//===-- AsmWriter.cpp - Printing LLVM as an assembly file -----------------===// // // This library implements the functionality defined in llvm/Assembly/Writer.h // // Note that these routines must be extremely tolerant of various errors in the // LLVM code, because of of the primary uses of it is for debugging // transformations. // //===----------------------------------------------------------------------===// #include "llvm/Assembly/CachedWriter.h" #include "llvm/Assembly/Writer.h" #include "llvm/SlotCalculator.h" #include "llvm/DerivedTypes.h" #include "llvm/Module.h" #include "llvm/Function.h" #include "llvm/GlobalVariable.h" #include "llvm/BasicBlock.h" #include "llvm/Constants.h" #include "llvm/iMemory.h" #include "llvm/iTerminators.h" #include "llvm/iPHINode.h" #include "llvm/iOther.h" #include "llvm/SymbolTable.h" #include "llvm/Argument.h" #include "Support/StringExtras.h" #include "Support/STLExtras.h" #include using std::string; using std::map; using std::vector; using std::ostream; static void WriteAsOperandInternal(ostream &Out, const Value *V, bool PrintName, map &TypeTable, SlotCalculator *Table); static const Module *getModuleFromVal(const Value *V) { if (const Argument *MA = dyn_cast(V)) return MA->getParent() ? MA->getParent()->getParent() : 0; else if (const BasicBlock *BB = dyn_cast(V)) return BB->getParent() ? BB->getParent()->getParent() : 0; else if (const Instruction *I = dyn_cast(V)) { const Function *M = I->getParent() ? I->getParent()->getParent() : 0; return M ? M->getParent() : 0; } else if (const GlobalValue *GV = dyn_cast(V)) return GV->getParent(); return 0; } static SlotCalculator *createSlotCalculator(const Value *V) { assert(!isa(V) && "Can't create an SC for a type!"); if (const Argument *FA = dyn_cast(V)) { return new SlotCalculator(FA->getParent(), true); } else if (const Instruction *I = dyn_cast(V)) { return new SlotCalculator(I->getParent()->getParent(), true); } else if (const BasicBlock *BB = dyn_cast(V)) { return new SlotCalculator(BB->getParent(), true); } else if (const GlobalVariable *GV = dyn_cast(V)){ return new SlotCalculator(GV->getParent(), true); } else if (const Function *Func = dyn_cast(V)) { return new SlotCalculator(Func, true); } return 0; } // If the module has a symbol table, take all global types and stuff their // names into the TypeNames map. // static void fillTypeNameTable(const Module *M, map &TypeNames) { if (M && M->hasSymbolTable()) { const SymbolTable *ST = M->getSymbolTable(); SymbolTable::const_iterator PI = ST->find(Type::TypeTy); if (PI != ST->end()) { SymbolTable::type_const_iterator I = PI->second.begin(); for (; I != PI->second.end(); ++I) { // As a heuristic, don't insert pointer to primitive types, because // they are used too often to have a single useful name. // const Type *Ty = cast(I->second); if (!isa(Ty) || !cast(Ty)->getElementType()->isPrimitiveType()) TypeNames.insert(std::make_pair(Ty, "%"+I->first)); } } } } static string calcTypeName(const Type *Ty, vector &TypeStack, map &TypeNames) { if (Ty->isPrimitiveType()) return Ty->getDescription(); // Base case // Check to see if the type is named. map::iterator I = TypeNames.find(Ty); if (I != TypeNames.end()) return I->second; // Check to see if the Type is already on the stack... unsigned Slot = 0, CurSize = TypeStack.size(); while (Slot < CurSize && TypeStack[Slot] != Ty) ++Slot; // Scan for type // This is another base case for the recursion. In this case, we know // that we have looped back to a type that we have previously visited. // Generate the appropriate upreference to handle this. // if (Slot < CurSize) return "\\" + utostr(CurSize-Slot); // Here's the upreference TypeStack.push_back(Ty); // Recursive case: Add us to the stack.. string Result; switch (Ty->getPrimitiveID()) { case