[LegalizeVectorOps][X86] Don't defer BITREVERSE expansion to LegalizeDAG.

[craig.topper]

Summary

By expanding early it allows the shifts to be custom lowered in
LegalizeVectorOps. Then a DAG combine is able to run on them before
LegalizeDAG handles the BUILD_VECTORS for the masks used.

v16Xi8 shift lowering on X86 requires a mask to be applied to a v8i16
shift. The BITREVERSE expansion applied an AND mask before SHL ops and
after SRL ops. This was done to share the same mask constant for both shifts.
It looks like this patch allows DAG combine to remove the AND mask added
after v16i8 SHL by X86 lowering. This maintains the mask sharing that
BITREVERSE was trying to achieve. Prior to this patch it looks like
we kept the mask after the SHL instead which required an extra constant
pool or a PANDN to invert it.

This is dependent on D112248 because RISCV will end up scalarizing the BSWAP
portion of the BITREVERSE expansion if we don't disable BSWAP scalarization in
LegalizeVectorOps first.

[RKSimon]

LGTM

llvm/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp

1124

Do we have any consistency on when we use a &Results vs just return the SDValue?

llvm/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp

if (TLI.isOperationLegalOrCustom(ISD::SHL, VT) &&
       TLI.isOperationLegalOrCustomOrPromote(ISD::OR, VT))
     return TLI.expandBSWAP(Node, DAG);
 
   // Otherwise unroll.
   return DAG.UnrollVectorOp(Node);
 }
 
 void VectorLegalizer::ExpandBITREVERSE(SDNode *Node,
                                        SmallVectorImpl<SDValue> &Results) {
   EVT VT = Node->getValueType(0);
 
   // If we have the scalar operation, it's probably cheaper to unroll it.
   if (TLI.isOperationLegalOrCustom(ISD::BITREVERSE, VT.getScalarType())) {
     SDValue Tmp = DAG.UnrollVectorOp(Node);
     Results.push_back(Tmp);
     return;
   }
if (ScalarSizeInBits > 8 && (ScalarSizeInBits % 8) == 0) {
     }
   }
 
   // If we have the appropriate vector bit operations, it is better to use them
   // than unrolling and expanding each component.
   if (TLI.isOperationLegalOrCustom(ISD::SHL, VT) &&
       TLI.isOperationLegalOrCustom(ISD::SRL, VT) &&
       TLI.isOperationLegalOrCustomOrPromote(ISD::AND, VT) &&
-      TLI.isOperationLegalOrCustomOrPromote(ISD::OR, VT))
-    // Let LegalizeDAG handle this later.
+      TLI.isOperationLegalOrCustomOrPromote(ISD::OR, VT)) {
+    Results.push_back(TLI.expandBITREVERSE(Node, DAG));
     return;
+  }
 
   // Otherwise unroll.
   SDValue Tmp = DAG.UnrollVectorOp(Node);
   Results.push_back(Tmp);
 }
 
 SDValue VectorLegalizer::ExpandVSELECT(SDNode *Node) {
   // Implement VSELECT in terms of XOR, AND, OR

llvm/test/CodeGen/X86/bitreverse.ll

 ;
 ; X64-LABEL: test_bitreverse_v2i16:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movdqa %xmm0, %xmm1
 ; X64-NEXT:    psrlw $8, %xmm1
 ; X64-NEXT:    psllw $8, %xmm0
 ; X64-NEXT:    por %xmm1, %xmm0
 ; X64-NEXT:    movdqa %xmm0, %xmm1
-; X64-NEXT:    psllw $4, %xmm1
-; X64-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
-; X64-NEXT:    psrlw $4, %xmm0
-; X64-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-NEXT:    psrlw $4, %xmm1
+; X64-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; X64-NEXT:    pand %xmm2, %xmm1
+; X64-NEXT:    pand %xmm2, %xmm0
+; X64-NEXT:    psllw $4, %xmm0
 ; X64-NEXT:    por %xmm1, %xmm0
 ; X64-NEXT:    movdqa %xmm0, %xmm1
 ; X64-NEXT:    psrlw $2, %xmm1
 ; X64-NEXT:    movdqa {{.*#+}} xmm2 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
 ; X64-NEXT:    pand %xmm2, %xmm1
 ; X64-NEXT:    pand %xmm2, %xmm0
 ; X64-NEXT:    psllw $2, %xmm0
 ; X64-NEXT:    por %xmm1, %xmm0

llvm/test/CodeGen/X86/combine-bitreverse.ll

 ; X86-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
 ; X86-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
 ; X86-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
 ; X86-NEXT:    punpcklbw {{.*#+}} xmm0 = 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]
 ; X86-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
 ; X86-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
 ; X86-NEXT:    packuswb %xmm2, %xmm0
 ; X86-NEXT:    movdqa %xmm0, %xmm1
-; X86-NEXT:    psllw $4, %xmm1
-; X86-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1
-; X86-NEXT:    psrlw $4, %xmm0
-; X86-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-NEXT:    psrlw $4, %xmm1
+; X86-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; X86-NEXT:    pand %xmm2, %xmm1
+; X86-NEXT:    pand %xmm2, %xmm0
+; X86-NEXT:    psllw $4, %xmm0
 ; X86-NEXT:    por %xmm1, %xmm0
 ; X86-NEXT:    movdqa %xmm0, %xmm1
 ; X86-NEXT:    psrlw $2, %xmm1
 ; X86-NEXT:    movdqa {{.*#+}} xmm2 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
 ; X86-NEXT:    pand %xmm2, %xmm1
 ; X86-NEXT:    pand %xmm2, %xmm0
 ; X86-NEXT:    psllw $2, %xmm0
 ; X86-NEXT:    por %xmm1, %xmm0

llvm/test/CodeGen/X86/vector-bitreverse.ll

; GFNIAVX512BW-NEXT:    retq
   %b = call i64 @llvm.bitreverse.i64(i64 %a)
   ret i64 %b
 }
 
 define <16 x i8> @test_bitreverse_v16i8(<16 x i8> %a) nounwind {
 ; SSE2-LABEL: test_bitreverse_v16i8:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    psllw $4, %xmm1
-; SSE2-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
-; SSE2-NEXT:    psrlw $4, %xmm0
-; SSE2-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE2-NEXT:    pand %xmm2, %xmm1
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    psllw $4, %xmm0
 ; SSE2-NEXT:    por %xmm1, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; SSE2-NEXT:    psrlw $2, %xmm1
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
 ; SSE2-NEXT:    pand %xmm2, %xmm1
 ; SSE2-NEXT:    pand %xmm2, %xmm0
 ; SSE2-NEXT:    psllw $2, %xmm0
 ; SSE2-NEXT:    por %xmm1, %xmm0

 define <8 x i16> @test_bitreverse_v8i16(<8 x i16> %a) nounwind {
 ; SSE2-LABEL: test_bitreverse_v8i16:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; SSE2-NEXT:    psrlw $8, %xmm1
 ; SSE2-NEXT:    psllw $8, %xmm0
 ; SSE2-NEXT:    por %xmm1, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    psllw $4, %xmm1
-; SSE2-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
-; SSE2-NEXT:    psrlw $4, %xmm0
-; SSE2-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE2-NEXT:    pand %xmm2, %xmm1
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    psllw $4, %xmm0
 ; SSE2-NEXT:    por %xmm1, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; SSE2-NEXT:    psrlw $2, %xmm1
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
 ; SSE2-NEXT:    pand %xmm2, %xmm1
 ; SSE2-NEXT:    pand %xmm2, %xmm0
 ; SSE2-NEXT:    psllw $2, %xmm0
 ; SSE2-NEXT:    por %xmm1, %xmm0

 ; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = 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]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
 ; SSE2-NEXT:    packuswb %xmm2, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    psllw $4, %xmm1
-; SSE2-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
-; SSE2-NEXT:    psrlw $4, %xmm0
-; SSE2-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE2-NEXT:    pand %xmm2, %xmm1
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    psllw $4, %xmm0
 ; SSE2-NEXT:    por %xmm1, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; SSE2-NEXT:    psrlw $2, %xmm1
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
 ; SSE2-NEXT:    pand %xmm2, %xmm1
 ; SSE2-NEXT:    pand %xmm2, %xmm0
 ; SSE2-NEXT:    psllw $2, %xmm0
 ; SSE2-NEXT:    por %xmm1, %xmm0

 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = 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]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
 ; SSE2-NEXT:    packuswb %xmm2, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    psllw $4, %xmm1
-; SSE2-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
-; SSE2-NEXT:    psrlw $4, %xmm0
-; SSE2-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE2-NEXT:    pand %xmm2, %xmm1
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    psllw $4, %xmm0
 ; SSE2-NEXT:    por %xmm1, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; SSE2-NEXT:    psrlw $2, %xmm1
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
 ; SSE2-NEXT:    pand %xmm2, %xmm1
 ; SSE2-NEXT:    pand %xmm2, %xmm0
 ; SSE2-NEXT:    psllw $2, %xmm0
 ; SSE2-NEXT:    por %xmm1, %xmm0

 ; GFNIAVX512BW-NEXT:    retq
   %b = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %a)
   ret <2 x i64> %b
 }
 
 define <32 x i8> @test_bitreverse_v32i8(<32 x i8> %a) nounwind {
 ; SSE2-LABEL: test_bitreverse_v32i8:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; SSE2-NEXT:    movdqa %xmm0, %xmm3
-; SSE2-NEXT:    psllw $4, %xmm3
-; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; SSE2-NEXT:    movdqa %xmm1, %xmm4
-; SSE2-NEXT:    pandn %xmm3, %xmm4
-; SSE2-NEXT:    psrlw $4, %xmm0
-; SSE2-NEXT:    pand %xmm1, %xmm0
-; SSE2-NEXT:    por %xmm4, %xmm0
+; SSE2-NEXT:    psrlw $4, %xmm3
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE2-NEXT:    pand %xmm2, %xmm3
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    psllw $4, %xmm0
+; SSE2-NEXT:    por %xmm3, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm4
 ; SSE2-NEXT:    psrlw $2, %xmm4
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
 ; SSE2-NEXT:    pand %xmm3, %xmm4
 ; SSE2-NEXT:    pand %xmm3, %xmm0
 ; SSE2-NEXT:    psllw $2, %xmm0
 ; SSE2-NEXT:    por %xmm4, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm5
 ; SSE2-NEXT:    psrlw $1, %xmm5
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [85,85,85,85,85,85,85,85,85,85,85,85,85,85,85,85]
 ; SSE2-NEXT:    pand %xmm4, %xmm5
 ; SSE2-NEXT:    pand %xmm4, %xmm0
 ; SSE2-NEXT:    paddb %xmm0, %xmm0
 ; SSE2-NEXT:    por %xmm5, %xmm0
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
-; SSE2-NEXT:    psrlw $4, %xmm2
-; SSE2-NEXT:    pand %xmm1, %xmm2
-; SSE2-NEXT:    pandn %xmm5, %xmm1
-; SSE2-NEXT:    por %xmm2, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm5
+; SSE2-NEXT:    psrlw $4, %xmm5
+; SSE2-NEXT:    pand %xmm2, %xmm5
+; SSE2-NEXT:    pand %xmm2, %xmm1
+; SSE2-NEXT:    psllw $4, %xmm1
+; SSE2-NEXT:    por %xmm5, %xmm1
 ; SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; SSE2-NEXT:    psrlw $2, %xmm2
 ; SSE2-NEXT:    pand %xmm3, %xmm2
 ; SSE2-NEXT:    pand %xmm3, %xmm1
 ; SSE2-NEXT:    psllw $2, %xmm1
 ; SSE2-NEXT:    por %xmm2, %xmm1
 ; SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; SSE2-NEXT:    psrlw $1, %xmm2

 ; GFNIAVX512BW-NEXT:    retq
   %b = call <32 x i8> @llvm.bitreverse.v32i8(<32 x i8> %a)
   ret <32 x i8> %b
 }
 
 define <16 x i16> @test_bitreverse_v16i16(<16 x i16> %a) nounwind {
 ; SSE2-LABEL: test_bitreverse_v16i16:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm1, %xmm2
-; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    psrlw $8, %xmm1
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
 ; SSE2-NEXT:    psllw $8, %xmm0
-; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    por %xmm2, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm3
-; SSE2-NEXT:    psllw $4, %xmm3
-; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; SSE2-NEXT:    movdqa %xmm1, %xmm4
-; SSE2-NEXT:    pandn %xmm3, %xmm4
-; SSE2-NEXT:    psrlw $4, %xmm0
-; SSE2-NEXT:    pand %xmm1, %xmm0
-; SSE2-NEXT:    por %xmm4, %xmm0
+; SSE2-NEXT:    psrlw $4, %xmm3
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE2-NEXT:    pand %xmm2, %xmm3
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    psllw $4, %xmm0
+; SSE2-NEXT:    por %xmm3, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm4
 ; SSE2-NEXT:    psrlw $2, %xmm4
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
 ; SSE2-NEXT:    pand %xmm3, %xmm4
 ; SSE2-NEXT:    pand %xmm3, %xmm0
 ; SSE2-NEXT:    psllw $2, %xmm0
 ; SSE2-NEXT:    por %xmm4, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm5
 ; SSE2-NEXT:    psrlw $1, %xmm5
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [85,85,85,85,85,85,85,85,85,85,85,85,85,85,85,85]
 ; SSE2-NEXT:    pand %xmm4, %xmm5
 ; SSE2-NEXT:    pand %xmm4, %xmm0
 ; SSE2-NEXT:    paddb %xmm0, %xmm0
 ; SSE2-NEXT:    por %xmm5, %xmm0
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
+; SSE2-NEXT:    movdqa %xmm1, %xmm5
 ; SSE2-NEXT:    psrlw $8, %xmm5
-; SSE2-NEXT:    psllw $8, %xmm2
-; SSE2-NEXT:    por %xmm5, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
-; SSE2-NEXT:    psrlw $4, %xmm2
-; SSE2-NEXT:    pand %xmm1, %xmm2
-; SSE2-NEXT:    pandn %xmm5, %xmm1
-; SSE2-NEXT:    por %xmm2, %xmm1
+; SSE2-NEXT:    psllw $8, %xmm1
+; SSE2-NEXT:    por %xmm5, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm5
+; SSE2-NEXT:    psrlw $4, %xmm5
+; SSE2-NEXT:    pand %xmm2, %xmm5
+; SSE2-NEXT:    pand %xmm2, %xmm1
+; SSE2-NEXT:    psllw $4, %xmm1
+; SSE2-NEXT:    por %xmm5, %xmm1
 ; SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; SSE2-NEXT:    psrlw $2, %xmm2
 ; SSE2-NEXT:    pand %xmm3, %xmm2
 ; SSE2-NEXT:    pand %xmm3, %xmm1
 ; SSE2-NEXT:    psllw $2, %xmm1
 ; SSE2-NEXT:    por %xmm2, %xmm1
 ; SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; SSE2-NEXT:    psrlw $1, %xmm2

