X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;ds=sidebyside;f=tools%2Flto%2Flto.cpp;h=02034bbf84660abfd4f8d883c901411bc772b729;hb=31895e73591d3c9ceae731a1274c8f56194b9616;hp=c4face2712abe163c71a25012719daa41bbefea8;hpb=dc4c38279f6bf3b001515e6723e7b6d79ed378b0;p=oota-llvm.git diff --git a/tools/lto/lto.cpp b/tools/lto/lto.cpp index c4face2712a..02034bbf846 100644 --- a/tools/lto/lto.cpp +++ b/tools/lto/lto.cpp @@ -2,414 +2,267 @@ // // The LLVM Compiler Infrastructure // -// This file was developed by Devang Patel and is distributed under -// the University of Illinois Open Source License. See LICENSE.TXT for details. +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. // //===----------------------------------------------------------------------===// // -// This file implementes link time optimization library. This library is +// This file implements the Link Time Optimization library. This library is // intended to be used by linker to optimize code at link time. // //===----------------------------------------------------------------------===// -#include "llvm/Module.h" -#include "llvm/PassManager.h" -#include "llvm/Linker.h" -#include "llvm/Constants.h" -#include "llvm/DerivedTypes.h" -#include "llvm/SymbolTable.h" -#include "llvm/Bytecode/Reader.h" -#include "llvm/Bytecode/Writer.h" -#include "llvm/Support/CommandLine.h" -#include "llvm/Support/FileUtilities.h" -#include "llvm/Support/SystemUtils.h" -#include "llvm/Support/Mangler.h" -#include "llvm/System/Program.h" -#include "llvm/System/Signals.h" -#include "llvm/Analysis/Passes.h" -#include "llvm/Analysis/Verifier.h" -#include "llvm/Target/SubtargetFeature.h" -#include "llvm/Target/TargetData.h" -#include "llvm/Target/TargetMachine.h" -#include "llvm/Target/TargetMachineRegistry.h" -#include "llvm/Transforms/IPO.h" -#include "llvm/Transforms/Scalar.h" -#include "llvm/Analysis/LoadValueNumbering.h" -#include "llvm/LinkTimeOptimizer.h" -#include -#include - -using namespace llvm; - -extern "C" -llvm::LinkTimeOptimizer *createLLVMOptimizer() +#include "llvm-c/lto.h" +#include "llvm-c/Core.h" + +#include "LTOModule.h" +#include "LTOCodeGenerator.h" + + +// holds most recent error string +// *** not thread safe *** +static std::string sLastErrorString; + + + +// +// returns a printable string +// +extern const char* lto_get_version() { - llvm::LTO *l = new llvm::LTO(); - return l; + return LTOCodeGenerator::getVersionString(); } +// +// returns the last error string or NULL if last operation was successful +// +const char* lto_get_error_message() +{ + return sLastErrorString.c_str(); +} -/// If symbol is not used then make it internal and let optimizer takes -/// care of it. -void LLVMSymbol::mayBeNotUsed() { - gv->setLinkage(GlobalValue::InternalLinkage); + +// +// validates if a file is a loadable object file +// +bool lto_module_is_object_file(const char* path) +{ + return LTOModule::isBitcodeFile(path); } -// Helper routine -// FIXME : Take advantage of GlobalPrefix from AsmPrinter -static const char *addUnderscore(const char *name) { - size_t namelen = strlen(name); - char *symName = (char*)malloc(namelen+2); - symName[0] = '_'; - strcpy(&symName[1], name); - return symName; + +// +// validates if a file is a loadable object file compilable for requested target +// +bool lto_module_is_object_file_for_target(const char* path, + const char* target_triplet_prefix) +{ + return LTOModule::isBitcodeFileForTarget(path, target_triplet_prefix); } -// Map LLVM LinkageType to LTO LinakgeType -static LTOLinkageTypes -getLTOLinkageType(GlobalValue *v) + +// +// validates if a buffer is a loadable object file +// +bool lto_module_is_object_file_in_memory(const void* mem, size_t length) { - LTOLinkageTypes lt; - if (v->hasExternalLinkage()) - lt = LTOExternalLinkage; - else if (v->hasLinkOnceLinkage()) - lt = LTOLinkOnceLinkage; - else if (v->hasWeakLinkage()) - lt = LTOWeakLinkage; - else - // Otherwise it is internal linkage for link time optimizer - lt = LTOInternalLinkage; - return lt; + return LTOModule::isBitcodeFile(mem, length); } -// Find exeternal symbols referenced by VALUE. This is a recursive function. -static void -findExternalRefs(Value *value, std::set &references, - Mangler &mangler) { - - if (GlobalValue *gv = dyn_cast(value)) { - LTOLinkageTypes lt = getLTOLinkageType(gv); - if (lt != LTOInternalLinkage && strncmp (gv->getName().c_str(), "llvm.", 5)) - references.insert(mangler.getValueName(gv)); - } - - // GlobalValue, even with InternalLinkage type, may have operands with - // ExternalLinkage type. Do not ignore these operands. - if (Constant *c = dyn_cast(value)) - // Handle ConstantExpr, ConstantStruct, ConstantArry etc.. - for (unsigned i = 0, e = c->getNumOperands(); i != e; ++i) - findExternalRefs(c->getOperand(i), references, mangler); + +// +// validates if a buffer is a loadable object file compilable for the target +// +bool lto_module_is_object_file_in_memory_for_target(const void* mem, + size_t length, const char* target_triplet_prefix) +{ + return LTOModule::isBitcodeFileForTarget(mem, length, target_triplet_prefix); } -/// InputFilename is a LLVM bytecode file. If Module with InputFilename is -/// available then return it. Otherwise parseInputFilename. -Module * -LTO::getModule(const std::string &InputFilename) + + +// +// loads an object file from disk +// returns NULL on error (check lto_get_error_message() for details) +// +lto_module_t lto_module_create(const char* path, LLVMContextRef Ctxt) { - Module *m = NULL; - - NameToModuleMap::iterator pos = allModules.find(InputFilename.c_str()); - if (pos != allModules.end()) - m = allModules[InputFilename.c_str()]; - else { - m = ParseBytecodeFile(InputFilename); - allModules[InputFilename.c_str()] = m; - } - return m; + return LTOModule::makeLTOModule(path, *llvm::unwrap(Ctxt), + sLastErrorString); } -/// InputFilename is a LLVM bytecode file. Reade this bytecode file and -/// set corresponding target triplet string. -void -LTO::getTargetTriple(const std::string &InputFilename, - std::string &targetTriple) + +// +// loads an object file from memory +// returns NULL on error (check lto_get_error_message() for details) +// +lto_module_t lto_module_create_from_memory(const void* mem, size_t length, + LLVMContextRef Ctxt) +{ + return LTOModule::makeLTOModule(mem, length, *llvm::unwrap(Ctxt), + sLastErrorString); +} + + +// +// frees all memory for a module +// upon return the lto_module_t is no longer valid +// +void lto_module_dispose(lto_module_t mod) +{ + delete mod; +} + + +// +// returns triplet string which the object module was compiled under +// +const char* lto_module_get_target_triple(lto_module_t mod) { - Module *m = getModule(InputFilename); - if (m) - targetTriple = m->getTargetTriple(); + return mod->getTargetTriple(); } -/// InputFilename is a LLVM bytecode file. Read it using bytecode reader. -/// Collect global functions and symbol names in symbols vector. -/// Collect external references in references vector. -/// Return LTO_READ_SUCCESS if there is no error. -enum LTOStatus -LTO::readLLVMObjectFile(const std::string &InputFilename, - NameToSymbolMap &symbols, - std::set &references) + +// +// returns the number of symbols in the object module +// +uint32_t lto_module_get_num_symbols(lto_module_t mod) { - Module *m = getModule(InputFilename); - if (!m) - return LTO_READ_FAILURE; - - // Use mangler to add GlobalPrefix to names to match linker names. - // FIXME : Instead of hard coding "-" use GlobalPrefix. - Mangler mangler(*m, "_"); - - modules.push_back(m); - - for (Module::iterator f = m->begin(), e = m->end(); f != e; ++f) { - - LTOLinkageTypes lt = getLTOLinkageType(f); - - if (!f->isExternal() && lt != LTOInternalLinkage - && strncmp (f->getName().c_str(), "llvm.", 5)) { - LLVMSymbol *newSymbol = new