+/// If we inlined an invoke site, we need to convert calls
+/// in the body of the inlined function into invokes.
+///
+/// II is the invoke instruction being inlined. FirstNewBlock is the first
+/// block of the inlined code (the last block is the end of the function),
+/// and InlineCodeInfo is information about the code that got inlined.
+static void HandleInlinedEHPad(InvokeInst *II, BasicBlock *FirstNewBlock,
+ ClonedCodeInfo &InlinedCodeInfo) {
+ BasicBlock *UnwindDest = II->getUnwindDest();
+ Function *Caller = FirstNewBlock->getParent();
+
+ assert(UnwindDest->getFirstNonPHI()->isEHPad() && "unexpected BasicBlock!");
+
+ // If there are PHI nodes in the unwind destination block, we need to keep
+ // track of which values came into them from the invoke before removing the
+ // edge from this block.
+ SmallVector<Value *, 8> UnwindDestPHIValues;
+ llvm::BasicBlock *InvokeBB = II->getParent();
+ for (Instruction &I : *UnwindDest) {
+ // Save the value to use for this edge.
+ PHINode *PHI = dyn_cast<PHINode>(&I);
+ if (!PHI)
+ break;
+ UnwindDestPHIValues.push_back(PHI->getIncomingValueForBlock(InvokeBB));
+ }
+
+ // Add incoming-PHI values to the unwind destination block for the given basic
+ // block, using the values for the original invoke's source block.
+ auto UpdatePHINodes = [&](BasicBlock *Src) {
+ BasicBlock::iterator I = UnwindDest->begin();
+ for (Value *V : UnwindDestPHIValues) {
+ PHINode *PHI = cast<PHINode>(I);
+ PHI->addIncoming(V, Src);
+ ++I;
+ }
+ };
+
+ // Forward EH terminator instructions to the caller's invoke destination.
+ // This is as simple as connect all the instructions which 'unwind to caller'
+ // to the invoke destination.
+ for (Function::iterator BB = FirstNewBlock, E = Caller->end(); BB != E;
+ ++BB) {
+ Instruction *I = BB->getFirstNonPHI();
+ if (I->isEHPad()) {
+ if (auto *CEPI = dyn_cast<CatchEndPadInst>(I)) {
+ if (CEPI->unwindsToCaller()) {
+ CatchEndPadInst::Create(CEPI->getContext(), UnwindDest, CEPI);
+ CEPI->eraseFromParent();
+ UpdatePHINodes(BB);
+ }
+ } else if (auto *CEPI = dyn_cast<CleanupEndPadInst>(I)) {
+ if (CEPI->unwindsToCaller()) {
+ CleanupEndPadInst::Create(CEPI->getCleanupPad(), UnwindDest, CEPI);
+ CEPI->eraseFromParent();
+ UpdatePHINodes(BB);
+ }
+ } else if (auto *TPI = dyn_cast<TerminatePadInst>(I)) {
+ if (TPI->unwindsToCaller()) {
+ SmallVector<Value *, 3> TerminatePadArgs;
+ for (Value *Operand : TPI->operands())
+ TerminatePadArgs.push_back(Operand);
+ TerminatePadInst::Create(TPI->getContext(), UnwindDest, TPI);
+ TPI->eraseFromParent();
+ UpdatePHINodes(BB);
+ }
+ } else {
+ assert(isa<CatchPadInst>(I) || isa<CleanupPadInst>(I));
+ }
+ }
+
+ if (auto *CRI = dyn_cast<CleanupReturnInst>(BB->getTerminator())) {
+ if (CRI->unwindsToCaller()) {
+ CleanupReturnInst::Create(CRI->getCleanupPad(), UnwindDest, CRI);
+ CRI->eraseFromParent();
+ UpdatePHINodes(BB);
+ }
+ }
+ }
+
+ if (InlinedCodeInfo.ContainsCalls)
+ for (Function::iterator BB = FirstNewBlock, E = Caller->end(); BB != E;
+ ++BB)
+ if (BasicBlock *NewBB =
+ HandleCallsInBlockInlinedThroughInvoke(BB, UnwindDest))
+ // Update any PHI nodes in the exceptional block to indicate that there
+ // is now a new entry in them.
+ UpdatePHINodes(NewBB);
+
+ // Now that everything is happy, we have one final detail. The PHI nodes in
+ // the exception destination block still have entries due to the original
+ // invoke instruction. Eliminate these entries (which might even delete the
+ // PHI node) now.