Type::FunctionTyID: { const FunctionType *FTy = cast(Ty); Result = calcTypeName(FTy->getReturnType(), TypeStack, TypeNames) + " ("; for (FunctionType::ParamTypes::const_iterator I = FTy->getParamTypes().begin(), E = FTy->getParamTypes().end(); I != E; ++I) { if (I != FTy->getParamTypes().begin()) Result += ", "; Result += calcTypeName(*I, TypeStack, TypeNames); } if (FTy->isVarArg()) { if (!FTy->getParamTypes().empty()) Result += ", "; Result += "..."; } Result += ")"; break; } case Type::StructTyID: { const StructType *STy = cast(Ty); Result = "{ "; for (StructType::ElementTypes::const_iterator I = STy->getElementTypes().begin(), E = STy->getElementTypes().end(); I != E; ++I) { if (I != STy->getElementTypes().begin()) Result += ", "; Result += calcTypeName(*I, TypeStack, TypeNames); } Result += " }"; break; } case Type::PointerTyID: Result = calcTypeName(cast(Ty)->getElementType(), TypeStack, TypeNames) + "*"; break; case Type::ArrayTyID: { const ArrayType *ATy = cast(Ty); Result = "[" + utostr(ATy->getNumElements()) + " x "; Result += calcTypeName(ATy->getElementType(), TypeStack, TypeNames) + "]"; break; } default: assert(0 && "Unhandled case in getTypeProps!"); Result = ""; } TypeStack.pop_back(); // Remove self from stack... return Result; } // printTypeInt - The internal guts of printing out a type that has a // potentially named portion. // static ostream &printTypeInt(ostream &Out, const Type *Ty, map &TypeNames) { // Primitive types always print out their description, regardless of whether // they have been named or not. // if (Ty->isPrimitiveType()) return Out << Ty->getDescription(); // Check to see if the type is named. map::iterator I = TypeNames.find(Ty); if (I != TypeNames.end()) return Out << I->second; // Otherwise we have a type that has not been named but is a derived type. // Carefully recurse the type hierarchy to print out any contained symbolic // names. // vector TypeStack; string TypeName = calcTypeName(Ty, TypeStack, TypeNames); TypeNames.insert(std::make_pair(Ty, TypeName));//Cache type name for later use return Out << TypeName; } // WriteTypeSymbolic - This attempts to write the specified type as a symbolic // type, iff there is an entry in the modules symbol table for the specified // type or one of it's component types. This is slower than a simple x << Type; // ostream &WriteTypeSymbolic(ostream &Out, const Type *Ty, const Module *M) { Out << " "; // If they want us to print out a type, attempt to make it symbolic if there // is a symbol table in the module... if (M && M->hasSymbolTable()) { map TypeNames; fillTypeNameTable(M, TypeNames); return printTypeInt(Out, Ty, TypeNames); } else { return Out << Ty->getDescription(); } } static void WriteConstantInt(ostream &Out, const Constant *CV, bool PrintName, map &TypeTable, SlotCalculator *Table) { if (const ConstantBool *CB = dyn_cast(CV)) { Out << (CB == ConstantBool::True ? "true" : "false"); } else if (const ConstantSInt *CI = dyn_cast(CV)) { Out << CI->getValue(); } else if (const ConstantUInt *CI = dyn_cast(CV)) { Out << CI->getValue(); } else if (const ConstantFP *CFP = dyn_cast(CV)) { // We would like to output the FP constant value in exponential notation, // but we cannot do this if doing so will lose precision. Check here to // make sure that we only output it in exponential format if we can parse // the value back and get the same value. // std::string StrVal = ftostr(CFP->getValue()); // Check to make sure that the stringized number is not some string like // "Inf" or NaN, that atof will accept, but the lexer will not. Check that // the string matches the "[-+]?