X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;f=include%2Fllvm%2FAnalysis%2FCallGraph.h;h=803c0dfaaba247cd43c8b0ea0d8e321cc72a396a;hb=54fec07ec00b4449393a66e8e2e62fd241781f80;hp=fde7c82abbe72a791da69dfe5ee37f5a31a4e711;hpb=3889a2cb05c36f30050941679d5fd55d45e6a3ed;p=oota-llvm.git diff --git a/include/llvm/Analysis/CallGraph.h b/include/llvm/Analysis/CallGraph.h index fde7c82abbe..803c0dfaaba 100644 --- a/include/llvm/Analysis/CallGraph.h +++ b/include/llvm/Analysis/CallGraph.h @@ -1,278 +1,459 @@ -//===- CallGraph.h - Build a Module's call graph -----------------*- C++ -*--=// +//===- CallGraph.h - Build a Module's call graph ----------------*- C++ -*-===// // -// This interface is used to build and manipulate a call graph, which is a very -// useful tool for interprocedural optimization. +// The LLVM Compiler Infrastructure // -// Every function in a module is represented as a node in the call graph. The -// callgraph node keeps track of which functions the are called by the function -// corresponding to the node. -// -// A call graph will contain nodes where the function that they correspond to is -// null. This 'external' node is used to represent control flow that is not -// represented (or analyzable) in the module. As such, the external node will -// have edges to functions with the following properties: -// 1. All functions in the module without internal linkage, since they could -// be called by functions outside of the our analysis capability. -// 2. All functions whose address is used for something more than a direct -// call, for example being stored into a memory location. Since they may -// be called by an unknown caller later, they must be tracked as such. -// -// Similarly, functions have a call edge to the external node iff: -// 1. The function is external, reflecting the fact that they could call -// anything without internal linkage or that has its address taken. -// 2. The function contains an indirect function call. -// -// As an extension in the future, there may be multiple nodes with a null -// function. These will be used when we can prove (through pointer analysis) -// that an indirect call site can call only a specific set of functions. -// -// Because of these properties, the CallGraph captures a conservative superset -// of all of the caller-callee relationships, which is useful for -// transformations. -// -// The CallGraph class also attempts to figure out what the root of the -// CallGraph is, which is currently does by looking for a function named 'main'. -// If no function named 'main' is found, the external node is used as the entry -// node, reflecting the fact that any function without internal linkage could -// be called into (which is common for libraries). +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. // //===----------------------------------------------------------------------===// +/// \file +/// +/// This file provides interfaces used to build and manipulate a call graph, +/// which is a very useful tool for interprocedural optimization. +/// +/// Every function in a module is represented as a node in the call graph. The +/// callgraph node keeps track of which functions the are called by the +/// function corresponding to the node. +/// +/// A call graph may contain nodes where the function that they correspond to +/// is null. These 'external' nodes are used to represent control flow that is +/// not represented (or analyzable) in the module. In particular, this +/// analysis builds one external node such that: +/// 1. All functions in the module without internal linkage will have edges +/// from this external node, indicating that they could be called by +/// functions outside of the module. +/// 2. All functions whose address is used for something more than a direct +/// call, for example being stored into a memory location will also have +/// an edge from this external node. Since they may be called by an +/// unknown caller later, they must be tracked as such. +/// +/// There is a second external node added for calls that leave this module. +/// Functions have a call edge to the external node iff: +/// 1. The function is external, reflecting the fact that they could call +/// anything without internal linkage or that has its address taken. +/// 2. The function contains an indirect function call. +/// +/// As an extension in the future, there may be multiple nodes with a null +/// function. These will be used when we can prove (through pointer analysis) +/// that an indirect call site can call only a specific set of functions. +/// +/// Because of these properties, the CallGraph captures a conservative superset +/// of all of the caller-callee relationships, which is useful for +/// transformations. +/// +/// The CallGraph class also attempts to figure out what the root of the +/// CallGraph is, which it currently does by looking for a function named +/// 'main'. If no function named 'main' is found, the external node is used as +/// the entry node, reflecting the fact that any function