 ; GFNIAVX512BW-NEXT:    retq
   %b = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> %a)
   ret <16 x i16> %b
 }
 
 define <8 x i32> @test_bitreverse_v8i32(<8 x i32> %a) nounwind {
 ; SSE2-LABEL: test_bitreverse_v8i32:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm1, %xmm2
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = 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]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE2-NEXT:    pxor %xmm2, %xmm2
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,2,1,0,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = 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]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    packuswb %xmm1, %xmm0
+; SSE2-NEXT:    packuswb %xmm3, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm4
-; SSE2-NEXT:    psllw $4, %xmm4
-; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    pandn %xmm4, %xmm5
-; SSE2-NEXT:    psrlw $4, %xmm0
-; SSE2-NEXT:    pand %xmm1, %xmm0
-; SSE2-NEXT:    por %xmm5, %xmm0
+; SSE2-NEXT:    psrlw $4, %xmm4
+; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE2-NEXT:    pand %xmm3, %xmm4
+; SSE2-NEXT:    pand %xmm3, %xmm0
+; SSE2-NEXT:    psllw $4, %xmm0
+; SSE2-NEXT:    por %xmm4, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm5
 ; SSE2-NEXT:    psrlw $2, %xmm5
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
 ; SSE2-NEXT:    pand %xmm4, %xmm5
 ; SSE2-NEXT:    pand %xmm4, %xmm0
 ; SSE2-NEXT:    psllw $2, %xmm0
 ; SSE2-NEXT:    por %xmm5, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm6
 ; SSE2-NEXT:    psrlw $1, %xmm6
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [85,85,85,85,85,85,85,85,85,85,85,85,85,85,85,85]
 ; SSE2-NEXT:    pand %xmm5, %xmm6
 ; SSE2-NEXT:    pand %xmm5, %xmm0
 ; SSE2-NEXT:    paddb %xmm0, %xmm0
 ; SSE2-NEXT:    por %xmm6, %xmm0
-; SSE2-NEXT:    movdqa %xmm2, %xmm6
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm3[8],xmm6[9],xmm3[9],xmm6[10],xmm3[10],xmm6[11],xmm3[11],xmm6[12],xmm3[12],xmm6[13],xmm3[13],xmm6[14],xmm3[14],xmm6[15],xmm3[15]
+; SSE2-NEXT:    movdqa %xmm1, %xmm6
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm2[8],xmm6[9],xmm2[9],xmm6[10],xmm2[10],xmm6[11],xmm2[11],xmm6[12],xmm2[12],xmm6[13],xmm2[13],xmm6[14],xmm2[14],xmm6[15],xmm2[15]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = 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]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    packuswb %xmm6, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    psllw $4, %xmm3
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
+; SSE2-NEXT:    packuswb %xmm6, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; SSE2-NEXT:    psrlw $4, %xmm2
-; SSE2-NEXT:    pand %xmm1, %xmm2
-; SSE2-NEXT:    pandn %xmm3, %xmm1
+; SSE2-NEXT:    pand %xmm3, %xmm2
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    psllw $4, %xmm1
 ; SSE2-NEXT:    por %xmm2, %xmm1
 ; SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; SSE2-NEXT:    psrlw $2, %xmm2
 ; SSE2-NEXT:    pand %xmm4, %xmm2
 ; SSE2-NEXT:    pand %xmm4, %xmm1
 ; SSE2-NEXT:    psllw $2, %xmm1
 ; SSE2-NEXT:    por %xmm2, %xmm1
 ; SSE2-NEXT:    movdqa %xmm1, %xmm2

 ; GFNIAVX512BW-NEXT:    retq
   %b = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> %a)
   ret <8 x i32> %b
 }
 
 define <4 x i64> @test_bitreverse_v4i64(<4 x i64> %a) nounwind {
 ; SSE2-LABEL: test_bitreverse_v4i64:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm1, %xmm2
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = 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]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE2-NEXT:    pxor %xmm2, %xmm2
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,0,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,2,1,0,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = 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]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    packuswb %xmm1, %xmm0
+; SSE2-NEXT:    packuswb %xmm3, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm4
-; SSE2-NEXT:    psllw $4, %xmm4
-; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    pandn %xmm4, %xmm5
-; SSE2-NEXT:    psrlw $4, %xmm0
-; SSE2-NEXT:    pand %xmm1, %xmm0
-; SSE2-NEXT:    por %xmm5, %xmm0
+; SSE2-NEXT:    psrlw $4, %xmm4
+; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE2-NEXT:    pand %xmm3, %xmm4
+; SSE2-NEXT:    pand %xmm3, %xmm0
+; SSE2-NEXT:    psllw $4, %xmm0
+; SSE2-NEXT:    por %xmm4, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm5
 ; SSE2-NEXT:    psrlw $2, %xmm5
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
 ; SSE2-NEXT:    pand %xmm4, %xmm5
 ; SSE2-NEXT:    pand %xmm4, %xmm0
 ; SSE2-NEXT:    psllw $2, %xmm0
 ; SSE2-NEXT:    por %xmm5, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm6
 ; SSE2-NEXT:    psrlw $1, %xmm6
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [85,85,85,85,85,85,85,85,85,85,85,85,85,85,85,85]
 ; SSE2-NEXT:    pand %xmm5, %xmm6
 ; SSE2-NEXT:    pand %xmm5, %xmm0
 ; SSE2-NEXT:    paddb %xmm0, %xmm0
 ; SSE2-NEXT:    por %xmm6, %xmm0
-; SSE2-NEXT:    movdqa %xmm2, %xmm6
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm3[8],xmm6[9],xmm3[9],xmm6[10],xmm3[10],xmm6[11],xmm3[11],xmm6[12],xmm3[12],xmm6[13],xmm3[13],xmm6[14],xmm3[14],xmm6[15],xmm3[15]
+; SSE2-NEXT:    movdqa %xmm1, %xmm6
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm2[8],xmm6[9],xmm2[9],xmm6[10],xmm2[10],xmm6[11],xmm2[11],xmm6[12],xmm2[12],xmm6[13],xmm2[13],xmm6[14],xmm2[14],xmm6[15],xmm2[15]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,3,0,1]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = 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]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,0,1]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    packuswb %xmm6, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    psllw $4, %xmm3
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
+; SSE2-NEXT:    packuswb %xmm6, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; SSE2-NEXT:    psrlw $4, %xmm2
-; SSE2-NEXT:    pand %xmm1, %xmm2
-; SSE2-NEXT:    pandn %xmm3, %xmm1
+; SSE2-NEXT:    pand %xmm3, %xmm2
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    psllw $4, %xmm1
 ; SSE2-NEXT:    por %xmm2, %xmm1
 ; SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; SSE2-NEXT:    psrlw $2, %xmm2
 ; SSE2-NEXT:    pand %xmm4, %xmm2
 ; SSE2-NEXT:    pand %xmm4, %xmm1
 ; SSE2-NEXT:    psllw $2, %xmm1
 ; SSE2-NEXT:    por %xmm2, %xmm1
 ; SSE2-NEXT:    movdqa %xmm1, %xmm2

