-//===- llvm/Analysis/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
//
-// This call graph represents a dynamic method invocation as a null method node.
-// A call graph may only have up to one null method node that represents all of
-// the dynamic method invocations.
-//
-// Additionally, the 'root' node of a call graph represents the "entry point"
-// node of the graph, which has an edge to every external method in the graph.
-// This node has a null method pointer.
+// 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 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 "llvm/ADT/GraphTraits.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/IR/CallSite.h"
+#include "llvm/IR/Function.h"
+#include "llvm/IR/ValueHandle.h"
#include "llvm/Pass.h"
-class Method;
+#include <map>
+
+namespace llvm {
+
+class Function;
class Module;
+class CallGraphNode;
-namespace cfg {
+/// \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;
-class CallGraph;
-class CallGraphNode {
- Method *Meth;
- std::vector<CallGraphNode*> CalledMethods;
+ typedef std::map<const Function *, CallGraphNode *> FunctionMapTy;
+
+ /// \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);
- CallGraphNode(const CallGraphNode &); // Do not implement
public:
- typedef std::vector<CallGraphNode*>::iterator iterator;
- typedef std::vector<CallGraphNode*>::const_iterator const_iterator;
+ CallGraph(Module &M);
+ ~CallGraph();
- // getMethod - Return the method that this call graph node represents...
- Method *getMethod() const { return Meth; }
+ void print(raw_ostream &OS) const;
+ void dump() const;
- inline iterator begin() { return CalledMethods.begin(); }
- inline iterator end() { return CalledMethods.end(); }
- inline const_iterator begin() const { return CalledMethods.begin(); }
- inline const_iterator end() const { return CalledMethods.end(); }
- inline unsigned size() const { return CalledMethods.size(); }
+ typedef FunctionMapTy::iterator iterator;
+ typedef FunctionMapTy::const_iterator const_iterator;
- inline CallGraphNode *operator[](unsigned i) const { return CalledMethods[i];}
+ /// \brief Returns the module the call graph corresponds to.
+ Module &getModule() const { return M; }
- void removeAllCalledMethods() {
- CalledMethods.clear();
- }
+ 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(); }
-private: // Stuff to construct the node, used by CallGraph
- friend class CallGraph;
+ /// \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;
+ }
- // CallGraphNode ctor - Create a node for the specified method...
- inline CallGraphNode(Method *M) : Meth(M) {}
-
- // addCalledMethod add a method to the list of methods called by this one
- void addCalledMethod(CallGraphNode *M) {
- CalledMethods.push_back(M);
+ /// \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; }
-class CallGraph : public Pass {
- Module *Mod; // The module this call graph represents
+ CallGraphNode *getCallsExternalNode() const { return CallsExternalNode; }
- typedef std::map<const Method *, CallGraphNode *> MethodMapTy;
- MethodMapTy MethodMap; // Map from a method to its node
+ //===---------------------------------------------------------------------
+ // Functions to keep a call graph up to date with a function that has been
+ // modified.
+ //
- CallGraphNode *Root;
-public:
- static AnalysisID ID; // We are an analysis, we must have an ID
+ /// \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);
+
+ /// \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);
+};
- CallGraph(AnalysisID AID) : Root(0) { assert(AID == ID); }
- ~CallGraph() { destroy(); }
+/// \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 {
+public:
+ /// \brief A pair of the calling instruction (a call or invoke)
+ /// and the call graph node being called.
+ typedef std::pair<WeakVH, CallGraphNode *> CallRecord;
- typedef MethodMapTy::iterator iterator;
- typedef MethodMapTy::const_iterator const_iterator;
+public:
+ typedef std::vector<CallRecord> CalledFunctionsVector;
- inline CallGraphNode *getRoot() { return Root; }
- inline const CallGraphNode *getRoot() const { return Root; }
- inline iterator begin() { return MethodMap.begin(); }
- inline iterator end() { return MethodMap.end(); }
- inline const_iterator begin() const { return MethodMap.begin(); }
- inline const_iterator end() const { return MethodMap.end(); }
+ /// \brief Creates a node for the specified function.
