void BasicBlockCounter::instrument() 
{
    InstrumentationTool::instrument();
    ASSERT(currentPhase == ElfInstPhase_user_reserve && "Instrumentation phase order must be observed"); 
    uint32_t temp32;

    LineInfoFinder* lineInfoFinder = NULL;
    if (hasLineInformation()){
        lineInfoFinder = getLineInfoFinder();
    }

    uint64_t lineArray = reserveDataOffset(getNumberOfExposedBasicBlocks() * sizeof(uint32_t));
    uint64_t fileNameArray = reserveDataOffset(getNumberOfExposedBasicBlocks() * sizeof(char*));
    uint64_t funcNameArray = reserveDataOffset(getNumberOfExposedBasicBlocks() * sizeof(char*));
    uint64_t hashCodeArray = reserveDataOffset(getNumberOfExposedBasicBlocks() * sizeof(uint64_t));
    uint64_t appName = reserveDataOffset((strlen(getApplicationName()) + 1) * sizeof(char));
    initializeReservedData(getInstDataAddress() + appName, strlen(getApplicationName()) + 1, getApplicationName());
    uint64_t instExt = reserveDataOffset((strlen(getInstSuffix()) + 1) * sizeof(char));
    initializeReservedData(getInstDataAddress() + instExt, strlen(getInstSuffix()) + 1, getInstSuffix());

    // the number blocks in the code
    uint64_t counterArrayEntries = reserveDataOffset(sizeof(uint32_t));

    // an array of counters. note that everything is passed by reference
    uint64_t counterArray = reserveDataOffset(getNumberOfExposedBasicBlocks() * sizeof(uint32_t));

    exitFunc->addArgument(counterArray);
    exitFunc->addArgument(appName);
    exitFunc->addArgument(instExt);

    InstrumentationPoint* p = addInstrumentationPoint(getProgramExitBlock(), exitFunc, InstrumentationMode_tramp);
    if (!p->getInstBaseAddress()){
        PRINT_ERROR("Cannot find an instrumentation point at the exit function");
    }

    temp32 = 0;
    for (uint32_t i = 0; i < getNumberOfExposedBasicBlocks(); i++){
        initializeReservedData(getInstDataAddress() + counterArray + i*sizeof(uint32_t), sizeof(uint32_t), &temp32);
    }

    // the number of inst points
    entryFunc->addArgument(counterArrayEntries);
    temp32 = getNumberOfExposedBasicBlocks();
    initializeReservedData(getInstDataAddress() + counterArrayEntries, sizeof(uint32_t), &temp32);

    // an array for line numbers
    entryFunc->addArgument(lineArray);
    // an array for file name pointers
    entryFunc->addArgument(fileNameArray);
    // an array for function name pointers
    entryFunc->addArgument(funcNameArray);
    // an array for hashcodes
    entryFunc->addArgument(hashCodeArray);

    p = addInstrumentationPoint(getProgramEntryBlock(), entryFunc, InstrumentationMode_tramp, FlagsProtectionMethod_full, InstLocation_prior);
    p->setPriority(InstPriority_userinit);
    if (!p->getInstBaseAddress()){
        PRINT_ERROR("Cannot find an instrumentation point at the entry block");
    }

    uint64_t noDataAddr = getInstDataAddress() + reserveDataOffset(strlen(NOSTRING) + 1);
    char* nostring = new char[strlen(NOSTRING) + 1];
    sprintf(nostring, "%s\0", NOSTRING);
    initializeReservedData(noDataAddr, strlen(NOSTRING) + 1, nostring);

    PRINT_DEBUG_MEMTRACK("There are %d instrumentation points", getNumberOfExposedBasicBlocks());
    Vector<BasicBlock*>* allBlocks = new Vector<BasicBlock*>();
    Vector<LineInfo*>* allLineInfos = new Vector<LineInfo*>();

    uint32_t noProtPoints = 0;
    uint32_t complexSelection = 0;
    for (uint32_t i = 0; i < getNumberOfExposedBasicBlocks(); i++){