+ UnwindDest->removePredecessor(InvokeBB);
+}
+
+/// When inlining a function that contains noalias scope metadata,
+/// this metadata needs to be cloned so that the inlined blocks
+/// have different "unqiue scopes" at every call site. Were this not done, then
+/// aliasing scopes from a function inlined into a caller multiple times could
+/// not be differentiated (and this would lead to miscompiles because the
+/// non-aliasing property communicated by the metadata could have
+/// call-site-specific control dependencies).
+static void CloneAliasScopeMetadata(CallSite CS, ValueToValueMapTy &VMap) {
+ const Function *CalledFunc = CS.getCalledFunction();
+ SetVector<const MDNode *> MD;
+
+ // Note: We could only clone the metadata if it is already used in the
+ // caller. I'm omitting that check here because it might confuse
+ // inter-procedural alias analysis passes. We can revisit this if it becomes
+ // an efficiency or overhead problem.
+
+ for (Function::const_iterator I = CalledFunc->begin(), IE = CalledFunc->end();
+ I != IE; ++I)
+ for (BasicBlock::const_iterator J = I->begin(), JE = I->end(); J != JE; ++J) {
+ if (const MDNode *M = J->getMetadata(LLVMContext::MD_alias_scope))
+ MD.insert(M);
+ if (const MDNode *M = J->getMetadata(LLVMContext::MD_noalias))
+ MD.insert(M);
+ }
+
+ if (MD.empty())
+ return;
+
+ // Walk the existing metadata, adding the complete (perhaps cyclic) chain to
+ // the set.
+ SmallVector<const Metadata *, 16> Queue(MD.begin(), MD.end());
+ while (!Queue.empty()) {
+ const MDNode *M = cast<MDNode>(Queue.pop_back_val());
+ for (unsigned i = 0, ie = M->getNumOperands(); i != ie; ++i)
+ if (const MDNode *M1 = dyn_cast<MDNode>(M->getOperand(i)))
+ if (MD.insert(M1))
+ Queue.push_back(M1);
+ }
+
+ // Now we have a complete set of all metadata in the chains used to specify
+ // the noalias scopes and the lists of those scopes.
+ SmallVector<TempMDTuple, 16> DummyNodes;
+ DenseMap<const MDNode *, TrackingMDNodeRef> MDMap;
+ for (SetVector<const MDNode *>::iterator I = MD.begin(), IE = MD.end();
+ I != IE; ++I) {
+ DummyNodes.push_back(MDTuple::getTemporary(CalledFunc->getContext(), None));
+ MDMap[*I].reset(DummyNodes.back().get());
+ }
+
+ // Create new metadata nodes to replace the dummy nodes, replacing old
+ // metadata references with either a dummy node or an already-created new
+ // node.
+ for (SetVector<const MDNode *>::iterator I = MD.begin(), IE = MD.end();
+ I != IE; ++I) {
+ SmallVector<Metadata *, 4> NewOps;
+ for (unsigned i = 0, ie = (*I)->getNumOperands(); i != ie; ++i) {
+ const Metadata *V = (*I)->getOperand(i);
+ if (const MDNode *M = dyn_cast<MDNode>(V))
+ NewOps.push_back(MDMap[M]);
+ else
+ NewOps.push_back(const_cast<Metadata *>(V));
+ }
+
+ MDNode *NewM = MDNode::get(CalledFunc->getContext(), NewOps);
+ MDTuple *TempM = cast<MDTuple>(MDMap[*I]);
+ assert(TempM->isTemporary() && "Expected temporary node");
+
+ TempM->replaceAllUsesWith(NewM);
+ }
+
+ // Now replace the metadata in the new inlined instructions with the
+ // repacements from the map.
+ for (ValueToValueMapTy::iterator VMI = VMap.begin(), VMIE = VMap.end();
+ VMI != VMIE; ++VMI) {
+ if (!VMI->second)
+ continue;
+
+ Instruction *NI = dyn_cast<Instruction>(VMI->second);
+ if (!NI)
+ continue;
+
+ if (MDNode *M = NI->getMetadata(LLVMContext::MD_alias_scope)) {
+ MDNode *NewMD = MDMap[M];
+ // If the call site also had alias scope metadata (a list of scopes to
+ // which instructions inside it might belong), propagate those scopes to
+ // the inlined instructions.