 ; GFNIAVX512BW-NEXT:    retq
   %b = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> %a)
   ret <4 x i64> %b
 }
 
 define <64 x i8> @test_bitreverse_v64i8(<64 x i8> %a) nounwind {
 ; SSE2-LABEL: test_bitreverse_v64i8:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    movdqa %xmm0, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
-; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; SSE2-NEXT:    movdqa %xmm3, %xmm6
-; SSE2-NEXT:    pandn %xmm5, %xmm6
-; SSE2-NEXT:    psrlw $4, %xmm0
-; SSE2-NEXT:    pand %xmm3, %xmm0
-; SSE2-NEXT:    por %xmm6, %xmm0
+; SSE2-NEXT:    psrlw $4, %xmm5
+; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE2-NEXT:    pand %xmm4, %xmm5
+; SSE2-NEXT:    pand %xmm4, %xmm0
+; SSE2-NEXT:    psllw $4, %xmm0
+; SSE2-NEXT:    por %xmm5, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm6
 ; SSE2-NEXT:    psrlw $2, %xmm6
-; SSE2-NEXT:    movdqa {{.*#+}} xmm8 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
-; SSE2-NEXT:    pand %xmm8, %xmm6
-; SSE2-NEXT:    pand %xmm8, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    pand %xmm5, %xmm6
+; SSE2-NEXT:    pand %xmm5, %xmm0
 ; SSE2-NEXT:    psllw $2, %xmm0
 ; SSE2-NEXT:    por %xmm6, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm7
 ; SSE2-NEXT:    psrlw $1, %xmm7
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm6 = [85,85,85,85,85,85,85,85,85,85,85,85,85,85,85,85]
 ; SSE2-NEXT:    pand %xmm6, %xmm7
 ; SSE2-NEXT:    pand %xmm6, %xmm0
 ; SSE2-NEXT:    paddb %xmm0, %xmm0
 ; SSE2-NEXT:    por %xmm7, %xmm0
 ; SSE2-NEXT:    movdqa %xmm1, %xmm7
-; SSE2-NEXT:    psllw $4, %xmm7
-; SSE2-NEXT:    movdqa %xmm3, %xmm5
-; SSE2-NEXT:    pandn %xmm7, %xmm5
-; SSE2-NEXT:    psrlw $4, %xmm1
-; SSE2-NEXT:    pand %xmm3, %xmm1
-; SSE2-NEXT:    por %xmm5, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    psrlw $2, %xmm5
-; SSE2-NEXT:    pand %xmm8, %xmm5
-; SSE2-NEXT:    pand %xmm8, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm7
+; SSE2-NEXT:    pand %xmm4, %xmm7
+; SSE2-NEXT:    pand %xmm4, %xmm1
+; SSE2-NEXT:    psllw $4, %xmm1
+; SSE2-NEXT:    por %xmm7, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm7
+; SSE2-NEXT:    psrlw $2, %xmm7
+; SSE2-NEXT:    pand %xmm5, %xmm7
+; SSE2-NEXT:    pand %xmm5, %xmm1
 ; SSE2-NEXT:    psllw $2, %xmm1
-; SSE2-NEXT:    por %xmm5, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    psrlw $1, %xmm5
-; SSE2-NEXT:    pand %xmm6, %xmm5
+; SSE2-NEXT:    por %xmm7, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm7
+; SSE2-NEXT:    psrlw $1, %xmm7
+; SSE2-NEXT:    pand %xmm6, %xmm7
 ; SSE2-NEXT:    pand %xmm6, %xmm1
 ; SSE2-NEXT:    paddb %xmm1, %xmm1
-; SSE2-NEXT:    por %xmm5, %xmm1
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
-; SSE2-NEXT:    movdqa %xmm3, %xmm7
-; SSE2-NEXT:    pandn %xmm5, %xmm7
-; SSE2-NEXT:    psrlw $4, %xmm2
-; SSE2-NEXT:    pand %xmm3, %xmm2
+; SSE2-NEXT:    por %xmm7, %xmm1
+; SSE2-NEXT:    movdqa %xmm2, %xmm7
+; SSE2-NEXT:    psrlw $4, %xmm7
+; SSE2-NEXT:    pand %xmm4, %xmm7
+; SSE2-NEXT:    pand %xmm4, %xmm2
+; SSE2-NEXT:    psllw $4, %xmm2
 ; SSE2-NEXT:    por %xmm7, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psrlw $2, %xmm5
-; SSE2-NEXT:    pand %xmm8, %xmm5
-; SSE2-NEXT:    pand %xmm8, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm7
+; SSE2-NEXT:    psrlw $2, %xmm7
+; SSE2-NEXT:    pand %xmm5, %xmm7
+; SSE2-NEXT:    pand %xmm5, %xmm2
 ; SSE2-NEXT:    psllw $2, %xmm2
-; SSE2-NEXT:    por %xmm5, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psrlw $1, %xmm5
-; SSE2-NEXT:    pand %xmm6, %xmm5
+; SSE2-NEXT:    por %xmm7, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm7
+; SSE2-NEXT:    psrlw $1, %xmm7
+; SSE2-NEXT:    pand %xmm6, %xmm7
 ; SSE2-NEXT:    pand %xmm6, %xmm2
 ; SSE2-NEXT:    paddb %xmm2, %xmm2
-; SSE2-NEXT:    por %xmm5, %xmm2
-; SSE2-NEXT:    movdqa %xmm4, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
-; SSE2-NEXT:    psrlw $4, %xmm4
-; SSE2-NEXT:    pand %xmm3, %xmm4
-; SSE2-NEXT:    pandn %xmm5, %xmm3
-; SSE2-NEXT:    por %xmm4, %xmm3
+; SSE2-NEXT:    por %xmm7, %xmm2
+; SSE2-NEXT:    movdqa %xmm3, %xmm7
+; SSE2-NEXT:    psrlw $4, %xmm7
+; SSE2-NEXT:    pand %xmm4, %xmm7
+; SSE2-NEXT:    pand %xmm4, %xmm3
+; SSE2-NEXT:    psllw $4, %xmm3
+; SSE2-NEXT:    por %xmm7, %xmm3
 ; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    psrlw $2, %xmm4
-; SSE2-NEXT:    pand %xmm8, %xmm4
-; SSE2-NEXT:    pand %xmm8, %xmm3
+; SSE2-NEXT:    pand %xmm5, %xmm4
+; SSE2-NEXT:    pand %xmm5, %xmm3
 ; SSE2-NEXT:    psllw $2, %xmm3
 ; SSE2-NEXT:    por %xmm4, %xmm3
 ; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    psrlw $1, %xmm4
 ; SSE2-NEXT:    pand %xmm6, %xmm4
 ; SSE2-NEXT:    pand %xmm6, %xmm3
 ; SSE2-NEXT:    paddb %xmm3, %xmm3
 ; SSE2-NEXT:    por %xmm4, %xmm3