+ inline CallGraphNode(Function *F) : F(F), NumReferences(0) {}
- inline const CallGraphNode *operator[](const Method *M) const {
- const_iterator I = MethodMap.find(M);
- assert(I != MethodMap.end() && "Method not in callgraph!");
- return I->second;
+ ~CallGraphNode() {
+ assert(NumReferences == 0 && "Node deleted while references remain");
}
- inline CallGraphNode *operator[](const Method *M) {
- const_iterator I = MethodMap.find(M);
- assert(I != MethodMap.end() && "Method not in callgraph!");
- return I->second;
+
+ typedef std::vector<CallRecord>::iterator iterator;
+ typedef std::vector<CallRecord>::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;
}
- // Methods to keep a call graph up to date with a method that has been
+ /// \brief Print out this call graph node.
+ void dump() const;
+ void print(raw_ostream &OS) const;
+
+ //===---------------------------------------------------------------------
+ // Methods to keep a call graph up to date with a function that has been
// modified
//
- void addMethodToModule(Method *Meth); // TODO IMPLEMENT
+ /// \brief Removes all edges from this CallGraphNode to any functions it
+ /// calls.
+ void removeAllCalledFunctions() {
+ while (!CalledFunctions.empty()) {
+ CalledFunctions.back().second->DropRef();
+ CalledFunctions.pop_back();
+ }
+ }
- // removeMethodFromModule - Unlink the method from this module, returning it.
- // Because this removes the method from the module, the call graph node is
- // destroyed. This is only valid if the method does not call any other
- // methods (ie, there are no edges in it's CGN). The easiest way to do this
- // is to dropAllReferences before calling this.
- //
- Method *removeMethodFromModule(CallGraphNode *CGN);
- Method *removeMethodFromModule(Method *Meth) {
- return removeMethodFromModule((*this)[Meth]);
+ /// \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);
}
- // run - Compute the call graph for the specified module.
- virtual bool run(Module *TheModule);
+ /// \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();
+ }
- // getAnalysisUsageInfo - This obviously provides a call graph
- virtual void getAnalysisUsageInfo(AnalysisSet &Required,
- AnalysisSet &Destroyed,
- AnalysisSet &Provided) {
- Provided.push_back(ID);
+ void removeCallEdge(iterator I) {
+ I->second->DropRef();
+ *I = CalledFunctions.back();
+ CalledFunctions.pop_back();
}
-private: // Implementation of CallGraph construction
- void destroy();
+ /// \brief Removes the edge in the node for the specified call site.
+ ///
+ /// Note that this method takes linear time, so it should be used sparingly.
+ void removeCallEdgeFor(CallSite CS);
+
+ /// \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);
+
+ /// \brief Removes one edge associated with a null callsite from this node to
+ /// the specified callee function.
+ void removeOneAbstractEdgeTo(CallGraphNode *Callee);
+
+ /// \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);
+
+private:
+ friend class CallGraph;
- // getNodeFor - Return the node for the specified method or create one if it
- // does not already exist.
- //
- CallGraphNode *getNodeFor(Method *M);
+ AssertingVH<Function> F;
- // addToCallGraph - Add a method to the call graph, and link the node to all
- // of the methods that it calls.
- //
- void addToCallGraph(Method *M);
+ std::vector<CallRecord> 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; }
+
+ /// \brief A special function that should only be used by the CallGraph class.
+ void allReferencesDropped() { NumReferences = 0; }
};
-} // end namespace cfg
+/// \brief An analysis pass to compute the \c CallGraph for a \c Module.
+///
+/// This class implements the concept of an analysis pass used by the \c
+/// ModuleAnalysisManager to run an analysis over a module and cache the
+/// resulting data.
+class CallGraphAnalysis {
+public:
+ /// \brief A formulaic typedef to inform clients of the result type.
+ typedef CallGraph Result;
+
+ static void *ID() { return (void *)&PassID; }
+
+ /// \brief Compute the \c CallGraph for the module \c M.
+ ///
+ /// The real work here is done in the \c CallGraph constructor.
