//===- CallGraph.cpp - Build a Module's call graph ------------------------===//
-//
-// The LLVM Compiler Infrastructure
-//
-// This file was developed by the LLVM research group and is distributed under
-// the University of Illinois Open Source License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This interface is 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 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.
+// The LLVM Compiler Infrastructure
//
-// 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.
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
//
-// 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 implements the CallGraph class and provides the BasicCallGraph
+// default implementation.
//
//===----------------------------------------------------------------------===//
#include "llvm/Analysis/CallGraph.h"
-#include "llvm/Constants.h" // Remove when ConstantPointerRefs are gone
#include "llvm/Module.h"
-#include "llvm/iOther.h"
-#include "llvm/iTerminators.h"
+#include "llvm/Instructions.h"
#include "llvm/Support/CallSite.h"
-#include "Support/STLExtras.h"
-
-static RegisterAnalysis<CallGraph> X("callgraph", "Call Graph Construction");
-
-static const char * const KnownExternalFunctions[] = {
- // Low-level system calls
- "open",
- "read",
- "write",
- "writev",
- "lseek",
- "poll",
- "ioctl",
-
- // Low-level stdc library functions
- "abort", "exit",
- "getenv", "putenv",
-
- // Standard IO functions
- "printf",
- "sprintf",
- "fopen",
- "freopen",
- "fclose",
- "fwrite",
- "puts",
- "fputs",
- "getc",
- "ungetc",
- "putc",
- "putchar",
- "fread",
- "fileno",
- "ftell",
- "fflush",
- "fseek",
- "fileno",
- "ferror",
- "feof",
- "fdopen",
- "__fxstat",
- "setbuf",
- "setbuffer",
- "etlinebuf",
- "setvbuf",
-
- // Memory functions
- "malloc",
- "free",
- "realloc",
- "calloc",
- "memalign",
-
- // String functions
- "atoi",
- "memmove",
- "memset",
- "memchr",
- "memcmp",
- "strchr",
- "strncpy",
- "strncmp",
- "strcmp",
- "strtok",
- "__strcoll_l",
- "__strxfrm_l",
- "__strftime_l",
- "__strtol_l",
- "__strtoul_l",
- "__strtoll_l",
- "__strtoull_l",
- "__strtof_l",
- "__strtod_l",
- "__strtold_l",
- "isalpha",
-
- // Math functions
- "exp", "sqrt", "cbrt", "hypot",
- "log", "log10", "pow",
- "sin", "cos", "tan",
- "asin", "acos", "atan", "atan2",
-
- // Locale functions
- "__uselocale",
- "__newlocale",
- "__freelocale",
- "__duplocale",
- "__nl_langinfo_l",
-
- // gettext functions used by libstdc++
- "gettext",
- "dgettext",
- "dcgettext",
- "textdomain",
- "bindtextdomain",
-
- // Random stuff
- "__assert_fail",
- "__errno_location",
- "clock", "time",
- "__main",
-};
+#include "llvm/Support/Compiler.h"
+#include "llvm/Support/Streams.h"
+#include <ostream>
+using namespace llvm;
+
+/// isOnlyADirectCall - Return true if this callsite is *just* a direct call to
+/// the specified function. Specifically return false if the callsite also
+/// takes the address of the function.
+static bool isOnlyADirectCall(Function *F, CallSite CS) {
+ if (!CS.getInstruction()) return false;
+ for (CallSite::arg_iterator I = CS.arg_begin(), E = CS.arg_end(); I != E; ++I)
+ if (*I == F) return false;
+ return true;
+}
+namespace {
-/// ExternalFunctionDoesntCallIntoProgram - This hack is used to indicate to the
-/// call graph that the specified external function is _KNOWN_ to not call back
-/// into the program. This is important, because otherwise functions which call
-/// "printf" for example, end up in a great big SCC that goes from the function
-/// through main.
-///
-static bool ExternalFunctionDoesntCallIntoProgram(const std::string &Name) {
- static std::vector<std::string> Funcs;
-
- // First time this is called?
- if (Funcs.empty()) {
- // Add a whole bunch of functions which are often used...
- Funcs.insert(Funcs.end(), KnownExternalFunctions,
- KnownExternalFunctions+
- sizeof(KnownExternalFunctions)/sizeof(KnownExternalFunctions[0]));
- // Sort the list for efficient access
- std::sort(Funcs.begin(), Funcs.end());
+//===----------------------------------------------------------------------===//
+// BasicCallGraph class definition
+//
+class VISIBILITY_HIDDEN BasicCallGraph : public CallGraph, public ModulePass {
+ // Root is root of the call graph, or the external node if a 'main' function
+ // couldn't be found.
