path: root/tools/gccld/GenerateCode.cpp

//===- GenerateCode.cpp - Functions for generating executable 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 file contains functions for generating executable files once linking
// has finished.  This includes generating a shell script to run the JIT or
// a native executable derived from the bytecode.
//
//===----------------------------------------------------------------------===//

#include "gccld.h"
#include "llvm/System/Program.h"
#include "llvm/Module.h"
#include "llvm/PassManager.h"
#include "llvm/Analysis/LoadValueNumbering.h"
#include "llvm/Analysis/Passes.h"
#include "llvm/Analysis/Verifier.h"
#include "llvm/Bytecode/Archive.h"
#include "llvm/Bytecode/WriteBytecodePass.h"
#include "llvm/Target/TargetData.h"
#include "llvm/Transforms/IPO.h"
#include "llvm/Transforms/Scalar.h"
#include "llvm/Support/SystemUtils.h"
#include "llvm/Support/CommandLine.h"

using namespace llvm;

namespace {
  cl::opt<bool>
  DisableInline("disable-inlining", cl::desc("Do not run the inliner pass"));

  cl::opt<bool>
  Verify("verify", cl::desc("Verify intermediate results of all passes"));

  cl::opt<bool>
  DisableOptimizations("disable-opt",
                       cl::desc("Do not run any optimization passes"));
}

/// CopyEnv - This function takes an array of environment variables and makes a
/// copy of it.  This copy can then be manipulated any way the caller likes
/// without affecting the process's real environment.
///
/// Inputs:
///  envp - An array of C strings containing an environment.
///
/// Return value:
///  NULL - An error occurred.
///
///  Otherwise, a pointer to a new array of C strings is returned.  Every string
///  in the array is a duplicate of the one in the original array (i.e. we do
///  not copy the char *'s from one array to another).
///
static char ** CopyEnv(char ** const envp) {
  // Count the number of entries in the old list;
  unsigned entries;   // The number of entries in the old environment list
  for (entries = 0; envp[entries] != NULL; entries++)
    /*empty*/;

  // Add one more entry for the NULL pointer that ends the list.
  ++entries;

  // If there are no entries at all, just return NULL.
  if (entries == 0)
    return NULL;

  // Allocate a new environment list.
  char **newenv = new char* [entries];
  if ((newenv = new char* [entries]) == NULL)
    return NULL;

  // Make a copy of the list.  Don't forget the NULL that ends the list.
  entries = 0;
  while (envp[entries] != NULL) {
    newenv[entries] = new char[strlen (envp[entries]) + 1];
    strcpy (newenv[entries], envp[entries]);
    ++entries;
  }
  newenv[entries] = NULL;

  return newenv;
}


/// RemoveEnv - Remove the specified environment variable from the environment
/// array.
///
/// Inputs:
///  name - The name of the variable to remove.  It cannot be NULL.
///  envp - The array of environment variables.  It cannot be NULL.
///
/// Notes:
///  This is mainly done because functions to remove items from the environment
///  are not available across all platforms.  In particular, Solaris does not
///  seem to have an unsetenv() function or a setenv() function (or they are
///  undocumented if they do exist).
///
static void RemoveEnv(const char * name, char ** const envp) {
  for (unsigned index=0; envp[index] != NULL; index++) {
    // Find the first equals sign in the array and make it an EOS character.
    char *p = strchr (envp[index], '=');
    if (p == NULL)
      continue;
    else
      *p = '\0';

    // Compare the two strings.  If they are equal, zap this string.
    // Otherwise, restore it.
    if (!strcmp(name, envp[index]))
      *envp[index] = '\0';
    else
      *p = '=';
  }
}

static void dumpArgs(const char **args) {
  std::cerr << *args++;
  while (*args)
    std::cerr << ' ' << *args++;
  std::cerr << '\n' << std::flush;
}

static inline void addPass(PassManager &PM, Pass *P) {
  // Add the pass to the pass manager...
  PM.add(P);

  // If we are verifying all of the intermediate steps, add the verifier...
  if (Verify) PM.add(createVerifierPass());
}

static bool isBytecodeLibrary(const sys::Path &FullPath) {
  // Check for a bytecode file
  if (FullPath.isBytecodeFile()) return true;
  // Check for a dynamic library file
  if (FullPath.isDynamicLibrary()) return false;
  // Check for a true bytecode archive file
  if (FullPath.isArchive() ) {
    std::string ErrorMessage;
    Archive* ar = Archive::OpenAndLoadSymbols( FullPath, &ErrorMessage );
    return ar->isBytecodeArchive();
  }
  return false;
}

static bool isBytecodeLPath(const std::string &LibPath) {
  sys::Path LPath(LibPath);