[0-9]" regex. // if ((StrVal[0] >= '0' && StrVal[0] <= '9') || ((StrVal[0] == '-' || StrVal[0] == '+') && (StrVal[0] >= '0' && StrVal[0] <= '9'))) // Reparse stringized version! if (atof(StrVal.c_str()) == CFP->getValue()) { Out << StrVal; return; } // Otherwise we could not reparse it to exactly the same value, so we must // output the string in hexadecimal format! // // Behave nicely in the face of C TBAA rules... see: // http://www.nullstone.com/htmls/category/aliastyp.htm // double Val = CFP->getValue(); char *Ptr = (char*)&Val; assert(sizeof(double) == sizeof(uint64_t) && sizeof(double) == 8 && "assuming that double is 64 bits!"); Out << "0x" << utohexstr(*(uint64_t*)Ptr); } else if (const ConstantArray *CA = dyn_cast(CV)) { // As a special case, print the array as a string if it is an array of // ubytes or an array of sbytes with positive values. // const Type *ETy = CA->getType()->getElementType(); bool isString = (ETy == Type::SByteTy || ETy == Type::UByteTy); if (ETy == Type::SByteTy) for (unsigned i = 0; i < CA->getNumOperands(); ++i) if (cast(CA->getOperand(i))->getValue() < 0) { isString = false; break; } if (isString) { Out << "c\""; for (unsigned i = 0; i < CA->getNumOperands(); ++i) { unsigned char C = (ETy == Type::SByteTy) ? (unsigned char)cast(CA->getOperand(i))->getValue() : (unsigned char)cast(CA->getOperand(i))->getValue(); if (isprint(C)) { Out << C; } else { Out << '\\' << (char) ((C/16 < 10) ? ( C/16 +'0') : ( C/16 -10+'A')) << (char)(((C&15) < 10) ? ((C&15)+'0') : ((C&15)-10+'A')); } } Out << "\""; } else { // Cannot output in string format... Out << "["; if (CA->getNumOperands()) { Out << " "; printTypeInt(Out, ETy, TypeTable); WriteAsOperandInternal(Out, CA->getOperand(0), PrintName, TypeTable, Table); for (unsigned i = 1, e = CA->getNumOperands(); i != e; ++i) { Out << ", "; printTypeInt(Out, ETy, TypeTable); WriteAsOperandInternal(Out, CA->getOperand(i), PrintName, TypeTable, Table); } } Out << " ]"; } } else if (const ConstantStruct *CS = dyn_cast(CV)) { Out << "{"; if (CS->getNumOperands()) { Out << " "; printTypeInt(Out, CS->getOperand(0)->getType(), TypeTable); WriteAsOperandInternal(Out, CS->getOperand(0), PrintName, TypeTable, Table); for (unsigned i = 1; i < CS->getNumOperands(); i++) { Out << ", "; printTypeInt(Out, CS->getOperand(i)->getType(), TypeTable); WriteAsOperandInternal(Out, CS->getOperand(i), PrintName, TypeTable, Table); } } Out << " }"; } else if (isa(CV)) { Out << "null"; } else if (ConstantPointerRef *PR = dyn_cast(CV)) { const GlobalValue *V = PR->getValue(); if (V->hasName()) { Out << "%" << V->getName(); } else if (Table) { int Slot = Table->getValSlot(V); if (Slot >= 0) Out << "%" << Slot; else Out << ""; } else { Out << ""; } } else { assert(0 && "Unrecognized constant value!!!"); } } // WriteAsOperand - Write the name of the specified value out to the specified // ostream. This can be useful when you just want to print int %reg126, not the // whole instruction that generated it. // static void WriteAsOperandInternal(ostream &Out, const Value *V, bool PrintName, map &TypeTable, SlotCalculator *Table) { Out << " "; if (PrintName && V->hasName()) { Out << "%" << V->getName(); } else { if (const Constant *CV = dyn_cast(V)) { WriteConstantInt(Out, CV, PrintName, TypeTable, Table); } else { int Slot; if (Table) { Slot = Table->getValSlot(V); } else { if (const Type *Ty = dyn_cast(V)) { Out << Ty->getDescription(); return; } Table = createSlotCalculator(V); if (Table == 0) { Out << "BAD VALUE TYPE!"; return; } Slot = Table->getValSlot(V); delete Table; } if (Slot >= 0) Out << "%" << Slot; else if (PrintName) Out << ""; // Not embeded into a location? } } } // WriteAsOperand - Write the name of the specified value out to the specified // ostream. This can be useful when you just want to print int %reg126, not the // whole instruction that generated it. // ostream &WriteAsOperand(ostream &Out, const Value *V, bool