT:    vpmovwb %zmm1, %ymm1
 ; AVX512BW-NEXT:    vpaddb %ymm0, %ymm1, %ymm1
 ; AVX512BW-NEXT:    vpsrlw $2, %ymm1, %ymm2
 ; AVX512BW-NEXT:    vpand {{.*}}(%rip), %ymm2, %ymm2
 ; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm3 = [32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32]
 ; AVX512BW-NEXT:    vpxor %ymm3, %ymm2, %ymm2
 ; AVX512BW-NEXT:    vpsubb %ymm3, %ymm2, %ymm2
 ; AVX512BW-NEXT:    vpsrlw $7, %ymm1, %ymm1
 ; AVX512BW-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1

test/CodeGen/X86/vector-idiv-udiv-128.ll

 ; AVX1-NEXT:    vpsubb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrlw $1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrlw $2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: test_div7_16i8:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX2-NEXT:    vpmullw {{.*}}(%rip), %ymm1, %ymm1
-; AVX2-NEXT:    vpsrlw $8, %ymm1, %ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX2-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpsubb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpsrlw $1, %xmm0, %xmm0
-; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
-; AVX2-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpsrlw $2, %xmm0, %xmm0
-; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
+; AVX2NOBW-LABEL: test_div7_16i8:
+; AVX2NOBW:       # BB#0:
+; AVX2NOBW-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; AVX2NOBW-NEXT:    vpmullw {{.*}}(%rip), %ymm1, %ymm1
+; AVX2NOBW-NEXT:    vpsrlw $8, %ymm1, %ymm1
+; AVX2NOBW-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2NOBW-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
+; AVX2NOBW-NEXT:    vpsubb %xmm1, %xmm0, %xmm0
+; AVX2NOBW-NEXT:    vpsrlw $1, %xmm0, %xmm0
+; AVX2NOBW-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
+; AVX2NOBW-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
+; AVX2NOBW-NEXT:    vpsrlw $2, %xmm0, %xmm0
+; AVX2NOBW-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
+; AVX2NOBW-NEXT:    vzeroupper
+; AVX2NOBW-NEXT:    retq
+;
+; AVX512BW-LABEL: test_div7_16i8:
+; AVX512BW:       # BB#0:
+; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; AVX512BW-NEXT:    vpmullw {{.*}}(%rip), %ymm1, %ymm1
+; AVX512BW-NEXT:    vpsrlw $8, %ymm1, %ymm1
+; AVX512BW-NEXT:    vpmovwb %zmm1, %ymm1
+; AVX512BW-NEXT:    vpsubb %xmm1, %xmm0, %xmm0
+; AVX512BW-NEXT:    vpsrlw $1, %xmm0, %xmm0
+; AVX512BW-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
+; AVX512BW-NEXT:    vpaddb %xmm1, %xmm0, %xmm0
+; AVX512BW-NEXT:    vpsrlw $2, %xmm0, %xmm0
+; AVX512BW-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
   %res = udiv <16 x i8> %a, <i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7>
   ret <16 x i8> %res
 }
 
 ;
 ; urem by 7
 ;
 

 ; AVX2NOBW-NEXT:    vzeroupper
 ; AVX2NOBW-NEXT:    retq
 ;
 ; AVX512BW-LABEL: test_rem7_16i8:
 ; AVX512BW:       # BB#0:
 ; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
 ; AVX512BW-NEXT:    vpmullw {{.*}}(%rip), %ymm1, %ymm1
 ; AVX512BW-NEXT:    vpsrlw $8, %ymm1, %ymm1
-; AVX512BW-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX512BW-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpmovwb %zmm1, %ymm1
 ; AVX512BW-NEXT:    vpsubb %xmm1, %xmm0, %xmm2
 ; AVX512BW-NEXT:    vpsrlw $1, %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpand {{.*}}(%rip), %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpaddb %xmm1, %xmm2, %xmm1
 ; AVX512BW-NEXT:    vpsrlw $2, %xmm1, %xmm1
 ; AVX512BW-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
 ; AVX512BW-NEXT:    vpmovsxbw %xmm1, %ymm1
 ; AVX512BW-NEXT:    vpmullw {{.*}}(%rip), %ymm1, %ymm1
 ; AVX512BW-NEXT:    vpmovwb %zmm1, %ymm1
 ; AVX512BW-NEXT:    vpsubb %xmm1, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
   %res = urem <16 x i8> %a, <i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7>
   ret <16 x i8> %res
 }

test/CodeGen/X86/vector-idiv-udiv-256.ll

 ; AVX2NOBW-NEXT:    vpand {{.*}}(%rip), %ymm0, %ymm0
 ; AVX2NOBW-NEXT:    retq
 ;
 ; AVX512BW-LABEL: test_div7_32i8:
 ; AVX512BW:       # BB#0:
 ; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
 ; AVX512BW-NEXT:    vpmullw {{.*}}(%rip), %zmm1, %zmm1
 ; AVX512BW-NEXT:    vpsrlw $8, %zmm1, %zmm1
-; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,4,5,2,3,6,7]
-; AVX512BW-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
-; AVX512BW-NEXT:    vpackuswb %ymm2, %ymm1, %ymm1
+; AVX512BW-NEXT:    vpmovwb %zmm1, %ymm1
 ; AVX512BW-NEXT:    vpsubb %ymm1, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpsrlw $1, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpand {{.*}}(%rip), %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpaddb %ymm1, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpsrlw $2, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpand {{.*}}(%rip), %ymm0, %ymm0
 ; AVX512BW-NEXT:    retq
   %res = udiv <32 x i8> %a, <i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7>

 ; AVX2NOBW-NEXT:    vpsubb %ymm1, %ymm0, %ymm0
 ; AVX2NOBW-NEXT:    retq
 ;
 ; AVX512BW-LABEL: test_rem7_32i8:
 ; AVX512BW:       # BB#0:
 ; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
 ; AVX512BW-NEXT:    vpmullw {{.*}}(%rip), %zmm1, %zmm1
 ; AVX512BW-NEXT:    vpsrlw $8, %zmm1, %zmm1
-; AVX512BW-NEXT:    vshufi64x2 {{.*#+}} zmm1 = zmm1[0,1,4,5,2,3,6,7]
-; AVX512BW-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
-; AVX512BW-NEXT:    vpackuswb %ymm2, %ymm1, %ymm1
+; AVX512BW-NEXT:    vpmovwb %zmm1, %ymm1
 ; AVX512BW-NEXT:    vpsubb %ymm1, %ymm0, %ymm2
 ; AVX512BW-NEXT:    vpsrlw $1, %ymm2, %ymm2
 ; AVX512BW-NEXT:    vpand {{.*}}(%rip), %ymm2, %ymm2
 ; AVX512BW-NEXT:    vpaddb %ymm1, %ymm2, %ymm1
 ; AVX512BW-NEXT:    vpsrlw $2, %ymm1, %ymm1
 ; AVX512BW-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1
 ; AVX512BW-NEXT:    vpmovsxbw %ymm1, %zmm1
 ; AVX512BW-NEXT:    vpmullw {{.*}}(%rip), %zmm1, %zmm1
 ; AVX512BW-NEXT:    vpmovwb %zmm1, %ymm1
 ; AVX512BW-NEXT:    vpsubb %ymm1, %ymm0, %ymm0
 ; AVX512BW-NEXT:    retq
   %res = urem <32 x i8> %a, <i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7,i8 7, i8 7, i8 7, i8 7>
   ret <32 x i8> %res
 }