  // Make sure it exists and is a directory
  sys::FileStatus Status;
  if (LPath.getFileStatus(Status) || !Status.isDir)
    return false;
  
  // Grab the contents of the -L path
  std::set<sys::Path> Files;
  if (LPath.getDirectoryContents(Files, 0))
    return false;
  
  // Iterate over the contents one by one to determine
  // if this -L path has any bytecode shared libraries
  // or archives
  std::set<sys::Path>::iterator File = Files.begin();
  std::string dllsuffix = sys::Path::GetDLLSuffix();
  for (; File != Files.end(); ++File) {

    // Not a file?
    if (File->getFileStatus(Status) || Status.isDir)
      continue;

    std::string path = File->toString();

    // Check for an ending '.dll', '.so' or '.a' suffix as all
    // other files are not of interest to us here
    if (path.find(dllsuffix, path.size()-dllsuffix.size()) == std::string::npos
        && path.find(".a", path.size()-2) == std::string::npos)
      continue;

    // Finally, check to see if the file is a true bytecode file
    if (isBytecodeLibrary(*File))
      return true;
  }
  return false;
}

/// GenerateBytecode - generates a bytecode file from the specified module.
///
/// Inputs:
///  M           - The module for which bytecode should be generated.
///  StripLevel  - 2 if we should strip all symbols, 1 if we should strip
///                debug info.
///  Internalize - Flags whether all symbols should be marked internal.
///  Out         - Pointer to file stream to which to write the output.
///
/// Returns non-zero value on error.
///
int llvm::GenerateBytecode(Module *M, int StripLevel, bool Internalize,
                           std::ostream *Out) {
  // In addition to just linking the input from GCC, we also want to spiff it up
  // a little bit.  Do this now.
  PassManager Passes;

  if (Verify) Passes.add(createVerifierPass());

  // Add an appropriate TargetData instance for this module...
  addPass(Passes, new TargetData(M));

  // 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.
  addPass(Passes, createFunctionResolvingPass());

  if (!DisableOptimizations) {
    // Now that composite has been compiled, scan through the module, looking
    // for a main function.  If main is defined, mark all other functions
    // internal.
    addPass(Passes, createInternalizePass(Internalize));

    // Propagate constants at call sites into the functions they call.  This
    // opens opportunities for globalopt (and inlining) by substituting function
    // pointers passed as arguments to direct uses of functions.
    addPass(Passes, createIPSCCPPass());
    
    // Now that we internalized some globals, see if we can hack on them!
    addPass(Passes, createGlobalOptimizerPass());

    // Linking modules together can lead to duplicated global constants, only
    // keep one copy of each constant...
    addPass(Passes, createConstantMergePass());

    // Remove unused arguments from functions...
    addPass(Passes, createDeadArgEliminationPass());

    if (!DisableInline)
      addPass(Passes, createFunctionInliningPass()); // Inline small functions

    addPass(Passes, createPruneEHPass());            // Remove dead EH info
    addPass(Passes, createGlobalOptimizerPass());    // Optimize globals again.
    addPass(Passes, 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.
    addPass(Passes, createArgumentPromotionPass());

    // The IPO passes may leave cruft around.  Clean up after them.
    addPass(Passes, createInstructionCombiningPass());
    addPass(Passes, createPredicateSimplifierPass());

    addPass(Passes, createScalarReplAggregatesPass()); // Break up allocas

    // Run a few AA driven optimizations here and now, to cleanup the code.
    addPass(Passes, createGlobalsModRefPass());      // IP alias analysis

    addPass(Passes, createLICMPass());               // Hoist loop invariants
    addPass(Passes, createLoadValueNumberingPass()); // GVN for load instrs
    addPass(Passes, createGCSEPass());               // Remove common subexprs
    addPass(Passes, createDeadStoreEliminationPass()); // Nuke dead stores

    // Cleanup and simplify the code after the scalar optimizations.
    addPass(Passes, createInstructionCombiningPass());

    // Delete basic blocks, which optimization passes may have killed...
    addPass(Passes, createCFGSimplificationPass());

    // Now that we have optimized the program, discard unreachable functions...
    addPass(Pa