+++ /dev/null
-//===-- BytecodeHandler.cpp - Parsing Handler -------------------*- C++ -*-===//
-//
-// The LLVM Compiler Infrastructure
-//
-// This file was developed by Reid Spencer and is distributed under the
-// University of Illinois Open Source License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This header file defines the BytecodeHandler class that gets called by the
-// AbstractBytecodeParser when parsing events occur.
-//
-//===----------------------------------------------------------------------===//
-
-#include "AnalyzerInternals.h"
-#include <iostream>
-
-using namespace llvm;
-
-
-namespace {
-
-class AnalyzerHandler : public BytecodeHandler {
- BytecodeAnalysis& bca;
- BytecodeAnalysis::BytecodeFunctionInfo* currFunc;
-public:
- AnalyzerHandler(BytecodeAnalysis& TheBca)
- : bca(TheBca)
- , currFunc(0)
- { }
-
- virtual bool handleError(const std::string& str ) {
- return false;
- }
-
- virtual void handleStart() {
- bca.ModuleId.clear();
- bca.numBlocks = 0;
- bca.numTypes = 0;
- bca.numValues = 0;
- bca.numFunctions = 0;
- bca.numConstants = 0;
- bca.numGlobalVars = 0;
- bca.numInstructions = 0;
- bca.numBasicBlocks = 0;
- bca.numOperands = 0;
- bca.numCmpctnTables = 0;
- bca.numSymTab = 0;
- bca.maxTypeSlot = 0;
- bca.maxValueSlot = 0;
- bca.numAlignment = 0;
- bca.fileDensity = 0.0;
- bca.globalsDensity = 0.0;
- bca.functionDensity = 0.0;
- bca.instructionSize = 0;
- bca.longInstructions = 0;
- bca.vbrCount32 = 0;
- bca.vbrCount64 = 0;
- bca.vbrCompBytes = 0;
- bca.vbrExpdBytes = 0;
- bca.FunctionInfo.clear();
- bca.BytecodeDump.clear();
- bca.BlockSizes[BytecodeFormat::Module] = 0;
- bca.BlockSizes[BytecodeFormat::Function] = 0;
- bca.BlockSizes[BytecodeFormat::ConstantPool] = 0;
- bca.BlockSizes[BytecodeFormat::SymbolTable] = 0;
- bca.BlockSizes[BytecodeFormat::ModuleGlobalInfo] = 0;
- bca.BlockSizes[BytecodeFormat::GlobalTypePlane] = 0;
- bca.BlockSizes[BytecodeFormat::BasicBlock] = 0;
- bca.BlockSizes[BytecodeFormat::InstructionList] = 0;
- bca.BlockSizes[BytecodeFormat::CompactionTable] = 0;
- }
-
- virtual void handleFinish() {
- bca.fileDensity = double(bca.byteSize) / double( bca.numTypes + bca.numValues );
- double globalSize = 0.0;
- globalSize += double(bca.BlockSizes[BytecodeFormat::ConstantPool]);
- globalSize += double(bca.BlockSizes[BytecodeFormat::ModuleGlobalInfo]);
- globalSize += double(bca.BlockSizes[BytecodeFormat::GlobalTypePlane]);
- bca.globalsDensity = globalSize / double( bca.numTypes + bca.numConstants +
- bca.numGlobalVars );
- bca.functionDensity = double(bca.BlockSizes[BytecodeFormat::Function]) /
- double(bca.numFunctions);
- }
-
- virtual void handleModuleBegin(const std::string& id) {
- bca.ModuleId = id;
- }
-
- virtual void handleModuleEnd(const std::string& id) { }
-
- virtual void handleVersionInfo(
- unsigned char RevisionNum, ///< Byte code revision number
- Module::Endianness Endianness, ///< Endianness indicator
- Module::PointerSize PointerSize ///< PointerSize indicator
- ) { }
-
- virtual void handleModuleGlobalsBegin(unsigned size) { }
-
- virtual void handleGlobalVariable(
- const Type* ElemType, ///< The type of the global variable
- bool isConstant, ///< Whether the GV is constant or not
- GlobalValue::LinkageTypes ///< The linkage type of the GV
- ) {
- bca.numGlobalVars++;
- bca.numValues++;
- }
-
- virtual void handleInitializedGV(
- const Type* ElemType, ///< The type of the global variable
- bool isConstant, ///< Whether the GV is constant or not
- GlobalValue::LinkageTypes,///< The linkage type of the GV
- unsigned initSlot ///< Slot number of GV's initializer
- ) {
- bca.numGlobalVars++;
- bca.numValues++;
- }
-
- virtual void handleType( const Type* Ty ) { bca.numTypes++; }
-
- virtual void handleFunctionDeclaration(
- Function* Func, ///< The function
- const FunctionType* FuncType ///< The type of the function
- ) {
- bca.numFunctions++;
- bca.numValues++;
- }
-
- virtual void handleModuleGlobalsEnd() { }
-
- virtual void handleCompactionTableBegin() { }
-
- virtual void handleCompactionTablePlane( unsigned Ty, unsigned NumEntries) {
- bca.numCmpctnTables++;
- }
-
- virtual void handleCompactionTableType( unsigned i, unsigned TypSlot,
- const Type* ) {}
-
- virtual void handleCompactionTableValue(
- unsigned i,
- unsigned ValSlot,
- const Type* ) { }
-
- virtual void handleCompactionTableEnd() { }
-
- virtual void handleSymbolTableBegin() { bca.numSymTab++; }
-
- virtual void handleSymbolTablePlane( unsigned Ty, unsigned NumEntries,
- const Type* Typ) { }
-
- virtual void handleSymbolTableType( unsigned i, unsigned slot,
- const std::string& name ) { }
-
- virtual void handleSymbolTableValue( unsigned i, unsigned slot,
- const std::string& name ) { }
-
- virtual void handleSymbolTableEnd() { }
-
- virtual void handleFunctionBegin( Function* Func, unsigned Size) {
- const FunctionType* FType =
- cast<FunctionType>(Func->getType()->getElementType());
- currFunc = &bca.FunctionInfo[Func];
- currFunc->description = FType->getDescription();
- currFunc->name = Func->getName();
- currFunc->byteSize = Size;
- currFunc->numInstructions = 0;
- currFunc->numBasicBlocks = 0;
- currFunc->numPhis = 0;
- currFunc->numOperands = 0;
- currFunc->density = 0.0;
- currFunc->instructionSize = 0;
- currFunc->longInstructions = 0;
- currFunc->vbrCount32 = 0;
- currFunc->vbrCount64 = 0;
- currFunc->vbrCompBytes = 0;
- currFunc->vbrExpdBytes = 0;
- }
-
- virtual void handleFunctionEnd( Function* Func) {
- currFunc->density = double(currFunc->byteSize) /
- double(currFunc->numInstructions+currFunc->numBasicBlocks);
- }
-
- virtual void handleBasicBlockBegin( unsigned blocknum) {
- bca.numBasicBlocks++;
- bca.numValues++;
- if ( currFunc ) currFunc->numBasicBlocks++;
- }
-
- virtual bool handleInstruction( unsigned Opcode, const Type* iType,
- std::vector<unsigned>& Operands, unsigned Size) {
- bca.numInstructions++;
- bca.numValues++;
- bca.instructionSize += Size;
- if (Size > 4 ) bca.longInstructions++;
- bca.numOperands += Operands.size();
- if ( currFunc ) {
- currFunc->numInstructions++;
- currFunc->instructionSize += Size;
- if (Size > 4 ) currFunc->longInstructions++;
- if ( Opcode == Instruction::PHI ) currFunc->numPhis++;
- }
- return Instruction::isTerminator(Opcode);
- }
-
- virtual void handleBasicBlockEnd(unsigned blocknum) { }
-
- virtual void handleGlobalConstantsBegin() { }
-
- virtual void handleConstantExpression( unsigned Opcode, const Type* Typ,
- std::vector<std::pair<const Type*,unsigned> > ArgVec ) {
- bca.numConstants++;
- bca.numValues++;
- }
-
- virtual void handleConstantValue( Constant * c ) {
- bca.numConstants++;
- bca.numValues++;
- }
-
- virtual void handleConstantArray( const ArrayType* AT,
- std::vector<unsigned>& Elements ) {
- bca.numConstants++;
- bca.numValues++;
- }
-
- virtual void handleConstantStruct(
- const StructType* ST,
- std::vector<unsigned>& ElementSlots)
- {
- bca.numConstants++;
- bca.numValues++;
- }
-
- virtual void handleConstantPointer( const PointerType* PT, unsigned Slot) {
- bca.numConstants++;
- bca.numValues++;
- }
-
- virtual void handleConstantString( const ConstantArray* CA ) {
- bca.numConstants++;
- bca.numValues++;
- }
-
- virtual void handleGlobalConstantsEnd() { }
-
- virtual void handleAlignment(unsigned numBytes) {
- bca.numAlignment += numBytes;
- }
-
- virtual void handleBlock(
- unsigned BType, const unsigned char* StartPtr, unsigned Size) {
- bca.numBlocks++;
- bca.BlockSizes[llvm::BytecodeFormat::FileBlockIDs(BType)] += Size;
- }
-
- virtual void handleVBR32(unsigned Size ) {
- bca.vbrCount32++;
- bca.vbrCompBytes += Size;
- bca.vbrExpdBytes += sizeof(uint32_t);
- if (currFunc) {
- currFunc->vbrCount32++;
- currFunc->vbrCompBytes += Size;
- currFunc->vbrExpdBytes += sizeof(uint32_t);
- }
- }
-
- virtual void handleVBR64(unsigned Size ) {
- bca.vbrCount64++;
- bca.vbrCompBytes += Size;
- bca.vbrExpdBytes += sizeof(uint64_t);
- if ( currFunc ) {
- currFunc->vbrCount64++;
- currFunc->vbrCompBytes += Size;
- currFunc->vbrExpdBytes += sizeof(uint64_t);
- }
- }
-};
-
-}
-
-void llvm::BytecodeAnalyzer::AnalyzeBytecode(
- const unsigned char *Buf,
- unsigned Length,
- BytecodeAnalysis& bca,
- const std::string &ModuleID
-)
-{
- bca.byteSize = Length;
- AnalyzerHandler TheHandler(bca);
- AbstractBytecodeParser TheParser(&TheHandler, true, true, true);
- TheParser.ParseBytecode( Buf, Length, ModuleID );
- TheParser.ParseAllFunctionBodies();
-}
-
-// vim: sw=2
+++ /dev/null
-//===-- ReaderInternals.h - Definitions internal to the reader --*- C++ -*-===//
-//
-// 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 header file defines various stuff that is used by the bytecode reader.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef ANALYZER_INTERNALS_H
-#define ANALYZER_INTERNALS_H
-
-#include "Parser.h"
-#include "llvm/Bytecode/Analyzer.h"
-#include "llvm/Constants.h"
-#include "llvm/DerivedTypes.h"
-
-
-namespace llvm {
-
-class BytecodeAnalyzer {
- BytecodeAnalyzer(const BytecodeAnalyzer &); // DO NOT IMPLEMENT
- void operator=(const BytecodeAnalyzer &); // DO NOT IMPLEMENT
-public:
- BytecodeAnalyzer() { }
- ~BytecodeAnalyzer() { }
-
- void AnalyzeBytecode(
- const unsigned char *Buf,
- unsigned Length,
- BytecodeAnalysis& bca,
- const std::string &ModuleID
- );
-
- void DumpBytecode(
- const unsigned char *Buf,
- unsigned Length,
- BytecodeAnalysis& bca,
- const std::string &ModuleID
- );
-};
-
-} // End llvm namespace
-
-#endif
-
-// vim: sw=2
+++ /dev/null
-//===- AnalyzerWrappers.cpp - Analyze bytecode from file or buffer -------===//
-//
-// The LLVM Compiler Infrastructure
-//
-// This file was developed by Reid Spencer and is distributed under the
-// University of Illinois Open Source License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This file implements loading and analysis of a bytecode file and analyzing a
-// bytecode buffer.