 ; GFNIAVX512BW-NEXT:    retq
   %b = call <64 x i8> @llvm.bitreverse.v64i8(<64 x i8> %a)
   ret <64 x i8> %b
 }
 
 define <32 x i16> @test_bitreverse_v32i16(<32 x i16> %a) nounwind {
 ; SSE2-LABEL: test_bitreverse_v32i16:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm3, %xmm4
-; SSE2-NEXT:    movdqa %xmm0, %xmm3
-; SSE2-NEXT:    psrlw $8, %xmm3
+; SSE2-NEXT:    movdqa %xmm0, %xmm4
+; SSE2-NEXT:    psrlw $8, %xmm4
 ; SSE2-NEXT:    psllw $8, %xmm0
-; SSE2-NEXT:    por %xmm3, %xmm0
+; SSE2-NEXT:    por %xmm4, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
-; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; SSE2-NEXT:    movdqa %xmm3, %xmm6
-; SSE2-NEXT:    pandn %xmm5, %xmm6
-; SSE2-NEXT:    psrlw $4, %xmm0
-; SSE2-NEXT:    pand %xmm3, %xmm0
-; SSE2-NEXT:    por %xmm6, %xmm0
+; SSE2-NEXT:    psrlw $4, %xmm5
+; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE2-NEXT:    pand %xmm4, %xmm5
+; SSE2-NEXT:    pand %xmm4, %xmm0
+; SSE2-NEXT:    psllw $4, %xmm0
+; SSE2-NEXT:    por %xmm5, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm6
 ; SSE2-NEXT:    psrlw $2, %xmm6
-; SSE2-NEXT:    movdqa {{.*#+}} xmm8 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
-; SSE2-NEXT:    pand %xmm8, %xmm6
-; SSE2-NEXT:    pand %xmm8, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    pand %xmm5, %xmm6
+; SSE2-NEXT:    pand %xmm5, %xmm0
 ; SSE2-NEXT:    psllw $2, %xmm0
 ; SSE2-NEXT:    por %xmm6, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm7
 ; SSE2-NEXT:    psrlw $1, %xmm7
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm6 = [85,85,85,85,85,85,85,85,85,85,85,85,85,85,85,85]
 ; SSE2-NEXT:    pand %xmm6, %xmm7
 ; SSE2-NEXT:    pand %xmm6, %xmm0
 ; SSE2-NEXT:    paddb %xmm0, %xmm0
 ; SSE2-NEXT:    por %xmm7, %xmm0
 ; SSE2-NEXT:    movdqa %xmm1, %xmm7
 ; SSE2-NEXT:    psrlw $8, %xmm7
 ; SSE2-NEXT:    psllw $8, %xmm1
 ; SSE2-NEXT:    por %xmm7, %xmm1
 ; SSE2-NEXT:    movdqa %xmm1, %xmm7
-; SSE2-NEXT:    psllw $4, %xmm7
-; SSE2-NEXT:    movdqa %xmm3, %xmm5
-; SSE2-NEXT:    pandn %xmm7, %xmm5
-; SSE2-NEXT:    psrlw $4, %xmm1
-; SSE2-NEXT:    pand %xmm3, %xmm1
-; SSE2-NEXT:    por %xmm5, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    psrlw $2, %xmm5
-; SSE2-NEXT:    pand %xmm8, %xmm5
-; SSE2-NEXT:    pand %xmm8, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm7
+; SSE2-NEXT:    pand %xmm4, %xmm7
+; SSE2-NEXT:    pand %xmm4, %xmm1
+; SSE2-NEXT:    psllw $4, %xmm1
+; SSE2-NEXT:    por %xmm7, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm7
+; SSE2-NEXT:    psrlw $2, %xmm7
+; SSE2-NEXT:    pand %xmm5, %xmm7
+; SSE2-NEXT:    pand %xmm5, %xmm1
 ; SSE2-NEXT:    psllw $2, %xmm1
-; SSE2-NEXT:    por %xmm5, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    psrlw $1, %xmm5
-; SSE2-NEXT:    pand %xmm6, %xmm5
+; SSE2-NEXT:    por %xmm7, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm7
+; SSE2-NEXT:    psrlw $1, %xmm7
+; SSE2-NEXT:    pand %xmm6, %xmm7
 ; SSE2-NEXT:    pand %xmm6, %xmm1
 ; SSE2-NEXT:    paddb %xmm1, %xmm1
-; SSE2-NEXT:    por %xmm5, %xmm1
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psrlw $8, %xmm5
+; SSE2-NEXT:    por %xmm7, %xmm1
+; SSE2-NEXT:    movdqa %xmm2, %xmm7
+; SSE2-NEXT:    psrlw $8, %xmm7
 ; SSE2-NEXT:    psllw $8, %xmm2
-; SSE2-NEXT:    por %xmm5, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
-; SSE2-NEXT:    movdqa %xmm3, %xmm7
-; SSE2-NEXT:    pandn %xmm5, %xmm7
-; SSE2-NEXT:    psrlw $4, %xmm2
-; SSE2-NEXT:    pand %xmm3, %xmm2
 ; SSE2-NEXT:    por %xmm7, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psrlw $2, %xmm5
-; SSE2-NEXT:    pand %xmm8, %xmm5
-; SSE2-NEXT:    pand %xmm8, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm7
+; SSE2-NEXT:    psrlw $4, %xmm7
+; SSE2-NEXT:    pand %xmm4, %xmm7
+; SSE2-NEXT:    pand %xmm4, %xmm2
+; SSE2-NEXT:    psllw $4, %xmm2
+; SSE2-NEXT:    por %xmm7, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm7
+; SSE2-NEXT:    psrlw $2, %xmm7
+; SSE2-NEXT:    pand %xmm5, %xmm7
+; SSE2-NEXT:    pand %xmm5, %xmm2
 ; SSE2-NEXT:    psllw $2, %xmm2
-; SSE2-NEXT:    por %xmm5, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psrlw $1, %xmm5
-; SSE2-NEXT:    pand %xmm6, %xmm5
+; SSE2-NEXT:    por %xmm7, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm7
+; SSE2-NEXT:    psrlw $1, %xmm7
+; SSE2-NEXT:    pand %xmm6, %xmm7
 ; SSE2-NEXT:    pand %xmm6, %xmm2
 ; SSE2-NEXT:    paddb %xmm2, %xmm2
-; SSE2-NEXT:    por %xmm5, %xmm2
-; SSE2-NEXT:    movdqa %xmm4, %xmm5
-; SSE2-NEXT:    psrlw $8, %xmm5
-; SSE2-NEXT:    psllw $8, %xmm4
-; SSE2-NEXT:    por %xmm5, %xmm4
-; SSE2-NEXT:    movdqa %xmm4, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
-; SSE2-NEXT:    psrlw $4, %xmm4
-; SSE2-NEXT:    pand %xmm3, %xmm4
-; SSE2-NEXT:    pandn %xmm5, %xmm3
-; SSE2-NEXT:    por %xmm4, %xmm3
+; SSE2-NEXT:    por %xmm7, %xmm2
+; SSE2-NEXT:    movdqa %xmm3, %xmm7
+; SSE2-NEXT:    psrlw $8, %xmm7
+; SSE2-NEXT:    psllw $8, %xmm3
+; SSE2-NEXT:    por %xmm7, %xmm3
+; SSE2-NEXT:    movdqa %xmm3, %xmm7
+; SSE2-NEXT:    psrlw $4, %xmm7
+; SSE2-NEXT:    pand %xmm4, %xmm7
+; SSE2-NEXT:    pand %xmm4, %xmm3
+; SSE2-NEXT:    psllw $4, %xmm3
+; SSE2-NEXT:    por %xmm7, %xmm3
 ; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    psrlw $2, %xmm4
-; SSE2-NEXT:    pand %xmm8, %xmm4
-; SSE2-NEXT:    pand %xmm8, %xmm3
+; SSE2-NEXT:    pand %xmm5, %xmm4
+; SSE2-NEXT:    pand %xmm5, %xmm3
 ; SSE2-NEXT:    psllw $2, %xmm3
 ; SSE2-NEXT:    por %xmm4, %xmm3
 ; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    psrlw $1, %xmm4
 ; SSE2-NEXT:    pand %xmm6, %xmm4
 ; SSE2-NEXT:    pand %xmm6, %xmm3
 ; SSE2-NEXT:    paddb %xmm3, %xmm3
 ; SSE2-NEXT:    por %xmm4, %xmm3