+ if (MDNode *CSM =
+ CS.getInstruction()->getMetadata(LLVMContext::MD_alias_scope))
+ NewMD = MDNode::concatenate(NewMD, CSM);
+ NI->setMetadata(LLVMContext::MD_alias_scope, NewMD);
+ } else if (NI->mayReadOrWriteMemory()) {
+ if (MDNode *M =
+ CS.getInstruction()->getMetadata(LLVMContext::MD_alias_scope))
+ NI->setMetadata(LLVMContext::MD_alias_scope, M);
+ }
+
+ if (MDNode *M = NI->getMetadata(LLVMContext::MD_noalias)) {
+ MDNode *NewMD = MDMap[M];
+ // If the call site also had noalias metadata (a list of scopes with
+ // which instructions inside it don't alias), propagate those scopes to
+ // the inlined instructions.
+ if (MDNode *CSM =
+ CS.getInstruction()->getMetadata(LLVMContext::MD_noalias))
+ NewMD = MDNode::concatenate(NewMD, CSM);
+ NI->setMetadata(LLVMContext::MD_noalias, NewMD);
+ } else if (NI->mayReadOrWriteMemory()) {
+ if (MDNode *M = CS.getInstruction()->getMetadata(LLVMContext::MD_noalias))
+ NI->setMetadata(LLVMContext::MD_noalias, M);
+ }
+ }
+}
+
+/// If the inlined function has noalias arguments,
+/// then add new alias scopes for each noalias argument, tag the mapped noalias
+/// parameters with noalias metadata specifying the new scope, and tag all
+/// non-derived loads, stores and memory intrinsics with the new alias scopes.
+static void AddAliasScopeMetadata(CallSite CS, ValueToValueMapTy &VMap,
+ const DataLayout &DL, AAResults *CalleeAAR) {
+ if (!EnableNoAliasConversion)
+ return;
+
+ const Function *CalledFunc = CS.getCalledFunction();
+ SmallVector<const Argument *, 4> NoAliasArgs;
+
+ for (Function::const_arg_iterator I = CalledFunc->arg_begin(),
+ E = CalledFunc->arg_end(); I != E; ++I) {
+ if (I->hasNoAliasAttr() && !I->hasNUses(0))
+ NoAliasArgs.push_back(I);
+ }
+
+ if (NoAliasArgs.empty())
+ return;
+
+ // To do a good job, if a noalias variable is captured, we need to know if
+ // the capture point dominates the particular use we're considering.
+ DominatorTree DT;
+ DT.recalculate(const_cast<Function&>(*CalledFunc));
+
+ // noalias indicates that pointer values based on the argument do not alias
+ // pointer values which are not based on it. So we add a new "scope" for each
+ // noalias function argument. Accesses using pointers based on that argument
+ // become part of that alias scope, accesses using pointers not based on that
+ // argument are tagged as noalias with that scope.
+
+ DenseMap<const Argument *, MDNode *> NewScopes;
+ MDBuilder MDB(CalledFunc->getContext());
+
+ // Create a new scope domain for this function.
+ MDNode *NewDomain =
+ MDB.createAnonymousAliasScopeDomain(CalledFunc->getName());
+ for (unsigned i = 0, e = NoAliasArgs.size(); i != e; ++i) {
+ const Argument *A = NoAliasArgs[i];
+
+ std::string Name = CalledFunc->getName();
+ if (A->hasName()) {
+ Name += ": %";
+ Name += A->getName();
+ } else {
+ Name += ": argument ";
+ Name += utostr(i);
+ }
+
+ // Note: We always create a new anonymous root here. This is true regardless
+ // of the linkage of the callee because the aliasing "scope" is not just a
+ // property of the callee, but also all control dependencies in the caller.
+ MDNode *NewScope = MDB.createAnonymousAliasScope(NewDomain, Name);
+ NewScopes.insert(std::make_pair(A, NewScope));
+ }
+
+ // Iterate over all new instructions in the map; for all memory-access
+ // instructions, add the alias scope metadata.
