+/// IsOnlyUsedInZeroEqualityComparison - Return true if it only matters that the
+/// value is equal or not-equal to zero.
+static bool IsOnlyUsedInZeroEqualityComparison(Value *V) {
+ for (Value::use_iterator UI = V->use_begin(), E = V->use_end();
+ UI != E; ++UI) {
+ if (ICmpInst *IC = dyn_cast<ICmpInst>(*UI))
+ if (IC->isEquality())
+ if (Constant *C = dyn_cast<Constant>(IC->getOperand(1)))
+ if (C->isNullValue())
+ continue;
+ // Unknown instruction.
+ return false;
+ }
+ return true;
+}
+
+static SDValue getMemCmpLoad(Value *PtrVal, MVT LoadVT, const Type *LoadTy,
+ SelectionDAGBuilder &Builder) {
+
+ // Check to see if this load can be trivially constant folded, e.g. if the
+ // input is from a string literal.
+ if (Constant *LoadInput = dyn_cast<Constant>(PtrVal)) {
+ // Cast pointer to the type we really want to load.
+ LoadInput = ConstantExpr::getBitCast(LoadInput,
+ PointerType::getUnqual(LoadTy));
+
+ if (Constant *LoadCst = ConstantFoldLoadFromConstPtr(LoadInput, Builder.TD))
+ return Builder.getValue(LoadCst);
+ }
+
+ // Otherwise, we have to emit the load. If the pointer is to unfoldable but
+ // still constant memory, the input chain can be the entry node.
+ SDValue Root;
+ bool ConstantMemory = false;
+
+ // Do not serialize (non-volatile) loads of constant memory with anything.
+ if (Builder.AA->pointsToConstantMemory(PtrVal)) {
+ Root = Builder.DAG.getEntryNode();
+ ConstantMemory = true;
+ } else {
+ // Do not serialize non-volatile loads against each other.
+ Root = Builder.DAG.getRoot();
+ }
+
+ SDValue Ptr = Builder.getValue(PtrVal);
+ SDValue LoadVal = Builder.DAG.getLoad(LoadVT, Builder.getCurDebugLoc(), Root,
+ Ptr, PtrVal /*SrcValue*/, 0/*SVOffset*/,
+ false /*volatile*/, 1 /* align=1 */);
+
+ if (!ConstantMemory)
+ Builder.PendingLoads.push_back(LoadVal.getValue(1));
+ return LoadVal;
+}
+
+
+/// visitMemCmpCall - See if we can lower a call to memcmp in an optimized form.
+/// If so, return true and lower it, otherwise return false and it will be
+/// lowered like a normal call.
+bool SelectionDAGBuilder::visitMemCmpCall(CallInst &I) {
+ // Verify that the prototype makes sense. int memcmp(void*,void*,size_t)
+ if (I.getNumOperands() != 4)
+ return false;
+
+ Value *LHS = I.getOperand(1), *RHS = I.getOperand(2);
+ if (!isa<PointerType>(LHS->getType()) || !isa<PointerType>(RHS->getType()) ||
+ !isa<IntegerType>(I.getOperand(3)->getType()) ||
+ !isa<IntegerType>(I.getType()))
+ return false;
+
+ ConstantInt *Size = dyn_cast<ConstantInt>(I.getOperand(3));
+
+ // memcmp(S1,S2,2) != 0 -> (*(short*)LHS != *(short*)RHS) != 0
+ // memcmp(S1,S2,4) != 0 -> (*(int*)LHS != *(int*)RHS) != 0
+ if (Size && IsOnlyUsedInZeroEqualityComparison(&I)) {
+ bool ActuallyDoIt = true;
+ MVT LoadVT;
+ const Type *LoadTy;
+ switch (Size->getZExtValue()) {
+ default:
+ LoadVT = MVT::Other;
+ LoadTy = 0;
+ ActuallyDoIt = false;
+ break;
+ case 2:
+ LoadVT = MVT::i16;
+ LoadTy = Type::getInt16Ty(Size->getContext());
+ break;
+ case 4:
+ LoadVT = MVT::i32;
+ LoadTy = Type::getInt32Ty(Size->getContext());
+ break;
+ case 8:
+ LoadVT = MVT::i64;
+ LoadTy = Type::getInt64Ty(Size->getContext());
+ break;
+ /*
+ case 16:
+ LoadVT = MVT::v4i32;
+ LoadTy = Type::getInt32Ty(Size->getContext());
+ LoadTy = VectorType::get(LoadTy, 4);
+ break;
+ */
+ }
+
+ // This turns into unaligned loads. We only do this if the target natively
+ // supports the MVT we'll be loading or if it is small enough (<= 4) that
+ // we'll only produce a small number of byte loads.
+
+ // Require that we can find a legal MVT, and only do this if the target
+ // supports unaligned loads of that type. Expanding into byte loads would
+ // bloat the code.
+ if (ActuallyDoIt && Size->getZExtValue() > 4) {
+ // TODO: Handle 5 byte compare as 4-byte + 1 byte.
+ // TODO: Handle 8 byte compare on x86-32 as two 32-bit loads.
+ if (!TLI.isTypeLegal(LoadVT) ||!TLI.allowsUnalignedMemoryAccesses(LoadVT))
+ ActuallyDoIt = false;
+ }
+
+ if (ActuallyDoIt) {
+ SDValue LHSVal = getMemCmpLoad(LHS, LoadVT, LoadTy, *this);
+ SDValue RHSVal = getMemCmpLoad(RHS, LoadVT, LoadTy, *this);
+
+ SDValue Res = DAG.getSetCC(getCurDebugLoc(), MVT::i1, LHSVal, RHSVal,
+ ISD::SETNE);
+ EVT CallVT = TLI.getValueType(I.getType(), true);
+ setValue(&I, DAG.getZExtOrTrunc(Res, getCurDebugLoc(), CallVT));
+ return true;
+ }
+ }
+
+
+ return false;
+}
+