        BasicBlock* bb = getExposedBasicBlock(i);
        LineInfo* li = NULL;
        if (lineInfoFinder){
            li = lineInfoFinder->lookupLineInfo(bb);
        }
        Function* f = bb->getFunction();

        (*allBlocks).append(bb);
        (*allLineInfos).append(li);

        if (i % 1000 == 0){
            PRINT_DEBUG_MEMTRACK("inst point %d", i);
            PRINT_MEMTRACK_STATS(__LINE__, __FILE__, __FUNCTION__);            
        }

        if (li){
            uint32_t line = li->GET(lr_line);
            initializeReservedData(getInstDataAddress() + lineArray + sizeof(uint32_t)*i, sizeof(uint32_t), &line);

            uint64_t filename = reserveDataOffset(strlen(li->getFileName()) + 1);
            uint64_t filenameAddr = getInstDataAddress() + filename;
            initializeReservedData(getInstDataAddress() + fileNameArray + i*sizeof(char*), sizeof(char*), &filenameAddr);
            initializeReservedData(getInstDataAddress() + filename, strlen(li->getFileName()) + 1, (void*)li->getFileName());

        } else {
            temp32 = 0;
            initializeReservedData(getInstDataAddress() + lineArray + sizeof(uint32_t)*i, sizeof(uint32_t), &temp32);
            initializeReservedData(getInstDataAddress() + fileNameArray + i*sizeof(char*), sizeof(char*), &noDataAddr);
        }
        uint64_t funcname = reserveDataOffset(strlen(f->getName()) + 1);
        uint64_t funcnameAddr = getInstDataAddress() + funcname;
        initializeReservedData(getInstDataAddress() + funcNameArray + i*sizeof(char*), sizeof(char*), &funcnameAddr);
        initializeReservedData(getInstDataAddress() + funcname, strlen(f->getName()) + 1, (void*)f->getName());

        uint64_t hashValue = bb->getHashCode().getValue();
        initializeReservedData(getInstDataAddress() + hashCodeArray + i*sizeof(uint64_t), sizeof(uint64_t), &hashValue);
        
        InstrumentationSnippet* snip = new InstrumentationSnippet();
        uint64_t counterOffset = counterArray + (i * sizeof(uint32_t));

        // snippet contents, in this case just increment a counter
        if (is64Bit()){
            snip->addSnippetInstruction(X86InstructionFactory64::emitAddImmByteToMem(1, getInstDataAddress() + counterOffset));
        } else {
            snip->addSnippetInstruction(X86InstructionFactory32::emitAddImmByteToMem(1, getInstDataAddress() + counterOffset));
        }

        // do not generate control instructions to get back to the application, this is done for
        // the snippet automatically during code generation
            
        // register the snippet we just created
        addInstrumentationSnippet(snip);            
            
        // register an instrumentation point at the function that uses this snippet
        FlagsProtectionMethods prot = FlagsProtectionMethod_light;
#ifndef NO_REG_ANALYSIS
        if (bb->getLeader()->allFlagsDeadIn()){
            prot = FlagsProtectionMethod_none;
            noProtPoints++;
        }
        for (uint32_t j = 0; j < bb->getNumberOfInstructions(); j++){
            if (bb->getInstruction(j)->allFlagsDeadIn() || bb->getInstruction(j)->allFlagsDeadOut()){
                complexSelection++;
                break;
            }
        }
#endif
        InstrumentationPoint* p = addInstrumentationPoint(bb, snip, InstrumentationMode_inline, prot);
    }
    PRINT_MEMTRACK_STATS(__LINE__, __FILE__, __FUNCTION__);
#ifdef NO_REG_ANALYSIS
    PRINT_WARN(10, "Warning: register analysis disabled");
#endif
    PRINT_INFOR("Excluding flags protection for %d/%d instrumentation points", noProtPoints, getNumberOfExposedBasicBlocks());
    //PRINT_INFOR("complex inst point selection: %d/%d instrumentation points", complexSelection, getNumberOfExposedBasicBlocks());

    printStaticFile(allBlocks, allLineInfos);

    delete[] nostring;
    delete allBlocks;
    delete allLineInfos;

