线程池

1.所谓线程池，就是程序的初始化阶段，就预先创建一批线程，每个线程都做好准备干活；

2.然后有一个任务列表，一开始为空，当有任务来了，就往任务列表里面添加；这个任务列表由于那些线程们都会进行操作，所以需要做好同步工作。

3.任务列表里面有任务了，这时候那些等待的线程们就要抢活干了，怎么抢，使用各种线程同步手段（互斥量，临界区等），人品好的线程抢到任务后，从任务列表取出任务，就可以开始干活了。干完以后，就又继续回到初始等待状态，准备抢夺下一个任务。

/**********************************************************************************************************************************************/
这样就好比你有一批小弟排队在那里等着，一旦有任务，他们会很守纪律的去抢着干，每个任务都会被一个小弟抢走，干完以后，小弟不用休息，继续等着抢下一个任务干活。这样当你的任务源源不断的到达，你的小弟们就一个个争先恐后的抢过来完成，绝不偷懒。
/**********************************************************************************************************************************************/

相反，如果不使用线程池，每次等到任务来了，再临时创建线程。这样就相当于每次有任务时，你再临时招聘一个小弟过来，小弟完成任务后，就回家了。然后下次再有任务，又招聘一个小弟过来，完成任务后，回家。相比线程池，中间招聘小弟的时间就要额外耗费时间和精力了（创建和销毁线程中，cpu的时间，内存的分配）。
/**********************************************************************************************************************************************/

所以，使用线程池，就省去了哪些额外的线程开销，从而连续的完成所有的任务。当然，线程池中用于线程同步的操作同样也有一定的消耗，但这个消耗是相对小的。另外，还可以对线程池中的线程根据当前的任务量进行动态的调整，从而更好的节省相关资源。

下面代码是网上找到的一个例子：

threadpool.h

/*
    Thread Pool implementation for unix / linux environments
    Copyright (C) 2008 Shobhit Gupta
	
    This program is free software: you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation, either version 3 of the License, or
    (at your option) any later version.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with this program.  If not, see <http://www.gnu.org/licenses/>.
*/

#include <pthread.h>
#include <semaphore.h>
#include <iostream>
#include <vector>

using namespace std;
/*
WorkerThread class
This class needs to be sobclassed by the user.
*/
class WorkerThread{
public:
    int id;

    unsigned virtual executeThis()
	{
		return 0;
	}

    WorkerThread(int id) : id(id) {}
    virtual ~WorkerThread(){}
};

/*
ThreadPool class manages all the ThreadPool related activities. This includes keeping track of idle threads and ynchronizations between all threads.
*/
class ThreadPool{
public:
    ThreadPool();
    ThreadPool(int maxThreadsTemp);
    virtual ~ThreadPool();
	
	void destroyPool(int maxPollSecs);

    bool assignWork(WorkerThread *worker);
    bool fetchWork(WorkerThread **worker);

	void initializeThreads();
	
    static void *threadExecute(void *param);
    
    static pthread_mutex_t mutexSync;
    static pthread_mutex_t mutexWorkCompletion;
    
    
private:
    int maxThreads;
    
    pthread_cond_t  condCrit;
    sem_t availableWork;
    sem_t availableThreads;

    //WorkerThread ** workerQueue;
    vector<WorkerThread *> workerQueue;

    int topIndex;
    int bottomIndex;
	
	int incompleteWork;

    
    int queueSize;

};

threadpool.cpp

/*
    Thread Pool implementation for unix / linux environments
    Copyright (C) 2008 Shobhit Gupta
	
    This program is free software: you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation, either version 3 of the License, or
    (at your option) any later version.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with this program.  If not, see <http://www.gnu.org/licenses/>.
*/

#include <stdlib.h>
#include "threadpool.h"

using namespace std;

pthread_mutex_t ThreadPool::mutexSync = PTHREAD_MUTEX_INITIALIZER;
pthread_mutex_t ThreadPool::mutexWorkCompletion = PTHREAD_MUTEX_INITIALIZER;



ThreadPool::ThreadPool()
{
	ThreadPool(2);
}

ThreadPool::ThreadPool(int maxThreads)
{
   if (maxThreads < 1)  maxThreads=1;
  
   //mutexSync = PTHREAD_MUTEX_INITIALIZER;
   //mutexWorkCompletion = PTHREAD_MUTEX_INITIALIZER; 
   
   pthread_mutex_lock(&mutexSync);
   this->maxThreads = maxThreads;
   this->queueSize = maxThreads;
   //workerQueue = new WorkerThread *[maxThreads];
   workerQueue.resize(maxThreads, NULL);
   topIndex = 0;
   bottomIndex = 0;
   incompleteWork = 0;
   sem_init(&availableWork, 0, 0);
   sem_init(&availableThreads, 0, queueSize);
   pthread_mutex_unlock(&mutexSync);
}

void ThreadPool::initializeThreads()
{
   for(int i = 0; i<maxThreads; ++i)
	{
		pthread_t tempThread;
		pthread_create(&tempThread, NULL, &ThreadPool::threadExecute, (void *) this ); 
		 //threadIdVec[i] = tempThread;
   }