-//
-//===----------------------------------------------------------------------===//
-
-#include "llvm/Bytecode/Analyzer.h"
-#include "AnalyzerInternals.h"
-#include "Support/FileUtilities.h"
-#include "Support/StringExtras.h"
-#include "Config/unistd.h"
-#include <cerrno>
-#include <iomanip>
-
-using namespace llvm;
-
-//===----------------------------------------------------------------------===//
-// BytecodeFileAnalyzer - Analyze from an mmap'able file descriptor.
-//
-
-namespace {
- /// BytecodeFileAnalyzer - parses a bytecode file from a file
- class BytecodeFileAnalyzer : public BytecodeAnalyzer {
- private:
- unsigned char *Buffer;
- unsigned Length;
-
- BytecodeFileAnalyzer(const BytecodeFileAnalyzer&); // Do not implement
- void operator=(const BytecodeFileAnalyzer &BFR); // Do not implement
-
- public:
- BytecodeFileAnalyzer(const std::string &Filename, BytecodeAnalysis& bca);
- ~BytecodeFileAnalyzer();
- };
-}
-
-static std::string ErrnoMessage (int savedErrNum, std::string descr) {
- return ::strerror(savedErrNum) + std::string(", while trying to ") + descr;
-}
-
-BytecodeFileAnalyzer::BytecodeFileAnalyzer(const std::string &Filename,
- BytecodeAnalysis& bca) {
- Buffer = (unsigned char*)ReadFileIntoAddressSpace(Filename, Length);
- if (Buffer == 0)
- throw "Error reading file '" + Filename + "'.";
-
- try {
- // Parse the bytecode we mmapped in
- if ( bca.dumpBytecode )
- DumpBytecode(Buffer, Length, bca, Filename);
- AnalyzeBytecode(Buffer, Length, bca, Filename);
- } catch (...) {
- UnmapFileFromAddressSpace(Buffer, Length);
- throw;
- }
-}
-
-BytecodeFileAnalyzer::~BytecodeFileAnalyzer() {
- // Unmmap the bytecode...
- UnmapFileFromAddressSpace(Buffer, Length);
-}
-
-//===----------------------------------------------------------------------===//
-// BytecodeBufferAnalyzer - Read from a memory buffer
-//
-
-namespace {
- /// BytecodeBufferAnalyzer - parses a bytecode file from a buffer
- ///
- class BytecodeBufferAnalyzer : public BytecodeAnalyzer {
- private:
- const unsigned char *Buffer;
- bool MustDelete;
-
- BytecodeBufferAnalyzer(const BytecodeBufferAnalyzer&); // Do not implement
- void operator=(const BytecodeBufferAnalyzer &BFR); // Do not implement
-
- public:
- BytecodeBufferAnalyzer(const unsigned char *Buf, unsigned Length,
- BytecodeAnalysis& bca, const std::string &ModuleID);
- ~BytecodeBufferAnalyzer();
-
- };
-}
-
-BytecodeBufferAnalyzer::BytecodeBufferAnalyzer(const unsigned char *Buf,
- unsigned Length,
- BytecodeAnalysis& bca,
- const std::string &ModuleID) {
- // If not aligned, allocate a new buffer to hold the bytecode...
- const unsigned char *ParseBegin = 0;
- if ((intptr_t)Buf & 3) {
- Buffer = new unsigned char[Length+4];
- unsigned Offset = 4 - ((intptr_t)Buffer & 3); // Make sure it's aligned
- ParseBegin = Buffer + Offset;
- memcpy((unsigned char*)ParseBegin, Buf, Length); // Copy it over
- MustDelete = true;
- } else {
- // If we don't need to copy it over, just use the caller's copy
- ParseBegin = Buffer = Buf;
- MustDelete = false;
- }
- try {
- if ( bca.dumpBytecode )
- DumpBytecode(ParseBegin, Length, bca, ModuleID);
- AnalyzeBytecode(ParseBegin, Length, bca, ModuleID);
- } catch (...) {
- if (MustDelete) delete [] Buffer;
- throw;
- }
-}
-
-BytecodeBufferAnalyzer::~BytecodeBufferAnalyzer() {
- if (MustDelete) delete [] Buffer;
-}
-
-//===----------------------------------------------------------------------===//
-// BytecodeStdinAnalyzer - Read bytecode from Standard Input
-//
-
-namespace {
- /// BytecodeStdinAnalyzer - parses a bytecode file from stdin
- ///
- class BytecodeStdinAnalyzer : public BytecodeAnalyzer {
- private:
- std::vector<unsigned char> FileData;
- unsigned char *FileBuf;
-
- BytecodeStdinAnalyzer(const BytecodeStdinAnalyzer&); // Do not implement
- void operator=(const BytecodeStdinAnalyzer &BFR); // Do not implement
-
- public:
- BytecodeStdinAnalyzer(BytecodeAnalysis& bca);
- };
-}
-
-BytecodeStdinAnalyzer::BytecodeStdinAnalyzer(BytecodeAnalysis& bca ) {
- int BlockSize;
- unsigned char Buffer[4096*4];
-
- // Read in all of the data from stdin, we cannot mmap stdin...
- while ((BlockSize = ::read(0 /*stdin*/, Buffer, 4096*4))) {
- if (BlockSize == -1)
- throw ErrnoMessage(errno, "read from standard input");
-
- FileData.insert(FileData.end(), Buffer, Buffer+BlockSize);
- }
-
- if (FileData.empty())
- throw std::string("Standard Input empty!");
-
- FileBuf = &FileData[0];
- if (bca.dumpBytecode)
- DumpBytecode(&FileData[0], FileData.size(), bca, "<stdin>");
- AnalyzeBytecode(FileBuf, FileData.size(), bca, "<stdin>");
-}
-
-//===----------------------------------------------------------------------===//
-// Wrapper functions
-//===----------------------------------------------------------------------===//
-
-// AnalyzeBytecodeFile - analyze one file
-void llvm::AnalyzeBytecodeFile(const std::string &Filename,
- BytecodeAnalysis& bca,
- std::string *ErrorStr)
-{
- try {
- if ( Filename != "-" )
- BytecodeFileAnalyzer bfa(Filename,bca);
- else
- BytecodeStdinAnalyzer bsa(bca);
- } catch (std::string &err) {
- if (ErrorStr) *ErrorStr = err;
- }
-}
-
-// AnalyzeBytecodeBuffer - analyze a buffer
-void llvm::AnalyzeBytecodeBuffer(
- const unsigned char* Buffer, ///< Pointer to start of bytecode buffer
- unsigned BufferSize, ///< Size of the bytecode buffer
- BytecodeAnalysis& Results, ///< The results of the analysis
- std::string* ErrorStr ///< Errors, if any.
- )
-{
- try {
- BytecodeBufferAnalyzer(Buffer, BufferSize, Results, "<buffer>" );
- } catch (std::string& err ) {
- if ( ErrorStr) *ErrorStr = err;
- }
-}
-
-
-/// This function prints the contents of rhe BytecodeAnalysis structure in
-/// a human legible form.
-/// @brief Print BytecodeAnalysis structure to an ostream
-namespace {
-inline static void print(std::ostream& Out, const char*title,
- unsigned val, bool nl = true ) {
- Out << std::setw(30) << std::right << title
- << std::setw(0) << ": "
- << std::setw(9) << val << "\n";
-}
-
-inline static void print(std::ostream&Out, const char*title,
- double val ) {
- Out << std::setw(30) << std::right << title
- << std::setw(0) << ": "
- << std::setw(9) << std::setprecision(6) << val << "\n" ;
-}
-
-inline static void print(std::ostream&Out, const char*title,
- double top, double bot ) {
- Out << std::setw(30) << std::right << title
- << std::setw(0) << ": "
- << std::setw(9) << std::setprecision(6) << top
- << " (" << std::left << std::setw(0) << std::setprecision(4)
- << (top/bot)*100.0 << "%)\n";
-}
-inline static void print(std::ostream&Out, const char*title,
- std::string val, bool nl = true) {
- Out << std::setw(30) << std::right << title
- << std::setw(0) << ": "
- << std::left << val << (nl ? "\n" : "");
-}
-
-}
-
-void llvm::PrintBytecodeAnalysis(BytecodeAnalysis& bca, std::ostream& Out )
-{
- print(Out, "Bytecode Analysis Of Module", bca.ModuleId);
- print(Out, "File Size", bca.byteSize);
- print(Out, "Bytecode Compression Index",std::string("TBD"));
- print(Out, "Number Of Bytecode Blocks", bca.numBlocks);
- print(Out, "Number Of Types", bca.numTypes);
- print(Out, "Number Of Values", bca.numValues);
- print(Out, "Number Of Constants", bca.numConstants);
- print(Out, "Number Of Global Variables", bca.numGlobalVars);
- print(Out, "Number Of Functions", bca.numFunctions);
- print(Out, "Number Of Basic Blocks", bca.numBasicBlocks);
- print(Out, "Number Of Instructions", bca.numInstructions);
- print(Out, "Number Of Operands", bca.numOperands);
- print(Out, "Number Of Compaction Tables", bca.numCmpctnTables);
- print(Out, "Number Of Symbol Tables", bca.numSymTab);
- print(Out, "Long Instructions", bca.longInstructions);
- print(Out, "Instruction Size", bca.instructionSize);
- print(Out, "Average Instruction Size",
- double(bca.instructionSize)/double(bca.numInstructions));
- print(Out, "Maximum Type Slot Number", bca.maxTypeSlot);
- print(Out, "Maximum Value Slot Number", bca.maxValueSlot);
- print(Out, "Bytes Thrown To Alignment", double(bca.numAlignment),
- double(bca.byteSize));
- print(Out, "File Density (bytes/def)", bca.fileDensity);
- print(Out, "Globals Density (bytes/def)", bca.globalsDensity);
- print(Out, "Function Density (bytes/func)", bca.functionDensity);
- print(Out, "Number of VBR 32-bit Integers", bca.vbrCount32);
- print(Out, "Number of VBR 64-bit Integers", bca.vbrCount64);
- print(Out, "Number of VBR Compressed Bytes", bca.vbrCompBytes);
- print(Out, "Number of VBR Expanded Bytes", bca.vbrExpdBytes);
- print(Out, "VBR Savings",
- double(bca.vbrExpdBytes)-double(bca.vbrCompBytes),
- double(bca.byteSize));
-
- if ( bca.detailedResults ) {
- print(Out, "Module Bytes",
- double(bca.BlockSizes[BytecodeFormat::Module]),
- double(bca.byteSize));
- print(Out, "Function Bytes",
- double(bca.BlockSizes[BytecodeFormat::Function]),
- double(bca.byteSize));
- print(Out, "Constant Pool Bytes",
- double(bca.BlockSizes[BytecodeFormat::ConstantPool]),
- double(bca.byteSize));
- print(Out, "Symbol Table Bytes",
- double(bca.BlockSizes[BytecodeFormat::SymbolTable]),
- double(bca.byteSize));
- print(Out, "Module Global Info Bytes",
- double(bca.BlockSizes[BytecodeFormat::ModuleGlobalInfo]),
- double(bca.byteSize));
- print(Out, "Global Type Plane Bytes",
- double(bca.BlockSizes[BytecodeFormat::GlobalTypePlane]),
- double(bca.byteSize));
- print(Out, "Basic Block Bytes",
- double(bca.BlockSizes[BytecodeFormat::BasicBlock]),
- double(bca.byteSize));
- print(Out, "Instruction List Bytes",
- double(bca.BlockSizes[BytecodeFormat::InstructionList]),
- double(bca.byteSize));
- print(Out, "Compaction Table Bytes",
- double(bca.BlockSizes[BytecodeFormat::CompactionTable]),
- double(bca.byteSize));
-
- std::map<const Function*,BytecodeAnalysis::BytecodeFunctionInfo>::iterator I =
- bca.FunctionInfo.begin();
- std::map<const Function*,BytecodeAnalysis::BytecodeFunctionInfo>::iterator E =
- bca.FunctionInfo.end();
-
- while ( I != E ) {
- Out << std::left << std::setw(0);
- Out << "Function: " << I->second.name << "\n";
- print(Out, "Type:", I->second.description);
- print(Out, "Byte Size", I->second.byteSize);
- print(Out, "Instructions", I->second.numInstructions);
- print(Out, "Long Instructions", I->second.longInstructions);
- print(Out, "Instruction Size", I->second.instructionSize);
- print(Out, "Average Instruction Size",
- double(I->second.instructionSize)/double(I->second.numInstructions));
- print(Out, "Basic Blocks", I->second.numBasicBlocks);
- print(Out, "Operand", I->second.numOperands);
- print(Out, "Function Density", I->second.density);
- print(Out, "Number of VBR 32-bit Integers", I->second.vbrCount32);
- print(Out, "Number of VBR 64-bit Integers", I->second.vbrCount64);
- print(Out, "Number of VBR Compressed Bytes", I->second.vbrCompBytes);
- print(Out, "Number of VBR Expanded Bytes", I->second.vbrExpdBytes);
- print(Out, "VBR Savings",
- double(I->second.vbrExpdBytes)-double(I->second.vbrCompBytes),
- double(I->second.byteSize));
- ++I;
- }
- }
-
- if ( bca.dumpBytecode )
- Out << bca.BytecodeDump;
-}
-// vim: sw=2
+++ /dev/null
-//===-- BytecodeDumper.cpp - Parsing Handler --------------------*- C++ -*-===//
-//
-// The LLVM Compiler Infrastructure
-//
-// This file was developed by Reid Spencer and is distributed under the
-// University of Illinois Open Source License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This header file defines the BytecodeDumper class that gets called by the
-// AbstractBytecodeParser when parsing events occur. It merely dumps the
-// information presented to it from the parser.