+ for (ValueToValueMapTy::iterator VMI = VMap.begin(), VMIE = VMap.end();
+ VMI != VMIE; ++VMI) {
+ if (const Instruction *I = dyn_cast<Instruction>(VMI->first)) {
+ if (!VMI->second)
+ continue;
+
+ Instruction *NI = dyn_cast<Instruction>(VMI->second);
+ if (!NI)
+ continue;
+
+ bool IsArgMemOnlyCall = false, IsFuncCall = false;
+ SmallVector<const Value *, 2> PtrArgs;
+
+ if (const LoadInst *LI = dyn_cast<LoadInst>(I))
+ PtrArgs.push_back(LI->getPointerOperand());
+ else if (const StoreInst *SI = dyn_cast<StoreInst>(I))
+ PtrArgs.push_back(SI->getPointerOperand());
+ else if (const VAArgInst *VAAI = dyn_cast<VAArgInst>(I))
+ PtrArgs.push_back(VAAI->getPointerOperand());
+ else if (const AtomicCmpXchgInst *CXI = dyn_cast<AtomicCmpXchgInst>(I))
+ PtrArgs.push_back(CXI->getPointerOperand());
+ else if (const AtomicRMWInst *RMWI = dyn_cast<AtomicRMWInst>(I))
+ PtrArgs.push_back(RMWI->getPointerOperand());
+ else if (ImmutableCallSite ICS = ImmutableCallSite(I)) {
+ // If we know that the call does not access memory, then we'll still
+ // know that about the inlined clone of this call site, and we don't
+ // need to add metadata.
+ if (ICS.doesNotAccessMemory())
+ continue;
+
+ IsFuncCall = true;
+ if (CalleeAAR) {
+ FunctionModRefBehavior MRB = CalleeAAR->getModRefBehavior(ICS);
+ if (MRB == FMRB_OnlyAccessesArgumentPointees ||
+ MRB == FMRB_OnlyReadsArgumentPointees)
+ IsArgMemOnlyCall = true;
+ }
+
+ for (ImmutableCallSite::arg_iterator AI = ICS.arg_begin(),
+ AE = ICS.arg_end(); AI != AE; ++AI) {
+ // We need to check the underlying objects of all arguments, not just
+ // the pointer arguments, because we might be passing pointers as
+ // integers, etc.
+ // However, if we know that the call only accesses pointer arguments,
+ // then we only need to check the pointer arguments.
+ if (IsArgMemOnlyCall && !(*AI)->getType()->isPointerTy())
+ continue;
+
+ PtrArgs.push_back(*AI);
+ }
+ }
+
+ // If we found no pointers, then this instruction is not suitable for
+ // pairing with an instruction to receive aliasing metadata.
+ // However, if this is a call, this we might just alias with none of the
+ // noalias arguments.
+ if (PtrArgs.empty() && !IsFuncCall)
+ continue;
+
+ // It is possible that there is only one underlying object, but you
+ // need to go through several PHIs to see it, and thus could be
+ // repeated in the Objects list.
+ SmallPtrSet<const Value *, 4> ObjSet;
+ SmallVector<Metadata *, 4> Scopes, NoAliases;
+
+ SmallSetVector<const Argument *, 4> NAPtrArgs;
+ for (unsigned i = 0, ie = PtrArgs.size(); i != ie; ++i) {
+ SmallVector<Value *, 4> Objects;
+ GetUnderlyingObjects(const_cast<Value*>(PtrArgs[i]),
+ Objects, DL, /* MaxLookup = */ 0);
+
+ for (Value *O : Objects)
+ ObjSet.insert(O);
+ }
+
+ // Figure out if we're derived from anything that is not a noalias
+ // argument.
+ bool CanDeriveViaCapture = false, UsesAliasingPtr = false;
+ for (const Value *V : ObjSet) {
+ // Is this value a constant that cannot be derived from any pointer
+ // value (we need to exclude constant expressions, for example, that
+ // are formed from arithmetic on global symbols).
+ bool IsNonPtrConst = isa<ConstantInt>(V) || isa<ConstantFP>(V) ||
+ isa<ConstantPointerNull>(V) ||
+ isa<ConstantDataVector>(V) || isa<UndefValue>(V);
+ if (IsNonPtrConst)
+ continue;
+
+ // If this is anything other than a noalias argument, then we cannot
+ // completely describe the aliasing properties using alias.scope
+ // metadata (and, thus, won't add any).
+ if (const Argument *A = dyn_cast<Argument>(V)) {
+ if (!A->hasNoAliasAttr())
+ UsesAliasingPtr = true;
+ } else {
+ UsesAliasingPtr = true;
+ }
+
+ // If this is not some identified function-local object (which cannot
+ // directly alias a noalias argument), or some other argument (which,
+ // by definition, also cannot alias a noalias argument), then we could
+ // alias a noalias argument that has been captured).
+ if (!isa<Argument>(V) &&
+ !isIdentifiedFunctionLocal(const_cast<Value*>(V)))
+ CanDeriveViaCapture = true;
+ }
+
+ // A function call can always get captured noalias pointers (via other
+ // parameters, globals, etc.).