without internal +/// linkage could be called into (which is common for libraries). +/// +//===----------------------------------------------------------------------===// #ifndef LLVM_ANALYSIS_CALLGRAPH_H #define LLVM_ANALYSIS_CALLGRAPH_H -#include "Support/GraphTraits.h" -#include "Support/STLExtras.h" +#include "llvm/ADT/GraphTraits.h" +#include "llvm/ADT/OwningPtr.h" +#include "llvm/ADT/STLExtras.h" +#include "llvm/IR/Function.h" #include "llvm/Pass.h" +#include "llvm/Support/CallSite.h" +#include "llvm/Support/IncludeFile.h" +#include "llvm/Support/ValueHandle.h" +#include + +namespace llvm { + class Function; class Module; class CallGraphNode; -//===----------------------------------------------------------------------===// -// CallGraph class definition -// -class CallGraph : public Pass { - Module *Mod; // The module this call graph represents +/// \brief The basic data container for the call graph of a \c Module of IR. +/// +/// This class exposes both the interface to the call graph for a module of IR. +/// +/// The core call graph itself can also be updated to reflect changes to the IR. +class CallGraph { + Module &M; typedef std::map FunctionMapTy; - FunctionMapTy FunctionMap; // Map from a function to its node - // Root is root of the call graph, or the external node if a 'main' function - // couldn't be found. ExternalNode is equivalent to (*this)[0]. - // - CallGraphNode *Root, *ExternalNode; + /// \brief A map from \c Function* to \c CallGraphNode*. + FunctionMapTy FunctionMap; + + /// \brief Root is root of the call graph, or the external node if a 'main' + /// function couldn't be found. + CallGraphNode *Root; + + /// \brief This node has edges to all external functions and those internal + /// functions that have their address taken. + CallGraphNode *ExternalCallingNode; + + /// \brief This node has edges to it from all functions making indirect calls + /// or calling an external function. + CallGraphNode *CallsExternalNode; + + /// \brief Replace the function represented by this node by another. + /// + /// This does not rescan the body of the function, so it is suitable when + /// splicing the body of one function to another while also updating all + /// callers from the old function to the new. + void spliceFunction(const Function *From, const Function *To); + + /// \brief Add a function to the call graph, and link the node to all of the + /// functions that it calls. + void addToCallGraph(Function *F); + public: + CallGraph(Module &M); + ~CallGraph(); + + void print(raw_ostream &OS) const; + void dump() const; - //===--------------------------------------------------------------------- - // Accessors... - // typedef FunctionMapTy::iterator iterator; typedef FunctionMapTy::const_iterator const_iterator; - // getExternalNode - Return the node that points to all functions that are - // accessable from outside of the current program. - // - CallGraphNode *getExternalNode() { return ExternalNode; } - const CallGraphNode *getExternalNode() const { return ExternalNode; } - - // getRoot - Return the root of the call graph, which is either main, or if - // main cannot be found, the external node. - // - CallGraphNode *getRoot() { return Root; } - const CallGraphNode *getRoot() const { return Root; } + /// \brief Returns the module the call graph corresponds to. + Module &getModule() const { return M; } - inline iterator begin() { return FunctionMap.begin(); } - inline iterator end() { return FunctionMap.end(); } + inline iterator begin() { return FunctionMap.begin(); } + inline iterator end() { return FunctionMap.end(); } inline const_iterator begin() const { return FunctionMap.begin(); } - inline const_iterator end() const { return FunctionMap.end(); } + inline const_iterator end() const { return FunctionMap.end(); } - - // Subscripting operators, return the call graph node for the provided - // function + /// \brief Returns the call graph node for the provided function. inline const CallGraphNode *operator[](const Function *F) const { const_iterator I = FunctionMap.find(F); assert(I != FunctionMap.end() && "Function not in callgraph!"); return I->second; } + + /// \brief Returns the call graph node for the provided function. inline CallGraphNode *operator[](const Function *F) { const_iterator I = FunctionMap.find(F); assert(I != FunctionMap.end() && "Function not in callgraph!"); return I->second; } + /// \brief Returns the \c CallGraphNode which is used to represent + /// undetermined calls into the callgraph. + CallGraphNode *getExternalCallingNode() const { return ExternalCallingNode; } + + CallGraphNode *getCallsExternalNode() const { return CallsExternalNode; } + + /// \brief Returns the root/main method in the module, or some other root + /// node, such as the externalcallingnode. + CallGraphNode *getRoot() { return Root; } + const CallGraphNode *getRoot() const { return Root; } + //===--------------------------------------------------------------------- // Functions to keep a call graph up to date with a function that has been - // modified + // modified. // - void addFunctionToModule(Function *Meth); - - // removeFunctionFromModule - Unlink the function from this module, returning - // it. Because this removes the function from the module, the call graph node - // is destroyed. This is only valid if the function does not call any other - // functions (ie, there are no edges in it's CGN). The easiest way to do this - // is to dropAllReferences before calling this. - // + /// \brief Unlink the function from this module, returning it. + /// + /// Because this removes the function from the module, the call graph node is + /// destroyed. This is only valid if the function does not call any other + /// functions (ie, there are no edges in it's CGN). The easiest way to do + /// this is to dropAllReferences before calling this. Function *removeFunctionFromModule(CallGraphNode *CGN); - Function *removeFunctionFromModule(Function *Meth) { - return removeFunctionFromModule((*this)[Meth]); + + /// \brief Similar to operator[], but this will insert a new CallGraphNode for + /// \c F if one does not already exist. + CallGraphNode *getOrInsertFunction(const Function *F); +}; + +/// \brief A node in the call graph for a module. +/// +/// Typically represents a function in the call graph. There are also special +/// "null" nodes used to represent theoretical entries in the call graph. +class CallGraphNode { + friend class CallGraph; + + AssertingVH F; + +public: + /// \brief A pair of the calling instruction (a call or invoke) + /// and the call graph node being called. + typedef std::pair CallRecord; + +private: + std::vector CalledFunctions; + + /// \brief The number of times that this CallGraphNode occurs in the + /// CalledFunctions array of this or other CallGraphNodes. + unsigned NumReferences; + + CallGraphNode(const CallGraphNode &) LLVM_DELETED_FUNCTION; + void operator=(const CallGraphNode &) LLVM_DELETED_FUNCTION; + + void DropRef() { --NumReferences; } + void AddRef() { ++NumReferences; } + +public: + typedef std::vector CalledFunctionsVector; + + /// \brief Creates a node for the specified function. + inline CallGraphNode(Function *F) : F(F), NumReferences(0) {} + + ~CallGraphNode() { + assert(NumReferences == 0 && "Node deleted while references remain"); } + typedef std::vector::iterator iterator; + typedef std::vector::const_iterator const_iterator; + + /// \brief Returns the function that this call graph node represents. + Function *getFunction() const { return F; } + + inline iterator begin() { return CalledFunctions.begin(); } + inline iterator end() { return CalledFunctions.end(); } + inline const_iterator begin() const { return CalledFunctions.begin(); } + inline const_iterator end() const { return CalledFunctions.end(); } + inline bool empty() const { return CalledFunctions.empty(); } + inline unsigned size() const { return (unsigned)CalledFunctions.size(); } + + /// \brief Returns the number of other CallGraphNodes in this CallGraph that + /// reference this node in their callee list. + unsigned getNumReferences() const { return NumReferences; } + + /// \brief Returns the i'th called function. + CallGraphNode *operator[](unsigned i) const { + assert(i < CalledFunctions.size() && "Invalid index"); + return CalledFunctions[i].second; + } + + /// \brief Print out this call graph node. + void dump() const; + void print(raw_ostream &OS) const; //===--------------------------------------------------------------------- - // Pass infrastructure interface glue code... + // Methods to keep a call graph up to date with a function that has been + // modified // - CallGraph() : Root(0) {} - ~CallGraph() { destroy(); } - // run - Compute the call graph for the specified module. - virtual bool run(Module &M); + /// \brief Removes all edges from this CallGraphNode to any functions it + /// calls. + void removeAllCalledFunctions() { + while (!CalledFunctions.empty()) { + CalledFunctions.back().second->DropRef(); + CalledFunctions.pop_back(); + } + } + + /// \brief Moves all the callee information from N to this node. + void stealCalledFunctionsFrom(CallGraphNode *N) { + assert(CalledFunctions.empty() && + "Cannot steal callsite information if I already have some"); + std::swap(CalledFunctions, N->CalledFunctions); + } - // getAnalysisUsage - This obviously provides a call graph - virtual void getAnalysisUsage(AnalysisUsage &AU) const { - AU.setPreservesAll(); + /// \brief Adds a function to the list of functions called by this one. + void addCalledFunction(CallSite CS, CallGraphNode *M) { + assert(!CS.getInstruction() || !CS.getCalledFunction() || + !CS.getCalledFunction()->isIntrinsic()); + CalledFunctions.push_back(std::make_pair(CS.getInstruction(), M)); + M->AddRef(); } - // releaseMemory - Data structures can be large, so free memory aggressively. - virtual void releaseMemory() { - destroy(); + void removeCallEdge(iterator I) { + I->second->DropRef(); + *I = CalledFunctions.back(); + CalledFunctions.pop_back(); } - /// Print the types found in the module. If the optional Module parameter