-//
-//===----------------------------------------------------------------------===//
-
-#include "AnalyzerInternals.h"
-#include "llvm/Constant.h"
-#include "llvm/Constants.h"
-#include "llvm/DerivedTypes.h"
-#include "llvm/Instruction.h"
-#include "llvm/Type.h"
-
-using namespace llvm;
-
-namespace {
-
-class BytecodeDumper : public llvm::BytecodeHandler {
-public:
-
- virtual bool handleError(const std::string& str )
- {
- std::cout << "ERROR: " << str << "\n";
- return true;
- }
-
- virtual void handleStart()
- {
- std::cout << "Bytecode {\n";
- }
-
- virtual void handleFinish()
- {
- std::cout << "} End Bytecode\n";
- }
-
- virtual void handleModuleBegin(const std::string& id)
- {
- std::cout << " Module " << id << " {\n";
- }
-
- virtual void handleModuleEnd(const std::string& id)
- {
- std::cout << " } End Module " << id << "\n";
- }
-
- virtual void handleVersionInfo(
- unsigned char RevisionNum, ///< Byte code revision number
- Module::Endianness Endianness, ///< Endianness indicator
- Module::PointerSize PointerSize ///< PointerSize indicator
- )
- {
- std::cout << " RevisionNum: " << int(RevisionNum)
- << " Endianness: " << Endianness
- << " PointerSize: " << PointerSize << "\n";
- }
-
- virtual void handleModuleGlobalsBegin()
- {
- std::cout << " BLOCK: ModuleGlobalInfo {\n";
- }
-
- virtual void handleGlobalVariable(
- const Type* ElemType, ///< The type of the global variable
- bool isConstant, ///< Whether the GV is constant or not
- GlobalValue::LinkageTypes Linkage ///< The linkage type of the GV
- )
- {
- std::cout << " GV: Uninitialized, "
- << ( isConstant? "Constant, " : "Variable, ")
- << " Linkage=" << Linkage << " Type="
- << ElemType->getDescription() << "\n";
- }
-
- virtual void handleInitializedGV(
- const Type* ElemType, ///< The type of the global variable
- bool isConstant, ///< Whether the GV is constant or not
- GlobalValue::LinkageTypes Linkage,///< The linkage type of the GV
- unsigned initSlot ///< Slot number of GV's initializer
- )
- {
- std::cout << " GV: Initialized, "
- << ( isConstant? "Constant, " : "Variable, ")
- << " Linkage=" << Linkage << " Type="
- << ElemType->getDescription()
- << " InitializerSlot=" << initSlot << "\n";
- }
-
- virtual void handleType( const Type* Ty ) {
- std::cout << " Type: " << Ty->getDescription() << "\n";
- }
-
- virtual void handleFunctionDeclaration(
- Function* Func,
- const FunctionType* FuncType) {
- std::cout << " Function: " << FuncType->getDescription() << "\n";
- }
-
- virtual void handleModuleGlobalsEnd() {
- std::cout << " } END BLOCK: ModuleGlobalInfo\n";
- }
-
- virtual void handleCompactionTableBegin() {
- std::cout << " BLOCK: CompactionTable {\n";
- }
-
- virtual void handleCompactionTablePlane( unsigned Ty, unsigned NumEntries ) {
- std::cout << " Plane: Ty=" << Ty << " Size=" << NumEntries << "\n";
- }
-
- virtual void handleCompactionTableType(
- unsigned i,
- unsigned TypSlot,
- const Type* Ty) {
- std::cout << " Type: " << i << " Slot:" << TypSlot
- << " is " << Ty->getDescription() << "\n";
- }
-
- virtual void handleCompactionTableValue(
- unsigned i,
- unsigned ValSlot,
- const Type* Ty ) {
- std::cout << " Value: " << i << " Slot:" << ValSlot
- << " is " << Ty->getDescription() << "\n";
- }
-
- virtual void handleCompactionTableEnd() {
- std::cout << " } END BLOCK: CompactionTable\n";
- }
-
- virtual void handleSymbolTableBegin() {
- std::cout << " BLOCK: SymbolTable {\n";
- }
-
- virtual void handleSymbolTablePlane(
- unsigned Ty,
- unsigned NumEntries,
- const Type* Typ) {
- std::cout << " Plane: Ty=" << Ty << " Size=" << NumEntries
- << " Type: " << Typ->getDescription() << "\n";
- }
-
- virtual void handleSymbolTableType(
- unsigned i,
- unsigned slot,
- const std::string& name ) {
- std::cout << " Type " << i << " Slot=" << slot
- << " Name: " << name << "\n";
- }
-
- virtual void handleSymbolTableValue(
- unsigned i,
- unsigned slot,
- const std::string& name ) {
- std::cout << " Value " << i << " Slot=" << slot
- << " Name: " << name << "\n";
- }
-
- virtual void handleSymbolTableEnd() {
- std::cout << " } END BLOCK: SymbolTable\n";
- }
-
- virtual void handleFunctionBegin(
- const Type* FType, GlobalValue::LinkageTypes linkage ) {
- std::cout << "BLOCK: Function {\n";
- std::cout << " Linkage: " << linkage << "\n";
- std::cout << " Type: " << FType->getDescription() << "\n";
- }
-
- virtual void handleFunctionEnd( const Type* FType) {
- std::cout << "} END BLOCK: Function\n";
- }
-
- virtual void handleBasicBlockBegin( unsigned blocknum) {
- std::cout << " BLOCK: BasicBlock #" << blocknum << "{\n";
- }
-
- virtual bool handleInstruction(
- unsigned Opcode,
- const Type* iType,
- std::vector<unsigned>& Operands,
- unsigned Size) {
- std::cout << " INST: OpCode="
- << Instruction::getOpcodeName(Opcode) << " Type="
- << iType->getDescription() << "\n";
- for ( unsigned i = 0; i < Operands.size(); ++i )
- std::cout << " Op#" << i << " Slot=" << Operands[i] << "\n";
-
- return Instruction::isTerminator(Opcode);
- }
-
- virtual void handleBasicBlockEnd(unsigned blocknum) {
- std::cout << " } END BLOCK: BasicBlock #" << blocknum << "{\n";
- }
-
- virtual void handleGlobalConstantsBegin() {
- std::cout << " BLOCK: GlobalConstants {\n";
- }
-
- virtual void handleConstantExpression(
- unsigned Opcode,
- const Type* Typ,
- std::vector<std::pair<const Type*,unsigned> > ArgVec
- ) {
- std::cout << " EXPR: " << Instruction::getOpcodeName(Opcode)
- << " Type=" << Typ->getDescription() << "\n";
- for ( unsigned i = 0; i < ArgVec.size(); ++i )
- std::cout << " Arg#" << i << " Type="
- << ArgVec[i].first->getDescription() << " Slot="
- << ArgVec[i].second << "\n";
- }
-
- virtual void handleConstantValue( Constant * c ) {
- std::cout << " VALUE: ";
- c->print(std::cout);
- std::cout << "\n";
- }
-
- virtual void handleConstantArray(const ArrayType* AT,
- std::vector<unsigned>& Elements ) {
- std::cout << " ARRAY: " << AT->getDescription() << "\n";
- for ( unsigned i = 0; i < Elements.size(); ++i )
- std::cout << " #" << i << " Slot=" << Elements[i] << "\n";
- }
-
- virtual void handleConstantStruct( const StructType* ST,
- std::vector<unsigned>& Elements) {
- std::cout << " STRUC: " << ST->getDescription() << "\n";
- for ( unsigned i = 0; i < Elements.size(); ++i )
- std::cout << " #" << i << " Slot=" << Elements[i] << "\n";
- }
-
- virtual void handleConstantPointer(
- const PointerType* PT, unsigned Slot)
- {
- std::cout << " POINT: " << PT->getDescription()
- << " Slot=" << Slot << "\n";
- }
-
- virtual void handleConstantString( const ConstantArray* CA ) {
- std::cout << " STRNG: ";
- CA->print(std::cout);
- std::cout << "\n";
- }
-
- virtual void handleGlobalConstantsEnd() {
- std::cout << " } END BLOCK: GlobalConstants\n";
- }
-};
-
-}
-
-void BytecodeAnalyzer::DumpBytecode(
- const unsigned char *Buf,
- unsigned Length,
- BytecodeAnalysis& bca,
- const std::string &ModuleID) {
- BytecodeDumper TheHandler;
- AbstractBytecodeParser TheParser(&TheHandler);
- TheParser.ParseBytecode( Buf, Length, ModuleID );
- if ( bca.detailedResults )
- TheParser.ParseAllFunctionBodies();
-}
-
-// vim: sw=2
+++ /dev/null
-##===- lib/Bytecode/Reader/Makefile ------------------------*- Makefile -*-===##
-#
-# 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.