PrintType, bool PrintName, SlotCalculator *Table) { map TypeNames; const Module *M = getModuleFromVal(V); if (M && M->hasSymbolTable()) fillTypeNameTable(M, TypeNames); if (PrintType) printTypeInt(Out, V->getType(), TypeNames); WriteAsOperandInternal(Out, V, PrintName, TypeNames, Table); return Out; } class AssemblyWriter { ostream &Out; SlotCalculator &Table; const Module *TheModule; map TypeNames; public: inline AssemblyWriter(ostream &o, SlotCalculator &Tab, const Module *M) : Out(o), Table(Tab), TheModule(M) { // If the module has a symbol table, take all global types and stuff their // names into the TypeNames map. // fillTypeNameTable(M, TypeNames); } inline void write(const Module *M) { printModule(M); } inline void write(const GlobalVariable *G) { printGlobal(G); } inline void write(const Function *F) { printFunction(F); } inline void write(const BasicBlock *BB) { printBasicBlock(BB); } inline void write(const Instruction *I) { printInstruction(I); } inline void write(const Constant *CPV) { printConstant(CPV); } inline void write(const Type *Ty) { printType(Ty); } void writeOperand(const Value *Op, bool PrintType, bool PrintName = true); private : void printModule(const Module *M); void printSymbolTable(const SymbolTable &ST); void printConstant(const Constant *CPV); void printGlobal(const GlobalVariable *GV); void printFunction(const Function *F); void printArgument(const Argument *FA); void printBasicBlock(const BasicBlock *BB); void printInstruction(const Instruction *I); // printType - Go to extreme measures to attempt to print out a short, // symbolic version of a type name. // ostream &printType(const Type *Ty) { return printTypeInt(Out, Ty, TypeNames); } // printTypeAtLeastOneLevel - Print out one level of the possibly complex type // without considering any symbolic types that we may have equal to it. // ostream &printTypeAtLeastOneLevel(const Type *Ty); // printInfoComment - Print a little comment after the instruction indicating // which slot it occupies. void printInfoComment(const Value *V); }; // printTypeAtLeastOneLevel - Print out one level of the possibly complex type // without considering any symbolic types that we may have equal to it. // ostream &AssemblyWriter::printTypeAtLeastOneLevel(const Type *Ty) { if (FunctionType *FTy = dyn_cast(Ty)) { printType(FTy->getReturnType()) << " ("; for (FunctionType::ParamTypes::const_iterator I = FTy->getParamTypes().begin(), E = FTy->getParamTypes().end(); I != E; ++I) { if (I != FTy->getParamTypes().begin()) Out << ", "; printType(*I); } if (FTy->isVarArg()) { if (!FTy->getParamTypes().empty()) Out << ", "; Out << "..."; } Out << ")"; } else if (StructType *STy = dyn_cast(Ty)) { Out << "{ "; for (StructType::ElementTypes::const_iterator I = STy->getElementTypes().begin(), E = STy->getElementTypes().end(); I != E; ++I) { if (I != STy->getElementTypes().begin()) Out << ", "; printType(*I); } Out << " }"; } else if (PointerType *PTy = dyn_cast(Ty)) { printType(PTy->getElementType()) << "*"; } else if (ArrayType *ATy = dyn_cast(Ty)) { Out << "[" << ATy->getNumElements() << " x "; printType(ATy->getElementType()) << "]"; } else { assert(Ty->isPrimitiveType() && "Unknown derived type!"); printType(Ty); } return Out; } void AssemblyWriter::writeOperand(const Value *Operand, bool PrintType, bool PrintName) { if (PrintType) { Out << " "; printType(Operand->getType()); } WriteAsOperandInternal(Out, Operand, PrintName, TypeNames, &Table); } void AssemblyWriter::printModule(const Module *M) { // Loop over the symbol table, emitting all named constants... if (M->hasSymbolTable()) printSymbolTable(*M->getSymbolTable()); for_each(M->gbegin(), M->gend(), bind_obj(this, &AssemblyWriter::printGlobal)); Out << "\n; Functions:\n"; // Output all of the