is - /// passed in, then the types are printed symbolically if possible, using the - /// symbol table from the module. + /// \brief Removes the edge in the node for the specified call site. /// - void print(std::ostream &o, const Module *M) const; + /// Note that this method takes linear time, so it should be used sparingly. + void removeCallEdgeFor(CallSite CS); -private: - //===--------------------------------------------------------------------- - // Implementation of CallGraph construction - // + /// \brief Removes all call edges from this node to the specified callee + /// function. + /// + /// This takes more time to execute than removeCallEdgeTo, so it should not + /// be used unless necessary. + void removeAnyCallEdgeTo(CallGraphNode *Callee); - // getNodeFor - Return the node for the specified function or create one if it - // does not already exist. - // - CallGraphNode *getNodeFor(Function *F); + /// \brief Removes one edge associated with a null callsite from this node to + /// the specified callee function. + void removeOneAbstractEdgeTo(CallGraphNode *Callee); - // addToCallGraph - Add a function to the call graph, and link the node to all - // of the functions that it calls. - // - void addToCallGraph(Function *F); + /// \brief Replaces the edge in the node for the specified call site with a + /// new one. + /// + /// Note that this method takes linear time, so it should be used sparingly. + void replaceCallEdge(CallSite CS, CallSite NewCS, CallGraphNode *NewNode); - // destroy - Release memory for the call graph - void destroy(); + /// \brief A special function that should only be used by the CallGraph class. + // + // FIXME: Make this private? + void allReferencesDropped() { NumReferences = 0; } }; +/// \brief The \c ModulePass which wraps up a \c CallGraph and the logic to +/// build it. +/// +/// This class exposes both the interface to the call graph container and the +/// module pass which runs over a module of IR and produces the call graph. The +/// call graph interface is entirelly a wrapper around a \c CallGraph object +/// which is stored internally for each module. +class CallGraphWrapperPass : public ModulePass { + OwningPtr G; -//===----------------------------------------------------------------------===// -// CallGraphNode class definition -// -class CallGraphNode { - Function *Meth; - std::vector CalledFunctions; - - CallGraphNode(const CallGraphNode &); // Do not implement public: - //===--------------------------------------------------------------------- - // Accessor methods... - // + static char ID; // Class identification, replacement for typeinfo - typedef std::vector::iterator iterator; - typedef std::vector::const_iterator const_iterator; + CallGraphWrapperPass(); + virtual ~CallGraphWrapperPass(); - // getFunction - Return the function that this call graph node represents... - Function *getFunction() const { return Meth; } + /// \brief The internal \c CallGraph around which the rest of this interface + /// is wrapped. + const CallGraph &getCallGraph() const { return *G; } + CallGraph &getCallGraph() { return *G; } - inline iterator begin() { return CalledFunctions.begin(); } - inline iterator end() { return CalledFunctions.end(); } - inline const_iterator begin() const { return CalledFunctions.begin(); } - inline const_iterator end() const { return CalledFunctions.end(); } - inline unsigned size() const { return CalledFunctions.size(); } + typedef CallGraph::iterator iterator; + typedef CallGraph::const_iterator const_iterator; - // Subscripting operator - Return the i'th called function... - // - CallGraphNode *operator[](unsigned i) const { return CalledFunctions[i];} + /// \brief Returns the module the call graph corresponds to. + Module &getModule() const { return G->getModule(); } + + inline iterator begin() { return G->begin(); } + inline iterator end() { return G->end(); } + inline const_iterator begin() const { return G->begin(); } + inline const_iterator end() const { return G->end(); } + + /// \brief Returns the call graph node for the provided function. + inline const CallGraphNode *operator[](const Function *F) const { + return (*G)[F]; + } + /// \brief Returns the call graph node for the provided function. + inline CallGraphNode *operator[](const Function *F) { return (*G)[F]; } + + /// \brief Returns the \c CallGraphNode which is used to represent + /// undetermined calls into the callgraph. + CallGraphNode *getExternalCallingNode() const { + return G->getExternalCallingNode(); + } + + CallGraphNode *getCallsExternalNode() const { + return G->getCallsExternalNode(); + } + + /// \brief Returns the root/main method in the module, or some other root + /// node, such as the externalcallingnode. + CallGraphNode *getRoot() { return G->getRoot(); } + const CallGraphNode *getRoot() const { return G->getRoot(); } //===--------------------------------------------------------------------- - // Methods to keep a call graph up to date with a function that has been - // modified + // Functions to keep a call