-#
-##===----------------------------------------------------------------------===##
-LEVEL = ../../..
-LIBRARYNAME = bcanalyzer
-
-include $(LEVEL)/Makefile.common
-
+++ /dev/null
-//===- Parser.cpp - Code to parse bytecode files --------------------------===//
-//
-// 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 library implements the functionality defined in llvm/Bytecode/Parser.h
-//
-// Note that this library should be as fast as possible, reentrant, and
-// threadsafe!!
-//
-// TODO: Allow passing in an option to ignore the symbol table
-//
-//===----------------------------------------------------------------------===//
-
-#include "AnalyzerInternals.h"
-#include "llvm/Module.h"
-#include "llvm/Bytecode/Format.h"
-#include "Support/StringExtras.h"
-#include <iostream>
-#include <sstream>
-
-using namespace llvm;
-
-// Enable to trace to figure out what the heck is going on when parsing fails
-//#define TRACE_LEVEL 10
-//#define DEBUG_OUTPUT
-
-#if TRACE_LEVEL // ByteCodeReading_TRACEr
-#define BCR_TRACE(n, X) \
- if (n < TRACE_LEVEL) std::cerr << std::string(n*2, ' ') << X
-#else
-#define BCR_TRACE(n, X)
-#endif
-
-#define PARSE_ERROR(inserters) { \
- std::ostringstream errormsg; \
- errormsg << inserters; \
- if ( ! handler->handleError( errormsg.str() ) ) \
- throw std::string(errormsg.str()); \
- }
-
-inline bool AbstractBytecodeParser::moreInBlock() {
- return At < BlockEnd;
-}
-
-inline void AbstractBytecodeParser::checkPastBlockEnd(const char * block_name) {
- if ( At > BlockEnd )
- PARSE_ERROR("Attempt to read past the end of " << block_name << " block.");
-}
-
-inline void AbstractBytecodeParser::align32() {
- BufPtr Save = At;
- At = (const unsigned char *)((unsigned long)(At+3) & (~3UL));
- if ( reportAlignment && At > Save ) handler->handleAlignment( At - Save );
- if (At > BlockEnd)
- throw std::string("Ran out of data while aligning!");
-}
-
-inline unsigned AbstractBytecodeParser::read_uint() {
- if (At+4 > BlockEnd)
- throw std::string("Ran out of data reading uint!");
- At += 4;
- return At[-4] | (At[-3] << 8) | (At[-2] << 16) | (At[-1] << 24);
-}
-
-inline unsigned AbstractBytecodeParser::read_vbr_uint() {
- unsigned Shift = 0;
- unsigned Result = 0;
- BufPtr Save = At;
-
- do {
- if (At == BlockEnd)
- throw std::string("Ran out of data reading vbr_uint!");
- Result |= (unsigned)((*At++) & 0x7F) << Shift;
- Shift += 7;
- } while (At[-1] & 0x80);
- if (reportVBR)
- handler->handleVBR32(At-Save);
- return Result;
-}
-
-inline uint64_t AbstractBytecodeParser::read_vbr_uint64() {
- unsigned Shift = 0;
- uint64_t Result = 0;
- BufPtr Save = At;
-
- do {
- if (At == BlockEnd)
- throw std::string("Ran out of data reading vbr_uint64!");
- Result |= (uint64_t)((*At++) & 0x7F) << Shift;
- Shift += 7;
- } while (At[-1] & 0x80);
- if (reportVBR)
- handler->handleVBR64(At-Save);
- return Result;
-}
-
-inline int64_t AbstractBytecodeParser::read_vbr_int64() {
- uint64_t R = read_vbr_uint64();
- if (R & 1) {
- if (R != 1)
- return -(int64_t)(R >> 1);
- else // There is no such thing as -0 with integers. "-0" really means
- // 0x8000000000000000.
- return 1LL << 63;
- } else
- return (int64_t)(R >> 1);
-}
-
-inline std::string AbstractBytecodeParser::read_str() {
- unsigned Size = read_vbr_uint();
- const unsigned char *OldAt = At;
- At += Size;
- if (At > BlockEnd) // Size invalid?
- throw std::string("Ran out of data reading a string!");
- return std::string((char*)OldAt, Size);
-}
-
-inline void AbstractBytecodeParser::read_data(void *Ptr, void *End) {
- unsigned char *Start = (unsigned char *)Ptr;
- unsigned Amount = (unsigned char *)End - Start;
- if (At+Amount > BlockEnd)
- throw std::string("Ran out of data!");
- std::copy(At, At+Amount, Start);
- At += Amount;
-}
-
-inline void AbstractBytecodeParser::readBlock(unsigned &Type, unsigned &Size) {
- Type = read_uint();
- Size = read_uint();
- BlockStart = At;
- if ( At + Size > BlockEnd )
- throw std::string("Attempt to size a block past end of memory");
- BlockEnd = At + Size;
- if ( reportBlocks ) {
- handler->handleBlock( Type, BlockStart, Size );
- }
-}
-
-const Type *AbstractBytecodeParser::getType(unsigned ID) {
-//cerr << "Looking up Type ID: " << ID << "\n";
-
-if (ID < Type::FirstDerivedTyID)
- if (const Type *T = Type::getPrimitiveType((Type::TypeID)ID))
- return T; // Asked for a primitive type...
-
-// Otherwise, derived types need offset...
-ID -= Type::FirstDerivedTyID;
-
-if (!CompactionTypeTable.empty()) {
- if (ID >= CompactionTypeTable.size())
- PARSE_ERROR("Type ID out of range for compaction table!");
- return CompactionTypeTable[ID];
-}
-
-// Is it a module-level type?
- if (ID < ModuleTypes.size())
- return ModuleTypes[ID].get();
-
- // Nope, is it a function-level type?
- ID -= ModuleTypes.size();
- if (ID < FunctionTypes.size())
- return FunctionTypes[ID].get();
-
- PARSE_ERROR("Illegal type reference!");
- return Type::VoidTy;
-}
-
-bool AbstractBytecodeParser::ParseInstruction(std::vector<unsigned> &Operands) {
- BufPtr SaveAt = At;
- Operands.clear();
- unsigned iType = 0;
- unsigned Opcode = 0;
- unsigned Op = read_uint();
-
- // bits Instruction format: Common to all formats
- // --------------------------
- // 01-00: Opcode type, fixed to 1.
- // 07-02: Opcode
- Opcode = (Op >> 2) & 63;
- Operands.resize((Op >> 0) & 03);
-
- switch (Operands.size()) {
- case 1:
- // bits Instruction format:
- // --------------------------
- // 19-08: Resulting type plane
- // 31-20: Operand #1 (if set to (2^12-1), then zero operands)
- //
- iType = (Op >> 8) & 4095;
- Operands[0] = (Op >> 20) & 4095;
- if (Operands[0] == 4095) // Handle special encoding for 0 operands...
- Operands.resize(0);
- break;
- case 2:
- // bits Instruction format:
- // --------------------------
- // 15-08: Resulting type plane
- // 23-16: Operand #1
- // 31-24: Operand #2
- //
- iType = (Op >> 8) & 255;
- Operands[0] = (Op >> 16) & 255;
- Operands[1] = (Op >> 24) & 255;
- break;
- case 3:
- // bits Instruction format:
- // --------------------------
- // 13-08: Resulting type plane
- // 19-14: Operand #1
- // 25-20: Operand #2
- // 31-26: Operand #3
- //
- iType = (Op >> 8) & 63;
- Operands[0] = (Op >> 14) & 63;
- Operands[1] = (Op >> 20) & 63;
- Operands[2] = (Op >> 26) & 63;
- break;
- case 0:
- At -= 4; // Hrm, try this again...
- Opcode = read_vbr_uint();
- Opcode >>= 2;
- iType = read_vbr_uint();
-
- unsigned NumOperands = read_vbr_uint();
- Operands.resize(NumOperands);
-
- if (NumOperands == 0)
- PARSE_ERROR("Zero-argument instruction found; this is invalid.");
-
- for (unsigned i = 0; i != NumOperands; ++i)
- Operands[i] = read_vbr_uint();
- align32();
- break;
- }
-
- return handler->handleInstruction(Opcode, getType(iType), Operands, At-SaveAt);
-}
-
-/// ParseBasicBlock - In LLVM 1.0 bytecode files, we used to output one
-/// basicblock at a time. This method reads in one of the basicblock packets.
-void AbstractBytecodeParser::ParseBasicBlock( unsigned BlockNo) {
- handler->handleBasicBlockBegin( BlockNo );
-
- std::vector<unsigned> Args;
- bool is_terminating = false;
- while ( moreInBlock() )
- is_terminating = ParseInstruction(Args);
-
- if ( ! is_terminating )
- PARSE_ERROR("Non-terminated basic block found!");
-
- handler->handleBasicBlockEnd( BlockNo );
-}
-
-/// ParseInstructionList - Parse all of the BasicBlock's & Instruction's in the
-/// body of a function. In post 1.0 bytecode files, we no longer emit basic
-/// block individually, in order to avoid per-basic-block overhead.
-unsigned AbstractBytecodeParser::ParseInstructionList() {
- unsigned BlockNo = 0;
- std::vector<unsigned> Args;
-
- while ( moreInBlock() ) {
- handler->handleBasicBlockBegin( BlockNo );
-
- // Read instructions into this basic block until we get to a terminator
- bool is_terminating = false;
- while (moreInBlock() && !is_terminating )
- is_terminating = ParseInstruction(Args ) ;
-
- if (!is_terminating)
- PARSE_ERROR( "Non-terminated basic block found!");
-
- handler->handleBasicBlockEnd( BlockNo );
- ++BlockNo;
- }
- return BlockNo;
-}
-
-void AbstractBytecodeParser::ParseSymbolTable() {
- handler->handleSymbolTableBegin();
-
- while ( moreInBlock() ) {
- // Symtab block header: [num entries][type id number]
- unsigned NumEntries = read_vbr_uint();
- unsigned Typ = read_vbr_uint();
- const Type *Ty = getType(Typ);
-
- handler->handleSymbolTablePlane( Typ, NumEntries, Ty );
-
- for (unsigned i = 0; i != NumEntries; ++i) {
- // Symtab entry: [def slot #][name]
- unsigned slot = read_vbr_uint();
- std::string Name = read_str();
-
- if (Typ == Type::TypeTyID)
- handler->handleSymbolTableType( i, slot, Name );
- else
- handler->handleSymbolTableValue( i, slot, Name );
- }
- }
- checkPastBlockEnd("Symbol Table");
-
- handler->handleSymbolTableEnd();
-}
-
-void AbstractBytecodeParser::ParseFunctionLazily() {
- if (FunctionSignatureList.empty())
- throw std::string("FunctionSignatureList empty!");
-
- Function *Func = FunctionSignatureList.back();
- FunctionSignatureList.pop_back();
-
- // Save the information for future reading of the function
- LazyFunctionLoadMap[Func] = LazyFunctionInfo(BlockStart, BlockEnd);
-
- // Pretend we've `parsed' this function
- At = BlockEnd;
-}
-
-void AbstractBytecodeParser::ParseNextFunction(Function* Func) {
- // Find {start, end} pointers and slot in the map. If not there, we're done.