+ if (IsFuncCall && !IsArgMemOnlyCall)
+ CanDeriveViaCapture = true;
+
+ // First, we want to figure out all of the sets with which we definitely
+ // don't alias. Iterate over all noalias set, and add those for which:
+ // 1. The noalias argument is not in the set of objects from which we
+ // definitely derive.
+ // 2. The noalias argument has not yet been captured.
+ // An arbitrary function that might load pointers could see captured
+ // noalias arguments via other noalias arguments or globals, and so we
+ // must always check for prior capture.
+ for (const Argument *A : NoAliasArgs) {
+ if (!ObjSet.count(A) && (!CanDeriveViaCapture ||
+ // It might be tempting to skip the
+ // PointerMayBeCapturedBefore check if
+ // A->hasNoCaptureAttr() is true, but this is
+ // incorrect because nocapture only guarantees
+ // that no copies outlive the function, not
+ // that the value cannot be locally captured.
+ !PointerMayBeCapturedBefore(A,
+ /* ReturnCaptures */ false,
+ /* StoreCaptures */ false, I, &DT)))
+ NoAliases.push_back(NewScopes[A]);
+ }
+
+ if (!NoAliases.empty())
+ NI->setMetadata(LLVMContext::MD_noalias,
+ MDNode::concatenate(
+ NI->getMetadata(LLVMContext::MD_noalias),
+ MDNode::get(CalledFunc->getContext(), NoAliases)));
+
+ // Next, we want to figure out all of the sets to which we might belong.
+ // We might belong to a set if the noalias argument is in the set of
+ // underlying objects. If there is some non-noalias argument in our list
+ // of underlying objects, then we cannot add a scope because the fact
+ // that some access does not alias with any set of our noalias arguments
+ // cannot itself guarantee that it does not alias with this access
+ // (because there is some pointer of unknown origin involved and the
+ // other access might also depend on this pointer). We also cannot add
+ // scopes to arbitrary functions unless we know they don't access any
+ // non-parameter pointer-values.
+ bool CanAddScopes = !UsesAliasingPtr;
+ if (CanAddScopes && IsFuncCall)
+ CanAddScopes = IsArgMemOnlyCall;
+
+ if (CanAddScopes)
+ for (const Argument *A : NoAliasArgs) {
+ if (ObjSet.count(A))
+ Scopes.push_back(NewScopes[A]);
+ }
+
+ if (!Scopes.empty())
+ NI->setMetadata(
+ LLVMContext::MD_alias_scope,
+ MDNode::concatenate(NI->getMetadata(LLVMContext::MD_alias_scope),
+ MDNode::get(CalledFunc->getContext(), Scopes)));
+ }
+ }
+}
+
+/// If the inlined function has non-byval align arguments, then
+/// add @llvm.assume-based alignment assumptions to preserve this information.
+static void AddAlignmentAssumptions(CallSite CS, InlineFunctionInfo &IFI) {
+ if (!PreserveAlignmentAssumptions)
+ return;
+ auto &DL = CS.getCaller()->getParent()->getDataLayout();
+
+ // To avoid inserting redundant assumptions, we should check for assumptions
+ // already in the caller. To do this, we might need a DT of the caller.
+ DominatorTree DT;
+ bool DTCalculated = false;
+
+ Function *CalledFunc = CS.getCalledFunction();
+ for (Function::arg_iterator I = CalledFunc->arg_begin(),
+ E = CalledFunc->arg_end();
+ I != E; ++I) {
+ unsigned Align = I->getType()->isPointerTy() ? I->getParamAlignment() : 0;
+ if (Align && !I->hasByValOrInAllocaAttr() && !I->hasNUses(0)) {
+ if (!DTCalculated) {
+ DT.recalculate(const_cast<Function&>(*CS.getInstruction()->getParent()
+ ->getParent()));
+ DTCalculated = true;
+ }
+
+ // If we can already prove the asserted alignment in the context of the
+ // caller, then don't bother inserting the assumption.
+ Value *Arg = CS.getArgument(I->getArgNo());
+ if (getKnownAlignment(Arg, DL, CS.getInstruction(),
+ &IFI.ACT->getAssumptionCache(*CS.getCaller()),
+ &DT) >= Align)
+ continue;
+
+ IRBuilder<>(CS.getInstruction())
+ .CreateAlignmentAssumption(DL, Arg, Align);
+ }
+ }
+}
+
+/// Once we have cloned code over from a callee into the caller,
+/// update the specified callgraph to reflect the changes we made.
+/// Note that it's possible that not all code was copied over, so only