 ; GFNIAVX512BW-NEXT:    retq
   %b = call <32 x i16> @llvm.bitreverse.v32i16(<32 x i16> %a)
   ret <32 x i16> %b
 }
 
 define <16 x i32> @test_bitreverse_v16i32(<16 x i32> %a) nounwind {
 ; SSE2-LABEL: test_bitreverse_v16i32:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    pxor %xmm8, %xmm8
-; SSE2-NEXT:    movdqa %xmm0, %xmm3
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm8[8],xmm3[9],xmm8[9],xmm3[10],xmm8[10],xmm3[11],xmm8[11],xmm3[12],xmm8[12],xmm3[13],xmm8[13],xmm3[14],xmm8[14],xmm3[15],xmm8[15]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,2,1,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; SSE2-NEXT:    movdqa %xmm0, %xmm5
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm8[8],xmm5[9],xmm8[9],xmm5[10],xmm8[10],xmm5[11],xmm8[11],xmm5[12],xmm8[12],xmm5[13],xmm8[13],xmm5[14],xmm8[14],xmm5[15],xmm8[15]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,2,1,0,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,5,4]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    packuswb %xmm3, %xmm0
+; SSE2-NEXT:    packuswb %xmm5, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm6
-; SSE2-NEXT:    psllw $4, %xmm6
-; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; SSE2-NEXT:    movdqa %xmm3, %xmm7
-; SSE2-NEXT:    pandn %xmm6, %xmm7
-; SSE2-NEXT:    psrlw $4, %xmm0
-; SSE2-NEXT:    pand %xmm3, %xmm0
-; SSE2-NEXT:    por %xmm7, %xmm0
+; SSE2-NEXT:    psrlw $4, %xmm6
+; SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE2-NEXT:    pand %xmm5, %xmm6
+; SSE2-NEXT:    pand %xmm5, %xmm0
+; SSE2-NEXT:    psllw $4, %xmm0
+; SSE2-NEXT:    por %xmm6, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm7
 ; SSE2-NEXT:    psrlw $2, %xmm7
-; SSE2-NEXT:    movdqa {{.*#+}} xmm9 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
-; SSE2-NEXT:    pand %xmm9, %xmm7
-; SSE2-NEXT:    pand %xmm9, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm6 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    pand %xmm6, %xmm7
+; SSE2-NEXT:    pand %xmm6, %xmm0
 ; SSE2-NEXT:    psllw $2, %xmm0
 ; SSE2-NEXT:    por %xmm7, %xmm0
-; SSE2-NEXT:    movdqa %xmm0, %xmm5
-; SSE2-NEXT:    psrlw $1, %xmm5
+; SSE2-NEXT:    movdqa %xmm0, %xmm4
+; SSE2-NEXT:    psrlw $1, %xmm4
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm7 = [85,85,85,85,85,85,85,85,85,85,85,85,85,85,85,85]
-; SSE2-NEXT:    pand %xmm7, %xmm5
+; SSE2-NEXT:    pand %xmm7, %xmm4
 ; SSE2-NEXT:    pand %xmm7, %xmm0
 ; SSE2-NEXT:    paddb %xmm0, %xmm0
-; SSE2-NEXT:    por %xmm5, %xmm0
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm8[8],xmm5[9],xmm8[9],xmm5[10],xmm8[10],xmm5[11],xmm8[11],xmm5[12],xmm8[12],xmm5[13],xmm8[13],xmm5[14],xmm8[14],xmm5[15],xmm8[15]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,2,1,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,5,4]
+; SSE2-NEXT:    por %xmm4, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm8[8],xmm4[9],xmm8[9],xmm4[10],xmm8[10],xmm4[11],xmm8[11],xmm4[12],xmm8[12],xmm4[13],xmm8[13],xmm4[14],xmm8[14],xmm4[15],xmm8[15]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,2,1,0,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,6,5,4]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    packuswb %xmm5, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
-; SSE2-NEXT:    movdqa %xmm3, %xmm6
-; SSE2-NEXT:    pandn %xmm5, %xmm6
-; SSE2-NEXT:    psrlw $4, %xmm1
-; SSE2-NEXT:    pand %xmm3, %xmm1
-; SSE2-NEXT:    por %xmm6, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    psrlw $2, %xmm5
-; SSE2-NEXT:    pand %xmm9, %xmm5
-; SSE2-NEXT:    pand %xmm9, %xmm1
+; SSE2-NEXT:    packuswb %xmm4, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
+; SSE2-NEXT:    psrlw $4, %xmm4
+; SSE2-NEXT:    pand %xmm5, %xmm4
+; SSE2-NEXT:    pand %xmm5, %xmm1
+; SSE2-NEXT:    psllw $4, %xmm1
+; SSE2-NEXT:    por %xmm4, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
+; SSE2-NEXT:    psrlw $2, %xmm4
+; SSE2-NEXT:    pand %xmm6, %xmm4
+; SSE2-NEXT:    pand %xmm6, %xmm1
 ; SSE2-NEXT:    psllw $2, %xmm1
-; SSE2-NEXT:    por %xmm5, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    psrlw $1, %xmm5
-; SSE2-NEXT:    pand %xmm7, %xmm5
+; SSE2-NEXT:    por %xmm4, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
+; SSE2-NEXT:    psrlw $1, %xmm4
+; SSE2-NEXT:    pand %xmm7, %xmm4
 ; SSE2-NEXT:    pand %xmm7, %xmm1
 ; SSE2-NEXT:    paddb %xmm1, %xmm1
-; SSE2-NEXT:    por %xmm5, %xmm1
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm8[8],xmm5[9],xmm8[9],xmm5[10],xmm8[10],xmm5[11],xmm8[11],xmm5[12],xmm8[12],xmm5[13],xmm8[13],xmm5[14],xmm8[14],xmm5[15],xmm8[15]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,2,1,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,5,4]
+; SSE2-NEXT:    por %xmm4, %xmm1
+; SSE2-NEXT:    movdqa %xmm2, %xmm4
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm8[8],xmm4[9],xmm8[9],xmm4[10],xmm8[10],xmm4[11],xmm8[11],xmm4[12],xmm8[12],xmm4[13],xmm8[13],xmm4[14],xmm8[14],xmm4[15],xmm8[15]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,2,1,0,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,6,5,4]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3],xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    packuswb %xmm5, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
-; SSE2-NEXT:    movdqa %xmm3, %xmm6
-; SSE2-NEXT:    pandn %xmm5, %xmm6
-; SSE2-NEXT:    psrlw $4, %xmm2
-; SSE2-NEXT:    pand %xmm3, %xmm2
-; SSE2-NEXT:    por %xmm6, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psrlw $2, %xmm5
-; SSE2-NEXT:    pand %xmm9, %xmm5
-; SSE2-NEXT:    pand %xmm9, %xmm2
+; SSE2-NEXT:    packuswb %xmm4, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm4
+; SSE2-NEXT:    psrlw $4, %xmm4
+; SSE2-NEXT:    pand %xmm5, %xmm4
+; SSE2-NEXT:    pand %xmm5, %xmm2
+; SSE2-NEXT:    psllw $4, %xmm2
+; SSE2-NEXT:    por %xmm4, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm4
+; SSE2-NEXT:    psrlw $2, %xmm4
+; SSE2-NEXT:    pand %xmm6, %xmm4
+; SSE2-NEXT:    pand %xmm6, %xmm2
 ; SSE2-NEXT:    psllw $2, %xmm2
-; SSE2-NEXT:    por %xmm5, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psrlw $1, %xmm5
-; SSE2-NEXT:    pand %xmm7, %xmm5
+; SSE2-NEXT:    por %xmm4, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm4
+; SSE2-NEXT:    psrlw $1, %xmm4
+; SSE2-NEXT:    pand %xmm7, %xmm4
 ; SSE2-NEXT:    pand %xmm7, %xmm2
 ; SSE2-NEXT:    paddb %xmm2, %xmm2
-; SSE2-NEXT:    por %xmm5, %xmm2
-; SSE2-NEXT:    movdqa %xmm4, %xmm5
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm8[8],xmm5[9],xmm8[9],xmm5[10],xmm8[10],xmm5[11],xmm8[11],xmm5[12],xmm8[12],xmm5[13],xmm8[13],xmm5[14],xmm8[14],xmm5[15],xmm8[15]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,2,1,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3],xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
+; SSE2-NEXT:    por %xmm4, %xmm2
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm8[8],xmm4[9],xmm8[9],xmm4[10],xmm8[10],xmm4[11],xmm8[11],xmm4[12],xmm8[12],xmm4[13],xmm8[13],xmm4[14],xmm8[14],xmm4[15],xmm8[15]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    packuswb %xmm5, %xmm4
-; SSE2-NEXT:    movdqa %xmm4, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3],xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,2,1,0,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; SSE2-NEXT:    packuswb %xmm4, %xmm3
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    psrlw $4, %xmm4
-; SSE2-NEXT:    pand %xmm3, %xmm4
-; SSE2-NEXT:    pandn %xmm5, %xmm3
+; SSE2-NEXT:    pand %xmm5, %xmm4
+; SSE2-NEXT:    pand %xmm5, %xmm3
+; SSE2-NEXT:    psllw $4, %xmm3
 ; SSE2-NEXT:    por %xmm4, %xmm3
 ; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    psrlw $2, %xmm4
-; SSE2-NEXT:    pand %xmm9, %xmm4
-; SSE2-NEXT:    pand %xmm9, %xmm3
+; SSE2-NEXT:    pand %xmm6, %xmm4
+; SSE2-NEXT:    pand %xmm6, %xmm3
 ; SSE2-NEXT:    psllw $2, %xmm3
 ; SSE2-NEXT:    por %xmm4, %xmm3
 ; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    psrlw $1, %xmm4
 ; SSE2-NEXT:    pand %xmm7, %xmm4
 ; SSE2-NEXT:    pand %xmm7, %xmm3
 ; SSE2-NEXT:    paddb %xmm3, %xmm3
 ; SSE2-NEXT:    por %xmm4, %xmm3

