一份C++线程池的代码，非常实用

#ifndef _ThreadPool_H_
#define _ThreadPool_H_
#pragma warning(disable: 4530)
#pragma warning(disable: 4786)
#include <cassert>
#include <vector>
#include <queue>
#include <windows.h>

using namespace std;

class ThreadJob  //工作基类
{
public:
  //供线程池调用的虚函数
  virtual void DoJob(void *pPara) = 0;
};
class ThreadPool
{
public:
  //dwNum 线程池规模
  ThreadPool(DWORD dwNum = 4) : _lThreadNum(0), _lRunningNum(0)
  {
    InitializeCriticalSection(&_csThreadVector);
    InitializeCriticalSection(&_csWorkQueue);
    _EventComplete = CreateEvent(0, false, false, NULL);
    _EventEnd = CreateEvent(0, true, false, NULL);
    _SemaphoreCall = CreateSemaphore(0, 0,  0x7FFFFFFF, NULL);
    _SemaphoreDel =  CreateSemaphore(0, 0,  0x7FFFFFFF, NULL);
    assert(_SemaphoreCall != INVALID_HANDLE_VALUE);
    assert(_EventComplete != INVALID_HANDLE_VALUE);
    assert(_EventEnd != INVALID_HANDLE_VALUE);
    assert(_SemaphoreDel != INVALID_HANDLE_VALUE);
    AdjustSize(dwNum <= 0 ? 4 : dwNum);
  }
  ~ThreadPool()
  {
    DeleteCriticalSection(&_csWorkQueue);
    CloseHandle(_EventEnd);
    CloseHandle(_EventComplete);
    CloseHandle(_SemaphoreCall);
    CloseHandle(_SemaphoreDel);

    vector<ThreadItem*>::iterator iter;
    for(iter = _ThreadVector.begin(); iter != _ThreadVector.end(); iter++)
    {
      if(*iter)
        delete *iter;
    }
    DeleteCriticalSection(&_csThreadVector);
  }
  //调整线程池规模
  int AdjustSize(int iNum)
  {
    if(iNum > 0)
    {
      ThreadItem *pNew;
      EnterCriticalSection(&_csThreadVector);
      for(int _i=0; _i<iNum; _i++)
      {
        _ThreadVector.push_back(pNew = new ThreadItem(this));
        assert(pNew);
        pNew->_Handle = CreateThread(NULL, 0, DefaultJobProc, pNew, 0, NULL);
        // set priority
        SetThreadPriority(pNew->_Handle, THREAD_PRIORITY_BELOW_NORMAL);
        assert(pNew->_Handle);
      }
      LeaveCriticalSection(&_csThreadVector);
    }
    else
    {
      iNum *= -1;
      ReleaseSemaphore(_SemaphoreDel,  iNum > _lThreadNum ? _lThreadNum : iNum, NULL);
    }
    return (int)_lThreadNum;
  }
  //调用线程池
  void Call(void (*pFunc)(void  *), void *pPara = NULL)
  {
    assert(pFunc);
    EnterCriticalSection(&_csWorkQueue);
    _JobQueue.push(new JobItem(pFunc, pPara));
    LeaveCriticalSection(&_csWorkQueue);
    ReleaseSemaphore(_SemaphoreCall, 1, NULL);
  }
  //调用线程池
  inline void Call(ThreadJob * p, void *pPara = NULL)
  {
    Call(CallProc, new CallProcPara(p, pPara));
  }
  //结束线程池, 并同步等待
  bool EndAndWait(DWORD dwWaitTime = INFINITE)
  {
    SetEvent(_EventEnd);
    return WaitForSingleObject(_EventComplete, dwWaitTime) == WAIT_OBJECT_0;
  }
  //结束线程池
  inline void End()
  {
    SetEvent(_EventEnd);
  }
  inline DWORD Size()
  {
    return (DWORD)_lThreadNum;
  }
  inline DWORD GetRunningSize()
  {
    return (DWORD)_lRunningNum;
  }
  bool IsRunning()
  {
    return _lRunningNum > 0;
  }
protected:
  //工作线程
  static DWORD WINAPI DefaultJobProc(LPVOID lpParameter = NULL)
  {
    ThreadItem *pThread = static_cast<ThreadItem*>(lpParameter);
    assert(pThread);
    ThreadPool *pThreadPoolObj = pThread->_pThis;
    assert(pThreadPoolObj);
    InterlockedIncrement(&pThreadPoolObj->_lThreadNum);
    HANDLE hWaitHandle[3];
    hWaitHandle[0] = pThreadPoolObj->_SemaphoreCall;
    hWaitHandle[1] = pThreadPoolObj->_SemaphoreDel;
    hWaitHandle[2] = pThreadPoolObj->_EventEnd;
    JobItem *pJob;
    bool fHasJob;

    for(;;)
    {
      DWORD wr = WaitForMultipleObjects(3, hWaitHandle, false, INFINITE);
      //响应删除线程信号
      if(wr == WAIT_OBJECT_0 + 1)
        break;