LLVMSymbol(lt, f, f->getName(), - mangler.getValueName(f)); - symbols[newSymbol->getMangledName()] = newSymbol; - allSymbols[newSymbol->getMangledName()] = newSymbol; - } - - // Collect external symbols referenced by this function. - for (Function::iterator b = f->begin(), fe = f->end(); b != fe; ++b) - for (BasicBlock::iterator i = b->begin(), be = b->end(); - i != be; ++i) - for (unsigned count = 0, total = i->getNumOperands(); - count != total; ++count) - findExternalRefs(i->getOperand(count), references, mangler); - } - - for (Module::global_iterator v = m->global_begin(), e = m->global_end(); - v != e; ++v) { - LTOLinkageTypes lt = getLTOLinkageType(v); - if (!v->isExternal() && lt != LTOInternalLinkage - && strncmp (v->getName().c_str(), "llvm.", 5)) { - LLVMSymbol *newSymbol = new LLVMSymbol(lt, v, v->getName(), - mangler.getValueName(v)); - symbols[newSymbol->getMangledName()] = newSymbol; - allSymbols[newSymbol->getMangledName()] = newSymbol; - - for (unsigned count = 0, total = v->getNumOperands(); - count != total; ++count) - findExternalRefs(v->getOperand(count), references, mangler); - - } - } - - return LTO_READ_SUCCESS; + return mod->getSymbolCount(); } -/// Optimize module M using various IPO passes. Use exportList to -/// internalize selected symbols. Target platform is selected -/// based on information available to module M. No new target -/// features are selected. -static enum LTOStatus lto_optimize(Module *M, std::ostream &Out, - std::vector &exportList) +// +// returns the name of the ith symbol in the object module +// +const char* lto_module_get_symbol_name(lto_module_t mod, uint32_t index) { - // Instantiate the pass manager to organize the passes. - PassManager Passes; - - // Collect Target info - std::string Err; - const TargetMachineRegistry::Entry* March = - TargetMachineRegistry::getClosestStaticTargetForModule(*M, Err); - - if (March == 0) - return LTO_NO_TARGET; - - // Create target - std::string Features; - std::auto_ptr target(March->CtorFn(*M, Features)); - if (!target.get()) - return LTO_NO_TARGET; - - TargetMachine &Target = *target.get(); - - // Start off with a verification pass. - Passes.add(createVerifierPass()); - - // Add an appropriate TargetData instance for this module... - Passes.add(new TargetData(*Target.getTargetData())); - - // Often if the programmer does not specify proper prototypes for the - // functions they are calling, they end up calling a vararg version of the - // function that does not get a body filled in (the real function has typed - // arguments). This pass merges the two functions. - Passes.add(createFunctionResolvingPass()); - - // Internalize symbols if export list is nonemty - if (!exportList.empty()) - Passes.add(createInternalizePass(exportList)); - - // Now that we internalized some globals, see if we can hack on them! - Passes.add(createGlobalOptimizerPass()); - - // Linking modules together can lead to duplicated global constants, only - // keep one copy of each constant... - Passes.add(createConstantMergePass()); - - // If the -s command line option was specified, strip the symbols out of the - // resulting program to make it smaller. -s is a GLD option that we are - // supporting. - Passes.add(createStripSymbolsPass()); - - // Propagate constants at call sites into the functions they call. - Passes.add(createIPConstantPropagationPass()); - - // Remove unused arguments from functions... - Passes.add(createDeadArgEliminationPass()); - - Passes.add(createFunctionInliningPass()); // Inline small functions - - Passes.add(createPruneEHPass()); // Remove dead EH info - - Passes.add(createGlobalDCEPass()); // Remove dead functions - - // If we didn't decide to inline a function, check to see if we can - // transform it to pass arguments by value instead of by reference. - Passes.add(createArgumentPromotionPass()); - - // The IPO passes may leave cruft around. Clean up after