+ //
+ CallGraphNode *Root;
+
+ // ExternalCallingNode - This node has edges to all external functions and
+ // those internal functions that have their address taken.
+ CallGraphNode *ExternalCallingNode;
+
+ // CallsExternalNode - This node has edges to it from all functions making
+ // indirect calls or calling an external function.
+ CallGraphNode *CallsExternalNode;
+
+public:
+ static char ID; // Class identification, replacement for typeinfo
+ BasicCallGraph() : ModulePass((intptr_t)&ID), Root(0),
+ ExternalCallingNode(0), CallsExternalNode(0) {}
+
+ // runOnModule - Compute the call graph for the specified module.
+ virtual bool runOnModule(Module &M) {
+ CallGraph::initialize(M);
+
+ ExternalCallingNode = getOrInsertFunction(0);
+ CallsExternalNode = new CallGraphNode(0);
+ Root = 0;
+
+ // Add every function to the call graph...
+ for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
+ addToCallGraph(I);
+
+ // If we didn't find a main function, use the external call graph node
+ if (Root == 0) Root = ExternalCallingNode;
+
+ return false;
}
- if (Name.size() > 7 && !memcmp("__llvm_", Name.c_str(), 7))
- return true;
-
- // Binary search for the function name...
- std::vector<std::string>::iterator I =
- std::lower_bound(Funcs.begin(), Funcs.end(), Name);
+ virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+ AU.setPreservesAll();
+ }
- // Found it?
- return I != Funcs.end() && *I == Name;
-}
+ void print(std::ostream *o, const Module *M) const {
+ if (o) print(*o, M);
+ }
+ virtual void print(std::ostream &o, const Module *M) const {
+ o << "CallGraph Root is: ";
+ if (Function *F = getRoot()->getFunction())
+ o << F->getName() << "\n";
+ else
+ o << "<<null function: 0x" << getRoot() << ">>\n";
+
+ CallGraph::print(o, M);
+ }
+ virtual void releaseMemory() {
+ destroy();
+ }
+
+ /// dump - Print out this call graph.
+ ///
+ inline void dump() const {
+ print(cerr, Mod);
+ }
-// getNodeFor - Return the node for the specified function or create one if it
-// does not already exist.
-//
-CallGraphNode *CallGraph::getNodeFor(Function *F) {
- CallGraphNode *&CGN = FunctionMap[F];
- if (CGN) return CGN;
+ CallGraphNode* getExternalCallingNode() const { return ExternalCallingNode; }
+ CallGraphNode* getCallsExternalNode() const { return CallsExternalNode; }
- assert((!F || F->getParent() == Mod) && "Function not in current module!");
- return CGN = new CallGraphNode(F);
-}
+ // 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; }
-static bool isOnlyADirectCall(Function *F, CallSite CS) {
- if (!CS.getInstruction()) return false;
- for (CallSite::arg_iterator I = CS.arg_begin(), E = CS.arg_end(); I != E; ++I)
- if (*I == F) return false;
- return true;
-}
+private:
+ //===---------------------------------------------------------------------
+ // Implementation of CallGraph construction
+ //
-// addToCallGraph - Add a function to the call graph, and link the node to all
-// of the functions that it calls.
-//
-void CallGraph::addToCallGraph(Function *F) {
- CallGraphNode *Node = getNodeFor(F);
-
- // If this function has external linkage, anything could call it...
- if (!F->hasInternalLinkage()) {
- ExternalNode->addCalledFunction(Node);
-
- // Found the entry point?
- if (F->getName() == "main") {
- if (Root)
- Root = ExternalNode; // Found multiple external mains? Don't pick one.
- else
- Root = Node; // Found a main, keep track of it!
+ // addToCallGraph - Add a function to the call graph, and link the node to all
+ // of the functions that it calls.
+ //
+ void addToCallGraph(Function *F) {
+ CallGraphNode *Node = getOrInsertFunction(F);
+
+ // If this function has external linkage, anything could call it.
+ if (!F->hasInternalLinkage()) {
+ ExternalCallingNode->addCalledFunction(CallSite(), Node);
+
+ // Found the entry point?
+ if (F->getName() == "main") {
+ if (Root) // Found multiple external mains? Don't pick one.
+ Root = ExternalCallingNode;
+ else
+ Root = Node; // Found a main, keep track of it!
+ }
}
- }
-
- // If this function is not defined in this translation unit, it could call
- // anything.
- if (F->isExternal() && !F->getIntrinsicID() &&
- !ExternalFunctionDoesntCallIntoProgram(F->getName()))
- Node->addCalledFunction(ExternalNode);
- // Loop over all of the users of the function... looking for callers...