    ASSERT(currentPhase == ElfInstPhase_user_reserve && "Instrumentation phase order must be observed"); 
}
void FlowGraph::computeDefUseDist(){
    uint32_t fcnt = function->getNumberOfInstructions();
    std::pebil_map_type<uint64_t, X86Instruction*> imap;
    std::pebil_map_type<uint64_t, BasicBlock*> bmap;
    std::pebil_map_type<uint64_t, LinkedList<X86Instruction::ReachingDefinition*>*> alliuses;
    std::pebil_map_type<uint64_t, LinkedList<X86Instruction::ReachingDefinition*>*> allidefs;

    for (uint32_t i = 0; i < basicBlocks.size(); i++){
        BasicBlock* bb = basicBlocks[i];
        for (uint32_t j = 0; j < bb->getNumberOfInstructions(); j++){
            X86Instruction* x = bb->getInstruction(j);
            for (uint32_t k = 0; k < x->getSizeInBytes(); k++){
                ASSERT(imap.count(x->getBaseAddress() + k) == 0);
                ASSERT(bmap.count(x->getBaseAddress() + k) == 0);
                imap[x->getBaseAddress() + k] = x;
                bmap[x->getBaseAddress() + k] = bb;
            }

            ASSERT(alliuses.count(x->getBaseAddress()) == 0);
            ASSERT(allidefs.count(x->getBaseAddress()) == 0);
            alliuses[x->getBaseAddress()] = x->getUses();
            allidefs[x->getBaseAddress()] = x->getDefs();
        }
    }

    // For each loop
    for (uint32_t i = 0; i < loops.size(); ++i) {
        BasicBlock** allLoopBlocks = new BasicBlock*[loops[i]->getNumberOfBlocks()];
        loops[i]->getAllBlocks(allLoopBlocks);
        uint64_t loopLeader = loops[i]->getHead()->getBaseAddress();
        // For each block
        for (uint32_t j = 0; j < loops[i]->getNumberOfBlocks(); ++j){
            BasicBlock* bb = allLoopBlocks[j];

            // For each instruction
            for (uint32_t k = 0; k < bb->getNumberOfInstructions(); ++k){
                X86Instruction* ins = bb->getInstruction(k);

                // Skip the instruction if it can't define anything
                if (!ins->isIntegerOperation() && !ins->isFloatPOperation() && !ins->isMoveOperation()) {
                    continue;
                }
                ASSERT(!ins->usesControlTarget());

                singleDefUse(this, ins, bb, loops[i], imap, bmap, alliuses, allidefs, k, loopLeader, fcnt);
            }
        }
        delete[] allLoopBlocks;
    }

    // handle all non-loop blocks
    for (uint32_t i = 0; i < getNumberOfBasicBlocks(); i++){
        BasicBlock* bb = basicBlocks[i];
        if (!isBlockInLoop(bb->getIndex())){
            for (uint32_t k = 0; k < bb->getNumberOfInstructions(); ++k){
                X86Instruction* ins = bb->getInstruction(k);
                
                if (!ins->isIntegerOperation() && !ins->isFloatPOperation() && !ins->isMoveOperation()) {
                    continue;
                }
                ASSERT(!ins->usesControlTarget());
                
                singleDefUse(this, ins, bb, NULL, imap, bmap, alliuses, allidefs, k, 0, fcnt);
            }
        }
    }

    for (uint32_t i = 0; i < basicBlocks.size(); i++){

        BasicBlock* bb = basicBlocks[i];
        for (uint32_t j = 0; j < bb->getNumberOfInstructions(); j++){

            X86Instruction* x = bb->getInstruction(j);
            uint64_t addr = x->getBaseAddress();

            ASSERT(alliuses.count(addr) == 1);
            LinkedList<X86Instruction::ReachingDefinition*>* l = alliuses[addr];
            alliuses.erase(addr);
            while (!l->empty()){
                delete l->shift();
            }
            delete l;

            ASSERT(allidefs.count(addr) == 1);
            l = allidefs[addr];
            allidefs.erase(addr);
            while (!l->empty()){
                delete l->shift();
            }
            delete l;
        }
    }

    ASSERT(alliuses.size() == 0);
    ASSERT(allidefs.size() == 0);
}