 ; GFNIAVX512BW-NEXT:    retq
   %b = call <16 x i32> @llvm.bitreverse.v16i32(<16 x i32> %a)
   ret <16 x i32> %b
 }
 
 define <8 x i64> @test_bitreverse_v8i64(<8 x i64> %a) nounwind {
 ; SSE2-LABEL: test_bitreverse_v8i64:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    pxor %xmm8, %xmm8
-; SSE2-NEXT:    movdqa %xmm0, %xmm3
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm8[8],xmm3[9],xmm8[9],xmm3[10],xmm8[10],xmm3[11],xmm8[11],xmm3[12],xmm8[12],xmm3[13],xmm8[13],xmm3[14],xmm8[14],xmm3[15],xmm8[15]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,0,1]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,2,1,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; SSE2-NEXT:    movdqa %xmm0, %xmm5
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm8[8],xmm5[9],xmm8[9],xmm5[10],xmm8[10],xmm5[11],xmm8[11],xmm5[12],xmm8[12],xmm5[13],xmm8[13],xmm5[14],xmm8[14],xmm5[15],xmm8[15]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,3,0,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,2,1,0,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,5,4]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    packuswb %xmm3, %xmm0
+; SSE2-NEXT:    packuswb %xmm5, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm6
-; SSE2-NEXT:    psllw $4, %xmm6
-; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; SSE2-NEXT:    movdqa %xmm3, %xmm7
-; SSE2-NEXT:    pandn %xmm6, %xmm7
-; SSE2-NEXT:    psrlw $4, %xmm0
-; SSE2-NEXT:    pand %xmm3, %xmm0
-; SSE2-NEXT:    por %xmm7, %xmm0
+; SSE2-NEXT:    psrlw $4, %xmm6
+; SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE2-NEXT:    pand %xmm5, %xmm6
+; SSE2-NEXT:    pand %xmm5, %xmm0
+; SSE2-NEXT:    psllw $4, %xmm0
+; SSE2-NEXT:    por %xmm6, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm7
 ; SSE2-NEXT:    psrlw $2, %xmm7
-; SSE2-NEXT:    movdqa {{.*#+}} xmm9 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
-; SSE2-NEXT:    pand %xmm9, %xmm7
-; SSE2-NEXT:    pand %xmm9, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm6 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    pand %xmm6, %xmm7
+; SSE2-NEXT:    pand %xmm6, %xmm0
 ; SSE2-NEXT:    psllw $2, %xmm0
 ; SSE2-NEXT:    por %xmm7, %xmm0
-; SSE2-NEXT:    movdqa %xmm0, %xmm5
-; SSE2-NEXT:    psrlw $1, %xmm5
+; SSE2-NEXT:    movdqa %xmm0, %xmm4
+; SSE2-NEXT:    psrlw $1, %xmm4
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm7 = [85,85,85,85,85,85,85,85,85,85,85,85,85,85,85,85]
-; SSE2-NEXT:    pand %xmm7, %xmm5
+; SSE2-NEXT:    pand %xmm7, %xmm4
 ; SSE2-NEXT:    pand %xmm7, %xmm0
 ; SSE2-NEXT:    paddb %xmm0, %xmm0
-; SSE2-NEXT:    por %xmm5, %xmm0
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm8[8],xmm5[9],xmm8[9],xmm5[10],xmm8[10],xmm5[11],xmm8[11],xmm5[12],xmm8[12],xmm5[13],xmm8[13],xmm5[14],xmm8[14],xmm5[15],xmm8[15]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,3,0,1]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,2,1,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,5,4]
+; SSE2-NEXT:    por %xmm4, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm8[8],xmm4[9],xmm8[9],xmm4[10],xmm8[10],xmm4[11],xmm8[11],xmm4[12],xmm8[12],xmm4[13],xmm8[13],xmm4[14],xmm8[14],xmm4[15],xmm8[15]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,0,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,2,1,0,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,6,5,4]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    packuswb %xmm5, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
-; SSE2-NEXT:    movdqa %xmm3, %xmm6
-; SSE2-NEXT:    pandn %xmm5, %xmm6
-; SSE2-NEXT:    psrlw $4, %xmm1
-; SSE2-NEXT:    pand %xmm3, %xmm1
-; SSE2-NEXT:    por %xmm6, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    psrlw $2, %xmm5
-; SSE2-NEXT:    pand %xmm9, %xmm5
-; SSE2-NEXT:    pand %xmm9, %xmm1
+; SSE2-NEXT:    packuswb %xmm4, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
+; SSE2-NEXT:    psrlw $4, %xmm4
+; SSE2-NEXT:    pand %xmm5, %xmm4
+; SSE2-NEXT:    pand %xmm5, %xmm1
+; SSE2-NEXT:    psllw $4, %xmm1
+; SSE2-NEXT:    por %xmm4, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
+; SSE2-NEXT:    psrlw $2, %xmm4
+; SSE2-NEXT:    pand %xmm6, %xmm4
+; SSE2-NEXT:    pand %xmm6, %xmm1
 ; SSE2-NEXT:    psllw $2, %xmm1
-; SSE2-NEXT:    por %xmm5, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    psrlw $1, %xmm5
-; SSE2-NEXT:    pand %xmm7, %xmm5