+ CallGraph run(Module *M) { return CallGraph(*M); }
+
+private:
+ static char PassID;
+};
+
+/// \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 {
+ std::unique_ptr<CallGraph> G;
+
+public:
+ static char ID; // Class identification, replacement for typeinfo
+
+ CallGraphWrapperPass();
+ virtual ~CallGraphWrapperPass();
+
+ /// \brief The internal \c CallGraph around which the rest of this interface
+ /// is wrapped.
+ const CallGraph &getCallGraph() const { return *G; }
+ CallGraph &getCallGraph() { return *G; }
+
+ typedef CallGraph::iterator iterator;
+ typedef CallGraph::const_iterator const_iterator;
+ /// \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();
+ }
+
+ //===---------------------------------------------------------------------
+ // Functions to keep a call graph up to date with a function that has been
+ // modified.
+ //
+
+ /// \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);
+ }
+
+ /// \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.
+ //
+
+ void getAnalysisUsage(AnalysisUsage &AU) const override;
+ bool runOnModule(Module &M) override;
+ void releaseMemory() override;
+
+ void print(raw_ostream &o, const Module *) const override;
+ void dump() const;
+};
+
+//===----------------------------------------------------------------------===//
+// GraphTraits specializations for call graphs so that they can be treated as
+// graphs by the generic graph algorithms.
+//
// Provide graph traits for tranversing call graphs using standard graph
// traversals.
-template <> struct GraphTraits<cfg::CallGraphNode*> {
- typedef cfg::CallGraphNode NodeType;
- typedef NodeType::iterator ChildIteratorType;
+template <> struct GraphTraits<CallGraphNode *> {
+ typedef CallGraphNode NodeType;
+
+ typedef CallGraphNode::CallRecord CGNPairTy;
+ typedef std::pointer_to_unary_function<CGNPairTy, CallGraphNode *>
+ CGNDerefFun;
+
+ static NodeType *getEntryNode(CallGraphNode *CGN) { return CGN; }
- static NodeType *getEntryNode(cfg::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<NodeType::iterator, CGNDerefFun> 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<const cfg::CallGraphNode*> {
- typedef const cfg::CallGraphNode NodeType;
+template <> struct GraphTraits<const CallGraphNode *> {
+ typedef const CallGraphNode NodeType;
typedef NodeType::const_iterator ChildIteratorType;
- static NodeType *getEntryNode(const cfg::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 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(); }
};
+template <>
+struct GraphTraits<CallGraph *> : public GraphTraits<CallGraphNode *> {
+ static NodeType *getEntryNode(CallGraph *CGN) {
+ return CGN->getExternalCallingNode(); // Start at the external node!
+ }
+ typedef std::pair<const Function *, CallGraphNode *> PairTy;
+ typedef std::pointer_to_unary_function<PairTy, CallGraphNode &> DerefFun;
-template<> struct GraphTraits<cfg::CallGraph*> :
- public GraphTraits<cfg::CallGraphNode*> {
- static NodeType *getEntryNode(cfg::CallGraph *CGN) {
- return CGN->getRoot();
+ // nodes_iterator/begin/end - Allow iteration over all nodes in the graph
+ typedef mapped_iterator<CallGraph::iterator, DerefFun> nodes_iterator;
+ static nodes_iterator nodes_begin(CallGraph *CG) {
+ return map_iterator(CG->begin(), DerefFun(CGdereference));
}
-};
-template<> struct GraphTraits<const cfg::CallGraph*> :
- public GraphTraits<const cfg::CallGraphNode*> {
- static NodeType *getEntryNode(const cfg::CallGraph *CGN) {
- return CGN->getRoot();
+ static nodes_iterator nodes_end(CallGraph *CG) {
+ return map_iterator(CG->end(), DerefFun(CGdereference));
}
+
+ static CallGraphNode &CGdereference(PairTy P) { return *P.second; }
};
+template <>
+struct GraphTraits<const CallGraph *> : public GraphTraits<
+ const CallGraphNode *> {
+ static NodeType *getEntryNode(const CallGraph *CGN) {
+ 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(); }
+};
-// Checks if a method contains any call instructions.
-// Note that this uses the call graph only if one is provided.
-// It does not build the call graph.
-//
-bool isLeafMethod(const Method* method, const cfg::CallGraph *callGraph = 0);
+} // End llvm namespace
#endif