- LazyFunctionMap::iterator Fi = LazyFunctionLoadMap.find(Func);
-
- // Make sure we found it
- if ( Fi == LazyFunctionLoadMap.end() ) {
- PARSE_ERROR("Unrecognized function of type " << Func->getType()->getDescription());
- return;
- }
-
- BlockStart = At = Fi->second.Buf;
- BlockEnd = Fi->second.Buf;
- assert(Fi->first == Func);
-
- LazyFunctionLoadMap.erase(Fi);
-
- this->ParseFunctionBody( Func );
-}
-
-void AbstractBytecodeParser::ParseAllFunctionBodies() {
- LazyFunctionMap::iterator Fi = LazyFunctionLoadMap.begin();
- LazyFunctionMap::iterator Fe = LazyFunctionLoadMap.end();
-
- while ( Fi != Fe ) {
- Function* Func = Fi->first;
- BlockStart = At = Fi->second.Buf;
- BlockEnd = Fi->second.EndBuf;
- this->ParseFunctionBody(Func);
- ++Fi;
- }
-}
-
-void AbstractBytecodeParser::ParseFunctionBody(Function* Func ) {
-
- unsigned FuncSize = BlockEnd - At;
- GlobalValue::LinkageTypes Linkage = GlobalValue::ExternalLinkage;
-
- unsigned LinkageType = read_vbr_uint();
- switch (LinkageType) {
- case 0: Linkage = GlobalValue::ExternalLinkage; break;
- case 1: Linkage = GlobalValue::WeakLinkage; break;
- case 2: Linkage = GlobalValue::AppendingLinkage; break;
- case 3: Linkage = GlobalValue::InternalLinkage; break;
- case 4: Linkage = GlobalValue::LinkOnceLinkage; break;
- default:
- PARSE_ERROR("Invalid linkage type for Function.");
- Linkage = GlobalValue::InternalLinkage;
- break;
- }
-
- Func->setLinkage( Linkage );
- handler->handleFunctionBegin(Func,FuncSize);
-
- // Keep track of how many basic blocks we have read in...
- unsigned BlockNum = 0;
- bool InsertedArguments = false;
-
- BufPtr MyEnd = BlockEnd;
- while ( At < MyEnd ) {
- unsigned Type, Size;
- BufPtr OldAt = At;
- readBlock(Type, Size);
-
- switch (Type) {
- case BytecodeFormat::ConstantPool:
- ParseConstantPool(FunctionTypes );
- break;
-
- case BytecodeFormat::CompactionTable:
- ParseCompactionTable();
- break;
-
- case BytecodeFormat::BasicBlock:
- ParseBasicBlock(BlockNum++);
- break;
-
- case BytecodeFormat::InstructionList:
- if (BlockNum)
- PARSE_ERROR("InstructionList must come before basic blocks!");
- BlockNum = ParseInstructionList();
- break;
-
- case BytecodeFormat::SymbolTable:
- ParseSymbolTable();
- break;
-
- default:
- At += Size;
- if (OldAt > At)
- PARSE_ERROR("Wrapped around reading bytecode");
- break;
- }
- BlockEnd = MyEnd;
-
- // Malformed bc file if read past end of block.
- align32();
- }
-
- handler->handleFunctionEnd(Func);
-
- // Clear out function-level types...
- FunctionTypes.clear();
- CompactionTypeTable.clear();
-}
-
-void AbstractBytecodeParser::ParseCompactionTable() {
-
- handler->handleCompactionTableBegin();
-
- while ( moreInBlock() ) {
- unsigned NumEntries = read_vbr_uint();
- unsigned Ty;
-
- if ((NumEntries & 3) == 3) {
- NumEntries >>= 2;
- Ty = read_vbr_uint();
- } else {
- Ty = NumEntries >> 2;
- NumEntries &= 3;
- }
-
- handler->handleCompactionTablePlane( Ty, NumEntries );
-
- if (Ty == Type::TypeTyID) {
- for (unsigned i = 0; i != NumEntries; ++i) {
- unsigned TypeSlot = read_vbr_uint();
- const Type *Typ = getGlobalTableType(TypeSlot);
- handler->handleCompactionTableType( i, TypeSlot, Typ );
- }
- } else {
- const Type *Typ = getType(Ty);
- // Push the implicit zero
- for (unsigned i = 0; i != NumEntries; ++i) {
- unsigned ValSlot = read_vbr_uint();
- handler->handleCompactionTableValue( i, ValSlot, Typ );
- }
- }
- }
- handler->handleCompactionTableEnd();
-}
-
-const Type *AbstractBytecodeParser::ParseTypeConstant() {
- unsigned PrimType = read_vbr_uint();
-
- const Type *Val = 0;
- if ((Val = Type::getPrimitiveType((Type::TypeID)PrimType)))
- return Val;
-
- switch (PrimType) {
- case Type::FunctionTyID: {
- const Type *RetType = getType(read_vbr_uint());
-
- unsigned NumParams = read_vbr_uint();
-
- std::vector<const Type*> Params;
- while (NumParams--)
- Params.push_back(getType(read_vbr_uint()));
-
- bool isVarArg = Params.size() && Params.back() == Type::VoidTy;
- if (isVarArg) Params.pop_back();
-
- Type* result = FunctionType::get(RetType, Params, isVarArg);
- handler->handleType( result );
- return result;
- }
- case Type::ArrayTyID: {
- unsigned ElTyp = read_vbr_uint();
- const Type *ElementType = getType(ElTyp);
-
- unsigned NumElements = read_vbr_uint();
-
- BCR_TRACE(5, "Array Type Constant #" << ElTyp << " size="
- << NumElements << "\n");
- Type* result = ArrayType::get(ElementType, NumElements);
- handler->handleType( result );
- return result;
- }
- case Type::StructTyID: {
- std::vector<const Type*> Elements;
- unsigned Typ = read_vbr_uint();
- while (Typ) { // List is terminated by void/0 typeid
- Elements.push_back(getType(Typ));
- Typ = read_vbr_uint();
- }
-
- Type* result = StructType::get(Elements);
- handler->handleType( result );
- return result;
- }
- case Type::PointerTyID: {
- unsigned ElTyp = read_vbr_uint();
- BCR_TRACE(5, "Pointer Type Constant #" << ElTyp << "\n");
- Type* result = PointerType::get(getType(ElTyp));
- handler->handleType( result );
- return result;
- }
-
- case Type::OpaqueTyID: {
- Type* result = OpaqueType::get();
- handler->handleType( result );
- return result;
- }
-
- default:
- PARSE_ERROR("Don't know how to deserialize primitive type" << PrimType << "\n");
- return Val;
- }
-}
-
-// ParseTypeConstants - We have to use this weird code to handle recursive
-// types. We know that recursive types will only reference the current slab of
-// values in the type plane, but they can forward reference types before they
-// have been read. For example, Type #0 might be '{ Ty#1 }' and Type #1 might
-// be 'Ty#0*'. When reading Type #0, type number one doesn't exist. To fix
-// this ugly problem, we pessimistically insert an opaque type for each type we
-// are about to read. This means that forward references will resolve to
-// something and when we reread the type later, we can replace the opaque type
-// with a new resolved concrete type.
-//
-void AbstractBytecodeParser::ParseTypeConstants(
- TypeListTy &Tab, unsigned NumEntries
-) {
- assert(Tab.size() == 0 && "should not have read type constants in before!");
-
- // Insert a bunch of opaque types to be resolved later...
- Tab.reserve(NumEntries);
- for (unsigned i = 0; i != NumEntries; ++i)
- Tab.push_back(OpaqueType::get());
-
- // Loop through reading all of the types. Forward types will make use of the
- // opaque types just inserted.
- //
- for (unsigned i = 0; i != NumEntries; ++i) {
- const Type *NewTy = ParseTypeConstant(), *OldTy = Tab[i].get();
- if (NewTy == 0) throw std::string("Couldn't parse type!");
- BCR_TRACE(4, "#" << i << ": Read Type Constant: '" << NewTy <<
- "' Replacing: " << OldTy << "\n");
-
- // Don't insertValue the new type... instead we want to replace the opaque
- // type with the new concrete value...
- //
-
- // Refine the abstract type to the new type. This causes all uses of the
- // abstract type to use NewTy. This also will cause the opaque type to be
- // deleted...
- //
- cast<DerivedType>(const_cast<Type*>(OldTy))->refineAbstractTypeTo(NewTy);
-
- // This should have replace the old opaque type with the new type in the
- // value table... or with a preexisting type that was already in the system
- assert(Tab[i] != OldTy && "refineAbstractType didn't work!");
- }
-
- BCR_TRACE(5, "Resulting types:\n");
- for (unsigned i = 0; i < NumEntries; ++i) {
- BCR_TRACE(5, (void*)Tab[i].get() << " - " << Tab[i].get() << "\n");
- }
-}
-
-
-void AbstractBytecodeParser::ParseConstantValue(unsigned TypeID) {
-
- // We must check for a ConstantExpr before switching by type because
- // a ConstantExpr can be of any type, and has no explicit value.
- //
- // 0 if not expr; numArgs if is expr
- unsigned isExprNumArgs = read_vbr_uint();
-
- if (isExprNumArgs) {
- unsigned Opcode = read_vbr_uint();
- const Type* Typ = getType(TypeID);
-
- // FIXME: Encoding of constant exprs could be much more compact!
- std::vector<std::pair<const Type*,unsigned> > ArgVec;
- ArgVec.reserve(isExprNumArgs);
-
- // Read the slot number and types of each of the arguments
- for (unsigned i = 0; i != isExprNumArgs; ++i) {
- unsigned ArgValSlot = read_vbr_uint();
- unsigned ArgTypeSlot = read_vbr_uint();
- BCR_TRACE(4, "CE Arg " << i << ": Type: '" << *getType(ArgTypeSlot)
- << "' slot: " << ArgValSlot << "\n");
-
- // Get the arg value from its slot if it exists, otherwise a placeholder
- ArgVec.push_back(std::make_pair(getType(ArgTypeSlot), ArgValSlot));
- }
-
- handler->handleConstantExpression( Opcode, Typ, ArgVec );
- return;
- }
-
- // Ok, not an ConstantExpr. We now know how to read the given type...
- const Type *Ty = getType(TypeID);
- switch (Ty->getTypeID()) {
- case Type::BoolTyID: {
- unsigned Val = read_vbr_uint();
- if (Val != 0 && Val != 1)
- PARSE_ERROR("Invalid boolean value read.");
-
- handler->handleConstantValue( ConstantBool::get(Val == 1));
- break;
- }
-
- case Type::UByteTyID: // Unsigned integer types...
- case Type::UShortTyID:
- case Type::UIntTyID: {
- unsigned Val = read_vbr_uint();
- if (!ConstantUInt::isValueValidForType(Ty, Val))
- throw std::string("Invalid unsigned byte/short/int read.");
- handler->handleConstantValue( ConstantUInt::get(Ty, Val) );
- break;
- }
-
- case Type::ULongTyID: {
- handler->handleConstantValue( ConstantUInt::get(Ty, read_vbr_uint64()) );
- break;
- }
-
- case Type::SByteTyID: // Signed integer types...