them. - Passes.add(createInstructionCombiningPass()); - - Passes.add(createScalarReplAggregatesPass()); // Break up allocas - - // Run a few AA driven optimizations here and now, to cleanup the code. - Passes.add(createGlobalsModRefPass()); // IP alias analysis - - Passes.add(createLICMPass()); // Hoist loop invariants - Passes.add(createLoadValueNumberingPass()); // GVN for load instrs - Passes.add(createGCSEPass()); // Remove common subexprs - Passes.add(createDeadStoreEliminationPass()); // Nuke dead stores - - // Cleanup and simplify the code after the scalar optimizations. - Passes.add(createInstructionCombiningPass()); - - // Delete basic blocks, which optimization passes may have killed... - Passes.add(createCFGSimplificationPass()); - - // Now that we have optimized the program, discard unreachable functions... - Passes.add(createGlobalDCEPass()); - - // Make sure everything is still good. - Passes.add(createVerifierPass()); - - FunctionPassManager *CodeGenPasses = - new FunctionPassManager(new ExistingModuleProvider(M)); - - CodeGenPasses->add(new TargetData(*Target.getTargetData())); - Target.addPassesToEmitFile(*CodeGenPasses, Out, TargetMachine::AssemblyFile, - true); - - // Run our queue of passes all at once now, efficiently. - Passes.run(*M); - - // Run the code generator, if present. - CodeGenPasses->doInitialization(); - for (Module::iterator I = M->begin(), E = M->end(); I != E; ++I) { - if (!I->isExternal()) - CodeGenPasses->run(*I); - } - CodeGenPasses->doFinalization(); - - return LTO_OPT_SUCCESS; + return mod->getSymbolName(index); } -///Link all modules together and optimize them using IPO. Generate -/// native object file using OutputFilename -/// Return appropriate LTOStatus. -enum LTOStatus -LTO::optimizeModules(const std::string &OutputFilename, - std::vector &exportList, - std::string &targetTriple) + +// +// returns the attributes of the ith symbol in the object module +// +lto_symbol_attributes lto_module_get_symbol_attribute(lto_module_t mod, + uint32_t index) { - if (modules.empty()) - return LTO_NO_WORK; - - std::ios::openmode io_mode = - std::ios::out | std::ios::trunc | std::ios::binary; - std::string *errMsg = NULL; - Module *bigOne = modules[0]; - Linker theLinker("LinkTimeOptimizer", bigOne, false); - for (unsigned i = 1, e = modules.size(); i != e; ++i) - if (theLinker.LinkModules(bigOne, modules[i], errMsg)) - return LTO_MODULE_MERGE_FAILURE; - -#if 0 - // Enable this when -save-temps is used - std::ofstream Out("big.bc", io_mode); - WriteBytecodeToFile(bigOne, Out, true); -#endif - - // Strip leading underscore because it was added to match names - // seen by linker. - for (unsigned i = 0, e = exportList.size(); i != e; ++i) { - const char *name = exportList[i]; - NameToSymbolMap::iterator itr = allSymbols.find(name); - if (itr != allSymbols.end()) - exportList[i] = allSymbols[name]->getName(); - } - - - std::string ErrMsg; - sys::Path TempDir = sys::Path::GetTemporaryDirectory(&ErrMsg); - if (TempDir.isEmpty()) { - std::cerr << "lto: " << ErrMsg << "\n"; - return LTO_WRITE_FAILURE; - } - sys::Path tmpAsmFilePath(TempDir); - if (!tmpAsmFilePath.appendComponent("lto")) { - std::cerr << "lto: " << ErrMsg << "\n"; - TempDir.eraseFromDisk(true); - return LTO_WRITE_FAILURE; - } - if (tmpAsmFilePath.createTemporaryFileOnDisk(&ErrMsg)) { - std::cerr << "lto: " << ErrMsg << "\n"; - TempDir.eraseFromDisk(true); - return LTO_WRITE_FAILURE; - } - sys::RemoveFileOnSignal(tmpAsmFilePath); - - std::ofstream asmFile(tmpAsmFilePath.c_str(), io_mode); - if (!asmFile.is_open() || asmFile.bad()) { - if (tmpAsmFilePath.exists()) { - tmpAsmFilePath.eraseFromDisk(); - TempDir.eraseFromDisk(true); - } - return LTO_WRITE_FAILURE; - } - - enum LTOStatus status = lto_optimize(bigOne, asmFile, exportList); - asmFile.close(); - if (status != LTO_OPT_SUCCESS) { - tmpAsmFilePath.eraseFromDisk(); - TempDir.eraseFromDisk(true); - return status; - } - - targetTriple = bigOne->getTargetTriple(); - - // Run GCC to assemble and link the program into native code. - // - // Note: - // We can't just assemble and link the file with the system assembler - // and linker because we don't know where to put the _start symbol. - // GCC mysteriously knows how to do it. - const sys::Path gcc = sys::Program::FindProgramByName("gcc"); - if (gcc.isEmpty()) { - tmpAsmFilePath.eraseFromDisk(); - TempDir.eraseFromDisk(true); - return LTO_ASM_FAILURE; - } - - std::vector args; - args.push_back(gcc.c_str()); - args.push_back("-c"); - args.push_back("-x"); - args.push_back("assembler"); - args.push_back("-o"); - args.push_back(OutputFilename.c_str()); - args.push_back(tmpAsmFilePath.c_str()); - args.push_back(0); - - if (sys::Program::ExecuteAndWait(gcc, &args[0], 0, 0, 1, &ErrMsg)) { - std::cerr << "lto: " << ErrMsg << "\n"; - return LTO_ASM_FAILURE; - } - - tmpAsmFilePath.eraseFromDisk(); - TempDir.eraseFromDisk(true); - - return LTO_OPT_SUCCESS; + return mod->getSymbolAttributes(index); } + + + + + +// +// instantiates a code generator +// returns NULL if there is an error +// +lto_code_gen_t lto_codegen_create(LLVMContextRef ContextRef) +{ + return new LTOCodeGenerator(*llvm::unwrap(ContextRef)); +} + + + +// +// frees all memory for a code generator +// upon return the lto_code_gen_t is no longer valid +// +void lto_codegen_dispose(lto_code_gen_t cg) +{ + delete cg; +} + + + +// +// add an object module to the set of modules for which code will be generated +// returns true on error (check lto_get_error_message() for details) +// +bool lto_codegen_add_module(lto_code_gen_t cg, lto_module_t mod) +{ + return cg->addModule(mod, sLastErrorString); +} + + +// +// sets what if any format of debug info should be generated +// returns true on error (check lto_get_error_message() for details) +// +bool lto_codegen_set_debug_model(lto_code_gen_t cg, lto_debug_model debug) +{ + return cg->setDebugInfo(debug, sLastErrorString); +} + + +// +// sets what code model to generated +// returns true on error (check lto_get_error_message() for details) +// +bool lto_codegen_set_pic_model(lto_code_gen_t cg, lto_codegen_model model) +{ + return cg->setCodePICModel(model, sLastErrorString); +} + +// +// sets the path to gcc +// +void lto_codegen_set_gcc_path(lto_code_gen_t cg, const char* path) +{ + cg->setGccPath(path); +} + +// +// sets the path to the assembler tool +// +void lto_codegen_set_assembler_path(lto_code_gen_t cg, const char* path) +{ + cg->setAssemblerPath(path); +} + +// +// adds to a list of all global symbols that must exist in the final +// generated code. If a function is not listed there, it might be +// inlined into every usage and optimized away. +// +void lto_codegen_add_must_preserve_symbol(lto_code_gen_t cg, const char* symbol) +{ + cg->addMustPreserveSymbol(symbol); +} + + +// +// writes a new file at the specified path that contains the +// merged contents of all modules added so far. +// returns true on error (check lto_get_error_message() for details) +// +bool lto_codegen_write_merged_modules(lto_code_gen_t cg, const char* path) +{ + return cg->writeMergedModules(path, sLastErrorString); +} + + +// +// Generates code for all added modules into one native object file. +// On sucess returns a pointer to a generated mach-o/ELF buffer and +// length set to the buffer size. The buffer is owned by the +// lto_code_gen_t and will be freed when lto_codegen_dispose() +// is called, or lto_codegen_compile() is called again. +// On failure, returns NULL (check lto_get_error_message() for details). +// +extern const void* +lto_codegen_compile(lto_code_gen_t cg, size_t* length) +{ + return cg->compile(length, sLastErrorString); +} + + +// +// Used to pass extra options to the code generator +// +extern void +lto_codegen_debug_options(lto_code_gen_t cg, const char * opt) +{ + cg->setCodeGenDebugOptions(opt); +} \ No newline at end of file