}

ThreadPool::~ThreadPool()
{
   workerQueue.clear();
}



void ThreadPool::destroyPool(int maxPollSecs = 2)
{
	while( incompleteWork>0 )
	{
	        //cout << "Work is still incomplete=" << incompleteWork << endl;
		sleep(maxPollSecs);
	}
	cout << "All Done!! Wow! That was a lot of work!" << endl;
	sem_destroy(&availableWork);
	sem_destroy(&availableThreads);
        pthread_mutex_destroy(&mutexSync);
        pthread_mutex_destroy(&mutexWorkCompletion);

}


bool ThreadPool::assignWork(WorkerThread *workerThread)
{
        pthread_mutex_lock(&mutexWorkCompletion);
		incompleteWork++;
		//cout << "assignWork...incomapleteWork=" << incompleteWork << endl;
	pthread_mutex_unlock(&mutexWorkCompletion);
    
	sem_wait(&availableThreads);
	
	pthread_mutex_lock(&mutexSync);
		//workerVec[topIndex] = workerThread;
		workerQueue[topIndex] = workerThread;
                //cout << "Assigning Worker[" << workerThread->id << "] Address:[" << workerThread << "] to Queue index [" << topIndex << "]" << endl;
		if(queueSize !=1 )
			topIndex = (topIndex+1) % (queueSize-1);
		sem_post(&availableWork);
	pthread_mutex_unlock(&mutexSync);
	return true;
}

bool ThreadPool::fetchWork(WorkerThread **workerArg)
{
	sem_wait(&availableWork);

	pthread_mutex_lock(&mutexSync);
		WorkerThread * workerThread = workerQueue[bottomIndex];
                workerQueue[bottomIndex] = NULL;
		*workerArg = workerThread;
		if(queueSize !=1 )
			bottomIndex = (bottomIndex+1) % (queueSize-1);
		sem_post(&availableThreads);
	pthread_mutex_unlock(&mutexSync);
    return true;
}

void *ThreadPool::threadExecute(void *param)
{
	WorkerThread *worker = NULL;
	
	while(((ThreadPool *)param)->fetchWork(&worker))
	{
		if(worker)
                {
			worker->executeThis();
                        //cout << "worker[" << worker->id << "]\tdelete address: [" << worker << "]" << endl;
                        delete worker;
                        worker = NULL;
                }

		pthread_mutex_lock( &(((ThreadPool *)param)->mutexWorkCompletion) );
                //cout << "Thread " << pthread_self() << " has completed a Job !" << endl;
	 	((ThreadPool *)param)->incompleteWork--;
		pthread_mutex_unlock( &(((ThreadPool *)param)->mutexWorkCompletion) );
	}
	return 0;
}

main.cpp

/*
    Thread Pool implementation for unix / linux environments
    Copyright (C) 2008 Shobhit Gupta

    This program is free software: you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation, either version 3 of the License, or
    (at your option) any later version.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with this program.  If not, see <http://www.gnu.org/licenses/>.
*/

#include <iostream>
#include "threadpool.h"

using namespace std;


#define ITERATIONS 200

class SampleWorkerThread : public WorkerThread
{
public:
        int id;

	unsigned virtual executeThis()
	{
	// Instead of sleep() we could do anytime consuming work here.
	//Using ThreadPools is advantageous only when the work to be done is really time consuming. (atleast 1 or 2 seconds)
		sleep(2);

		return(0);
	}


        SampleWorkerThread(int id) : WorkerThread(id), id(id)
        {
//           cout << "Creating SampleWorkerThread " << id << "\t address=" << this << endl;
        }

        ~SampleWorkerThread()
        {
//           cout << "Deleting SampleWorkerThread " << id << "\t address=" << this << endl;
        }
};


int main(int argc, char **argv)
{
	//ThreadPool(N);
	//Create a Threadpool with N number of threads
	ThreadPool* myPool = new ThreadPool(25);
	myPool->initializeThreads();

	//We will count time elapsed after initializeThreads()
    time_t t1=time(NULL);

	//Lets start bullying ThreadPool with tonnes of work !!!
	for(unsigned int i=0;i<ITERATIONS;i++){
		SampleWorkerThread* myThread = new SampleWorkerThread(i);
//cout << "myThread[" << myThread->id << "] = [" << myThread << "]" << endl;
		myPool->assignWork(myThread);
	}
	
	// destroyPool(int maxPollSecs)
	// Before actually destroying the ThreadPool, this function checks if all the pending work is completed.
	// If the work is still not done, then it will check again after maxPollSecs
	// The default value for maxPollSecs is 2 seconds.
	// And ofcourse the user is supposed to adjust it for his needs.
	
    myPool->destroyPool(2);

    time_t t2=time(NULL);
    cout << t2-t1 << " seconds elapsed\n" << endl;
	delete myPool;
	
    return 0;
}