graph up to date with a function that has been + // modified. // - void removeAllCalledFunctions() { - CalledFunctions.clear(); + /// \brief Unlink the function from this module, returning it. + /// + /// Because this removes the function from the module, the call graph node is + /// destroyed. This is only valid if the function does not call any other + /// functions (ie, there are no edges in it's CGN). The easiest way to do + /// this is to dropAllReferences before calling this. + Function *removeFunctionFromModule(CallGraphNode *CGN) { + return G->removeFunctionFromModule(CGN); } -private: // Stuff to construct the node, used by CallGraph - friend class CallGraph; - - // CallGraphNode ctor - Create a node for the specified function... - inline CallGraphNode(Function *F) : Meth(F) {} - - // addCalledFunction add a function to the list of functions called by this - // one - void addCalledFunction(CallGraphNode *M) { - CalledFunctions.push_back(M); + /// \brief Similar to operator[], but this will insert a new CallGraphNode for + /// \c F if one does not already exist. + CallGraphNode *getOrInsertFunction(const Function *F) { + return G->getOrInsertFunction(F); } -}; + //===--------------------------------------------------------------------- + // Implementation of the ModulePass interface needed here. + // + + virtual void getAnalysisUsage(AnalysisUsage &AU) const; + virtual bool runOnModule(Module &M); + virtual void releaseMemory(); + void print(raw_ostream &o, const Module *) const; + void dump() const; +}; //===----------------------------------------------------------------------===// // GraphTraits specializations for call graphs so that they can be treated as -// graphs by the generic graph algorithms... +// graphs by the generic graph algorithms. // // Provide graph traits for tranversing call graphs using standard graph // traversals. -template <> struct GraphTraits { +template <> struct GraphTraits { typedef CallGraphNode NodeType; - typedef NodeType::iterator ChildIteratorType; + + typedef CallGraphNode::CallRecord CGNPairTy; + typedef std::pointer_to_unary_function + CGNDerefFun; static NodeType *getEntryNode(CallGraphNode *CGN) { return CGN; } - static inline ChildIteratorType child_begin(NodeType *N) { return N->begin();} - static inline ChildIteratorType child_end (NodeType *N) { return N->end(); } + + typedef mapped_iterator ChildIteratorType; + + static inline ChildIteratorType child_begin(NodeType *N) { + return map_iterator(N->begin(), CGNDerefFun(CGNDeref)); + } + static inline ChildIteratorType child_end(NodeType *N) { + return map_iterator(N->end(), CGNDerefFun(CGNDeref)); + } + + static CallGraphNode *CGNDeref(CGNPairTy P) { return P.second; } }; -template <> struct GraphTraits { +template <> struct GraphTraits { typedef const CallGraphNode NodeType; typedef NodeType::const_iterator ChildIteratorType; static NodeType *getEntryNode(const CallGraphNode *CGN) { return CGN; } - static inline ChildIteratorType child_begin(NodeType *N) { return N->begin();} - static inline ChildIteratorType child_end (NodeType *N) { return N->end(); } + static inline ChildIteratorType child_begin(NodeType *N) { + return N->begin(); + } + static inline ChildIteratorType child_end(NodeType *N) { return N->end(); } }; -template<> struct GraphTraits : public GraphTraits { +template <> +struct GraphTraits : public GraphTraits { static NodeType *getEntryNode(CallGraph *CGN) { - return CGN->getExternalNode(); // Start at the external node! + return CGN->getExternalCallingNode(); // Start at the external node! } - typedef std::pair PairTy; - typedef std::pointer_to_unary_function DerefFun; + typedef std::pair PairTy; + typedef std::pointer_to_unary_function DerefFun; // nodes_iterator/begin/end - Allow iteration over all nodes in the graph typedef mapped_iterator nodes_iterator; static nodes_iterator nodes_begin(CallGraph *CG) { return map_iterator(CG->begin(), DerefFun(CGdereference)); } - static nodes_iterator nodes_end (CallGraph *CG) { + static nodes_iterator nodes_end(CallGraph *CG) { return map_iterator(CG->end(), DerefFun(CGdereference)); } - static CallGraphNode &CGdereference (std::pair P) { - return *P.second; - } + static CallGraphNode &CGdereference(PairTy P) { return *P.second; } }; -template<> struct GraphTraits : - public GraphTraits { + +template <> +struct GraphTraits : public GraphTraits< + const CallGraphNode *> { static NodeType *getEntryNode(const CallGraph *CGN) { - return CGN->getExternalNode(); + return CGN->getExternalCallingNode(); } // nodes_iterator/begin/end - Allow iteration over all nodes in the graph typedef CallGraph::const_iterator nodes_iterator; static nodes_iterator nodes_begin(const CallGraph *CG) { return CG->begin(); } - static nodes_iterator nodes_end (const CallGraph *CG) { return CG->end(); } + static nodes_iterator nodes_end(const CallGraph *CG) { return CG->end(); } }; +} // End llvm namespace + +// Make sure that any clients of this file link in CallGraph.cpp +FORCE_DEFINING_FILE_TO_BE_LINKED(CallGraph) + #endif