- //
- bool isUsedExternally = false;
- for (Value::use_iterator I = F->use_begin(), E = F->use_end(); I != E; ++I) {
- if (Instruction *Inst = dyn_cast<Instruction>(*I)) {
- if (isOnlyADirectCall(F, CallSite::get(Inst)))
- getNodeFor(Inst->getParent()->getParent())->addCalledFunction(Node);
- else
- isUsedExternally = true;
- } else if (ConstantPointerRef *CPR = dyn_cast<ConstantPointerRef>(*I)) {
- // THIS IS A DISGUSTING HACK. Brought to you by the power of
- // ConstantPointerRefs!
- for (Value::use_iterator I = CPR->use_begin(), E = CPR->use_end();
- I != E; ++I)
- if (Instruction *Inst = dyn_cast<Instruction>(*I)) {
- if (isOnlyADirectCall(F, CallSite::get(Inst)))
- getNodeFor(Inst->getParent()->getParent())->addCalledFunction(Node);
- else
- isUsedExternally = true;
- } else {
+ // If this function is not defined in this translation unit, it could call
+ // anything.
+ if (F->isDeclaration() && !F->isIntrinsic())
+ Node->addCalledFunction(CallSite(), CallsExternalNode);
+
+ // Loop over all of the users of the function... looking for callers...
+ //
+ bool isUsedExternally = false;
+ for (Value::use_iterator I = F->use_begin(), E = F->use_end(); I != E; ++I){
+ if (Instruction *Inst = dyn_cast<Instruction>(*I)) {
+ CallSite CS = CallSite::get(Inst);
+ if (isOnlyADirectCall(F, CS))
+ getOrInsertFunction(Inst->getParent()->getParent())
+ ->addCalledFunction(CS, Node);
+ else
isUsedExternally = true;
- }
- } else { // Can't classify the user!
- isUsedExternally = true;
+ } else if (GlobalValue *GV = dyn_cast<GlobalValue>(*I)) {
+ for (Value::use_iterator I = GV->use_begin(), E = GV->use_end();
+ I != E; ++I)
+ if (Instruction *Inst = dyn_cast<Instruction>(*I)) {
+ CallSite CS = CallSite::get(Inst);
+ if (isOnlyADirectCall(F, CS))
+ getOrInsertFunction(Inst->getParent()->getParent())
+ ->addCalledFunction(CS, Node);
+ else
+ isUsedExternally = true;
+ } else {
+ isUsedExternally = true;
+ }
+ } else { // Can't classify the user!
+ isUsedExternally = true;
+ }
}
- }
- if (isUsedExternally)
- ExternalNode->addCalledFunction(Node);
+ if (isUsedExternally)
+ ExternalCallingNode->addCalledFunction(CallSite(), Node);
- // Look for an indirect function call...
- for (Function::iterator BB = F->begin(), BBE = F->end(); BB != BBE; ++BB)
- for (BasicBlock::iterator II = BB->begin(), IE = BB->end(); II != IE; ++II){
+ // Look for an indirect function call.
+ for (Function::iterator BB = F->begin(), BBE = F->end(); BB != BBE; ++BB)
+ for (BasicBlock::iterator II = BB->begin(), IE = BB->end();
+ II != IE; ++II) {
CallSite CS = CallSite::get(II);
if (CS.getInstruction() && !CS.getCalledFunction())
- Node->addCalledFunction(ExternalNode);
- }
-}
+ Node->addCalledFunction(CS, CallsExternalNode);
+ }
+ }
-bool CallGraph::run(Module &M) {
- destroy();
+ //
+ // destroy - Release memory for the call graph
+ virtual void destroy() {
+ /// CallsExternalNode is not in the function map, delete it explicitly.
+ delete CallsExternalNode;
+ CallsExternalNode = 0;
+ CallGraph::destroy();
+ }
+};
- Mod = &M;
- ExternalNode = getNodeFor(0);
- Root = 0;
+} //End anonymous namespace
- // Add every function to the call graph...