- case Type::ShortTyID:
- case Type::IntTyID: {
- case Type::LongTyID:
- int64_t Val = read_vbr_int64();
- if (!ConstantSInt::isValueValidForType(Ty, Val))
- throw std::string("Invalid signed byte/short/int/long read.");
- handler->handleConstantValue( ConstantSInt::get(Ty, Val) );
- break;
- }
-
- case Type::FloatTyID: {
- float F;
- read_data(&F, &F+1);
- handler->handleConstantValue( ConstantFP::get(Ty, F) );
- break;
- }
-
- case Type::DoubleTyID: {
- double Val;
- read_data(&Val, &Val+1);
- handler->handleConstantValue( ConstantFP::get(Ty, Val) );
- break;
- }
-
- case Type::TypeTyID:
- PARSE_ERROR("Type constants shouldn't live in constant table!");
- break;
-
- case Type::ArrayTyID: {
- const ArrayType *AT = cast<ArrayType>(Ty);
- unsigned NumElements = AT->getNumElements();
- std::vector<unsigned> Elements;
- Elements.reserve(NumElements);
- while (NumElements--) // Read all of the elements of the constant.
- Elements.push_back(read_vbr_uint());
-
- handler->handleConstantArray( AT, Elements );
- break;
- }
-
- case Type::StructTyID: {
- const StructType *ST = cast<StructType>(Ty);
- std::vector<unsigned> Elements;
- Elements.reserve(ST->getNumElements());
- for (unsigned i = 0; i != ST->getNumElements(); ++i)
- Elements.push_back(read_vbr_uint());
- handler->handleConstantStruct( ST, Elements );
- break;
- }
-
- case Type::PointerTyID: { // ConstantPointerRef value...
- const PointerType *PT = cast<PointerType>(Ty);
- unsigned Slot = read_vbr_uint();
- handler->handleConstantPointer( PT, Slot );
- break;
- }
-
- default:
- PARSE_ERROR("Don't know how to deserialize constant value of type '"+
- Ty->getDescription());
- }
-}
-
-void AbstractBytecodeParser::ParseGlobalTypes() {
- ParseConstantPool(ModuleTypes);
-}
-
-void AbstractBytecodeParser::ParseStringConstants(unsigned NumEntries ){
- for (; NumEntries; --NumEntries) {
- unsigned Typ = read_vbr_uint();
- const Type *Ty = getType(Typ);
- if (!isa<ArrayType>(Ty))
- throw std::string("String constant data invalid!");
-
- const ArrayType *ATy = cast<ArrayType>(Ty);
- if (ATy->getElementType() != Type::SByteTy &&
- ATy->getElementType() != Type::UByteTy)
- throw std::string("String constant data invalid!");
-
- // Read character data. The type tells us how long the string is.
- char Data[ATy->getNumElements()];
- read_data(Data, Data+ATy->getNumElements());
-
- std::vector<Constant*> Elements(ATy->getNumElements());
- if (ATy->getElementType() == Type::SByteTy)
- for (unsigned i = 0, e = ATy->getNumElements(); i != e; ++i)
- Elements[i] = ConstantSInt::get(Type::SByteTy, (signed char)Data[i]);
- else
- for (unsigned i = 0, e = ATy->getNumElements(); i != e; ++i)
- Elements[i] = ConstantUInt::get(Type::UByteTy, (unsigned char)Data[i]);
-
- // Create the constant, inserting it as needed.
- ConstantArray *C = cast<ConstantArray>( ConstantArray::get(ATy, Elements) );
- handler->handleConstantString( C );
- }
-}
-
-
-void AbstractBytecodeParser::ParseConstantPool( TypeListTy &TypeTab) {
- while ( moreInBlock() ) {
- unsigned NumEntries = read_vbr_uint();
- unsigned Typ = read_vbr_uint();
- if (Typ == Type::TypeTyID) {
- ParseTypeConstants(TypeTab, NumEntries);
- } else if (Typ == Type::VoidTyID) {
- ParseStringConstants(NumEntries);
- } else {
- BCR_TRACE(3, "Type: '" << *getType(Typ) << "' NumEntries: "
- << NumEntries << "\n");
-
- for (unsigned i = 0; i < NumEntries; ++i) {
- ParseConstantValue(Typ);
- }
- }
- }
-
- checkPastBlockEnd("Constant Pool");
-}
-
-void AbstractBytecodeParser::ParseModuleGlobalInfo() {
-
- handler->handleModuleGlobalsBegin();
-
- // Read global variables...
- unsigned VarType = read_vbr_uint();
- while (VarType != Type::VoidTyID) { // List is terminated by Void
- // VarType Fields: bit0 = isConstant, bit1 = hasInitializer, bit2,3,4 =
- // Linkage, bit4+ = slot#
- unsigned SlotNo = VarType >> 5;
- unsigned LinkageID = (VarType >> 2) & 7;
- bool isConstant = VarType & 1;
- bool hasInitializer = VarType & 2;
- GlobalValue::LinkageTypes Linkage;
-
- switch (LinkageID) {
- case 0: Linkage = GlobalValue::ExternalLinkage; break;
- case 1: Linkage = GlobalValue::WeakLinkage; break;
- case 2: Linkage = GlobalValue::AppendingLinkage; break;
- case 3: Linkage = GlobalValue::InternalLinkage; break;
- case 4: Linkage = GlobalValue::LinkOnceLinkage; break;
- default:
- PARSE_ERROR("Unknown linkage type: " << LinkageID);
- Linkage = GlobalValue::InternalLinkage;
- break;
- }
-
- const Type *Ty = getType(SlotNo);
- if ( !Ty ) {
- PARSE_ERROR("Global has no type! SlotNo=" << SlotNo);
- }
-
- if ( !isa<PointerType>(Ty)) {
- PARSE_ERROR("Global not a pointer type! Ty= " << Ty->getDescription());
- }
-
- const Type *ElTy = cast<PointerType>(Ty)->getElementType();
-
- // Create the global variable...
- if (hasInitializer) {
- unsigned initSlot = read_vbr_uint();
- handler->handleInitializedGV( ElTy, isConstant, Linkage, initSlot );
- } else
- handler->handleGlobalVariable( ElTy, isConstant, Linkage );
-
- // Get next item
- VarType = read_vbr_uint();
- }
-
- // Read the function objects for all of the functions that are coming
- unsigned FnSignature = read_vbr_uint();
- while (FnSignature != Type::VoidTyID) { // List is terminated by Void
- const Type *Ty = getType(FnSignature);
- if (!isa<PointerType>(Ty) ||
- !isa<FunctionType>(cast<PointerType>(Ty)->getElementType())) {
- PARSE_ERROR( "Function not a pointer to function type! Ty = " +
- Ty->getDescription());
- // FIXME: what should Ty be if handler continues?
- }
-
- // We create functions by passing the underlying FunctionType to create...
- const FunctionType* FTy =
- cast<FunctionType>(cast<PointerType>(Ty)->getElementType());
- Function* Func = new Function(FTy, GlobalValue::ExternalLinkage);
-
- // Save this for later so we know type of lazily instantiated functions
- FunctionSignatureList.push_back(Func);
-
- handler->handleFunctionDeclaration(Func, FTy);
-
- // Get Next function signature
- FnSignature = read_vbr_uint();
- }
-
- if (hasInconsistentModuleGlobalInfo)
- align32();
-
- // Now that the function signature list is set up, reverse it so that we can
- // remove elements efficiently from the back of the vector.
- std::reverse(FunctionSignatureList.begin(), FunctionSignatureList.end());
-
- // This is for future proofing... in the future extra fields may be added that
- // we don't understand, so we transparently ignore them.
- //
- At = BlockEnd;
-
- handler->handleModuleGlobalsEnd();
-}
-
-void AbstractBytecodeParser::ParseVersionInfo() {
- unsigned Version = read_vbr_uint();
-
- // Unpack version number: low four bits are for flags, top bits = version
- Module::Endianness Endianness;
- Module::PointerSize PointerSize;
- Endianness = (Version & 1) ? Module::BigEndian : Module::LittleEndian;
- PointerSize = (Version & 2) ? Module::Pointer64 : Module::Pointer32;
-
- bool hasNoEndianness = Version & 4;
- bool hasNoPointerSize = Version & 8;
-
- RevisionNum = Version >> 4;
-
- // Default values for the current bytecode version
- hasInconsistentModuleGlobalInfo = false;
- hasExplicitPrimitiveZeros = false;
- hasRestrictedGEPTypes = false;
-
- switch (RevisionNum) {
- case 0: // LLVM 1.0, 1.1 release version
- // Base LLVM 1.0 bytecode format.
- hasInconsistentModuleGlobalInfo = true;
- hasExplicitPrimitiveZeros = true;
- // FALL THROUGH
- case 1: // LLVM 1.2 release version
- // LLVM 1.2 added explicit support for emitting strings efficiently.
-
- // Also, it fixed the problem where the size of the ModuleGlobalInfo block
- // included the size for the alignment at the end, where the rest of the
- // blocks did not.
-
- // LLVM 1.2 and before required that GEP indices be ubyte constants for
- // structures and longs for sequential types.
- hasRestrictedGEPTypes = true;
-
- // FALL THROUGH
- case 2: // LLVM 1.3 release version
- break;
-
- default:
- PARSE_ERROR("Unknown bytecode version number: " << RevisionNum);
- }
-
- if (hasNoEndianness) Endianness = Module::AnyEndianness;
- if (hasNoPointerSize) PointerSize = Module::AnyPointerSize;
-
- handler->handleVersionInfo(RevisionNum, Endianness, PointerSize );
-}
-
-void AbstractBytecodeParser::ParseModule() {
- unsigned Type, Size;
-
- FunctionSignatureList.clear(); // Just in case...
-
- // Read into instance variables...
- ParseVersionInfo();
- align32(); /// FIXME: Is this redundant? VI is first and 4 bytes!
-
- bool SeenModuleGlobalInfo = false;
- bool SeenGlobalTypePlane = false;
- BufPtr MyEnd = BlockEnd;
- while (At < MyEnd) {
- BufPtr OldAt = At;
- readBlock(Type, Size);
-
- switch (Type) {
-
- case BytecodeFormat::GlobalTypePlane:
- if ( SeenGlobalTypePlane )
- PARSE_ERROR("Two GlobalTypePlane Blocks Encountered!");
-
- ParseGlobalTypes();
- SeenGlobalTypePlane = true;
- break;
-
- case BytecodeFormat::ModuleGlobalInfo:
- if ( SeenModuleGlobalInfo )
- PARSE_ERROR("Two ModuleGlobalInfo Blocks Encountered!");
- ParseModuleGlobalInfo();
- SeenModuleGlobalInfo = true;
- break;
-
- case BytecodeFormat::ConstantPool:
- ParseConstantPool(ModuleTypes);
- break;
-
- case BytecodeFormat::Function:
- ParseFunctionLazily();
- break;
-
- case BytecodeFormat::SymbolTable:
- ParseSymbolTable();
- break;
-
- default:
- At += Size;
- if (OldAt > At) {
- PARSE_ERROR("Unexpected Block of Type" << Type << "encountered!" );
- }
- break;
- }
- BlockEnd = MyEnd;
- align32();
- }
-
- /// Make sure we pulled them all out. If we didn't then there's a declaration
- /// but a missing body. That's not allowed.
- if (!FunctionSignatureList.empty())
- throw std::string(
- "Function declared, but bytecode stream ended before definition");
-}
-
-void AbstractBytecodeParser::ParseBytecode(
- BufPtr b, unsigned Length,
- const std::string &ModuleID) {
-
- At = MemStart = BlockStart = b;
- MemEnd = BlockEnd = b + Length;
- handler->handleStart();
-
- // Read and check signature...
