[X86] Fix two places that appear to misuse peekThroughOneUseBitcasts

[craig.topper]

Summary

peekThroughOneUseBitcasts checks the use count of the operand of the bitcast. Not the bitcast itself. So I think that means we need to do any outside haseOneUse checks before calling the function not after.

I was working on another patch where I misused the function and did a very quick audit to see if I there were other similar mistakes.

[RKSimon]

LGTM - cheers

llvm/lib/Target/X86/X86ISelLowering.cpp

case X86ISD::PSHUFHW:
 
     // Look for shuffle patterns which can be implemented as a single unpack.
     // FIXME: This doesn't handle the location of the PSHUFD generically, and
     // only works when we have a PSHUFD followed by two half-shuffles.
     if (Mask[0] == Mask[1] && Mask[2] == Mask[3] &&
         (V.getOpcode() == X86ISD::PSHUFLW ||
          V.getOpcode() == X86ISD::PSHUFHW) &&
         V.getOpcode() != N.getOpcode() &&
-        V.hasOneUse()) {
+        V.hasOneUse() && V.getOperand(0).hasOneUse()) {
       SDValue D = peekThroughOneUseBitcasts(V.getOperand(0));
-      if (D.getOpcode() == X86ISD::PSHUFD && D.hasOneUse()) {
+      if (D.getOpcode() == X86ISD::PSHUFD) {
         SmallVector<int, 4> VMask = getPSHUFShuffleMask(V);
         SmallVector<int, 4> DMask = getPSHUFShuffleMask(D);
         int NOffset = N.getOpcode() == X86ISD::PSHUFLW ? 0 : 4;
         int VOffset = V.getOpcode() == X86ISD::PSHUFLW ? 0 : 4;
         int WordMask[8];
         for (int i = 0; i < 4; ++i) {
           WordMask[i + NOffset] = Mask[i] + NOffset;
           WordMask[i + VOffset] = VMask[i] + VOffset;
if (isTargetShuffle(N->getOpcode())) {
     if (TLI.SimplifyDemandedVectorElts(Op, DemandedElts, KnownUndef, KnownZero, DCI))
       return SDValue(N, 0);
   }
 
   // Pull subvector inserts into undef through VZEXT_MOVL by making it an
   // insert into a zero vector. This helps get VZEXT_MOVL closer to
   // scalar_to_vectors where 256/512 are canonicalized to an insert and a
   // 128-bit scalar_to_vector. This reduces the number of isel patterns.
-  if (N->getOpcode() == X86ISD::VZEXT_MOVL && !DCI.isBeforeLegalizeOps()) {
+  if (N->getOpcode() == X86ISD::VZEXT_MOVL && !DCI.isBeforeLegalizeOps() &&
+      N->getOperand(0).hasOneUse()) {
     SDValue V = peekThroughOneUseBitcasts(N->getOperand(0));
 
-    if (V.getOpcode() == ISD::INSERT_SUBVECTOR && V.hasOneUse() &&
+    if (V.getOpcode() == ISD::INSERT_SUBVECTOR &&
         V.getOperand(0).isUndef() && isNullConstant(V.getOperand(2))) {
       SDValue In = V.getOperand(1);
       MVT SubVT =
           MVT::getVectorVT(VT.getSimpleVT().getVectorElementType(),
                            In.getValueSizeInBits() / VT.getScalarSizeInBits());
       In = DAG.getBitcast(SubVT, In);
       SDValue Movl = DAG.getNode(X86ISD::VZEXT_MOVL, dl, SubVT, In);
       return DAG.getNode(ISD::INSERT_SUBVECTOR, dl, VT,