+; SSE2-NEXT:    por %xmm4, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
+; SSE2-NEXT:    psrlw $1, %xmm4
+; SSE2-NEXT:    pand %xmm7, %xmm4
 ; SSE2-NEXT:    pand %xmm7, %xmm1
 ; SSE2-NEXT:    paddb %xmm1, %xmm1
-; SSE2-NEXT:    por %xmm5, %xmm1
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm8[8],xmm5[9],xmm8[9],xmm5[10],xmm8[10],xmm5[11],xmm8[11],xmm5[12],xmm8[12],xmm5[13],xmm8[13],xmm5[14],xmm8[14],xmm5[15],xmm8[15]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,3,0,1]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,2,1,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,5,4]
+; SSE2-NEXT:    por %xmm4, %xmm1
+; SSE2-NEXT:    movdqa %xmm2, %xmm4
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm8[8],xmm4[9],xmm8[9],xmm4[10],xmm8[10],xmm4[11],xmm8[11],xmm4[12],xmm8[12],xmm4[13],xmm8[13],xmm4[14],xmm8[14],xmm4[15],xmm8[15]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,0,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,2,1,0,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,6,5,4]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3],xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,0,1]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    packuswb %xmm5, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
-; SSE2-NEXT:    movdqa %xmm3, %xmm6
-; SSE2-NEXT:    pandn %xmm5, %xmm6
-; SSE2-NEXT:    psrlw $4, %xmm2
-; SSE2-NEXT:    pand %xmm3, %xmm2
-; SSE2-NEXT:    por %xmm6, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psrlw $2, %xmm5
-; SSE2-NEXT:    pand %xmm9, %xmm5
-; SSE2-NEXT:    pand %xmm9, %xmm2
+; SSE2-NEXT:    packuswb %xmm4, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm4
+; SSE2-NEXT:    psrlw $4, %xmm4
+; SSE2-NEXT:    pand %xmm5, %xmm4
+; SSE2-NEXT:    pand %xmm5, %xmm2
+; SSE2-NEXT:    psllw $4, %xmm2
+; SSE2-NEXT:    por %xmm4, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm4
+; SSE2-NEXT:    psrlw $2, %xmm4
+; SSE2-NEXT:    pand %xmm6, %xmm4
+; SSE2-NEXT:    pand %xmm6, %xmm2
 ; SSE2-NEXT:    psllw $2, %xmm2
-; SSE2-NEXT:    por %xmm5, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    psrlw $1, %xmm5
-; SSE2-NEXT:    pand %xmm7, %xmm5
+; SSE2-NEXT:    por %xmm4, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm4
+; SSE2-NEXT:    psrlw $1, %xmm4
+; SSE2-NEXT:    pand %xmm7, %xmm4
 ; SSE2-NEXT:    pand %xmm7, %xmm2
 ; SSE2-NEXT:    paddb %xmm2, %xmm2
-; SSE2-NEXT:    por %xmm5, %xmm2
-; SSE2-NEXT:    movdqa %xmm4, %xmm5
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm8[8],xmm5[9],xmm8[9],xmm5[10],xmm8[10],xmm5[11],xmm8[11],xmm5[12],xmm8[12],xmm5[13],xmm8[13],xmm5[14],xmm8[14],xmm5[15],xmm8[15]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,3,0,1]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,2,1,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3],xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
+; SSE2-NEXT:    por %xmm4, %xmm2
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm8[8],xmm4[9],xmm8[9],xmm4[10],xmm8[10],xmm4[11],xmm8[11],xmm4[12],xmm8[12],xmm4[13],xmm8[13],xmm4[14],xmm8[14],xmm4[15],xmm8[15]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,0,1]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,2,1,0,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,6,5,4]
-; SSE2-NEXT:    packuswb %xmm5, %xmm4
-; SSE2-NEXT:    movdqa %xmm4, %xmm5
-; SSE2-NEXT:    psllw $4, %xmm5
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3],xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,0,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,2,1,0,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; SSE2-NEXT:    packuswb %xmm4, %xmm3
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    psrlw $4, %xmm4
-; SSE2-NEXT:    pand %xmm3, %xmm4
-; SSE2-NEXT:    pandn %xmm5, %xmm3
+; SSE2-NEXT:    pand %xmm5, %xmm4
+; SSE2-NEXT:    pand %xmm5, %xmm3
+; SSE2-NEXT:    psllw $4, %xmm3
 ; SSE2-NEXT:    por %xmm4, %xmm3
 ; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    psrlw $2, %xmm4
-; SSE2-NEXT:    pand %xmm9, %xmm4
-; SSE2-NEXT:    pand %xmm9, %xmm3
+; SSE2-NEXT:    pand %xmm6, %xmm4
+; SSE2-NEXT:    pand %xmm6, %xmm3
 ; SSE2-NEXT:    psllw $2, %xmm3
 ; SSE2-NEXT:    por %xmm4, %xmm3
 ; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    psrlw $1, %xmm4
 ; SSE2-NEXT:    pand %xmm7, %xmm4
 ; SSE2-NEXT:    pand %xmm7, %xmm3
 ; SSE2-NEXT:    paddb %xmm3, %xmm3
 ; SSE2-NEXT:    por %xmm4, %xmm3