- unsigned Sig = read_uint();
- if (Sig != ('l' | ('l' << 8) | ('v' << 16) | ('m' << 24))) {
- PARSE_ERROR("Invalid bytecode signature: " << Sig);
- }
-
- handler->handleModuleBegin(ModuleID);
-
- unsigned Type, Size;
- readBlock(Type, Size);
- if ( Type != BytecodeFormat::Module ) {
- PARSE_ERROR("Expected Module Block! At: " << unsigned(intptr_t(At))
- << ", Type:" << Type << ", Size:" << Size);
- }
- if ( At + Size != MemEnd ) {
- PARSE_ERROR("Invalid Top Level Block Length! At: "
- << unsigned(intptr_t(At)) << ", Type:" << Type << ", Size:" << Size);
- }
- this->ParseModule();
-
- handler->handleModuleEnd(ModuleID);
-
- handler->handleFinish();
-}
-
-//===----------------------------------------------------------------------===//
-//=== Default Implementations of Handler Methods
-//===----------------------------------------------------------------------===//
-
-bool BytecodeHandler::handleError(const std::string& str ) { return false; }
-void BytecodeHandler::handleStart() { }
-void BytecodeHandler::handleFinish() { }
-void BytecodeHandler::handleModuleBegin(const std::string& id) { }
-void BytecodeHandler::handleModuleEnd(const std::string& id) { }
-void BytecodeHandler::handleVersionInfo( unsigned char RevisionNum,
- Module::Endianness Endianness, Module::PointerSize PointerSize) { }
-void BytecodeHandler::handleModuleGlobalsBegin() { }
-void BytecodeHandler::handleGlobalVariable(
- const Type* ElemType, bool isConstant, GlobalValue::LinkageTypes ) { }
-void BytecodeHandler::handleInitializedGV(
- const Type* ElemType, bool isConstant, GlobalValue::LinkageTypes,
- unsigned initSlot) {}
-void BytecodeHandler::handleType( const Type* Ty ) {}
-void BytecodeHandler::handleFunctionDeclaration(
- Function* Func, const FunctionType* FuncType) {}
-void BytecodeHandler::handleModuleGlobalsEnd() { }
-void BytecodeHandler::handleCompactionTableBegin() { }
-void BytecodeHandler::handleCompactionTablePlane( unsigned Ty,
- unsigned NumEntries) {}
-void BytecodeHandler::handleCompactionTableType( unsigned i, unsigned TypSlot,
- const Type* ) {}
-void BytecodeHandler::handleCompactionTableValue( unsigned i, unsigned ValSlot,
- const Type* ) {}
-void BytecodeHandler::handleCompactionTableEnd() { }
-void BytecodeHandler::handleSymbolTableBegin() { }
-void BytecodeHandler::handleSymbolTablePlane( unsigned Ty, unsigned NumEntries,
- const Type* Typ) { }
-void BytecodeHandler::handleSymbolTableType( unsigned i, unsigned slot,
- const std::string& name ) { }
-void BytecodeHandler::handleSymbolTableValue( unsigned i, unsigned slot,
- const std::string& name ) { }
-void BytecodeHandler::handleSymbolTableEnd() { }
-void BytecodeHandler::handleFunctionBegin( Function* Func,
- unsigned Size ) {}
-void BytecodeHandler::handleFunctionEnd( Function* Func) { }
-void BytecodeHandler::handleBasicBlockBegin( unsigned blocknum) { }
-bool BytecodeHandler::handleInstruction( unsigned Opcode, const Type* iType,
- std::vector<unsigned>& Operands, unsigned Size) {
- return Instruction::isTerminator(Opcode);
- }
-void BytecodeHandler::handleBasicBlockEnd(unsigned blocknum) { }
-void BytecodeHandler::handleGlobalConstantsBegin() { }
-void BytecodeHandler::handleConstantExpression( unsigned Opcode,
- const Type* Typ, std::vector<std::pair<const Type*,unsigned> > ArgVec ) { }
-void BytecodeHandler::handleConstantValue( Constant * c ) { }
-void BytecodeHandler::handleConstantArray( const ArrayType* AT,
- std::vector<unsigned>& Elements ) { }
-void BytecodeHandler::handleConstantStruct( const StructType* ST,
- std::vector<unsigned>& ElementSlots) { }
-void BytecodeHandler::handleConstantPointer(
- const PointerType* PT, unsigned Slot) { }
-void BytecodeHandler::handleConstantString( const ConstantArray* CA ) {}
-void BytecodeHandler::handleGlobalConstantsEnd() {}
-void BytecodeHandler::handleAlignment(unsigned numBytes) {}
-void BytecodeHandler::handleBlock(
- unsigned BType, const unsigned char* StartPtr, unsigned Size) {}
-void BytecodeHandler::handleVBR32(unsigned Size ) {}
-void BytecodeHandler::handleVBR64(unsigned Size ) {}
-
-// vim: sw=2
+++ /dev/null
-//===-- Parser.h - Abstract Interface To Bytecode Parsing -------*- C++ -*-===//
-//
-// The LLVM Compiler Infrastructure
-//
-// This file was developed by Reid Spencer and is distributed under the
-// University of Illinois Open Source License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This header file defines the interface to the Bytecode Parser and the
-// Bytecode Handler interface that it calls.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef BYTECODE_PARSER_H
-#define BYTECODE_PARSER_H
-
-#include "llvm/Constants.h"
-#include "llvm/DerivedTypes.h"
-#include "llvm/GlobalValue.h"
-#include "llvm/Module.h"
-#include <utility>
-#include <vector>
-#include <map>
-
-namespace llvm {
-
-class BytecodeHandler; ///< Forward declare the handler interface
-
-/// This class defines the interface for parsing a buffer of bytecode. The
-/// parser itself takes no action except to call the various functions of
-/// the handler interface. The parser's sole responsibility is the correct
-/// interpretation of the bytecode buffer. The handler is responsible for
-/// instantiating and keeping track of all values. As a convenience, the parser
-/// is responsible for materializing types and will pass them through the
-/// handler interface as necessary.
-/// @see BytecodeHandler
-/// @brief Abstract Bytecode Parser interface
-class AbstractBytecodeParser {
-
-/// @name Constructors
-/// @{
-public:
- AbstractBytecodeParser(
- BytecodeHandler* h,
- bool repAlignment = false,
- bool repBlocks = false,
- bool repVBR = false
- ) {
- handler = h;
- reportAlignment = repAlignment;
- reportBlocks = repBlocks;
- reportVBR = repVBR;
- }
-
- ~AbstractBytecodeParser() { }
-
-/// @}
-/// @name Types
-/// @{
-public:
- /// @brief A convenience type for the buffer pointer
- typedef const unsigned char* BufPtr;
-
- /// @brief The type used for vector of potentially abstract types
- typedef std::vector<PATypeHolder> TypeListTy;
-
- /// @brief
-
-/// @}
-/// @name Methods
-/// @{
-public:
-
- /// @brief Main interface to parsing a bytecode buffer.
- void ParseBytecode(const unsigned char *Buf, unsigned Length,
- const std::string &ModuleID);
-
- /// The ParseBytecode method lazily parses functions. Use this
- /// method to cause the parser to actually parse all the function bodies
- /// in the bytecode buffer.
- /// @see ParseBytecode
- /// @brief Parse all function bodies
- void ParseAllFunctionBodies ();
-
- /// The Parsebytecode method lazily parses functions. Use this
- /// method to casue the parser to parse the next function of a given
- /// types. Note that this will remove the function from what is to be
- /// included by ParseAllFunctionBodies.
- /// @see ParseAllFunctionBodies
- /// @see ParseBytecode
- /// @brief Parse the next function of specific type
- void ParseNextFunction (Function* Func) ;
-
-/// @}
-/// @name Parsing Units For Subclasses
-/// @{
-protected:
- /// @brief Parse whole module scope
- void ParseModule ();
-
- /// @brief Parse the version information block
- void ParseVersionInfo ();
-
- /// @brief Parse the ModuleGlobalInfo block
- void ParseModuleGlobalInfo ();
-
- /// @brief Parse a symbol table
- void ParseSymbolTable ();
-
- /// This function parses LLVM functions lazily. It obtains the type of the
- /// function and records where the body of the function is in the bytecode
- /// buffer. The caller can then use the ParseNextFunction and
- /// ParseAllFunctionBodies to get handler events for the functions.
- /// @brief Parse functions lazily.
- void ParseFunctionLazily ();
-
- /// @brief Parse a function body
- void ParseFunctionBody (Function* Func);
-
- /// @brief Parse a compaction table
- void ParseCompactionTable ();
-
- /// @brief Parse global types
- void ParseGlobalTypes ();
-
- /// @brief Parse a basic block (for LLVM 1.0 basic block blocks)
- void ParseBasicBlock (unsigned BlockNo);
-
- /// @brief parse an instruction list (for post LLVM 1.0 instruction lists
- /// with blocks differentiated by terminating instructions.
- unsigned ParseInstructionList();
-
- /// @brief Parse an instruction.
- bool ParseInstruction (std::vector<unsigned>& Args);
-
- /// @brief Parse a constant pool
- void ParseConstantPool (TypeListTy& List);
-
- /// @brief Parse a constant value
- void ParseConstantValue (unsigned TypeID);
-
- /// @brief Parse a block of types.
- void ParseTypeConstants (TypeListTy &Tab, unsigned NumEntries);
-
- /// @brief Parse a single type.
- const Type *ParseTypeConstant();
-
- /// @brief Parse a string constants block
- void ParseStringConstants (unsigned NumEntries);
-
-/// @}
-/// @name Data
-/// @{
-private:
- BufPtr MemStart; ///< Start of the memory buffer
- BufPtr MemEnd; ///< End of the memory buffer
- BufPtr BlockStart; ///< Start of current block being parsed
- BufPtr BlockEnd; ///< End of current block being parsed
- BufPtr At; ///< Where we're currently parsing at
-
- bool reportAlignment; ///< Parser should report alignment?
- bool reportBlocks; ///< Parser should report blocks?
- bool reportVBR; ///< Report VBR compression events
-
- // Information about the module, extracted from the bytecode revision number.
- unsigned char RevisionNum; // The rev # itself
-
- // Flags to distinguish LLVM 1.0 & 1.1 bytecode formats (revision #0)
-
- // Revision #0 had an explicit alignment of data only for the ModuleGlobalInfo
- // block. This was fixed to be like all other blocks in 1.2
- bool hasInconsistentModuleGlobalInfo;
-
- // Revision #0 also explicitly encoded zero values for primitive types like
- // int/sbyte/etc.
- bool hasExplicitPrimitiveZeros;
-
- // Flags to control features specific the LLVM 1.2 and before (revision #1)
-
- // LLVM 1.2 and earlier required that getelementptr structure indices were
- // ubyte constants and that sequential type indices were longs.
- bool hasRestrictedGEPTypes;
-
-
- /// CompactionTable - If a compaction table is active in the current function,
- /// this is the mapping that it contains.
- std::vector<Type*> CompactionTypeTable;
-
- // ConstantFwdRefs - This maintains a mapping between <Type, Slot #>'s and
- // forward references to constants. Such values may be referenced before they
- // are defined, and if so, the temporary object that they represent is held
- // here.
- //
- typedef std::map<std::pair<const Type*,unsigned>, Constant*> ConstantRefsType;
- ConstantRefsType ConstantFwdRefs;
-
- // TypesLoaded - This vector mirrors the Values[TypeTyID] plane. It is used
- // to deal with forward references to types.