- for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
- addToCallGraph(I);
+static RegisterAnalysisGroup<CallGraph> X("Call Graph");
+static RegisterPass<BasicCallGraph>
+Y("basiccg", "Basic CallGraph Construction", false, true);
+static RegisterAnalysisGroup<CallGraph, true> Z(Y);
- // If we didn't find a main function, use the external call graph node
- if (Root == 0) Root = ExternalNode;
-
- return false;
-}
+char CallGraph::ID = 0;
+char BasicCallGraph::ID = 0;
-void CallGraph::destroy() {
- for (FunctionMapTy::iterator I = FunctionMap.begin(), E = FunctionMap.end();
- I != E; ++I)
- delete I->second;
- FunctionMap.clear();
+void CallGraph::initialize(Module &M) {
+ Mod = &M;
}
-static void WriteToOutput(const CallGraphNode *CGN, std::ostream &o) {
- if (CGN->getFunction())
- o << "Call graph node for function: '"
- << CGN->getFunction()->getName() <<"'\n";
- else
- o << "Call graph node <<null function: 0x" << CGN << ">>:\n";
-
- for (unsigned i = 0; i < CGN->size(); ++i)
- if ((*CGN)[i]->getFunction())
- o << " Calls function '" << (*CGN)[i]->getFunction()->getName() << "'\n";
- else
- o << " Calls external node\n";
- o << "\n";
+void CallGraph::destroy() {
+ if (!FunctionMap.empty()) {
+ for (FunctionMapTy::iterator I = FunctionMap.begin(), E = FunctionMap.end();
+ I != E; ++I)
+ delete I->second;
+ FunctionMap.clear();
+ }
}
-void CallGraph::print(std::ostream &o, const Module *M) const {
- o << "CallGraph Root is: ";
- if (getRoot()->getFunction())
- o << getRoot()->getFunction()->getName() << "\n";
- else
- o << "<<null function: 0x" << getRoot() << ">>\n";
-
+void CallGraph::print(std::ostream &OS, const Module *M) const {
for (CallGraph::const_iterator I = begin(), E = end(); I != E; ++I)
- WriteToOutput(I->second, o);
+ I->second->print(OS);
}
+void CallGraph::dump() const {
+ print(cerr, 0);
+}
//===----------------------------------------------------------------------===//
// Implementations of public modification methods
//
-// Functions to keep a call graph up to date with a function that has been
-// modified
-//
-void CallGraph::addFunctionToModule(Function *Meth) {
- assert(0 && "not implemented");
- abort();
-}
-
// 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
return F;
}
-void CallGraph::stub() {}
+// changeFunction - This method changes the function associated with this
+// CallGraphNode, for use by transformations that need to change the prototype
+// of a Function (thus they must create a new Function and move the old code
+// over).
+void CallGraph::changeFunction(Function *OldF, Function *NewF) {
+ iterator I = FunctionMap.find(OldF);
+ CallGraphNode *&New = FunctionMap[NewF];
+ assert(I != FunctionMap.end() && I->second && !New &&
+ "OldF didn't exist in CG or NewF already does!");
+ New = I->second;
+ New->F = NewF;
+ FunctionMap.erase(I);
+}
+
+// getOrInsertFunction - This method is identical to calling operator[], but
+// it will insert a new CallGraphNode for the specified function if one does
+// not already exist.
+CallGraphNode *CallGraph::getOrInsertFunction(const Function *F) {
+ CallGraphNode *&CGN = FunctionMap[F];
+ if (CGN) return CGN;
+
+ assert((!F || F->getParent() == Mod) && "Function not in current module!");
+ return CGN = new CallGraphNode(const_cast<Function*>(F));
+}
+
+void CallGraphNode::print(std::ostream &OS) const {
+ if (Function *F = getFunction())
+ OS << "Call graph node for function: '" << F->getName() <<"'\n";
+ else
+ OS << "Call graph node <<null function: 0x" << this << ">>:\n";
+
+ for (const_iterator I = begin(), E = end(); I != E; ++I)
+ if (I->second->getFunction())
+ OS << " Calls function '" << I->second->getFunction()->getName() <<"'\n";
+ else
+ OS << " Calls external node\n";
+ OS << "\n";
+}
+
+void CallGraphNode::dump() const { print(cerr); }
+
+void CallGraphNode::removeCallEdgeTo(CallGraphNode *Callee) {
+ for (unsigned i = CalledFunctions.size(); ; --i) {
+ assert(i && "Cannot find callee to remove!");
+ if (CalledFunctions[i-1].second == Callee) {
+ CalledFunctions.erase(CalledFunctions.begin()+i-1);
+ return;
+ }
+ }
+}
+
+/// removeCallEdgeFor - This method 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 CallGraphNode::removeCallEdgeFor(CallSite CS) {
+ for (unsigned i = CalledFunctions.size(); ; --i) {
+ assert(i && "Cannot find callee to remove!");
+ if (CalledFunctions[i-1].first == CS) {
+ CalledFunctions.erase(CalledFunctions.begin()+i-1);
+ return;
+ }
+ }
+}
+
+
+// removeAnyCallEdgeTo - This method removes any 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 CallGraphNode::removeAnyCallEdgeTo(CallGraphNode *Callee) {
+ for (unsigned i = 0, e = CalledFunctions.size(); i != e; ++i)
+ if (CalledFunctions[i].second == Callee) {
+ CalledFunctions[i] = CalledFunctions.back();
+ CalledFunctions.pop_back();
+ --i; --e;
+ }
+}
+
+// Enuse that users of CallGraph.h also link with this file
+DEFINING_FILE_FOR(CallGraph)