functions... for_each(M->begin(), M->end(), bind_obj(this,&AssemblyWriter::printFunction)); } void AssemblyWriter::printGlobal(const GlobalVariable *GV) { if (GV->hasName()) Out << "%" << GV->getName() << " = "; if (GV->hasInternalLinkage()) Out << "internal "; if (!GV->hasInitializer()) Out << "uninitialized "; Out << (GV->isConstant() ? "constant " : "global "); printType(GV->getType()->getElementType()); if (GV->hasInitializer()) writeOperand(GV->getInitializer(), false, false); printInfoComment(GV); Out << "\n"; } // printSymbolTable - Run through symbol table looking for named constants // if a named constant is found, emit it's declaration... // void AssemblyWriter::printSymbolTable(const SymbolTable &ST) { for (SymbolTable::const_iterator TI = ST.begin(); TI != ST.end(); ++TI) { SymbolTable::type_const_iterator I = ST.type_begin(TI->first); SymbolTable::type_const_iterator End = ST.type_end(TI->first); for (; I != End; ++I) { const Value *V = I->second; if (const Constant *CPV = dyn_cast(V)) { printConstant(CPV); } else if (const Type *Ty = dyn_cast(V)) { Out << "\t%" << I->first << " = type "; // Make sure we print out at least one level of the type structure, so // that we do not get %FILE = type %FILE // printTypeAtLeastOneLevel(Ty) << "\n"; } } } } // printConstant - Print out a constant pool entry... // void AssemblyWriter::printConstant(const Constant *CPV) { // Don't print out unnamed constants, they will be inlined if (!CPV->hasName()) return; // Print out name... Out << "\t%" << CPV->getName() << " ="; // Write the value out now... writeOperand(CPV, true, false); printInfoComment(CPV); Out << "\n"; } // printFunction - Print all aspects of a function. // void AssemblyWriter::printFunction(const Function *M) { // Print out the return type and name... Out << "\n" << (M->isExternal() ? "declare " : "") << (M->hasInternalLinkage() ? "internal " : ""); printType(M->getReturnType()) << " \"" << M->getName() << "\"("; Table.incorporateFunction(M); // Loop over the arguments, printing them... const FunctionType *MT = M->getFunctionType(); if (!M->isExternal()) { for_each(M->getArgumentList().begin(), M->getArgumentList().end(), bind_obj(this, &AssemblyWriter::printArgument)); } else { // Loop over the arguments, printing them... const FunctionType *MT = M->getFunctionType(); for (FunctionType::ParamTypes::const_iterator I = MT->getParamTypes().begin(), E = MT->getParamTypes().end(); I != E; ++I) { if (I != MT->getParamTypes().begin()) Out << ", "; printType(*I); } } // Finish printing arguments... if (MT->isVarArg()) { if (MT->getParamTypes().size()) Out << ", "; Out << "..."; // Output varargs portion of signature! } Out << ")\n"; if (!M->isExternal()) { // Loop over the symbol table, emitting all named constants... if (M->hasSymbolTable()) printSymbolTable(*M->getSymbolTable()); Out << "begin"; // Output all of its basic blocks... for the function for_each(M->begin(), M->end(), bind_obj(this, &AssemblyWriter::printBasicBlock)); Out << "end\n"; } Table.purgeFunction(); } // printArgument - This member is called for every argument that // is passed into the function. Simply print it out // void AssemblyWriter::printArgument(const Argument *Arg) { // Insert commas as we go... the first arg doesn't get a comma if (Arg != Arg->getParent()->getArgumentList().front()) Out << ", "; // Output type... printType(Arg->getType()); // Output name, if available... if (Arg->hasName()) Out << " %" << Arg->getName(); else if (Table.getValSlot(Arg) < 0) Out << ""; } // printBasicBlock - This member is called for each basic block in a methd. // void AssemblyWriter::printBasicBlock(const BasicBlock *BB) { if (BB->hasName()) { // Print out the label if it exists... Out << "\n" << BB->getName() << ":"; } else { int Slot = Table.getValSlot(BB); Out << "\n;