- //
- TypeListTy ModuleTypes;
- TypeListTy FunctionTypes;
-
- // When the ModuleGlobalInfo section is read, we create a FunctionType object
- // for each function in the module. When the function is loaded, this type is
- // used to instantiate the actual function object.
-
- std::vector<Function*> FunctionSignatureList;
-
- // Constant values are read in after global variables. Because of this, we
- // must defer setting the initializers on global variables until after module
- // level constants have been read. In the mean time, this list keeps track of
- // what we must do.
- //
- std::vector<std::pair<GlobalVariable*, unsigned> > GlobalInits;
-
-/// @}
-/// @name Implementation Details
-/// @{
-private:
- /// This stores the parser's handler. It makes virtual function calls through
- /// the BytecodeHandler to notify the handler of parsing events. What the
- /// handler does with the events is completely orthogonal to the business of
- /// parsing the bytecode.
- /// @brief The handler of bytecode parsing events.
- BytecodeHandler* handler;
-
- /// For lazy reading-in of functions, we need to save away several pieces of
- /// information about each function: its begin and end pointer in the buffer
- /// and its FunctionSlot.
- struct LazyFunctionInfo {
- const unsigned char *Buf, *EndBuf;
- LazyFunctionInfo(const unsigned char *B = 0, const unsigned char *EB = 0)
- : Buf(B), EndBuf(EB) {}
- };
- typedef std::map<Function*, LazyFunctionInfo> LazyFunctionMap;
- LazyFunctionMap LazyFunctionLoadMap;
-
-private:
-
- /// Is there more to parse in the current block?
- inline bool moreInBlock();
-
- /// Have we read past the end of the block
- inline void checkPastBlockEnd(const char * block_name);
-
- /// Align to 32 bits
- inline void align32();
-
- /// Reader interface
- inline unsigned read_uint();
- inline unsigned read_vbr_uint();
- inline uint64_t read_vbr_uint64();
- inline int64_t read_vbr_int64();
- inline std::string read_str();
- inline void read_data(void *Ptr, void *End);
-
- /// Read a block header
- inline void readBlock(unsigned &Type, unsigned &Size);
-
- const Type *AbstractBytecodeParser::getType(unsigned ID);
- /// getGlobalTableType - This is just like getType, but when a compaction
- /// table is in use, it is ignored. Also, no forward references or other
- /// fancy features are supported.
- const Type *getGlobalTableType(unsigned Slot) {
- if (Slot < Type::FirstDerivedTyID) {
- const Type *Ty = Type::getPrimitiveType((Type::TypeID)Slot);
- assert(Ty && "Not a primitive type ID?");
- return Ty;
- }
- Slot -= Type::FirstDerivedTyID;
- if (Slot >= ModuleTypes.size())
- throw std::string("Illegal compaction table type reference!");
- return ModuleTypes[Slot];
- }
-
- unsigned getGlobalTableTypeSlot(const Type *Ty) {
- if (Ty->isPrimitiveType())
- return Ty->getTypeID();
- TypeListTy::iterator I = find(ModuleTypes.begin(),
- ModuleTypes.end(), Ty);
- if (I == ModuleTypes.end())
- throw std::string("Didn't find type in ModuleTypes.");
- return Type::FirstDerivedTyID + (&*I - &ModuleTypes[0]);
- }
-
- AbstractBytecodeParser(const AbstractBytecodeParser &); // DO NOT IMPLEMENT
- void operator=(const AbstractBytecodeParser &); // DO NOT IMPLEMENT
-
-/// @}
-};
-
-/// This class provides the interface for the handling bytecode events during
-/// parsing. The methods on this interface are invoked by the
-/// AbstractBytecodeParser as it discovers the content of a bytecode stream.
-/// This class provides a a clear separation of concerns between recognizing
-/// the semantic units of a bytecode file and deciding what to do with them.
-/// The AbstractBytecodeParser recognizes the content of the bytecode file and
-/// calls the BytecodeHandler methods to determine what should be done. This
-/// arrangement allows Bytecode files to be read and handled for a number of
-/// purposes simply by creating a subclass of BytecodeHandler. None of the
-/// parsing details need to be understood, only the meaning of the calls
-/// made on this interface.
-///
-/// Another paradigm that uses this design pattern is the XML SAX Parser. The
-/// ContentHandler for SAX plays the same role as the BytecodeHandler here.
-/// @see AbstractbytecodeParser
-/// @brief Handle Bytecode Parsing Events
-class BytecodeHandler {
-
-/// @name Constructors And Operators
-/// @{
-public:
- /// @brief Default constructor (empty)
- BytecodeHandler() {}
- /// @brief Virtual destructor (empty)
- virtual ~BytecodeHandler() {}
-
-private:
- BytecodeHandler(const BytecodeHandler &); // DO NOT IMPLEMENT
- void operator=(const BytecodeHandler &); // DO NOT IMPLEMENT
-
-/// @}
-/// @name Handler Methods
-/// @{
-public:
-
- /// This method is called whenever the parser detects an error in the
- /// bytecode formatting. Returning true will cause the parser to keep
- /// going, however this is inadvisable in most cases. Returning false will
- /// cause the parser to throw the message as a std::string.
- /// @brief Handle parsing errors.
- virtual bool handleError(const std::string& str );
-
- /// This method is called at the beginning of a parse before anything is
- /// read in order to give the handler a chance to initialize.
- /// @brief Handle the start of a bytecode parse
- virtual void handleStart();
-
- /// This method is called at the end of a parse after everything has been
- /// read in order to give the handler a chance to terminate.
- /// @brief Handle the end of a bytecode parse
- virtual void handleFinish();
-
- /// This method is called at the start of a module to indicate that a
- /// module is being parsed.
- /// @brief Handle the start of a module.
- virtual void handleModuleBegin(const std::string& id);
-
- /// This method is called at the end of a module to indicate that the module
- /// previously being parsed has concluded.
- /// @brief Handle the end of a module.
- virtual void handleModuleEnd(const std::string& id);
-
- /// This method is called once the version information has been parsed. It
- /// provides the information about the version of the bytecode file being
- /// read.
- /// @brief Handle the bytecode prolog
- virtual void handleVersionInfo(
- unsigned char RevisionNum, ///< Byte code revision number
- Module::Endianness Endianness, ///< Endianness indicator
- Module::PointerSize PointerSize ///< PointerSize indicator
- );
-
- /// This method is called at the start of a module globals block which
- /// contains the global variables and the function placeholders
- virtual void handleModuleGlobalsBegin();
-
- /// This method is called when a non-initialized global variable is
- /// recognized. Its type, constness, and linkage type are provided.
- /// @brief Handle a non-initialized global variable
- virtual void handleGlobalVariable(
- const Type* ElemType, ///< The type of the global variable
- bool isConstant, ///< Whether the GV is constant or not
- GlobalValue::LinkageTypes ///< The linkage type of the GV
- );
-
- /// This method is called when an initialized global variable is recognized.
- /// Its type constness, linkage type, and the slot number of the initializer
- /// are provided.
- /// @brief Handle an intialized global variable.
- virtual void handleInitializedGV(
- const Type* ElemType, ///< The type of the global variable
- bool isConstant, ///< Whether the GV is constant or not
- GlobalValue::LinkageTypes,///< The linkage type of the GV
- unsigned initSlot ///< Slot number of GV's initializer
- );
-
- /// This method is called when a new type is recognized. The type is
- /// converted from the bytecode and passed to this method.
- /// @brief Handle a type
- virtual void handleType( const Type* Ty );
-
- /// This method is called when the function prototype for a function is
- /// encountered in the module globals block.
- virtual void handleFunctionDeclaration(
- Function* Func,
- const FunctionType* FuncType ///< The type of the function
- );
-
- /// This method is called at the end of the module globals block.
- /// @brief Handle end of module globals block.
- virtual void handleModuleGlobalsEnd();
-
- /// This method is called at the beginning of a compaction table.
- /// @brief Handle start of compaction table.
- virtual void handleCompactionTableBegin();
-
- /// @brief Handle start of a compaction table plane
- virtual void handleCompactionTablePlane(
- unsigned Ty,
- unsigned NumEntries
- );
-
-
- /// @brief Handle a type entry in the compaction table
- virtual void handleCompactionTableType(
- unsigned i,
- unsigned TypSlot,
- const Type*
- );
-
- /// @brief Handle a value entry in the compaction table
- virtual void handleCompactionTableValue(
- unsigned i,
- unsigned ValSlot,
- const Type*
- );
-
- /// @brief Handle end of a compaction table
- virtual void handleCompactionTableEnd();
-
- /// @brief Handle start of a symbol table
- virtual void handleSymbolTableBegin();
-
- /// @brief Handle start of a symbol table plane
- virtual void handleSymbolTablePlane(
- unsigned Ty,
- unsigned NumEntries,
- const Type* Ty
- );
-
- /// @brief Handle a named type in the symbol table
- virtual void handleSymbolTableType(
- unsigned i,
- unsigned slot,
- const std::string& name
- );
-
- /// @brief Handle a named value in the symbol table
- virtual void handleSymbolTableValue(
- unsigned i,
- unsigned slot,
- const std::string& name
- );
-
- /// @brief Handle the end of a symbol table
- virtual void handleSymbolTableEnd();
-
- /// @brief Handle the beginning of a function body
- virtual void handleFunctionBegin(
- Function* Func, unsigned Size
- );
-
- /// @brief Handle the end of a function body
- virtual void handleFunctionEnd(
- Function* Func
- );
-
- /// @brief Handle the beginning of a basic block
- virtual void handleBasicBlockBegin(
- unsigned blocknum
- );
-
- /// This method is called for each instruction that is parsed.
- /// @returns true if the instruction is a block terminating instruction
- /// @brief Handle an instruction
- virtual bool handleInstruction(
- unsigned Opcode,
- const Type* iType,
- std::vector<unsigned>& Operands,
- unsigned Length
- );
-
- /// @brief Handle the end of a basic block
- virtual void handleBasicBlockEnd(unsigned blocknum);
-
- /// @brief Handle start of global constants block.
- virtual void handleGlobalConstantsBegin();
-
- /// @brief Handle a constant expression
- virtual void handleConstantExpression(
- unsigned Opcode,
- const Type* Typ,
- std::vector<std::pair<const Type*,unsigned> > ArgVec
- );
-
- /// @brief Handle a constant array
- virtual void handleConstantArray(
- const ArrayType* AT,
- std::vector<unsigned>& ElementSlots
- );
-
- /// @brief Handle a constant structure
- virtual void handleConstantStruct(
- const StructType* ST,
- std::vector<unsigned>& ElementSlots
- );
-
- /// @brief Handle a constant pointer
- virtual void handleConstantPointer(
- const PointerType* PT,
- unsigned Slot
- );
-
- /// @brief Handle a constant strings (array special case)
- virtual void handleConstantString(
- const ConstantArray* CA
- );
-
- /// @brief Handle a primitive constant value
- virtual void handleConstantValue( Constant * c );
-
- /// @brief Handle the end of the global constants
- virtual void handleGlobalConstantsEnd();
-
- /// @brief Handle an alignment event
- virtual void handleAlignment(unsigned numBytes);
-
- virtual void handleBlock(
- unsigned BType, ///< The type of block
- const unsigned char* StartPtr, ///< The start of the block
- unsigned Size ///< The size of the block
- );
- virtual void handleVBR32(unsigned Size );
- virtual void handleVBR64(unsigned Size );
-/// @}
-
-};
-
-} // End llvm namespace
-
-// vim: sw=2
-#endif