[X86] Add X86ISD::VSRAI to computeKnownBitsForTargetNode.

[RKSimon]

LGTM

llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

case X86ISD::PEXTRW: {
     EVT SrcVT = Src.getValueType();
     APInt DemandedElt = APInt::getOneBitSet(SrcVT.getVectorNumElements(),
                                             Op.getConstantOperandVal(1));
     Known = DAG.computeKnownBits(Src, DemandedElt, Depth + 1);
     Known = Known.zextOrTrunc(BitWidth);
     Known.Zero.setBitsFrom(SrcVT.getScalarSizeInBits());
     break;
   }
+  case X86ISD::VSRAI:
   case X86ISD::VSHLI:
   case X86ISD::VSRLI: {
     if (auto *ShiftImm = dyn_cast<ConstantSDNode>(Op.getOperand(1))) {
       if (ShiftImm->getAPIntValue().uge(VT.getScalarSizeInBits())) {
         Known.setAllZero();
         break;
       }
 
       Known = DAG.computeKnownBits(Op.getOperand(0), DemandedElts, Depth + 1);
       unsigned ShAmt = ShiftImm->getZExtValue();
       if (Opc == X86ISD::VSHLI) {
         Known.Zero <<= ShAmt;
         Known.One <<= ShAmt;
         // Low bits are known zero.
         Known.Zero.setLowBits(ShAmt);
-      } else {
+      } else if (Opc == X86ISD::VSRLI) {
         Known.Zero.lshrInPlace(ShAmt);
         Known.One.lshrInPlace(ShAmt);
         // High bits are known zero.
         Known.Zero.setHighBits(ShAmt);
+      } else {
+        Known.Zero.ashrInPlace(ShAmt);
+        Known.One.ashrInPlace(ShAmt);
       }
     }
     break;
   }
   case X86ISD::PACKUS: {
     // PACKUS is just a truncation if the upper half is zero.
     APInt DemandedLHS, DemandedRHS;
     getPackDemandedElts(VT, DemandedElts, DemandedLHS, DemandedRHS);

llvm/trunk/test/CodeGen/X86/combine-shl.ll

; AVX-NEXT:    retq
   ret <8 x i32> %3
 }
 
 define <8 x i32> @combine_vec_shl_ext_shl1(<8 x i16> %x) {
 ; SSE2-LABEL: combine_vec_shl_ext_shl1:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pmullw {{.*}}(%rip), %xmm0
 ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
-; SSE2-NEXT:    psrad $16, %xmm1
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
-; SSE2-NEXT:    psrad $16, %xmm0
-; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    pslld $31, %xmm2
-; SSE2-NEXT:    pslld $30, %xmm0
-; SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE2-NEXT:    psrad $16, %xmm1
 ; SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; SSE2-NEXT:    pslld $29, %xmm2
 ; SSE2-NEXT:    pslld $28, %xmm1
 ; SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE2-NEXT:    pslld $30, %xmm0
+; SSE2-NEXT:    xorpd %xmm2, %xmm2
+; SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: combine_vec_shl_ext_shl1:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pmullw {{.*}}(%rip), %xmm0
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
 ; SSE41-NEXT:    pmovsxwd %xmm1, %xmm1
 ; SSE41-NEXT:    pmovsxwd %xmm0, %xmm0