      //从队列里取得用户作业
      EnterCriticalSection(&pThreadPoolObj->_csWorkQueue);
      if(fHasJob = !pThreadPoolObj->_JobQueue.empty())
      {
        pJob = pThreadPoolObj->_JobQueue.front();
        pThreadPoolObj->_JobQueue.pop();
        assert(pJob);
      }
      LeaveCriticalSection(&pThreadPoolObj->_csWorkQueue);
      //受到结束线程信号 确定是否结束线程(结束线程信号 && 是否还有工作)
      if(wr == WAIT_OBJECT_0 + 2 && !fHasJob)
        break;
      if(fHasJob && pJob)
      {
        InterlockedIncrement(&pThreadPoolObj->_lRunningNum);
        pThread->_dwLastBeginTime = GetTickCount();
        pThread->_dwCount++;
        pThread->_fIsRunning = true;
        pJob->_pFunc(pJob->_pPara); //运行用户作业
        delete pJob;
        pThread->_fIsRunning = false;
        InterlockedDecrement(&pThreadPoolObj->_lRunningNum);
      }
    }
    //删除自身结构
    EnterCriticalSection(&pThreadPoolObj->_csThreadVector);
    pThreadPoolObj->_ThreadVector.erase(find(pThreadPoolObj->_ThreadVector.begin(), pThreadPoolObj->_ThreadVector.end(), pThread));
    LeaveCriticalSection(&pThreadPoolObj->_csThreadVector);
    delete pThread;
    InterlockedDecrement(&pThreadPoolObj->_lThreadNum);
    if(!pThreadPoolObj->_lThreadNum)  //所有线程结束
      SetEvent(pThreadPoolObj->_EventComplete);
    return 0;
  }
  //调用用户对象虚函数
  static void CallProc(void *pPara)
  {
    CallProcPara *cp = static_cast<CallProcPara *>(pPara);
    assert(cp);
    if(cp)
    {
      cp->_pObj->DoJob(cp->_pPara);
      delete cp;
    }
  }
  //用户对象结构
  struct CallProcPara
  {
    ThreadJob* _pObj;//用户对象
    void *_pPara;//用户参数
    CallProcPara(ThreadJob* p, void *pPara) : _pObj(p), _pPara(pPara) { };
  };
  //用户函数结构
  struct JobItem
  {
    void (*_pFunc)(void  *);//函数
    void *_pPara; //参数
    JobItem(void (*pFunc)(void  *) = NULL, void *pPara = NULL) : _pFunc(pFunc), _pPara(pPara) { };
  };
  //线程池中的线程结构
  struct ThreadItem
  {
    HANDLE _Handle; //线程句柄
    ThreadPool *_pThis;  //线程池的指针
    DWORD _dwLastBeginTime; //最后一次运行开始时间
    DWORD _dwCount; //运行次数
    bool _fIsRunning;
    ThreadItem(ThreadPool *pthis) : _pThis(pthis), _Handle(NULL), _dwLastBeginTime(0), _dwCount(0), _fIsRunning(false) { };
    ~ThreadItem()
    {
      if(_Handle)
      {
        CloseHandle(_Handle);
        _Handle = NULL;
      }
    }
  };

  std::queue<JobItem *> _JobQueue;  //工作队列
  std::vector<ThreadItem *>  _ThreadVector; //线程数据
  CRITICAL_SECTION _csThreadVector, _csWorkQueue; //工作队列临界, 线程数据临界
  HANDLE _EventEnd, _EventComplete, _SemaphoreCall, _SemaphoreDel;//结束通知, 完成事件, 工作信号， 删除线程信号
  long _lThreadNum, _lRunningNum; //线程数, 运行的线程数
};
#endif //_ThreadPool_H_

转载自 http://blog.csdn.net/pjchen/archive/2004/11/06/170606.aspx

基本上是拿来就用了，对WIN32 API不熟，但对线程池的逻辑还是比较熟的，认为这个线程池写得很清晰，我拿来用在一个多线程下载的模块中。很实用的东东。

调用方法

void threadfunc(void *p)

{

     YourClass* yourObject = (YourClass*)    p;

//... } ThreadPool tp; for(i=0; i<100; i++)   tp.Call(threadfunc);

ThreadPool tp(20);//20为初始线程池规模

tp.Call(threadfunc, lpPara);     

使用时注意几点：

1. ThreadJob  没什么用，直接写线程函数吧。

2. 线程函数（threadfunc）的入口参数void* 可以转成自定义的类型对象，这个对象可以记录下线程运行中的数据，并设置线程当前状态，以此与线程进行交互。

3. 线程池有一个EndAndWait函数，用于让线程池中所有计算正常结束。有时线程池中的一个线程可能要运行很长时间，怎么办？可以通过线程函数threadfunc的入口参数对象来处理，比如：

class YourClass {

  int cmd; // cmd = 1是上线程停止计算，正常退出。

};

threadfunc(void* p) {

  YourClass* yourObject = (YourClass*)p;

  while (true) {

    // do some calculation

    if (yourClass->cmd == 1)

      break;

  }

}

在主线程中设置yourClass->cmd = 1，该线程就会自然结束。

很简洁通用的线程池实现。